8th International Symposium
on NDT in Aerospace
3rd – 5th November, 2016, Bangalore, India
8th International Symposium on Non-Destructive Testing in Aerospace
3rd – 5th November, 2016, Bangalore, India
Contents
Overview
Programme
About the Symposium
About Indian Institute of Science, Bangalore
Papers
Overview
0800 - 0830
0830 - 1005
rd
3 November, 2016,
Thursday
th
4 November, 2016,
Thursday
th
5 November, 2016,
Thursday
1005 - 1035
1035 - 1235
1235 – 1330
1330 – 1520
1520 – 1550
1550 – 1740
1930 onwards
0830 - 1000
1000 - 1030
1030 - 1230
1230 – 1330
1330 – 1520
1520 – 1550
1550 – 1740
1830 – 1930
2000 onwards
0830 - 1000
1000 - 1030
1030 - 1230
1230 – 1330
1330 – 1520
1520 – 1545
1600 – 1800
REGISTRATION STARTS
SYMPOSIYM INAUGURATION and PLENARY
SESSION 1
COFFEE BREAK
PLENARY SESSION 2
LUNCH
PARALLEL SESSIONS
COFFEE BREAK
PARALLEL SESSIONS
BANQUET DINNER
PLENARY SESSION 3
COFFEE BREAK
PLENARY SESSION 4
LUNCH
PARALLEL SESSIONS
COFFEE BREAK
PARALLEL SESSIONS
CULTURAL EVENING
DINNER
PLENARY SESSION 5
COFFEE BREAK
PLENARY SESSION 6
LUNCH
PARALLEL SESSIONS
COFFEE BREAK
VISIT TO IISc HYPERSONIC TEST FACILITY
Schedule
November 3, 2016
8.30-10.05 am – Symposium Inauguration and Plenary Session 1
Session Chair: Dr. Christian Boller, University of Saarland, Germany
8.30-8.35am: Welcome and about the Conference, Prof. S Gopalakrishnan
8.35-8.40 am: About NDT in Aerospace Symposium, Dr. Christian Boller
8.40-9.20 am: NDT Initiatives in Indian Aerospace programs, Dr. Kota
Harinarayana, Former Program Director, ADA, India
9.20-10.05 am: Plenary Talk: Hand Operated Shock Tube (Reddy Tube) for
Non Destructive Testing, Prof. K P J Reddy, Department of Aerospace
Engineering, Indian Institute of Science, Bangalore, India
10.05 am-10.35 am: Coffee/Tea
10.35 am-12.35 pm: Plenary Session 2
NDE Aspects for Military Aviation and Defense
Session Chair: Dr. Kota Harinarayana, India
10.35-11.05am: New NDE Technologies and Solutions from CNDE@IITM for
Aerospace Industries, Krishnan Balasubramaniam, Indian Institute of
Technology, Madras, Chennai , India (Invited Talk)
11.05-11.35 am: Role of NDT in ensuring Structural Integrity of Aircraft, Dr. P
D Mangalgiri, Visiting professor, Indian Institute of Technology Kanpur,
India (Invited Talk)
11.35 am-12.05 pm: Aviation Safety, Prof. B Dattaguru, Jain University,
Bangalore, India (Invited Talk)
12.05-12.35 pm: Electromechanical Impedance Method for Assessment of
Adhesive Bonds of CFRP at the Production and Repair Stage, Prof. Wiesław M
Ostachowicz, Polish Academy of Sciences, Poland (Invited Talk)
12.35 pm-1.30 pm: Lunch
1.30 pm-3.20pm: Parallel Sessions
Composites-I
Metallic Materials-I
Applications-I
3.20 pm-3.50 pm: Tea/Coffee
4.00 pm-6.00 pm: Parallel Sessions
Composites-II
Metallic Materials -II
Applications-II
7.30 pm: Banquet Dinner
November 4, 2016
8.30 am-10.00 am: Plenary Session 3
Chair: Prof. Krishnan Balasubramaniam, IITM, India
8.30-9.15 am : Use of NDE Methodologies in Marine Composites, Dr. Yapa
Rajapakse, Programme Manager, Solid Mechanics, Office of Naval Research,
USA
9.15-10.00am: A Permanent Inspection System for Damage Detection at
Composite Laminates, Based on Distributed Fibre Optics, Prof. Alfredo
Gümes, Universidad Politécnica de Madrid, Spain
10.00 am-10.30 am: Tea/Coffee
10.30 am-12.30 pm: Plenary Session 4
SHM Simulation Platform
Chair: Prof. B Dattaguru, IIAEM, India
10.30-11.00 am: Simulation as a Prerequisite in Structural Health
Monitoring, Prof. Christian Boller, University of Saarland, Germany (Invited
Talk)
11.00-11.30 am: SHM System Simulation Based Design Considering
Composite Patch Repaired Stiffened Panel of Aircrafts, Rakesh Shivamurthy,
Keerthy M Simon, Nitin Balajee Ravi, Nibir Chakraborty, Debiprosad Roy
Mahapatra, IISc, Bangalore, India (Invited Talk)
11.30 am-12.00 noon: Feasibility Study of SHM Simulation Based Design of
Accelerated Fatigue Tests, Nitin Balajee Ravi, Nibir Chakraborty, Rakesh
Shivamurthy, Keerthy M Simon, Ramanan Sridaran Venkat, Mirko Steckel,
Debiprosad Roy Mahapatra, Christian Boller, IISc, Bangalore, India (Invited
Talk)
12.00-12.30 pm : Simulation as Key Enabler to Support ISHM Certification,
Matthias Buderath, Partha Adhikari, Harsha Gururaja Rao, Airbus,
Bangalore, India (Invited Talk)
12.30 pm-1.30 pm: Lunch
1.30 pm-3.20 pm: Parallel Sessions
Composites -III
Guided Waves-I Emerging Technologies
3.20 pm-3.50 pm: Tea/Coffee
3.50 pm-5.40 pm: Parallel Sessions
Composites-IV
Guided Waves-II Applications-III
6.30 pm-7.30pm: Cultural Evening - Demonstration of Indian Art
forms: Mrs. Shyla Prasad
8.00 pm: Dinner
November 5, 2016
8.30 am-10.00 am: Plenary Session 5
Chair: S V Suresh, HAL Helicopters, India
8.30-9.15 am: Real-Time Tracking of Damage Growth in CFRP Composites
during Testing for Durability and Fracture, Dr. R Sunder, BISS Research,
Bangalore, India
9.15-10.00 am: Integrated System Health Monitoring and Management
Initiatives and Challenges in Military Aviation, Matthias Buderath, Airbus
Defence & Space, Manching, Germany
10.00 am-10.30 am: Tea/Coffee
10.30-12.30 pm: Plenary Session 6
SHM Simulation Platform
Chair: Dr. P D Mangalgiri, IITK, India
10.30-11.0 am: Accelerated Fatigue Data Evaluation for SHM Validation
Activities, Peter Starke and Christian Boller, University of Saarland, Germany
(Invited Talk)
11.00-11.30 am: Simulation of Ultrasonic Inspection of Defects in Thick
Structural Components, Nitin Balajee Ravi, Nibir Chakraborty, Rakesh
Shivamurthy, Ramanan Sridaran Venkat, Debiprosad Roy Mahapatra,
Christian Boller, IISc, India
11.30-12.0 noon: Simulation of Guided Waves Inspection, from NDE to SHM,
Pierre Calmon, Bastien Chapuis CEA LIST, France (Invited Talk)
12.0 noon-12.30 pm: Structural Health Monitoring of Repaired Metallic
Aircraft Panel-Modeling Based Approach, Ramanan Sridaran Venkat, Adrià
Taltvull, Christian Boller, Christian Durager, University of Saarland,
Germany (Invited Talk)
12.30 pm-1.30pm: Lunch
1.30 pm-3.20pm: Parallel Sessions
Composites-V
Applications-IV
3.20 pm 3.45 pm: Tea/Coffee
4.0 pm-6.0pm: Visit to IISc Hypersonic Test Facility
Session Composites-I
Date: November 3, 2016 Time: 1.30-3.20pm
Session Chair:
1.30-2.00 pm: How to Make NDE of Composites More Deterministic, Dr. B V S R
Murthy, Advanced Systems Lab, DRDO, India, (Invited paper)
2.00-2.20 pm: A Novel Nonlinear Acoustics Technique for the Detection of
Defects in Composites, A K Singh, National University of Singapore, Singapore
2.20-2.40 pm: Comparison of Fibre Angles Between Hand Draped Carbon
Fibres and Draping Simulation, C Frommel, DLR, Augsburg, Germany
2.40-3.00 pm: Acoustic Emission Monitoring of the Cracking Load in Composite
Materials, G Nardoni, P Nardoni, M Turconi, I&T Nardoni Institute, Leonardo
Finmeccanica Group, Italy
3.00-3.20 pm: Guided Lamb Wave Based Multi-Level Disbond Detection in a
Honeycomb Composite Sandwich Structure, Shirsendu Sikdar, Sauvik Banerjee,
IIT Mumbai, India
Session Metallic Materials-I
Date: November 3, 2016 Time: 1.30-3.10pm
Session Chair: Dr. R Ganguli, IISc, India
1.30-1.50 pm: NDE Assessment of Aerospace Components – IGCAR Experiences,
Dr B. Venkatraman , IGCAR, Bangalore, India (Invited Talk)
1.50-2.10 pm: Subwavelength Resolution of Delaminations, Amireddy Kiran
Kumar, Prabhu Rajagopal and Krishnan Balasubramaniam, IIT Madras, India
2.10-2.30 pm: Anisotropic Magnetic Properties of Grain-oriented and Nonoriented Si-Fe Electrical Steel, Gholamhossein Shirkoohi, London Southbank
Univ., England
2.30-2.50 pm: Sonic Analysis System as an Effective Means of NDT Analysis for
Cast Iron Nodularity Measurement, Kalyan Ram B, S Arun Kumar, Prajval M S,
Electrono Solutions Pvt. Ltd, India
2.50-3.10 pm: Spectral Finite Element Method for Inspection of Adhesively
Bonded Metallic Joints Using Guided Waves, Shweta Paunikar, S
Gopalakrishnan, IISc, Bangalore, India
Session Applications-I
Date: November 3, 2016 Time: 1.30-3.20pm
Session Chair: Ramesh Sundaam, CSIR-National Aerospace Laboratories,
India
1.30-2.00 pm: NDE for Aircraft Engines: Challenges, Opportunities and
Approaches, Shyam Sundar, GE, Bangalore, India (Invited paper)
2.00-2.20 pm: Non Destructive Evaluation Methodologies to Predict the Strain
Hardening Effect in Landing Gear Components: A Review, Arun Dinesh,
Chanakesava Reddy, Bharath Marappan, Seshadri Venkatadri, UTC Aerospace
Systems, Bangalore, India
2.20-2.40 pm: Main Gear Box Testing for Light Utility Helicopter, Bhawani
Singh Rathore, Muralidhar B S, RWR&DC, HAL, India
2.40-3.00 pm: Experimental Validation of Random Packs for Composite Solid
Propellants Using X- Ray Computed Tomography, Chaitanya V, Raghuvarun K,
Sai Bhargav K V, Krishnan Balasubramaniam, P A Ramakrishna, IIT Chennai,
India
3.00-3.20 pm: Fault Detection in Electrical Wires and Cables, Gholamhossein
Shirkoohi, London South Bank Univ., England
Session: Composites-II
Date: November 3, 2016 Time: 3.50-5.30pm
Session Chair: Dr. Mira Mitra, IITB, India
3.50-4.20 pm: Structural Health Monitoring of Aircraft Composite Structures:
Offline & Online Approach, Ramesh Sundaram, Nitesh Gupta, Augustin M J,
Amitabha Datta, CSIR – National Aerospace Laboratories, Bangalore, India
(Invited Talk)
4.20-4.40 pm: Characterization of Ceramic Composites by High Frequency
Eddy Current Techniques, H Heuer, M Schulze, S Hillmann, Fraunhofer IKTS,
Germany
4.40-5.10 pm: Thermal Properties of Basalt Fibre Epoxy Composites by
Focused Gaussian Illumination Using Infrared Thermography, Kalyanavalli V,
T K Abilasha Ramadhas, and D SastiKumar, Nat. Inst. of Technology,
Tiruchirapalli, India
5.10-5.30 pm: In-situ Nondestructive Evaluation of Composite Panels in an
Aircraft using InfraRed Thermography, S Kshama and S Kalyana Sundaram,
NAL Bangalore, India
Session: Metallic Materials-II
Date: November 3, 2016 Time: 3.50-5.40pm
Session Chair: Dr. Makarand Joshi, DRDO - R&DE(Engrs), India
3.50-4.20 pm: NDE of Aerospace Structural Materials and Components Using
Acoustic Wave Based Tool, M R Bhat, Department of Aerospace Engineering,
IISc, Bangalore, India (Invited paper)
4.20-4.40 pm: A Non-Destructive Methodology of Estimating Single Crystal
Elastic Constants, Phani Mylavarapu, Karthik Karuparthi, Jai Prakash
Gautam, Defence Metallurgical Research Laboratory, Univ. Hyderabad, India
4.40-5.00 pm: Detection of Material Inhomogeneity Using Inverse Time
Domain Spectral Finite Element Method, Raghavendra B Kulkarni, S
Gopalakrishnan and Manish Trikha, ISRO, Bangalore, India
5.00-5.20 pm: Evaluation of Elastic Modulus of Metals Using Acoustic Emission
Technique, R Gopikrishna, M Padma Amani, M Varadanam, Defence Research
Development Laboratory, India
5.20-5.40 pm: Non-Destructive Evaluation of Friction Stir Weld Defects in
Aluminum Alloy, V D Ragupathy, M R Bhat and M V N Prasad, LPSC, ISRO, India
Session: Applications-II
Date: November 3, 2016 Time: 3.50-5.30 pm
Session Chair: Dr. Suresh Bhalla, IITD, India
3.50-4.10 pm: Detection of Blocked Cooling Holes in a Gas Turbine Nozzle using
Infrared Imaging, Joel Jon Bosco, Dheepa Srinivasan, Debabrata
Mukhppadhyay and Paul Dimascio, GE Power, India
4.10-4.30 pm: Studies on Multi Site Damages in Riveted Joints under Fatigue
Cycles with Acoustic Emission Approximate Entropy Approach, S Kalyana
Sundaram, V R Ranganath, M R Bhat, NAL, Bangalore, India
4.30-4.50 pm: Hula Seal Stellite6 Coating Thickness on Service Returned
Combustion Liner Using Ultrasonic Gauge, Santhosh Bangera, Dheepa
Srinivasan, Mohammed Anwar, James Baummer, Francis Reed, GE Power,
India
4.50-5.10 pm: Effects And Corrective Techniques of Helium as Signal Noise in
the MSLD Based Leak Detection Methods, Soumikh Sarkhel, V D Ragupathy and
G Naryanan, Liquid Propulsion System Centre, India
5.10-5.30 pm: Application of Image based technique in Qualification Testing of
Spacecraft Structures & Components, Sriranga T S, Harsh Kumar, Ananthan A,
Raghunatha Behara, C S Varghese, ISRO, Satellite Centre, India
Session: Composites-III
Date: November 4, 2016 Time: 1.30-3.20pm
Session Chair: Prof. D Roy Mahapatra, IISc, India
1.30-2.00 pm: Uncertainty Handling Using Fuzzy Logic in Structural Health
Monitoring, Prof. R Ganguli, Department of Aerospace Engineering, IISc,
Bangalore, India (Invited paper)
2.00-2.20 pm: Damage Assessment of Single Blade Stiffened CFRP Specimen
Subjected to Axial Compression Using AE And DIC Techniques, N R Kolanu, L B
Andraju, M Ramji, G Raju,IIT Hyderabad, India
2.20-2.40 pm: Real Time Monitoring of Interfacial Delamination of Sandwich
Composite Panels Using Optical Sensors, Nilanjan Mitra, IIT Kharagpur, India
2.40-3.00 pm: Locating Delamination in a Composite Laminate Using
Nonlinear Response of Lamb Waves, Nitesh P Yelve, Mira Mitra, and Prasanna
M Mujumdar, IIT Mumbai, India
3.00-3.20 pm: Investigation of Anisotropic Propagation of Shear Horizontal
Modes in Composite Laminates Using Fiber Bragg Grating Sensors, P Ray, P
Rajagopal, B Srinivasan and K Balasubramaniam, IIT Chennai, India
Session: Guided Waves-I
Date: November 4, 2016 Time: 1.30-3.20pm
Session Chair: Dr. Peter Starke, University of Saarland, Germany
1.30-2.00 pm: Damage Detection Using Nonlinear Interaction with Breathing
Crack in 1-D Beam, Mira Mitra, IIT Mumbai, India (Invited paper)
2.00-2.20pm: Corrosion Detection on Aerospace Structural Materials by Lamb
Wave Visualization Method - Advantages & Challenges, Aparna Aradhye, S
Kalyana Sundaram, NAL, Bangalore, India
2.20-2.40 pm: Laser Based Non-Contact Method for Nonlinear Mixing of
Rayleigh waves, Chaitanya Bakre, Prabhu Rajagopal and Krishnan
Balasubramaniam, IIT Chennai, India
2.40-3.00pm: Localization of a Breathing Crack in a Stepped Beam Using a
Method of Harmonic Separation, D M Joglekar, IIT Roorkee, India
3.00-3.20 pm: The Interfacial Stiffness Evaluation of Single Lap Joint
Assemblies Using the Transmission of Lamb Waves, E Siryabe, M Renier, A
Meziane, M Castaings, Univ. Bordeaux, France
Session: Emerging Technologies
Date: November 4, 2016 Time: 1.30-3.20pm
Session Chair: Dr. Prabhu Rajagopal, IITM, India
1.30-2.00 pm: Issues in Piezoelectric Energy Harvesting from Civil Structures,
Dr. Suresh Bhalla, Indian Institute of Technology, Delhi, India (Invited paper)
2.00-2.20 pm: Modeling of Elastic Wave Scattering in Polycrystalline Materials,
Abhishek Pandala, S Shivaprasad, C V Krishnamurthy, Krishnan
Balasubramaniam, IIT Chennai, India
2.20-2.40 pm: Potential for Synchrotron Radiation Based NDE in Aerospace
Industry, C V Krishnamurthy, IIT Chennai, India
2.40-3.00 pm: Effect of Surface Asperity on Thin Film Adhesion using Laser
Induced Stress Waves, S S Singh, R Kitey, IIT Kanpur, India
3.00-3.20 pm: Effect of Shock Induced Acoustic Emission and Shock Wave
Impact on Polyurethane Foam, V Jayaram and G Arvind Raj, IISc, India
Session: Composites-IV
Date: November 4, 2016 Time: 3.50-5.30pm
Session Chair: Wieslaw Ostochowicz, Polish Academy of Sciences, Poland
3.50-4.10 pm: Manufacturing Aspects on Fabrication of Composite Reference
Standard for NDT Ultrasonic Inspection, Pranab Biswal, B N Srinivasa Reddy,
Pratim. M Baruah, HAL, Bangalore, India
4.10-4.30 pm: Wave Propagation in a Delaminated Thin Anisotropic Strip,
Punith P, M Mitra and P J Guruprasad, IIT Mumbai, India
4.30-4.50 pm: Rapid Ultrasonic Inspection of Stiffened Composite Ailerons,
Rabi Sankar Panda, Prabhu Rajagopal and Krishnan Balasubramaniam, IIT
Chennai, India
4.50-5.10 pm: Evaluation of Composite Structures-Using Computed
Tomography (CT) Emerging NDE Methodology, Ramesh Babu G, Pranab
Biswal, B N Srinivasa Reddy, P Mukhopadhyay, Hindustan Aerospace Ltd.
Bangalore, India
5.10-5.30 pm: Non Destructive Evaluation of In-situ Skirt Joint of Filament
Wound Composite Pressure Vessel, Sanjeev Kumar and P J Thakar, ASL, DRDO,
Hyderabad, India
Session: Guided Waves-II
Date: November 4, 2016 Time: 3.50-5.40pm
Session Chair: Prof. Alfredo Güemes, Universidad Politécnica de Madrid,
Spain
3.50-4.20 pm: Feature-Guided Waves: New Paradigm For Inspection and
Health Monitoring of Aerospace Composite Structures and Components, Dr.
Prabhu Rajagopal, IIT Madras, India (Invited paper)
4.20-4.40 pm: Multiphysics Simulation of Guided Wave Propagation under
Load Condition, Lei Qiu, University of Saarland, Germany
4.40-5.00 pm: Lamb Wave Based Damage Detection using Orthogonal
Matching Pursuit and Artificial Neural Network, Navjeet Kumar, Mira Mitra,
IIT Mumbai, India
5.00-5.20 pm: Attenuation of Fundamental Anti-Symmetric Lamb Mode (A0) in
Isotropic Plates, C Ramadas, Irfan Khan and Makarand Joshi , R & DE(Engrs),
Pune, India
5.20-5.40 pm: Interaction of Fundamental S0 Lamb Mode with Delaminations
in Composite Plate Structures, Saurabh Gupta and Prabhu Rajagopal, IIT
Madras, India
Session: Applications-III
Date: November 4, 2016 Time: 3.50-5.30pm
Session Chair: M R Bhat, IISc, India
3.50-4.10 pm: Detecting Onset of Combustion Instability in Gas Turbines
through Flame Visualisation using Fiber Optic Bundle, Suma H, Joel Vasanth,
Balaji Srinivasan and S R Chakravarthy, IIT Chennai, India
4.10-430 pm: Practical Experiences in POD Determination for Airframe ET
Inspection- Virkkunen I, Ylitalo M, Pirtola J and Patronen J, Patria Aviation,
Finland
4.30-4.50 pm: Post Impact Damage Evaluation of High Velocity Impacted EGlass Composites Using Immersion Ultrasonic C-Scan Technique, M Srinivasa
Rao, T Sreekantha Reddy, P Rama Subba Reddy, V Madhu, B V S R Murthy, ASL,
DRDO, Hyderbad, India
4.50-5.10 pm: Model Assisted Probability of Detection for Lognormally
Distributed Defects, Vamsi Krishna Rentala, Phani Mylavarapu ,K.Gopinath ,
J.P.Gautam , Vikas Kumar, University of Hyderabad, India
5.10-5.30 pm: A Scanning Spatial-Wavenumber Filter based On-Line Damage
Imaging Method of Composite Structure, Lei Qiu, Shenfang Yuan, Yuanqiang
Ren, Qiao Bao, NUAA Nanjing, China
5.30-5.50 pm: Non-destructive Evaluation of Wrinkles, Sivaramanivas R,
Megha Navalgund, Debasish Mishra and Richard Klaassen, GE Global
Research, India and GE Aviation, USA
Session: Composites-V
Date: November 5, 2016 Time: 1.30-3.30pm
Session Chair: Dr. B V S R Murthy, DRDO-ASL, India
1.30-1.50 pm: A Study on Highly Porous Carbon-Carbon Aircraft Brake Disc
using Air Coupled Ultrasonic, Suresh Chand Jangir, Srinivasa V, Ramesh
Kumar M, Ramesh Sundaram, Gururaja Rao J, NAL Bangalore, India
1.50-2.10 pm: Evaluation of Thermography and Ultrasonic NDT Techniques for
Detecting Resin Rich and Resin Starved Defects in Composites, YLVD Prasad,
Dr. SK Majee, A O Siddiqui, B V S R Murthy, ASL, DRDO, Hyderabad, India
2.10-2.30 pm: Aging study of Coated Fabric Materials, H Nanjundegowda, R
Sivaraman, K Vembar, B Marappan & V Seshadiri, UTC Aerospace Systems,
India
2.30-2.50 pm: Online Structural Health Monitoring of Composites, Harsh Shah,
Prabhu Rajagopal, Krishnan Balasubramaniam, IIT Madras, India
2.50-3.10 pm: Integrated Approach to Demonstrate Optimum Sensor Positions
in a Guided Wave Based SHM System Using Numerical Simulation Ramanan
Sridaran Venkat, Christian Boller, Lei Qiu, Nitin Balajee Ravi , Debiprosad Roy
Mahapatra, Nibir Chakraborty, University of Saarland, Germany
3.10-3.30 pm: Wave Scattering Analysis in a Delaminated Cross-ply Laminate
due to Incident S0 Wave, Rajendra Kumar Munian, G Kolappan Geetha, D Roy
Mahapatra, S Gopalakrishnan, Indian Institute of Science, India
Session: Applications-IV
Date: November 5, 2016 Time: 1.30-3.10pm
Session Chair: Dr. Lei Qiu, Nanjing University, China
1.30-150 pm: SHM - Prognostic Analysis of Crack Growth retardation in
Fastener Joints with Bush around the Pin- L Chikmath, B V Sravan, K Bharath
and B Dattaguru, Jain University, Bangalore, India
1.50-2.10 pm: Multiple NDE Methods for Crack Characterization on Spur Gear,
M R Vijaya Lakshmi, A K Mondal, Shubhanjali, M V Subbaraju, V Thangavelu
Sreelal Sreedhar, Gas Turbine Research Establishment, Bangalore, India
2.10-2.30 pm: On Health Monitoring of Insert Joints in Spacecraft Structures, P
Sathish Kumar, A Ananthan, T S Srirangaa, Krushna Chandra Dakua S
Shankar Narayan, ISRO, Satellite Centre, India
2.30-2.50 pm: Thermo Elastic Deformation Measurements on Spacecraft
Components using Non-contact Target-less Image Based Technique,
Raghunatha Behara, T S Srirangaa, CS Varghese, C Koteshwar Rao, Swapnil
Pathak,Pravesh Mathur, Krushna Chandra Dakua, ISRO, Satellite Centre, India
2.50-3.10 pm: Advances in Phased Array Ultrasonic Testing Analysis Software,
Bharath Kodumuru and Jay Amos, Textron Aviation, India and USA
About the Symposium
Non-Destructive Testing and Evaluation (NDT&E) is one of the major requirements in aerospace structural
design. Almost all the components have to pass through various NDT quality assurance procedures. For
damage tolerant design used in aviation, NDT is a prerequisite. Appropriate use of NDT guarantees safety in
aerospace and is thus a subject of highest attention.
Although aerospace industries have to clearly stick to certified NDT procedures, there is a large amount of
technology being currently developed within the engineering science discipline, which is worth discussion
with respect to their application potential. The widespread use of emerging materials such as CFRP
composites as well as new manufacturing processes for aerospace structural components has also given rise to
many new questions regarding the materials’ reliability and manufacturing methods in the light of nondestructive evaluation, both from a research as well as a manufacturing and assembly point of view. To
communicate latest R&D achievements to industrial applicants on the one hand and to discuss advanced and
improved methods both with scientists as well as industrial researchers on the other, the Symposium for NDT
in Aerospace has been established in 2008 and has been organised on an annual basis since 2010. It is run in
places worldwide all around the globe where aerospace has a home including Hamburg, Germany, Montreal,
Canada, Augsburg, Germany, Singapore, Madrid, Spain, and Bremen, Germany so far. The 8th Symposium for
NDT in Aerospace is now being held in November 2016 in Bangalore/ India.
This Symposium of NDT in Aerospace is intended to provide the delegates the latest trends in this area
especially in the context of Aerospace Engineering. This year's symposium will have two theme oriented
plenary sessions addressing NDT aspects in Military aviation and in Structural Health Monitoring Simulation
Platform. Well selected Plenary and invited speakers will address some novel topics of high relevance and this
symposium will in particular showcase the experience of Indian Scientists and Engineers in the NDT field
through a number of invited and contributed papers.
Prof. S Gopalakrishnan
Indian Institute of Science,
Bangalore, India
Prof. K Balasubramaniam
Indian Institute of Technology Madras,
Chennai, India
Scientific Advisory Committee
George Akhras
Christian Boller
S Gopalakrishnan
Alfredo Güemes
Randolf Hanke
Johann Kastner
Patrice Masson
William H. Prosser
Matthias Purschke
Christian U. Große
Shenfang Yuan
Wieslaw Ostachowicz
Robert Smith
Gary Georgeson
Philippe Benoist
Esmeralda Cuevas
Royal Military College of Canada, Kingston, Canada
Saarland University, Saarbrücken, Germany
Indian Institute of Science, Bangalore, India
Universidad Politecnica de Madrid, Spain
Fraunhofer IZFP, Saarbrücken, Fraunhofer EZRT, Fürth &
University of Würzburg, Germany
University of Applied Sciences Upper Austria, Wels, Austria
GAUS – Univ. de Sherbrooke, Canada
NASA, Hampton, USA
DGZfP, Berlin, Germany
TU Munich, Germany
Nanjing University of Aeronautics and Astronautics, Nanjing,
China
Polish Academy of Sciences, Gdansk, Poland
University of Bristol, United Kingdom
Boeing Commercial Aircrafts, Seattle, USA
M2M, Les Ulis, France
Tecnatom, Madrid, Spain
National Advisory Committee
Ramesh Sundaram
Roy Mahapatra
M Ramachandra Bhat
Mira Mitra
National Aerospace Laboratories, Bangalore, India
Indian Institute of Science, Bangalore, India
Indian Institute of Science, Bangalore, India
Defence Research and Development Organisation, Hyderabad,
India
Indian Institute of Technology Bombay, Mumbai, India
B Venkataraman
Indira Gandhi Centre for Atomic Research, Kalpakkam, India
Prabhu Rajagopal
Shyamsundar
K Vijayaraju
S Kalyana Sundaram
Indian Institute of Technology Madras, Chennai, India
GE Global Research Centre, Bangalore, India
Aeronautical Development Agency, Bangalore, India
National Aerospace Laboratories, Bangalore, India
Liquid Propulsion Systems Centre, Indian Space Research
Organisation, India
Structural Engineering Research Centre, Chennai, India
B V S R Murthy
Raghupathy
Santosh Kapuria
About the Indian Institute of Science, India
The Indian Institute of Science (IISc) was founded in 1909 as a result of the joint efforts of Jamsetji
Nusserwanji Tata, the Government of India, and the Maharaja of Mysore in Bangalore, India. The name
Bangalore encompasses so many sobriquets like Garden City, Silicon Valley of India, and Pub City and so on.
One of the prettiest cities in India, known for its salubrious climate all year round, Bangalore is bestowed with
tall tree lines streets and several parks adding to its greenery. With its gorgeous parks, boulevards,
multiplexes, bustling shopping bazaars and historical monuments, Bangalore is indubitably sparkling life with
vigour.
Over the 105 years since its establishment, IISc has become the premier institute for advanced scientific and
technological research and education in India. Beginning with 2 departments and 21 students in 1911, today
IISc has 39 departments, units, or centres, 3500 students, and about 500 academic and scientific staff,
supported by 600 administrative personnel. Out of this population of students, about 2200 are in various PhD
programs, almost 900 are enrolled for various Masters degrees, whereas about 400 are registered in the newly
established, research oriented, four-year Bachelor of Science (Research) programme, of which the first batch
graduated in 2015.
In the recent years, with new centres such as CiSTUP (The Centre of Infrastructure, Sustainable
Transportation, and Urban Planning), the Divecha Centre for Climate Change, the Centre for Earth Sciences,
the Centre for NeuroScience, the Centre for Excellence in Nano Science and Engineering (CeNSE), the
Robert Bosch Centre for Cyber Physical Systems (RBCCPS), a new Centre for Brain Research the Institute
has vigorously promoted inter-disciplinary research. With the recent construction of modern buildings,
acquisition of new research facilities, and induction of a large number of new faculty members, the Institute
can look forward to an era of high productivity, increasing impact, and rising prominence in the world.
Papers
Hand Operated Shock Tube (Reddy Tube) for Non Destructive Testing
K P J Reddy
Department of Aerospace Engineering,
Indian Institute of Science, Bangalore, India
Contact: laser@aero.iisc.ernet.in
The field of shock waves has attracted the attention of many researchers for more than a century and the
physics behind the shock wave phenomenon has been well understood. However, applications of shock waves
are relatively unexplored in spite of realizing the potential applications in all fields of science and engineering
in the laboratory. More surprising is the lack of application of shock waves in Non Destructive Testing
(NDT). Similar to light pulse shaping in lasers the pressure pulses of required shape, strength and frequency
can be generated using shock tubes. These pulses can ideally be used for many NDT applications. It is
surprising that there are no reports of using shock waves in NDT except for the pioneering paper published in
the year 2002 (NDT&E Int., 35, 399-406, 2002). This could be due to the bulkiness of the shock tubes capable
of producing the required overpressures and complexity of their operation. Our recent invention of hand
operated shock tube named Reddy tube, shown in Figure 1, has overcome all these limitations and hence is an
ideal devise for NDT. Reddy tube is capable of producing shock Mach numbers exceeding 2 manually without
any high pressure gas supply and it can be made in any size starting from few centimeters to meters with
diameter varying from one millimeter to tens of millimeters. Reddy tube has already found applications in
many fields including science, engineering and medicine. We will describe the basic principle and design of
Reddy tube along with typical examples of few applications which are already in practice. We will also
present a tentative plan for using the Reddy tube for NDT. We feel that the versatility and easy portability of
Reddy tube will eventually make it a preferred tool for NDT. Finally we will describe the manually operated
Reddy tunnel driven by Reddy tube, shown in Figure 2, capable of producing hypersonic flows with flow
Mach numbers exceeding 6 suitable for testing the aerodynamics of hypersonic flight vehicles.
Figure 1: Hand operated shock tube:
Reddy Tube
Figure 2: Hand operated hypersonic shock tunnel:
Reddy Tunnel
New NDE Technologies and Solutions from CNDE@IITM for Aerospace Industries
Prof. Krishnan Balasubramaniam
Centre for Non-destructive Evaluation & Department of Mechanical Engineering
IIT Madras, Chennai-600036, Tamil Nadu, India
Contact: balas@iitm.ac.in
Several NDT methods are currently employed in the quality assurance and in-service inspection of aerospace
components. Most of these techniques are focused on the detection, sizing, and characterization of flaws such
as cracks, at pre-determined critical locations, that lead to fractures and hence failures in the component.
Advanced NDE methods are being developed by the Centre for NDE at the Indian Institute of Technology
Madras (CNDE@IITM) that may potentially influence the fabrication, inspection, safety, costing, and
maintainability of the aerospace components and its fleets in the strategic sector as well as the commercial
sector. Some of the techniques that will be discussed here includes: (a) Use of ultrasonic guided waves for the
improved inspection of complex structures and components including hidden areas, (b) Structural health
monitoring of components and structures using attached and embedded sensor networks, (c) Use of new and
novel active thermography techniques such as induction and laser based, and (d) Waveguide sensors for
process condition measurements. Using the methods discussed here, the operator now has the opportunity to
take vital decisions such as component integrity and propose necessary repair/replacement or estimate the
remaining life of the component.
Role of NDT in Ensuring Structural Integrity of Aircraft
Prakash D Mangalgiri
Formerly, Visiting Professor, Department of Aerospace Engineering,
IIT Kanpur, India 208016
Structural design philosophy of aircraft has evolved over the years, learning often from mishaps and service
experiences, so as to ensure structural integrity and safety. Increasing complexity of aircraft structure to meet
the ever increasing demands on the performance of the aircraft – be it a small combat aircraft or a large longrange passenger aircraft – has brought in the need for constant vigilance for defects and damages during the
manufacture and during the service. Also, the increasing sizes of fleets with shrinking budgets and
competitiveness in commercial sector have brought in the need to keep the costs for diagnosis and prognosis
of defects and damages at an economical level. This has given rise to a constant evolution and refinement of
the Non-destructive Testing (NDT) techniques, as well as devising suitable methodologies and protocols to
use them efficiently. It has also brought in new paradigms of looking at the entire life cycle of the aircraft with
the participation of the Design, Manufacture, Operations and Maintenance agencies. The modern concepts of
Structural Health Management (SHM) thus involve a spread of conventional and new NDT techniques and
their customisation into a strategy to ensure safety and mission-readiness of the aircraft at an affordable cost.
The development of the Light Combat Aircraft (LCA), Tejas, in India has seen a concurrent development of
NDT techniques and their use for ensuring structural integrity. The design philosophy of damage tolerance,
use of fibre-reinforced polymer composites and the need for reducing the maintenance costs – all these have
been the drivers for evolving a suitable strategy for use of NDT techniques leading to a Structural Health
Management (SHM) scheme. Also, National initiatives of R&D Programmes on Smart Technologies and
MEMS based sensors have contributed significantly to development of NDT technologies useful for SHM.
These developments have spanned from NDT at manufacturing stage, such as use of ultrasonics for
composites, to monitoring systems on-board aircraft and rapid NDE systems for reducing maintenance time
and costs.
The talk will highlight some of these developments and issues involved in use of NDT techniques in ensuring
aircraft safety, based on the experiences in the above programmes.
Aviation Safety
B Dattaguru
International Institute of Aerospace Engineering & Management
Department of Aerospace Engineering,
Jakkasandra, Bangalore, India-562112
Contact: datgurb@gmail.com
Aviation safety in aero-vehicles is to ensure safety till the vehicle completes its mission. For example in civil
airlines the requirement is to ensure safety of all passengers from the time they enter the airport till they land
at the destination and leave the airport. In other types of aero-vehicles the most critical mission of the aircraft
is to be clearly defined. This is a multi-dimensional issue covering structural safety during the flight,
continuous monitoring of engine health monitoring, air traffic control to help landing and take-off of a number
of aircraft in a short time avoiding collisions and any other incidents and/or accidents, smooth logistics
operations to ensure availability of aircraft at any instant of time, airport design for user friendly business
operations and introducing anti-terror mechanisms. At Jain University the research groups are progressing on
the basis of “Aviation Safety” as the main theme. The strong structures group is already advanced in their
efforts. The other groups are initiated and are being oriented towards this theme.
The current paper deals with prognostic aspect of Structural Health Monitoring (SHM). There are well known
programs such as NASGRO to study the growth of defects (or cracks) in standard configurations. On the
other hand it is necessary to develop special equations for fracture parameters while dealing with non-linear
problems and novel materials such as composites and its variants. There is considerable progress has been
achieved in the study of fracture parameters at critical fastener and bonded joints, delamination in layered
composites and cracks merging in aging structures. For accurate estimation of fracture parameters special
post-processing techniques such Modified Virtual Crack Closure Integral (MVCCI) is developed. Some of
the results in these areas are presented in this paper.
Most of the problems addressed are non-linear in any one or more of the types of non-linearity. Fastener joints
are moving boundary value and contact problems during load transfer, bonded joints undergo large
deformation and adhesive material has low yielding strength, delaminated composite undergoes large
deformation of sub-laminate and cracks are deemed to have merged when plastic zones of the cracks meet. In
all these problems the main concentration of the work is on prognostic approach estimating the defect growth
life and estimating the remaining life at any instant. Engine health monitoring is initiated by the propulsion
group. Inter-relations between engine parameters and their effect on the engine thrust are being developed.
Several components of the engine are also susceptible to fatigue damage. Growth of three-dimensional cracks
are analysed with and without interaction with neighbouring cracks.
The recent ventures are in the area of “Big Data Analytics” and collision avoidance. Big Data has been
addressed for 2 – different requirements. One is towards analysis of data from Structural Health Monitoring
where the main aim is to identify damage initiation and growth so that estimation can be carried out in the
remaining strength and remaining life. The second is to identify any unusual activity to signify terror activity
which can be prevented. In the area of engine health monitoring the dependence of thrust on the other engine
measured parameters is being developed. Research plans in these directions is found to be essential for safe
installations. Integration of these various activities would go a long way to ensure Aviation safety.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Electromechanical impedance method for assessment of adhesive bonds
of CFRP at the production and repair stage
Paweł H. MALINOWSKI1, Tomasz WANDOWSKI1, Wiesław M. OSTACHOWICZ1,2
1
Institute of Fluid–Flow Machinery, Polish Academy of Sciences
14 Fiszera Street, 80–231, Gdansk, Poland
Phone: +48 58 341 12 71, Fax: +48 58 341 61 44; e-mail: pmalinowski@imp.gda.pl, tomaszw@imp.gda.pl
2
Warsaw University of Technology, Faculty of Automotive and Construction Machinery, Warsaw, Poland;
e-mail: wieslaw@imp.gda.pl
Abstract
Numerous techniques of non-destructive testing (NDT) of structural parts of CFRP are investigated. In this paper
electromechanical impedance (EMI) technique is studied as an NDT tool for assessment of adhesive bonds. In
order to perform the assessment a surface mounted piezoelectric sensor is used. Due to the piezoelectric effects
the electrical response of the sensor is related to mechanical response of the inspected object. The electrical
quantities of the sensor are tracked in order to find a relation with the mechanical state of the object. In the reported
research adhesively bonded CFRP samples were investigated. The adhesive bonds were modified simulating the
conditions in manufacturing and repair stages. Sample surface was contaminated before bonding with release agent
in order to simulate a manufacturing stage threat to the quality of the bond. Pre-bond thermal treatment, pre-bond
contamination with de-icing fluid, and faulty curing of the adhesive were considered as repair stage threats to the
quality of the bond. The electromechanical impedance spectra were investigated searching for anomalies and
changes caused by modification of the adhesive bond. These spectra for different cases were compared with
reference measurement results gathered from pristine samples. Numerical indexes for comparison of the EMI
characteristics were proposed. The sensitivity of the EMI method to modified bonds was observed.
Keywords: electromechanical impedance, EMI, adhesive bonding, carbon fiber composite, CFRP
1. Introduction
The paper focuses on the assessment of the CFRP adhesive bonds. It is important to have a
reliable tool that allows to verify the integrity of the bond during manufacturing or after repair.
The joints between structural elements should ensure safe usage of a structure. The performance
of adhesive bonds depends on the properties of the adhered surfaces. Improper preparation of
the surface may lead to weak bond and in consequence to failure. Mechanical tests showed that
the bond weakness can be related to contamination or bad curing of the adhesive.
Authors in [1] studied the effect of pre-bond release agent and moisture absorption on mode-I
fracture toughness of CFRP bonded joints. A mode-I double cantilever beam (DCB) specimen
was used to determine the mode-I energy release rate (GIC) referring to crack initiation. The
lowest considered contamination with release agent do not seem to affect the bond. The increase
of the contamination results in 60 % and higher drop in GIC in comparison to samples with
uncontaminated bond. In [2] researchers showed that lower temperature of adhesive curing
have great impact on the bond performance. The GIC for sample cured at 120⁰C drops nearly
by 95% in relation to GIC for properly cured sample at 180⁰C. In research reported here authors
focus on monitoring of adhesive bonds for both manufacturing and repair-related scenarios.
The paper is structured as follows. Firstly, the EMI method is described. Secondly, the samples
with manufacturing-related modification of the adhesive bonds are described and respective
results are presented. Thirdly, the samples with repair-related modifications are described
together with the assessment results. The paper ends with a concluding section.
2. Electromechanical impedance (EMI)
EMI uses measurements of electrical parameters of piezoelectric transducer bonded to a host
structure. Due to electromechanical coupling of the transducer and the structure, mechanical
resonances have their influence on electrical characteristics of the piezoelectric transducer. EMI
is an active method, because transducer works both as actuator and as sensor. Data analysis for
this method is performed in the frequency domain. One of the first publications about this
method is [3]. In experimental research laboratory impedance analysers like HP4194 or HIOKI
IM3570 are utilized. Researchers also utilise impedance analyser AD5933 in the form of
integrated circuit. In research reported here authors utilize the IM3570 analyser, because the
research was focused on thickness mode of the transducer that appears about 4 MHz. The
promising results of using this approach for composites adhesive bonds were presented in [4].
Measurements were performed with a piezoelectric disc transducer that was bonded at the
middle of investigated samples. The disc had 10 mm diameter and 0.5 mm thickness, and was
manufactured by CeramTec out of SONOX P502 material.
3. Samples and results
The material used for samples manufacturing was Hexcel M21E. The size of samples was 10
cm x 10 cm. The samples comprised of two plates bonded together. Each plate was made with
8 plies and layup sequence: [0, 0, 45, -45, -45, 45, 0, 0]. The plates were bonded with film
adhesive. In the case of samples for manufacturing stage the adhesive was FM 300 K cured at
171˚C, while for repair stage the adhesive was FM 300-2 cured at 121˚C.
3.1 Adhesive bonding assessment for manufacturing stage
The manufacturing stage samples comprised of samples contaminated with silicon-based
release agent, the Frekote 700-NC. It was applied in three defined concentrations on clean
CFRP plates by means of dip-coating. Then these plates were bonded to clean ones. The
concentrations of release agent on the surfaces was expressed by silicon concentracion
measured with XPS (3.2, 5.1, 6.2 at% Si). There were 3 samples prepared for each level of
contamination. For comparison three reference samples were prepared (PRE). In total there
were 12 samples in the manufacturing stage scenario. The results are depicted in figure 1. The
results were averaged, because there were four groups of the samples prepared in the same way.
In the results, one observes a slight increase of the RMS value for the first level of
contamination case (PRA1). For the higher levels (PRA2, PRA3) of contamination there is a
sudden drop indicating a significantly different adhesive bond.
Fig. 1. Results of the assessment of samples with pre-bond contamination with release agent
3.2 Adhesive bonding assessment for repair stage
The repair stage samples comprised of four kinds of samples. The investigations were
conducted on 24 samples: 3 reference samples, 9 samples with pre-bond thermal treatment, 9
samples with pre-bond contamination and 3 samples with faulty adhesive curing. The thermal
treatment was performed by exposure of samples to elevated temperatures (220, 260 and
280˚C) for two hours and grinding afterwards before bonding. The pre-bond contamination
with de-icing fluid was prepared by dip coating of the plates in a solution of SAFEWAY KF
de-icer with three concentrations. The contamination was expressed by potassium presence at
the surface measured by XPS (6.4, 10.9, 12.0 at% K). The case of faulty curing of adhesive was
prepared by local pre-curing of the adhesive by IR-spot. This pre-curing was made at the middle
of the sample.
Fig. 2. Results of the assessment of samples with pre-bond thermal treatment
Fig. 3. Results of the assessment of samples with pre-bond contamination with de-icing fluid
Fig. 4. Results of the assessment of samples with faulty curing of the adhesive
Figure 2 depicts the result for the thermally treated samples. An increasing value of RMS is
observed. The reference (RRE) case has the lowest value and the samples treated at 280˚C
(RTD3) have the largest value. Although, there were 3 levels of de-icing fluid contamination
prepared, the second (RDI2) and third (RDI3) level resulted in approximately the same level of
surface contamination, if the error of concentration estimation was considered. For this reason
the results for this two cases were averaged and treated as one level. The result for the de-icing
fluid were depicted in figure 3. A RMS value increase is observed as the contamination level
increases.The last scenario was related to faulty curing of adhesive. Only one level (RFC1) of
modified bond was available at the time of preparing this manuscript. The RMS value increases
by more than 30% (figure 4).
4. Conclusions
It as shown that the EMI method indicates some sensitivity to the modification of adhesive
bonds of CFRP samples for manufacturing and repair stage scenarios. There is a decrease of
RMS as the contamination with release agent increases. In the repair stage samples, there is an
increase of RMS as the modification severity is higher (RTD and RDI samples), so the bond
quality level assessment is possible. The detection of faulty cured adhesive for RFC samples
was also possible. The main difference between the two stages is that for manufacturing case
RMS drop was observed, while an increase was seen for repair stage samples. It should be
remembered that samples for this two stages were prepared in different way. There was
different film adhesive used and the repair stage samples were grinded down to fibers before
bonding.
The research will be continued taking into account the bond performance assessment obtained
from destructive testing and referential nondestructive methods. Moreover other indices and
other scenarios will be also investigated.
Acknowledgements
This project has received funding from the European Union’s Horizon 2020 research and
innovation programme under grant agreement No 636494.
This work was partially supported by the grant no. PBS1/B6/8/2012 (project KOMPNDT) of
Polish National Centre for Research and Development (NCBIR).
References
1. D N Markatos, et al. ‘The effects of manufacturing-induced and in-service related bonding
quality reduction on the mode-I fracture toughness of composite bonded joints for
aeronautical use’, Composites: Part B, 45, pp 556–564, 2013.
2. D N Markatos, et al. ‘ Degradation of Mode-I fracture toughness of CFRP bonded joints
due to moisture and release agent and moisture pre-bond contamination’, Journal of
Adhesion, 90, pp 156-173, 2014.
3. C Liang, et al. ‘An impedance method for dynamic analysis of active material system’,
Journal of Vibration and Acoustics, 116(1), pp 120-128, 1994.
4. P Malinowski, et al. ‘The use of electromechanical impedance conductance signatures for
detection of weak adhesive bonds of carbon fibre–reinforced polymer’, Structural Health
Monitoring, 14(4), pp 332-344, 2015.
NDT Aspects in Marine Composites under Extreme Environments
Yapa D. S. Rajapakse
Office of Naval Research (ONR 332)
875 North Randolph Street (Suite 1425)
Arlington, VA 22203-1995, U.S.A.
Contact: yapa.rajapakse@navy.mil
The Solid Mechanics Program of the Office of Naval Research (ONR) provides the scientific basis for
the effective design and utilization of affordable and reliable naval structures operating in severe
environments. Current emphasis on energy efficient, reliable, agile structures with enhanced
capabilities and reduced life-cycle costs, has led to the increased use of composite materials in ship
structures. The marine environment is hostile because of the presence of high humidity levels, sea
water, wave loading, hydrostatic pressure, and temperature extremes. Marine structures are subjected
to fluid-structure loading, and Naval structures are designed to resist highly dynamic loading (shock,
blast, implosions).The performance of composite structures in this environment, the challenges
encountered, and the physics/mechanics underlying these processes are the central themes addressed
by the ONR Solid Mechanics Program.
The current research focus is on the mechanics of marine composite materials and composite
sandwich structures. The program seeks to establish physically-based models for the processes
involved in the thermo-mechanical response of glass-fiber and carbon-fiber reinforced marine
composite materials and composite sandwich structures, subjected to static, cyclic, and dynamic
loading conditions, in severe environments. The establishment of these of these validated models,
with predictive capabilities, requires multi-scale, multi-physics analyses. Avenues for enhancing the
performance of marine composite structures through the introduction of nanoparticles and nanotubes,
and through the incorporation of novel design concepts, are also being explored in this program.
Research on multi-functional composites seeks to enhance performance through the incorporation of
additional beneficial attributes, without compromising on the mechanical properties.
Several recent accomplishments will be summarized, including: Failure criteria under multi-axial
loading; Effects of loads on sea water absorption, and degradation in mechanical properties; Effect of
extreme cold temperature on dynamic response; Shock/blast effects on composite and sandwich
structures; Implosions in submerged composite structures; Blast mitigation in composite structures;
and Mapping interior deformations using digital volumetric speckle photography
The presentation will include a discussion of future directions of research in mechanics of marine
composites and sandwich structures for affordable naval structures, with enhanced performance and
reduced life-cycle costs. Areas of increased emphasis include: Behavior of composite materials and
structures in extreme cold (Arctic) environments; Mitigation of shock/blast effects in composite
structures; combined effects of sea water, cold temperatures, and highly dynamic loading; and Nondestructive evaluation techniques for internal damage in composites. This research will contribute to
the design of affordable naval structures with enhanced performance and reduced life-cycle costs.
8th International Symposium on NDT in Aerospace, November 3-5,
2016
A permanent inspection system for damage detection at composite
laminates, based on distributed fiber optics sensing
Alfredo GUEMES, A. FERNÁNDEZ-LÓPEZ, P. FERNÁNDEZ DÍAZ-MAROTO
1
Dpt Aeronautics, Polytechnics University of Madrid, Spain
Phone: +34913366327, Fax: +34913366334; e-mail: alfredo.guemes@upm.es
Abstract
Distributed fibre optic systems allow to get information of the local strain all along the optical fibre.
Advanced systems may get strains readings with a spatial resolution of a few mm, and accuracy of a few
micro strains. This new technology offer a wide range of possibilities for structural testing, but also for
Damage detection, because a crack crossing or growing nearby the optical fiber path will be detected as a
local spot with high strains.
Concerning composite laminates, the most feared and common damage are VBID ( Barely visible Impact
Damage) because they may go undetected to visual inspection, but they cause a drop up to 50 % for the
compressive strength. These internal delaminations are also accompanied by some local residual strains at
the damage area, which are easily detectable if a strain sensor was installed just there.
Usually, the laminate edges, like manholes, or the area surrounding doors, are high risk areas for impact
damage, so the proposed approach is to bond or embed optical fiber along this perimeter, that may be
checked regularly for delaminations. The concept was experimentally demonstrated on laminate edges,
and also on internal delaminations, and on stringer run-away. The reliability of the technique has been
found to be 100%, always that the damage happened close to the fiber path.
Keywords: Structural Health Monitoring, SHM, aerospace, carbon fiber composite, CFRP
1. Introduction
Impact damage is considered to be the highest threat for composite structures during its
service life. Low/medium energy impacts (called BVID = Barely Visible Impact
Damage) do not leave any external visible marks, but cause internal delaminations that
drop the compressive strength by nearly 50 %. They need to be identified and repaired
as soon as possible to avoid the growth of the damaged area under repeated loads.
A permanent inspection, without disassembly the structure, even without the need of
moving a probe on the surface, is what may be achieved by SHM, which allows a
continuous automatic surveillance of the structural parts.
Structural Health Monitoring (SHM) is defined [1] as “the process of acquiring and
analyzing data from on-board sensors to evaluate the health of a structure”. The three
elements of an SHM system are:
1. A network of sensors, permanently attached to the structure; this aspect is a
main differentiation with conventional NDT procedures.
2. On-board data handling and computing facilities. Due to the high number of
sensors, data have to be processed on real time; wireless is a desired capability
(avoid cabling, repairability). SHM was feasible when large capacity portable
computers were available ( mid ‘80)
3. Algorithms that collect data from sensors, clean data from environmental
effects, compare to former data from the pristine structure and inform about
occurrence, localization and damage type.
Several kinds of sensors are used for SHM, but the two most commonly employed are
Piezoelectric wafers (PZT) and Fiber Optic sensors (FOS). Piezoelectrics offers a high
sensitivity, with a direct conversion of dynamic strain into voltage, and the opposite, so
they may act also as actuators.
Fiber Optic sensors offer a very low size, the optical fiber has a diameter of 150
microns, so it can be embedded within a ply into the composite material during
manufacturing. Other benefits for FOS are EMI/RFI immunity, wide temperature range,
very long cabling if needed, because of the low attenuation, and the multiplexing
capability (several sensors on the same optical fiber). As sketched at figure 1, three
topologies are possible:
•
•
•
Point sensor: detect measurand variation only in the vicinity of the sensor.
Example: micromirror at fiber tip. This is mostly used for chemical sensors
Multiplexed sensor: Multiple localized sensors are placed at intervals along the
fiber length. i.e . FBG (sensor length 10 mm typical). About 10 sensors/fiber if
multiplexed by wavelength, to 1000 sensors by using OFDR
Distributed sensor: Sensing is distributed along the length of the fiber, the
optical fiber works simultaneously for transmitting the information and for
sensing the local external variables (temperature, strain).
Figure 1 Topology of Fiber Optic Sensors
Fiber Optic Sensors have built a confidence at their performances as strain/temperature
sensors, equaling conventional sensors, and their reliability is now fully proven and
accepted. As damage sensors, the following considerations must be taken into account:
•
•
•
Strain changes caused by damage are very small a few cms away from the crack
tip, and masked by temperature drifting, load changes or any other
environmental factor.
Getting information about damage occurrence from strain measurements is then
a difficult task. Of course the larger the damage and the proximity to some
sensor, the higher the probability to be detected.
It drives to the need to include a large number of sensors into the structure,
which is feasible when using optical fiber sensors,
•
Data processing has to be done in a fully automated approach, algorithms are
needed to compare and extract information from the multiple strain
measurements.
2. Fiber Optic Distributed Sensing
Several kinds of Fiber Optic Distributed sensing systems are available, in dependence of
the wavelength they are working with [2]. Table 1 summarizes its performances.
Rayleigh system working with OFDR (Optical Frequency Domain Reflectometry) is the
only one to offer spatial resolution in the range of mm, as needed for aeronautic
applications (for civil engineering applications, a long measurement range may be the
preferred quality, which may drive to other choices)
Table 1. Comparison of Distributed Fiber Optic sensor systems
This OFDR technology has proven to be very useful for structural tests of large
structures [3]. At figure 2 left, it is shown the static test of a large wind turbine blade,
instrumented with just 4 continuous optical fibers (and 200 conventional electric strain
sensors, for comparison purposes).
Figure 2. Left, static test of a large wind turbine; right: strain measurements
obtained by distributed sensing (pink line), and strain gages (blue dots)
Results are shown at figure 2 right: pink line are the measurements obtained with the OFDR,
blue dots are measurements obtained from the strain gages. Not only if was much simpler to
prepare the test, but a more detailed information is obtained (as the sudden strain changes due
to the ply drops), and also, and very important, information on the initial phases of local
buckling.
3. Residual strains at impacted laminates.
A sixteen plies crossply CFRP laminate was built from UD prepreg material by OOA (Out of
Autoclave) procedures. A poliimide coated optical fiber was embedded inside the laminate
during layup. (Figure 4 A)
The laminate was impacted by a drop weight test, and a delamination was produced, as was
verified by ultrasonic C-Scan (Figure 4 B, green spot). The white line show the position of the
optical fibre, and the lower image (figure 4 C) show the strains measured by the optical fiber
along this line. It can be seen the appearance of residual strains at the delaminated area. In
fact, the delaminated area can nearly be plotted if the optical fiber follows a crooked path,
with parallel fibers every 5 mm. the strain field map of the area can be obtained with
relatively high accuracy (fig 4 D).
Figure 4. Upper left A. Composite laminate with an optical fiber following parallel paths
Upper right B. Ultrasonic C-Scan of the impacted laminate
Lower right C. Strains measured along the with line of the upper figure by an O.F.
Lower left D Strain mapping at the delaminated área, obtained by a set of parallel O.F.
4. Detection of delaminations at laminate edges
The former approach may be used for SHM of similar of structures, like small cylindrical
pressure vessels, or for structural details, like monitoring stringer debondings, but for
practical reasons the whole surface of the aircraft can not be covered with a continuous optical
fiber, the maximum inspectable length is about 100 meters. The proposed approach is to
reduce the covered area to critical regions with a higher risk of damage.
Laminate edges, like surroundings of cargo doors and man holes, are areas of high risk for
accidental impacts, and consequently require a more frequent inspection; a permanent
automated inspection system is highly desirable. The following experiments were done to
demonstrate the validity and reliability of the approach.
Several identical CFRP 16 plies laminates were built from UD prepreg material, by OOA
procedures, with the lay-up (04, 904)s. This special layup sequence was used for simplicity, to
have only two delaminations interfaces; nevertheless the concept is also working for any other
general laminate. Dimension of the cured laminate was 200 mm X 100 mm. An optical fiber
was bonded at the surface of the cured laminate, as sketched at figure 5.
The laminates were submitted to impacts of controlled energy, by using a drop weight test
machine, in both directions, perpendicular to the laminate and on-edge direction (figure 6).
The energy was gradually increased until a visible damage was produced (figure 7), and the
residual strains were recorded after every impact (figure 8)
Figure 5. Laminate with an optical fiber bonded at the surface, for the edge
delamination tests
Figure 6. Local impacts, using a drop weight test machine
Figure 7. Delamination caused at a (04, 904)s CFRP laminate after a 5 J on-edge impact
Figure 8. Strain recorded by the optical fiber after successive impacts of increasing
energy, from 2 to 5 joules. The two identifiable peaks recorded at 100 and 400 mm are
for the first and second loops of the optical fiber, respectively
From these data an strain map of residual strains may be plotted, as shown at figure 9; It may
be seen how the delaminated area increase in size for higher impacts, and that strains are
higher at the surface closer to the delamination, as expected.
Figure 9. Plot of the residual strains caused by on-edge impacts, obtained by the OFDR
system
Figure 10 Up.
Delamination caused by a normal impact, as obtained by ultrasonic
C-Scan ( at the right of the image, a B-Scan is given)
Figure 10 Down. Map of the residual strains after the impact, as measured by an OF.
Similar findings were obtained when impacts were done at the direction normal to the
laminate, figure 10. For these testss, an embedded O.F. was used, located at the second ply of
the surface opposed to the impact. The energy needed to cause a BVID was slightly higher
than in the former case.
5. Conclusions
The following conclusions can be drawn from the former experiments:
•
•
•
•
•
•
•
•
Fiber optics are excellent, fully proven strain sensors.
They are not ‘damage sensors’. Damage can be inferred only by comparing the
collected data to those coming from pristine structure
A crack may drop significantly the strength of the structure, acting as the failure
initiation point, but it only alters locally the strain field around the crack; damage may
go undetected unless a sensor was located just there.
Distributed sensing will detect the crack as far as it crosses the optical fiber, or when it
produces a strain change on the O.F
Delaminations in laminates produce a liberation of residual strain field, and they will
be detected by an O.F located there.
This damage detection method is robust and reliable (at least equivalent to CVM), for
impact prone areas, like doors surroundings areas and man-holes
This technique (detection of impact-created residual strains) is based on detectable
strain changes at the optical fiber. Structure does not need to be loaded (overnight
inspections)
Damage has to happen on the optical fiber path
This approach is fully independent, and conceptually different, from those based on
submitting the structure to external loads, and comparing the response of the structure to
these loads, before and after damage, that were also developed by our group and presented
elsewhere [4]. Algorithms to get information about damage from the slight changes
produced at other sensors are being developed, but improvements in the resolution (size of
damage, distance to sensors) are still needed.
References
1. SAE “Guidelines for Implementation of Structural Health Monitoring on Fixed Wing
Aircraft”, SAE Standard: ARP6461 (2013)
2. Güemes, A, Fernandez-Lopez, A, Soller B. “Optical Fiber Distributed Sensing.
Physical Principles and Applications” J. Structural Health Monitoring, Vol. 9, No.
3, 233- 245. (2010)
3. J. Sierra, M.-A. Torres-Arredondo, A. Guemes, L. Mujica and C.P. Fritzen “Structural
Health Monitoring of Wind Turbine Blades From Distributed Strain Measurements”.
In Proceedings of the 6th World Conference on Structural Control and Monitoring,
2014, Barcelona, Spain.
4. J. Sierra, A. Güemes and L. Mujica. Damage Detection by Using FBGs and Strain
Field Pattern Recognition Techniques. Smart Materials and Structures. 2501125020. 2013
Simulation as a Prerequisite in Structural Health
C. Boller1, D.R. Mahapatra2, R. Sridaran Venkat1, N.B. Ravi2, N.Chakraborty2
1
Chair of Non-Destructive Testing and Quality Assurance (LZfPQ)
Saarland University, Saarbrücken/Germany
2
Department of Aerospace Engineering, Indian Institute of Science, Bangalore 560012, India
Contact: c.boller@mx.uni-saarland.de
Inspection performed today needs to work at highest reliability. In non-destructive testing (NDT) this
is done by achieving the highest level of probability of detection (POD) with the inspection method of
choice, moving the respective transducer as much around until hopefully 100% coverage of the
volume to be inspected is achieved. With structural health monitoring (SHM) this is not achievable
that way since transducers are generally fixed regarding their position once an SHM system has been
implemented onto or into a structure. Hence the optimum transducer configuration has to be found by
other means and a prospective means in that regard is simulation. Different simulation approaches
have been proposed in the past with some being very much focused on and limited to the application
considered while others are gradually finding their way to become generally recognised and a part of
simulation packages on the verge of commercialization. However, those simulation tools need to work
in close to real time in case they should be accepted by the scientific and engineering community and
should make the appropriate contribution regarding a scheme on how to determine POD for SHM
systems. Within the INDEUS project there has been a first attempt to establish a platform which not
only allows to cluster different tools applied for SHM simulation but also to align those into a
structural simulation process chain starting from simulation of loads, the geometry, strength including
fatigue behaviour and a resulting probability of damage and finally ending up with the ability to be
inspected and a resulting SHM system meeting the optimum in terms of POD. In that regard different
case studies will be addressed that will be presented in more detail in further presentations.
SHM System Simulation based Design Considering Composite Patch Repaired
Stiffened Panel of Aircrafts
Rakesh Shivamurthy, Keerthy M Simon, Nitin Balajee Ravi, Nibir Chakraborty Debiprosad Roy
Mahapatra*
Department of Aerospace Engineering,
Indian Institute of Science, Bengaluru, India
Contact: rakeshs@aero.iisc.ernet.in, keerthy@civil.iisc.ernet.in, nitinb@aero.iisc.ernet.in,
nibir.chakraborty@aero.iisc.ernet.in, droymahapatra@aero.iisc.ernet.in
Sometimes repair of a structural component is a cost-effective option as compared to complete dropout of the component or the entire structural assembly. Combining with this option certain health
monitoring capability to enhance the reliability of the repaired structure is an attractive choice. The
present paper reports on the verification and validation effort and on the effectiveness of a structural
health monitoring (SHM) sensor system designed to monitor fatigue crack growth in a post-repaired
structural component such as aircraft fuselage stiffened panel. The SHM system design is developed
through SHM simulation including probabilistic modeling of fatigue damage. The damage tolerant
stiffened panel is subjected to crack growth followed by a composite patch repair. We simulate
probability of fatigue damages (PODm) and the probable locations of damage growth are analyzed for
SHM sensor placement. This consideration is further analyzed for its effectiveness for monitoring
growth of pre-existing crack and any new cracks after repair of the structure. The behavior of the unrepaired and repaired panel with crack is investigated. Complexity of crack growth behavior increases
in a repaired panel and the associated advantages in deploying SHM sensor system on it are evaluated
with example simulation of sensor signals and lab-scale testing with the SHM system as integrated in
the test panel. Further, attempts are made to obtain insight regarding the influence of repair on
damage resistance and fatigue life of repaired stiffened panel from the SHM sensor data.
Feasibility Study of SHM Simulation based Design of Accelerated Fatigue Tests
Nitin Balajee Ravi1, Nibir Chakraborty1, Rakesh Shivamurthy1, Keerthy M Simon1, Ramanan Sridaran
Venkat2, Mirko Steckel3, Debiprosad Roy Mahapatra1, Christian Boller 2
1
Department of Aerospace Engineering, Indian Institute of Science, Bengaluru, India
2
Chair of Non-Destructive Testing and Quality Assurance (LZfPQ), University of Saarland, Campus
Dudweiler, 66125 Saarbrücken, Germany
3 IMA Materialforschung und Anwendungstechnik GmbH, 01109 Dresden, Germany
Email: nitinb@aero.iisc.ernet.in, nibir.chakraborty@aero.iisc.ernet.in, rakeshs@aero.iisc.ernet.in,
keerthy@civil.iisc.ernet.in, ramanan.sridaran@uni‐saarland.de,
Mirko.Steckel@ima-dresden.de, droymahapatra@aero.iisc.ernet.in,
c.boller@mx-uni‐saarland.de
This paper studies the advantages of simulation in extrapolating the crack growth curve based on
applied load and its influence in accelerating the fatigue testing. The studies were conducted on a
stiffened panel subjected to fatigue loading. The crack growth rate was simulated using physics based
model. The current challenges faced in this simulation were twofold – (i) to estimate the number of
cycles required for the propagating crack to reach a particular size when the structural response and
load path is complicated and (ii) to predict the crack growth rate reliably from the data generated by
the SHM system where the SHM system has been designed with the help of simulation. In this study,
the loading condition was based on design load spectrum used to validate damage tolerance design
criteria of a stiffened aircraft. Crack initiation and propagation have been investigated in detail based
on fatigue simulation results obtained within a probabilistic framework. The crack growth curves
were obtained from simulation based correlation derived from data obtained along an actual test
carried out at the IMA Dresden test facility on a stiffened panel while actual crack growth was
monitored visually and from strain gauge data. The SHM system used involved a network of piezo
transducers, which were placed on the panel surface to generate guided waves and were deployed
during the test to monitor fatigue crack growth.
Comparison of predicted crack growth and crack lengths observed have been used as an attempt to
understand the reliability of such an SHM system. The crack growth rate as monitored has been
incorporated in an analytical form to obtain a relationship with the stress amplitude applied.
Following this type of parametric estimation, we performed fatigue simulation at a different stress
level on the identical component with same load case to validate the accelerated crack growth from an
accelerated fatigue testing point of view.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Simulation as Key Enabler to Support ISHM Certification
Dipl.-Ing. Matthias BUDERATH1, Partha ADHIKARI2, Harsha GURURAJA RAO2
1
Airbus Defence and Space, Rechliner Straße, 85077 – Manching, Germany
E-mail: matthias.buderath@airbus.com
2
Airbus India Engineering, Bangalore, 560 048, India;
E-mail: partha.p.adhikari@airbus.com, harsha.rao@airbus.com
Abstract
Integrated system health monitoring and management (ISHM) is a field of research and development where lot
of different industries and academia are highly focused on. Since the ISHM technology itself is still evolving,
the standards available for certification and successful qualification of the systems are yet to be fully matured.
And these different qualification methods and processes have to be included in the early stages of the
development of the respective systems. This paper highlights the survey of different certification methodologies,
provides an insight into Airbus Defence & Space’s Certification Roadmap, the role of ISHM Simulation
Framework in certification and also the lessons learnt. The paper will also provide an outlook how Structural
Health Monitoring (SHM) technologies can be addressed in the ISHM simulation environment to define the
monitoring concept to contribute to the requirement of probability of detection.
Keywords: ISHM, SHM, Simulation, Certification, CBM, Prognostics, Enhanced Diagnostics
1. Introduction
With growing financial uncertainty, air vehicle operators (both commercial and military) are
under tremendous pressure to reduce operational and support costs. It is accepted across the
aerospace industry that ISHM is a potentially valuable strategy for the manufacturing and
management of vehicle platforms. At the same time, ISHM has not yet fully matured as a
technology in several key functional areas. Research and development to address this
shortfall is occurring across both the automobile and aerospace industries. Although
technologies related to Built–In-Test (BIT) and diagnostics have advanced greatly and
research into enhanced diagnostics are progressing very fast, prognostics technology for all
types of aircraft sub-systems are still in a very nascent stage.
Validation & Verification (V&V) method leading to the qualification and certification of
ISHM is a key area of development. Although there has been considerable effort in this
direction, ISHM system at the aircraft level is yet to be certified. Certification agencies
(EASA, FAA, SAE, etc.) are yet to establish comprehensive certification regulation for
Integrated System Health Monitoring system.
Deployment of ISHM in an aircraft and the resulting qualification process demands a huge
investment. Verification and validation of these ISHM technologies is an important step in
building confidence, both qualitatively and quantitatively. Practically, the cost of correcting
an error after fielding an ISHM system is dramatically greater than that of in the testing
phase, thus highlighting the need for appropriate verification and validation techniques.
Certification considerations must be addressed during the very early stages of technology
development in order to successfully meet any significant qualification goals. Appropriate
guidelines and strategies should be followed in ISHM technology development to ensure
successful certification within the desired time frame. Additionally, trade studies in the
selection of V&V platforms reduce the eventual cost of V&V processes. This paper focuses
on development of such guidelines for the V&V process while emphasizing the relevance of
ISHM simulation frameworks, and a well devised certification roadmap.
2. Main Objective of Simulation Framework
Airbus DS has developed a comprehensive integrated ISHM simulation framework which
contributes in the following areas:
 Integrated demonstration of Proof of Enablers (PoE)
 Training & maturity of PHM functions
 Maturation of ISHM requirements (KPIs)
 V&V of ISHM functions
This ISHM framework is used primarily for demonstrating Proof of Enablers (PoE) and
System Integration Laboratory (SIL) testing, including S/W and H/W in loop, which is the
goal of concept refinement and technology development. For end-to-end demonstration of
ISHM, simulation framework hosts simulation of aircraft system with fault injection
provision, on-board health assessment function, off-board analytics related to prognostics,
operational risk assessment, database management, fleet planning, maintenance/logistics
planning, etc. in enterprise level. For SHM, Airbus Defence & Space is addressing the
virtualization of structural components to be monitored by SHM and facing the challenge to
model the sensing system and the insertion of failure modes.
During early stage of ISHM development for new aircraft platforms, sufficient amount of inservice or test flight data (both nominal behaviour & fault behaviour, run-to-fail data) is not
available. Physics based simulation of aircraft system and fault progression model plays an
important role for modelling enhanced diagnostics & prognostics modules/functions.
Simulation models and PHM functions undergo continuous evolution of maturity with data
(rig data, test flight, in-service flight data) available through progress of development
lifecycle. Simulation framework has the mechanism to accommodate additional correction
factor related to modelling imperfection.
User objective and metrics related to ISHM can be refined through exhaustive Monte-Carlo
simulation of off-nominal scenarios, which is not a viable solution with real flight tests.
Simulation framework supports functional analysis related to selection of candidate subsystems, faults, sensors and performance matrices of enhanced diagnostics and prognostics.
This will enrich performance requirements of key algorithms mainly related to enhanced
diagnostics, prognostics, etc.
With the increase in maturity of the Simulation framework, it plays different roles of V& V
platforms viz. Engineering simulator, System-subsystem Test setup, Integration Test Setup,
etc. Ground based ISHM systems can be deployed in this environment. This framework with
high fidelity modelling of sub-systems and sensor data provides enough confidence in
installation of on-board ISHM non-critical systems before controlled introduction to service
for further tuning & refinement of algorithm. Integrated HILS will have simulation of
Aircraft Dynamics, Aircraft Subsystem H/W and adverse environmental effects. Also, there
is the capability to inject system faults. This facility can expedite the validation process of
ISHM and reduce validation time period during Controlled Introduction to Service. However
this capability demands a huge investment of time and capital. These investments can be
greatly reduced in case of V&V of aircraft’s ISHM by utilization of Simulation Framework.
3. ISHM Certification Guideline
3.1
Certification Basis
Certification agencies (EASA, FAA, SAE, etc.) are yet to establish comprehensive
certification regulation for Integrated System Health Monitoring system. This section
summarizes existing efforts for certification basis which will act as an overall guideline for
ISHM development.
Kevin R. Wheeler et al. [15] contribute to an extensive survey of recent ISHM programs and
mentions that vast differences in user objectives with regard to engineering development is
the major barrier for successful V&V. The paper identifies in detail the objectives and
associated metrics across operational, regulatory and engineering domains for diagnosis and
prognosis algorithms and systems.
James E. Dzakowic et al. [13] introduces a methodology for verifying and validating the
capabilities of detection, diagnostic and prognostic algorithms through an on-line metrics
based evaluation.
Martin S. Feather [16] mentions in his publication that state-of-the-practice V&V and
certification techniques will not suffice for emerging forms of ISHM systems. However, a
number of maturing software engineering assurance technologies show particular promise for
addressing these ISHM V&V challenges.
Dimitry Gorinevsky et al. [8] describes the importance of a NASA-led effort in open system
IVHM architecture. Detailed functional decompositions of IHM systems with respect to
criticality, on/off board operation and development cost are presented and certification
standards are mapped accordingly. This paper also addresses the current NASA IVHM test
bed along with development and deployment steps corresponding to increasing TRL.
The FAA’s advisory circular (AC), AC 29-2C MG-15, provides guidance in achieving
airworthiness approval for rotorcraft Health and Usage Monitoring System (HUMS)
installations. It also outlines the process of credit validation, and Instructions for Continued
Airworthiness (ICA) for the full range of HUMS applications.
Brian D Larder et al. [7] converted the text of AC 29-2C MG-15 into a flow chart. His
intention was to define the generic end-to-end certification process for HUMS CBM credit.
Further, he sought to identify the relationships and interactions between different elements of
the certification process that are contained in the three separate sections of the AC
(installation, credit validation, and Instructions for Continued Airworthiness). This paper also
mentions that HUMS have achieved very few credits, and that the material in the AC is
largely untested. However HUMS in-service experience shows that the potential for future
credits does exist.
ADS-79E HDBK [3] describes the US Army‘s Condition Based Maintenance (CBM) system
and defines the overall guidance necessary to achieve CBM goals for Army aircraft systems
and Unmanned Aircraft Systems (UAS).
Praneet Menon et al. [19] published a paper, which summarizes the work of a Vertical Lift
Consortium Industry Team to provide the detailed guidance for the Verification and
Validation (V&V) of CBM Maintenance Credits.
Existing ARPs (viz. ARP 5783, ARP 4761, etc. ) published by SAE already supports some
aspect of guidance in different stages of ISHM development.
SAE formed an Integrated Vehicle Health Management (IVHM) Steering Group to explore
the needs for standardization to support IVHM technology towards the following objectives.




the development of a single definition and taxonomy of IVHM to be used by the
aerospace and IVHM communities,
the identification of how and where IVHM could be implemented,
the development of a roadmap for IVHM standards,
and the identification of future IVHM technological and regulatory needs.
The following figure summarizes existing ARPs and standards in the different stages of
ISHM process flow.
Figure 3-1: ISHM Process Flow mapping with ARPs and standards
3.2
Guideline for Life Cycles
ISHM development process steps are mapped onto the ARP 4754A aircraft/system
development life cycle as baseline. The ISHM development process steps can be mapped to
the V-model of the ARP 4754A process with the exception of “Concept & Technology
Development” and “Controlled Introduction to Service”, “Instructions for Continued
Airworthiness (ICA)”, and “In-Service Validation” processes steps, which are outside of the
V-model. It is assumed that after the transfer from R&T the ISHM development will be part
of an aircraft system development. This concludes that the ISHM development is part of an
overall aircraft and system development process. The detailed guidelines for all processes
will be available in the respective standards as mentioned in Figure 3-1.
Figure 3-2: Mapping of ISHM development process steps on to ARP4754 Aircraft
development process
3.3
ISHM Simulation Framework
The goal of ISHM system are preparation of intelligent Maintenance Plan, intelligence
Mission Plan and automatic logistics function for enhancing availability, maintainability and
mission capabilities. These functions are achieved through Condition Based Maintenance
(CBM). The Simulation Framework, which is built around OSA-CBM and OSA-EAI
architecture, simulates all ISHM functional layers through different sub-system models
Prognostic Health Management (PHM) is the core of ISHM technology. Like in any other
domain, challenges in the introduction of PHM systems in the aerospace domain are twofold.
On one hand, there are individual challenges in developing sensor technology, state detection
and health assessment methodologies and models for determining the future life span of a
(possibly deteriorated) component. On the other hand, there are integration challenges when
turning heterogeneous data from disparate and distributed sources into consolidated
information and dependable decision support on aircraft and fleet level. It has therefore been
recognized in the community that standardized and open data management solutions are
crucial to the success of PHM. Such a standard should introduce a commonly accepted
framework for data representation, data communication and data storage.
This simulation framework supports key features, viz. demonstration of end-to-end value
chain of ISHM, real-time simulation for the on-board computation, almost real-time for offboard, having features of simulating lifetime (with time acceleration mode), provision for
refinement of physical models with data from test rigs, test flights with seeded faults and inservice data.
Figure 3-3: ISHM Simulation Framework
ISHM Simulation Framework simulates following modules:

Aircraft System Model

On-board ISHM

On-ground ISHM

Supply Chain (Enterprise Level)

Simulation Management
Simulation of Aircraft system model and supply chain (Enterprise Level) create simulation
environment for ISHM system models and simulation management controls the operation of
complete ISHM Simulation Framework.
3.3.1 Aircraft System Model
Aircraft System Model simulates those systems and their sensors for which we intend to
develop ISHM capabilities. Aircraft System Model have high fidelity modelling of Aircraft
aerodynamics model, Hydraulics / Actuator System Model, Landing Gear, Fuel, ECS and
Aircraft Structure, etc. Each sub-system implements physics based modelling of dynamic
behaviour, physics of fault, and computation of states or parameters for deriving sensor data
for each sub-system. Sensor data for each sub-system are generated from computed states and
parameters after corrupting with all possible errors that might occur in real-life scenario, as
well as with noise specific to those sensors. All faults are injected from simulation control
GUI. Any system, for which ISHM specific monitoring and prediction capabilities should be
validated and verified, needs to be modelled with a high level of detail. This should enable
the realistic simulation of failures to support the validation of diagnostic and prognostic
functions. Respective controller model simulates Built-in-Test (BIT) and Reactive Health
Assessment (RHA) of the sub-system.
3.3.2 On-board ISHM
On-board ISHM function includes a central ISHM data processor. Sensors push their data to
the IVHM data processor via an OSA-CBM implementation. The underlying message
protocol is optimized for embedded systems. The ISHM data processor calculates ISHM
information according to the OSA-CBM layer specifications, up to health assessment layer.
As per OSA-CBM, there are seven functional layers. Central ISHM data processor has
following functions:




First four functions of OSA-CBM
o Data Acquisition
o Data Manipulation
o State Detection
o Health Assessment
High Level Reasoning
BIT Function
Storing of on-board health data
Several seeded fault tests under fixed conditions are sufficient to enable the model-based
development of diagnostic functions. The development of prognostic functions (to be part of
ground based ISHM) needs also to cover the development of suitable failure mode specific
degradation models. Once the degradation models have been developed, it is possible to
verify the diagnostic and prognostic functions through Monte-Carlo simulations. These
simulations should include stochastic fault insertion for so-called "hard faults" (stochastically
occurring failures without impacts on observable system parameters before the specified
failure threshold is exceeded) and the usage of degradation models for "soft faults"
(stochastically occurring degradations with impacts on observable system parameters before
the specified failure threshold is exceeded). This concept is illustrated in Figure 3-4.
Figure 3-4: Fault simulation concept for Simulation Framework
3.3.3 On-ground ISHM
Major functionalities towards enhancing availability, maintainability and mission capabilities
related to ISHM system are realized by ground based sub-systems. On-board ISHM function
includes only data acquisition and diagnostic function of equipment health along with
intermediate processing of data. Ground based ISHM system has significant amount of
processing related to the following prime functions (Fatih Camci et al. [10] ):









On Ground Heath Management function
Operational Risk Assessment / Fleet High Level Reasoning
Maintenance Management
Maintenance Planer
Resource / Logistic Management
Mission Planer
Learning Agent
Simulation of Enterprise System
Presentation Layer
On Ground Health Management function:
On ground health management function consists of advanced diagnostic and predictive
analysis. Advanced diagnostic validates further on-board diagnostic result with historical data
of same aircraft and fleet wide fault data base and refine diagnostic decision. Advanced
prognostic computes RUL & Confidence for CBM candidate. Predictive Analysis (Trend
analysis) identifies impending failure using trend analysis of historically collected data, but
does not predict when failure will occur.
Maintenance Management:
Maintenance Management functions finds one of the following maintenance solutions for a
sub-system depending upon RCM process:
 Corrective

Preventive

CBM
Figure 3-5: Maintenance Strategies including Predictive / Condition based Maintenance
Maintenance Management executes the following functions:
 Identification of Maintenance task corresponding to sub-system / functional failure
 Rank of optimal maintenance task is computed as a function of maintenance
effectiveness for the failure mode, maintenance downtime and cost.
 Execute Maintenance (work order generation, Track Maintenance action, Receive
feedback and close work order) as per approved maintenance plan
Maintenance Planer:
Opportunistic Maintenance agent finds opportunistic maintenance time and task using rank of
maintenance task, mission capability of sub-system / function for future mission, RUL for
future missions. Maintenance planner schedules the intelligent maintenance plan, validates
with feedback from Resource Management and publishes maintenance plan after getting
approval from decision support system.
Resource / Logistic Management:
This function tracks the availability along with configuration parameters of LRUs, tools,
parts, consumables and personnel, etc. (configurable items). On the receipt of maintenance
plan, Resource / Logistic management function sends feedback on validity of maintenance
plan to Maintenance Planner on the basis of resource availability. This function finally
generates a plan for resource / inventory and generates order for parts or LRUs to OEMs or
suppliers as per present and projected status of inventory.
Mission Planner:
Mission Plans & Flying Programmes are entered using digital map and editing GUI. Mission
planner instructs user to reschedule the Mission Plan if performance of aircraft exceeds as per
mission plan entered and edited. Flying programs are asked to reschedule if approved
maintenance plan superimposes with mission plan. Applicability of mission segments of a
particular aircraft is checked further with respect to operational capabilities of the aircraft for
the segment, computed by Operational Risk Assessment (ORA). If capability of flight
segment or complete mission is less than critical threshold, Mission Planner instructs user to
reschedule or cancel the mission for particular Aircraft.
Learning Agent:
As experience is accumulated, some of the parameters within the model can be learned
automatically by analyzing the feedback from the maintainer, OEM industry, Mission
Commander, Resource Manager. The parameters to be learned are opportunistic maintenance
threshold, required maintenance threshold, resource lead time, maintenance effectiveness and
different co-efficient related to diagnostics & prognostics, etc.
Simulation of Enterprise System:
This module simulates supply of specific LRUs or parts from OEM, Service/Industry Support
organization, Wholesale Stock point accounting appropriate accumulated delay attributed
due to order process by resource management function, manufacturing (if applicable),
shipping process, etc. related to Supply Chain Management.
Presentation Layer:
Decision support personal interacts through Presentation Layer which consists of following
GUIs distributed across different terminals.
 Health Management & Monitoring

Interactive GUI for Maintenance Management

Resource Management & Monitoring

Maintenance Planner

Mission Planner
High Level Reasoning / Operational Risk Assessment:
High Level Reasoning (HLR) is the capability that can estimate an airplane’s (or vehicle’s)
functional availability. The purpose of HLR concept is used to estimate the functional
availability of a vehicle based on the health assessment results from lower level systems and
subsystems. Both concepts are part of the HLR development and integration into the
simulation framework. RUL & confidence is recomputed for each component failure for all
future missions and used by HLR. ORA finally determines and quantifies remaining
functional / operational availability at the subsystem, vehicle levels and mission levels.
3.3.4 Key Tools related to ISHM Design
3.3.4.1
Functional Failure Analysis Tool
The first step in ISHM capability design is to clearly define how the vehicle and its
subsystems function and how they can potentially fail. A clear understanding and
representation of the functions to be accomplished provides the framework for capturing how
a system can fail, the manifestations of the failure, its consequences, and its impact on the
vehicle as a whole. FFA tool plays an important role in the development of ISHM System,
CBM candidate selection, refining system performance matrices, trade-off study in the design
of ISHM architecture.
Figure 3-5: Concept of Functional Failure Analysis
3.3.4.2
Maintenance Strategy Tool
Maintenance Strategy [1] aims to map all fault modes at individual and LRU levels to
different maintenance categories. RCM analysis is the foundation to establish a framework
for candidate selection. The decision logic is based on existing guidelines: SAE JA1011, SAE
JA1012, NAVAIR 00-25-403 and ATA MSG-3 with suitable modification. After fault
consequence check, maintenance options for each fault type of a LRU are short listed based
on technical feasibility only. Cost effectiveness and risk are computed for each selected
option of the fault type. Best maintenance option or combinations of options are selected for
LRU by solving optimization problem which maximizes availability, ROI of selected option
and minimizes risk at the LRU level.
Figure 3-6: Different Categories of Optimization for ISHM
4. Role of ISHM Framework in Certification
The Figure 4-1 depicts the V&V road map of ISHM with increasing Technology Readiness
Level. On the basis of earlier discussion, V&V process towards airworthiness certification of
ISHM will be spread over the following phases:




Concept Refinement & Technology Development
Development
Controlled Introduction to Service
Instruction for Continuous Airworthiness
From the V&V roadmap [6], it is very much evident that different facilities are needed
towards V&V, certification & qualification of ISHM technologies. ISHM simulation
Framework plays multi-role being as a single platform.
Figure 4-1: V&V Roadmap of ISHM and role of Simulation Framework
5. Lessons Learnt
Key findings through the development of Airbus DS ISHM program and Simulation
Framework as validation platform are summarized here.

Enhanced Diagnostics (to compute heath index of degrading sub-system component),
enhanced fault isolation (to isolate incipient faults), heath aggregation in higher levels
(sub-system, system, operation) and prognostics (to compute remaining useful life) are
key enablers for PHM functions.

During early stage of ISHM life cycle, in the absence of in-service/test flight data, physics
based model plays an important role. Physics based model undergoes continuous maturity
with the help of in-service data

Once put into service, the PHM functions have to be continuously validated to detect
potential drifts between initial design / implementation and real life behaviour [14] .

Initial PHM design/model (technical choices: algorithms, learning database, etc.) is
subjected to potential changes in operational (change in the operational cycles, loading of
the system) and environmental (heat, humidity for instance) conditions [14].

The potential evolution of the maintenance operations may have an impact on the validity
of PHM functions [14].

The integration of structural health monitoring is still at a low TRL compared to system
because of the complexity of modeling the characteristics of the SHM sensing system, the
damage characteristic of composite material and required processing, diagnostic and
prognostic capabilities.
Few of the important insights related to Simulation and ISHM certification are given in the
Figure 5-1.
Figure 5-1: ISHM Simulation Framework challenges & solution approaches
6. Conclusion
The survey of works towards ISHM certification, suggested customization and experiences
support SHM development as well. However, there exists a significant challenge in
certification of SHM, particularly for composite structures. Modelling of sensors, Fault
Progression model are some of the key challenging areas. In general, it is evident that nature
of challenges in V&V and certification of ISHM is different compared to standard standalone system. One of the major challenges in certification of ISHM system is due to nonavailability of comprehensive regulatory standards for ISHM. V&V also poses challenges
mainly due to the fact that ISHM has to handle a large number of off-nominal scenarios, has
to ensure performance, safety, and reliability across the entire performance envelope and has
to reliably avoid ‘false alarm’. Moreover, V&V has to deal with multidisciplinary aspects of
ISHM. Most prominent aspect is gathering of direct evidence for faults effects related to
V&V of enhanced diagnostics and prognostics. To handle these issues, the key aspects of
ISHM V&V mentioned above are summarized here:

V&V maturity starts from concept refinement and technology development phase.

If specific sub-system / function of ISHM, is classified as Hazardous/Severe Major, then
direct evidence must be gathered. (FAA’s advisory circular AC 29-2C MG-15).

If specific sub-system / function of ISHM, is classified as Major or Lower, then indirect
evidence is sufficient. (FAA’s advisory circular AC 29-2C MG-15).

During ‘Controlled Introduction to Service’, CBM maintenance credit is considered as
maintenance benefit. i.e. CBM output is compared with maintenance instructions
suggested by conventional RCM process.
 After maturation of algorithm and certification, CBM obtains maintenance credit.
 Appropriate sequence of V&V process of ISHM function layers are to be considered.
 It must be noted that the V&V of ISHM functionalities in Simulation Framework do not
completely address defects created by designer. It is evident from Figure 4-1 (V&V
Roadmap with increasing TRL) that subsequent V&V phases (i.e. V&V in integration
RIG, Integrated HILS, V&V during controlled introduction to the service and ICA) are
suggested in order to achieve maintenance credit.
 Since ISHM simulation framework plays vital role in V&V process, simulation framework
has to be qualified (Robert G. Sargent. [20]).
This study may give enough confidence to ISHM community towards achieving maintenance
credit through implementation of this technology.
7. Acknowledgments
The authors are grateful to Ernst Tobias & Andreas Loehr for their valuable inputs. We are
also grateful to reviewers for many of the improvements to the document.
Nomenclature
AC
Advisory Circular
AMC Acceptable Means of Compliance
ARP
Aerospace Recommended Practice
AWR Airworthiness Report
BIT
Build-In Test
CBM Condition Based Maintenance
CC
Certification Coordinator
CS
Certification Specification
EAI
Enterprise Application Integration
FHA
Functional Hazard Analysis
FMECA Failure Modes, Effects, and Criticality Analysis
GUI
Graphical User Interface
HILS
Hardware in Loop Simulation
HLR
High Level Reasoning
HUMS Heath Usage Monitoring System
IA
Integrity Assessment
ICA
Instruction for Continued Airworthiness
ISHM Integrated System Health Monitoring
IVHM Integrated Vehicle Heath Monitoring
LRU
Line Replaceable Unit
OEM
Original Equipment Manufacturer
ORA
Operational Risk Assessment
OSA
Open System Architecture
PHM
Prognostic Health Management
RCM
Reliability Centered Maintenance
RUL
Remaining Useful Life
SHM
Structural Health Monitoring
TRL
Technology Readiness Level
References
1. Adhikari P. P., Buderath M, 'A Framework for Aircraft Maintenance Strategy
including CBM', European Conference of the Prognostics and Health Management
Society, Bilbao, Spain, 2016.
2. Adhikari P. P., Makhecha D., Buderath M., ‘A Certifiable Approach towards
Integrated Solution for Aircraft Readiness Management’, Second European
Conference of the Prognostics and Health Management Society, Nantes, France,2014.
3. ADS-79E-HDBK, ‘Aeronautical Design Standard Handbook. Condition Based
Maintenance System for. US Army’, 8 February 2016.
4. ATA MSG-3, ‘Operator/Manufacturer Scheduled Maintenance Development’, issued
in 1980, revised in 1993.
5. A. Hess, G. Calvello, and T. Dabney, ‘PHM a key enabler for the JSF autonomic
logistics support concept’, IEEE Aerospace Conference,2004.
6. Buderath M. & Adhikari P. P., ‘Simulation Framework and Certification Guidance for
Condition Monitoring and Prognostic Health Management’, European Conference of
Prognostics and Health Management Society 2012.
7. Brian D Larder, Mark W Davis, ‘HUMS Condition Based Maintenance Credit
Validation’, American Helicopter Society 63rd Annual Forum, Virginia Beach,
VA,2007
8. Dimitry Gorinevsky, Azary Smotrich, Robert Mah, Ashok Srivastava, Kirby Keller
& Tim Felke, ‘Open Architecture for Integrated Vehicle Health Management’, AIAA
Infotech @ Aerospace, Atlanta, GA,2010.
9. FAA Advisory Circular 29-2C MG 15, ‘Airworthiness Standards Transport Category
Rotorcraft’.
10. Fatih Camci, G. Scott Valentine, Kelly Navarra, ‘Methodologies for Integration of
PHM Systems with Maintenance Data’, IEEEAC paper #1191, Version 1, 2006.
11. Gorinevsky, D., Smotrich, A., Mah, R., Srivastava A., Keller, K., & Felke, T, ‘Open
Architecture for Integrated Vehicle Health Management’, AAIA Infotech@Aerospace
Conference, 2010.
12. HAHN Spring Limited. Development, ‘Validation, Qualification and Certification of
Structural Health Monitoring Systems’, HAHN Spring Report 1/B002, 2011.
13. James E. Dzakowic, G. Scott Valentine, ‘Advanced Techniques for the verification
and validation of prognostics & health management capabilities. Impact
Technologies’, LLC, 2004.
14. Jayant Sen Gupta, Christain Trinquier, Kamal Medjaher, Noureddine Zerhouni,
'Continuous validation of the PHM function in aircraft industry', The First
International Conference on Reliability Systems Engineering & 2015 Prognostics and
System Health Management Conference-Beijing (2015 ICRSE & PHM-Beijing).
15. Kevin R. Wheeler, Tolga Kurtoglu, and Scott D. Poll. A, ‘Survey of Health
Management User Objectives Related to Diagnostic and Prognostic Metrics’, 2010.
16. Martin S. Feather, Lawrence Z. Markosian, ‘Emerging Technologies for V&V of
ISHM Software for Space Exploration’, IEEE Aerospace Conference paper #1441, V2, 2005.
17. MASAAG Paper 123, 'Development, Validation, Verification and Certification of
Structural Health Monitoring Systems for Military Aircraft', Issue 2a, 6th January
2016
18. O Benedettini, T S Baines, HWLightfoot, and RMGreenough, ‘State-of-the-art in
integrated vehicle health management’, 2008.
19. Praneet Menon, Bob Robinson, Mike August, Terry Larchuk & Jack Zhao,
‘Verification and Validation Process for CBM Maintenance Credits’, American
Helicopter Society 67th Annual Forum, 2011.
20. Robert G. Sargent, ‘Verification and Validation of Simulation Model’, Simulation
Research Group, Syracuse University, 1998.
21. SAE International, ‘Aerospace Standards News Letter’, Volume II, Issue 1,2010.
22. SAE JA1011, ‘Evaluation Criteria for Reliability-Centered Maintenance (Rcm)
Processes’, issued in 1999, revised in 2009.
23. SAE JA1012, ‘A Guide to the Reliability-Centered Maintenance (RCM) Standard’,
issued in January 2002, revised in 2011.
BIOGRAPHIES
Matthias Buderath - Aeronautical Engineer with more than 30 years of experience in
structural design, system engineering and product- and service support. Main expertise and
competence is related to system integrity management, service solution architecture and
integrated system health monitoring and management, Today he is head of Airbus Defence
and Space R&T Strategy and Senior Expert Integrated System / Aircraft Health Monitoring
and Management He is member of international Working Groups covering Through Life
Cycle Management, Integrated System Health Management and Structural Health
Management. He has published more than 80 papers in the field of Structural Health
Management, Integrated Health Monitoring and Management, Structural Integrity
Programme Management and Maintenance and Fleet Information Management Systems.
Partha Pratim Adhikari - has more than 18 years of experience in the field of IVHM,
Simulation of Aircraft Systems and Avionics. Partha has Bachelor’s degrees in Physics (H)
and B. Tech in Opto-electronics from Calcutta University and a Master’s degree in Computer
Science from Bengal Engineering and Science University. In his tenure across various
aerospace organizations, Partha made significant contributions in the fields of IVHM,
Navigation systems, Avionics and Simulation technologies. Partha published several papers
in the fields of estimation, signal processing and IVHM in national as well as international
conferences and journals. Partha, in his current role at Airbus Group India, Bangalore is
working on devising ISHM technologies for aviation systems with focus on complete vehicle
health, robust implementation and certification of the developed technologies.
Harsha Gururaja Rao – Software Engineer with more than 5 years of experience in
designing & developing enterprise level Software Applications. Harsha has a Bachelor of
Engineering Degree in Computer Science from the Visvesaraya Technological University.
Harsha, in his current role as an Engineer at Airbus India, Bangalore is working on
developing Software for IVHM technologies and other enterprise applications to meet
Aerospace & Defense Business Systems requirements.
Real-Time Tracking of Damage Growth in CFRP Composites during Testing for
Durability and Fracture
R Sunder, Murali Mohan, K Ramesh, Vishal Raina
BISS (P) Ltd, 497E, 14th Cross, 4th Phase,
Peenya Industrial Area, Bangalore 560058, India
Contact: rs@biss.in
Fatigue research on metallic materials was for over a century, hampered by the availability of just two
parameters at failure - cycle count and two fracture surfaces. The realization that fatigue is about
crack growth and the emergence of techniques to measure crack extension in fatigue and fracture
testing, combined with advances in Fracture Mechanics have completely transformed the manner in
which contemporary experiments on metal fatigue are conducted.
A variety of NDT techniques are available to characterize damage in composites. Techniques like
ultrasound C-Scan and X-ray tomography can deliver a direct, quantifiable estimate of damage by
way of area of delamination or extent of local fibre or matrix fracture. Unfortunately, these demand
the removal of the test specimen from the test system as a pre-requisite to damage characterization.
There are a number of indirect techniques such as compliance, acoustic emission and electrical
resistance measurement that are amenable to implementation while a test is in progress. However,
being essentially indirect, these merely characterize a damage parameter that does not carry physical
representation, such as area of damage. Infrared imaging does provide a direct picture of damage in
progress; however, it also is subject to an element of interpretation. Besides, it requires a certain time
interval for differential thermal response to emerge that may not necessarily be acceptable from the
viewpoint of test scheduling.
This paper describes a real-time in-situ scanning system that permits quantitative characterization of
damage whilst a mechanical test, be it static or cyclic is in progress. The scanning system is fully
integrated with the test system controls so that individual images of propagating damage can be
automatically recorded with the associated load, strain, cycle count or time. In addition, a digital
signal processing algorithm is developed that permits translation of acquired damage image into
damage size represented in terms of area. These together permit the characterization of damage
kinetics in much the same manner as conventional crack growth test software permits characterization
of crack growth rate.
Work is in progress to collect damage growth rate data on CFRP coupons carrying controlled initial
drop-weight impact damage in an attempt to characterize damage kinetics under a variety of loading
and initial damage conditions. The study is expected to conclude well in time to present substantive
results at the November meeting.
Accelerated fatigue data evaluation for SHM
validation activities
Peter Starke and Christian Boller
Chair of Nondestructive Testing and Quality Assurance, Saarland University
Campus Dudweiler, Am Markt Zeile 4, 66125 Saarbrücken/Germany
Keywords: Fatigue life calculation, short-time procedure, non-destructive testing, Structural
Health Monitoring
Structures made from advanced as well as traditional materials especially in the field of
aeronautics are loaded up to levels where they can easily fracture in a controlled way
meeting the damage tolerance principle and its criteria to guarantee a safe operation.
Structural health monitoring (SHM) is a means to safeguard this damage tolerance principle
in a way that it reliably detects a damage of tolerable size in an automated way. However,
what happens if the detection algorithms in the SHM systems applied do contain a prognostic
feature and the true material condition must be principally considered unknown in the first
place? Such cases may occur in situations where aircraft have been flown over a longer
period of time and under severe conditions but where the true condition of resulting damage
is rather unknown and hence a true residual life can hardly be determined reliably. Such
situations become increasingly popular specifically with ageing military aircraft or aircraft
where the operational mission has suddenly changed. For those aircraft and their structures
it becomes increasingly difficult to define damage tolerance criteria and with this to configure
appropriate SHM systems. Principally when an SHM system has to be configured for a
specific structure it requires the damage condition of the structure to be known. This is
clearly defined when a structure is in its pristine condition. However, when it comes to ageing
aircraft or structures the damage condition becomes an unknown and this damage condition
needs to be determined in a fairly efficient way.
The essential data characterising a structure and material under fatigue is fatigue-life (S-N)
and crack propagation data. While crack propagation data is comparatively less sensitive to
an ageing process - as long as the operational (including environmental) conditions do not
change dramatically, which can be excluded for aeronautical applications - S-N data
definitely is. Generating S-N-data for an aged aeronautical structure therefore needs to be
performed in a quick and cost effective way. Furthermore the material used for
characterisation needs to come from the structure considered to be assessed. This does
therefore not allow much material to be required which all comes down to the fact that the full
dataset required has to be obtained with a minimum of effort in terms of material, time and
cost.
The reason why S-N-curves are used for structural fatigue design today is mainly devoted to
tradition and the fact that loads can be broken down to stresses and strains and it is
specifically the strains which have been able to be sensed over the past decades,
represented by electrical resistance strain gauges and other sensors of a similar nature.
However, is a strain determined from a stress a true and sufficient parameter to be used to
characterise damage in the end in a representative way? The continuous discussion for
nearly a century now with respect to the validity of the Palmgren-Miner linear damage
accumulation rule and the desire to understand the true non-linearities within a damage
accumulation process is a continuous proof to bring fatigue life evaluation out of its crude
attitude in having to argue in factors instead of percentages with respect to the fatigue lives it
can predict. In this context the discussion about appropriate damage parameters in fatigue
life evaluation and monitoring as such needs to be rest to a level of broader physical
understanding. With the option to live in a century where an increasing number of
parameters can be sensed nowadays and this even under the condition of low cost and
being an integral part of a structural component itself is like a paradigm shift in structural
mechanics being provided through SHM. The first step in that regard is to make better use of
non-destructive testing (NDT) in structural mechanics and assessment and to take
advantage of the various parameters NDT provides to find out how those parameters are
related to the non-linear damage accumulation process and how those could be sensed and
used hopefully in a more appropriate way than traditional strains would do.
In this paper an approach is described on how a complete fatigue-life dataset is determined
not just based on stresses and strains only but also on various other NDT parameters with a
comparatively limited effort in terms of material, time and cost and which can be adequately
associated with the parameters used in an SHM system, allowing the SHM system to be
‘calibrated’ to the appropriate material and structural condition it is being used for.
During the last couple of decades, different short-time procedures have been developed,
which offer the possibility to reduce the effort for generating materials data for cyclic loading
enormously. Among others the PHYsically Based FAtigue Life calculation method “PHYBAL”
is one of these [1]. PHYBAL is a short-time method for the fast calculation of S-N curves, for
the estimation of the fatigue limit and for the evaluation of the relationship between applied
load and materials’ response to this load. The application of PHYBAL leads to a reduction in
time and cost for experimentation by up to 90 % because it only requires three instead of the
traditional 30 experiments being required to get a full set for an S-N curve and therefore
provides enormous scientific and economic advantages. The three tests required include a
load increase test (LIT), where an un-notched specimen is loaded at a block of cycles under
constant amplitude loading starting from a stress level around the endurance limit and being
gradually increased block wise until it fractures. The relationship between stress applied and
a damaging parameter considered (i.e. plastic strain, temperature, electrical resistance) is
determined, which generally turns out to be linear. This is followed by two traditional constant
amplitude tests (CAT) where the one is performed around the endurance limited and the
stress level the LIT started and the other at the stress level where the LIT fractured. Again
the damaging parameter considered is recorded at around half fatigue life and plotted into
the applied stress versus damaging parameter diagramme, where the LIT results have
already been plotted. Since a linear relationship can also be assumed between the two
results obtained from the CATs, the LIT results can be proportionally converted leading to
additional fatigue life data, which allow with the help of some conversion algorithms the
complete S-N-curve to be determined. Further details regarding the approach can be found
in [1].
During the last years, this approach has been further modified at the Chair for NonDestructive Testing at Saarland University [2,3], which includes new measurement methods
to be included such as temperature by means of infrared cameras allowing different fatigue
mechanisms to be differentiated as well as new strategies for data analysis and advanced
interpolation algorithms in the calculation procedure to be developed. Figure 1 provides an
example where an infrared camera technique has been applied in a LIT on a normalized
SAE 1045. The stress amplitude has been kept constant for a block of 9000 cycles each and
has been gradually increased stepwise by 25 MPa until fracture. The red course shows the
change in temperature, which is the material response due to the applied load in that, LIT.
Therefore five measuring fields (1×1 pixel) were defined along the specimen, one along the
gauge length and two at each shaft. The difference from ambient temperature is determined
in a way that the temperature recorded on a short black painted specimen of the same
material placed next to the specimen being loaded is taken as a reference. The change in
temperature was calculated in accordance to equation (1) below:
!
ΔΘ = ΔΘ! − ∙ ΔΘ! + ΔΘ! ……………………………..(1)
!
After a first linear decrease of the temperature values of ΔΘeff at 7.2·104 cycles caused by
thermo-elastic effects a change in the slope of the temperature related data can be observed
at σa = 300 MPa, which can be related to the first plastic deformations in the specimen’s
gauge length
Figure 1: Load increase test with the different levels of stress amplitude versus change in
temperature and related thermographs for normalized SAE 1045
Furthermore the approach has been extended in a way that a new stepped bar principally
‘un-notched’ specimen has been proposed which represents different cross-sections (Figure
2a), where each cross-section again represents a material volume being stressed under a
specific stress level.
20
12,0
11,1
9,1
7,9
7,1
R4
R4
R4
R4
R4
R4
43
R1
10
47,43
73,33
64,72
57,73
51,73
156,65
6,5
6,0
M 1 : 2
Figure 2: (a) Specimen geometry in accordance to the SteBLife approach and (b)
experimental lifetimes (CATexp.) and SteBLife curve calculated on the basis of the change in
temperature for normalized SAE 1045
Hence different sections of the specimen are loaded at different stress levels and therefore
different fatigue tests are represented on a single specimen and test. Although the fatigue life
can only be determined in the section with the highest applied stress and hence smallest
cross-section the remaining sections still deliver valuable data in a way that values of the
non-controlled damaging parameter can be recorded that will allow the complete S-N curve
of the material tested to be determined as to the approach mentioned above. Principally this
allows the determination of a material’s S-N curve to be reduced to a single specimen only
and the approach developed has been named the stepped-bar fatigue life (SteBLife)
approach [4]. This approach takes into account that there is no linear relation between the
elastic, elastic-plastic and plastic portion of the material reaction over the fatigue life as it is
pronounced by Palmgren-Miner. When compared to the traditional way for determining S-N
data sets where 10 to 15 fatigue tests are required easily, SteBLife is a remarkable step
forward in reducing the effort for determining a material’s fatigue data by a factor of 10 or
even more. Some result of the SteBLife approach is provided as an example for the
normalized SAE 1045 steel in Figure 2 b.
This paper gives examples how quantifiable numbers determined from NDT techniques can
be combined with short-time procedures for the advanced evaluation of fatigue data in an
‘intelligent’ way, which consequently will provide the possibility for a non-destructive based
monitoring system by means of SHM platforms to be used twofold, on the one hand as input
parameters for simulation activities with regard to residual life assessment and on the other
as values for validating those simulations along an SHM system installed in a structure such
as an aircraft.
[1]
[2]
[3]
[4]
P. Starke, F. Walther, D. Eifler, Fatigue assessment and fatigue life calculation of
quenched and tempered SAE 4140 steel based on stress-strain hysteresis,
temperature and electrical resistance measurements, Fatigue & Fracture of
Engineering Materials & Structures 30 (2007) 1044-1051.
P. Starke, D. Eifler, C. Boller, Fatigue assessment of metallic materials beyond strain
measurement, Int. J. Fat. 82 (2016) 274-279.
P. Starke, H. Wu, C. Boller, Short time evaluation of metallic materials’ fatigue
potential combining destructive and non-destructive testing methods, NDT in
Aerospace 2016, NDT.net Vol. 4 (2016) 1-8.
P. Starke, H. Wu, C. Boller, Advanced evaluation of fatigue phenomena using nondestructive testing methods, Proceedings Thermec 2016 (in press).
Simulation of Ultrasonic Inspection of Defects in Thick Structural Components
Nitin BALAJEE RAVI 1, Nibir CHAKRABORTY 1, Rakesh SHIVAMURTHY 1, Ramanan
SRIDARAN VENKAT 2, Debiprosad ROY MAHAPATRA 1, Christian BOLLER 2
rakeshs@aero.iisc.ernet.in, keerthy@civil.iisc.ernet.in, nitinb@aero.iisc.ernet.in,
nibir.chakraborty@aero.iisc.ernet.in, ramanan.sridaran@uni‐saarland.de,
droymahapatra@aero.iisc.ernet.in, c.boller@mx-uni‐saarland.de
1
Department of Aerospace Engineering, Indian Institute of Science, Bengaluru 560012,
India
2
Chair of Non-Destructive Testing and Quality Assurance (LZfPQ), University of Saarland,
Campus Dudweiler, 66125 Saarbrücken, Germany
Keywords: Ultrasonic, C-Scan, Bulk Wave, Weld, Defects, Simulation
We present here a simulation scheme developed for ultrasonic inspection of bulk or thick
structural components, which includes wave field scattering from complex boundaries
and defects. In this simulation studies, we consider defects such as crack originating from
a drilled hole[rsv1][D2] and another component with weld joints having voids and cracks in
the weld region. To model different types of defects within the weld, we model scattering
sources as material inhomogeneity within the framework of ray traced finite element
simulation involving a finite element mesh and CAD geometry as background information.
The bulk wave-field and propagation is then modelled and C-Scan images from simulation
results are visualized. We then compare the signal modelled with other commercially
available modeling tools such as CIVA. This simulation will help in predicting the wave
propagation and scattering due to defects and thereby optimizing scanning scheme and
transducer parameters as a simulation driven optimization of ultrasonic NDE inspection
process.
Simulation of Guided Waves Inspection: From NDE to SHM
Pierre Calmon, Bastien Chapuis
CEA LIST, France
Contact: pierre.calmon@cea.fr
Over the years the role played by simulation in the NDE field has been continuously increasing.
Simulation is involved at the various stages of inspection from the probe design to the data
interpretation and establishment of diagnosis. One of the main reasons of this evolution is the progress
realized by the models and codes. The CEA LIST has been contributing to this evolution by
developing the CIVA software platform which proposes to the NDE practitioners in the same
framework simulation and processing tools for different techniques (UT, ECT, XT/CT, GWT). The
performance of the NDE technique is expressed by statistical quantities such as the probability of
detection (POD) which requires a large number of sample and experiments. One of the important
applications of simulation is to contribute to the performance demonstration of inspection techniques
in complement to validation experiments and consequently to reduce costs and to improve reliability
of studies. The numerical performances of current modelling tools make it now possible to perform
intensive computational campaigns.
Guided Wave (GWT) is a potentially attractive solution in NDE and SHM for the inspection of large
structures such as tank, plates or pipes. Indeed, only a limited number of transducers is necessary
thanks to the capability of these waves to propagate over large distances. GWT can be used actively
(i.e. transducers are used to emit and to detect waves) or in passive mode, called “acoustic emission”
(i.e. sensors are used to detect the waves emitted by the propagation of a defect (e.g. crack or
corrosion).
We will present here the capabilities of the simulation tools for guided wave technique. The current
developments that are in progress at CEA LIST to deal with the specificities of SHM will be
described with latest results from the project SARISTU (European project) on the determination of
POD curves of composite plates monitoring systems. Recent developments of imaging algorithms
dedicated to corrosion monitoring system on metallic plates will also be presented.
Structural Health Monitoring of Repaired Metallic Aircraft Panel-Modeling
Based Approach
Ramanan Sridaran Venkat1, Adrià Taltavull1, Christian Boller1, Christian Dürager2
1
Universität des Saarlandes, 66125 Saarbrücken, Germany
2
NDT Centre, SR-Technics, 8058 Zurich-Kloten Airport, Switzerland
Contact: Ramanan.sridaran@uni-saarland.de, c.boller@mx.uni-saarland.de,
christian.duerager@srtechnics.com
Aircraft structures are subjected to damage, which as a consequence will have to be repaired. Some of
the repairs can be made in accordance to a standard procedure while others have to be designed
according to best engineering knowledge and practise while some uncertainty may remain with
respect to the repaired structure’s safety that may only be compensated through a high amount of
inspection. This can easily result in the structure not being eco-nomically interesting anymore and
subject to scrapping except the inspection can be auto-mated through a structural health monitoring
(SHM) system. The main objective of any air-craft repairs is to retrieve the damaged parts to their
service condition. SHM using guided waves has been considered as an option for monitoring repairs
in both metallic and non-metallic aircraft structures. Usually, the repair in the non-pressurized area of
a metallic skin is to remove the damaged area and to replace it with a doubler plate riveted to the
structure. Guided waves using piezoelectric actuators/sensors can therefore be considered as a means
suitable to be used as a method for monitoring large structures with respect to their integrity.
However, it requires optimum sensor locations and appropriate inspection parameters to monitor the
repair patches with regard to the patches’ integrity such that a high probability of detection (POD) is
obtained. Determination of the optimum SHM system as well as the optimum POD could possibly be
achieved by means of numerical simulations. In this paper, a standard repair on a skin of a large
fuselage structure has been taken to demonstrate SHM of the repairs using guided waves. Two and
three-dimensional guided wave FEM models using COMSOL-Multiphysics have been developed to
study the wave behaviour with and without repairs. Various damage detection schemes using
frequency based methods have been described in detail. Initially, stress/strain simulations have been
shown to locate the high stress zones on the repair models. By taking advantage of the wave patterns
from the guided wave FEM simulation, a strategy has been proposed to identify the optimum sensor
locations using differential imaging. The simulation approach shown in this paper is considered to
design a reliable guided wave SHM system for repaired patches. The results from the simulations are
validated with the experiment.
How to Make NDE of Composites More Deterministic
Ashvin Hadnoor, G Satyanarayana, Siddiqui A O, Sanjith G Zacharia, and Murthy B V S R
ASL, DRDO, Kanchanbagh PO,
Hyderabad-500058 India
Contact: bvsrmurthy@asl.drdo.in
It is well known that composites are heterogeneous, multi-layered, anisotropic and highly attenuative
materials. Composites are highly process and material dependent in terms of the resultant properties
and are sensitive to the variations in the process adopted. Conventional NDE techniques like
ultrasonic, radiography, thermography and other techniques are employed in identifying defects such
as voids, porosity, delaminations, inclusions and so on. There are a variety of processes which include
filament winding, tape wrapping, compression moulding, matched-die moulding, vacuum bag
moulding and so on. Each process has its benefits but each of them has a different signature in terms
of the incidence of defects that are characteristic to the process. Some of the composites structures in
aerospace are so large that they are not readily amenable for corroborative study because of
limitations facilities which calls for different strategies.
Quite often the investigation teams are confronted with difficulty in accurate interpretation of the
defect identified in composites merely by observation. In line with current practices it is natural to
resort to corroboration using different techniques with some definite success. Still there will be issues
in accuracy of interpretation because understanding of NDE in the ever evolving field of composites
is a continuous process. There are some specific cases where a certain feature in the morphology of
the composite is essential for desired performance which needs to be determined non-destructively.
All these efforts need representative samples which could be prepared as closely as possible to the
actual hardware and conduct experiments that may be destructive or non-destructive in order to
improve the understanding which improves accuracy of interpretation in the long-term.
The purpose of this presentation is to bring out the approaches taken in improving accuracy of
interpretation through corroborative NDE and by studying representative coupons specifically
prepared to evaluate the NDE which makes the interpretation more deterministic.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
A novel nonlinear acoustics technique for the detection of defects in
composite laminates
Ashish Kumar SINGH 1, Vincent B.C. TAN 1, Tong-Earn TAY 1, Heow-Pueh LEE1
1
Department of Mechanical Engineering, 9 Engineering Drive 1, National University of Singapore, 117575
E-mail: ashish@u.nus.edu
Abstract
Nonlinear acoustic methods have shown potential to identify damages in structures which are difficult to detect
using the conventional ultrasonic techniques. Most of these methods typically involve exciting the damaged
structure with a particular frequency (or multiple frequencies) and looking out for the nonlinear behavior near the
defect region. However, before the nonlinear acoustic methods can be put to use as an effective NDT technique, a
number of issues need to be resolved, one of which is the effect of excitation frequency on the success of the
method. It is known that while the tests seem to be successful at one frequency they may yet fail at some other
frequency leading to a NDT method which is highly dependent on the excitation frequency. This paper proposes
a new excitation method to overcome this problem using sweep or chirp signals. To illustrate the technique,
experiments were conducted on two plate specimens; one intact and one damaged with both made of carbon fiber
reinforced composite material. It was shown that the method was able to distinguish between the pristine and
damaged samples using the proposed technique.
Keywords: Nonlinear acoustics, composites, damage detection, NDT
1. Introduction
Composite materials are making more headways into the modern aerospace structures due to
their high stiffness to weight ratio leading to more efficient structures. However, composite
materials are much more prone to internal defects such as delaminations, which can severely
degrade the strength of the structure. Traditionally, linear acoustic methods have been used
effectively to identify the delaminations in the structure. Linear acoustic methods involve
sending sound wave signals into the structure and based on the reflection, dissipation, and
transmission of the sound wave from the defect, the damage can be predicted fairly accurately.
However, in recent times there has been an interest in nonlinear acoustic methods which are
more sensitive to damage than the linear methods and have shown potential to identify early
stage defects. An introduction to nonlinear acoustic methods as a damage detection technique
can be found in [1,2]. Non-linear acoustic experiments are highly dependent on the frequency
of excitation and further research efforts are required to overcome this dependency. Most of the
experimental works in nonlinear acoustics choose a suitable frequency at which there is
maximum nonlinear behaviour observed and the technique is then employed at that particular
frequency. This study aims to assess the effect of the excitation frequencies on the applicability
of the method and also proposes a new method of excitation which partially reduces the
dependence of nonlinear acoustics methods on the excitation frequency.
2. The experimental method
2.1 Preparation of samples
Two square plate specimens with dimensions 10 cm x 10 cm x 2 mm made up of carbonfiber/epoxy-matrix were fabricated by hand laying of prepregs and curing in the autoclave. In
one of these specimen a delamination was introduced by inserting a Teflon sheet in the middle
at a depth of 0.5 mm from the surface while doing the prepreg layup. Two piezoelectric
transducers (PI Ceramic PIC255) were then permanently bonded on the plates, one for actuation
and another one for sensing. The actuating transducer (2 cm diameter, 2 mm thick) was bonded
at one of the corners and the sensing transducer (1 cm diameter, 2 mm thick) was bonded at the
top of the delamination defect. A schematic of the specimen and the transducers is shown figure
below:
Figure 1: Plate specimen with transducers
2.2 The experimental setup
The experimental setup comprises of a function generator, an amplifier, an oscilloscope, and a
computer for processing of the results. The input excitation signal is generated from the function
generator and is being amplified by a power amplifier before being sent to the actuating
transducer. The signal from the output transducer is displayed on the oscilloscope and the output
signal data is then acquired from the oscilloscope. The output data in time domain is then
processed on a computer to convert it into frequency domain using the Fast Fourier Transform
(FFT) algorithm. The schematic of the experiment is shown in figure below:
Input signal
Power amplifier
Function generator
Input transducer
Output
signal
Oscilloscope
Amplified
Input signal
Output transducer
Data acquisition
PC
Figure 2: Experimental setup
3. Single frequency excitation
The plate specimens were excited by single frequency sinusoidal signals with peak-to-peak
voltage amplitude of 240 V. The output signals were extracted from the sensing transducer at a
sampling frequency of 5 MHz for a time period of 0.05 s. A typical response from a damaged
and intact specimen is shown in figure below for an excitation frequency of 48 kHz.
Figure 3: Time response of intact (left) and damaged (right) specimen for 5 cycles of 48 kHz signal
Higher harmonics
Figure 4:Frequency response of intact (left) and damaged (right) specimen for 48 kHz signal
It can be noticed that for intact specimen the output signal is mostly sinusoidal with almost no
(or very little) disturbances in the signal. For the damaged specimen the output signal is a bit
distorted and the amplitude in positive and negative parts of the sinusoid is also different. These
variations are due to the nonlinear behaviour of the damaged region. However, this is not true
at all the frequencies and many frequencies show almost linear behaviour for damaged
specimen also. The intact specimen is also not perfectly intact and higher harmonics are present
in its frequency spectrum, however their amplitudes were significantly smaller compared to the
higher harmonics amplitude of the damaged specimen. Due to this, there arises a need for
quantification of the nonlinearity associated with the appearance of higher harmonics. For this
purpose, a higher harmonics index (HH index) was created using the amplitudes of the
nonlinear frequencies (higher harmonics) and the amplitude of the linear frequency i.e. the
excitation frequency. HH index is defined by the below formula [5]:
𝐻𝐻 𝑖𝑛𝑑𝑒𝑥 =
𝑆𝑢𝑚 𝑜𝑓 𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒𝑠 𝑜𝑓 ℎ𝑖𝑔ℎ𝑒𝑟 ℎ𝑎𝑟𝑚𝑜𝑛𝑖𝑐𝑠
𝐴𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑥𝑐𝑖𝑡𝑎𝑡𝑖𝑜𝑛 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦
It is expected that the HH index should be minimal for the intact specimen while providing
significant values for the damaged specimen. The test was performed for frequencies between
20 kHz to 80 kHz at an interval of 1 kHz and the HH index was calculated for each of the test.
The results are shown in figure below:
1
Damaged
Intact
HH index
0.8
0.6
0.4
0.2
0
20
24
28
32
36
40
44
48
52
56
60
64
68
72
76
80
Frequency (kHz)
Figure 5: HH indices for various frequencies
It can be observed that the HH index is generally higher for damaged specimen for most of the
frequencies as was expected. However, for a few frequencies (24, 52, 58, 73, 75) the HH index
for the damaged specimen is lower or almost equal to HH index for intact specimen. This
phenomenon can be explained if we have look at the amplitude of the vibrations at these
frequencies in the figure below.
Amplitude (V)
5.0
Damaged
Intact
4.0
3.0
2.0
1.0
0.0
20
24
28
32
36
40
44
48
52
56
60
64
68
72
76
80
Frequency (kHz)
Figure 6: Excitation frequency amplitude for various frequencies
The frequencies at which intact specimen have higher HH index than that of the damaged
specimen have low vibration amplitudes meaning that the output signal is extracted at a nodal
point. And since the linear frequency amplitude term is in the denominator of the HH index
expression, the magnitude of the HH index bumps up at these frequencies. This is one of the
issue with nonlinear acoustic method based on single frequency excitation. Another issue is that
at a particular frequency the difference between the damage indices of the intact specimen is
not so clear. At some frequencies the difference is large, and some other frequencies the
difference is smaller. This is not an ideal situation and we would like a method in which the
difference between the intact and damaged specimen is more clearly marked.
4. Sweeping harmonics method
To remove the frequency effects a nonlinear acoustic method is proposed in which in place of
exciting the structure with a single frequency a sweep signal will be used instead. A typical
expression for a linear sweep signal can be represented by the below equation:
𝑥 = 𝐴 sin [2𝜋 (𝑓0 𝑡 +
𝑓1 − 𝑓0 2
𝑡 )]
2𝑇
Where f0 is the start frequency f1 is the end frequency and T is the time it takes to sweep from
frequency f0 to frequency f1. Sweep signals described above were now used for the input
excitation. A peak-to-peak amplitude of 240 V was used for the sweep signals and the time for
sweep was chosen as 0.05 s. A typical output response in frequency spectrum for the intact and
damaged specimen is shown in figure below.
Higher harmonic observed
in this range
Figure 7: Frequency response for a typical frequency sweep
Sweeping harmonics index
It can be observed in the figure above that the damaged specimen has the presence of higher
harmonics in the frequency range which is double of the range of the excitation frequency. In
above figure the range of excitation frequency was 60-70 kHz and the first higher harmonics
are observed in 2 x (60-70 kHz) = 120-140 kHz range. The second higher harmonics were also
observed in 3 x (60-70 kHz) = 180-210 kHz range (not shown in figure). Similar to the single
frequency excitation tests, the appearance of higher harmonics in corresponding ranges of the
excitation frequency provides a clear distinction between the intact and damaged specimen. A
sweeping harmonics index (SH index) similar to the HH index of the single frequency excitation
tests was introduced by measuring the average values of higher harmonics in respective ranges
(non-linear component) and the average value of excitation amplitude (linear response). The
SH index consists of the ratio of the nonlinear component to the linear component. A
comparison of the SH indices for intact and damaged specimens for various frequency ranges
has been shown in figure below.
0.12
Damaged
Intact
0.1
0.08
0.06
0.04
0.02
0
20-30
30-40
40-50
50-60
60-70
70-80
Frequency range (kHz)
Figure 8: Damage indices obtained from sweeping harmonics method for various frequency ranges
It can be seen that using the sweep signals as excitation source there is a clear distinction
between the damaged and intact specimen for all the frequency ranges. Additionally, the
distinction becomes even more clearer in higher frequency ranges.
5. Effect of excitation amplitude
It is known through various experiments that the nonlinear acoustic experiments are highly
dependent on the amplitude of the excitation signal as well. In this section a comparison has
been made between the effect of excitation amplitude on the single frequency excitation tests
and the sweeping excitation tests. The HH index for a 65 kHz signal and the SH index for 6070 kHz frequency range, for various values of excitation voltage have been shown in figure
below.
Damaged
0.12
Intact
0.75
SH index
HH index
1
0.5
0.25
0
Damaged
Intact
0.09
0.06
0.03
0
7.5
30
120
Excitation voltage (V)
7.5
30
120
Excitation amplitude (V)
It can be seen that for both the methods (single frequency and sweep) the intact and damaged
specimen could not be distinguished at lower excitation amplitudes. However, for higher values
of excitation voltage the distinction was much clearer for both the methods.
6. Conclusion
It was shown that nonlinear acoustic tests based on single frequency excitation were dependent
on the excitation frequency and the damaged and pristine specimens could not be distinguished
at several values of the excitation frequencies. Also the difference between the damage index
for the intact and damaged specimen was not clearly marked. A novel excitation method which
utilizes sweep signals for excitation was introduced and it was shown that the method was more
efficient for whole range of excitation frequencies. Additionally, the distinction was more
clearly marked due to large differences between the damage indices for the intact and damaged
specimens, especially at higher frequency ranges. The new sweep excitation method was
checked for the effect of excitation amplitude and it was observed that no significant
improvement could be provided by the sweep method in comparison to single frequency
excitation.
Acknowledgement
The support in the form of research scholarship for the first author and the resources from
strategic grant numbers R-265-000-463-112 and R-265-000-523-646 of NUS are gratefully
acknowledged.
References
1. I Y Solodov, N Krohn, and G Busse, ‘CAN: an example of nonclassical acoustic nonlinearity
in solids’, Ultrasonics, Vol 40, No 1, pp 621-625, May 2002.
2. A Klepka, T Stepinski, T Uhl and W Staszewski, ‘Nonlinear acoustics’, Advanced structural
damage detection: From theory to Engineering applications, John Wiley & Sons, 2013.
3. N Krohn, R Stoessel and G Busse, ‘Acoustic non-linearity for defect selective imaging’,
Ultrasonics, Vol 40, No 1, pp 633-637, May, 2002.
4. M Meo and G Zumpano, ‘Nonlinear elastic wave spectroscopy identification of impact
damage on a sandwich plate’, Composite structures, Vol 71, No 3, pp 469-474, Dec 2005.
5. B Y Chen, S K Soh, H P Lee, T E Tay, V B C Tan, ‘A vibro-acoustic modulation method
for the detection of delamination and kissing bond in composites’, Journal of Composite
Materials. Nov 2015.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Comparison of fibre angles between hand draped carbon fibres and
draping simulation
Christoph FROMMEL, Marian KÖRBER
German Aerospace Center, Institute of Structures and Design, Center for Lightweight-Production-Technology
(ZLP), Am Technologiezentrum 4, 86159 Augsburg, Germany
Phone: +49-821-319874-1074, Fax: +49-821-319874-1028; e-mail: christoph.frommel@dlr.de
Abstract
The mechanical properties of carbon fibre reinforced plastics, short CFRP, are highly sensitive to its respective
fibre angles. Small deviations result in a high decrease of stiffness in the manufactured parts. In the first steps of
industrial engineering for CFRP parts, draping simulations are used to give an approximation of fibre angles that
the real draped fabrics adjust to when draped in its three dimensional state. Since the draping simulations are
based on mathematical models they do not match the draping results of the manual performed production
processes. In order to compensate these manufacture errors, tolerance ranges for fibre angle deviations are set
very high. One way to control the fibre angle after draping is to use optical camera systems for quality
assurance.
The approach at DLR Augsburg to measure fibre angles is realised with a CCD camera, which is mounted on an
industrial robot. The measurement system is highly sensitive regarding the reference orientation. Every
deviation from the normal axis results in deviation of the measured angle. The setup with the robot guaranties
high reproducibility in measurements and a high accuracy to the reference.
In this paper a specified cut piece was used to perform draping simulations with different mathematical models.
The cut piece was manually draped onto a tooling. The resulting fibre angles were measured with the robot
controlled measurement system. In the end a comparison between the simulated and draped fibre angles was
done. The result provides hints which solver delivers the lowest errors for the draped cut piece. With this
information the tolerance ranges could be reduced which lead to fewer plies of carbon fibres. Less carbon fibres
lead to less material cost, weight and manufacturing time.
Keywords: Draping simulation, fibre angle measurement, robotic controlled measurement
1. Introduction
Carbon fibre reinforced plastics nowadays are an established material for high tenacity
manufacturing parts. Carbon fibres have extremely good mechanical properties along their
fibre orientation. Unfortunately they decrease rapidly when force is applied out of the fibre
longitudinal axis. The advantages from carbon fibres result in very complex manufacturing
and design processes. In a first step the designer needs to know the load cases of the
manufacturing part. The engineer then has to create a stack up of carbon fibre cut pieces that
can handle the forces and no boundary conditions are harmed. Boundary conditions for
example are the maximum deviation in fibre angle between each ply, symmetrical stack ups,
maximum width of fibre material, force inducing areas and many more [1]. A crucial
influence on the stiffness lies in the tolerance ranges that are applied on each ply regarding the
fibre orientation.
The fibre materials used in this case are carbon fibre woven fabrics which are advantageous
for handling and draping processes. These fabrics consist of warp and weft rovings which are
in plane state perpendicular to each other [2]. If a double curved deformation is applied to
these fibres, for example a 3D-shape of a tooling in production, they have to shear and change
their respective angle to each other. To compensate these changes in fibre orientation the
tolerance ranges are expanded. To estimate the shear in draped fabrics, draping simulations
can be done. The draping simulation estimates the fibre angles with the help of mathematical
models. Two common methods are kinematic algorithms and finite elements methods.
Kinematic algorithms are a fast way to estimate the fibre orientation but have relatively high
errors [3]. The finite element methods are more difficult to calculate and require much more
input and time to calculate, thus are more precisely.
After the engineering the manufacturing can begin. The worker has to place the fabric cut
pieces exactly where the engineer has calculated it. Therefore a laser projector generates the
outlines of the cut pieces on the tooling for orientation. To have a defined start mostly a line
on the side of the cut pieces are placed to its projected line and gets fixated. From this line the
worker drapes the cut piece till it fits the complete projected outline. The start line and the
outline also are boundary conditions and results of the draping simulation.
In this paper a double curved generic tooling was used to drape a generic shaped cut piece
manually on it. The tooling and cut piece were also used for draping simulation. The draping
simulation was performed with several kinematic solvers. At the German Aerospace Center in
Augsburg an optical fibre angle measurement system is used to measure fibre angles on multi
curved surfaces. With the help of this system we want to compare the fibre angles from
simulation and real draped fibres at defined points on the cut piece to get better knowledge of
the relation between draping simulation and manually draped carbon fibres.
1.1 Main issues
The question that will be answered in this paper is, to what extent the fibre angle results of the
manual process can be compared with the simulation results. First, the robot system needs to
be calibrated to ensure the measurement is conducted at the correct positions. When the
connection between reality and simulation was established the measured values had to be
compared between the simulated ones. Since the material was a fabric, the fibre angles for
warp and weft and the resulting shear between them had to be compared and evaluated. The
last step is to check which solver results in the lowest deviations from the measured ones for
this set up.
2. Experiments
The experiments can be divided into four parts. At first an experimental set-up had to be
established. To get the knowledge of how the fibres propagate on the surface the draping
simulation had to be done. After finishing the set-up and the simulation the experiment can be
executed with draping the cut piece and then measuring the fibre angles with the fibre angle
measurement system. In our work a total of six identical cut pieces were used.
2.1 Experimental set-up
The tooling that was used in this research has a double curved concave shape (Figure 1).
Figure 1 Double curved Tooling
Since the tooling is not symmetrical and not circular, the radius is different for each curvature.
The depth is approximately 1000mm. The red outlined contour was used for the cut piece.
The boundary conditions to this contour were that the piece must be as large as possible.
There needs to be a short distance (white line) from the starting edge (yellow line) to the other
side. It also inherits a long distance (purple line) from the starting edge. Also a curved edge
(green line) and a corner (blue line) should be included. The flat cut piece was approximately
2000mm long and 1200mm width.
The robot is a standard KUKA KR 120 R2700 extra HA. This robot has a very high
repeatability accuracy which ensures that it always measures at the same points for every
draped cut piece.
The measurement system is an optical system from the company PROFACTOR. The system
is based on the reflective behaviour of carbon fibres. The system takes pictures of the fibres
with different directions of illumination. The software combines the pictures and calculates
the fibre angles out of their respective reflections (Figure 2) [4].
Figure 2 Light reflection of carbon fibres
(Source: Profactor CF-Sensoren November 2010)
The movement strategy for the robot and the measurement plan were generated in CATIA and
the FAST Suite. With the help of the offline programming it can be guaranteed that the robot
moves parallel to the starting edge and levels our measurement system correct. The
measurement system mounted onto the robot, the tooling and the measurement plan can be
seen in Figure 3.
Figure 3 CATIA FAST Suite set up
The measurement plan covers 100% of the cut piece. The measurement system makes a
picture of 40 mm by 40 mm, therefore the measurement points have a distance of 40 mm from
each other. A total amount of 903 measurement points were scheduled.
2.2 Draping simulation
The draping simulation for this research was performed with CATIA and the respective
toolboxes. The standard toolbox for composite design inherits two solvers called Symmetric
and Minimum Distortion. With the addition of the toolbox Composites Fiber Modeler (CFM)
three additional solvers called Optimized Energy, Optimized MaxShear and FEFlatten could
be used. All of these solvers are kinematic solvers. The Symmetric solver forces the fibres to
propagate symmetrical which is for usage of symmetric toolings. In the Minimum Distortion
solver the propagation of fibres is done with the lowest deformation of the fibres itself. From
the more advanced toolbox CFM the Optimized Energy solver tries to minimize the shear
strain energy on the extending edge. For the Optimized MaxShear solver the maximum shear
gets limited. The FEFlatten solver is the most advanced one. It considers strain along the
fibres and contains finite element methods which reduce errors from the geometrical solvers.
While the lines for the other solvers leave the surface for reaching its next shear point the
FEFlatten solver always stays on the surface while propagating. Every calculation needs a
material, a rosette, a seed point or a seed line, the size of the net and a solver. For the material
the material data from our cut piece was put. The rosette defines the fibre directions, for our
material 0 degree and 90 degree. The 0 degree is perpendicular to the starting edge and the 90
degree orientation is parallel to the starting edge. The seed point/line is the definition from
where the simulation starts to propagate. The seed point was in the middle of the starting edge
and the seed line was the starting edge. The net size defines how long a line from the
kinematic approach will be, the shorter the lines the higher the calculation time but the more
accurate the solver gets. For our calculation a net size of 4 mm was used which corresponds
with the weave of our fabric. These inputs were then used to perform a calculation for each
solver with a seed point and seed line. The calculations took about 1 minute for every solver
except the FEFlatten. It took about 20 minutes for one calculation with that solver. After the
calculation the measurement plan was loaded into the calculation model. A report function of
the toolbox generates a list of each point and its respective coordinates and fibre angle. This
list just gives you the information for the main direction. Therefore the export had to be done
a second time for the 90 degree direction and the resulting shear was then calculated
manually. The result for the Minimum Distortion (left side) and the Optimized MaxShear
(right side), both for the 90 degree orientation, can be seen in Figure 4. The green areas are
areas with fibre angle deviation below 1 degree, the yellow areas are above 1 and below 3
degrees and the red areas are above 3 degrees [5].
Figure 4 Results of draping simulations for
Minimum Distortion and Optimized MaxShear
2.3 Draping
The cut pieces were provided by an industrial cutting machine and then stored on a table.
Now the cut piece was manually laid into the tooling. At first the starting edge was levelled to
the drawn contour from the tooling. After the cut piece starting edge matched with the shape
on the tooling it was fixated. Like the behaviour of the kinematic algorithms the draping was
started perpendicular to the starting edge at the height of the rosette. You could see that the
cut pieces were very accurate in reaching the other side of the tooling shape since the 0 degree
line suffers almost no angle deviation. After that the further draping was into the corners of
the cut piece starting from the rosette. After reaching the tooling shape the complete cut piece
was fixated onto the tooling. In the lower part were the distance to the rosette is short the cut
piece matched fairly well. In the upper part with the high distances a lot of draping had to be
done to get the cut piece into the shape of the tooling shape. When considering the time it
needed to get the cut piece into the shape we assumed that in the lower side would be less
change in fibre directions and a lot in the upper part. We also assumed that the resulting shear
angle in the lower part raised and in the upper part decreased because we had to sweep over
them in the lower side and on the upper part had to push them into the tooling shape. An
example of a draped cut piece can be seen in Figure 5.
Figure 5 Draped cut piece
2.4 Fibre angle measurement
After each draped cut piece a fibre angle measurement was made. Before the measurement
started a reference point on the tooling was measured if the robot and the measurement system
are still calibrated. The robot moved to the tooling and the respective measurement point and
then remained still for 2 seconds. This ensured that the robot is not vibrating and the camera
has enough time to refocus and gather enough pictures. The measurement data is saved and
that the robot moved to the next point. After finishing the measurement a list of data was
saved. The list contained the number of measurement point, the robot positions to reach the
respective point and its fibre angle. The measuring system can be seen in Figure 6.
Figure 6 Measuring system while measuring
3. Discussion
In the first step the repeatability of the manual draping results controlled. Therefore the
standard deviation for each point over the 6 cut pieces was calculated. This was done for the 0
degree, 90 degree and resulting shear angle. The results showed that the standard deviations
were relatively high but this was expected for only 6 samples. The standard deviation was
2.17°, 1.53° and 2.06° for 0°, 90° and resulting shear respectively. With these high standard
deviations there was no use to check every cut piece one by one. Therefore the mean fibre
angle for each point over the 6 cut pieces was used for further calculations.
After the repeatability was checked the assumptions made while draping had to be controlled.
Therefore plots of the fibre angle from the measurements over their coordinates were made.
These plots can be seen in Figure 7 were the left plot is the 0 degree orientation, the middle is
for 90 degree and the right for resulting shear.
Figure 7 Measured fibre angles
Here it could be seen that our assumptions were right. The lower part had less change in fibre
angle than the upper part. Also the resulting shear looked as expected. The right side of the
lower part suffered an increase in the resulting shear angle and the upper part a decrease. One
very important aspect is that the fibre angle of the 90 degree fibres changed much more than
the 0 degree orientation. Our explanation for this behaviour is the weave of our fabric and the
roughness of the tooling. The fabric has an atlas weave, which means that the warp fibre lies
under a minimum of two weft fibres before crossing one weft fibre on top, and therefore more
fibres of one direction on each side than the other. In this case the 0 degree fibres were lying
on the bottom and therefor in touch with the tooling surface. Since the tooling is made out of
Ureol it has a relatively high roughness. Therefore we assume that the 0 degree fibres were
hold in place due to friction between fabric and tooling and caused the 90 degree fibres to
deflect more to reach the necessary shear angle.
The next step was to compare the measured angles with the simulated ones. To get values for
comparison, the difference in fibre angle between reality and simulation was calculated for
each point. This was done for the 0 degree and 90 degree orientation and the resulting shear.
These differences were used to calculate the highest error in negative and positive direction of
fibre angle deviation. To get a better overview about the distribution over the whole cut piece
the absolute values from these deltas were added up to an error sum. The differences for
positive and negative deviation and the error summation were compared between seed point
and seed line. For the positive deviation the seed point had the lower errors. The best result
for negative deviation was achieved with a seed line. Compared with the error sum the seed
point was better. Over all the seed point had the lower errors. To evaluate which solver has
the lowest difference to the reality the lowest delta for positive deviation, the lowest delta for
negative deviation and the lowest error sum was searched for each fibre direction and
resulting shear. The results can be seen in Table 1.
Table 1 Results for the lowest errors
Positive deviation
Negative deviation
Error sum
0 degree direction
Optimized MaxShear
Seed Line 5,17°
Minimum Distortion
Seed Line -1,54°
FEFlatten Seed Point
1335,37°
90 degree direction
FEFlatten Seed Line
0,84°
Minimum Distortion
Seed Line -4,17°
FEFlatten Seed Point
1149,58°
Resulting shear angle
FEFlatten Seed Point
1,26°
Optimized Energy
Seed Point -9,72°
Optimized MaxShear
Seed Point 2201,40°
The result shows that no clear favourite could be generated. For every boundary condition the
solver with the lowest error changes. This shows that with the high standard deviation in hand
draping the definition of which solver would be the best is near impossible. What gives a
better view about the outcome of the comparison can be seen in Figure 8, were the difference
between solver and reality is plotted over their coordinates for each point.
Figure 8 Delta plots for the solvers with the lowest errors
With help of the colour scale the results of the solvers are more apparent. The pictures are
sorted in the same way as in Table 1. If you look on the left column for the 0 degree direction
the Optimized MaxShear solver has the lowest difference in positive direction but also inherits
a high difference for negative direction. For the negative deviation the Minimum Distortion
solver has the lowest difference but has a high positive deviation. Looking now on the error
sum the FEFlatten solver has the lowest summation of differences and looks much smoother
without that high peaks in it. The exact same behaviour can be seen for the 90 degree
direction and the resulting shear. This would lead to the conclusion that the solvers with the
lowest error sum tend to have the same trend as the reality but lack in accuracy and would
then be more suitable. With this comparison the FEFlatten solver with a seed point could be
declared as the best fitting solver to our draped cut piece. It has the lowest error sums for
positive and negative direction and also a low error sum for shear. Only the Optimized
MaxShear solver was better for the resulting shear but had much higher error sums for 0 and
90 degree orientation.
At last a check if the solver would always have higher fibre angle deviations than the reality
was made. This would ensure that the engineering can use the simulation for calculations
since the error would be higher than in reality. The comparison can be seen in Table 2.
Table 2 Comparison of fibre angle ranges
0° Max
0° Min
90° Max
90° Min
Shear angle Max
Shear angle Min
Reality
FEFlatten
Seed Point
3,46°
-3,31°
91,73°
82,59°
90,21°
83,11°
1,27°
-5,96°
91,16°
85,37°
94,42°
88,33°
Minimum
Distortion
Seed Line
1,54°
-8,56°
90,74°
85,72°
95,15°
85,19°
Optimized
Energy Seed
Point
2,62°
-5,75°
91,21°
87,02°
93,64°
86,40°
Optimized
MaxShear
Seed Point
3,59°
-5,50°
91,04°
86,67°
93,49°
86,45°
Here it can be seen that over all the solvers do have wider ranges and bigger errors. Due to the
fact that in reality the 90 degree fibers suffered more angle deviation than the 0 degree
direction the reality can have bigger deviations then the simulation. This would lead to weaker
mechanical properties at some areas of the part than expected from the simulation.
4. Conclusion
Within this research an experimental set up was created that can measure fibre angles in a
way that they can be compared to draping simulations. A process to evaluate the data of the
experimental set up to find the best fitting solver was established. The boundary conditions
for the draping simulations could be understood much better. It was possible to declare a best
fitting solver for our experimental set up. The best fitting solver can change for different
materials and toolings. It could be shown that the kinematic solvers used for draping
simulation neglect crucial parts of real draping like roughness of the tooling which lead to
high errors. The conditions to realize manual draping in production have huge disadvantages
for working with fibre reinforced plastics and thus generate higher fibre angle deviations.
Knowing these circumstances the next approach will be to use a kinematic end effector for
draping. This means the end effector can pick up a cut piece in its flat position drape it by
changing its own shape while holding the cut piece. The end effector also can place the cut
piece into the tooling. This would ensure that each cut piece would be draped the same and
has no random influences like draping manually. Since this would eliminate boundary
conditions made for manual draping like the starting edge, the other boundary conditions
could be changed to suit the draping process better. The rosette for example could be placed
in the middle of each cut piece which would reduce errors drastically due to the shorter
distances to the edges. The biggest improvement would be the repeatability of the robotic
system which would lower the standard deviation and would deliver a much more accurate
declaration of which solver generates the lowest errors.
Acknowledgements
This work had been founded by fundamental funding of DLR. Special thanks go to our
colleagues of ZLP Augsburg that helped on this research and made things happening.
References
1.
Helmut Schürmann, 'Konstruieren mit Faser-Kunststoff-Verbunden', Springer-Verlag
Berlin Heidelberg 2005.
2.
DIN 60000: Textilien. Grundbegriffe. January 1969
3.
S.G. Hancock and K.D. Potter, 'The use of kinematic drape modelling to inform the
hand lay-up of complex composite components using woven reinforcements',
Composites: Part A 37 (2006) 413–422, May 2005
4.
Thomas Schmidt, 'Non-destructive testing for composites – part 1', Script of lecture
Fatigue at the University of Applied Science Augsburg, October 2013.
5.
Dassault Systèmes, 'CATIA Documentation', [Online]. Available:
http://docs.css.herts.ac.uk/CATIA%20documentation/. [Accessed 10 2016]
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Acoustic Emission testing to evaluate a new parameter in composite
materials: Delay cracking Load
Giuseppe NARDONI1, Nasim FALLAHI1, Pietro NARDONI1, Mario TURCONI2, Franco
MONTI2
1
2
I&T NARDONI INSTITUTE, Brescia, 25124 , Italy
Phone: +39030 266582, E-mail: Nardoni.campus@gmail.com
Center of integrated training systems, LEONARDO Aircraft, Vengono Superiore (VA), 21040, Italy
E-mail: mario.turconi@leonardocompany.com
Abstract
The aim of this paper is to investigate a new parameter in the behavior of carbon-epoxy composite materials: “Delay
Cracking Load (DCL)”. The delay cracking load is a constant load that applied at a three point flexural bending test
give up to a delay cracking process up to the complete fracture in composite specimens.
Only Acoustic Emission technique is able to predict the delay cracking in constant load. DCL is an important
parameter in the characterization of the composite materials. The interval between the load steps based on this
research shall not be less than 2-5 days.
Keywords: delay crack load, Acoustic emissions, Carbon-epoxy composite, fracture mechanisms.
1. Introduction
Carbon fiber composites [1, 2] in the early 60s were used because of high strength and stiffness
coupled with low weight and high resistance to corrosion and fatigue compared with traditional
materials such as metals (see figure 1). Carbon fibers are a new generation of high strength fibers
that used in composites with light weight matrix such resins. For this reason Fiber composites are
suitable to make aircraft parts (Fig 2).
Till today, knowledge about plastic materials failure limit is more complex than for linear
elastic materials like metals. Plastics can be dimensioned by means of critical strain as failure
limit. So, acoustic emission method is a valuable analysis in a tensile test, to get quick value of
critical strain of a new material [3].
Interface fracture is one of the critical failure modes for composite structures, which may lead
to the separation of plies and eventually to the fault of the component: it is therefore necessary to
strengthen the interlaminar fracture toughness for highly reliable composite materials and
structures. Knowledge about the fracture mechanisms is vitally important in different kinds of
loading test. As some of the authors have already demonstrated, interleaving small diameter
fibers between one or more interfaces, improves strength, toughness and delamination resistance
of composites without reducing in the in-plane properties or adding weight [4-9].
For composite materials design, knowledge of fracture mechanisms is vitally important, so AE is
one of the suitable techniques to investigate the online fracture mechanisms [5, 10-14]. The most
important advantages of AE is on-line monitoring for fracture or friction between the layers or
fibers and the capability to detection of the damage mechanisms [15]. Most common analyzing
method for AE, uses certain features extracted from signals such as amplitude, energy, count,
rise time [16].
Ratio of
strength/stiffness
to weight is high
Stiff
Strong
Light weight
Corrosion
Figure 1. Benefits of using composite materials
Figure 2. Different kinds of composite materials in aerospace field
The successful development and design of continues fiber reinforced carbon epoxy composites
are dependent on a thorough understanding of basic properties such as fracture and delay failure,
slow crack growth or damage accumulation[17].
2. Materials and Methods
2.1. Composite fabrications
Panels were fabricated by hand lay-up process, after the lay-up, panels were cured by vacuum
bag in an autoclave in 185°C (365 °F) according to process specification provided by supplier.
Carbon unidirectional preprag fiber-epoxy composite (table 1) were used in constant load to
investigate the delay crack in carbon epoxy composite in sequence loading (Fig 3).
Table 1: unidirectional 0/90 preprag of carbon epoxy composite
Properties
Percentage of resin
Value
33%
Percentage of fibers
Preprage (weight per unit area)
Fiber type
67%
134 g/m2
Unidirectional carbon fibers 0°/90°
2.2. Mechanical flexural bending testing
The three point flexural bending test has appointed to perform flexural test for the
characterization of composite materials and two AE sensors have been applied on the specimens
(Fig 4).
Mechanical hydraulic test has been set up to minimize the noise level on the acoustic emission
system during the step loading on the specimens.
The flexural test has been calculated according to the ASTM D790[18]:
AE features such as count, rise time, energy and amplitude,
3𝑃𝐿
𝜎𝑓 = 2𝑏𝑑2
(1)
𝜎𝑓 : Stress in the outer fibers at midpoint, MPa (psi).
P: load at a given point on the load-deflection curve, N (Newton).
L: support spans (mm).
b: width of beam tested (mm) and,
d: depth of beam tested (mm).
(A)
(B)
Fig 4: (A) Three-Point bending flexural test by AE, (B) Three-points bending test on carbon
epoxy composite through silent hydraulic pressurized
2.3. Acoustic emission equipment
The AMSY-6 acoustic emission system1 was used for damage monitoring during monotonic
3-point bending examination of composites in mechanical hydraulic test.
The AMSY-6 was equipped with Vallen software for data acquisition and analysis. Loading
data were transferred directly to AMSY-6 system allowing for correlation with acoustic signals
obtained from examined GFRP and CFRP specimens. The AE signals were registered using two
broad-band sensors attached to specimen’s surface with rubber band and vacuum grease as a
coupling agent (Fig 4B). The Vallen AEP4 preamplifiers had the gain set to 34 𝑑𝐵. System
threshold setting varied from 34 to 46 𝑑𝐵. Sensitivity of the sensors and placement of the sensors
was calibrated by Hsu-Nielsen source preceded the actual measurements.
After the calibration step, AE signals were recorded during the mechanical testing. Signal
features such as count, rise time, energy and amplitude were used for investigating of the failure
mechanisms (Fig 5).
Figure. 5 Amplitude vs. time presentation of AE signals during the delay cracking load. The first
and second load applied before in the period of 7 days does not provoke any delay cracking.
3. Results and Discussion
In the Carbon epoxy specimens, the same procedure has been applied as for the glass-epoxy
composites. The delay cracking load has been identified at the third applied load. This load has
been kept constant for all the test duration. After 184 hours the first high amplitude clusters of
1
Manufactured by VALLEN System GmbH, 82057 Icking, Germany. http://www.vallen.de/
AE signals representing the micro cracks in fiber and matrix (see Fig 5).
From 222 hours a scattered high amplitude signals ranging from (60-95 dB) appear in the
monitor of AE. The pattern distribution of these signals was different from the glass epoxy
composites, not in periodical cluster but in the “sky stars shape” signals. At the end time (264
hours) through the thickness surface breaking crack emerged in the specimens (see figure 5). The
crack was along the axis of the high stressed area (figure 6). At the end stage SEM picture
captured from the surface of fracture (Fig 7).
Figure. 6 Main surface breaking fracture appeared after continues process of delay cracking
Figure. 7 SEM from the fracture in carbon epoxy composite: (A),(C): matrix cracking and fiber
breakage (B) fiber breakage (D) matrix cracking[19, 20]
4. Future work
The important of the phenomena of delay cracking load in the project and manufacturing
composite materials speed up a research to correlate the final ration of strength data with the
delay cracking load data.
Most probably this will have as a result more reliable composite material for the construction
which are more and more complex and extended in the application.
The recent introduction of nanofibers (Fig 8) in composite material manufacture will give
ongoing research may give a contribution improving the mechanical properties of the composite
materials [19, 20].
5. Conclusion
Delay cracking load is a new investigation in composite materials. In 3-point bending by apply
the constant load for a long time, fracture mechanisms occur. Knowledge of the critical load is an
important parameter to prevent delay crack loading. Acoustic Emission technique is a suitable
method to detect the delay cracking load in composite materials. For the better indication of
delay cracking load, attention to the amplitude of the signals, strongly provides. so the
knowledge of critical constant load could help to prevent the failure mechanisms by the time in
each composite material.
Figure. 14 (a) nanofibers of nylon 6,6 (b) fracture mechanisms in Nano nylon 6,6 in carbon
epoxy composite
Reference
[1] Carboni M, Gianneo A, Giglio M. A Lamb waves based statistical approach to structural health
monitoring of carbon fibre reinforced polymer composites. Ultrasonics. 2015;60:51-64.
[2] Carboni M, Gianneo A, Giglio M. A LOW FREQUENCY LAMB-WAVES BASED STRUCTURAL
HEALTH MONITORING OF AN AERONAUTICAL CARBON FIBRE REINFORCED POLYMER
COMPOSITE.
[3] F. Willems JB, Ch. Bonten. Detection the critical strain of fiber reinforced plastics by means of
acoustic emission analysis. 32th European Conference on Acoustic Emission Testing. Czech Czech
Society for Nondestructive Testing; 2016. p. 525-34.
[4] Bacon R. Growth, Structure, and Properties of Graphite Whiskers. Journal of Applied Physics.
1960;31:283-90.
[5] Dzenis YA, Reneker DH. Delamination resistant composites prepared by small diameter fiber
reinforcement at ply interfaces. Google Patents; 2001.
[6] Kim J-s, Reneker DH. Mechanical properties of composites using ultrafine electrospun fibers.
Polymer Composites. 1999;20:124-31.
[7] Refahi Oskouei A, Zucchelli A, Ahmadi M, Minak G. An integrated approach based on acoustic
emission and mechanical information to evaluate the delamination fracture toughness at mode I in
composite laminate. Materials & Design. 2011;32:1444-55.
[8] Arumugam V, Sajith S, Stanley AJ. Acoustic Emission Characterization of Failure Modes in GFRP
Laminates Under Mode I Delamination. Journal of Nondestructive Evaluation. 2011;30:213-9.
[9] Wevers M. Listening to the sound of materials: Acoustic emission for the analysis of material
behaviour. NDT & E International. 1997;30:99-106.
[10] Philippidis TP, Nikolaidis VN, Anastassopoulos AA. Damage characterization of carbon/carbon
laminates using neural network techniques on AE signals. NDT & E International. 1998;31:329-40.
[11] Marec A, Thomas JH, El Guerjouma R. Damage characterization of polymer-based composite
materials: Multivariable analysis and wavelet transform for clustering acoustic emission data. Mechanical
Systems and Signal Processing. 2008;22:1441-64.
[12] Bar HN, Bhat MR, Murthy CRL. Parametric Analysis of Acoustic Emission Signals for Evaluating
Damage in Composites Using a PVDF Film Sensor. Journal of Nondestructive Evaluation. 2005;24:12134.
[13] Ramesh C, Ragesh H, Arumugam V, Stanley AJ. Effect of Hydrolytic Ageing on Kevlar/Polyester
Using Acoustic Emission Monitoring. Journal of Nondestructive Evaluation. 2012;31:140-7.
[14] Yan J, Heng-hu Y, Hong Y, Feng Z, Zhen L, Ping W, et al. Nondestructive Detection of Valves
Using Acoustic Emission Technique. Advances in Materials Science and Engineering. 2015;2015:9.
[15] Inasaki I. Application of acoustic emission sensor for monitoring machining processes. Ultrasonics.
1998;36:273-81.
[16] Lu C, Ding P, Chen Z. Time-frequency Analysis of Acoustic Emission Signals Generated by
Tension Damage in CFRP. Procedia Engineering. 2011;23:210-5.
[17] Knauss WG. Delayed failure — the Griffith problem for linearly viscoelastic materials. International
Journal of Fracture Mechanics. 1970;6:7-20.
[18] ASTM. Standard Test Methods for Flexural Properties of Unreinforced and Reinforced Plastics and
Electrical Insulating Materials. ASTM: astm; 2007.
[19] Palazzetti R, Zucchelli A, Gualandi C, Focarete ML, Donati L, Minak G, et al. Influence of
electrospun Nylon 6,6 nanofibrous mats on the interlaminar properties of Gr–epoxy composite laminates.
Composite Structures. 2012;94:571-9.
[20] N. Fallahi G. Nardoni, H. Heidary, R. Palazzetti, X.T. Yan, A. Zucchelli. Supervised and Nonsupervised AE Data Classification of Nanomodified CFRP During DCB Tests. FME TRANSACTIONS.
2016;44.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Guided Lamb wave based multi-level disbond detection in a honeycomb
composite sandwich structure
Shirsendu Sikdar1, Sauvik Banerjee 2
1,2
Civil Engineering Department, Indian Institute of Technology Bombay; Mumbai-400076, India
Phone: +91-22-25767343, Fax: +91-22-2576 7302; e-mail: 1shisu.iitkgp@gmail.com, 2sauvik@civil.iitb.ac.in
Abstract
The objective of this study is to detect hidden disbonds in a honeycomb composite sandwich structure (HCSS)
using ultrasonic guided Lamb waves and bonded piezoelectric wafer transducers (PWTs). To achieve this, a
combined theoretical, numerical and experimental study has been carried out to understand the guided wave
(GW) propagation characteristics in HCSS. A fast and efficient global matrix based two dimensional (2D) semianalytical model is used to study dispersion behaviors and transient response in a healthy HCSS under PWTexcitations. Finite element (FE) based numerical simulation of GW propagation in a disbonded HCSS is then
carried out in ABAQUS. Experiments are then carried out in the laboratory to validate the numerical results. A
good agreement is found among the theoretical, numerical and experimental results. A substantial amplification
in the primary anti-symmetric mode (A0) is observed due to the presence of disbond. Eventually, based on these
changes in modal amplitudes, the location and size of hidden disbond within the PWT-network is experimentally
determined using a probability based damage detection algorithm. The study is further broadened for multiple
disbond identification in HCSS, using pseudo-experimental signals.
Keywords: Honeycomb composite sandwich structure (HCSS), guided wave (GW), disbond, piezoelectric wafer
transducer (PWT), group velocity
1. Introduction
HCSS is a special kind of composite structure in which, thin fiber reinforced composite skins
are bonded to the top and bottom faces of relatively thick and substantially lightweight
aluminum honeycomb core using adhesive. These novel materials are extensively used in
aeronautic, aerospace and marine industries as a specialized lightweight construction material
[1]. The higher strength-to-weight ratio makes it suitable for construction of some of the
major structural components such as flight wings, fuselage, blades, etc. and the high energyabsorption capability makes it attractive for impact protection and mitigation related
applications. Unfortunately, due to aging, repeating loading or an intensive load, the
honeycomb core tends to induce debonding along the skin-core interface, jeopardizing the
safety and integrity of the whole structure [2]. The in-situ inspection methods are the subjects
of recent studies to overcome the limitations of the conventional, time-consuming and costly
off-line non-destructive examinations, which require the disassembly of the large
structures/sub-systems. Guided Lamb wave based inspection techniques have the potential to
accurately detect such disbonds in composite structures [3,4,5,6]. The GW mode tuning plays
a significant role in the non-destructive evaluation (NDE) and structural health monitoring
(SHM) of composite structures employing piezoelectric transducers (PZTs) [7,8]. By using
the piezoelectric actuators/sensor system, Su et al. [9,10] used the time of flight (TOF)
calculation technique to triangulate the delaminations in composite laminates. The prestackreverse migration technique was used by Wang and Yuan [11]. In this work, a linear PZT disk
network was used to image the delaminations in composite structures. The wave propagation
in honeycomb sandwich structures can be characterized as leaky GWs at sufficiently higher
frequencies, owing to the striking acoustic impedance difference between the skin and core
and the high core/skin thickness-ratios [12,13]. A significant amount of wave attenuation
occurred due to the energy dissipation mechanism of GWs into the honeycomb core. The GW
propagation through the debonded area of honeycomb sandwich structures, a substantial
increase in amplitude of the received output signal was noticed [6,14,15]. A global matrix
method based solution of wave propagation problems for a multilayered anisotropic plate due
to the time harmonic loadings was presented by Mal [16]. Wave propagation studies in
laminated composite plate using the spectral finite element method and the first order
laminated theory was presented in [17]. Dispersion characteristics of the propagating GW
modes in a multilayered laminated composite plate subjected to transient surface excitations
ware studied using a global matrix method based simplified two-dimensional (2D) semianalytical model [18]. A robust 2D semi-analytical model is established by Banerjee and Pol
[19] for rapid calculations of the elasto-dynamic field in a laterally unbounded honeycomb
composite sandwich plate, subjected to time-dependent transient surface excitation. The 2D
semi-analytical model has shown the potential to accurately represent the theoretical output
signal for the HCSS. Theoretical simulation of leaky Lamb waves in the composite skin was
shown by Hay et al. [6]. The sensitivity analysis of various Lamb wave modes for the
composite skin-Nomex core debonding was analyzed by the frequency sweeping. Qi et al.
[20] compared the ultrasonic wave transmission energy for the specimen at normal conditions
and the debonded specimen in order to identify the skin-core debonding in the honeycomb
composite by using the leaky surface wave propagation. Nevertheless, the information for the
quantitative assessment of debonding was not provided. The authors used clusters of sensor
arrangements to accurately locate the skin-core debonding, though, no information was
provided about the size of debonding. Recently, the probability analysis-based algorithms
have been proposed, in order to image the corrosion and cracks in aircraft wings and
composite laminates [21,22]. A three-dimensional numerical simulation and experimental
study have been done by Song et al. [13] for damage detection. In the work, a signal
difference coefficient (SDC) was used to represent the differential features of debonding.
Recently, Sikdar et al. [23] has shown a damage detection technic based on the modeconversion in the received signal, using the PWT actuator/sensor network for a HCSS plate,
based on the experimental results. Though, there is a deficiency of a fast, systematic and
efficient dis-bond detection technique for HCSS by using a sparse actuator-receiver network
of PWTs.
In this research work, the dispersion curve and the baseline theoretical response of the
HCSS is obtained, using a global matrix method based 2D semi-analytical model [19] to
guide the numerical and experimental results. A 3D FE method based numerical simulation in
ABAQUS is carried out, in order to simulate the disbond influences on GW propagation in the
HCSS. Experimental studies are then conducted to verify the numerical simulation and for
disbond detection. Eventually, based on changes in modal characteristics, the location and
size of unknown disbonds within the PWT array are experimentally (for single disbond) and
pseudo-experimentally (for multiple disbond) determined using an appropriate SDC
algorithm. The differential features of disbond are represented by using an appropriate SDC.
Probabilistic analysis of the differential features of the propagated GWs in the HCSS with and
without disbond for each transmitter-receiver pair is conducted to form the disbond
localization image at each frequency. By superimposing the images from each transmitterreceiver itinerary, an imaging area is reconstructed. The final image of the whole plate is
obtained by using image fusion technique, demonstrating that the proposed approach is able
to provide reliable quantitative information about the size and location of hidden disbonds in
the HCSS.
2. Experimental setup
The HCSS sample plate (600mm × 450mm × 13.5mm) used in this study is comprised of a
aluminum honeycomb core (12mm thick) embedded between two Graphite/Epoxy fiber
reinforced composite (GFRC) skins (0.74mm thin) at the top and bottom. Each GFRC skin
consists of seven composite layers, in which five unidirectional (UD) twill composite layers
are present in-between two cross-ply (CP) layers. The disbond region (30mm × 30mm)
between the skin-core interphase is generated during the manufacture of the sample HCSS
plate. The detailed elastic properties of the HCSS are given in Table 1. Where, „E’ is the
Young‟s modulus, „G’ is the shear modulus, „’ is the Poisson‟s ratio, „t‟ is the layer thickness
and „‟ is the mass density.
Table 1: Elastic properties of the HCSS
E1
(GPa)
E2
(GPa)
E3
(GPa)
G12
(GPa)
G23
(GPa)
G13
(GPa)
12
UD-lamina
CP-lamina
Soft-core
60.212
110.31
0.0804
60.212
110.31
0.0804
10.252
18.247
1.6121
18.20
42.41
0.0321
3.611
4.136
0.0964
3.611
4.136
0.0964
0.20
0.30
0.25
0.03
0.12
0.025
Adhesive
0.0486
0.0486
0.0486
0.0174
0.0174
0.0174
0.40
0.40
Material
13
23

(kg/m3)
t
(mm)
0.03
0.12
0.025
142
165
32
0.08
0.17
12
0.40
125
0.01
In order to verify the theoretical and numerical results, experiments are carried out on the
HCSS sample plate using surface-bonded PWTs (20mm × 20mm × 0.4mm). The
actuator/sensor PWTs are operated through an NI-instrument (PXI system), as clearly
described in [24].
Normalized load amplitude
1.0
0.5
0.0
-0.5
-1.0
0
5
10
15
20
25
30
35
Time (s)
(a)
(b)
(c)
Fig. 1: (a) Sample HCSS, (b) PWT arrangements on the HCSS against the disbond region and
(c) 150kHz input signal
A proper transducer arrangement is crucial for the success of GW based NDE of HCSS [15].
In Fig. 1(b), the schematic diagram of the PWT positions on the experimental plate is shown.
It is useful to note that the disbond falls directly along the GW propagation path comprised of
transducer pairs 5-6 and 1-8.
Selection of an appropriate driving frequency is essential for the success of GW-based
SHM and NDE [25]. In order to counter the dispersive effects of the propagating GW signal,
an optimum input signal for the available SP-5H PWTs is selected as 150kHz five–cycle sine
pulse in a Hanning window (Fig. 1(c)) and used for the experiment and subsequent theoretical
and numerical studies unless otherwise stated.
3. Theoretical modeling of HCSS
In order to study the dispersion characteristics and response of the healthy HCSS plate, a
recently developed 2D semi-analytical theoretical model is considered. The schematic
representation of the 2D semi-analytical model of the given HCSS for a horizontal surface
excitation (representing the PWT loading behavior) is shown in Fig. 2.
Fig. 2: 2D schematic representation of the layered HCSS
A detailed formulation of 2D semi-analytical model for wave field calculations in HCSS can
be found in Banerjee and Pol [19] and will not be repeated here for brevity. In this model, the
horizontal surface displacement on the HCSS can be expressed in frequency-wavenumber
domain in the form:
uˆ1 
F (1 ,  )
G (1 ,  )
(2)
where, ω and ξ1 comes from the complicated matrix operation.
The group velocity-dispersion curves can be obtained from the given dispersion condition as:
(3)
G(1 ,  )  0
The values of ξ1 can be determined for a range of values of ω and the corresponding group
velocity-dispersion plots can be sought by using
cg 

1
(4)
where cg represent the group velocity of the propagating GW modes.
The solution for the horizontal surface displacement in the spatial domain can be obtained
through
F (1 ,  ) 1x1
e d1
 G ( ,  )
1
u1 ( x1 , x3 ,  )  

(5)
The integration in equation (5) can be solved by applying the Cauchy‟s residue theorem [26]
on the real roots of ξ1 (ξr, where r = 1 to NR, NR = no. of real roots) as
NR
F ( r ,  )
u1 ( x1 , x3 ,  )  2 i 
eir x1
(6)
r 1 dG (1 ,  )
d1
 
1
r
Finally, an inverse Fourier transform can be performed to the Eq. (6) to get the time-domain
results.
4. Numerical Simulation
A 3D numerical simulation of PWT induced GW propagation in the HCSS with multiple
disbond regions is carried out using FE based explicit and implicit codes in ABAQUS, in
order to study the disbond effects on the propagating GW modes and to generate pseudoexperimental signals for the pseudo-experimental identification of multiple disbonds in a
single HCSS. In the explicit code, the HCSS sample plate is modeled, using the 8-noded
ABAQUS C3D8R elements. The layer-wise element sizes in the HCSS are: GFRC skin: 5mm
× 5mm × 0.05mm, adhesive: 5mm × 5mm × 0.01mm and aluminum core: 5mm × 5mm ×
0.5mm. Then, in the implicit code eight-PWTs (actuator/receiver) are placed at a distance of
400mm on the top surface of the explicit-HCSS model (Fig. 3(a)). The PWTs comprises of
two parts, one is the electrode part (at the bottom) and the other is the piezoelectric part (at the
top), as shown in Fig. 3(b). The 8-noded standard C3D8E elements (1 mm × 1mm × 0.1 mm)
are considered to model the electrode part. In which, zero voltage is assigned on both the top
and bottom surface nodes. The 8-noded standard C3D8E linear piezoelectric brick elements
(1 mm × 1mm × 0.13 mm) are selected for the piezoelectric part (piezo). This C3D8E element
is capable to grasp the electro-mechanical coupling phenomenon, and the electrical voltage is
the additional DOF in the coupling element. The selected input signal (voltage) applied at the
nodes on the top surface of the transmitter-piezo, and zero voltage is applied at the bottom
nodes of both the transmitter and receiver-piezo to model the grounding operation. The output
voltage collected at the top surface of the receiver-piezo. In this study, the SP-5H
piezoelectric wafer transducers (manufacturer: SPARKLER Ceramics Pvt. Ltd., India) are
used [27]. Finally, the ABAQUS „Standard-Explicit Co-simulation‟ option is implemented in
order to create the link between explicit and implicit simulation [28].
(a)
(b)
Fig. 3: (a) Numerical model of HCSS with three disbond regions showing the PWT positions
and (b) an enlarged view of the PWT
5. Results and discussions
5.1 Dispersion curves and baseline response: The theoretical frequency versus group
velocity dispersion curve for the HCSS is theoretically obtained and presented in Fig. 4(a).
1.0
7
A0
6
S0
Theoretical
Numerical
Experimental
S2
A0
5
A1
4
S1
3
2
A2
1
S2
0
0
A2
0.5
50
100
Frequency (kHz)
150
200
S0/A1/S1
Amp*
Group Velocity (mm/s)
8
0.0
-0.5
-1.0
0
50
100
150
Time (s)
200
250
(a)
(b)
Fig. 4: (a) Theoretical group velocity-dispersion curve and the (b) comparison of baseline
theoretical, numerical and experimental time-history responses for the HCSS
The dispersion plot clearly shows the existence of multiple propagating wave modes that
correspond to the real roots of the Eq. (3). The steepest descent method is applied to draw out
these roots [29]. The anti-symmetric modes are designated as A0, A1, A2, etc., while the
symmetric modes are designated as S0, S1, S2, etc. [25]. The theoretical obtained output
response is obtained for a actuator-receiver distance of 200mm on the healthy HCSS and
compared with the numerical and experimental baseline responses, which shows a good
agreement with the presence of six independent GW modes.
5.2 Study of disbond effect: The numerical and experimental output responses are obtained
and compared for with and without disbond cases, in order to understand the disbond-effect
on the propagating GWs (Fig. 5(a,b)). It is observed that due to the presence of disbond in the
HCSS significantly amplifies the A0 mode amplitudes of the received signals for both
numerical simulation as well as experiments.
1.0
1.00
1.0
A2
S2
0.5
A2
Without Disbond
With Disbond
0.5
A0
S2
0.75
S0/A1/S1
0.0
Amp*
Amp*
Amp*
-0.5
Without Disbond
With Disbond
A0
S0/A1/S1
0.0
A2
Without Disbond
With Disbond
-0.5
0.50
S2
0.25
Amplification
of A0-mode
S0/A1/S1
-1.0
0
50
100
150
Time (s)
200
250 -1.0
0
50
100
150
Time (s)
200
250
0.00
0
50
100
150
Time (s)
200
250
(a)
(b)
(c)
Fig. 5: Comparison of the with and without disbond signals from (a) numerical simulation,
(b) laboratory experiment and the (c) WT of the experimental signals at (b)
5.3 Disbond detection algorithm: An SDC based disbond imaging algorithm is used [25],
which uses the Wavelet transform (WT) [18] of the experimental sensor signals in the timedomain. The damage localization probability, Dd, of any arbitrary position (x, y), within the
sensor network is expressed [22] as:
   Aij ( x, y) 
Dd ( x, y)   iN11  Nj i 1Dij ( x, y)   iN11  Nj i 1sdcij ( x, y) 
  1 

(1)
where, Dij(x,y) represent the damage distribution probability, measured from actuator-sensor
pair: i-j and sdcij(x,y) is the signal different coefficient, which is the difference in amplitude
area with disbond and without disbond for a particular GW mode. The SDC can be
represented as:

sdcij 
t2
t1
( s d  s b )dt

t2
t1
(2)
[ s d ]2 dt
d
where, s and sb are the GW signals correspond to the with disbond and without disbond, t1 is
the time of arrival for a particular wave mode and t2 = (t1 + bandwidth of signal),
is the spatial distribution function, which has contour in the shape of ellipse with a positive
value, and
Aij ( x, y) 

Pij ( x , y ), Pij ( x , y )  
,
Pij ( x , y )  
where, Pij ( x, y)  [ ( x  xi )2 ( y  yi )2  ( x  x j )2 ( y  y j ) 2 ] / pij
(3)
(4)
where the β is a small scaling parameter that reduces the size of the disbond zone and it is
independent of propagation velocity and Pij is the distance between actuator „i‟ and receiver
„j‟.
5.4 Disbond identification using experimental signals: The size and location of a hidden
disbond region is identified using the WT of the A0 mode based experimental received
signals (e.g. Fig. 5(c)) in the SDC algorithm. In this study, the 1-2 sensor configuration (path)
is taken as baseline for 3-4, 5-6, and 7-8 sensor path, as the 1-2 path is considerably away
from the disbond region (ref. Fig. 1(b)). Similarly, the sensor path: 2-7 is considered as
baseline for the sensor path: 1-8. In order to obtain the SDC maps, the transformed A0 mode
received signals are collected from the sensor configuration: 1-2, 3-4, 5-6, 7-8, 1-8 and 2-7,
and applied as input to the SDC algorithm. The SDC maps are shown in Fig. 6, which
represents the maximum damage intensity value close to the disbond location. A 3D
representation of the disbond region corresponding to the SDC magnitudes is also shown in
Fig. 6(c) for better understanding of the disbond effect on the SDC. It is expected that the
availability of baseline data will significantly improve the disbond detection capability and
possibly size it with some degree of confidence. However, it is observed that the algorithm is
capable to identify the hidden disbond location using the A0 mode based signal with a
minimum number of sensor paths available.
(a)
(b)
(c)
Fig. 6: SDC maps in (a) contour pattern (b) grid pattern and (c) the 3D representation of the
disbond effects on the SDC-magnitudes
5.5 Multiple-disbond identification using pseudo-experimental signals: The pseudo
experimental results are considered for multiple disbond identification due to the absence of
experimental sample plate with multiple disbond zones. Towards this, a 3D numerical
simulation of the HCSS with presence of three disbond zones (50 mm × 50 mm) is carried out
in ABAQUS. In the PWT-network shown in Fig. 3(a), the „with disbond‟ and „without
disbond‟ signals are collected from different receiver PWTs. The actual experimental system
noise is considered, as shown in Fig. 7(a). A histogram (Fig. 7(b)) of the experimental-noise
is obtained in MATLAB. The relationship between number of data and number of bins can be
represented [30] as: W = 1 + 3:3 log10 (NS)
(5)
where, W and NS are the number of bins and data points, respectively.
Displacement (mm)
0.002
0.001
0.000
-0.001
-0.002
0
50
100
150
200
250
Time (s)
(a)
(b)
Fig. 7: (a) Actual experimental-noise and (b) Histogram showing a normal distribution
In the study, a 250µs noise with 748 data points is considered for generating the
histogram, where the number of bins was obtained as 11. The normal distribution is fitted
well to the histogram. Once, the noise is characterized, random numbers are generated based
on the parameters (mean = 0.000519065 and variance = 2.18356e-05) of the distribution. The
pseudo-experimental signals are then generated by adding this simulated noise to the
numerically obtained signals.
(a)
(b)
(c)
Fig. 8: SDC map in (a) contour pattern, (b) grid pattern and (c) 3D representation, showing
the exact location of the disbond zones in reference to Fig. 3(a)
The pseudo-experimental signals correspond to different actuator-sensor paths are
applied in the SDC algorithm to examine the potential of the proposed NDE technique for
multiple disbond identification in the HCSS. It is observed that the SDC-maps (Fig. 8)
obtained by using the pseudo-experimental signals in the algorithm have shown the exact
location and approximate size of the hidden disbond zones. Fig. 8(c) shows a 3D
representation of the SDC map with prominent effects from the multiple disbond zones.
6. Conclusions
The GW propagation mechanism in the HCSS is multi-modal in nature. Existences of six
independent GW modes (A0, S0, A1, S1, A2 and S2) are found in the received signals at
150kHz frequency. The presence of disbond in the HCSS leads to significant amplification of
the A0 mode. The proposed SDC-algorithm based NDE technique is capable to
experimentally identify the approximate size and exact location of hidden disbond region in
the HCSS within an actuator/sensor network, using minimum number of sensor paths. The
proposed technique has also shown its potential to efficiently identify multiple disbond
regions in the HCSS, using the transformed pseudo-experimental signals.
References
1. J Fatemi and M H J Lemmen, „Effective Thermal/Mechanical Properties of Honeycomb
Core Panels for Hot Structure Applications‟, Journal of Spacecraft and Rockets, Vol 46,
No 3, pp 514–525, 2009.
2. G A O Davies, D Hitchings and J Ankersen, „Predicting Delamination and
Debonding in Modern Aerospace Composite Structures‟, Composites Science and
Technology, Vol 66, No 6, pp 846–854, 2006.
3. M Mitra and S Gopalakrishnan, „Guided wave based structural health monitoring: A
review‟, Smart Materials and Structures, Vol 25, No 5, p 053001, 2016 Mar 30.
4. C Ramdas, K Balsubrmanyam, M Joshi and C V Krishnamurthy, „Interacton of guided
lamb waves with an asymmetrically located delamination in a laminated composite
plate‟, Smart Mater. Struct., Vol 19, No 6, p 065009, 2010.
5. H Baid, S Banerjee, A K Mal and S Joshi, 2008. Detection of disbonds in a honeycomb
composite structure using guided waves. In: Proceedings of SPIE, Smart Structures
and Materials & Nondestructive Evaluation and Health Monitoring VI, March 9–13,
San Deigo, California.
6. T R Hay, L Wei and j L Rose, „Rapid Inspection of Composite Skin-Honeycomb Core
Structures with Ultrasonic Guided Waves‟, Journal of Composite Materials, Vol 37,
No 10, pp 929–939, 2003.
7. K I Maslov and T Kundu, „Selection of lamb modes for detecting internal defects in
laminated composites‟, Ultrasonics, Vol 35, pp 141–150, 1997.
8. B Xu and V Giurgiutiu, „Single mode tuning effects on Lamb wave time reversal with
piezoelectric wafer active sensors for structural health monitoring‟, J. Nondestructive.
Eval., Vol 26, pp 123–134, 2007.
9. Z Su and L Ye, L., “Fundamental Lamb Mode-Based Delamination Detection for
CF/EP Composite Laminates Using Distributed Piezoelectrics,” Structural Health
Monitoring: An International Journal, Vol. 3, No. 1, 2004, pp. 43–68. doi:
10.1177/1475921704041874.
10. Z Su, C Li, X M Wang, L Yu and C Zhou, „Predicting Delamination of Composite
Laminates Using an Imaging Approach‟, Smart Materials and Structures, Vol 18, No
7, p 074002, 2009.
11. L Wang and F G Yuan, „Damage Identification in a Composite Plate Using Prestack
Reverse- Time Migration Technique‟, Structural Health Monitoring: An International
Journal, Vol 4, No 3, pp 195–211, 2005.
12. N Bourasseau, E Moulin, C Delebarre, and P Bonniau, „Radome Health Monitoring
with Lamb Waves: Experimental Approach‟, Nondestructive Testing and Evaluation,
Vol 33, No 6, pp 393– 400, 2000.
13. F Song, G L Huang and K Hudson, „Guided Wave Propagation in Honeycomb
Sandwich Structures Using a Piezoelectric Actuator/Sensor System‟, Smart Materials
and Structures, Vol 18, No 12, p 125007, 2009.
14. F He, Z Zhou and Z Feng, „Research on an inspection method for de-bond defects in
aluminum skin-honeycomb core sandwich structure with guided waves‟, In17 th world
conference on nondestructive testing, Shanghai, China, Oct. 2008.
15. S Sikdar, S Banerjee and G Ashish, „Ultrasonic guided wave propagation and disbond
identification in a honeycomb composite sandwich structure using bonded
piezoelectric wafer transducers‟, Journal of Intelligent Material Systems and
Structures, Vol 27, No 3, pp 1767-1779.
16. A K Mal, „Wave propagation in layered composite laminates under periodic surface
loads‟, Wave Motion, Vol 10, pp 257–266, 1988.
17. A Chakraborty and S Gopalkrishnan, „A spectrally formulated finite element for wave
propagation analysis in laminated composite media‟, Int. J. Solids Struct., Vol 41, pp
5155-5183, Northampton, England, 2008.
18. C Pol and S Banerjee, „Modeling and analysis of propagating guided wave modes in a
laminated composite plate subject to transient surface excitations‟, Wave Motion, Vol
50, pp 964-978, 2013.
19. S Banerjee and C Pol, „Theoretical modeling of guided wave propagation in a
Sandwich plate subjected to transient surface excitations‟, Int. J. Solids Struct., Vol
49, pp 3233–3241, 2012.
20. X Qi, J L Rose and C G Xu, „Ultrasonic Guided Wave Nondestructive Testing for
Helicopter Rotor Blades‟, 17th World Conference on Nondestructive Testing, British
Institute of Non- Destructive Testing, Northampton, England, 2008.
21. D Wang, L Ye, Y Lu and F Li, „A Damage Diagnostic Imaging Algorithm Based on
the Quantitative Comparison of Lamb Wave Signals, Smart Materials and Structures‟,
Vol 19, No 6, p 065008, 2010.
22. X L Zhao, H D Gao, G F Zhang, B Ayhan, F Yan, C Kwan and J L Rose, „Active
Health Monitoring of an Aircraft Wing with Embedded Piezoelectric Sensor/Actuator
Network: 1. Defect Detection, Localization and Growth Monitoring‟, Smart Materials
and Structures, Vol 16, No 4, pp 1208–1217, 2007.
23. S Sikdar, S Banerjee and S M Subhani, „Detection of Disbond in a Honeycomb
Composite Sandwich Structure Using Ultrasonic Guided Waves and Bonded PZT
Sensors‟, Proceedings of the 14th Asia Pacific Conference on Non-destructive Testing
(APCNDT), Mumbai, India, 18-22 November, 2013.
24. S Sikdar and S Banerjee, „Wave Propagation in a Honeycomb Composite Sandwich
Structure in the Presence of High-Density Core Using Bonded PZT-Sensors‟,
Vibration engineering and technology of machinery, Mechanisms and machine
science, Vol 23, pp 619-628, 2014.
25. S Sikdar and S Banerjee, „Identification of disbond and high density core region in a
honeycomb composite sandwich structure using ultrasonic guided waves‟, Composite
Structures, Vol 152, pp 568–578, 2016.
26. L V Ahlfors, „Complex analysis‟, McGraw Hill, 1979, ISBN 0-07-085008-9.
27. S Sikdar and S Banerjee, „Guided Wave Propagation in a Honeycomb Composite
Sandwich Structure in Presence of a High Density Core‟, Ultrasonics, Vol 71, pp 8697, 2016.
28. M H Soorgee, C J Lissenden, J L Rose, et al., „Planar guided waves for SHM of plate
structures using piezoelectric fiber transducers‟, In: Review of Progress in
Quantitative Nondestructive Evaluation, Vol. 32, pp. 254–261, American Institute of
Physics Proc. 1511, July 2012.
29. S Banerjee, W H Prosser and A K Mal, „Calculation of the response of a composite
plate to localized dynamic surface loads using a new wave number integral method‟,
ASME J Appl Mech, Vol 72, pp 18–24, 2005.
30. H Sturges, „The choice of a class interval‟, Journal of the American Statistical
Association, Vol 21, No 153, pp 65-66, 1926.
NDE Assessment of Aerospace Components – IGCAR Experiences
B Venkatraman
Indira Gandhi Centre for Atomic Research
Kalpakkam – 603 102, India
Contact: bvenkat@igcar.gov.in
The desire to non-invasively evaluate the quality and fitness for purpose of materials and structures
has a long history starting from the tap test by ancient potters. While the physical basis of the present
day Non Destructive Testing (NDT) methods had their origin in the 18th and 19th century, actual
large scale application of NDT in the industry was only after the World Wars. Radiography,
ultrasonics, penetrant and magnetic particle inspection along with visual testing formed the core NDT
methods. During this period, NDT was also qualitative and a “go no go” technique. The renaissance
in NDT actually occurred during the 1980s with the development of fracture mechanics concepts and
advances in sensors, instrumentation coupled with image and signal processing techniques. NDT
transformed to Non Destructive Evaluation (NDE) becoming quantitative. NDE today is an
indispensable part of any industry, being used right from design through validation, fabrication, pre
and in-service inspections including remnant life prediction and extension. In India and
internationally, it was the strategic industries – aerospace, nuclear and defence that have contributed
extensively for the growth and application of NDE.
Non Destructive Evaluation plays an important and catalytic role in the aerospace industry. While
NDE techniques at the prefabrication stage have been well established, it is the in service inspection,
structural health monitoring (SHM) and ageing aircraft inspection that continue to provide the
challenges as well as spur the technological developments in NDE.
From the materials perspective, while metallic materials continue to be used for aero structures, the
content of composites is continuously increasing due to their unique features such as excellent
strength/weight ratio, corrosion resistance and the possibility to manufacture samples of complicated
shapes. For the metallic materials, standard NDE methods are well established. However, inspection
of composites pose challenge as the inherent physical characteristics, anisotropic nature, damage
mechanisms and the types of flaws to be evaluated is different.
It is a well realized fact that each NDE method has its advantages as well as limitations and no single
method can provide all the necessary information demanded. Multi modal NDE approaches coupled
with advanced signal and image analysis offer greater probability of detection, better sensitivity and
robust reliability of inspection. IGCAR has contributed significantly for Indian Air Force and Indian
Space Research Organisation (ISRO) for solving challenging problems such as life extension of
defence aircrafts and helicopters, detection of blockage in GSLV fuel flow lines, qualification of
satellite gas bottles, earth sensors of INSAT, development of procedure and techniques for neutron
radiographic inspection of critical pyro devices using the KAMINI (mini reactor) at IGCAR and
phased array ultrasonic inspection of maraging steel rocket motor casings and silicon steel motor
casings. Pulsed Phase thermography is a well-established advanced NDE technique for composite
inspection. Apart from studying the damage in composites using thermography, it has been adapted
for qualification of the repair work carried on the composites. Stress corrosion cracking in primary
structural components is one of the most pressing issues associated with aging aircraft. In order to
improve the corrosion resistance retrogression and re-aging (RRA) heat treatment have been
developed. Thermal property measurement of these retrogressed and re-aged Al alloy samples were
determined non-invasively using novel Photothermal methods. Vibrothermography for turbine blade
for crack detection in the turbine blade is another upcoming area which helps in defect select imaging
with zero contribution of unwanted signal. This presentation focusses on all the above mentioned
works that have been undertaken at the author’s lab as well as in IGCAR.
While the application of NDE has been progressing for quality and performance assessment of
materials and components, the latest domains of research are in the application of forward and inverse
modeling coupled with experimental NDE for materials characterization, material property
correlation, and online detection of early microstructural damage and understand damage mechanics.
Non-linear ultrasonic techniques based on second and third harmonic analysis, digital image
correlation coupled with thermal microscopy etc. are being employed especially for creep damage
characterisation and for studying the strain rate effect on deformation kinetics under monotonic
loading conditions. A number of new developments have also been made in the areas of transient and
lock in thermography for defect characterisation, ultrasound thermography for linear defect detection
and thermal signal reconstruction techniques. This presentation would also dwell on these and
highlight the significant role of innovations and synergistic multi-disciplinary approaches for
enhanced structural integrity assessment.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Subwavelength resolution of delaminations
Kiran Kumar AMIREDDY, Prabhu RAJAGOPAL and Krishnan BALASUBRAMANIAM
Centre for Non-destructive Evaluation, Department of Mechanical Engineering,
Indian Institute of Technology- Madras, Chennai 600036, Tamil Nadu, India. Phone: +91-442257-4662; e-mail: amireddykiran@gmail.com; prajagopal@iitm.ac.in; balas@iitm.ac.in
Abstract
In this paper, we demonstrate the application of a holey structured metamaterial lens for sub-wavelength imaging
of defects in a laminated metallic sample in the ultrasonic regime. Finite Element (FE) simulations are used to
study longitudinal wave interaction with ideal delaminations in isotropic elastic materials. Holey-structured
metamaterial lens is then used to transmit the scattered wave field. This paper discusses the subwavelength
resolution of λ/7 delamination in a laminated aluminium sample which to the best of our knowledge this is the
highest resolution achieved in the ultrasonic regime.
Keywords: Diffraction limit, Delamination, Subwavelength imaging, and Holey-structured metamaterial lens.
1. Introduction
Delamination is a common and critical failure mechanism in aerospace structures, created by
fatigue, impact stresses during in-service applications [1]. In these applications, the most
commonly used non-destructive test is ultrasonic inspection. Diffraction sets a natural limit on
resolution, by any wave modality which limits to half the wavelength of the wave used for
inspection [2]. To inspect small delaminations, it is necessary to use high frequency (for small
wavelength), but in composites high frequency waves are highly attenuated [3]. Hence it is
difficult to characterize the small delaminations in the composites. Detection and
characterization of subwavelength delaminations are most important in aerospace structural
applications otherwise these leads to the sudden catastrophic failure at the high speed. Here
we present a technique to characterize the subwavelength delaminations in metals by using
periodic holey structured metamaterials. We report the resolution of a subwavelength (λ/7)
delamination in a bonded 2 layer metallic (aluminium) sample.
This paper is organised as follows. We begin with an introduction to delamination in layered
media, and its detection and characterization importance during the in-service applications
fallowed by its background. Then we present the problem studied for analysis and the detailed
explanation of the procedure followed for Finite Element (FE) simulation. Then finally we
present the results and discussions with an implication to the future work.
2. Background
Holey structured metamaterials work on the principle of Fabry-Perot resonances inside the
holes when the scattered waves are propagating through it. With this concept holey-structured
meta-lens transfers both the propagating and evanescent waves from input to output surfaces
and hence subwavelength information carried by the evanescent waves helps to create the
image with high resolution [4, 5].
2.1 Problem studied
We consider a laminated aluminium sample having a subwavelength delamination of length
1.8 mm as shown in Figure 1. The delamination present in the laminated aluminium sample is
considered as the defect for imaging purpose. The length of the crack (1.8 mm) is about λ/7
for a frequency of 500 kHz. This object is imaged with ultrasonic immersion through
transmission to resolve the subwavelength delamination in its image with the help of periodic
holey-structured metamaterial lens.
Figure 1. Aluminium sample with the details of subwavelength delamination presented in it (dimensions are in
mm).
3. FE simulations
Commercially available Finite Element (FE) package [6] is used to model the wave
propagation through the Aluminium sample and metamaterial. This was modelled by two
parts, one is aluminium sample with delamination and the other is metamaterial immersed in
water as shown in Figure 2. A 2-D FE model was created with dimensions 200x200 mm2
chosen to avoid reflections from the boundaries. Mechanical properties of aluminium were
assigned to the model, with density ρ = 2700 kg/m3, Young’s modulus of elasticity E = 69
GPa, and Poisson’s ratio υ = 0.334. Delamination was created in the aluminium by selecting
the nodes of required dimension (1.8mm) and applied rigid boundary conditions
(displacement is zero). A 2-D model of size 240x120 mm2 was created and assigned water
(acoustic medium) properties, with density ρ = 1000 kg/m3 and Bulk modulus K = 2.2 GPa.
2.2 Gpa.
Figure 2. Snapshot of FE model with mesh.
A 2-D holey structured metamaterial of length 13 mm with a hole size of 1.5 mm and a
periodicity of 2 mm is created by setting rigid (pressure is zero) boundary conditions on
selected nodal lines. CPE4R- A four-node bilinear plane strain quadrilateral, reduced
integration, hourglass control mesh was used for aluminium sample and acoustic element
AC2D4R was chosen for meshing of metamaterial region with a seed size of 0.15mm. This is
about λ/20 for 500 kHz in water for mesh convergence [7]. Tie constraint was given between
the two models to allow the wave propagation from one part to other. The model was excited
by applying a periodic force of 3 cycle Hanning windowed tone burst signal of central
frequency 500 kHz. An iteration step time of 1µs was used. This analysis is run for a total
time of 90 µs which was enough for the waves to reach the end of the model once.
4. Results and Discussion
After completion of the line scan the maximum amplitude variation across the measurement
positions (for both experiment and simulation) are plotted as shown in Figure 3. In this plot
we can clearly see that at the position of the delamination the amplitude drops down which is
indicated with dotted rectangular box. The two edge diffractions from the delamination in the
results are clearly matching with the exact dimension in the sample.
Figure 3. Experimental and simulated results for normalized amplitude variation with the measurement position
across the sample. The dashed lines represent the position of the subwavelength (λ/7) delamination in the
laminated aluminium sample.
This clearly shows that the holey-structured metamaterial is resolved the subwavelength (λ/7)
sized delamination in the laminated sample. Hence the proposed technique can useful to
detect and characterize the subwavelength sized in-service defects like delaminations,
disbanding, in the composite laminates. By changing the parameters of the holey structured
metamaterial the resolution can be greatly improve and the authors are presently working on
it.
References
1.
2.
3.
4.
5.
6.
7.
M.R. Wisnom, ‘The role of delamination in failure of fibre-reinforced composites’, Phil.
Trans. R. Sco. A, 370, 1850-1870, 2012.
X. Zhang and Z. Liu, ‘Super lenses to overcome the diffraction limit’, Nat. Mater. 7(6),
435–441, 2008.
S. Biwa, Y. Watanabe, S. Idekoba, N. Ohno, ‘Wave Scattering and Attenuation in
Polymer-Based Composites: Analysis and Measurements’, IUTAM Symposium on
Dynamics of Advanced Materials and Smart Structures, Volume 106 of the series Solid
Mechanics and Its Applications, Pp. 19-28, 2003.
J. Zhu, J. Christensen, J. Jung, L. Martin-Moreno, X. Yin, L. Fok, X. Zhang, and F. J.
Garcia-Vidal, ‘A holey-structured metamaterial for acoustic deep-subwavelength
imaging’, Nat. Phys. 7(1), 52–55, 2011.
K. K. Amireddy, K. Balasubramaniam, and P. Rajagopal, ‘Holey-structured
metamaterial lens for subwavelength resolution in ultrasonic characterization of metallic
components’, Appl. Phys. Lett. 108, 224101, 2016.
See
http://www.3ds.com/products/simulia/portfolio/abaqus/abaqus-portfolio
for
ABAQUS. Analysis User’s Manual. Version 6.10-1; accessed 28 July, 2015.
A. Ramdhas, R. K. Pattanayak, K. Balasubramaniam, and P. Rajagopal, ‘Symmetric
low-frequency feature-guided ultrasonic waves in thin plates with transverse bends’,
Ultrasonics 56, 232–242 (2015).
Anisotropic Magnetic Properties of Grain-oriented and Non-oriented Si-Fe
Electrical Steel
Gholamhossein Shirkoohi
School of Engineering
London South Bank University
103 Borough Road, London SE1 0AA, England
Contact: maziar.shirkoohi@lsbu.ac.uk
In earlier studies, Global anisotropic variation of the steels was shown to be directly proportional to
that of the intrinsic anisotropy energy of the cubic single crystal [1]. Non-oriented electrical
steels are usually used in construction of rotating machines and in other devices, where rotational
fields frequently occur [2]. Grain-oriented electrical steels are on the other hand used in construction
of the magnetic core of power transformers, reactors and Generators, where unidirectional field
propagation is enforced by design. During their non-oriented steels acquire mild anisotropic properties
simply due to their cold rolling manufacturing process; whereas manufacturing process is designed to
encourage grain growth, such that, the easy axes of the cubic crystals are generally ordered in the
rolling direction of the sheet steel. This results in a favourable behaviour where the magnetic
properties are best along this direction. The orientation of the crystals in grain-oriented steels involves
a cube-on-edge configuration, and since the cubic crystals of silicon-iron exhibit intrinsic anisotropy
[3], the result is a high deterioration of the magnetic properties in other directions. Epstein size (30
mm x 300 mm) samples of 0.27 mm thick conventional and high permeability grain- oriented
electrical steel materials were cut between 0º and 90º. Magnetic properties of these materials were
measured in these directional samples under pure sinusoidal flux conditions, ranging between 0 and
2.0 T, 50 Hz. the maximum induction level was limited to 1.3 T, for magnetisation directions near
50º and 60º due to the lower saturation magnetization limits near the global hard axis of the sheet
steel. The properties are also investigated under very high field regime, close to near full saturation
[4].
References
1. G. Shirkoohi, M. Arikat, Anisotropic properties of high permeability grain-oriented 3.25% Si-Fe
electrical steel, IEEE Transactions on Magnetics 30(2):928-930 · April 1994, DOI:
10.1109/20.312448
2. G.H. Shirkoohi, Anisotropic Properties of Low Grade Non-Oriented Electrical Steels,
ELMECO'94 International conference on Electromagnetic Devices and Processes in
Environmental Protection, Lublin, Poland, Sept. 1994:141-146. ISBN 83-86333-60-X
F. Brailsford "Physical Principles of Magnetism" Van Nostrand, London, 1966:121G. Shirkoohi,
Dependence of Magnetisation near Saturation on Alloying Content in Ferromagnetic Steel, IEEE
Transactions on Magnetics, 51(7):1-10, July 2015, DOI: 10.1109/TMAG.2015.2405056
Sonic Analysis System as an effective means of NDT
analysis for cast iron nodularity measurement
1
Kalyan Ram B., 2S.Arun Kumar, 3Prajval M S.
1,2,3
Electrono Solutions Pvt Ltd.
1
2
kalyan@electronosolutions.com
arun@electronosolutions.com
3
prajvalms@electronosolutions.com
Abstract— Sonic Analysis System is a Non Destructive Testing
Technique to differentiate between the nodularity of Cast iron in
a casting environment
The objective of this solution is to achieve 100% inline testing
of cast iron components in a manufacturing environment with a
defined Pass/Fail criteria for their nodularity levels which is
critical to certain applications such as power transmission,
brakes and other components subjected to high torsion and
friction loads. This facilitates eliminating failed components from
going into the next levels of Manufacturing process thereby
improve the overall supply quality, reduce rejections and save
costs.
Sonic Analysis mechanism operates in the audio frequency
range. Every material produces a different/unique signature
frequency when struck with a solid object, such as gong. The
captured signal is analyzed for all of its signal parameters such as
Fundamental Frequency, Amplitude, Harmonics, Disturbance
settling time and so on. Critical automotive parts such as Power
transmission units, Brake units and the like that require
structural frequency response characteristics by real time modal
analysis in the manufacturing line is need of the hour. Such
analysis in an inline process needs to be accurate, reliable and
should support high throughput.
Sonic Analysis method helps address the above mentioned
requirements and beyond.
A typical chemical analysis of this material:

Carbon 3.2 to 3.6%

Silicon 2.2 to 2.8%

Manganese 0.1 to 0.5%

Magnesium 0.03 to 0.05%

Phosphorus 0.005 to 0.04%

Sulfur 0.005 to 0.02%

Copper <0.40%

Iron balance %
As indicated above, Magnesium is the nodulizing element and
is critical in defining a Ductile Cast Iron.
Why bother about Nodularity?
Cast iron contains high carbon content, and during
solidification of the metal, the carbon content can form
irregularly shaped graphite flakes that disrupt the crystalline
structure, causing cracks and brittleness.
In ductile iron, the graphite forms spherical, rounded nodules
that inhibit the formation of cracks and provide enhanced
ductility and machinability. The higher the nodularity, the
higher the ductility.
Keywords— Sonic Analysis System, Nodularity, Ductile, Gray,
SG, cast iron, NDT for automotive components, frequency.
amplitude, harmonic analyis for component testing.
I.
INTRODUCTION
Sonic Analysis System in this context is being used to
differentiate between Ductile Cast and Gray Cast components.
Ductile iron is not a single material but is part of a group of
materials which can be produced to have a wide range of
properties through control of the microstructure. The common
defining characteristic of this group of materials is the shape of
the graphite. In ductile irons, the graphite is in the form
of nodules rather than flakes as it is in gray iron. The sharp
shape of the flakes of graphite create stress concentration
points within the metal matrix and the rounded shape of the
nodules less so, thus inhibiting the creation of cracks and
providing the enhanced ductility that gives the alloy its
name. The formation of nodules is achieved by the addition
of nodulizing elements, most commonly magnesium (note
magnesium boils at 1100°C and iron melts at 1500°C)
Fig1: 91% nodularity
Fig 2: 54% nodularity
Compared to gray iron, nodular iron (interchangeably referred
to as ductile iron) has an absolute advantage in intensity. The
max tensile strength of nodular iron is 90k psi, while the max
tensile strength of gray iron is only 35k psi.
Nodular Irons are generally superior to gray irons, regarding
their yield strength. The max yield strength of ductile iron is
40k psi; Gray iron is not very malleable or strong, it fractures
easily.
Nodular iron is more flexible and elastic than other cast irons.
Nodular iron has higher strengths, greater elongation and better
resistance to impact than gray iron.
observed that evaporation of magnesium is a continuous
process and the reduction in nodularity is a continuous function
of time.
For these features of Nodular iron (Ductile iron), it is used for
critical applications such as Power transmission, Brakes,
Suspensions and so on.
On testing the nodularity in each of the ladels from the first to
the last, it could be noticed that the nodularity drops from the
first to the last. The objective is to ensure that the given
pouring and solidification process should be completed within
the time when nodularity drops below 85%. This depends on
the cast being made and the pouring/solidification process
being followed. Typically it lies in the scope of manufacturing
process. Several checks and control could be put in place to
ensure quality assurance at this level.
History of Ductile Iron Development
In spite of the progress achieved during the first half of 19th
century in the development of Gray and Malleable Irons,
foundrymen continued to search for the ideal cast iron - an ascast "gray iron" with mechanical properties equal or superior to
Malleable Iron.
J.W. Bolton, speaking at the 1943 Convention of the American
Foundrymen's Society (AFS), made the following statements.
"Your indulgence is requested to permit the posing of one
question. Will real control of graphite shape be realized in gray
iron? Visualize a material, possessing (as-cast) graphite flakes
or groupings resembling those of malleable iron instead of
elongated flakes."
A few weeks later, in the International Nickel Company
Research Laboratory, Keith Dwight Millis made a ladle
addition of magnesium (as a copper-magnesium alloy) to cast
iron and justified Bolton's optimism - the solidified castings
contained not flakes, but nearly perfect spheres of graphite.
Ductile Iron was born!
Five years later, at the 1948 AFS Convention, Henton Morrogh
of the British Cast Iron Research Association announced the
successful production of spherical graphite in hypereutectic
gray iron by the addition of small amounts of cerium.
At the time of Morrogh's presentation, the International Nickel
Company revealed their development, starting with Millis'
discovery in 1943, of magnesium as a graphite spherodizer. On
October 25, 1949, patent 2,486,760 was granted to the
International Nickel Company, assigned to Keith D. Millis,
Albert P. Gegnebin and Norman B. Pilling. This was the
official birth of Ductile Iron, and, as shown in Figure 2.6, the
beginning of 40 years of continual growth worldwide, in spite
of recessions and changes in materials technology and usage.
What are the reasons for this growth rate, which is especially
phenomenal, compared to other ferrous castings?
How does Nodularity
Manufacturing Process?
get
affected
during
However, from the end of line testing and dispatch perspective,
an independent testing process is required to ensure Supply
Quality and to reduce Rejections. It is in this context that the
Sonic Analysis System is critical to the process.
II. RELATED WORKS
With the demand for zero rejections becoming the norm in
automotive manufacturing sector, several NDT techniques
have been emerging to the fore front since over 2 decades.
Though over several years earlier, there have been many papers
published about defect identification in metals and non-metals
using several techniques, during the PANNDT (Pan-American
conference for Non-Destructive Testing) 2003 in Rio, Brazil,
Ingolf Hertlin (Germany) and Detlev Schultze (Brazil)
presented a paper on "Acoustic Resonance Testing: the
upcoming volume-oriented NDT method".
Further to this, there have been several white papers and
conference proceedings published by "The Modal Shop Inc."
of Cincinnati, Ohio.
III.
ARCHITECTURE OF SONIC ANALYSIS
SYSTEM
The architecture of Sonic Analysis System could be
categorized under 2 major types of systems:
1) Stand Alone Architecture for Sonic Analysis System
The block diagram below represents the Acoustic chamber
of the Sonic Analysis system.
the
As mentioned above, the process of making Ductile Iron
involves addition of Magnesium into the molten iron (at 1450
deg C). It is then poured into the ladels of cast and allowed to
cool to form the castings as desired. Typically, these casting
are moved over a conveyor and the pouring happens into each
of the castings.
It is to be observed that at the high temperatures of 1450 deg C
of the molten metal where magnesium is added, over a period
of time, magnesium evaporates and thus causes the - so - called
- Ductile Iron to get converted back to Gray Iron. So, it is
critical that the pouring and solidification should happen before
the magnesium evaporates from the mixture. It is also to be
Fig 3: Block diagram of Acoustic Chamber
Acoustic chamber is central to the Sonic Analysis System
that helps provide the appropriate information about the
component under test to the controller to facilitate the pass/fail
decision making accordingly. This involves two important
aspects:
a. Ensure the obtained sonic information from the
component strike is the signature of the component under test
alone and is appropriately isolated from the rest of the system.
b. The external noise reduced to maintain the acoustics in
the system thereby maintaining appropriate signal to noise
ratio.
A Sonic Analysis System could be categorized under
several sub-systems such as:
iv.
B/W laser Printer
v.
Sonic Analysis System Software
e) User Interface/ HMI
i.
32" LED Display monitor
ii.
Emergency Stop Switch
iii.
Start push button
iv.
Cycle completion indicator
A standalone system block diagram integrated with its sub
systems would be as shown below.
a) Mechanical Subsystem
i.
Workstation to hold the Acoustic chamber
ii.
Acoustic chamber
iii.
Component mounting assembly
iv.
Load cell mounting assembly
v.
Striker and its mounting assembly
vi.
Vibration sensor mounting
Fig 4: Stand Alone Architecture - Sonic Analysis System
vii.
Microphones mounting
2) Inline Architecture for Sonic analysis System
viii.
Pass/Fail indication on component - Marking or
Paint etc.
b) Sensors and Actuator Subsystem
i.
Load Cell
ii.
Microphones
iii.
Vibration sensor
iv.
Striker motor
v.
Striker position feedback sensor
vi.
Loading/Unloading door open/close sensor
vii.
Camera (Web cam)
viii.
Appropriate light source inside the chamber
ix.
In an Inline Architecture, the objective is to feed the system
with the components for test through a conveyor. This will
eliminate the need to put in either an operator or a robotic
system for loading and unloading of components. In this
architecture, in addition to the components of sub-systems
mentioned in the Stand alone architecture, some components
such as a conveyor, Component detect sensor on the conveyor,
Conveyor control system and software update for the same are
some of the add-ons needed.
Paint Level check for Pass/Fail indication system
c) Electrical Subsystem
i.
Panel Box with accessories
ii.
Striker motor driver
iii.
Fixed Power Supply SMPS 24Vdc/12Vdc/5Vdc
iv.
UPS
v.
Switching system with relays/Opto Isolators
d) Electronic Controls and Software
i.
Advanced Digital Multimeter system
ii.
Striker motor controller
iii.
Computer system with keyboard and mouse
Fig 5:Inline Architecture - Sonic Analysis System
The introduction of a conveyor system would reduce the
cycle time and increase productivity which is the critical
requirement in a mass production application. The typical cycle
time in an Inline architecture could be optimized to the levels
of 25% to 50% of the cycle time in a Stand alone architecture.
However, this architecture also comes with the limitation
that the component weight would not be available as the
components are moving over a conveyor and is stuck for sonic
information without moving the component out of the
conveyor system. Even the vibration sensor is required to be
moved to the striker. Though the shifting of vibration sensor to
the striker does not affect the system performance, it certainly
adds to the complexity in the manufacturing of the system and
difficulty in its maintenance.
Further, in case of the Inline architecture system, in
addition to the signal processing, the timing and
synchronization of conveyor control with pass/fail decision
making, component marking and segregation requires to be
addressed.
IV. SYSTEM IMPLEMENTATION AND FEATURES
Sonic Analysis System has been developed from the
concept and architectures mentioned above to help
manufacturers of ductile cast products ensure and maintain the
quality of supply of these critical high nodularity parts to
support the applications requiring it.
As explained in the architectures, both the Stand alone and
Inline architectures were implemented.
The Stand alone system was designed to establish the proof
of concept. The subsystems as mentioned in the architecture
are integrated to for the system shown in the image below.
Fig 8:Inline Sonic Analysis System
Fig 6:Standalone Sonic Analysis System
The signals received from the Sonic Analysis System is
conditioned accordingly. In the screen shot shown below, both
the time domain and frequency domain signals are shown
below.
Fig 9:Acoustic chamber of Inline Sonic Analysis System
Fig 10:GUI of Inline Sonic Analysis System
Fig 7:Time domain and Frequency domain test result
The time domain output are analyzed for the number of
peaks and the settling time. After applying fourier transform to
the time domain signal, the obtained frequency domain signal
is analyzed for frequency information and is evaluated
accordingly.
Features of Sonic Analysis System

Capable to differentiate between objects made of
different types of Cast Iron such as Ductile Cast Iron
and Gray Cast Iron

It can be extended to any other type of metallic
objects as well with appropriate calibration

The end-to-end cycle time is less than 45 seconds.
Could be optimized further on case to case basis.

User friendly interface for differentiating objects as
Pass/Fail

Non-Contact type measurement detection mechanism

This mechanism doesn’t use any consumables

No replaceable parts due to wear and tear of the
measurement system as the measurement mechanism
is non-contact type

Maintain the traceability of parts tested

Requires very little maintenance
V. USAGE AND RESULTS
The system is currently being used and has tested over
60,000 components over a span of 2 months. In order to define
the operation of the system, a given component needs to be
identified for its strike point as different strike points results in
different frequencies. A sample of such test is shown below for
identical Ductile and Grey Cast Iron component.
Date
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
19-02-16
Time
15:34:03
15:34:05
15:35:01
15:35:06
15:35:16
15:35:27
15:35:38
15:35:44
15:35:49
15:35:54
15:36:04
15:36:14
15:36:26
15:36:37
15:36:47
15:36:52
15:37:02
15:37:13
15:37:23
15:37:34
15:37:39
15:37:50
15:38:00
15:38:05
Component Type Ladel Frequency Result
800
10
2078
Pass
800
10
2078
Pass
800
10
2086
Pass
800
10
2078
Pass
800
10
2075
Pass
800
10
2080
Pass
800
10
2078
Pass
800
10
2088
Pass
800
10
2086
Pass
800
10
2077
Pass
800
10
2075
Pass
800
10
2080
Pass
800
10
2079
Pass
800
10
2088
Pass
800
10
2086
Pass
800
10
2078
Pass
800
10
2075
Pass
800
10
2080
Pass
800
10
2078
Pass
800
10
2088
Pass
800
10
2086
Pass
800
10
2078
Pass
800
10
2074
Pass
800
10
2080
Pass
Fig 12:Test results of a Inline Sonic Analysis System
The system is tested for repeatability and the data thus
obtained is plotted with the serial number of the component
tested. The first dataset shows a variation of 1Hz over the range
of 50 tested components.
Fig 11:Test results at several location of a given part
In addition to the application discussed herewith, Sonic
Analysis System could be used for
 cracks inside of the part,
 crack detection during shearing of bars,
 nodularity (of cementite or graphite),
 inclusions,
 density differences in sinter metal products,
 hardness differences (heat treated or aged),
 bonding (welding, friction welding).
Further in the system implementation, the live data
captured from the Inline Sonic Analysis System is shown in the
figure below with several parameters including datetimestamp. Such data helps in maintaining traceability of each
of the components supplied to the customer.
Fig 12:Data of Frequency Vs Component serial number
The second dataset shows no variation of frequency over
the range of 50 tested components. This concludes that the
results are repeatable and consistent.
Fig 13:Data of Frequency Vs Component serial number
In order to establish a correlation between Nodularity and
Frequency, the components are tested for nodularity using
Image Analyzer System (which is the existing process to
determine nodularity of a given part). The parts from the same
sample of ladel is also evaluated using Sonic Analysis System.
It is evident from the above comparison that Sonic Analysis
System is the preferred method for many tasks including
differentiating ductile and grey cast components.
The results that obtained are tabulated over averages
obtained from several tests.
VII. CONCLUSION
Analyses of results obtained from the components' test
suggests that Sonic Analysis System demonstrates excellent
and flawless capability to differentiate between Ductile and
Gray cast iron with
Avg.Frequency Avg.Nodularity
2924
91,26
2940
85,51
2947
89,65
2952
90,58
2955
89,68
2962
91,6
2963
91,26
2966
90,48
2969
91,64
2978
91,14
2983
89,26
2995
90,99
3000
91,63
3005
91,27
3009
91,66
3033
92,08
Fig 14:Data of Avg.Frequency Vs Avg.Nodularity
VI.
COMPARISON WITH OTHER METHODS

Good repeatability

Good Accuracy

High speed

Low probability of erroneous detection

Applicable for all components independent of its
shape and size

High data storage capability with good data
retrieving capability

Facility to provide statistical analysis and reports
Further, tests also reveal that Nodularity and Frequency
have a linear relation considering other dependant parameters
such as weight and profile of the component to be constant.
The data obtained from various research tests and field data
shows that Sonic Analysis System serves as an effective means
of NDT analysis for cast iron nodularity measurement.
VIII. REFERENCES
[1] Ingolf Hertlin, Detlev Schultze "Acoustic Resonance
Testing: the upcoming volume-oriented NDT method"
[2] Juran, Joseph M. and Godfrey, A. Blanton, Fifth
Edition, "Juran’s Quality Handbook", McGraw-Hill.
[3] Neil Baker, "Techniques for Optimisation and Analysis
of Composite Structures for Damage Tolerance and Buckling
Stiffness"
[4] Gail R Stultz, Richard W Bono, Mark I Schiefer "
Fundamentals of Resonant Acoustic Method NDT"
[5] Ewins, D.J. Modal Testing: Theory and Practice.
Fig 15:Comparison of features and capability of Sonic
Analysis System with other methods
Spectral Finite Element Method for Inspection of Adhesively Bonded
Metallic Joints Using Guided waves
Shweta Paunikar, S. Gopalakrishnan
Department of Aerospace Engineering,
Indian Institute of Science, Bangalore, India
Contact: shweta.paunikar@aero.iisc.ernet.in
With the advent of damage tolerance philosophy for designing an aircraft, the need of adhesive joints
over mechanical joints is increased manifolds. In this work, Spectral Finite Element Method (SFEM)
is used to obtain the wave responses in adhesively bonded metallic joints of varying health. SFEM is
tailor made technique formulated in frequency domain to study wave propagation problems
exclusively. Ultrasonic guided waves are used for inspection here. The entire metallic lap joint is
represented using three elements only. Two spectrally formulated Timoshenko beam elements are
used to model the metallic adherands and spectrally formulated double beam element models the
overlapping joint portion. Two spectral Timoshenko beam elements coupled using continuously
distributed elastic spring foundation are used to develop the double beam element. Spectrum relation
obtained in the frequency domain yields the wavenumbers corresponding to incident and reflected
components of axial, bending and shear modes propagating in the elements. It is evident from the
spectrum relation that the shear mode starts propagating only after the cut-off frequency. For the
double beam element there are six forward and backward propagating modes each and additional cutoff frequencies depending on the value of foundation stiffness (called the foundation cut-off
frequency). On increasing the value of foundation stiffness, the foundation cut-off frequency
approaches the shear cut-off frequency and at a particular value of foundation stiffness, the two cutoff frequencies have same value. The adhesive bond quality is studied by varying the value of
foundation stiffness.
NDE for Aircraft Engines: Challenges, Opportunities and Approaches
Dr. Shyamsunder Mandayam
Manufacturing & Materials Technologies,
GE Global Research, Bangalore, INDIA
Contact: Mt.shyamsunder@ge.com
An aircraft engine is one of the most complexes of systems with several components being integrated
to create a safety critical system which must perform with absolute integrity in extreme harsh
operating conditions. The complexity arises from the variety of materials being used varying from
alloys to composites to ceramics, complex shapes and designs which sometimes pose a major
challenge to apply the NDE technique. The nature of damages which can occur to both static and
rotating parts vary considerably and needs a good physics understanding of energy-material
interactions to develop the appropriate inspection technique. The components or structure is designed
for a specific functionality over its design life with due consideration of the influencing factors such
as the operating conditions and environment, safety and of course its core objective. In addition, the
knowledge related to the health of a component is a vital parameter for philosophies of life extension,
remaining life assessment and asset management. There is a continual degradation in material’s
structure and properties and it is crucial to be able to predict and/or measure this degradation, and
assess the associated reduction in useful life.
A few of the important manifestations include stresses, defects and microstructure (changes)/ damage.
Currently, the integrity, availability and safety are achieved by means of physical inspections using
conventional and advanced Non-destructive Evaluation (NDE) techniques, and more recently, through
condition and structural health monitoring that are useful in detecting and preventing catastrophic
failures. The concept of gathering information about the current state of the material awareness to
determine the remaining life of a component is gaining significance importance. Characterizing the
current state of material damage before the onset of macro damage (cracks and other damage
characteristics detectable and measurable by state-of-the-art NDE), sensing of a material state at the
microstructure level, precursor damage at the dislocation level, and fatigue-crack size population.
Thus detecting, measuring and characterizing the root cause of material deterioration (defects,
microstructure and stresses) in aircraft engine components becomes very important and necessary and
that too in an entirely non-destructive manner.
The paper will provide an overview of the challenges, opportunities and approaches which exist in the
NDE of aircraft engines at various stages starting from the raw material, machining and
manufacturing components, and assembled engines and in-service. The overview will comprise of
technologies in use and those which are actively being pursued, researched and developed to mature
them for actual applications. Modalities of interest vary from Ultrasound, Electromagnetics, infrared,
X-ray, Optical and many others. Work reported here comprise of both those happening at GE as well
as those being utilized and studied by other agencies.
Non Destructive Evaluation methodologies to predict the strain hardening effect
in Landing Gear components: A review
Arun Dinesh, Chanakesava Reddy, Bharath Marappan, Seshadri Venkatadri
UTC Aerospace systems, Global Engineering Centre, Bangalore - India
Landing Gear systems are large and complex in shape having multiple load bearing materials.
Residual stresses are an important factor to determine the functional performance and integrity of
a part or an assembly. Landing Gear experiences strain hardening over a period of time and also
experiences high strains during hard landings; at such occurrences, it is a challenge to predict the
life of Landing Gear when the magnitude of strain induced on the components are not known
during service. It is the need of an hour to evaluate the existing methods and best
practices/techniques, to predict the magnitude and the extent of the stress experienced by a
component without removal / disassembly from assembly. This paper highlights/discusses the
current trends, and research needs in structural health monitoring (SHM) using NDE techniques.
Keywords: Landing Gear, Strain hardening, Non Destructive Evaluation (NDE)
1. Introduction:
Landing system is an integration of the landing gear and the wheels and brakes of an aircraft. It is
one of the most critical sub-systems which carry the entire load of an aircraft during taxiing on the
runway while take-off and landing. The need to develop a landing gear with minimum weight,
Minimum volume, and reduced life cycle cost poses multiple challenges to the designers and
practitioners. These challenges can be addressed by introducing advanced technologies, next
generation materials, improving the processes and production methods and analysis tools to
extrapolate the effects the aircraft components undergoes to be represented virtually.
Aftermarket service is an important stage in any product industry. A balance between the life of
the component and the service time frame of the component should be established to extract the
maximum life out of it, since the servicing or replacing a component is a cost driven factor.
Currently 27% of an average aircraft’s life cycle cost, both for commercial and military vehicles, is
spent on inspection and repair; a figure that excludes the opportunity cost associated with the time
the aircraft is grounded for scheduled maintenance [1] While most aircraft structures are made
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
from ductile alloys that can endure crack growth over time; landing gear use very high strength,
but brittle, alloys of steel, aluminum or titanium. These differences and other unique issues require
distinct approaches with landing gear structural health monitoring methods [2] Landing Gear
Actuator Health Monitoring Application
Landing gears are an essential part of any aircraft; even though they remain redundant for most of
the flight. The main task of the landing gear is to absorb the horizontal and vertical energy of the
aircraft as it touches down on the runway. During flight most aircraft have their landing gears
retracted and stowed, only extending them during the approach to landing.
Extension and
retraction of aircraft landing gears is commonly achieved using either hydraulic or pneumatic
drives. A standard landing gear contains three such actuator drives; the largest of which is the
retraction actuator that generates a force about a pivotal axis in order to raise the gears against
weight and aerodynamic loading. The other two actuators are the lock-stay actuator, which locks
the landing gear in place once extended and the door actuator that ensure that the landing gear
bay doors are successfully opened and closed for landing gear deployment. Figure (1) shows a
standard landing gear arrangement undergoing a retraction cycle.
Figure 1 Retraction cycle for a standard landing gear [37]
Hydraulic actuation systems have found popular use in aerospace as a whole due to their
reliability, relative simplicity and their wide spread use has generated engineering experience and
familiarity. They are ideally suited for a landing gear application as the hydraulic fluid provides a
constant lubrication and natural damping.
There are however disadvantages; when used in
aircraft they are heavy, require large volumes of space; operate noisily and require the correct
disposal of hydraulic fluid in accordance with environmental legislation. There is currently a move
within the aerospace industry to move away from these hydraulic drives and to replace them with
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
lighter, smaller, cleaner and more efficient electrically powered actuation as part of what is now
commonly termed the More Electric Aircraft [3].
Electric motor driven actuation is now very widespread within the transport industry. In automotive
products, for example, electric windows, locks, aerials, seat/lamp/mirror adjustment are common.
Drive-by-wire introduces motor-actuated steering and the starting circuit is a heavy motor-driven
actuation system. Similar situations are encountered in railway point motor mechanisms, heavy
electrical switch gear, and valve actuation. The motivation for the use of Electro-Mechanical
Actuators (EMA) is driven by the desire to reduce aircraft weight arising from a combination of
increasing fuel costs and environmental concerns.
For example, environmental damage
associated with air traffic has created the need to reduce aircraft fuel consumption and polluting
emissions, a key factor in achieving this is the reduction of weight. Another key motivation for
utilising electrically powered actuators is that there is a real possibility that with the move towards
optimisation of engine efficiency, future aircraft engines will not be designed to produce hydraulic
power. Another benefit of replacing landing gear hydraulics is to reduce aircraft turn-around
times. Aircraft currently have the requirement that the brakes must be allowed to cool before the
wheels can be stowed due to the fire risk associated with hot brakes coming into contact with
leaked flammable hydraulic fluid. EMA would mean that these toxic hydraulic fluids which require
significant maintenance efforts to maintain are no longer needed.
A key challenge associated with the introduction of EMA for the current range of civil aircraft in
operation are currently design issues associated with space constraints (it has proven difficult to fit
electrical motors into the available space designed for a hydraulic actuator). Aerospace requires
high reliability, and these new drives must be proved robust and as reliable as the current drives
in use. The aerospace industry has a long history of using hydraulic actuator drives, and customer
confidence must also be gained in replacement technologies.
The move towards electrical
actuation in aircraft therefore will utilise the support of health monitoring to guarantee reliability
and increase confidence [4].
Development of automated health monitoring systems are also being encouraged for aerospace
systems as a way of reducing unexpected failures and reducing over engineering of systems
removing to some extent the amount of redundant (backup) mechanical elements and finally to
aid in maintenance optimisation. EMA are complex mechanical systems and inspection of
individual subsystems and components is not as straight forward as is the case with hydraulics
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
and simple on-wing maintenance becomes more problematic. Most of the key components (i.e.
gears, bearings, wiring) are often sealed within housing units making access to them difficult. The
use of traditional inspections for damage would require a larger degree of landing gear
dismantling. There is the possibility therefore that the use of EMA could even increase the time
an aircraft spends in the maintenance hangar, increasing aircraft down time costs. The use of
automated systems to identify and/or predict component damage in the form of health monitoring
systems would offer a reduced risk of this being the case.
2. Methods:
NDE techniques with the highest likelihood of success were thoroughly examined, including
frequency response, Lamb wave, acoustic emission and strain monitoring methods. For each of
these methods, an analytical and experimental procedure was followed to optimize the testing
parameters and data interpretation. Their strength, limitations and SHM implementation potential
were evaluated, and suggested roles for each are presented.
2.1
Data based approach:
An organising principle for a data-based approach to SHM can be formulated in terms of a fourstage process [5]
(i)
Operational evaluation.
(ii)
Data acquisition, normalisation and cleansing.
(iii)
Feature selection and information condensation.
(iv)
Statistical model development for feature discrimination.
Past
History
Performance
Data
Maintenance
Database
CBM System
Aircraft
Health
Status
Management
Requirements
Management
Database
Maintenance
Schedule
Maintenance
Scheduler
Maintenance
Action
Figure 2 Predictive maintenance concept. [37]
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
2.1.1 Data Fusion and HUMS (Health and Usage Monitoring System) Integration [6]
This landing gear SHM system also allows data fusion of direct loads monitoring data into fatigue
life assessments. This feature is provided via utilization of communication to platform HUMS and
associated flight records for data assurance purposes. The interface and communication of the
SHM control units to the aircraft HUMS equipment provides the ability to synchronize loads data,
allowing for elimination/reduction of estimates on landing gear loads usage and service life. An
example of the benefits of direct loads monitoring is shown in Figure 3.
Figure 3 NLG strut bottom out during landing. [6]
In this example, the aircraft systems or aircrew did not detect a hard landing. However, the nose
landing gear (NLG) became fully compressed for several cycles during the landing. An
integrated SHM system could have alerted the aircrew and/or ground crew that the NLG should
be inspected, thereby enhancing operational safety. The other scenario is similar. The aircrew
could experience what they believe was a hard landing. With an integrated SHM system directly
measuring landing gear loads, the system could confirm the hard landing or indicate the
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
measured loads were within safe operating parameters. Thus, the SHM system can reduce
maintenance costs and increase aircraft availability by reducing unnecessary maintenance.
New military fighter-craft such as the Eurofighter, the Joint Strike Fighter and the F-22 all
incorporate Health Usage Monitoring Systems (HUMS), which record peak stress, strain and
acceleration experienced in key components of the vehicle [7]. While these measurements
provide useful information about the state of the vehicle between flights, the value of such a
system could be greatly increased if continuous data could be accessed instantaneously.
2.2
Algorithm Based Approach: [23]
In a collection of papers written by Zimmerman, he suggests that an algorithmic approach could
be used to enhance the model correlation and health monitoring capabilities using frequency
response methods [8].
Other researchers have developed algorithms to attempt to correlate modal response under
arbitrary excitation to models using a probabilistic sub-space based approach [9].
Giurgiutiu used Lamb wave techniques to compare changes in thin aluminum aircraft skins after
various levels of usage to detect changes, and used finite element techniques to attempt to
predict the level of damage with some success [10].
More detailed work was performed by Cawley’s group at Imperial College, who used Lamb waves
to experimentally examine representative metallic aircraft components such as lap joints, painted
sections and tapered thickness [11].
2.2.1 Frequency response methods: [23]
During the present research, several damage detection methods were tested that showed
encouraging implementation potential for an SHM system. A set of narrow rectangular quasiisotropic [90/±45/0] s laminates were manufactured of the AS4/3501-6 graphite/epoxy system
with various forms of damage introduced to them, including matrix-cracks, delaminations and
through-holes. These specimens were then reused for each test method by using PZT
piezoceramic patches as sensors, which were affixed using 3M ThermoBond thermoplastic
tape. The first methods surveyed were the frequency response methods. Detailed results for
these experiments have been presented in previous papers [12-14]. Experimentally, an
impedance meter was used to measure the natural frequencies, and the mode shapes were
deduced used a scanning laser vibrometer. A finite element analysis was also performed to
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
predict the frequency response of each specimen up to 20 KHz. From both sets of results it is
evident that all the forms of damage investigated in this study caused detectable changes in the
natural frequencies of a simple coupon. These changes are present in each of the lower normal
frequencies discovered, and become more pronounced at higher frequencies, however
coalescing modes made comparison impractical. A representative plot comparing a control and
damaged specimen can be seen in Figure 4. A strong correlation existed between relative
frequency reduction and the area damaged by a particular mechanism, however it is difficult to
draw any conclusions about the criticality of the damage since there is no information regarding
the form of the damage or its orientation. Based on these results, it is likely that an observer can
discern whether a structure has been damaged by observing its frequency response, however it
would be difficult to differentiate reliably between damage types, locations and orientations. This
method appears to be appropriate for detecting global changes in stiffness for relatively large
structures at a low power and weight cost
Figure 4 Frequency response transfer plot function plot from I-DEAS, range of 0-500 Hz. [23]
2.2.2 Lamb wave methods: [23]
Next the utility of using Lamb waves for damage detection was explored. Again, detailed results
for this Lamb wave research has been presented in previous papers [14-18]. The experimental
procedures followed a building block approach, and the first set of experiments conducted on
narrow composite coupons presented in the previous section
[19]. Both the actuation and the data acquisition were performed using a portable NI-DaqpadTM
6070E data acquisition board, and a laptop running LabviewTM as a virtual controller, and the
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
results were compared by performing a Morlet wavelet decomposition centered at the driving
frequency [20]. This procedure was also carried out for beam specimens, laminated plates with
bonded stiffeners, and a sandwich construction cylinder. Finite element models were produced
to simulate the small changes in time of flight caused by damage for each of these tests as well.
The results from the narrow coupon tests clearly show the presence of damage in all of the
specimens; this was made most obvious by comparing the wavelet decomposition plots. The
control specimens retained over twice as much energy at the peak frequency as compared to all
of the damaged specimens, as demonstrated in Figure 5. The loss of energy in the damaged
specimens was due to reflection energy and dispersion. Similar effects of damage were
observed in each of the built-up composite structure cases. Similar to frequency response
methods, their results are limited at higher frequencies, however their low frequency results
should provide sufficient data to predict damage. The disadvantage of Lamb wave methods is
that they require an active driving mechanism, and the resulting data can be more complicated
to interpret. Overall, Lamb wave techniques have the potential to provide more information than
other methods since they are sensitive to the local effects of damage in composite materials,
and have proven effective for the in-situ determination of the presence and severity of damage.
Figure 5 Wavelet coefficients for beam "blind test"; compares 50 kHz energy content for control beam specimen
and 2 specimens with delamination [23]
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
2.2.3 Multiple sensor approach: [30]
The SHM system is based on a multiple sensor approach:
(i)
Ultrasonic waves activated and sensed by piezoelectric patches bonded to the most
critical structural component.
(ii) MEMS accelerometers, measuring in differential system parameters evolution at crucial
locations of the landing gear.
(iii) Optical fibers (FBG) bonded to the most stressed locations of the landing gear.
2.2.3.1 Sensors: [23]
Sensors are used to record variables such as strain, acceleration, sound waves, electrical or
magnetic impedance, pressure or temperature. In the literature it has been estimated that a
SHM system for an aerospace vehicle would require between 100 and 1000 sensors,
depending on its size and desired coverage area [21]. Sensing systems can generally be
divided into two classes: passive or active sampling. Passive sampling systems are those that
operate by detecting responses due to perturbations of ambient conditions without any
artificially introduced energy. The simplest forms of a passive system are witness materials,
which use sensors that intrinsically record a single value of maximum or threshold stress, strain
or displacement. Examples of this can be phase change alloys that become magnetized beyond
a certain stress level, shape memory alloys, pressure sensitive polymers, or extensometers.
Another type of passive sensing is strain measurement by piezoelectric wafers. Lastly, several
vibrational techniques can be performed passively, such as some accelerometers, ambient
frequency response and acoustic emission with piezoelectric wafers. Active sampling systems
are those that require externally supplied energy in the form of a stress or electromagnetic wave
to properly function. A few strain-based examples of active systems include electrical and
magnetic impedance measurements, eddy currents and optical fibers which require a laser light
source. Active vibrational techniques include the transfer function modal analysis and Lamb
wave propagation. Good references for selection of actuators for various active systems can be
found in a review paper in the literature [22]. Passive techniques tend to be simpler to
implement and operate within a SHM system and provide useful global damage detection
capabilities, however generally active methods are more accurate in providing localized
information about a damaged area. A comparison of the sensing methods can be seen in Table
1 [23].
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Table 1 comparison of strengths, limitations and SHM implementation potential for various sensing systems [23]
Sensor selection charts plotting size of detectable damage against sensor size and power
requirement for various coverage areas, can be found in Figure 6 and Figure 7. It can be seen
that they are all generally capable of detecting the same size of damage and can be
implemented with similar size and power sensors, however frequency response and Lamb wave
techniques are the only ones that can offer full surface coverage for a 1 x 1 m plate. While other
methods, such as eddy currents, can offer better resolution, they are only capable of detecting
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
damage directly below the sensor, which would drive the system to use either very large
sensors or a large volume of sensors.
Figure 6 Sensor selection space comparing size of detectable damage with sensor size for various
sensing methods [23]
Figure 7 Sensor selection space comparing size of detectable damage with sensor size power for
various sensing methods [23]
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
The goal of the Health Monitoring and Management (HMM) system is to develop advanced
technologies for the health management of structural component to enable effective detection,
diagnosis, prognosis, and mitigation of damage into structures.
At this level it is fundamental to select the right sensors which are able to carry out the wanted
information. The present HMM system consists of different sensors:
(i) MEMS accelerometers;
(ii) Piezo-electric patches;
(iii) Optical fibers (FBG).
Microelectromechanical systems (MEMS) refer to devices having dimensions in the range
from a micron to a millimeter that combine electrical and mechanical components and that are
fabricated using the same technologies of integrated circuits [24]. The accelerometers are made
by a polysilicon surface with a micro machined structure built on top of a silicon wafer.
Silicon springs suspend the structure over the surface of the wafer and plates attached on the
fixed and moving mass form a differential capacitor. Acceleration deflects the moving mass and
unbalances the differential capacitor resulting in a sensor output with amplitude proportional to
acceleration. Phase-sensitive demodulation techniques determine the magnitude and direction
of the acceleration.
A piezoelectric material is a material able to generate an electric field when it is subjected to a
mechanical strain (direct effect), or, conversely, to generate a mechanical strain in response to
an applied electric field [25-27]. This means that piezoelectric material devices can be used as
both source and receiver. Piezoelectric sensors have been used to detect small dimension
damages by means of ultrasonic waves in the frequency range 30 kHz – 350 kHz; MEMS,
indeed, have been adopted in the low frequency range say 20Hz – 5 kHz.
A Fiber Bragg grating (FBG) is a periodic modulation of the refractive index in the core of a
little length of an optical fiber (FO). We all know that FBG works like a narrowband reflecting
mirror. The Bragg reflection wavelength (λB) of an FBG is directly dependent by effective
refractive index of the grating and the imposed grating period (Λ). So, the FBG sensors are
sensitive to both temperature and strain [28]. The strain response arises mainly from the
physical elongation of the length of the sensor. Being spectrally encoded, the FBGs are
insensible to EM interference, intensity modulation of the optical carrier and losses along all the
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
length of connection between the sensor location and the opto-electronic readout apparatus.
This allows to have high multiplexing capability on a single FO connection (more than 60 for a
and high time bandwidth response (up to MHz), confer to such kind of sensing system to be one
of the most eligible monitoring system for the SHM field [29].
The MEMS accelerometers and the FBG sensors are suitable for measurements in the low
frequency range, where a global behaviour of the structure is measured and hence damage of
certain dimensions should be identified. To detect smaller damages ultrasonic waves have been
used, at the appropriate frequency to have a wavelength comparable with dimension of the
damage. For this purpose piezoelectric sources and receivers have been adopted using the
pitch-catch technique on a component of a landing gear. Known the generated and the
measured signals it is possible to calculate the group velocity, the amplitude of the wave packet,
the transmission factor and by comparing these parameters before and after an event occurs it
is possible to report if a damage occurred or not.
2.2.3.2 Costs and benefits [30]:
The implementation of the HMM system will provide benefits especially for the maintenance,
which actually is scheduled. By using a HMM system, in fact, it can be possible to change this
approach and try to intervene in constructive and proactive way in the regular program of the
maintenance when the HMM system inform the maintainer. This approach will provide
substantial savings in terms of time and costs.
In order to have an idea in this section an approximate estimation of the costs associated with
inspection with and without HHM system is made.
Actually a programmed inspection of a landing gear, without HMM system, after 100000 hours
of flight is pairs to 1,67 M€. In the case of a landing gear equipped with a HMM system the
same inspection (always referred to 100000 hours of flight) will have a cost pairs to 435 k€, i.e.
four times less of the actual programmed inspection. This cost consists of [30]:
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Thus for a life cycle, estimated at 100000 hours of flight, an approximately 28% reduction of
maintenance costs will be provided [30].
In terms of weight the HMM system will increase the weight of a single landing gear to 13,5 kg
(consisting to sensors=0.5 kg, electronics= 8 kg, cables and other components= 5 kg), and
hence pairs to 40,5 kg for three landing gears [30].
3.
Value potential of Predictive Maintenance: [37]
The value of incorporating health monitoring systems is most likely to arise in savings in overall
operating costs. The use of health monitoring systems for landing gear retraction mechanisms,
or other aircraft systems, will offer a very competitive advantage in maintenance decisionmaking, which is crucial for both military and commercial aerospace users.
This will help
manufacturers retain customers and attract new business; these aspects will mean that health
monitoring solutions will become a key part of formulating future maintenance strategies. The
airline industry has seen a rapid increase in operators over the past decade, particularly in low
cost short haul operators. The nature of the budget airlines business succeeds in the ability to
operate large aircraft fleets, coupled with high aircraft availability and short turn-around times
whilst keeping ticket costs low. For such factors to remain and for airlines to create a business
winning advantage, then strategic maintenance management has to become one of the
significant factors in their operations management.
The adoption of health monitoring and
overall predictive maintenance can help push an aircraft operators business forward as
illustrated in Figure 8.
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Redefine
expectations
Increasing contribution
of predictive
maintenance
Be clearly the
best in the
industry
Link maintenance with
operations strategy
Be as good as
competitors
Adopt best practice
Organisation
held back
STAGE 1
The ability
to implement
strategy
Give an Operations and
business winning
Advantage
Correct the worst problems
STAGE 2
STAGE 3
The ability
to support
strategy
STAGE 4
The ability
to drive
strategy
Figure 8 Potential effects of predictive maintenance on an aircraft operators business [37]
4.
Technical Challenges to Integrating Health Monitoring: [37]
Large-scale integration of health monitoring will cause disruptive changes within well-defined
and established maintenance related practices, such as logistics, parts management, and
manufacturing and information management. If the integration of health monitoring technology
is efficiently managed with the necessary support and infrastructure requirements in place then
the technology will become firmly established, becoming a fully performance/business
competitive innovation. Health monitoring systems are aimed at improving the performance of
the aircraft, which will be achieved on the lines of ‘evolutionary’ changes whilst demonstrating
reliability, validated cost benefits [31] and reduces operational risks. It is also necessary to
ensure that appropriate parameters are selected for monitoring [32] and that there is robustness
in the modelling of the health management process [33].The integration of new technologies
inevitably face difficulties and a number of challenges face the community of engineers and
technical specialists as they seek to utilise health monitoring for aerospace usage, a nonexhaustive list of these difficulties include:
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
(i) The technology and frameworks are available but underutilized.
(ii) Performance characteristics are usually untested, leading to a lack of confidence
(iii) There is often a wealth of data available from the end users, but access to this data can be
limited and much is yet to be converted to ‘meaningful information’
Health monitoring systems for aerospace applications differ from those for other applications
such as industrial machine monitoring or the monitoring of civil structures due to hardware
restrictions and the difficulties associated with certification. Also, in many areas of aerospace
health monitoring system development, often the state-of-the-art monitoring technique being
developed are restricted by a variety of limitations. This affects their use in a real operational
situation’, for example, many of the sensor based methods under development for the
monitoring of fuselage structures, based upon such methods as acoustics or vibration patterns
require vast sensor arrays. Much of the information gained requires high levels of signal
processing with the results being very subjective and consequently they may not be applicable
for an on-line real time aerospace monitoring system, even though the fundamentals of the
techniques work well in other applications. This will potentially lead to a case where the state of
the art has difficulties in matching the necessary requirements for aerospace integration. This is
possibly the reason for the current slow integration of health monitoring on civil aircraft, despite
the vast wealth of academic research detailing monitoring methods, industry drive and potential
areas for application [37].
Figure 9 illustrates this hypothesis it demonstrates how the current health monitoring state-ofthe-art trend is progressing with respect to the capability requirements for health monitoring for
aerospace usage [34]. The hypothesis indicates that the current state-of-the-art is advanced
enough for most industry uses; offering leaps in performance and capabilities. But is far below
what is required for aerospace applications, and will require further innovations, amongst others,
in terms of hardware minimisation, data reduction techniques and the use of fusion to merge
multiple techniques to reduce individual limitations and maximise advantages [37].
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
HM system requirements for
aerospace applications
Current HM state of the art
trend
Capability
HM system requirements for an
‘enabling’ technology
Time
Figure 9 Aerospace health monitoring requirements as compared to the state-of-the-art [37]
4.1
New Sensor Technology and Systems Integration: [37]
This research has put a strong emphasis on the need to keep a health monitoring systems affects
on the actuators weight and complexity to a minimum. This instantly begins to restrict the health
monitoring approach away from more popular sensor intensive techniques.
Future research
should focus on monitoring solutions using wireless smart sensors with inbuilt signal processing
that can perform all monitoring tasks on board the sensor itself [35]. The benefits of these are that
they are lightweight; do not require additional cabling and they can create a reduction in the
demand for aircraft computing resources [36]. These advantages will open new doors for the
investigation of monitoring techniques which have been deemed inapplicable with conventional
sensor technology.
Systems integration research would need to be undertaken to ensure that any future monitoring
system can not only integrate into the landing gear subsystems, but can also integrate seamlessly
into the aircraft systems as a whole and if necessary work alongside aircraft BIT and any other
local monitoring systems. This is essential if the concept of an aircraft IVHM is ever to be realised.
5.
Conclusion
Health monitoring technology is intractably tied up with aerospace maintenance activities as a
whole. The aerospace maintenance industry is currently facing a time of unprecedented
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
demand for spare parts, complete overhauls and general servicing. This is due to, amongst
other reasons, a sudden increase in aircraft numbers in the last decade or so which now have
key systems such as landing gears reaching the end of their life. This is therefore putting a
strain on overhaul providers and manufacturers. This has begun to force these key players to
begin seeking new innovative maintenance solutions, to meet rising demands and costs [37].
References
1. Hall S.R. and T.J. Conquest. “The Total Data Integrity Initiative—Structural Health Monitoring, The
Next Generation.” Proceedings of the USAF ASIP, 1999.
2. R. Schmidt, P. Sartor, Landing Gear. Encyclopedia of Structural Health Monitoring, 2009)
3. Rosero et al., 2007
4. Phillips et al. 2008
5. Farrar and Worden 2007
6. Landing Gear Structural Health Prognostic/Diagnostic System Chad FORREST 1, Robert HUOT 1,
Arianne MOLINA.
7. Neumair M, “Requirements on Future Structural Health Monitoring Systems.” Proceedings of the
7th RTO
Meetings, May 1998
8. Zimmerman D.C., Simmermacher T. and M. Kaouk. “Model Correlation and System Health
Monitoring using Frequency Domain Measurements.” AIAA Journal, 3318-3325, 1995
9. Abdelghani M., Goursat M. and T. Biolchini. “On-Line Modal Monitoring of Aircraft Structures under
Unknown Exication.” Mechanical Systems and Signal Processing, v.13, 839-853, 1999.
10. Giurgiutiu V., Bao J. and W. Zhao. “Active Sensor Wave Propagation Health Monitoring of Beam
and Plate Structures.” Proceedings of the 8th International SPIE Symposium on Smart Structures
and Materials, Newport Beach, CA, 2001.
11. Dalton R.P., Cawley P. and M.J.S. Lowe. “The Potential of Guided Waves for Monitoring Large
Areas of Metallic Aircraft Fuselage Structure.” Journal of Nondestructive Evaluation, v.20, 29-46,
2001.
12. Kessler S.S., Spearing S.M., Atalla M.J., Cesnik C.E.S. and C. Soutis. “Damage Detection in
Composite Materials using Frequency Response Methods.” Proceedings of the SPIE’s 8th
International Symposium on Smart Structures and Materials, 4-8 March 2001, Newport Beach, CA,
NDE 4336-01.
13. Kessler S.S., Spearing S.M., Atalla M.J., Cesnik, C.E.S. and C. Soutis. “Structural Health
Monitoring in Composite.Materials using Frequency Response Methods.” Accepted for publication
by Composites Part B, June 2001.
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
14. Kessler S.S. “Piezoelectric-Based In-Situ Damage Detection of Composite Materials for Structural
Health Monitoring Systems.” Massachusetts Institute of Technology, Ph.D. thesis, January 2002.
15. Kessler S.S., Spearing, S.M. and C. Soutis. “Damage Detection in Composite Materials using Lamb
Wave Methods.” Proceedings of the American Society for Composites, 9-12 September 2001,
Blacksburg, VA.
16. Kessler S.S., Spearing S.M. and C. Soutis. “Optimization of Lamb Wave Methods for Damage
Detection in Composite Materials.” Proceedings of the 3rd International Workshop on Structural
Health Monitoring, 12-14 September 2001, Stanford University.
17. Kessler S.S., Spearing S.M. and C. Soutis. “Structural Health Monitoring in Composite Materials
using Lamb Wave Methods.” Submitted for publication to Smart Materials and Structures, July
2001.
18. Kessler S.S., and S.M. Spearing. “Damage Detection in Built-Up Composite Structures using Lamb
Wave Methods.” Submitted for publication to Journal of Intelligent Materials Systems and
Structures, December 2001.
19. “The Composite Materials Handbook MIL-17 Vol. 1” Guidelines for Characterization of Structural
Materials.” MIL-HDBK-1E, Department of Defense, 1999.
20. Strang G. and T. Nguyen Wavelets and Filter Banks. Wellesley-Cambridge Press, Wellesley, Ma,
1996.
21. Marantidis C., Van Way C.B. and J.N. Kudva. “Acoustic-Emission Sensing in an On-Board Smart
Structural Health Monitoring System for Military Aircraft.” Proceedings of the SPIE Conference on
Smart Structures and Integrated Systems, v. 2191, 258-264, 1994.
22. Huber J.E., Fleck N.A. and M.F. Ashby. “The Selection of Mechanical Actuators based on
Performance Indices.” Proceedings of the Royal Society of London, 2185-2205, 1997.
23. Design of a piezoelectric-based structural health monitoring system for damage detection in
composite materials Seth S. Kessler and S. Mark Spearing.
24. M. Gad-el-Hak, The MEMS Handbook, Boca Raton, Fla.: CRC Press, 2006.
25. Maio, L., Memmolo, V., Ricci, F., Boffa, N.D., Monaco, E., Pecora, R., Ultrasonic wave propagation
in composite laminates by numerical simulation (2015) Composite Structures, 121, pp. 64-74.
26. V. Giurgiutiu, Structural health monitoring with piezoelectric active wafer sensors, 2007.
27. J.L. Rose, Ultrasonic waves in solid media. Cambridge University Press, Cambridge, 2004.
28. A.D. Kersey, M.A. Davis, H.J. Patrick, M. LeBlanc, K.P. Koo, C.G. Askins, M.A. Putnam, and E.J.
Friebele, Fiber grating sensors, J. Lightwave Technol., 15, pp. 1442 –1463, 1997.
29. G. Breglio, A. Irace, A. Cusano, A. Cutolo, Optical Fiber Technology, 12-1, pp. 71-86 2006.
30. An Innovative health monitoring system for aircraft landing gears - Alfonso Nocella Astronomical
Observatory of Napoli, Napol, G. Petrone, M. Bruno, F. Bocchetto, G. Breglio, M. Pugliese, A.
Caldara, A. Cavallari, S. Schiano lo Moriello, G. Capuano, D. Rossetti.
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
31. Banks, J., Merenich, J., 2005. Cost benefit analyses for asset health management technology’,
Proceedings of the Reliability and Maintainability Symposium, Orlando FL, pp. 1-10
32. Kumar et al. 2010
33. Kumar and Pecht, 2010
34. Phillips et al 2009
35. Pietruszkiewicz et al. 2006
36. Starr et al. 2007
37. Perspectives on the Commercial development of landing gear health monitoring systems by Paul
Phillips, Dominic Diston & Andrew starr
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Main Gearbox Testing for Light Utility Helicopter of HAL
MURALIDHAR B.S. 1, Bhawani Singh RATHORE 1
1
Ground Testing Design Group, Rotary Wing Research and Design Centre,
Hindustan Aeronautics Limited (HAL), Bengaluru, India
Phone: +91 9242112123, E-mail: muralidhar.bs@hal-india.com
Abstract
Non-destructive testing forms a very important part of development of systems during the design and
development phase of a helicopter. Helicopter gearboxes are considered critical systems, as they transmit power
from the engines to the rotors and accessories of the helicopter at the required speed and direction. The
individual components of the gearboxes like gears, shafts, housings are subjected to various NDT techniques in
HAL such as Magnetic particle inspection, X-ray and liquid penetrate checks. Final checks on the components
include checking of all dimensions, surface finish, checking of special features such as splines, hardness checks,
ensuring compliance to the specified manufacturing processes and finishing, prior to the subsequent release of
acceptance tags for the accepted components.
While the above activities take care of the checks at the component level, checks at the assembly level are
essential in order to ensure that the gearbox, as an assembly, meets the design requirements with regard to its
performance under loads at the rated speed. The design and development of special test stands plays a crucial
role in the assembly level testing of gearboxes. The tests are carried out at the development stage, wherein loads
at the rated speeds are applied to check the proper functioning of the gearboxes, as well as in the prototype
acceptance stage, wherein each gearbox identified for ground runs and prototype flight testing undergoes Nondestructive testing on the gearbox test stands. This paper presents the details of Main Gear Box tests performed
for HAL’s Light Utility Helicopter on special test stands developed and commissioned at Ground Test Centre,
RWR&DC, HAL.
Keywords: Main Gear Box, Closed Loop Torques, Mast loads, Test stands
1. Introduction
Power transmission from engine to rotors and accessories in helicopters is carried out by the
Main Gearbox (MGB) via interconnecting shafts and gears. The MGB typically consists of
the gearbox housing, gears and pinions, support bearings, freewheels, lubrication and cooling
systems, accessory drive tap offs, monitoring sensors, and some of the components of rotor
control mechanisms that are integral with gearboxes, such as swash plate mechanisms and
control linkages [1].
MGB : Main Gearbox
MR
: Main Rotor
RB
: Rotor Brake
TR
: Tail Rotor
TGB : Tail Gearbox
Figure 1 : General schematic arrangement of gearboxes in a helicopter
The MGB in a helicopter is designed based on :
- Engine Installation Features & Adequate Engine-to-Tail Drive Clearance
- Required power capability
- Suspension on Vibration Isolation System (VIS)
- Drive for Accessories
- Mounting of Main Rotor Actuators for Main rotor controls
The individual components of the gearboxes like gears, shafts, housings are subjected to
various NDT techniques such as Magnetic particle inspection, X-ray and liquid penetrant
checks. Final checks on the components include checking of all dimensions, surface finish,
checking of special features such as splines, hardness checks, ensuring compliance to the
specified manufacturing processes and finishing, prior to the subsequent release of acceptance
tags for the accepted components. While these activities take care of the checks at the
component level, checks at the assembly level are essential in order to ensure that the
gearbox, as an assembly, meets the design requirements with regard to its performance under
loads at the rated speed. Towards this, special tests were carried out on the MGB as follows.
2. Tests carried out on Main Gearbox of Light Utility Helicopter (LUH)
The LUH MGB has a single engine input, with outputs for main and tail rotor and accessories.
Development and prototype Main Gear Boxes were designed and built by HAL for its LUH
program, and tests carried were out as follows, for checking the conformance to design [3].
2.1 MGB Oil distribution checks
Oil distribution checks are necessary for checking the adequacy of the oil flow rates at the
various gear meshes of the MGB. The Oil distribution checks set up comprised of an oil
reservoir and an oil pump to feed the oil at the required flow rate to the MGB. The resulting
oil pressure and flow at MGB inlet, TPTO (Tail Power Take-off) oil pressure, oil quantity
collected from each jet, and the oil temperature were measured during the checks.
Figure 2 : Views of the MGB specimen with cut outs for Oil Distribution Checks
Apart from measurement of oil flow rates, the Oil Distribution Checks were carried out for
determination of direction of oil jets and determination of pressure drop within the gearbox
unit. The tests carried out were :
 Determination of oil flow rate at the outlet of each oil jet
 Determination of effect on oil flow rate and pressure by blocking of any one oil jet hole
PRESSURE
INLET POINT
OIL SUPPLY TO
FILTER INLET
AUXILIARY
TANK
TPTO PRESSURE POINT
MGB SPECIMEN
FLOWMETER
FOR OIL FLOW
RATE
MEASUREMENT
MGB OIL INLET
HOSE
CONNECTION
POINT
HOSE
CONNECTION TO
AUXILIARY
TANK
EXTERNAL OIL
PUMP FOR OIL
DELIVERY
TO
MGB
MAIN TANK
Figure 3 : Test Set-up for Oil Distribution Checks
The Oil Distribution Checks test set-up comprised of a Main tank filled with oil. The oil was
fed to the inlet of MGB (at MGB filter) by using an external oil pump. The MGB was placed
on an Auxiliary tank which was inter-connected with the Main tank. Photographs of test setup are shown in figure 3. The oil was fed to the inlet of MGB, sprayed out from the MGB oil
jets to the MGB internals, and got collected in the Auxiliary tank. This oil flowed back to the
Main oil tank by gravity, thus completing the oil circuit. During the test, Oil Flow rate, MGB
inlet pressure, MGB TPTO pressure, Oil temperature and ambient temperatures were
measured. There were a total of over a dozen oil jets in the MGB. The oil used was JSD OX38 oil as per DERD 2487. For the test, the external oil pump was switched on, and the flow
rate was adjusted in steps as per the minimum flow rating of the MGB oil pump. After
completion of the tests at minimum flow, the oil flow rate was increased to mean flow, and
the directions of spray were observed for each case. The oil temperature was then increased
using the oil heater for the high temperature checks. The pressure drop within the MGB was
determined at different inlet oil flow rates by adjusting the flow and measuring the pressures
at the MGB inlet and TPTO. The difference between the inlet pressure and TPTO pressure
gives the required pressure drop. For the oil flow rate measurement at still higher temperature,
the test was carried out with alternate oil simulating the viscosity of OX-38 oil at 120°C, in
order to avoid hazards of handling hot oil. The MGB inlet oil flow rate was then adjusted to
the maximum flow, and oil from various branches spraying out from the oil jets were
collected one by one to measure the oil flow rate for the respective jet.
2.2 Wipe tests for contact pattern development of the spiral bevel gear mesh
Iterations of contact pattern tests were carried out till the achieving of satisfactory contact
pattern with the required shimming and grinding corrections. The Tooth contact pattern
development test (Wipe Test) was conducted in Quasi Static conditions, to arrive at the
acceptable contact pattern. The objective of the test was to arrive at the acceptable contact
pattern for gear & pinion in terms of size & location, and to establish the required machine
setting data for production of pinion for gear pair. This was a pre-requisite for further
dynamic and endurance load tests. The contact patterns were checked for each step after disassembly of the gearboxes from the rig.
No-load contact pattern for the gear teeth was prepared at the assembly shop at the gearbox
build stage. After this, the MGB was subjected to quasi-static wipe tests on Wipe Test Rig
(see figure 4). The test rig was designed for application of and measurement of torques (at
quasi-static speed by using a motor with speed reduction stage), as well as rotor load
application and measurement for the various steps as per test requirements.
The test rig set up consisted of a Rotor mast loading block, Input and TPTO Bracket
assemblies, Brake Motor with speed reduction gear box, Pulley arrangement and Specimen
mounting fixture (Test cart). A variable speed drive motor was integrated with the speed
reduction gear box connected to a flexible cardan shaft which in turn connected the rotor mast
loading unit of the rig. The specimen MGB was assembled on the trolley with pylon struts and
tie rods. The trolley with MGB specimen was assembled with rotor mast loading block, input
bracket and TPTO bracket of the Wipe test rig.
LOADING PULLEY
WITH WIRE ROPE
MGB SPECIMEN
STRUCTURAL FRAME
Figure 4 : Wipe Test set-up fpr LUH MGB contact pattern checks
The torque loading system consisted of rope & pulley arrangement assembled on the Input
bracket and TPTO brackets of the Wipe test rig. The ropes on Input and TPTO were locked
with wire lock adapters assembled on Input and TPTO brackets. The other end of the wire
ropes of Input and TPTO brackets were hung with weighing pans and calibrated weights were
added to apply the required torque, after which the drive motor was switched on for slow
speed/quasi-static rotation. Torque transducers were connected at flanges of Input, TPTO and
Hydraulic pinion for Torque measurement on MGB. Strain gauging was done on four pylon
struts & three Tie rods, and load calibration was carried out in in both directions. Strain values
were measured by means of strain measurement system using data acquisition software.
Moments & thrust loads were applied using hydraulic jacks. The applied moments & thrust
loads were measured through load cells attached to the hydraulic actuators of the Rotor mast
loading block. After completion of each load case, the MGB was sent to the gearbox assembly
shop for modular level disassembly and recording of the contact pattern.
2.3 Functional tests for establishing the lube oil parameters
A functional test rig was designed and fabricated for functional testing of the MGB as per
requirements of measuring the basic parameters like oil pressures, oil flow rate, gearbox
housing and sump oil temperatures and vibrations, before subjecting the MGB to full load
tests. The rig comprised structural items, speed step up rig gearbox, drive system and motor.
The test rig also comprised of a flushing unit for MGB, rig lubrication tank and a structure to
mount AC Drive motor and speed step up rig gearbox. The structure was isolated from floor
by providing shock mounts. The required power to MGB input was obtained from the AC
Drive motor and speed increasing adapter gearbox. The Test rig Lubrication system provided
lubrication to the speed increasing adapter gearbox through hydraulic hose connections. A
high speed helicopter flexible shaft was assembled in between the speed increasing adapter
Gearbox shaft to MGB Input shaft. The test set up sketch is shown in figure 5.
Figure 5 : Schematic arrangement for MGB Functional Test Rig
The measurement set up consisted of Flow meter, Thermocouples, accelerometers and
Pressure sensors for the measurement of Flow, Temperatures, Vibrations and Pressures on the
MGB respectively. Virtual instruments via software on PC monitor were used as indicators to
measure the following parameters during the test - Body temperatures on Conical housing
Bearing locations on Freewheel housing, Input and TPTO housing, MGB housing, Accessory
housing, Cooler fan, Temperature at inlet to the Oil Cooler unit, Sump oil temperature,
Temperature at outlet to the Oil Cooler unit, Flow measurement at inlet to cooler unit,
pressure at Outlet of the Lube oil Pump, Tail Power Take off, Input to MGB, Oil cooler inlet,
Oil cooler outlet, Vibration Pickups on Conical housing in X,Y,Z directions, Input housing in
Radial directions, TPTO housing in Radial directions, Accessory housing in Radial directions,
Cooler fan Bearing in Radial directions and RPM at MGB input. A PC based data acquisition
system was installed to acquire and save the test data continuously during the run. As a safety
measure for the test rig and the specimen under test, all the test rig parameters and specimen
parameters were interlocked with the drive system to safely trip the test when the parameters
exceed the limits specified.
First the speed increasing test was carried out wherein the MGB was filled with oil and run by
increasing the speed in steps of 5%, 10% to reach 100% speed. Both Flow and Pressure
started building simultaneously with increase in speed. During the test run the absence of
leakages at all interfaces, and the MGB parameters were monitored and recorded. After
achieving satisfactory initial conditions, MGB was run at 100% speed for 1 hour. During the
run all MGB parameters were monitored and recorded for every 2 minutes of interval till the
temperature reached stabilization. The MGB parameters like temperature of housing surfaces
at various bearing locations, sump & ambient temperatures, Vibrations, Time trace and FFT,
oil flow rate and pressures were recorded.
Subsequently an accelerated cycle test run was carried out wherein the MGB was accelerated
to 125% speed and held for 30 seconds. Then, it was decelerated to 100% and held for 60
seconds. A total 50 cycles of test run was carried out and parameters were recorded. During
the test, the MGB was observed for any abnormal noise. During the run, the leakages at all
interfaces and MGB parameters were monitored and recorded.
Further, an oil quantity optimization test was carried out on the MGB with inclinations of
forward tilt, rear tilt, right tilt and left tilt with reference to the Rotor axis. For each of the
above inclinations, the MGB was run till the stabilization of sump oil temperature was
reached. During the run, the leakages at all interfaces and MGB parameters were monitored
and recorded for every 2 minutes. The test was further continued for determination of
undrainable volume and unused oil quantity in MGB housing, oil filter, cooler and hoses.
2.4 Endurance Load Testing of MGB
The test was carried out on an MGB test stand specially designed for conduction of load test
by application of closed loop torques, mast loads and accessory loads at the required rpm. The
rig was housed in a separate test facility, and comprised of the Test Rig Structure, a Multidrive system, Rig Gear boxes, Rotor Mast loading block, Hydraulic Actuators, Accessory
loading systems, Control & monitoring system, Data acquisition system, Sensors for
pressures, flow, temperatures, mast loads, torque, speed & vibration measurements.
Initially, static strain survey tests & dynamic wipe tests were conducted on MGB assembly to
evaluate the strain values on MGB with the actual mounting plates and Torque reaction parts.
During the test, various combinations of mast loads and thrust loads were applied on MGB.
The test stand sketch is shown in figure 6.
During the endurance load test the MGB speed was increased gradually in steps by increasing
the speed of an AC motor used as prime mover. Prior to applying loads on the MGB, a warm
up run was carried at no load to observe functional parameters of MGB. The rotor mast loads,
torques and accessory loads were applied as per the load spectrum to carry out the endurance
runs.
For running the MGB, an AC induction motor controlled by variable frequency drive (VFD)
was used as a prime mover. The AC motor was mechanically coupled to the MGB input
flange via suitable flexible couplings and Engine line test rig gearbox. By varying the input
frequency to the motor, speed was varied from 0 RPM to the required RPM. The torque
application on Input and TPTO was through an electrically closed loop [2] loading system
where the electrical energy is re-circulated in a closed loop. For this, two AC induction
generators were mechanically coupled to MGB main rotor shaft & Tail power output shaft via
suitable flexible couplings, and controlled by separate AC drives connected to a common DC
bus used for producing electrical power proportional to the torque required by regenerative
braking, where mechanical energy was converted into electrical energy. The powers thus
generated by regenerative braking during the test were fed back to the mains/grid, thus
constituting an electrically closed loop system. Hence, the power drawn from the mains
during testing is required only to overcome frictional losses in the mechanical system, thereby
significantly reducing mains power demand and operating cost of the testing. Since the test rig
is of electrically closed loop type, it can accommodate changes in teeth ratios in the Main
Gearbox during the development phase, thus making it flexible in nature. The torques were
measured by torque sensors mounted in line with input & TPTO shafts.
The speed increasing/ decreasing test rig gear boxes with suitable gear ratios ensured that
required RPMs at MGB input and at AC generators were obtained. A specimen fixing
carriage used for MGB assembly, thus completing the test set up. The generators used for
application of torque at TPTO were mechanically coupled to the MGB flanges via flexible
cardan shafts.
The specimen carriage was used to assemble and prepare the MGB outside the rig. After
assembly the carriage was pushed inside to align the MGB shafts with test rig gear box shafts
(input & TPTO) and the cardan shaft connecting to upper gear box. The entire structure was
isolated from ground by providing mechanical shock mounts. The test stand was positioned
inside sound proof room with sound proof sliding doors and viewing windows. The
lubrication system for intermediate gear boxes, hydraulic system for application of rotor mast
loads & for loading hydraulic accessory pumps were placed at basement below the test rig for
easy return of rig gearbox oil by gravity flow. External forced air cooling was used to cool the
specimen under test during the load test to maintain the MGB temperatures within the
specified limit.
Figure 6 : Schematic arrangement for MGB Load Test Rig
For rotor mast load & thrust loads electronic pressure controllers were used, which regulated
the proportional pressure valves mounted in hydraulic lines for each hydraulic jacks so that
each jack was controlled independently such that any combination of loads could be applied
on each jack. A programmable logic controller (PLC) was used in order to automate/control
the sequence of activities up to the readiness to start the actual run. These activities included
starting of cooling pumps for motor & generators, switching on lubrications for intermediate
gearboxes, switching on hydraulic system for MGB accessory pumps, switching on hydraulic
pumps for Rotor Mast Loading system, confirming the on-line status of the data acquisition
system and bringing the multi-drive system to the “ready” mode. The PLC ensured that the
“drive ready” status was reached only after the switching on all the motors, pumps, pressure
& flow switches and ensuring the online status, the data acquisition system etc. The current
feedback from generators to the drives maintained the torques always steady during the test.
The encoder feedback from the motor to the drive always maintained constant speeds without
fluctuations. As a safety factor, the control section also included protective functions such as
over current, over voltage & over temperature protection circuitry. During the load test all the
flight conditions were simulated as per the load spectrum. Since the testing involved the
loading of MGB under dynamic conditions, a variable speed AC motor was used to rotate the
MGB in the required speed and direction. The speed at the MGB input & TPTO were
measured by speed sensor mounted at MGB input & TPTO shafts.
Simulation of rotor mast loads & moments: Simulation of rotor loads & moments involves
designing of a mechanical system to make the application of loads on a rotating part possible
with the help of loading device fixed onto the static test rig structure. This rotor mast loading
system was therefore visualized in the form of rotor mast loading block with thrust bearings,
having four lugs at which hydraulic jacks are attached. By having independent control over
the hydraulic jack load, it became possible to apply the required combination of rotor mast
loads & also the moments. The rotor mast loads were measured by the jack load cells.
Simulation of accessories loads: The accessories loads like alternator, two hydraulic pumps
and oil cooler fan were simulated during the test. For this the original helicopter items were
used on the MGB. The alternator was loaded by using external resistance load bank, where
the energy was dissipated in the form of heat. The hydraulic pumps were loaded by using an
external pump loading system.
Safety features: An interlocking system was provided for each of the measured parameters
for tripping the drive and releasing all the loads in case of safety limit of any parameter
exceeds during the test. Also a video monitoring system was provided to monitor & record the
run continuously during test. All the sensors, conditioners & data acquisition systems were
connected to UPS power to take care of power fluctuations as well as to prevent loss of
important test data in the event of power failures. As a safety factor, the control section also
included protective functions such as over current, over voltage & over temperature protection
circuitry. A shear neck flange with supporting bracket assembly was introduced between
MGB flange and the rig gearbox. This shear neck flange acted as a mechanical fuse to prevent
sudden load spikes or over load, if any.
The MGB load test called for continuous monitoring of the following various parameters
during the test: RPM of MGB specimen, Pressures at inlet, TPTO & outlet, Oil flow rate,
Strain values at different locations on housing & MGB mounting parts like Z plates, Torque
strips, Temperatures at various locations like bearings, housings and oil sump areas, Vibration
levels at various locations, Rotor mast loads applied by each of the 4 rotor mast loading jacks,
Torque values at MGB input & TPTO, Loads on Tie-rods, pylon struts, and FFT & Time trace
for vibration parameters. In addition to the parameters called for the test specifications, it was
equally important to continuously monitor the satisfactory performance of the test rig by
monitoring various key parameters.
In view of the large number of parameters involved a PC based data acquisition system was
used to monitor and acquire data continuously by means of data acquisition via Lab-view
program. For the safety of test rig and specimen, all the instruments/sensors were calibrated
prior to the start of tests.
Following the above endurance tests, acceptance functional and load testing was carried out
on new gearboxes made to the final configuration for the prototype tests. The typical gear
tooth contact patterns obtained are shown in figure 7.
Contact pattern for gear tooth at no load
Contact pattern for pinion tooth at no load
Contact pattern for gear tooth –
Contact pattern for pinion tooth –
after load test
after load test
Figure 7 : Typical gear tooth contact patterns obtained
The endurance load tests were carried out according to a load spectrum comprising of steps
corresponding to the loads experienced by the MGB during flight, including maximum
continuous power (MCP) and take off power (TOP) steps. After passing these tests, the MGB
was assembled onto the LUH prototypes for further helicopter system and flight tests.
3.0 Conclusion
Special test set ups were prepared for the testing of the Main Gearbox of Light Utility
Helicopter as per the test programs, and extensive testing was carried out in the various test
stands indigenously developed at HAL. The feedback from all these tests was used to finalise
the design of the helicopter main gearbox for use during the prototype testing phase of the
helicopter.
References:
1. DEF STAN 00-970 (Transmission system)
2. www.renktestsystems.com (Gearbox testing)
3. FAR 27 document
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Experimental Validation of Random Packs for Composite Solid Propellants
Using X-Ray Computed Tomography
Chaitanya VIJAY 1, K RAGHUVARUN 2, KV Sai BHARGAV 1, Krishnan
BALASUBRAMANIAM 2, PA RAMAKRISHNA 1
1
Department of Aerospace Engineering, IIT Madras, Chennai, India
Phone: +91 2257 4018, Fax: +91 2257 4002; e-mail: chaitu89@gmail.com, parama@iitm.ac.in
2
Department of Mechanical Engineering, IIT Madras, Chennai, India; E-mail: balas@iitm.ac.in
Abstract
This paper presents the first of its kind comparison of an actual solid propellant with a computationally
generated random pack, which is intended for use in combustion modelling studies. X-ray computed
tomography is used to obtain a 3D reconstruction of an actual propellant sample. Random distribution
of spherical particles of aluminium and ammonium perchlorate (AP) have been used to model a
composite solid propellant (random pack). The computed random pack and 3D reconstructed
propellant are compared by slicing them both at 10 µm intervals and computing an average area of
exposed AP particles. The average areas of exposed AP particles calculated in simulated random
packs and X-ray Computed Tomography (XCT) scan of a real propellant are in good agreement within
the limitations of the resolution of X-ray Computed Tomography images. Thus, it is inferred that the
random propellant pack simulated matches well with reality.
Keywords: Composite solid propellants, X-ray computed tomography, random packs.
Nomenclature:
Al = Aluminium
AP = Ammonium Perchlorate (NH4ClO4)
HTPB = Hydroxyl Terminated Poly Butadiene
XCT = X-ray Computed Tomography
C:F ratio = Coarse-to-fine ratio of AP particles
1. Introduction
Solid propellants are used in the launch vehicle industry apart from having applications in missile
systems. Solid propellants can be broadly classified into two categories based on their composition,
viz., homogeneous and heterogeneous propellants. The current study restricts itself to heterogeneous
propellants, which are also known as composite solid propellants. The constituents of composite
propellants, viz., ammonium perchlorate(AP) and hydroxyl terminated polybutadiene (HTPB) are
mixed mechanically. A composite propellant essentially consists of crystalline AP particles in a matrix
of HTPB, which acts as the binder as well as the fuel. In certain applications aluminium (Al) powder
is used as a fuel in addition to HTPB in order to enhance the specific impulse of the solid propellant
Error! Reference source not found..
The combustion field of a solid propellant is not amenable to direct measurement of gas phase
properties since the combustion processes have extremely small time scales and geometry around each
AP particle is three dimensional. Also, large gradient of properties makes measurements complex.
Thus, modelling solid propellant combustion is one method of attempting to understand the underlying
physics. Literature Error! Reference source not found. has results of 3D computations performed
on random packs which are computationally expensive. They have therefore tried to bring down the
computational time by homogenizing the fine particles.
The main issue addressed in this paper is to study the propellant to see if the random pack being
used to model propellant combustion is indeed representative of a propellant. In order to model
combustion phenomenon, random sphere packings have been used to represent the particles of AP and
Al in a propellant due to its simplicity. However, it is essential to examine whether this assumption is
really valid with regard to AP particles which are typically non-spherical. This paper compares
computationally generated random packs to actual propellants using X-ray Computed Tomography
(XCT) as the diagnostic tool in order to examine the assumption.
Solid propellant modelling has made steady progress over the last few decades. Computationally
generated three dimensional propellant packs have been used to study propellant combustion by
several researchers. Ramakrishna [3,4] attempted a statistical comparison between generated
propellant packs and an actual propellant. Since 2000, UIUC group Error! Reference source not
found. has been pursuing computational simulation of 3D combustion field of propellants. The basis
for development of their propellant packs was the Lubachevsky-Stillinger packing algorithm Error!
Reference source not found.. This algorithm follows a concurrent packing method, where a
randomly assigned set of points in a predefined cube grow into particles according to a growth
function, while also colliding with each other. Multi-modal distributions have been developed from
this algorithm by allocating different growth rates to different set of particles based on the particle
sizing and volume packing fraction to be achieved in the propellant pack. Due to the extensive number
of events which need tracking in such an algorithm, it is computationally expensive. Sankaralingam
and Chakravathy Error! Reference source not found. presented a method to develop propellant
packs to study the flame-let distribution over a composite solid propellant. They considered a bi-modal
distribution of AP particles with a coarse to fine ratio of 5.6:1. They optimized their pack with
concepts borrowed from simulated annealing. The coarse-to-fine (C:F) ratio, by mass or equivalently
by volume, of AP used is not representative of industrial class of propellants, which vary from 4:1 to
1:1. Further, Maggi et al. Error! Reference source not found. devised an improved algorithm to
generate propellant packs which was an improvement over the method presented by Kochevets et
al.Error! Reference source not found.. Recently, Baietta and MaggiError! Reference source
not found. presented a study on parallel programming to generate propellant packs. All the above
methods, barring Error! Reference source not found.,have resorted to comparison of their models
with previously published experimental studies by researchers on sphere packings [10-12]. All these
experiments study the packing of spherical metal/plastic shots which are not representative of AP
particles in a propellant. The current study is the first to present a comparison between an actual
propellant and a simulated model propellant using the XCT technique.
Also, the simulated three dimensional packs which have been developed are seen to have a C:F
ratio of 3.4:1, 5.6:1, 6.5:1 [5,7]. These choices of C:F ratios are not indicative of any industrial class
of propellant, which have coarse-to-fine ratios of 1:1 or 4:1. Since, it is an established fact Error!
Reference source not found. that particle size effects play a definite role in determining the burn
rate, pressure index and other important parameters of the propellant, it is imperative that the random
packs generated to simulate the propellants are representative of industrial packs.
2. Methodology
The exposed area of AP and binder to the combustion field in a propellant is key to propellant
combustion and is an important factor in deciding combustion characteristics of the propellant. An
analysis of exposed surface area will thus throw more light on burning of propellants. Thus, the
approach taken in this study is to build a random pack and compare it with a 3D XCT reconstruction
of a real propellant. Slices obtained from both XCT scan and the simulated random pack are compared
for exposed area of AP or Al particles.
For the purpose of this study, four representative packs and propellants were prepared with a
view to closely represent propellants used in the industry. The method of preparation of the propellant
is similar to the procedure detailed by Sumit and Ramakrisna0. The composition, particle size
distribution, theoretical and actual densities are listed in Table 1. As can be seen, the densities of
propellants prepared are close to theoretical densities achieved in the model.
Table 1 Propellant composition used for the study
Composition
Mass loading
Sample 1
Sample 2
Sample 3
AP-HTPB
AP-HTPB
AP-Al-HTPB
Sample 4
AP-Al-HTPB
80%
80%
86% (68%
AP+18% Al)
86%(68% AP +
18% Al)
1:1
2:1
1:1
Measured
Density
(kg/m3)
1540
1600
1776
Calculated
Density
(kg/m3)
1593
1616
1781
2:1
1779
1781
C:F
ratio
2.1 Random Propellant Pack
The basic algorithm used to generate the propellant pack is similar to the one used by
RamakrishnaError! Reference source not found.. The AP particles are assumed to be hard spheres
and they do not intersect each other. The co-ordinates of AP particles are assigned from a random
number generator. They are placed in a cube of fixed length subject to a constraint that it does not
intersect any of the other particles which have a priory been placed. On the edge of the cube, only the
portions of a particle which lie within the cube are taken into account to calculate the volume occupied
by the same. In case of multi-modal distributions, AP and Al particles are filled in a descending order
of particle size and each particle is filled up to a pre-calculated volumetric loading, based on the mass
loading of the particle size in the overall propellant. The volume not occupied by the AP or Al
particles in the cube is taken to be the binder in the propellant.
The average particle sizes considered in the study for bimodal propellant is 325 μm and 54
μm. The particles are distributed in the band 300 - 355 μm for coarse and 45 - 63 μm for fine AP to
account for variation in particle sizes in an actual propellant. A normal distribution is assumed for the
same. These particular sizes were primarily chosen since the above mentioned meshes were available.
As the aim of the study is to validate the random pack with an actual propellant, this choice of particle
size distribution is most suitable. A 3D model of the propellant pack having a cube size of 1 mm x 1
mm x 1mm is generated using MATCAD® and is shown in Fig. 1(a). The coarse particles are
coloured red and fine particles yellow.
Fig. 1 (a) Simulated random pack (Left). (b) XCT reconstruction of a real propellant (Right).
2.2 X-Ray Computed Tomography
XCT technique identifies density differential and reconstructs a 3D model based on that. The process
of reconstruction is detailed in 0. In an attempt to build a 3D representation of the propellant sample,
XCT scan was performed on the samples using General Electric – Phoenix V|tome|xs. It is fitted with
a micro focus tube and nano focus tube. The capacity of micro focus tube is 240 kV and the resolution
is 10 µm and the capacity of Nano focus tube is 180 kV and the resolution is 0.5 µm. Phoenix datosx
software is used for data acquisition and reconstruction. Volume graphics studiomax (VG Studiomax
software) is used for analysis. The scan was performed on the entire region of the sample using a nano
focus tube. The energy required to perform the scan is 60kV and 250 µA and the resolution is around 5
µm. The averaging rate is 4 and the time required to acquire one image is 1000ms and 1000such
images were acquired. The 3D representation of the propellant sample was reconstructed from the
XCT data obtained. The reconstructed propellant sample is shown in Fig. 1(b).
2.3 Analysis of Exposed AP Surface Area
In order to compare exposed area of AP and Al particles, slices at fixed intervals were passed through
the random pack which was generated. The exposed surface area of AP obtained through slicing is an
indicator of the burning characteristics of the overall propellant. In this study, a propellant pack of
dimension 1 mm x 1 mm x 1 mm was used and slices 10 μm apart were used to calculate the average
area of AP and aluminium exposed on the surface. The particle sizes were kept the same as mentioned
earlier.
Figure. Fig. 2(a) and Fig. 2(b) shows one of the slices for a sample 1 and sample 3 respectively.
The coarse, fine AP particles and aluminium particles are coloured as red, yellow and blue
respectively. The aluminium particle size here is considered to be 18μm - 25μm. The exposed area was
averaged over 100 planes to represent the average percentage exposed area of AP or aluminium. The
average exposed area remained invariant with number of slicing planes greater than 20.
Fig. 2 (a) Representative slice through Sample 1(Left) (b) Representative slice through Sample 3
(Right)
A similar method of passing slices through the XCT reconstruction is done. An image from the XCT
scan is shown in Fig. 3(a). As can be seen, Fig. 3(a) is similar in nature to Fig. 2(a) which was
generated by means of passing slices through the generated random propellant pack. Therefore,
analysing the surface by means of image processing and averaging the exposed AP surface area across
many slices allows us to make a comparison between an actual propellant and the random propellant
pack. It is pertinent to note that the images of AP in Fig. 3(a) are not circles as in Fig. 2(a). This
indicates that the spherical particles assumed may not be accurate. Further discussion is deferred.
2.2 Image Processing
Images of slices obtained from XCT scan were converted to binary images with the threshold being
calculated using the classical Otsu's algorithm0 for unsupervised image segmentation. A representative
image is shown in Fig. 3(b). The representative area of AP particles from the XCT scan image shown
in Fig. 3(a) is captured quite well through the image segmentation algorithm as can be seen from Fig.
3(b). The corresponding exposed AP areas are calculated as in Eq. 1.
𝐴𝑟𝑒𝑎 𝑜𝑓 𝐴𝑃 =
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑖𝑥𝑒𝑙𝑠 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑖𝑛𝑔 𝐴𝑃 𝑝𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠
𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑖𝑥𝑒𝑙𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑖𝑚𝑎𝑔𝑒
(1)
Fig. 3 (a) Slice through XCT reconstruction showing exposed AP particles (b) Binary image calculated
using threshold obtained from Otsu’s algorithm 0
The area of AP was also averaged across a minimum of 100 slices for each real propellant sample.
3. Results and Discussion
3.1 Analysis of Exposed AP Area
The results of AP and Al area obtained from the simulated random packs and processed XCT images
are as depicted in Table 2.
Table 2 Comparison of average % AP area exposed computed by both methods
Sample 1
Sample 2
Sample 3
Sample 4
Exposed %AP Area
XCT
Computed
53.7%
53.5 %
59.5%
59.6 %
41.6%
63.9 %
38.3%
60.1 %
The variation of exposed area, along the length of sample 1 is shown in Fig. 4. The variations
of exposed area for other samples show a similar trend. The standard deviation for the random pack
and XCT scan is 5.78 and 2.18 respectively.
Fig. 4 Comparison of exposed surface area along length of the propellant sample calculated by both
XCT and random pack
There is a difference between prediction and experiments in the average percentage area of AP
and Al exposed obtained in samples 3 and 4 as seen from Table 2. This difference can be explained if
the resolution of XCT scan images, which is around 5.3μm, is taken into account. Fig. 5 shows an
illustration of the effect of resolution on the boundaries of particles. The lighter shaded region
represents the area lost due to pixel resolution. This fact is illustrated in Fig. 5, which shows that some
aluminium particles may be consumed within the pixel and end up with a sub-pixel areas which will
remain undetected. The interval of slicing also plays a role in smudging of aluminium particles. Fig. 6
shows an illustration of the same. Even if the slicing interval is 5-6 μm, a maximum of 3 slices pass
through aluminium particles since their particle size ranges from 18 – 25 μm. Thus, aluminium
particles contribute significantly less area or no area in some cases. This comparison is seen quite well
in Table 2.
Fig. 5 Illustration of area and particles lost due to resolution of XCT scan
The average AP area exposed for samples 1 and 2 match very closely as is evident from Table
2. However, having said that, the estimates of the model are conservative in nature and it is likely to
under-predict the exposed area although not by much. This is due to the assumption that all particles
are hard spheres, and it is evident from XCT scan image (Fig. 3(a)) that particles are not exact spheres.
As sphere has the lowest surface area per unit volume, having non-spherical particles would mean that
surface area could be more than the predicted.
Al particle
Slicing planes
Fig. 6 Effect of slicing plane on detection of Al particles
In order to test the hypothesis regarding resolution of XCT scan, sample 1 was reconstructed
with a coarser resolution of 11μm. A correction was applied to the slices from the random pack similar
to what is shown in Fig. 5. The diameter (d) of AP particles were reduced by value of the resolution of
the image. The recomputed average exposed area of AP for random packs was found to be 40.71%.
This area is 13 % lower than the value corresponding to sample in Table 2. The average exposed area
of AP computed from XCT scan with 11 μm resolution was 38.92%. This area is 14 % lower than the
value corresponding to sample in Table 2. The percentage area obtained through both methods shows
a good match. This further establishes that if the resolution of the XCT images are taken into account,
the exposed area of AP particles is in good agreement with both the methods. This also validates the
random pack developed as it has been compared with real propellants.
4. Conclusion
An experimental comparison between a real propellant and a simulated random pack was performed
using XCT as the diagnostic tool. Simulated random packs assuming AP and Al particles as perfect
spheres were generated and then compared with the re-constructed XCT data. In order to compare
experimentally obtained XCT data and the simulated random pack, slices at definite intervals were
passed and an exposed area of AP and Al particles were calculated by averaging over a minimum of
100 slices. Random packs achieve a conservative estimate of the exposed area of AP and predict an
area exposed area for 80% solid loading propellant between 53-59 %. The exposed area of AP
calculated, matches well with that obtained experimentally. The difference in the results in case of
sample 3 and 4 is attributed to the limitation in resolution of XCT images and no comment can be
made on whether the exposed areas match with an actual propellant as of now and further study is
necessary in that direction. However, the random pack for Sample 1 and 2 are suitable for further
modelling studies as it compares with the average exposed area of AP particles.
References
1.
F.A. Williams, H. Barare and N.C. Huang, ‘Fundamental Aspects of Solid Propellant
Rockets’, Technivision Services, Slough, England, 1969. p.192.
2.
L. Massa, T.L. Jackson, J. Buckmaster and M. Campbell, ‘Three-Dimensional
Heterogeneous Propellant Combustion’, Proceedings of Combustion Institute, Vol 29,
No 2, pp 2975-2983, 2002.
3.
P.A. Ramakrishna, P.J. Paul and H.S. Mukunda, ‘Statistical analysis of Composite Solid
Propellants’, in: R.R. Panyam, S. Soundranayagam, S. Anantharam (Eds.), Air
Breathing Engines and Aerospace Propulsion, Interline publishing, pp 1-5, 1996.
4.
P.A. Ramakrishna, ‘Sandwich Propellant Combustion: Computational Studies and
Experimental Comparisons’, PhD Thesis, Indian Institute of Science, Bangalore, India,
2003.
5.
S. Kochevets, J. Buckmaster, T.L. Jackson and A. Hegab., ‘Random Packs and Their
Use in Modeling Heterogeneous Solid Propellant Combustion’, Journal of Propulsion
and Power, Vol 17, No 4, pp 883-891, 2001.
6.
B.D. Lubachevsky and F.H. Stillinger, ‘Geometric properties of Random Disk Packings,
Journal of Statistical Physics’, Vol 6, No 5, pp 561-583, 1990.
7.
K. Sankaralingam and S.R. Chakravarthy, ‘A Computer Model of Flamelet Distribution
on the Burning Surface of a Composite Solid Propellant’, Combustion Science &
Technology, Vol 161, No 1, pp 49-68, 2000.
8.
F. Maggi, S. Stafford, T.L.Jackson and J. Buckmaster, ‘Nature of Packs Used in
Propellant Modeling’, Physical Review, Vol 77, pp 046107-1-17, 2008.
9.
A. Baietta, F. Maggi, ‘Parallel Packing Code for Propellant Microstructure Analysis’,
Aerospace Science and Technology, Vol 46, pp 484-492, 2015.
10.
T. Aste, M. Saadatfar, T.J. Senden, ‘Geometrical Structure of Disordered Sphere
Packings’, Physical Review E, Vol 71, pp 061302, 2005.
11.
G. Mason, ‘Radial Distribution Functions from Small Packings of Spheres’, Nature, Vol 217,
No 5130, pp 733–735, 1968.
12.
R.K. McGeary, ‘Mechanical Packing of Spherical Particles’, Journal of American
Ceramics Society, Vol 44, No 10, pp 513–522, 1961.
13.
S. Verma and P.A. Ramakrishna, ‘Dependence of Density and Burning Rate Of Composite
Solid Propellant on Mixer Size’, Acta Astronautica, Vol 93, pp 130-137, 2014.
14.
G.T. Herman, ‘Fundamentals of Computerized Tomography: Image Reconstruction from
Projection’, 2ndEd., Springer Publishing Corp., NY, 2009.
15.
N. Otsu, ‘A Threshold Selection Method from Gray-Level Histograms’, IEEE Transactions on
Systems, Man and Cybernetics. Vol 9, No 1, pp 62–66, 1979.
Fault Detection in Electrical Wires and Cables
Gholamhossein Shirkoohi
School of Engineering
London South Bank University
103 Borough Road, London SE1 0AA, England
Contact: maziar.shirkoohi@lsbu.ac.uk
Computer models of electrical wires have been investigated in order to develop systems and test
procedures that could lead to identification and inspection for all types of fault, particularly small
defects in insulation of electrical wires and cables. The technique was first introduced by reporting
the results for the measurements in 5 mm and 10 mm faults obtained from the simulations are similar
to those reported in measurements carried for the coaxial cables [1]. The technique being developed
aims to develop systems that could provide means of detection of small defects in insulation and
shielding of cables. The method involves injection of a high frequency pulses into the wires, and the
analysis of the subsequent returning reflections. The technique has been investigated using the
modern electromagnetic modelling tool, Concerto [2], which was initially developed for rf and
microwave applications [3]. The 3D model was also used to represent the twisted pair cable [4] which
was the twisted version of the twin cable modelled earlier [5]. In order to achieve this, the straight
wire pairs were twisted along its length with a precise twist node periodicity of an actual twisted pair
cable of similar cross sectional dimensions. The full length model would require a much finer mesh
that would demand allocation of far more RAM, CPU time, and would subsequently run progressively
slower within the real-time post processor module. Perfect conductor (PEC) boundary condition is
constructed around the wires, in addition to those of the source and the load ports were assigned,
where pulses in the form of radiation could enter the model (from input port). The output port in most
cases was set to short circuit as termination. For the twisted wires a much finer mesh is needed which
reduces the model size to a much smaller, 500 mm length of cable for precipitous analysis of the data.
This means that the twisted pair model was some twenty times smaller than the models previously
investigated. The final analysis of the host of different types of wires and cables show the technique is
reliable and could be developed for tangible applications.
References
1. Shirkoohi, G., Hasan, K., “Enhanced TDR Technique for Fault Detection in Electrical Wires and
Cables,” 2nd International Symposium on NDT in Aerospace 2010, tu2b3, Hamburg, Germany,
Nov 2010, pp. 1-6
2. CONCERTO 7 Reference Manual, Vector Fields Ltd, Oxford, 2008
3. Suidong Yang, Michaelides, A., Riley, C., Archer, J., Hook, M., Simkin, J, "Optimizing the
design examples of wide-band antennas," Wideband and Multi-band Antennas and Arrays, 2005,
IEE (Ref. No. 2005/11059), Sept. 2005, pp. 185-190
4. Shirkoohi, G., “Fault detection in aircraft wiring using enhanced multi-pulse TDR technique”7th
International Symposium on NDT in Aerospace 2015, mo5a6, Bremen, Germany, Nov. 2015, pp.
1-8
5. Shirkoohi, G., “Modelling and simulation of fault detection in Shielded Twisted Pair cables,”
2016 IEEE International Conference on Industrial Technology (ICIT), Taipei, Taiwan, April2016,
pp. 1039-1044, DOI: 10.1109/ICIT.2016.7474897
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Structural Health Monitoring of Aircraft Composite Structures:
Offline & Online Approach
Ramesh SUNDARAM1, Nitesh GUPTA2, Augustin MJ2, Amitabha DATTA2
1
Head, Advanced Composites Division, CSIR-NAL, Bangalore
Scientist, Advanced Composites Division, CSIR-NAL, Bangalore
Phone: +91 80 25086400, Fax: +91 80 25267352; e-mail: rameshs@nal.res.in
2
Abstract
Composite materials have been in use for aerospace applications for more than two decades These materials
exhibit the unique properties of high strength-to-weight ratio, high stiffness and corrosion resistance. Aircraft
and spacecraft are typically weight sensitive and hence composite materials are well suited for these applications.
Damage tolerant and fail-safe design of aerospace structures requires a substantial amount of inspection and
maintenance adding significant amounts to their lifecycle cost. To detect and repair various structural damages
that can occur during the service life of the aircraft, a thorough inspection schedule is implemented through
conventional visual and Non Destructive Evaluation (NDE) methods. Such scheduled inspections lead to
considerable increase in maintenance cost & down-time of the aircraft. The lifecycle cost of aircraft and
aerospace structures can be reduced significantly if continuous and autonomous condition based structural health
monitoring (SHM) systems can be integrated into their design.
A structural health monitoring (SHM) system consisting of well-designed sensor networks incorporated in the
structure along with necessary hardware and software can provide information about the structure, thereby
leading to reporting of flaws or damages.
This paper discusses about the online and offline approach of SHM system being developed at Advanced
Composites Division of CSIR-NAL using fiber optic sensors.
Keywords: Structural Health Monitoring (SHM), Fiber Optic Sensors, Artificial Neural Network (ANN), Barely
Visible Impact Damage (BVID)
1. Introduction
Continuous growth of the air traffic in last few decades has generated the requirement of
manufacturing low budget aircrafts without compromising on the safety issues. Composite
materials are increasingly used in airframe structures owing to their high specific strength and
stiffness, resistance to corrosion and possibility of structural tailoring. Composites offer the
designer the flexibility of combining multiple smaller components into a single large
component using co-curing technology. This is an attractive option as it eliminates the need
for mechanically fastened subassemblies and the associated assembly time and cost. However,
to reap the structural benefits, a proper bonding between skin and substructure should be
assured throughout the service life of the aircraft. Foreign object damage (FOD) can lead to
disbonding of the skin and stiffener interface without any visible evidence. Skin-stiffener
disbond in composites is an area of concern among aircraft structural designers [1].
The complex nature of damages in composite structures necessitates periodic checking, which
is usually carried out through visual inspection and ultrasonic NDE. These methods are
sometimes limited by the inaccessibility of interior parts which may leave damages
unidentified. Any solution, which continuously monitors the status of the structure and
informs the concerned personnel, can lead to a timely and cost effective solution to this
problem. In this regard, various aircraft industries and research labs are pursuing development
of systems and methods for structural health monitoring of composite aircraft structures.
Structural Health Monitoring (SHM) has been defined in the literature as the “acquisition,
validation and analysis of technical data to facilitate life-cycle management decisions” [2].
The use of SHM systems has the potential to provide greater confidence to the user on the
integrity of the structure.
1. Components of Structural Health Monitoring System
Basic concept of SHM has been evolved essentially from the human system as shown in Fig.
1. The basic idea of having an on-board SHM system is to make a structure intelligent akin to
a human being [3-5].
Brain
Nervous System
Human Body
SHM Algorithms
Sensor System
Structure
Fig 1. SHM analogy for
Human body and Structures
Fig 2. Life cycle of SHM system
Having structures equipped with an SHM system will enable the structure to inform the
maintenance about its overall health based on the information gathered from the built-in
sensors & processed by SHM algorithms. The life cycle of an SHM system is shown in Fig. 2
[6]. The enabling components to realize an SHM system are detailed in the following sections.
1.1 Structural Design & Analyses
Monitoring an entire structure would prove to be very expensive and would probably never
see the light of day. Hence, it is important that an SHM system development for an aircraft
programme must be incorporated right from the structural design phase. Designing of an onboard SHM system necessitates an optimization of sensor location on the aircraft structure,
where strain needs to be monitored. This optimization should be arrived at after a detailed
finite element (FE) analysis considering various types of damages which may occur during the
entire life of aircraft; such as manufacturing process induced damages, damage occurring
during the assembly, damages during the service life of aircraft. A damage threat perception
analyses would be extremely important in this case which would result into the optimized
sensor location. In case of already built-up structure the final sensor location may also include
the service history information of the structure. This should also take into consideration of the
manufacturing wherein embedment or surface bonding of sensors is plausible.
1.2 Sensor Integration & Interrogation
A robust SHM framework requires the installation of a distributed sensor network so that
damage measurements can be made quickly and frequently without significant effort or
expense. Sensor technology has matured enough to have highly sensitive, reliable and
miniaturized strain sensors which can be embedded or surface bonded to the composite
structure during/after the manufacturing. Several types of sensor networks are being
investigated, including strains gauges, Piezo transducers and fiber optic sensors [8-15].
The sensor installation process is specific to the structure and manufacturing technology.
Hence, this process needs to be developed so as to be incorporated at the shop floor level for
large scale production of aircraft components. Ruggedized embedment schemes should ensure
that the sensors survive for the life of structure.
Sensor measurement is instrument specific and must cater to the operating and service
environment within the aircraft. As large number of sensors can be installed on the structure,
grid instrumentation schemes may also be adopted wherein each system will measure the
sensor data from specific part of the aircraft.
The data needs to be stored at the local level as well as in central processing computer thereby
ensuring redundancy. It is also important to have synchronization of data storage with flight
manoeuvres. This will not only ensure the efficient data management but also reduces the
complexities in sensor data pre-processing prior to the use of SHM algorithms.
1.3 SHM Methodologies & Algorithms
The key to an effective SHM system for aircraft structures is not only the appropriate sensor
selection and installation but also the processing of the sensor data to predict the flight load
and the damages, if any.
SHM algorithms are crucial elements in the implementation and operation of any damage
identification system. The generic system requires the availability of appropriate signal
processing technology to extract features from different types of sensors and use this
information for diagnostics. The pre-processing, which filters out all unwanted features, is the
first element of this chain. This pre-processing must involve basic signal processing
techniques such as zero detection, down sampling, filtering etc. This is followed by various
procedures for feature extraction and selection. It is important to isolate data features which
are sensitive to damage and insensitive to the loading and operational conditions. This is the
most difficult task in damage identification. Since data features are often combined in
patterns, the entire process of damage identification can be considered as a pattern recognition
procedure.
In order to develop the SHM algorithms, various techniques, such as artificial neural networks
(ANN), fuzzy logic, signal processing and genetic algorithms, have been looked into
worldwide [16-19]. The outcome of having an on-board SHM system will provide valuable
inputs to the maintenance personnel and aircraft designer. SHM technology would lead to a
‘condition based maintenance’/‘maintenance on demand’ paradigm that would increase
inspection intervals, leading to lower maintenance costs. SHM have a great impact on
structural design that would result in lower weight, thus reducing fuel costs or increased
payload. SHM would help in better fleet management leading to better resource utilization
through its prognostic capabilities.
3. Classification of Structural Health Monitoring System
The above mentioned components of a SHM system and their respective maturity level, from
their implementation point of view on an aircraft structure, classifies the SHM system into
following broad categories namely (i) On-line SHM, (ii) Off-line SHM and (iii) Hybrid SHM
system
3.1 On-line SHM System
An on-line SHM system; as the name suggests; requires the sensors to be the integral part of
the structure. The necessary measurement system to acquire and store the sensors data would
also be part of the on-board instrumentation. This necessitates the instrumentation to be
qualified for the flight regime in terms of environmental screening specifications (ESS). The
damage detection and load estimation algorithms may also become part of the onboard
processing.
The on-line SHM system will enforce the certification and approval of the previously
discussed SHM system components namely sensor integration, measurement system and
SHM algorithms from respective approving authorities.
3.2 Off-line SHM System
An off-line SHM system does not necessitate any of the SHM system components to become
part of the structure. Non-Destructive Evaluation technologies for structural inspection is one
of the best example for this category.
3.3. Hybrid SHM System
However as various components of a SHM system might not be at the same Technology
Maturity Level simultaneously hence depending upon their readiness level, there could be a
combination of these components wherein some of them could be either on-board or off-line.
This type of system can be categorised as hybrid SHM system.
4. SHM System Developmental Activities at Advanced Composites Division
The group at Advanced Composites Division (ACD) in National Aerospace Laboratories
(CSIR-NAL) has been pursuing the development of an aircraft SHM system using fiber optic
sensors, that attempts to address above aspects. Fiber optic sensors have several advantages
such as low weight, high sensitivity, immunity to electromagnetic interference, multiplexing
capability, etc. The various aspects of SHM system development such as sensor
characterisation, rugged embedment, sensor measurement system and algorithms for damage
and load estimation, have been addressed. The technology has been demonstrated on ground
using both quasi distributed Fiber Bragg Grating (FBG) sensors and fully distributed
Rayleigh’s scattering based bare fiber optic sensors for ground test box structures. FBG
technology has also been demonstrated at flight test level wherein the sensor measurement
was carried out in real time and data processing for damage and load estimation has been done
post flight test. In the hybrid SHM system approach, division has developed the technology
using bare fiber optic sensors. In this case the sensors would be the integral part of the
structure. However, the measurement would be done between the flights to assess the
structural integrity. Further section of this paper discusses about the brief of these
developmental activities.
5. SHM System Development using Fiber Bragg Grating (FBG) Sensors
This section of the paper focuses on the use of FBG sensors for SHM system development for
both off-line and on-line demonstrations.
FBG sensor relies on the narrowband reflection from a region having a periodic variation in
the core refractive index. The shift in the reflected wavelength from the central wavelength of
the FBG gives a direct measure of external perturbation. Necessary sensor characterisation
studies have been carried out to determine strain and temperature sensitivity of the sensor.
Rugged embedment schemes have been devised to sort out the ingress/egress issues ensure the
sensor installation at shop floor environment. A novel sensor patch scheme was developed to
ensure the quick and hassle free installation of the FBG sensors without affecting the existing
manufacturing process, as shown in Figure 3.
Fig 3. Sensor patch Embedment
Various experimental tests such as Tensile, compression and ILSS tests at room temperature,
with specimens having embedded optical fiber, were conducted to showcase that installation
or embedment of these sensors does not degrade the structural properties.
5.1 Test box studies using FBG sensors
The principal types of damages in composites are - Failure of fibre bundles, matrix cracking,
delaminations and de-bonding at joints. Delaminations have been thought of as critical to
structural integrity. It is known that the delaminations have practically little effect on the
tensile strength but the compressive strength is reduced. Various studies have shown that
among the different damage modes in co-cured and co-bonded composite structures, the
critical area of concern is the skin-stiffener debond/disbond.
In order to addresses the issue of skin-stiffener debonds in composite structures and its effect
on the strains, typical composite aircraft test boxes were fabricated with bolted, bonded and
co-cured constructions. The first testbox for this purpose was fabricated with five spars and
one central rib. Both top and bottom skins were fastened to the spars. In order to study the
effect of disbonds of different lengths on strain patterns, the disbond was created by bolt
removal studies. It has been shown experimentally that the proximity of the sensor location to
the disbond is important to have a faithful detection of the disbond [6,20].
Another testbox was further fabricated wherein the disbond was created by using a teflon
insert at the time of secondary bonding of the spar with the top skin after manufacturing.
Details of this composite box are shown in Figure 4 and Figure 5. The test box consists of top
and bottom skins and three spars. Both top and bottom skins are stiffened with co-cured hat
shaped stiffeners to eliminate buckling of skin during loading. The end spars are C-sections
which are fastened to top and bottom skins. The middle spar is an I-section which is co-cured
with the bottom skin and subsequently secondary bonded to the top skin. During this
secondary bonding operation, a 200mm long disbond is deliberately created between top skin
and middle spar using a Teflon insert. Subsequently, using a novel technique this disbond is
‘closed’ in multiple phases allowing us to conduct tests for various disbond sizes until
eventually the disbond is closed completely to obtain a ‘healthy’ structure.
510mm
760mm
Fig 4. Model of the Composite Testbox
Fig 5. Composite Testbox prior to asembly
Three sizes of disbonds, which are 200mm, 150mm and 100mm length, were considered as
part of this study. The disbond width of 90mm was constant for all the cases
The location of the fiber optic sensors was decided based on strain distribution in the
composite box obtained from numerical simulations and disbond location. The location of the
FBG sensors in the top skin is shown in Figure 6. Numerical model of the box was validated
through limited experimental tests on the testbox. Multiple loading scenarios (Figure 7), were
considered during tests: (a) Up-bending case (Max Load 2870Kg), (b) Down-bending case
(Max Load 1200Kg) & (c) Bending coupled with torsion case (Max Load 2400Kg).
P/4
P/2
P/4
P/4
P/2
P/4
(b)
(a)
Disbond
P/2
P/2
(c)
*not to scale
Fig 6. Sensor layout on Composite Testbox
Fig 7. Testbox Loading scenarios
Each of this scenario was repeated for all three disbond sizes and finally on the healthy
structure as well. FBG data was recorded using Micron Optics sm130 interrogator and sm041
channel multiplexer. All sensors were recorded simultaneously at each load step for all
loading schemes during the tests. Test data from 200mm disbond size case was considered for
validating the numerical model. Figure 8 and 9 present the variation in strains on the top skin
above mid-spar flange along the spanwise direction for 200mm disbond case for up bending,
down bending respectively. These figures show that there is good correlation between test
data and numerical results, giving confidence in the numerical model developed.
Following validation, the numerical model was used to generate required data for training the
Artificial Neural Network (ANN). In the present study, 26 different load cases and 4 different
disbond sizes (i.e. 200mm, 150mm, 100mm and 50mm) were considered for generating
training data from numerical model for different load cases. This training data set was divided
so as to be handled by five networks to avoid the problem of over-fitting and leading to better
generalization. Once trained, this set of five networks formed the damage and load estimator.
Fig 8. Strain variation for Up bending
Fig 9. Strain variation for Down bending
When an unseen strain pattern is fed to the damage estimator, a pattern recognition algorithm
determines the best ANN (out of the 5 networks) to be used for estimating disbond size,
disbond location and total applied load. Figures 10 to 12 show the estimation of ANN for 200
mm debond case. The study has validated for 150mm, 100 mm disbond cases also and found
to be in good agreement.
3000
400
Actual load
Load predicted by ANN_DE
Actual location of disbond
Location predicted by ANN_DE
Load (kgf)
2000
1500
1000
390
380
370
360
500
0
0
500
1000
1500
2000
Actual load (kgf)
2500
Fig 10. Load Estimation
3000
Actual disbond size
Disbond size predicted by ANN_DE
220
Disbond size (mm)
Disbond location (mm)
2500
350
0
210
200
190
180
0
500
1000
1500
2000
Actual Load (kgf)
2500
3000
Fig 11. Location Estimation
500
1000
1500
2000
Actual load (kgf)
2500
Fig 12. Size Estimation
5.2 Impact Detection studies using FBG sensors
Detection of such impact events using strain response of the structure obtained during impact
event is important for aerospace Structural Health Monitoring (SHM) applications. The
impact event for SHM applications need to be informed in terms of the location and the
resultant energy. Hardware and software triggered data acquisition schemes for the data
acquisition and sensor network based on the directional strain sensitivity and operational
safety were developed and implemented. In order to predict the location and estimate the
impact energy, different algorithms were developed based on the strain profile acquired
during the impact. Strain Amplitude, Strain Scan Based, Energy index based, Centroid based
and Correlation based algorithms were developed for the location estimation and validated on
structures with different level of structural complexity [21]. Support Vector Regression,
System Identification based approaches are being developed and validated for the energy or
force estimation and is validated through experiments on laminates and skin stringer panel .
The performance index which is a measure of prediction accuracy as a function of number of
sensors, dimension/area of the laminate and the maximum error, is also defined. Cumulative
Distribution Function (CDF) which is a measure of confidence over estimation was defined
for both laminates and skin-stringer panel. The CDF for the location estimations on laminates
and skin stiffener panels are shown in Figure 13 and 14 below.
It can be observed that, confidence of getting error of 70mm or less in any impact test is about
90% for laminates and that of getting 73mm or less is about 90% for skin stringer panel.
Fig 13. CDF for location estimation on laminates
Fig 14. CFD for location estimation on panels
5.3 Flight Test Results Using FBG Sensors
In order to move towards a flight worthy SHM system, flight tests were conducted on 2-seater
HANSA aircraft and Nishant UAV. In case of Nishant UAV FBG sensors were embedded in
the booms and were measured during the flight. The data was recorded on the certified onboard instrumentation. In-house developed software was used to estimate the flight loads postflight using the ANN methodology as explained earlier. Figure 15 shows the flight data
alongwith estimated load. A flightdaya playback software QuickVIEW© was also developed
to post process the FBG sensors data alongwith other flight parameters such as pitch, roll,
engine speed, altitude level etc. to provide the flight status immediately after the landing
(Figure 16) [22].
3000
Catapult Launch
Parachute
Recovery
Figure 15. Flight Test Results
on UAV
Fig 16. QuickVIEW© GUI
6. SHM System Development using Plain Fiber Optic Sensors
The quasi distributed sensing of FBG sensors picks up the strain signatures at unique points
on the structure (e.g. along the center spar of the box under study). However, if the disbond is
at other location along the spar then it may get missed. In order to overcome this problem,
Rayleigh’s Scattering based fiber optic sensor was used. This section of the paper focuses on
the use of bare/plain fiber optic cable as sensors for SHM system development for off-line
demonstrations.
In this case entire length of the fiber serves as the sensor. Luna Inc. ODiSi-B system was used
to interrogate the sensor at every 5mm. This has resulted into the total of 2000 sensors on a
fiber of length 10 meters. Necessary sensor characterisation studies were carried out to
determine the strain and temperature sensitivities coefficients of the sensor [23].
6.1 Testbox Studies Using Bare Fiber Optic Sensor
The composite testbox which was fabricated for the FBG ground test studies as discussed
previously, was used to study the use of bare fiber optic sensor for SHM application.
However, as the top skin has already made healthy by closing the debond in a controlled
manner during FBG sensor studies, hence the top skin has been debonded completely with the
center spar and then fastened back. This has provided the option of studying the debonds of
different lengths (26mm pitch between two consecutive fastener/bolts) along the center spar.
The bare fiber optic sensor was bonded on the top side of the top skin as shown in Figure 17.
To create a disbond of full flange width between center spar and top skin, a pair of adjacent
bolts is removed – one from each row.
Testbox was subjected to static tests in cantilever configuration. In order to estimate the load
from developed strain, model based approach was followed. The model was built using the
raw strain data acquired from the instrument. The structure was first loaded from 0 Kgf to
1200Kgf (all bolts fastened) and strain data was recorded in steps of 200Kgf and the acquired
data is segmented (taking only the line of interest; in this case the center line of the optical
fiber between the bolts). By analysing the segmented data, it was conirmed that the slope of
the strain vs. length plot is linearly proportional to the applied load. Based on this, a reference
model is created with loads vs slopes for healthy signatures of different loads. Slope values
are obtained by fitting a first order polynomial to each segmented data. Using this method, it
was possible to estimate the load which matched well with applied load values. In addition to
this Area based method (wherein area under the strain vs fiver length curve was used to
generate the reference model) Support Vector Regression (SVR) method (wherein feature
extraction of slope values, intercept values and area under the curve to create the model) was
also developed for load estimation. Figure 18 shows the result of load estimation of these
approaches w.r.t. actual applied load.
Figure 17. Bare Fiber Layout on the testbox
Fig 18. Load Estimation
Damage detection algorithm is developed based on reference based method, where strain
signature/profile is available for healthy structure (without any damage). This approach can be
used to monitor a structure from its given state i.e., to find any changes in the structure with
respect to original state of the structure. The flow diagram for Damage detection approach is
show in Figure 19.
Healthy strain signature along with respective load was recorded when all bolts are fastened.
An experimental simulated disbond was created by removing bolts (in pair, on both sides of
the fiber Refer Fig 17). Strain signatures were recorded in unhealthy case as well for
respective load values during the static test of the box. The successive difference of strain
signals obtained from difference of health reference signal and an unhealthy signal for a given
load was determined. The Gaussian fit was obtained on the envelope of this successive
difference strain signal along the length of the fiber, which provides the location and size of
the disbond as shown in Figure 20.
Fig 19. Bare Fiber Layout on the testbox
Fig 20. Damage Estimation
The disbond estimation approach was validated for both static and dynamic load cases and
results obtained are shown in Figure 21 and 22. The end dotted lines in these figures shows
the actual disbond length as per the bolts removal. The red dots in Figure 21 shows the
disbond estimation for different load cases in case of static testing. The green line shows the
disbond estimation under dynamic loading conditions.
It can be observed from these figures that the estimated values for disbond size and location is
matching well with the actual values for different load values and type of loading.
Figure 21. Disbond estimation (Static Test)
Fig 22. Disbond estimation (Dynamic Test)
7. Conclusion & Future Scope of Work
In order to have a fully qualified SHM system, it is important that the midway approach needs
to be taken. This necessitates the inclusion of the sensor and measurement technology for
FBG sensors to be adopted for the flight tests in order to collect the valuable flight data.
Additionally, bare fiber optic sensors can also be installed onto the structure to assess the
structural integrity information post flight. The data collected over period of time would be
useful to develop the model based approach for damage and load estimation. However, it is
equally important to look into the physics based approach for damage detection and load
estimation. The flight data collected over period of time would prove to be extremely useful in
this regard and would pave the way for fully qualified SHM system.
Acknowledgements
The work discussed herein this paper is the cumulative results of the various projects executed
at the division over last decade. Authors would like to acknowledge the support of
Aeronautical Development Agency under the Development Initiative for Smart Aircraft
Structures programme and the Aeronautical Research and Development Board under the
Centre of Excellence for Composite Structures Technology (Phase II & Phase III) programme.
The authors acknowledge the extensive support of Council of Scientific & Industrial Research
(CSIR), New Delhi under 11th Five Year Plan to demonstrate the technology at testbox level.
Authors express their thanks to Aeronautical Development Establishment (DRDO-ADE) for
their help and support to conduct the flight trial. Authors thanks the Director, NAL and staff
of ACD for their support and encouragement.
References
1 D. Roach & s. Neidigk ‘Does the Maturity of Structural Health Monitoring Technology
Match User Readiness?’ International Workshop on Structural Health Monitoring
(IWSHM 2011)
2 Charles R. Farrar, F Hemez and J Czarnecki ‘A Review of Structural Health Monitoring
Literature 1996 – 2001’ Hoon Sohn, , Los Alamos National Laboratory repor t(2003)
3 B. Beral and H. Speckmann ‘Structural health monitoring for aircraft structures: A
challenge for system developers and aircraft manufacturer’, 2003,Proc. 4th Int.
Workshop Structural Health Monitoring, Stanford, California, USA..
4 HJ. Schmidt and B.S Brandecker ‘Management of aging civil aircraft-The challenge of
the aerospace industry ‘. 2002., Proc. 8th Int. Fatigue Cong., Stockholm, Sweden
5 C. Boller, ‘Ways and options for aircraft structural health management, Smart Materials
and Structures’. 2001, 10, 432-439. http://dx.doi.org/10.1088/0964-1726/10/3/302
6 Gupta et al, ‘Structural Health Monitoring of Composite Structures - Issues and
Challenges’. 2012. Int. J. Vehicle Structures & Systems, 4(3), 74-85
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Roach, D, ‘Real time crack detection using mountable comparative vacuum monitoring
sensors.’ 2009, Smart Structures and Systems. 5(4), pp. 317–328.
S.C. Galea, S. Velden, I. Powlesland, Q. Nguyen, P. Ferrarotto, M. Konak , ‘Flight
demonstrator of a self-powered SHM system on a composite bonded patch attached to
an F/A-18 aileron hinge’ Pacific Workshop on SHM, Yokohama, Japan, 2006
Steven W. Arms, Christopher P. Townsend, Jacob H. Galbreath, Stephen J. DiStasi,
Daniel Liebschutz and Nam Phan, ‘Flight Testing of Wireless sensing Networks for
Rotorcraft Structural Health and Usage Management Systems’, 14 Fourteenth
Australian International Aerospace Congress, 7th DSTO International Conference on
Health & Usage Monitoring (HUMS 2011)
Eric Kosters and Thomas J. van EIS, ‘Structural health monitoring and impact detection
for primary aircraft structures’, http://spie.org/x39419.xml.
Betz DC Thursby G, et.al. , ‘Structural damage location with fiber Bragg grating
rosettes and Lamb waves’, Structural Health Monitoring 2007. - 299–308: Vol. 6(4).
C. Fu-Kuo, ‘Composite structures with built in sensors’, ICCM12 Europe.-ICCM, 1999.
Tsutsui H, et.al., ‘Detection of impact damage of stiffened composite panels using
embedded small-diameter optical fibers’, 2004 Smart Mater. Struct. 13 1284–90
Botsev Y et al, ‘Fiber Bragg grating sensing in smart composite patch repairs for aging
aircraft’ 2004 Proc. SPIE 5502 100
Jang B W et al ‘Real‐time impact identification algorithm for composite structures using
fiber Bragg grating sensors’ 2012 Struct. Control Health Monit. 19 580–91
C. Ramadas, K. Balasubramaniam, M. Joshi and C.V. Krishnamurthy, ‘Detection of
transverse cracks in a composite beam using combined features of Lamb wave and
vibration techniques in ANN environment’. 2008., Int. J. Smart Sensing and Intelligent
Systems, 1(4), 970-984.
N. Martin Jr, A. Ghoshal, M.J. Sundaresan, G. Lebby, P.R. Pratap and M.J. Schulz., ‘An
artificial neural receptor system for structural health monitoring’ 2005., Int. J. Structural
Health Monitoring, 4(3), 229-245. http://dx.doi.org/10.1177/1475921705055241
G.M. Kamath, et.al., ‘A neural network based health monitoring methodology for cocured/co-bonded composite aircraft structures’ 2006., Proc. 3rd European Workshop
Structural Health Monitoring, Granada, Spain
Li, Z.X., and Yang, X.M, ‘Damage identification for beams using ANN based on
statistical property of structural responses’, Computers & Structures 2008, Vol. 86, No.
1-2, pp 64-71
Sundaram R et.al., ‘Structural health monitoring of co-cured composite structures using
FBG sensors’. Proceedings of the SPIE International Conference on Smart Structures
and Systems, Newport Beach, USA. (2005)
Amitabha Datta, MJ Augustin, Nitesh Gupta, SR Viswamurthy, Kaushik N, Pitchai P,
Gaddikeri Kotresh and Ramesh Sundaram, ‘Performance comparison of impact event
localization algorithms using FBG sensors for CFRP structures’,ISSS National
Conference on MEMS Smart Materials, Structures and Systems, 2015, Kochi, India
I Kressel1, et.al., ‘Flight validation of an embedded structural health monitoring system
for an unmanned aerial vehicle’, Smart Mater. Struct. 24 (2015) 075022 (9pp),
doi:10.1088/0964-1726/24/7/075022
Amitabha Datta, SR Viswamurthy, Sakthi Sathya, Augustin MJ, N Gupta, Kotresh
Gaddikeri and Ramesh Sundaram, ‘Experimental Studies using Distributed Fiber Optic
Sensor for Aircraft Structural Health Monitoring Application,’ International conference
on Fiber optics and Photonics -2014
8th International Symposium on NDT in Aerospace, November 3-5, 2016
NDT investigations on C/SiC samples from different manufacturing steps
Mykhailo KYRYCHENKO1, Susanne HILLMANN2, Frank MACHER2, Martin Schulze2,
Henning HEUER1,2, Cesar CAMERINI3, Gabriela RIBEIRO PEREIRA3
1
Institute of Electronic Packaging Technology, Technische Universität Dresden; Dresden, Germany
Phone: +49 351 462 42924, Fax: +49 351 463 37035; e-mail: Mykhailo.Kyrychenko@tu-dresden.de
2
Fraunhofer Institute for Ceramic Technologies and Systems IKTS; Dresden, Germany; e-mail:
henning.heuer@ikts.fraunhofer.de
3
Laboratory of Nondestructive Testing, Corrosion and Welding of the Federal University of Rio de Janeiro
(LNDC / UFRJ)
Abstract
This paper is focused on eddy current measurements of C/SiC samples in different status of the manufacturing
process. The measurements are compared with results of X-Ray, Scanning Acoustic Microscope (SAM) and
optical images and discussed.
Keywords: C/SiC, eddy current, NDT, X-Ray, carbon fiber composite
1. Introduction
Using of eddy currents is state of art for nondestructive testing of electrical conductive
materials [1]. The developed high-frequency eddy current technology “EddyCus®” (with
frequency ranges up to 100 MHz) made it even possible to extend the classical fields of
application towards less conductive materials like CFRP [2]. More details to this technology
can be found in [3], [4].
C/SiC is ceramic matrix composite (CMC) and belongs to a class of materials developed for
aeronautics and space applications, in a domain where superalloys cannot be used
anymore [5]. In comparison to metallic structures C/SiC-composites have many advantages as
excellent thermal and mechanical properties by lower weight, high and stable friction
coefficient, long life, low wear rate, and lower sensibility to surroundings and oxidation [6].
They have potential applications in structures (air intakes, structural panels with stiffners,
etc.), in turbines and for brakes [5].
EddyCurrent (EC) testing was already performed on C/SiC samples in [7]. In this paper are
provided NDT examinations for three C/SiC manufacturing process steps.
The idea is to transfer the collected experience by investigations on CFRP to up-and-coming
C/SiC.
2. Experimental Setup and Practice
2.1 Samples
The samples, shown in Figure 1 represent different steps of the manufacturing process of
C/SiC and are numbered with 99, 100 and 101. These samples were provided by the working
group of Ceramic Matrix Composites at the Universität Bayreuth.
Figure 1: Optical view of C-SiC specimens
Samples dimensions are presented in Table 1:
Table 1: Dimensions of C/SiC samples
99
100,1*81,5*(3,3-4,1) mm3
100
99,7*81,1*(3-3,7) mm3
101
99,9*(81-81,7)*(3-3,3) mm3
Sample "99" is a pure CFRP (carbon fiber reinforced plastic). In "100" the resins were burned
out, so carbon fibres and carbon particles remain. In “101” the cavities were filled with Si and
after that the sample was sintered. Thus all samples have different states with different
properties and different defects (if existent).
2.2 Nondestructive Test Methods
2.2.1 Optical evaluation
There was a digital photo performed. Displays in photos and surface optical review were
evaluated and documented. The inconstant illumination intensity of the test and the
subjectivity of the assessment must be mentioned as disadvantages.
2.2.2 Eddy currents
Eddy current testing is a well-established nondestructive method for the characterization of
surfaces or materials by analyzing conductivity and permeability variations [3].
Alternating current passing the excitation coil induces a primary magnetic field. It excites
eddy currents in a specimen. They create a secondary magnetic field that opposes the primary
magnetic field. Changing the specimen material properties leads to changing the path of eddy
currents and correspondingly complex signal changing at the pick-up coil.
Measurements were done with EddyCus® MPECS – Multi Parameter Eddy Current Scanner.
A semi-transmission coil with middle frequency 4 MHz was used. All eddy current images in
this paper are done with 2.3 MHz.
Figure 2: EddyCus® MPECS
Figure 3: EddyCus® MPECS axes
Accordingly to axes on Figure 3, coil angles were defined: 0° along the x-axis, 90° along the
y-axis and 45° was the position, as in the Table 2.
Table 2: Coil angle definition
0°
45°
90°
Almost all eddy current measurements in this paper were done in contact-mode because of
better resolution. To prevent wearout of coil, samples were coated with polyethylene film.
By choosing (rotating) a Complex Phase Angle (CPA), noise signals can be filtered out.
2.2.3 X-Ray radiography
The use of X-rays for nondestructive testing is widely spread and offers numerous different
methods for the characterization of chemical and structural properties of the samples. The Xray radiation has wavelengths of about 10-6 m to 10-12 m and therefore belongs to the highenergy, ionizing radiation, which can penetrate a large number of substances. The simple Xray transmission is most widely used in the NDT because it is simple and fast to perform.
In the case of radioscopy, the structures of the sample are imaged as a function of their
attenuation, with objects lying one behind the other in the beam direction overlaid. Defects
within structure can thus be made visible [8].
For X-Ray radiography was used a phoenix nanomex (Figure 4).
2.2.4 Scanning acoustic microscopy (SAM)
Ultrasound refers to sound waves with frequencies above 20 kHz. It can be applied for
material nondestructive investigations.
An acoustic microscope uses the ultrasound propagation possibility in solids and liquids. An
ultrasonic pulse is generated in a transducer using a piezoelectric material. The generated
sound wave, focused via a lens, passes through the coupling medium water into a sample. The
interaction at interfaces between different materials (including inclusions or defects) is
investigated.
Using the focusing lenses allows receive higher resolution as normal ultrasound testing.
Scanning method makes it possible to get a two-dimensional image.
A scanning acoustic microscope assembled at Fraunhofer IKTS-MD was used for
measurements (Figure 5).
Figure 4: C/SiC specimen in X-Ray microscope
Figure 5: Scanning acoustic microscope at
Fraunhofer IKTS-MD
3. Results
Some of measuring effects are marked with *number* in text and with correspondent number
in figure.
3.1 Sample 99
Sample 99 is a pure CFRP (carbon fiber reinforced plastic).
The X-ray image on Figure 6 shows material particles that have a higher density *1* (dark
spots). They could not be seen with other tested methods. Also regions with smaller densities
can be observed *2* e.g. less filled. These areas were also seen visually on Figure 8 – “areas
with hollows”. They were not visible in the eddy current images due to an edge effect.
In the eddy current image on Figure 9 can be seen fiber structures. Noticeable spots are
marked on the picture. Two dark spots *3* resemble two hollow spots in the photo. Other
abnormalities couldn’t be captured with other methods.
In the optical view can be seen slightly swollen areas. One of them is also to see on Figure 7
(EC-image).
Different semi-transmission coil angles or other complex phase angle by eddy current testing
can influence a result image. You might get new signals, as *4* on Figure 10. Different phase
angles allow some signals to be displayed more clearly.
3
1
2
Figure 6: Sample 99, X-Ray
image
Figure 7: Sample 99, EC image:
2.3 MHz; 90°; 180° CPA
Figure 8: Sample 99, optical view,
front side
3
Figure 9: Sample 99, EC image:
2.3 MHz; 45°; 180° CPA
4
Figure 10: Sample 99, EC image:
2.3 MHz; 45°; 315° CPA
3.2 Sample 100
In Sample 100 the resins were burned out, so carbon fibres and carbon particles remain.
With X-Ray radiography on Figure 11 can be seen similar indications as in sample 99. At the
bottom right is marked a dark area *5*. It means greater weakening.
In the optical view on Figure 13, this area looks as if metal was infiltrated there. The area is
also visible on Figure 16 from the back side. It has to be discussed, why the area is not visible
in the eddy current image.
Some displays in the eddy current image on Figure 12 match optical view, some are new.
5
Figure 11: Sample 100, X-Ray
image
Figure 12: Sample 100, EC image:
2.3 MHz; 0°; 235° CPA
Figure 13: Sample 100, optical
view, front side
Figure 14: Sample 100, EC image:
2.3 MHz; 45°; 180° CPA
Figure 15: Sample 100, EC image:
2.3 MHz; 90°; 315° CPA
Figure 16: Sample 100, optical
view, back side
Depending on the angle of the semi-transmission coil, different displays appear more clearly
(compare Figure 12, Figure 14 and Figure 15). Optimization of the image contrast allows
most favorable image for the eyes.
3.3 Sample 101
In “101” the cavities were filled with Si and after that the sample was sintered.
8
7
6
9
Figure 17: Sample 101, X-Ray
image
Figure 18: Sample 101, EC image:
2.3 MHz; 45°; 0° CPA
Figure 19: Sample 101, optical
view, front side
Figure 20: Sample 101, SAM
image: 1 MHz; 20mm focal length
Figure 21: Sample 101, EC image,
with Lift-off: 2.3 MHz; 45°;
180° CPA
Figure 22: Sample 101, EC image:
2.3 MHz; 90°; 180° CPA
Scratch, that visually can be seen on Figure 19, is also visible in X-Ray image on Figure 17.
With eddy current it was not detected.
It is supposed that big circular indication *6* in eddy current images (see Figure 18, Figure
21, Figure 22) is silicon infiltration border. This border is partly in SAM image on Figure 20
visible. In X-Ray image can be seen two parts: the darker in the middle *7* and ring-shaped
brighter left *8*. The second part is barely visible.
On Figure 18 at the bottom in the middle are circular measurements-artefacts *9* visible.
Such artefacts were visible only by this sample in contact mode. By measurement with
minimal Lift-off (Figure 21) they were not detected. This artefact comes from illustration of
coil itself on surface irregularities.
On Figure 22 marked displays come clearly.
4. Conclusion and Outlook
C/SiC samples from three manufacturing steps were investigated with eddy current testing
and compared with X-Ray, SAM und visual images of the samples. Eddy current is promising
solution for material controlling during C/SiC manufacturing steps.
Supposedly a silicon infiltration border was seen by sample 101.
Direction of semi-transmission coil has big influence on defects detectability with eddy
current testing.
In future are planned systematic defect inductions during production of the same sample and
evaluations with nondestructive testing. In this way has to be proven suitability of NDTmethods for detection of specific defects.
Acknowledgements
We would like to thank the working group of Ceramic Matrix Composites at the Universität
Bayreuth, in particular Dr. Langhof, for providing the examined samples.
References
1.
2.
3.
4.
5.
6.
7.
8.
'Non-destructive testing – Eddy current testing – General principles', ISO 15549, 2008
M. Schulze, H.Heuer, 'Textural analyses of carbon fiber materials by 2D-FFT of
complex images obtained by high frequency eddy current imaging', Nondestructive
Characterization for Composite Materials, Aerospace Engineering, Civil Infrastructure,
and Homeland Security, SPIE Proceedings Vol. 8347, edited by A.L. Gyenyesi, pp 1825, 2012
H. Heuer, M. Schulze, M. Pooch, S. Gäbler, A. Nocke, G. Bardl, Ch. Cherif, M. Klein,
R. kupke, R. Vetter, F. Lenz, M. Kliem, C. Bülow, J. Goyavaerts, T. Mayer, S. Petrenz,
'Review on Quality Assurance along the CFRP Value Chain – Nondestructive Testing of
Fabrics, Preforms and CFRP by HF Radio Wave Techniques', Elsevier, Composites
Part B Engineering 27, compositesb.2015.03.022, March 2015
G. Bardl, A. Nocke, C. Cherif, M. Pooch, M. Schulze, H. Heuer, M. Schiller, R. Kupke,
M. Klein, 'Automated detection of yarn orientation in 3D-draped carbon fiber fabrics
and preforms from eddy current data', Elsevier, Composites Part B Engineering 96, pp
312-324, 2016
G. Boitier, S. Darzens, J.-L. Chermant, J. Vicens, 'Microstructural investigation of
interfaces in CMCs', Elsevier, Composites: Part A 33, pp 1467–1470, 2002
Shangwu Fan, Litong Zhang, Yongdong Xu, Laifei Cheng, Jianjun Lou, Junzhan Zhang,
Lin Yu, 'Microstructure and properties of 3D needle-punched carbon/silicon carbide
brake materials', Elsevier, Composites Science and Technology 67, pp 2390–2398,
2007
V. K. Srivastava, A. Udoh, H.-P. Maier, P. Knoch, K. Maile, 'Eddy current
nondestructive mapping of C/C–SiC composites', Springer-Verlag, Forschung im
Ingenieurwesen 68, pp 169 – 172, 2004
K.-J. Wolter, M. Bieberle, H. Budzier, G. Gerlach, T. Zerna, 'Zerstörungsfreie Prüfung
elektronischer Baugruppen mittels bildgebender Verfahren', Verlag Dr. Markus A.
Detert, 2012
Thermal Properties of Basalt Fibre Epoxy Composites by Focused Gaussian
Illumination Using Infrared Thermography
Kalyanavalli V, T K Abilasha Ramadhas, and D SastiKumar,
Department of Physics, National Institute of Technology,
Tiruchirappalli, India
Contact: itsabi@gmail.com
In recent years much attention is focused towards feasible composites which are reinforced with
naturally available fibers. Basalt is one such natural material that is found in volcanic rocks and it has
gained increasing attention as a reinforcing material when compared to traditional glass fibers due to
its ecological safety and natural longevity. Basalt fiber composites have high potential and are getting
a lot of attention due to its high temperature and abrasion resistance. This study investigates the
thermal properties of basalt composites prepared by hand lay-up followed by vacuum bagging
technique. Focused Gaussian Illumination method was adopted to measure the in-plane, in-depth
diffusivity of the composite. The Infrared Signals emitted from the specimen showed the surface
temperatures like a Gaussian profile. A standard laser flash method was adopted to validate the indepth diffusivity measurement. The change in thermal properties like conductivity and diffusivity
during manufacturing was characterized using different fiber orientation.
This work is supported by Technical Education Quality Improvement Programme (TEQIP), India.
In-situ Non-destructive Evaluation of Composite Panels in an Aircraft using
Infrared Thermography
S Kshama and S Kalyana Sundaram
Structural Lifing & Monitoring Group
Structural Technologies Division
CSIR-National Aerospace Laboratories
Bengaluru, India
Contact: kshama@nal.res.in
Composite materials are finding widespread application in construction of aircraft structures due to its
inherent merits of high strength to weight ratio and inertness for corrosion etc. But they are sensitive
to impact damages and hygro-thermal degradation etc. that are encountered during aircraft operations
in adverse environment. In-situ inspection of these kinds of damages on composite aircraft structures
having box type construction with conventional NDT techniques, like ultrasonics, are tedious and
often lead to poor interpretation of results. Research work presented here details the advantages of
infrared thermography with convective thermal excitation developed for in-situ inspection of
operational damages. A way of overcoming certain limiting hurdles in employing the reflection mode
active IR thermography for the in-situ inspection arising from accessibility of surfaces, lack of
thermal excitation, geometrical complexities, thickness of laminates etc. are discussed here. Besides,
quantification of damages commonly encountered in composite laminates is also detailed.
NDE of Aerospace Structural Materials and Components using Acoustic Wave
Based Tools
M R Bhat
Department of Aerospace Engineering
Indian Institute of Science
Bangalore, India 560 012
Contact: mrb@aero.iisc.ernet.in
With increased requirements of safety and reliability of performance, health monitoring of structures,
in particular in aerospace industry, is currently an area under extensive research and development.
Acoustic wave based techniques owing to their obvious advantages have established themselves as a
major set of tools for NDE and SHM of aerospace structural materials and components. While the
basic principles utilized for the purpose remain the same, the concepts of health monitoring require
extraction of comprehensive information integrating passive and active methods with the long term
objectives of making the structures smart. Thus, the aspects that gain importance are the sensors,
methods of embedding these sensors, data acquisition and processing to extract comprehensive
information. Detection and monitoring of crack initiation and propagation in metallic structures
though is still of prime importance, advanced composite materials replacing the aerospace structural
applications in a big way, naturally have been the materials under extensive investigations for
detection, evaluation and characterisation of different types of defects, damage and failure
mechanisms encountered in this category of materials.
This paper presents an overview of research work being carried out over the last decade through
experimental approach in the area of NDE & Structural Health Monitoring (SHM) using Acoustic
Methods at the department of Aerospace Engineering, Indian Institute of Science, Bangalore, India.
Through well planned set of experimental investigations different techniques based on Acoustic
Methods have been employed to detect, monitor and characterize different types of defects and
damage in aerospace structural materials and components. The studies involved laboratory
experiments on coupons as well as some field applications on composite structural components.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
A Non Destructive Methodology of Estimating Single Crystal Elastic
Constants.
Phani Mylavarapu 1, Karthik Karuparthi 2, Jai Prakash Gautam 2
1
Defence Metallurgical Research Lab, Hyderabad, India; E-mail: phanimylavarapu@dmrl.drdo.in
2
School of Engineering Science and Technology, University of Hyderabad, Indi
Abstract
Single crystal elastic constants (SCEC) of materials occupies an important role in the
estimation of Young’s modulus (E), bulk modulus (G) and Poisson’s ratio (ν) of materials.
SCEC are usually determined destructively by tensile and shear loading a single crystal
specimen. However, one of the major limitations in that procedure happens to be the
availability of single crystal (in case of destructive testing). Hence, in this study, a nondestructive procedure has been developed to estimate SCEC from polycrystalline specimens.
Ultrasonic longitudinal and shear velocities, longitudinal attenuation coefficient and
backscattered grain noise (FOM) are measured on pure Copper (forged condition) and pure
Aluminium (virgin ingot condition) specimens. In addition to the experimental measurement,
ultrasonic parameters are also estimated analytically using the existing relationships involving
SCEC and probabilistic grain size. SCEC were further estimated by minimizing the difference
between experimentally measured and analytically calculated ultrasonic parameters.
Combinations of velocities, attenuation coefficient and velocities, backscattered noise (FOM)
are considered in the minimization function. Global minimum of this minimization function
resulted in SCEC. It is observed that the estimation of SCEC using the procedure adopted in
this study is highly repeatable with less than 10% absolute error between estimated value and
literature value of SCEC. Moreover, it is also observed that equi-axed microstructure of
Aluminium resulted in lesser error in the predicted SCEC compared to copper with random
microstructure.
Single Crystal Elastic Constants, Ultrasonic Velocity, Attenuation Coefficient,
Backscattered Grain Noise (FOM), Minimization.
Keywords:
1. Introduction
Structural materials are crystalline in nature with constituent atoms or molecules arranged
in a specific order with repeated units called as “Unit Cell”. On the basis of the atoms
arrangement in crystals, materials are further classified into Cubic, Hexagonal, Tetragonal,
Orthorhombic, Rhombohedral, Monoclinic and Triclinic. In solids, the linear elastic
properties depend on the magnitude of interatomic force between the particles whereas the
response of the arrangement of atoms to the applied load influences the plastic deformation in
the material. This response in all three orthogonal directions determines the strength and
failure of the material [1]. The strength of the material within the elastic region is expressed
in terms of elastic properties such as Young’s modulus (E), bulk modulus (G) and Poisson’s
ratio (ν). These elastic moduli are in-turn related to single crystal elastic constants (SCEC) of
material.
1.1 Single Crystal Elastic Constants (SCEC) for Cubic Crystal Symmetric materials.
In solids linear dependence of the strain on the stress applied can be explained using
Hooke’s Law,
σ =C ε ;ε =S σ
(1)
where,σ is stress applied, ε is strain and Cijkl is Stiffness (Elastic) constant, Sijkl is
compliance constant. Using stress and strain symmetry relations, 81 elastic constants defined
for any crystal structure are reduced to 21 independent elastic constants. Further, cubic
crystals which possess four threefold rotational symmetry are defined by three independent
single crystal elastic constants (C11, C12 and C44) [1].
For a cubic symmetric material, single crystal elastic constants can be related to bulk
modulus B, Young’s modulus E, shear modulus G, Poisson’s ratio (ν) by Voigt-Reuss-Hill
average model [2]. Due to the importance of estimating the mechanical properties of the
material with the help of single crystal elastic constants, predicting or measuring these
constants is of utmost significance.
Several researchers have determined SCEC either by destructive methods or using nondestructive testing (NDT) techniques. Zhang et al [3] and Safarik et al [4] applied mechanical
testing on single crystal specimen and resonant ultrasound spectroscopy (RUS), respectively
on single crystal, to estimate SCEC. Patel et al [5] estimated single-crystal elastic constants
from the polycrystalline material using nano-indentation and orientation measurements. With
the increasing usage of NDT techniques in various fields of material science, including
material characterization, the current study deals with developing new approaches using
ultrasonic testing to estimate single crystal elastic constants.
2. Experimental Ultrasonic Testing
Estimation of SCEC of copper and aluminium requires experimental and analytical
measurement of ultrasonic parameters such as ultrasonic velocity, attenuation coefficient and
figure of merit (FOM). In this study, immersion based ultrasonic pulse-echo experimental
techniques are utilized for measuring ultrasonic parameters such as longitudinal velocity,
attenuation coefficient and backscattered grain noise and contact based through transmission
technique is used for shear wave velocity.
2.1 Sample preparation
Both copper and aluminium possess face centred cubic structure (FCC).Copper sample
with dimensions of 10 mm cube is cut from a forged ingot while aluminium sample with
dimensions of 20 mm cube is sectioned from an ingot in virgin condition (i.e. without any
mechanical or thermal treatment.). Samples are sectioned using EDM wire cut machine.
2.2 Experimental measurements of ultrasonic parameters
Ultrasonic signals required to extract ultrasonic parameters are captured using
immersion based pulse-echo technique. 6-axis ultrasonic immersion equipment with
dimensions of 1*1*1 m3 supplied by M/s Dhvani Research, India along with Panametrics
5900 pulser/receiver from M/s Olympus International, USA is used for ultrasonic inspection.
A-scan signals are captured using acqUT software provided with the immersion equipment.
Ultrasonic signals are captured with atleast two backwall reflections to measure velocity
and attenuation as shown in Figure 1(a). To measure backscattered grain noise, signal
between frontwall and first backwall is captured with enhanced gain as shown in Figure 1(b).
Captured ultrasonic A-scan signals are processed using Matlab algorithm to extract ultrasonic
parameter values. Details of ultrasonic frequency used, waterpath distance maintained and the
thickness of the samples are shown in Table 1.
(a)
(b)
Grain Noise
Figure 1: (a) Representative A-scan signals (b) Signal with enhanced gain indicating
backscattered noise.
Table 1: Corresponding probe frequencies, water path and thickness for Copper and
Aluminium samples.
Frequency of
Water path
Thickness of
Material
Probe(MHz)
(mm)
specimen (mm)
Copper
2.25
55.4
10
Aluminium
5
23.5
20
2.2.1 Ultrasonic Wave Velocity
Ultrasonic wave velocity in the material is
measured as the ratio between twice the
thicknesses of the sample to the time taken
for travel as shown in Equation 2
2d
V =
(2)
t
where, d= thickness of the sample, t =is the
time taken for one round trip of the
ultrasonic wave.
Time of arrival for the front wall, first and
second back wall reflections is obtained
using Hilbert transform. Figure 2, shows the
Hilbert response with the corresponding
time of arrival of echoes. The time
corresponding to the maximum peak
amplitude is considered as the time of
arrival of the corresponding echo.
2.2.2 Longitudinal Attenuation Coefficient
Figure 2: Hilbert transform of gated echoes
(α)
The attenuation coefficient using focused probes (for aluminum) is calculated using a ratio of
frequency responses of the successive reflectors as shown in Equation 3. In the case of
unfocused probe (for copper), attenuation coefficient is calculated by considering the
diffraction correction of the ultrasonic beam [6] as shown in Equation 4.
For focused transducers,
α(f) =
ln
()
()
(3)
For unfocused transducers,
α(f) =
ln
( )|
( )|
( )|
(4)
( )|
where, A1 (f), A2 (f) are the amplitude of frequency spectrum for the 1st and 2nd backwalls,
respectively. D1(f), D2(f) are diffraction coefficients to reduce the ultrasonic beam divergence
effect.
Normalized diffraction coefficient is given by
D (f) = 1 − exp(−iq ) J (q ) + iJ (q )
(5)
where, qm= πfa2 / (V0z0+mV1h).
J0, J1 are cylindrical Bessel function, transducer radius ‘a’ (mm), water path z0(mm).V0, V1
ultrasonic wave velocity in water (m/sec) and specimen material (m/sec), frequency of probe
f,(Hz), thickness of the specimen is h (mm)
Reflection coefficient, Rtw=
Single-Sided Amplitude Spectrum
2.2.3 Backscattered Noise (FOM)
5
First back wall
Second back wall
4
|Y(f)|
where, z1, z2 are the acoustic
impedances of water and specimen
material, respectively.
As per Equation 3 and 4, the
amplitude of frequency spectra for
successive backwall reflections is
required. The amplitude of frequency
spectra is obtained by applying Fourier
transform using FFT on the ultrasonic
signal. Typical frequency spectra
indicating amplitudes of front and back
wall echoes is as shown in Figure 3.
3
2
1
0
0
0.5
1
1.5
2
2.5
3
Frequency (Hz)
3.5
4
4.5
x 10
5
7
Figure 3: Representative frequency spectra of
reflected echoes.
In polycrystalline materials, grains are randomly oriented with each other resulting in a
variation of ultrasonic velocity from grain to grain. Due to variation in velocity, localized
acoustic impedance mismatch is manifested as small reflections between the front wall and
back wall echoes. These peaks are usually termed as “noise or grass” and as shown in Figure
1(b). This noise response, also called as backscattered noise is an indicator about the localized
microstructure and is termed as Figure of merit (FOM) by Thompson [7].Ultrasonic
backscattered signal captured is a combination of material response as well as the electronic
noise generated due to amplification of the signal. In order to extract the true response of the
microstructure without the influence of electronic noise, backscattered signal are captured at
minimum 200 locations through the length and width of the sample.
Processing signals from 200 locations, the instrumentation background level or
electronic noise, b(t) can be estimated as
∑ V (t)
b(t) =
(6)
Subtracting this instrumental background noise from each signal, root mean square
(rms) noise can be minimized as,
∑ V (t) −
∑
n
=
V (t)
(7)
In order to normalize the effect of gain or amplification of the signal on the rms value
of noise, the nrms value is divided with E max, which is one -half of the peak-to-peak amplitude
of the reference signal.
Nrms =
(8)
Nrms relates the material properties and the experimental parameters by a factor called
as Figure of Merit. Relation among these factors is given by Margetan et al [7] and is as
follows,
where, R00=
N
, T01 =
= FOM ∗
(
√
(
(
)
))
∗
(9)
,ρ , ρ are densities of water and specimen, respectively
v , v velocities of ultrasonic wave in water and specimen, respectively, k1 = wave number of
ultrasonic wave = .
2.2.4 Probabilistic Grain Size Estimation
Estimation of SCEC
(b)
(b)
(a)
involves
probabilistic
estimation of grain size. This
could be achieved using a high
resolution micrograph or a
EBSD
image
indicating
distinct grain boundaries.
Figure 4 shows representative
EBSD unique grain mapping
images
of
copper
in
longitudinal direction and
aluminium
in
radial-1
direction. Skeletonized binary Figure 4: Unique grain map longitudinal (a), in Copper
image of the grain boundaries radial-1(b) directions in Aluminium
of microstructure as shown in Figure 5 is obtained from unique grain map (Figure 4),
indicating distinct interface of grains. Probability of a particular line falling within a single
grain is calculated for all the lengths of line segment placed randomly on grain boundary
binary image as shown in Figure 5. Probability of line segment within single grain follows an
exponential trend with the length of line segment. Exponential curve fitting is performed on
data plotted between the length of the line segment and its probability of presence within a
single grain as shown in Figure 6.
From exponential fitting of P (L) =e ⁄ , ‘b’ parameter can be estimated [8]. Mean grain size
diameter can be calculated as 2*b*scale factor.
Fitting plot of P(L)
1
Ploted data
Fitted data
0.9
0.8
0.7
0.6
P(L)
0.5
0.4
0.3
0.2
0.1
Figure 5: Placing random lines
on grain boundary binary image
0
0
100
200
300
400
Length of line segment
500
600
Figure 6: Fitting plot data with P(L)= Exp(-L/b)
700
3. Results and discussions
3.1. Microstructure of specimen
Estimated grain sizes of aluminium and copper are as shown in Table2. From Table 2,
it can be observed that aluminum specimen possesses coarser grains compared to that of
copper in all three orthogonal directions. Coarse grain structure can be attributed to the cast
structure of aluminum. In addition, the presence of twins observed in copper is attributed to
its forged condition. Moreover, the wide distribution of grains is observed in the copper
sample compared to that of aluminum.
Table 2: Estimated grain size in three orthogonal directions.
Direction
Longitudinal
Short transverse
Transverse
Copper
120 µm
80 µm
110 µm
Direction
Axial
Radial-1
Radial-2
Aluminium
340 µm
360 µm
350 µm
3.2 Ultrasonic wave velocity
Ultrasonic longitudinal velocity is measured using Equation 2. Ultrasonic velocity
measurements of aluminium and copper samples in three orthogonal directions are tabulated
in Table 3 and Table 4. From Table 3, it can be observed that the ultrasonic velocity in
Aluminum is same irrespective of the orthogonal direction or the procedure followed to
measure time difference due to its equiaxed grain structure. In the case of copper, ultrasonic
velocity is found to be dependent on the orientation of the sample due to forged structure.
Table 3: Ultrasonic longitudinal wave velocities propagating in three orthogonal directions of
the Aluminium sample.
Longitudinal wave velocities of ultrasonic wave in Aluminium (m/s)
Axial
Radial-1
Radial- 2
Velocity 1 Velocity 2 Velocity 1 Velocity 2 Velocity 1 Velocity 2
Average
6454±65
6461±62
6421±15
6400±25
6409±10
6409±14
Table 4: Ultrasonic longitudinal wave velocities propagating in three orthogonal directions of
the copper sample.
Average
Longitudinal wave velocities of ultrasonic wave in Copper (m/s)
Longitudinal
Short Transverse
Transverse
Velocity 1 Velocity 2 Velocity 1 Velocity 2 Velocity 1 Velocity 2
4831±125 4854±115 5012±149 4977±130 5005±131 5090±134
Ultrasonic shear wave velocity in aluminium and copper is measured in three
orthogonal directions in through transmission contact mode technique with a 5 MHz shear
wave probe on a 2mm thick specimen. Shear wave velocity values are as shown in Table 5.
Table 5: Ultrasonic shear wave velocities of Copper and Aluminium in three orthogonal
directions.
Shear wave velocities in
Shear wave velocities in Copper (m/s)
Average
Aluminium (m/s)
Axial
Radial-1
Radial-2
Longitudinal Short
Transverse
transverse
3157±33
3214±22
3130±22
2170±287
2249±6
2108±9
3.3 Ultrasonic Wave Attenuation Coefficient (α) & backscattered noise (FOM)
Ultrasonic attenuation coefficient values in this study are calculated using Equation 3
and 4 and are tabulated in Table 6. FOM measured using Equation 9 are tabulated in Table 7.
Table 6: Ultrasonic wave attenuation coefficient (α) (dB/mm) propagating in three orthogonal
directions of the Aluminium and Copper samples.
Average
Aluminium
Axial
0.2426±0.03
Radial-1 0.2360±0.03
Radial-2
0.201±0.04
Copper
Longitudinal
0.308±0.03
Short transverse
0.33±0.10
Transverse
0.144±0.02
Table 7: Ultrasonic wave backscattered noise (FOM) in Aluminium and Copper samples.
Direction
Nrms
0.077
0.093
0.099
Axial
Radial-1
Radial -2
Aluminium
Direction
FOM(1/mm1/2)
0.109
Longitudinal
0.133
Short transverse
0.14
Transverse
Nrms
0.296
0.086
0.258
Copper
FOM(1/mm1/2)
0.303
0.084
0.261
It can be observed that the attenuation coefficient and FOM values in Aluminum is
similar irrespective of the orthogonal direction of scanning, unlike copper. This can be
attributed to the equiaxed grain structure of Aluminum compared to copper. Copper
attenuation coefficient and FOM values measured in longitudinal direction match with that of
reported in the literature [9] and thus validating the procedure adopted in this study.
4. Estimation of Single Crystal Elastic Constants (SCEC)
Least square error between experimentally measured and analytically estimated
ultrasonic parameters is considered as minimization function (Ω). SCEC are estimated using
this minimizing function (Ω). Analytical estimation of ultrasonic parameters is in-turn a
function of single crystal elastic constants.
The minimising function when used is a combination of ultrasonic parameters such as
velocities and attenuation is called as “attenuation approach” and, minimisation function (Ω)
is defined as,
Ω
=
+
+
(10)
When velocities and backscattered noise is used in the minimizing function, it is called
as “noise approach”.
Ω
=
+
(11)
+
The terms
,
,
can be obtained from analytical equations as defined in published
literature [10], [11], [12].
Minimization is performed by following the elastic stability criteria as,
+2
0;
0;
−
0
Extraction of global minima required for obtaining the minimum value of functions listed in
Equations 10, 11 is performed with the help of genetic algorithm (‘GA’) available in Matlab
optimization toolbox. Over successive generations, the population evolves toward an optimal
solution. Stopping criteria for minimizing function is defined in such a way that the function
fitness value reaches 1e-3 irrespective of a number of generations and time limit. Figure 7,
shows the function value obtained by iterating single crystal elastic constants within the lower
and upper bounds for fixed non-linear constraints.
GA toolbox gives the estimated SCEC on
reaching the stopping criteria. In order to
validate repeatability of the estimated SCEC
with every approach, iterations are repeated
consecutively for 10 times in the optimisation
toolbox as shown in Figure 8.The average
values of SCEC of copper and aluminium
with standard deviation are shown in Table 8.
It can be observed that irrespective of the
procedure followed the standard deviation in
estimated SCEC values is less than 10% of
average value. Figure shows the plot of
average values of SCEC of copper and Figure 7: Representative plot of function
aluminium. From Figure 9, it can be observed value to number of generations obtained
that average SCEC of aluminium by two from genetic algorithm toolbox
approaches has less scatter compared to that
of copper due to equiaxed grain structure of aluminium.
Table 8: Averaged values of SCEC of every approach for Copper and Aluminium
Average SCEC of Copper from all
Average SCEC of Aluminium from
runs for each approach in GPa
all runs for each approach in GPa
Approach
C11(168.4)
C12(121.4)
C44(75.4)
C11(108.3)
148.45±13
107.57±6.8 73.43±2.1
99.97±5.4
Attenuation
163.45±5.8 118.60±4.7 77.52±3.7 104.44±2.1
approach
Noise
approach
C12(61.3)
C44(28.5)
58.76±3.9
29.75±1.3
56.33±4.9
28.02±3.8
Copper
Aluminium
Figure 8: Graphical representation of estimated elastic constants of Copper and
Aluminium by (a) attenuation approach (b) Noise approach.
Estimated SCEC for Copper
120
100
80
60
C11
C12
C44
Attenuation approach
Noise approach
100
Elastic constants in GPa
Elastic constants in GPa
140
Estimated SCEC for Aluminium
110
Attenuation approach
Noise approach
160
90
80
70
60
50
40
30
20
C11
C12
C44
Figure 9: Average SCEC estimated using three approaches for Copper and Aluminium
5. Conclusions
Knowledge of single crystal elastic constants (SCEC) would enable the estimation of
elastic moduli of materials. However, due to the complexity involved in estimating SCEC’s,
such as requirement of single crystal grown in particular direction, relation between SCEC
and elastic moduli is of limited use. Optimum advantage of relation between SCEC and
elastic moduli could be availed with enhanced ease in estimation of SCEC’s. In this study,
ultrasonic testing has been used effectively to estimate SCEC from polycrystalline materials.
By minimizing the difference between experimentally measured and analytically estimated
ultrasonic parameters such as velocity, attenuation coefficient and backscattered noise,
SCEC’s are estimated. Polycrystalline pure copper in forged condition and pure aluminium in
ingot condition are used to measure ultrasonic parameters in immersion testing. Using the
relationship between ultrasonic parameters and single crystal elastic constants, a minimization
function has been developed. Grain size which is an important factor in analytical estimation
of figure of merit is estimated probabilistically. SCEC’s are estimated by minimizing the
function (combinations of velocity, attenuation coefficient and velocity, figure of merit).
Estimated SCEC are found to be within 10% of those SCEC reported in literature. Compared
to aluminium, SCEC values in copper are found to have scatter due to the presence of twins
and unequiaxed grains thus resulting in scattered estimation of grain size.
6. Acknowledgements
The authors express their gratitude to Dr. S. V. Kamat, Director, DMRL for the
encouragement provided to publish this work. The funding provided by Defence Research and
Development Organization (DRDO) to carry out the work is acknowledged.
7. References
1. George. E. Dieter, Mechanical Metallurgy, Mc.Graw-Hill book company chapter 2, 17-62,
1988.
2. Tokuteru Uesugi, Yorinobu Takigawa and Kenji Higashi, “Elastic Constants of Al Li”,
Materials Transactions, Vol. 46, No. 6, 1117- 1121, 2005
3. Y.W. Zhang, S.J. Li, E.G. Obbard, H. Wang, ‘Elastic properties of Ti–24Nb–4Zr–8Sn
single crystals with bcc crystal structure’, Acta Materialia, 59, 3081-3090, 2011.
4. D.J.Safarik, R.B.Schwarz, ‘Elastic constants of amorphous and single –crystal
Pd40Cu40P20’, Acta materialia, 55, 736-746, 2007.
5. Dipen K. Patel, Hamad F. Al-Harbi, Surya R. Kalidindi, “Extracting single-crystal elastic
constants from polycrystalline samples using spherical nanoindentation and orientation
measurements”, Acta materialia, 79, 108-116 ,2014.
6. A. Bhattacharjee, A.L. Pilchak, O.I. Lobkis, J.W. Foltz, S.I. Rokhlin, J.C. Williams,
“Correlating Ultrasonic Attenuation and Microtexture in a Near-Alpha Titanium Alloy,”
Metallurgical & Materials Transactions A, vol 42A, 2358-2372, 2011.
7. P. Haldipur, F. J. Margetan and R. B. Thompson, “Estimation Of Single-Crystal Elastic
Constants Of Polycrystalline Materials From Back-Scattered Grain Noise”, Review of
Quantitative Non-destructive Evaluation Vol. 25, 1133-1140, 2006.
8. F.Margetran, R.B.Thomson, “Detectability of small flaws in advanced engine alloys”,
CNDE, Iowa State University, 1993.
9. T. Stepinski and P. Wu, “Evaluation of Ultrasonic Attenuation and Estimation of
Ultrasonic Grain Noise in Copper”, AIP Conf. Proc. 497, 431-436,1999;
10. J.M.J.Toonder, J.A.W. Van Dommelen, “The relation between single crystal elasticity and
the effective elastic behaviour of polycrystalline materials: theory, measurement and
computation”, Modelling Simulation in Material Science and Engineering, Vol.7, No.6,
909, 1999.
11. H. Wasan, F. J. Margetan, “Backscattered Ultrasonic Noise Measurements in Jet-Engine
Nickel Alloys”, AIP Conf. Proc. 557, 1314, 2001.
12. Fred E. Stanke and G. S. Kino, “A unified theory for elastic wave propagation in
polycrystalline materials”, Journal of Acoustic Society of America, 75, 665, 1984.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Detection of material inhomogeneity using inverse time domain spectral
finite element method
Raghavendra B Kulkarni1, S Gopalakrishnan2 and Manish Trikha1
ISRO satellite centre, Bengaluru, India,
Phone:25084359; e-mail:ragm@isac.gov.in, mtrikha@isac.gov.in
2
Department of Aerospace engineering, Indian institute of science, Bengaluru, India; E-mail:
krishnan@aero.iisc.ernet.in
1.
Abstract
During the fabrication of a structure, there may be imperfections or flaws in terms of material in homogeneity
like blow holes in case of casting, flaws, impurity in material etc. Hence it becomes very important to identify
these imperfections as it affects the overall strength of the material. There are different methods for identifying
the material inhomogeneity analytically. However, the time domain spectral element method offers a very good
solution for such problems. This paper illustrates the method on a Timoshenko beam having a lumped mass as
imperfection which is taken as an example. A time domain spectral finite element method using an eight noded
element having Gauss-Lobatto Legendre (GLL) points as nodes are formulated using Timoshenko beam theory.
The mass, damping and stiffness matrices are formulated. Then this is solved using Newmark time integration
technique. When an input pulse is given to the beam, the pulse is not distorted and an undistorted response is felt
at the far end of the beam. When a lumped mass is added which simulates the material inhomogeneity, we get a
distorted response. Knowing the distorted pulse, we can predict the location and the mass of the lumped mass.
This illustrates the technique for inverse problem for detecting the material inhomogeneity
Keywords: Spectral element; time domain; material inhomogeneity
1. Introduction
In many applications, there may be imperfections in the material itself such as blowholes in
casting or a defect during the fabrication of the material. This in many cases has an impact on
the intended performance as against the design. Hence it becomes very important that one
needs to identify these flaws and its location so that it is within tolerable limits as it is
impossible to have a defect free material.
Recently many researchers like [1], worked on the defect detection using longitudinal waves
by using the fact that the apparent sound velocity decreases owing to the diffraction at the
defects. Many NDT techniques are also available for this kind of identification like X-ray is
in practice. However, there are many disadvantages like the penetration depth and the human
safety issues are of importance. Out of these, the method using the physics of wave
propagation shows promise. These kinds of problems to identify the flaws in the material are
also called as inverse problems in contrast with the forward problem where the input forces
as well as the material properties are known and responses are predicted. In this paper, time
domain spectral element method is used to identify the material in homogeneity or flaws in a
Timoshenko beam. This is demonstrated in this paper by using a lumped mass to represent
the flaw at the centre of a homogeneous aluminium beam and waves are generated by
exciting the tip of the beam.
2. Spectral finite element method
There are many methods to determine the change in the material properties using frequency
domain spectral methods as shown by [2].
When a high frequency wave is generated for example using an instrumented impact
hammer, elastic waves are generated and these waves travel across the structure and in case
of any foreign material, flaw or change in cross section, it gets reflected and diffracted and
the instrument at the far end of the structure captures this distorted signal. If we know the
response at that location for a structure without any anomalies, we can compare it with the
structure having in homogeneity. The shape of the response gives the indication of the size
and distance of the embedded anomaly. To do this, the spectral element method suits best.
For a high frequency response, the inverse problem becomes very difficult using
conventional finite element method as the element size should be as small in comparison with
the wave length of the frequency under consideration. Hence alternate methods which are
accurate and computationally effective should be used. One of the methods used is the
spectral finite element method. There are two different kinds of spectral finite element
approaches, one is based on frequency approach where fast Fourier transform (FFT) [3] is
used and another is based on time domain where orthogonal functions like Legendre and
Chebyshev is used along with discretisation typical for the FEM [4] This results in its high
accuracy and excellent convergence properties. This method uses these polynomials, the
nodes are placed not equidistant and these functions are orthogonal leading to the inertia
matrix to be diagonal and computationally easy and accurate. The FFT methods provide
excellent convergence and using single element, one can easily calculate the response.
However, this method requires the use of throw off elements due to the periodic nature of
FFT. For complex geometries, this becomes more complicated. Hence some other methods
using Laplace transforms are also described in literature [5]. However all these work well for
simple geometries. In this paper, time domain approach is followed where complex
geometries can also be solved taking advantage of finite element techniques.
2.1 Element formulation
As an illustration of the method, a spectrally formulated eight noded Timoshenko beam with
Gauss Lobatto Legendre points as nodes are used. The reason for this number is that this
provides good accuracy [6]. The stiffness, mass and damping matrices are derived using
Gauss Lobatto Legendre weighing functions. The individual matrices are assembled and the
resulting equation containing stiffness matrix along with mass and damping matrices are then
solved using time integration methods. Here Newmark integration scheme is used as it gives
good stability while solving. Also, the imperfection in the structure here is modelled as a
lumped mass.
2.2 Displacement fields
Displacement fields ( , ), ( , ), ( , ) for a Timoshenko beam which considers the
shear deformation with the central axis as x axis and along the thickness of the beam as z.
Also in this theory, it is assumed that plane section remains plane but may not be normal. The
total rotations of the plane originally normal to the neutral axis of the beam is given by the
rotation of the tangent to the neutral axis and shear deformation, . Also, the initial axial
displacements aregiven by ( ), ( ) and is given in equation (a).
( )=
( , )=
−
( )−
( )
(a)
( , )=0
( , )=
( )
The Strain displacement relations are obtained by differentiating equation (a) which are
shown in equation (b)
( , )
=
=
( )
( , )=
( )
−
( , )
+
=
+
( , )
=− ( )+
(b)
( )
2.3 Steps used in formulating the spectral finite element
2.3.1 Step1: Placement of nodes
(1 − )
′
( )=0
(1)
Here ξ ∈ [1, −1]are the roots of the Legendre polynomial. P′ (ξ) denotes the first derivative
of the Legendre polynomial of degree N. For example if we consider an element having eight
nodes, the corresponding Legendre polynomial will be P and this can be got by using
recurrence relations and is given by equation (2). Solving equation (1) we get the GLL points.
1
(2)
(429 − 693 + 315 − 35 )
=
16
The node points when distributed according to GLL points give very high interpolation
accuracy [6]
2.3.2 Step2: Interpolation
Displacement field approximation for the node placement according to equation (1) is carried
out based on the Lagrange interpolation function equation (3) which gives the linear
interpolation.
( − )( − ) … ( − )
(3)
=
( − )( − ) … ( − )
Same interpolation field is used for both displacement and rotation field in case of a beam.
For an eight noded element, we have from equation (4) the linear combination of the shape
functions.
=
(4)
2.3.3 Step 3: Strain displacement relationship
B matrix formulations which is differentiation of displacement field as derived from equation
(b). Assuming linear systems is shown in equation (5)
(5)
=
=
Where
is the strain displacement matrix.
2.3.4 Step 4: Matrix derivation
The mass, damping, stiffness and force matrices is derived using the numerical integration
using weighted residual techniques. Here Lobatto integration weights wi are used to evaluate
the integral numerically. ρ and D are the density and elasticity matrix respectively. These are
given by the equation (6-9). In the mass matrix, the lumped mass, mii which depicts the
addition of mass due to flaw is added in this case. Whereas for a beam having no anomaly
this term is absent
=
=
2
( − 1)[
( )]
( )
+
( )
≈
( )
( ) det ( ) +
=
( )
≈
=
≈
( )det( ( ))
(6)
(7)
(8)
( ) ( )
( ) ( )det( ( ))
(9)
2.3.5 Step 5: Solving the differential equation
The resulting differential equation is solved using Newmark integration. This implicit
integration method is used as it gives good stability and accuracy
3. Numerical experiment
Two cases are studied in this numerical experiment
3.1 Case-1: A beam without flaw
A cantilever beam having of aluminium material which is 5m long so that the wave
reflections are kept far away from the original pulse is chosen. The beam has a depth of 5mm
and width of 50mm. This beam is given a triangular pulse of 0.5ms width at the tip of the
cantilever beam as shown in Figure-1. The beam is modelled as having 8 spectral elements.
The corresponding response at the tip of the beam is shown in Figure-2
Triangular pulse with zero padding
1
0.9
0.8
force in newton
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
5
10
time (milliseconds)
15
20
25
20
25
Figure-1 Force time history for a beam
5
Response of Timoshenko beam for a lateral impulse load
-3
x 10
4
Velocity in m/s
3
2
1
0
-1
-2
0
5
10
Time in ms
15
Figure-2 Response of the beam at the tip without flaw
3.2 Case-2: A beam with flaw
A cantilever beam of the same dimension as in case-1 but with a centre mass representing a
flaw in the beam is shown in Figure-3
Figure-3 Beam with tip transient force and material imperfection at the centre of beam
This cantilever beam is excited by the same triangular input pulse as in Figure-1 and the
response studied. Figure-4 shows the response at the tip of the beam due to the force at the tip
of the beam which is having a centre mass simulating the flaw.
5
-3
x 10
Response of Timoshenko beam with center mass for a lateral impulse load
4
Velocity in m/s
3
2
1
0
-1
-2
0
5
10
Time in ms
15
20
Figure-4 Response at the tip of a beam with flaw at the center of the beam
From the response of Figure-4, it is clear that the reflections in case-2 starts early due to the
presence of the imperfection which is located at the centre of the beam. If this imperfection is
moved towards the end of the beam, the response will tend towards case-1.
Comparison with case-1 and case-2 is plotted in Figure-5. If the response of the beam with
flaw is used as input, the difference in the original mass matrix and the new mass matrix
gives the exact location and the mass of the flaw. In this case at the centre of the beam.
25
5
x 10
Comparison of a Timoshenko beam with and without defect
-3
without defect
With defect
4
Velocity in m/s
3
2
1
0
-1
-2
0
5
10
Time in ms
15
20
25
Figure-5 Comparison of responses at the tip of the beam with and without flaw at the center of beam
4. Conclusion
The time domain spectral finite element gives a very good indication of the flaw with fewer
numbers of elements. If we measure a response at a particular location and this is compared
with the beam without flaw, the difference in the mass matrices gives the exact location and
mass of the flaw.
References
1. Hideto Mitsui, Koichi Mizutani and Naoto Wakatsuki, ‘Defect detection in square billet
Using Time of flight of longitudinal waves’, Japanese Journal of applied physics, vol49,
number 7S, July 2010.
2. James F. Doyle, ‘Wave propagation in structures, Spectral analysis using fast discrete
fourier transforms’, second edition,1997.
3. M.T Martin and J.F. Doyle, ‘Impact force identification from wave propagation
responses’, Int. J. Impact Engg Vol.18.No.1,pp.65-77,1996.
4. P.Kudela, M.Krawczuk, Ostachowicz, ‘Wave propagation modelling in 1D structures
using spectral finite elements’, Journal of sound and vibration 300, pp 88-100, 2007
5. H. Igawa, K. Komasu, I. Yamaguchi, T. Kasai, ‘Wave propagation analysis of frame
structures using the spectral element method’, Journal of sound and vibration 277,
pp.1071-1081,2004.
6. C. Pozrikidis, ‘Introduction to Finite and Spectral Element Methods using MATLAB’
second edition, CRC press A Chapman and hall book 154, 2014
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Evaluation of elastic modulus of metals using Acoustic Emission technique
R. Gopikrishna1,a, M. Padma Amani1, M. Varadanam1
1
Defence Research Development Laboratory, Hyderabad, India.
a
Phone: +914024583516, e-mail: tisipog557@gmail.com
Abstract
Elastic modulus is an important mechanical property of a material. In order to determine this property, mostly
tensile load test is carried out on samples. When it comes to determining characteristics in-situ, existing
techniques like IR-guns can give only material compositions but not the mechanical parameters. In this paper, an
alternative non-destructive, Acoustic Emission (AE) technique is proposed to be used as a tool to evaluate the
elastic modulus of any given isotropic material. AE works on the principle of release and transmission of
acoustic waves within the material due to the growth of active flaws. Correlation in 1-D and 3-D scheme of
measurements exists relating the wave velocity to the Young’s modulus and density of the material. A pencil
lead break over the surface of the material and measurement of the time delay in the acquisition of signal by two
acoustic sensors, enable the computation of the wave velocity. In the present paper, study is made to establish the
method and results compared to 1-D and 3-D correlations for Young’s modulus, of known metals. This
technique will extensively help in evaluating elastic modulus using a non-destructive approach and also in
providing an instant measurement of the elastic modulus without the need for fabrication of specimens for tensile
tests. Studies are also carried out to determine the necessary input parameters to obtain consistent velocity
measurements on different metal plates.
Keywords: Elastic modulus, Wave velocity, Acoustic emission, Non-destructive technique
1. Introduction
Material characterisation is an important field of engineering. The properties of the material
define the strength of the structure with which it is made of. Over the many years of
engineering excellence, a variety of materials have been characterised using standard
techniques. ASTM E8 [1] defines the methodology to carry out tensile testing on material
samples to obtain necessary parameters of the material under study. Among the material
properties, Elastic modulus (popularly known as Young’s modulus) is a primary parameter
used to calculate the stiffness of any structure. The static and dynamic deformations of the
structure are functions of the elastic modulus of the material used. Although a huge amount
of experiments have been done to determine the elastic modulus of different metals, it is a
standard practise to carry out sample level testing to estimate the elastic modulus of a
particular material specimen before its use. Sample specimens need to be fabricated every
time a material has to be characterised. This paper proposes a new approach of estimating the
elastic modulus of a material in-situ, without the need to manufacture specimens to carry out
material tests.
Sound wave travels at different velocities in different medium. The velocity of sound in
solids, liquids and gases are different from each other. Furthermore, in solids, the velocity of
sound varies from metal to metal based on the material properties. This attribute has been
exploited from time immemorial to determine the material characteristics, if we have
knowledge of the wave velocity in that material. The velocity of sound waves in different
metals can be found using a variety of techniques. Database of sound velocities in different
metals is available in literature. However, the velocity is a function of the material grain
structure, presence of defects, purity of the metal etc. and hence it is better to measure the
wave velocity for every individual specimen. Acoustic emission technique is a powerful tool
used for structural health monitoring. This tool can also be employed to determine the wave
velocity in different metals. Using the wave velocity, the elastic modulus is obtained as a
derived quantity, provided the density of the material is known. The same formulation can be
used to determine the density of a given material, with prior knowledge of the elastic
modulus.
2. Acoustic Emission Technique & Terminologies
Acoustic emission is defined as a transient elastic wave generated from within the material
though rapid release of energy. These elastic waves are nothing but sound waves travelling
within the solid material. The technique developed to capture these elastic waves, analyse and
interpret them is called Acoustic Emission Technique. It is a non-destructive technique which
has been effectively used in fracture mechanics and material characterisation. Piezo-electric
sensors are employed to convert the elastic waves into electrical signal for further processing.
The common terminologies used in this technique, relevant to the current work are given
below [2].
Channel: Each Acoustic emission sensor and related equipments for transmitting,
conditioning, detecting and measuring signals that come from it.
Hit: The process of detecting and measuring an AE signal on a channel. When an elastic
wave is sensed by a channel, it is called a hit.
Activation: The onset of AE due to the application of a stimulus such as a force, pressure,
heat etc.
Attenuation: Loss of amplitude with distance as the wave travels through the structure.
Event: An elastic wave when detected by two or more channels is considered an event.
Source: The physical origin of one or more AE events
Time of Hit: The time at which a hit has been detected by the sensor. This is detected at the
instant the AE signal exceeds the AE threshold.
3. Nature of Acoustic Wave & Simulation of Artificial Source
Acoustic wave is a type of longitudinal wave. The direction of vibration is same as that of the
direction of travel. They are basically pressure waves, which experience compressions and
rarefactions along the length of propagation. Characterising the acoustic wave as such, is not
easy as the wave propagation depends on the nature of the material in which it travels. This is
the concern with the acoustic wave generated within the material which is under loading.
However, for velocity measurement studies, the source of the acoustic wave is artificially
given i.e. externally.
Acoustic emission is called a pseudo-non-destructive technique mainly because the structure
needs to be excited or stressed to activate the flaws in it, which release energy in the form of
waves. The elastic wave can also be artificially generated using the standard Pencil Lead
Fracture (PLF) technique [3]. This is a commonly used technique to easily reproduce the
same signal in different materials and different specimens. A 2H pencil with diameter 0.3mm
and protruded length of 3 mm is used to carry out PLF. A teflon guide, popularly known as
Hsu-Nielson shoe is used while carrying out the pencil break. The fracture is carried out by
keeping the pencil lead at 30° to the surface of the structure. Double break is avoided using
this teflon guide. A typical PLF signal recorded by the AE monitoring system is shown in
Figure-1.
Figure-1 Typical AE signal generated by Pencil Lead Fracture
In order to estimate the velocity, there is a requirement to generate an artificial source for the
acoustic sensor to pick up. The elastic wave thus generated, behaves more like a flexural
dispersive wave. However, the current work considers the assumption that the generated
wave through PLF is a longitudinal acoustic wave.
4. Relationship between Velocity & Elastic modulus
The velocity of an acoustic wave in a solid is a function of the elastic modulus, density and
temperature of the material. When an acoustic wave is generated, the elastic modulus defines
the force applied by the wave on the particles of the solid, while the density defines the
acceleration experienced by the particles on being subjected to this force. In 1-D solids, where
the wavelength of propagation is greater than the diameter of the solid, the relation between
the longitudinal velocity cL and the elastic modulus E is given as [4]
𝑬
𝒄𝑳 = √𝝆
(1)
where ρ is the density of the solid. In 3-D solids, the effect of poisson's ratio ν comes into
consideration, as the velocity of the wave becomes a function of the poisson's ratio also [4].
This is because the acoustic wave causes the dynamic change of the lateral dimension of the
solid, with respect to position of the source in the solid.
𝑬(𝟏−𝝂)
𝒄𝑳 = √𝝆(𝟏+𝝂)(𝟏−𝟐𝝂)
(2)
In this research work, the mathematical model for wave motion in rod is considered to
determine the elastic modulus i.e. the wave travels only as a longitudinal wave. The
attenuation in the amplitude of the signal picked up between the sensors is less than 1 dB and
therefore, this approximation of longitudinal wave holds good for the present work for both
the bar and plates.
5. Experimental set-up
The velocity-modulus correlation given in Eqn. (1) is valid for 1-D rods where the mode of
wave propagation is only a longitudinal wave and the diameter or thickness is less than the
wavelength of the wave. Experiments were conducted on rod specimens of different
dimensions and material. The specifications of the samples are Maraging steel square rod of
dimension 7 x 7 x 544 mm; Mild Steel rod of 70 mm diameter, 750 mm long; Mild steel
hexagonal rod of side 55 mm and length 540 mm. Figure-2 shows photograph of the set-up
for Maraging steel. Two piezo-electric AE sensors were positioned at the centre of the
specimens with a sensor spacing of 100 mm. The AE sensor used has a resonant frequency of
150 kHz. A threshold amplitude of 45 dB was pre-set to the AE system in order to filter out
the background noise, if any.
Figure-2 Experimental set-up of Maraging steel square rod
Experiments were also carried out on thin sheets and plates to verify the effectiveness of this
technique to determine elastic modulii of 2-D structures. Aluminium sheet of 3 mm thickness
was the first sample used. Two narrow bandwidth piezo-electric sensors were positioned at a
distance of 300 mm at the centre of the plate. Figure-3 shows a photograph of the
experimental set-up. The sheet was positioned on two wooden planks rested on a table.
Figure-3 Experimental set-up for the Aluminium sheet
The same experiment was carried out using a Titanium alloy plate (Ti6Al4V) and a pure
Titanium plate, both of 5 mm thickness. Sensor spacing maintained was 100 mm in this case.
Figure-4 shows the experimental set-up for the Titanium plate.
Figure-4 Experimental set-up for Titanium plate
6. Precautions during experimentation
i.
ii.
iii.
iv.
A teflon shoe is must while carrying out PLF. This shoe will ensure same length of
lead being broken and also avoid double break.
The specimen should not be resting directly on any solid surface during the test.
PLF should be carried out exactly in the line of the sensors. The velocity calculation
formula becomes invalid if there is any deviation in position of the simulated source
from the sensor line.
PLF shall be carried out outside the sensors and not between them.
7. Results & Discussion
Pencil Lead Fracture was carried out on each specimen at a minimum distance of 50 mm from
any one sensor, in the same line, outside the sensors. When a PLF is carried out, the elastic
wave is detected by each sensor as an individual hit. However, if the same wave is detected by
both the sensors, an event is formed. The amplitude of the detected signals by each sensor are
of the order of 90 dB. The velocity of the acoustic wave is calculated from the distance
between the sensors (d) and difference between the time of hit (Δt) between the sensors.
𝑐=
𝑑
𝛥𝑡
(3)
An assumption in this study is that the obtained velocity from the AE measurement is the
longitudinal velocity of the wave. As mentioned in the previous paragraphs, the sensor
spacing for the rods was 100 mm, aluminium sheet was 300 mm and for that of Titanium
plates was 100 mm.
The results obtained on the Maraging steel rod are given in Table-1. The density and
theoretical elastic modulus of the Maraging steel specimen were estimated using the standard
tensile test and found to be 8100 kg/m3 and 182.7 GPa respectively. 1-D rod formulation for
estimating elastic modulus is used and tabulated in Table-1.
Table-1 Experimental results obtained on Maraging steel rod
Trial
Δt
Velocity, c
Elastic modulus
No.
(in micro seconds)
(in m/s)
(GPa)
1
0.0000223
4484.30
162.88
2
0.0000226
4424.78
158.59
3
0.0000211
4739.34
181.94
4
0.0000226
4424.78
158.59
5
0.0000213
4694.84
178.54
% Error
10.85
13.20
0.42
13.20
2.28
Interestingly, the velocities obtained using the same procedure on mild steel rods, which were
of higher diameter yielded lesser velocities, of the order of 2800 m/s. The longitudinal
velocity of steel is usually of the order of 5000 m/s [5]. However, the surface velocity of steel
is around 2870 m/s. It was observed that in solid rods of dimensions very large as compared
to the wavelength of the acoustic wave, PLF generates a surface acoustic wave only and not a
longitudinal wave travelling through the material.
The material properties of the plate samples are given in Table-2. From the velocity
measured, the elastic modulus is calculated using both 1-D and 3-D formulations as given in
Eqn (1) and Eqn (2). The density and poisson's ratio for the above materials were evaluated
experimentally. The results obtained on the plates are given in Table-3 to Table-5.
Table-2 Density & poisson's ratio experimentally evaluated
Material
Density (kg/m3)
Aluminium alloy
2700
Ti6Al4V
4430
Pure Titanium alloy
4430
Poisson's ratio
0.33
0.31
0.33
Table-3 Velocity measurement on Aluminium sheet (sensor spacing = 300 mm)
Elastic modulus
Elastic modulus
Trial
Δt
Velocity, c
(1-D formulation)
(3-D formulation)
No.
(seconds)
(m/s)
GPa
GPa
1
0.0000567
5291.0
75.6
51.0
2
0.0000573
5235.6
74.0
50.0
3
0.0000573
5235.6
74.0
50.0
4
0.0000575
5217.4
73.5
49.6
5
0.0000568
5281.7
75.3
50.8
6
0.0000595
5042.0
68.6
46.3
7
0.0000575
5217.4
73.5
49.6
Table-2 Velocity measurement on Ti6Al4V plate (sensor spacing = 100 mm)
Elastic modulus
Elastic modulus
Trial
Δt
Velocity, c
(1-D formulation)
(3-D formulation)
No.
(seconds)
(m/s)
GPa
GPa
1
0.0000200
5000.0
111
82.3
2
0.0000207
4830.9
103
76.8
3
0.0000202
4950.5
109
80.7
4
0.0000200
5000.0
111
82.3
5
0.0000202
4950.5
109
80.7
6
0.0000202
4950.5
109
80.7
Table-3 Velocity measurement on pure Titanium plate (sensor spacing = 100 mm)
Elastic modulus
Elastic modulus
Trial
Δt
Velocity, c
(1-D formulation)
(3-D formulation)
No.
(seconds)
(m/s)
GPa
GPa
1
0.0000203
4926.1
108
79.9
2
0.0000198
5050.5
113
83.9
3
0.0000192
5208.3
120
89.3
4
0.0000203
4926.1
108
79.9
5
0.0000195
5128.2
117
86.5
6
0.0000190
5263.2
123
91.2
The elastic modulii of all the three materials were experimentally estimated by carrying out
standard tensile tests. The elastic modulii of aluminium alloy, Ti6Al4V and pure titanium
were found to be 70 GPa, 110 GPa and 110 GPa respectively. The percentage errors in the
average values for 1-D and 3-D formulation results with respect to experimental values are
tabulated in Table-8.
Table-8 Accuracy of results
Aluminium
Ti6Al4V
Pure titanium
1-D formulation
Average E, GPa
% Error
73.5
5.0
108.0
1.8
115.0
4.6
3-D formulation
Average E, GPa
% Error
49.6
29.1
80.5
26.8
85.1
22.6
From the above results, it is observed that 1-D formulation gives better estimate of the elastic
modulus than the 3-D formulation. The error in the estimated elastic modulus is less than 5%,
which is admissible to get an approximate estimate of the property. The accuracy of this result
can be improved by reconsidering the longitudinal behaviour of the simulated elastic wave
and also by optimising the sensor spacing.
Effect of sensor spacing
Additionally, a study was carried out to see the effect of sensor spacing on the accuracy of the
results. This study was carried out only on the Aluminium sheet. Sensor spacing was varied
between 100 mm to 400 mm and 7 trials were carried out for each sensor spacing. The
velocities estimated for different sensor spacing are shown in Figure-5. From the plot, it is
observed that the consistency of results for 100 mm sensor spacing is not as good as the
results from 300 or 400 mm sensor spacing. Therefore, a minimum sensor spacing need to be
maintained for the given structure, in order to get accurate results for velocity and elastic
modulus too.
Figure-5 Effect of sensor spacing on velocity measurement
8. Future scope of work
In the current research, available specimens of Maraging steel, Mild steel, Aluminium and
Titanium alloys were used to evaluate their elastic modulii. This can be extended to any
metallic material. The wave nature of the acoustic wave may be studied in detail and a more
appropriate lamb wave technique may be adopted to estimate the elastic modulus [6]. In plate
specimens, the governing wave equation for flexure of plates can be considered to obtain
more accurate experimental results.
8. Conclusion
The non-destructive approach to evaluate elastic modulus is detailed elaborately. This can be
used to exclude the need to destroy a material and make samples to carry out standard
material testing. The maximum error in the elastic modulus estimation on rods is about 13%
in this work. However, this error can be reduced further by carrying out sensor spacing studies
on rods of different dimensions and also by optimisation of minimum distance between the
artificial source and the sensor. The error in the experimental results on plates is about 5%
using 1-D formulation. Sensor spacing plays a vital role to obtain consistent results in both
rods and plates. Batter modelling of the acoustic wave and its relation with the elastic
modulus can yield more accurate results.
Acknowledgements
The authors would like to acknowledge Mr. Chandu Gopi, Engineer, DRDL for his
contributions in carrying out the experiments and handling the acquisition system. We also
thank Shri. S Narendar, Scientist, DRDL, for his valuable suggestions during the research.
References
[1] ASTM Standard E8/E8M – 16a
[2] ASTM Standard E569/E569M – 13
[3] ASTM Standard E976-15
[4] Albert E Brown, Rationale and Summary of Methods for Determining Ultrasonic Property
of Materials, NDT.net March 1996, Vol 1, No 03
[5] www.engineeringtoolbox.com, www.hyperphysics.com
[6] K. Vignesh, V. Malolan, M. Premakumar, A. S . Srinivasa Gopal, “Non-destructive
evaluation of elastic modulus in metals using Lamb wave technique”, e-Journal of NDT, Vol.
20, No 06, 2015.
Non-destructive Evaluation of Friction Stir Weld Defects in Aluminium Alloy
V D Ragupathy1, M R Bhat2 and M V N Prasad1
Liquid Propulsion Systems Centre, Indian Space Research Organisation
2
Department of Aerospace Engineering, Indian Institute of Science, Bangalore, India
Contact: ragupathy_v_d@rediffmail.com, mrb@aero.iisc.ernet.in
1
AA2219 Aluminum alloy is extensively used for fabrication of aerospace components. This material
is weldable compared to other alloys in the same series. Friction stir welding (FSW) is a solid state
metal joining process being introduced to weld the AA2219 material. The defects and anomalies
encountered in this type of welded joint differ from that of conventional fusion weld. The defects
present in the FSW are oriented in different direction due to the unique formation process involved
and are observed to be dispersed over the welded region width wise as well as thickness wise. Typical
defects such as void, tunnel defect, incomplete penetration, forging and crack that are encountered
during the friction stir welding process were intentionally induced and then were subjected to
different types of non-destructive evaluation. The methods adopted were penetrant testing,
radiography testing, ultrasonic testing and eddy current testing. The outcome of these experimental
investigations in terms of signal response obtained from the defects; to different NDE methods are
presented.
Detection of Blocked Cooling Holes in a Gas Turbine Nozzle using Infrared
Imaging
Joel Jon Bosco, Dheepa Srinivasan, Debabrata Mukhppadhyay and *Paul Dimascio
GE Power, GE India Technology Centre,
Whitefield, Bangalore, India
*GE Power, Greenville, South Carolina, USA
Contact: Dheepa.Srinivasan@ge.com
Gas turbine components such as blades and nozzles undergo repair and refurbishment after service
exposure. The starting step in a repair process involves grit blasting the component during incoming
inspection, in order to be able to clean the surface oxides and other contaminants as a result of gas
turbine operation. Grit blasting is carried out using Al2O3 particles (180-220 mesh) to enable a surface
that is now cleaned up and ready for welding or coating and other related repair operations. Many a
time the grit blasting results in the Al2O3 particles inadvertently getting entrapped in the component
cooling holes. If left undetected, this could end up being catastrophic as the part goes into service,
since the blocked cooling holes will end up raising the part temperature in that particular location,
leading to damage of the whole part and/or the system. A novel and simple method of identifying the
blocked cooling holes in a turbine component was carried out using the infrared (IR) camera. The
process comprised flowing compressed air into the nozzle and imaging the convective fluid flow on
wet tissue paper using the IR camera. The paper is a poor conductor of heat and acts as a membrane
that effectively differentiates blocked verses unblocked holes very effectively, by the rate at which the
paper dries in specific locations. A careful optimization of the pallet in the IR camera, results in a
quick and easy method of precisely detecting the blocked cooling holes.
Studies on Multi Site Damages in Riveted Joints under Fatigue Cycles with
Acoustic Emission Approximate Entropy Approach
S Kalyana Sundaram1, V R Ranganath1, M R Bhat2
1
Structural Technologies Division, CSIR-NAL, Bangalore, India
2
Department of Aerospace Engineering, IISC, Bangalore, India
Contact: kalyan@nal.res.in
Occurrence of multi-site damage (MSD) and multi element damage (MED) due to fatigue are very
common on riveted joints extensively used in aircraft structural construction. MSD starts with
formation of crack from rivet holes simultaneously due to fatigue. In the beginning, size of the cracks
/ MSD are sub millimeter levels and hidden under rivet heads. These cracks grow under cyclic
operational loads, coalesce and lead to threat structural integrity. Detection and understanding the
growth behavior of this damage are extremely important to maintain the structural integrity of the
airframe. Experimental work detailed here discusses evaluation of MSD on riveted butt joints under
fatigue cycles with online AE monitoring. Specimens used in study were fabricated from aluminum
alloy AA 2024 T3 widely used for the airframe construction.
Approximate entropy on detected AE signal parameters have been estimated for the peak, trough and
transient load conditions of fatigue cycles applied on the riveted joint specimens. Analysis is focused
on discriminating various stages of multisite damages such as single & multiple crack initiation, stable
crack growth, accelerated crack growth, coalescence of cracks etc. Results provide positive trends on
detection capabilities and discrimination of stages of MSD propagation of the methodologies
developed here with acoustic emission entropy analysis.
Hula Seal Stellite6 Coating Thickness on Service Returned Combustion Liner
Using Ultrasonic Gauge
Santhosh Bangera, Dheepa Srinivasan, Mohammed Anwar, James Baummer, Francis Reed
GE Power, GE India Technology Centre, Whitefield, Bangalore, India
Gas Turbine Services, Abu Dhabi, UAE
GE Power, Greenville, USA
GE Power, Schnectady, USA
Contact: Dheepa.Srinivasan@ge.com
Recently the Hula seal life extension program was implemented, permitting reuse of the Hula seal
during heat treatments applied to the combustion liner in gas turbines. The Hula seal is a spring seal
made of a Ni based superalloy (Inconel X750) with a wear resistant coating made of Stellite6. In some
cases the Stellite6 wear coating on the Hula seal is worn away after service exposure. In order to
implement the Hula seal reuse during repair, it is imperative to restore the Stellite6 coating back to the
required thickness/dimensions, to serve another interval of operation. Not carrying out the Stelllite6
coating restoration would imply that in spite of the life extension given to the Hula seal, the same
cannot be utilised to its full potential. This necessitated an estimate of the extent of wear in the Hula
seal, and more importantly assessing this wear, in situ on the liner without removing the Hula seal
during incoming inspection. A hand held ultrasonic gauge was used in order to assess the Stellite6
coating thickness, by differentiating the thickness of the coating in the unworn locations of the Hula
seal, and comparing with the worn regions along the centre of the Hula seal leaf. Calibration blocks
were made out of INX750 with and without the stellite6 coating to serve as a standard prior to testing.
This method has now paved the way for being able to extend the life of the Hula seal for several
intervals of service operation.
Effects and Corrective Techniques of Helium as Signal Noise in the MSLD based
Leak Detection Methods
Soumikh Sarkhel, V D Ragupathy and G Naryanan
Liquid Propulsion Systems Centre,
80 Feet Road, HAL II Stage,
Bangalore, India
Contact: ragupathyvd@lpscb.gov.in
Leak detection is primary NDT method for screening and acceptance of mechanical joints used in the
fluid circuits. Out of leak detection methods employed for evaluation of joint, several leak detection
techniques water immersion test, soap solution test are widely carried out depending upon the
application area. Leak detection using mass Spectrometer leak detector is an advanced technique used
for accessing the leak tightness of the aerospace fluid components. Helium mass spectrometer based
leak detection is the technique used extensively in the rocket propulsion system elements and
components as the sensitivity and repeatability is very good, in spite of cost of Helium. As Helium is
a rare gas in our atmosphere, the choice of Helium mass spectrometer is adapted widely for the rocket
engine testing. However there are certain situations when the tracer gas used for testing gets entrapped
in the testing zone may be due to leaking joint or unwanted leak from the test setup. This entrapped
Helium remains in the testing environment due to its lower molecular weight and takes sufficient time
to dissipate in the surrounding environment. The presence of such Helium in the atmosphere leads to
generation of noise in the leak detection process. This noise is also termed as background noise
leading to erroneous leak rate reported for a particular joint. The effect of source of this noise and the
methods to suppress the background noise are discussed in this paper.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Application of Image based technique in Qualification Testing of
Spacecraft Structures & Components.
Sriranga T S1 , Harsh Kumar, Ananthan. A, Raghunatha Behara, C.S Varghese. Structures
Group, ISRO Satellite center, Vimana Pura P.O. HAL Airport road. Bangalore 560017.
1
Correspondence: Sriranga.T. S. rangan@isac.gov.in;
ABSTRACT
Keywords: Structural Health Monitoring, Data evaluation and Image processing.
Several image based techniques are employed to ascertain strength, process quality, and dimensional stability of
spacecraft components. Details of efforts, in using specifically technique of Digital Image Correlation, for
Spacecraft structure and components are given here. To visualize strain distribution at selected small areas of
Composite cylinder during static load tests DIC was used. The distribution gave insight into the uncertainty of
strain gauge measurements due to its location on the honeycomb cell; whether it is on the cell wall or in the
middle of the cell. Also presented in the paper is image data analysis to extract information of interest such as
strain distribution around structural cutouts introduced by design to accommodate plumb lines, harness or even
test load application elements. A case study of measurement of large strains on a propellant tank is presented.
Effect of speckle pattern, sub set size selection is discussed.
1. Introduction
Digital Image Correlation (DIC) is commonly used technique to perform full field strain
measurements. In practice a carefully composed ‘speckle pattern’ is applied on the specimen
and camera images are taken first in the undeformed stage, which is the reference and then in
various deformed states. The technique consists of comparing the image of deformed pattern
with reference image and determining the displacements of the ‘subsets’. The software
eliminates rigid rotations of the subset before determining the strain field.
This technique was originally designed for large-strain measurements and as such, it works
very well when large strains are present, but when determining small strain fields, especially
in combination with large (rigid body) deformations or large strain gradients, this technique
becomes a lot more sensitive to the boundary conditions of the experimental setup and the
speckle pattern [1].
In this paper, studies done on capturing strains during various types of strain measurement
viz. strain gradients, large strains etc are presented. Efforts made in implementation of proper
speckle pattern (speckle size, density) for accurate measurement. Issues involved in picking
up Strain gradient around holes, cut-outs were studied.
Specifically the following case studies have been presented:
•
•
•
Accurate measurement of surface strains due to Static loading to visualize strain
distribution at selected small areas of CFRP cylinder due to Static loading on static
test rig.
CFRP honeycomb sandwich coupon variations absorbed in honeycomb cell walls and
cell centre.
DIC measurement was made on Aluminum Interface ring.
1
•
•
CFRP and Aluminum Honey comb sandwich coupons with circular cut-out are static
loaded and results were presented. An attempt is made to establish correlation
between Strain gauge Data and strain measured by DIC system.
DIC measurement was made on propellant tank, while Proof pressure loading.
2.
Large Strain Gradient Measurements
2.1
Short Cylinder under Compression loading
Any new design of CFRP layup of the cylinder involves a static load qualification. Prior to
fabrication of full size thrust cylinder, a cylinder of shorter height but identical diameter and
layups is fabricated and loaded on a test rig. During such a test, the test article is
instrumented with many strain gauges, displacement transducers to study behavior of the
cylinder.
2.2 Experimental results
The images were acquired using a digital camera having a resolution of 5 Mega pixels.
Continuous illumination was provided using LED lamps ensuring a reasonably uniform
illumination without any heating of test specimen. Fig.1 shows DIC experimental setup used
on the short cylinder. Camera was placed at a distance of half a meter from the specimen and
images were recorded continuously. The images were later analyzed using digital image
correlation software, VIC-3D to obtain strain fields. An important advantage of the method is
the whole field nature of the measurement data, providing quantitative evidence of important
features such as strain localization during the loading process.
Fig 1: shows DIC experimental
setup used on the short cylinder
Fig 2: Strain field (eyy, loading direction
strain) in composite cylinder
Since the composite cylinder is black in colour, random white colour speckle pattern was
applied on the component. Compression load 50 ton was applied on the composite cylinder.
2
Loading sequence was 0-25-50-25-0 ton. DIC Measurement image on Composite Cylinder
at 50 ton compressive load was shown in Fig. 2.
2.3 Honeycomb sandwich coupon
Random speckle pattern was applied on Honeycomb sandwich coupon and strain gauge also
mounted on it. A small load of 25 Kg of compression was applied on the coupon. DIC
measurement and Strain gauge monitoring was done simultaneously. Variation in the strain
gradients was absorbed in the cell centre and edges, which is shown in Fig.3a. A strain
gradient along the vertical line drawn on the coupon is shown in Fig.3b. Maximum Strain
measured in Strain gauges was -245 micro strains and corresponding DIC measurement was
-250 micro strains. The measured strain reading shows a good correlation between the Strain
gauge and DIC measurements.
Fig 3a: Honeycomb sandwich
coupon
Fig 3b: Strain gradients along the
vertical line drawn on coupon
2.4 DIC measurement was made on Aluminum Interface ring.
DIC measurement was made on Aluminum Interface ring. First white paint was coated on
the interface ring; later black random speckles were sprayed over it. Compression load 50 ton
was applied on the Interface Ring. Loading sequence
was 0-25-50-25-0 ton. DIC
measurement Image is shown in Fig. 4. Strain values at top, middle and bottom locations are
shown in Table 1. The DIC measured strain reading shows a good agreement with FEM
analysis results.
3
Fig 4 : DIC Image on the Aluminum
Interface Ring
2.5
Table1: Strain values at top, middle
and bottom locations on Interface Ring
Honey comb sandwich plate with a circular cut-out.
A CFRP sandwich panel with aluminum honeycomb core with a hole is prepared. This
coupon is tested with 500 Kgf load for getting strain gradient around the Circular cut-out. For
getting strain gradient different optics and speckle pattern has been tried. Images were
recorded and analyzed.
2000
PLATE WITHOUT DOUBLER
µ strains
1500
1000
FEM
500
DIC
0
0
20
40
60
Distance from cut-out edge in mm
Fig 5a: strain (eyy) variation
along cutout
Fig 6a: Strain measurement
using VIC 2D
Fig 5b: comparison of strain with FEM
data
Fig 6b: Data correlation along cutout with
fill boundary extrapolation
4
The fabricated test coupons are tested in a UTM to study the strain distribution with DIC.
Speckle pattern on the surface of test coupons is done by using Acrylic paints for Digital image
correlation. Test coupons with speckle pattern are mounted to the UTM with the help of loading
fixtures and the test is conducted by applying tensile load. The images are taken at regular
intervals while loading and post processing of the images is done.
The graphs are plotted for strain distribution at cut-out section for results obtained
from both FEM and DIC technique. While comparing it is found fairly good agreement between
FEM and DIC technique results (Fig 5a & b). Selection of sub set size affects the data
correlation. Smaller the sub set better is the spatial resolution but decorrelation chances are
more. So a trade off is required. Hence to have data correlation and strain value near cut-out
edge many subset sizes were tried and shown in Fig. 7 iteratively with a starting value as
suggested by the DIC software itself. In the range of 29 x29 to 75 x 75, it was found smaller
subset resulted in some of the area not being correlated, while with higher subset area near cutout edge was not correlated. To overcome this situation different optics and different patterns
have been tried. The lack of measurement at edge of holes or cut outs was recognized. For 2D
measurements, the data is extrapolated up to edge of holes /cut-outs (Fig 6a & b).
Fig 7 Various subset sizes for correlation at the edge
of the cut-out
2.6
Large strain monitoring during Proof pressure testing on RLV tank.
Propellant tanks are by design thin walled structures meant for carrying pressures in the order
of 35 bars. During handling of the tank, a small strain concentration zone had developed and
a need for pressurisation test arose. DIC technique was used to monitor in real time the strain
build-up around the local area of interest (AOI) is shown in the Fig.8. Proof pressure test was
conducted on the tank and Strain monitoring during multiple pressure cycles was done on the
5
AOI. Mounting of strain gauge was not possible in that area. The AOI was painted with
white paint and random black speckles pattern was sprayed over the white paint area.15
cycles of pressurization was done on the propellant tank. The highest strain noticed was
around 7600 micro strains and
DIC images for 1st cycle was shown in Fig.9. Strain
measurements were correlated with FE analysis. Both test and analysis strain data showed
similar strain distribution
DIC SPECKLE PATTERN
Fig 8: Strain monitoring on RLV tank proof pressure testing (35 bars
MEOP)
Figure 9: DIC image pattern of the Propellant tank at 35 bar
pressure
3.
Results & Discussion
The case studies have provided valuable inputs for future measurements and have
exposed some of the limitations in the DIC technique. The finite size of the subset
prevents measurement of the strain value at the highest stress concentration points. An
estimator for linearly extrapolating the values is required in 3D measurements. The
speckle size and density of the pattern need to be so chosen as to be able to select a small
subset without encountering de-correlation problem. To some extent proper choice of
optical lens with a larger focal length and larger aperture number helps in zooming on to
smaller region on the component[2]. A good Variations of Strain across honeycomb cell
walls and cell centre was captured. Large strain measurements of the order of 7600 micro
strains as encountered in the propellant tank case called for a speckle pattern of lower
6
density and larger speckle size. Issues arising during measurement of large strain
gradients, large strains, are handled by speckle pattern.
4.
Conclusions
During the course of implementation of DIC for structural components the following
objectives have been achieved.



Gradients around Stress concentration zone.
o Circular cut-out in a sandwich plate
Large strain monitoring.
o Proof pressure testing on RLV tank.
During the implementation effect on uncertainty:
o Pattern speckle size and density
o Selection of subset size and step size for image processing.
The DIC technique has been used in a variety of situations presenting unique measurement
challenges. The curvature of short cylinder was addressed by limiting the field of view to
overcome depth-of-field problems of optics. In situations of picking up strain gradients
around structural cut-outs, need for selecting sub set size and optics was addressed. The lack
of measurement at edge of holes or cut outs was recognized. For 2D measurements, the data
is extrapolated up to edge of holes /cut-outs . Speckle size and density of pattern used for
large strains as in the case of propellant tanks was adequate to capture the strain field reliably.
5.
ACKNOWLEDGEMENT
Authors wish to thank Dr.Anand Kumar Sharma, Deputy Director of Mechanical systems
area, ISRO Satellite center, and Dr.K Renji, Group Director, for kind permission to present
this work. We thank Mr. S Shankar Narayan, Head of Experimental Structures division, for
technical guidance and suggestions.
We also would like to thank Ms.Chithra S, Technician, R.Rajesh, Lokesh A.H,
ChaluvaRaja.C.M, Contract Assistants in carrying out the experimental work.
6.
References.
[1] Image correlation for shape,motion, and deformation measurement : Basic concept, theory
and applications by Michael A Sutton, Jean-Jose Orteu & Hubert W. Schreier
[2] Assesment of Complex Aerospace Design through optical Techniques , Digendranath
Swain, Jeby Philip, Applied mechanics and Materials, vols 592-594 (2014) pp 1006-1010
7
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Uncertainty Handling using Fuzzy Logic in Structural Health Monitoring
Ranjan GANGULI 1
1
Department of Aerospace Engineering, Indian Institute of Science; Bangalore, India
Phone: +91 8022933017, Fax: +91 8023600134; e-mail: ganguli@aero.iisc.ernet.in
Abstract
Uncertainty emanating from models and measurements are a source of inaccuracies in fault diagnosis and
structural health monitoring. Signal processing algorithms can smooth data and remove noise to some extent
from measurements, but model uncertainty remains. Model uncertainty can affect the accuracy of model based
diagnostics methods. Among the methods for handling uncertainty in engineering systems, fuzzy logic has
emerged as an attractive framework. Fuzzy logic is based on the logic of computing with words, rather than
computing with numbers, and is found to be closer to human reasoning. The application of fuzzy logic for
damage detection in structures is discussed. Several case studies are used to illustrate the application of the
methods to progressively complicated structures and with progressively complicated fuzzy logic architectures.
First, fuzzy logic is demonstrated on a cantilever beam with frequencies as the damage indicators. Then mode
shapes and modal curvatures are used as damage indicators. The damage indicators are contaminated with noise
to simulate measurement noise. The physics based model is also noisy and accounts for the aleatory or random
uncertainty in the material properties. The fuzzy logic based damage detection system is developed for these
structures and is found to perform very well. A new sliding window defuzzification method is created to improve
the accuracy of the fuzzy system. Finally, a hybrid soft computing architecture known as genetic fuzzy system is
developed for damage detection and demonstrated on a rotor blade. In this system, a genetic algorithm is used to
maximize the damage detection success rate by optimizing the design parameters of the fuzzy system such as the
standard deviations of the Gaussian fuzzy sets.
Keywords: Fuzzy logic, structural health monitoring, genetic algorithm, uncertainty, pattern recognition,
vibrations
1. Introduction
NDT is susceptible to uncertainty due to noisy measurements. Structural health monitoring
(SHM) can be considered to be a subset of NDT as it aims to estimate the health of a structure
without doing any damage to it. SHM uses sensors to estimate the state of the structure and
often uses a baseline “undamaged” model of the system. Many SHM methods are model
based and use information from mathematical models of the damaged systems to detect and
isolate the damage in the real system. A key assumption made in this process is that the
changes in measured behaviour of the damaged system are similar to the changes in the
measured behaviour of the mathematical model. While this assumption may seem limiting,
the changes in measured behaviour are often better predicted by models than the absolute
values. Furthermore, it is not possible to get experimental data for many damaged aerospace
structures in flight condition. For example, while the loss of a lag damper in a helicopter rotor
blade can be mathematically simulated, it would not be possible to fly such a helicopter due to
airworthiness requirements. However, mathematical models of dissimilar rotor blades can be
used for understanding the behaviour of damaged helicopter rotors which can be used to
develop health monitoring systems using pattern recognition algorithms such as those based
on neural networks [1] and fuzzy logic.
The damage detection problem can be viewed as a pattern recognition problem. The changes
in the measurements from an “undamaged” system are mapped to the damage type and
damage size. The structural health monitoring system is then expected to identify the type of
damage and size of damage when presented with damaged system measurements. If this
problem had no noise or uncertainty in measurement of noise, it would be easy to handle.
Unfortunately, real systems have noise in measurements and mathematical models. For
example, model uncertainty can come from the fact that the material properties used in the
model can be different from that of the real system because of dispersion due to
manufacturing errors, fabrication processes, voids and defects, etc. [2]. Other sources of
mathematical uncertainty can be weaknesses on physical modelling, incomplete convergence
of numerical solution etc. Thus, uncertainty must be handled well when performing structural
health monitoring. We show that fuzzy logic is a powerful method for handling uncertainty in
structural health monitoring. This paper summarizes some of the authors past work in this
area and presents some ideas for future research.
2. Fuzzy Logic System
A fuzzy logic system (FLS) maps an input feature vector into a scalar output [3]. Typically,
this is a nonlinear mapping. An FLS is a linear combination of fuzzy basis functions and has
been shown to be a universal function approximater. A multi-input single-output (MISO) FLS
maps an n dimensional input space to a one dimensional output space. A typical FLS
performs this mapping using four basic components: rules, fuzzifier, inference engine and
defuzzifier, as shown in Figure 1. The main component of an FLS is the rule base. Rules can
be created based on expert knowledge, numerical simulations, empirical evidence or a
combination of these sources. Typically, the rules can be expressed as if-then statements of
the form ubiquitous in computer programming languages. For example, a rule can be: “If u is
high and v is low then w is low” where u and v are inputs and w is the output. Rules need an
understanding of the following concepts:
1. Numerical values such as 3.5% are converted to linguistic values or words such as
low.
2. The quantifying linguistic variables such as u have a finite number of linguistic terms
associated with them, for example ranging from negligible to very high.
3. Logical connections between linguistic variables example and/or. Again, this is
similar to programming languages.
4. Implications such as If a then b.
The FLS received numerical values as inputs. It converts these numbers into words in the
fuzzifier. These words are modelled as fuzzy sets. The inference engine operates on fuzzy sets
using the rules. In some applications, numerical values are needed as outputs. In these cases, a
defuzzifier converts the words back into numerical variables. In other problems, the fuzzy set
output is better suited for the problem. The FLS has also been called “computing by words”.
A fuzzy set F is defined over a universe of discourse U and has a degree of membership (x)
which can take values between 0 and 1. This generalizes the concept of a typical element of a
set which can be within a set (degree of membership one) or out of the set (degree of
membership zero). For example, concepts such as high temperature or tall person use fuzzy
concepts. A given person can have a degree of membership in the tall fuzzy set. For a person
with a height of 5 feet six inches, the degree of membership in the tall set may be 0.3 and for
a person with a height of 6 feet six inches, the degree of membership in the tall set may be
0.9. Thus, fuzzy logic approach of computing with words is analogous to human reasoning.
Each fuzzy set if typically associated with a membership function. The commonly used
membership functions are triangular, trapezoidal, piecewise linear of Gaussian. The user
needs to select the fuzzy set used for the input and output variables. Further details for FLS
are given in [3].
3. Fuzzy Logic System for Damage Detection
The application of FLS for damage detection is illustrated by a simple example using
frequencies as input variables and damage location and size as output variables (Figure 1) [3].
The beam is divided into five equal segments, each having a length equal to 20 percent of the
beam length (Figure 2). These five locations are named as root, inboard, center, outboard,
and tip. The measurements or input variables are the changes in the first four natural
frequencies from the baseline undamaged condition. These input measurement deltas are split
into fuzzy sets. The symbols N, VL, L, LM, M, MH, H, VH and VVH are linguistic variables
which stand for negligible, very low, low, low-medium, medium, medium-high, high, very high
and very-very high, respectively. Structural damage is modelled as a loss of stiffness at the
damaged stiffness. Gaussian fuzzy sets with a standard deviation of 0.35 are used and the
midpoints of the set and symbols are allotted. The output damage locations can have damage
at several different levels. These are classified as undamaged, slight damage, moderate
damage, severe damage and catastrophic damage.


Fuzzy inference
engine
Fuzzifier
Damage
Detection,
Defuzzifier
Location
and Size
Fuzzy rule base
Fuzzy Logic System (FLS)
Figure 1 Schematic of fuzzy logic system for structural damage detection
20%
ROOT
20%
20%
INBOARD
CENTER
20%
OUTBOARD
20%
TIP
Figure 2 Cantilever beam divided into five segments for damage location
Rules of the FLS are obtained by fuzzification of the numerical values which are obtained
using a finite element based mathematical model of the damaged beam. A set of four
frequency measurement deltas corresponding to a given damage location and size are fed into
the FLS and the degree of membership of each of the four frequency measurement deltas are
obtained. Each measurement delta is allotted to the fuzzy set which has the maximum degrees
of membership. One rule is extracted for each damage level and location by correlating the
measurement delta with the maximum degree of membership to a fault. The rules are easy to
interpret. For example, one rule is: If  is medium high and 2 is low and 3 is very low
and 4 is low then moderate damage at root. Other such rules can be generated as given in
detail in [3]. Here i stands for the percentage change in the ith natural frequency relative to
the undamaged structure. The fuzzy rules are similar to those a human expert can discover
and learn after years of experience in the field, as shown in Table 1. Fuzzy logic solves a
pattern recognition problem of the type shown in Figure 3. Fuzzy system is also an open
system is contrast to neural networks which are closed architectures.
Application of FLS to damage detection becomes more complex as the system becomes more
complex. For example, [4] used fuzzy logic for solving a helicopter rotor damage detection
problem. Here, measurement deviations of blade tip bending and torsion response and
vibration from a “good” undamaged helicopter rotor are considered as inputs to the FLS.
Rules for the FLS are developed using an aeroelastic model of the helicopter rotor with
implanted damage. The damages modelled are moisture absorption, loss of trim mass,
damaged lag damper, damaged pitch control system, misadjusted pitch link, and damaged
flap. The damage was implanted into one blade and the effect of dissimilar helicopter rotor
dynamics was also addressed (Figure 4). The FLS is found to perform with an accuracy of 90100 percent with simulated data with added noise. The FLS is found to be superior to an
expert system which is based on the same rule base but does not use fuzziness to handle
uncertainty. The fuzzy approach was found to be easier to develop, implement and understand
than neural network based methods [5-6].
Fuzzy logic systems are only as good as the underlying rules. The process of developing rules
for the FLS can be cumbersome for a human designer. There is an effort in the soft computing
community to create hybrid methods to enhance the capability of fuzzy systems. One such
approach is the genetic fuzzy system which uses genetic algorithms to create an optimal rule
base. Pawar and Ganguli [7] developed a genetic fuzzy system for damage detection in a
cantilever beam. They considered natural frequencies and mode shape changes as the input
variables and found that complex structures are more amenable to the genetic fuzzy system
approach as the number of measurements and output damage location and size can become
large. This avoids the curse of dimensionality problem which hampers fuzzy logic systems.
Table 1 Rules for fuzzy logic system
Damage

Undamaged
N
Slight damage at root
VL
Slight damage at inboard
VL
Slight damage at center
N
Slight damage at outboard
N
Slight damage at tip
N
Moderate damage at root
MH
Moderate damage at inboard LM
Moderate damage at center
VL
Moderate damage at outboard N
Moderate damage at tip
N
Severe damage at root
VVH
Severe damage at inboard
MH
Severe damage at center
L
Severe damage at outboard
N
Severe damage at tip
N
2
N
VL
N
VL
VL
N
L
VL
LM
L
N
M
VL
H
M
N
3
N
N
VL
N
VL
N
VL
L
VL
LM
VL
LM
M
VL
MH
VL
4
N
N
N
VL
N
VL
L
VL
L
L
L
LM
LM
M
LM
LM
Pattern 1
Damage 2
Ideal
Data
Noisy
Data
Damage 1
Pattern 3
Noisy
Data
Noisy
Data
Ideal
Data
Ideal
Data
Pattern 2
Damage 3
Figure 3 Damage detection as a pattern recognition problem
Modern aerospace structures are often made of composite materials because of high stiffness
to weight ratio, superior damage tolerance properties and high fatigue life. A genetic fuzzy
system was developed for detecting matrix cracks in a composite pole [8]. The genetic fuzzy
system automates the rule development process in the development of a fuzzy system.
Typically, rules developed by human designers are suboptimal. Using a genetic algorithm to
maximize the success rate for a fuzzy system can result in tailored mid-points and standard
deviations of fuzzy sets and such architecture was found to give very good damage detection
results.
Pawar and Ganguli [9] applied the fuzzy logic method for damage detection in composite
helicopter rotor blades. A key element of this model based approach was developing
mathematical models which can simulate a damaged composite helicopter rotor in forward
flight [10-12]. Damage modelled included matrix cracking, debonding/delamination and fiber
breakage. A box-beam model of the rotor blade was used and the effect of damage on rotating
frequency, tip bending and torsion response and blade loads was studied. These measurements
are used for global fault detection. The effect of damage is also studied on axial and shear
strain for local damage detection. The mathematical model used is that of a dissimilar
composite helicopter rotor in forward flight and is based on the finite element method in space
and time.
Blade 2
(undamaged)
Blade 3
(undamaged)
X

Blade 1
(damaged)
Blade 4
(undamaged) Y
Figure 4 Damage in a helicopter rotor simulated by damage on one blade
The damage detection approach is extended to life estimation for composite rotor blades [13].
The cross sectional stiffness reduction obtained by physics based models is expressed as a
function of the life of the structure using the Mao and Mahadevan [14] phenomenological
damage growth model for composites. It is found that the life consumption in the matrix
cracking zone is about 12-15 percent and in the debonding-delamination zone is about 45-55
percent. It is postulated that the structure should be removed at the end of the debondingdelamination phase as the more dangerous and potentially catastrophic fiber breakage phase
begins thereafter. Genetic fuzzy systems are developed for global online prediction of
physical damage and life consumption. It is found that the success rate of the genetic fuzzy
system depends on the number of measurements and type of measurements. The genetic fuzzy
system works well with noisy data and was recommended for online SHM of composite
helicopter rotor blades (Figures 5 and 6).
In recent years, the importance of material uncertainty has become clear in structural health
monitoring. Different specimens of a structure have different material properties due to
differences in fabrication process, etc. Thus, mathematical models created to develop damage
detection systems need to factor in the material uncertainty present in the structure. This is
important for composite structures as such materials have higher levels of uncertainty in
parameters such as Young’s modulus, Poisson ratio etc. compared to metal structures.
Chandrasekhar and Ganguli [15-17] studied the importance of material uncertainty in SHM
using natural frequencies, mode shapes and modal curvatures as the inputs and location and
size of damage as the output. They also used Monte Carlo simulations of the measurements to
guide the selection of the fuzzy sets, thereby connecting the disparate fields of probability and
fuzzy logic. Numerical results showed that a new sliding window defuzzification system
performed very well for damage detection. Results were obtained for isotropic materials and
composite materials. For isotropic material, a coefficient of variation (standard
deviation/mean) of 3 percent was considered in the elastic modulus and a noise level of 0.15
was included as measurement noise. The fuzzy system was able to identify damage to an
accuracy of 94 percent using natural frequencies.
Physics based damage
Undamaged blade
Damage Modeling in Composite
U
Linking
Phenomenological
Damage Model
Life
Consumption
Composite
Blade Model
Aeroelastic
Analysis
D
Damaged blade
Z (D)
Z (U)
Data Reduction
 Z (D) = Z (D) -Z (U)
GFS
Physics
based
Damage Prediction
Residual
Prediction
Life
Figure 5 Schematic of development of genetic fuzzy system (GFS) for composite rotor blade
4. Concluding Remarks
Fuzzy logic is a powerful tool which can be exploited to enhance non-destructive testing
methods. Almost any problem which can be solved using neural networks can also be solved
using fuzzy logic. In addition, fuzzy systems are easier to understand, implement and deploy
compared to neural networks. This paper summarizes some of the research published by the
author and his co-worker in this area and provides references for further information. The
work will help researchers in using fuzzy logic for NDT applications and may lead to further
innovations at the edge of engineering and computer science, as some of the studies discussed
here have done. As sensors become ubiquitous on systems and large amounts of data (big
data) become available, fuzzy logic becomes valuable as an efficient machine learning and
artificial intelligence method for extracting more information from the sensors for performing
SHM and NDT functions. In conclusion, fuzzy logic is a useful tool for NDT researchers.
Readers can learn more about fuzzy logic from Zadeh [18] and the books by Kosko [19] and
Ross [20]. An introduction to the use of fuzzy logic for structural health monitoring is given
by Pawar and Ganguli [20] and a review of different fuzzy logic and other methods for
helicopter health monitoring is given in [21].
Forces
Strains
Deflections
Data Reduction
Physical Damage
GFS
Z
Undamaged Blade
Data
Prognostics
Residual Life
Maintenance Action
Figure 6 Implementation of genetic fuzzy system for rotor blade
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
L Udpa and S S Udpa, 'Neural Networks for the Classification of Non-Destructive
Evaluation Signals', IEE Proceedings-F, Vol 138, No 1, pp 201-205, February 1991.
P Gayathri, K Umesh and R Ganguli, 'Effect of Matrix Cracking and Material
Uncertainty on Composite Plates', Reliability Engineering and System Safety, Vol 95,
No 7, pp 716-728, 2010.
R Ganguli, 'A Fuzzy Logic System for Ground Based Structural Health Monitoring of a
Helicopter Rotor using Modal Data', Journal of Intelligent Material Systems and
Structures, Vol 12, No 6, pp 397-407, 2001.
R Ganguli, 'Health Monitoring of a Helicopter Rotor in Forward Flight Using Fuzzy
Logic’, AIAA Journal, Vol 40, No 1, pp 2373-2382, 2002.
R Ganguli, I Chopra and D J Haas, 'Detection of Helicopter Rotor System Simulated
Faults Using Neural Networks', Journal of the American Helicopter Society, Vol 42, No
2, pp 161-171, 1997.
R Ganguli, I Chopra, and D J Haas, 'Helicopter Rotor System Fault Detection Using
Physics Based Model and Neural Networks', AIAA Journal, Vol 36, No 6, pp 10781086, 1998.
P M Pawar and R Ganguli, 'Genetic Fuzzy System for Damage Detection in Beams and
Helicopter Rotor Blades', Computer Methods in Applied Mechanics and Engineering,
Vol 192, No 16, pp 2031-2057, 2003.
P M Pawar and R Ganguli, 'Matrix Crack Detection in Thin-Walled Composite Beam
Using Genetic Fuzzy System', Journal of Intelligent Material Systems and Structures,
Vol 16, No 5, pp. 395-409, 2005.
P N Pawar and R Ganguli, 'Genetic Fuzzy System for Online Structural Health
Monitoring of Composite Helicopter Rotor Blades', Mechanical Systems and Signal
Processing, Vol 21, No 5, pp 2212-2236, 2007.
P M Pawar and R Ganguli, 'Modeling Multi-Layer Matrix Cracking in Thin Walled
Composite Helicopter Rotor Blades', Journal of the American Helicopter Society, Vol.
50, No. 3, 2005, pp. 354-366.
P M Pawar and R Ganguli, 'On the Effect of Matrix Cracks in Composite Helicopter
Rotor Blades', Composites Science and Technology, Vol 65, No 3-4, pp. 581-594,
2005.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
P M Pawar and R Ganguli, 'Modeling Progressive Damage Accumulation in Thin
Walled Composite Beams for Rotor Blade Applications', Composites Science and
Technology, Vol 66, No 13, pp 2337-2349, 2006.
P M Pawar and R Ganguli, 'Fuzzy Logic Based Health Monitoring and Residual Life
Prediction of Composite Helicopter Rotor', Journal of Aircraft, Vol 44, No 3, pp 981995, 2007.
H Mao and S Mahadevan, 'Fatigue Damage Modeling of Composite Materials',
Composite Structures, Vol 58, No 4, pp 405-410, 2002.
M Chandrasekhar and R Ganguli, 'Structural Damage Detection using Modal Curvature
and Fuzzy Logic', Structural Health Monitoring, Vol 8, No 9, pp 267-282, 2009.
M Chandrasekhar and R Ganguli, 'Uncertainty Handling in Structural Damage Detection
using Fuzzy Logic and Probabilistic Simulation', Mechanical Systems and Signal
Processing, Vol 23, No 2, pp 384-404, 2009.
M Chandrasekhar and R Ganguli, 'Damage Assessment of Composite Plate Structures
with Material and Measurement Uncertainty', Mechanical Systems and Signal
Processing, Vol 75, pp 76-93, 2016.
L Zadeh, 'Fuzzy Logic = Computing with Words', IEEE Transactions on Fuzzy
Systems, Vol 4, No 2, pp 103-111, 2000.
B Kosko, Fuzzy Engineering, Prentice-Hall, Englewood Cliffs, NJ, USA, 1997.
T J Ross, Fuzzy Logic with Engineering Applications, John Wiley and Sons, New
Jersey, USA, 2009.
P M Pawar and R Ganguli, Structural Health Monitoring using Genetic Fuzzy Systems,
Springer, London, 2011.
P M Pawar and R Ganguli, 'Helicopter Rotor Health Monitoring – a Review ', Journal of
Aerospace Engineering, part G, Vol 221, No 5, pp 631-647, 2007.
Damage Assessment of Single Blade Stiffened CFRP Specimen Subjected to
Axial Compression using AE and DIC Techniques
Naresh R Kolanu, Lala Bahadur Andraju, M Ramji, Gangadharan Raju
Department of Mechanical and Aerospace Engineering,
IIT Hyderabad, Hyderabad, India
Contact: gangadharanr@iith.ac.in
Damage tolerant design of stiffened composite panels requires a clear understanding of the various
damage mechanisms like matrix cracking, fiber breakage and delamination. In this work, both
acoustic emission (AE) and digital image correlation (DIC) techniques are used for the assessment of
damage initiation and progression in a quasi-isotropic CFRP laminates under uniaxial compression.
The blade stiffener is fabricated using vacuum infusion process and later co-bonded to the skin of
quasi-isotropic laminate. End blocks are cast at the loading edges of the stiffened panel for applying
uniform compression load with no support along longitudinal edges. Experiments are carried out on
CFRP stiffened panels to capture the buckling load, displacement and strain field over the specimen
and also various AE parameters like hits, amplitude, relative energy, counts are determined. Both the
AE and DIC data are analyzed to extract qualitative and quantitative information about the
progressing damage mechanisms in the stiffened panel. DIC can provide whole field displacement
and strain over the critical regions such as the skin-stiffener interface where usually de-bonding
initiation and progression takes place. Acoustic emission parameters can be used to identify and
classify the various damage mechanisms in stiffened panel in a progressive manner and could be used
connected with the surface strain field obtained from DIC. Both AE and DIC techniques complement
each other and makes them suitable for the health monitoring of composite structures.
Real Time Monitoring of Interfacial Delamination of Sandwich Composite
Panels using Optical Sensors
Nilanjan Mitra
Department of Civil Engineering,
Indian Institute of Technology Kharagpur,
Kharagpur, India
Contact: nilanjan@civil.iitkgp.ernet.in
Sandwich composites are used in many different industrial sectors (ranging from infrastructure to
space vehicles) primarily due to light weight and high strength feature. Depending upon application,
the components of the sandwich composite varies with regards to the core (which can either be foam
or honeycomb materials made of polymers such as PVC, PMMA; metals and its alloys such as Al, Ti6Al-4V; ceramics such as SiC, B4C etc.) and face-sheet material (glass fiber, carbon fiber, Kevlar,
different metallic and ceramic sheets etc.) along with the glue system (epoxy, polyester, vinyl-ester
etc.) used and the method of fabrication (hand layup, vacuum resin infusion, autoclave etc.). Light
weight and high strength typically equates to easy transportability and modular construction (for
infrastructural purpose), low fuel consumption and higher payloads (for vehicles which maybe
automobiles, marine, aerospace, space vessels) and so on. In-spite of numerous benefits, one of the
major problems of these structures are propensity for interfacial delamination between the face sheet
and the core material which may be either due to manufacturing defects or while in service (such as a
result of some low velocity impact). These interfacial defects might eventually propagate as a result of
normal loading conditions which the structure or component might encounter in service.
Sandwich composites that are being considered in this research are that of glass fiber face sheets, poly
vinyl chloride foam core and epoxy resin system. These types of sandwich composites are typically
utilized in manufacture of different components of aircrafts, naval vessels as well as wind turbine
blades. It should be mentioned that for laminated composites there are various non-destructive means
by which delamination in between fibers are identified. The methodologies suitable for laminated
composites such as ultrasonic C-scan, x-ray, thermography and eddy current are unfortunately not
suitable for sandwich composite structures with a polymer foam core. The primary reason for nonsuitability of these methodologies is the absorbance of the generated signal by the core layer which is
typically a polymeric foam material. Apart from that, neither of the above described NDT
mechanisms is capable of real time in-situ determination of delamination since all of the above
described NDT measures require substantial amount of time and resources for post-damage inspection
and assessment. Thereby this means that the structure needs to be decommissioned and assessed
which reduces operation time of the vehicle resulting in an increase in life-cycle cost of the structure.
The Digital Image Correlation (DIC) technique can be used for real time determination but these are
unable to probe inside the structure of the material, especially for determination of interfacial
delamination in sandwich composites. The strain gauges typically provide a continuous and in-situ
monitoring of the structure but these are susceptible to electromagnetic and electrical interferences
and thereby cannot be used for many different structures especially in the defense domain. Moreover
the applicability of strain gauges in an embedded form is also not that reliable and because of size
effects might induce initial damages to the area where it is installed. Acoustic emission could
potentially be an effective method, although it suffers from low signal-to-noise ratio. Fibre optic
sensors on the other hand have a proven strain measuring capability with such characteristics as light
weight, small size, immunity to electromagnetic interference and easy integrability into composite
structures to produce so-called smart composite structures so as to access interior material and
structure locations where other sensing methods cannot easily probe. Amongst the fiber-optic sensors
currently being used for structural health monitoring, the most common is that of Fiber-Bragg grating
sensors. Bt cost is an issue of this FBG sensors and also interdependency of the strain and temperature
becomes a major bottleneck for use of these sensors. A new methodology known as Singlemultimode-single mode (SMS) fibers is being proposed which performs a distributed sensing of
strains being embedded at the interlaminar region of the sandwich composite samples.
A double cantilever beam sample with embedded SMS fibers are prepared and loaded experimentally.
The distributed strains are measured using this method and compared with numerical simulations.
Similar experiments are also done using FBG sensors and results have been reported and compared
with SMS methodology.
Locating Delamination in a Composite Laminate Using Nonlinear
Response of Lamb Waves
Nitesh P. Yelve1, Mira Mitra2, and Prasanna M. Mujumdar2
1
Deaprtment of Mechanical Engineering
Fr. C. Rodrigues Institute of Technology, Vashi, Navi Mumbai, India
2
Department of Aerospace Engineering
Indian Institute of Technology Bombay, Mumbai, India
1
nitesh@aero.iitb.ac.in
Abstract:
Present work deals with experimental and Finite Element simulation studies for locating a
delamination in composite laminates. A Lamb wave based nonlinear method is used for this
purpose, wherein Lamb waves produce higher harmonics after interacting with the
delamination as a result of contact nonlinearity. A hybrid method is formulated here which
uses frequency and time domain information from the response signal in order to locate the
delamination. The proposed method is simple and robust, and found to locate small
delamination in both the experiments and simulation with a fair accuracy which is generally
difficult to achieve using a Lamb wave based linear method.
Key word: Composite laminates, Delamination, Nonlinear response of Lamb waves, Contact
nonlinearity, Higher harmonics, Locating delamination, Temporal and spectral response.
1. Introduction
Composite laminates are widely used for making aerospace structures as these offer
advantages such as high specific strength, high specific stiffness, corrosion resistance, etc.
These composite laminates are susceptible to delamination which causes great reduction in
their strength related properties. Therefore, detection of delamination is quite essential, before
it causes any catastrophic failure. Ultrasonic testing using guided waves, typically Lamb
waves is an effective method for detecting delamination in the composite laminates, as Lamb
wave can scan relatively large area even in materials with high attenuation ratio [1].
Based on the characteristics of wave-damage interaction, Lamb wave based methods are
generally classified into two groups, linear and nonlinear [2]. In linear methods, parameters
like attenuation, transmission, and reflection coefficients indicate the presence of
delamination [2] in a composite laminate. Many researchers [3-15] have used Lamb wave
based linear methods for detecting delamination in composite laminates. However, the linear
Lamb wave methods are not sensitive to smaller delaminations [2], thus literature does not
show any work on locating small delamination in the composite laminates. In the case of
Lamb wave based nonlinear methods, higher harmonics, sub-harmonics, shift of resonance
frequency, and mixed frequency response indicate the presence delamination in a composite
laminate [2]. The nonlinearity introduced by wave-damage interaction is either classical or
non-classical [16]. Defects such as distributed micro-cracks in the material continuum during
fatigue damage progression make the material more complaint and in turn make the Hooke’s
law nonlinear by introducing higher order elastic terms into it. Such a nonlinearity is termed
as classical one [16]. There are many mechanisms proposed in the literature to characterize a
delamination which clap or breathe in response to the propagating wave. These mechanisms
include rough surfaces contact, stress-strain hysteresis, Luxemburg-Gorky effect, contact
nonlinearity, etc. and are considered to be responsible for the non-classical nonlinearity [16].
Each mechanism produces different nonlinear effects such as higher harmonics, subharmonics, mixed frequency response, etc. In the present case, contact nonlinearity is
emphasized as it is more prominent in the delamination kind of damages. As the delamination
closes and opens in spatially distributed modes due to the propagating wave, stiffness
changes at the contact interface. This gives rise to the stronger contact nonlinearity at the
delamination and in turn produces higher harmonics in Lamb waves excited at a relatively
lower frequency.
Higher harmonics are shown to be producing in Lamb waves as a result of the classical
nonlinear mechanism in [17-21]. Bermes et al. [17] and Deng and Yang [18] used this
method for material characterization purpose, whereas, Pruell et al. [19-20] and Deng and Pei
[21] used it for detecting damages such as plasticity driven material damage and accumulated
fatigue damage. Yelve et al. [22-23] used a non-classical mechanism, the contact
nonlinearity, with Lamb waves to detect and estimate size of a breathing crack in an
aluminium plate and to show the presence of higher harmonics in an intact aluminium plate
because of disbonding of piezoelectric wafer (PW) actuator respectively. As far as
delamination is concerned, very few researchers [24-25] have reported the work. Shkerdin
and Glorieux [24] modeled the nonlinear interaction between high frequency Lamb waves
and a metallic bilayer containing a delamination using a quasi-stationary approach. Sarens et
al. [25] investigated delamination-induced effects on the vibration of a harmonically excited
composite plate using laser Doppler vibrometry and shearography. Though, Lamb wave
excitation is claimed at a sinusoidal frequency of 20 kHz, only local vibration responses
confined to the delamination area are given. However, to the best of the authors’ knowledge,
nonlinear Lamb wave method involving higher harmonic generation as a result of contact
nonlinearity, has not yet been used for locating a delamination in composite laminates.
In the present study, experiments and Finite Element (FE) simulations are carried out to
locate a delamination in a composite laminate using Lamb wave based nonlinear-higher
harmonics generation method. E-glass/epoxy type of composite laminate with woven fiber
geometry is considered for the study. Woven fiber composites (WFCs) have balanced inplane
properties and better transverse tensile strength as compared to the unidirectional composites
because of the integrated nature of woven fibers [26]. The specimen laminates are prepared in
the laboratory. Teflon strip is used to create delamination in the specimen laminate. In the FE
simulation, contact nonlinearity is modeled using the contact elements and Augmented
Lagrangian algorithm. Piezoelectric wafer (PW) transducers are used for actuating and
sensing Lamb waves in both the experiments and simulation. These transducers are compact,
cost effective, and have wider operating frequency range. Further, a delamination is located
in a WFC laminate using the new hybrid method which uses both frequency and time domain
information from the response signal.
2. Nonlinear Interaction Between Lamb Waves and Delamination in a Composite
Laminate
2.1 Experimental Study
Firstly, experiments are carried out on a WFC laminate with the objective of studying
nonlinear interaction between Lamb waves and delamination. The WFC laminate is made in
the laboratory using woven E-glass fibers, LY556 epoxy resin, and HY951 hardener. The
plan form dimensions of the laminate are 425 mm × 265 mm. It has six layers and its total
thickness is 2.1 mm. Teflon strip is used for creating delamination in the laminate between its
second and third layers. Teflon strip is very thin, and it does not change any structural
property of the laminate. The width of the delamination is 5 mm, and it spans over a distance
of 72.5 mm to 77.5 mm from the PW actuator as shown in Fig. 1. The PW sensor is located at
a distance of 180 mm from the actuator. The size of the PW transducers is 10 mm × 7 mm ×
0.5 mm and the material type is SP-5H. A specimen is also made with no delamination. The
fiber volume fraction of the laminate is 0.55, density is 1850 Kg/m3, in-plane modulus of
elasticity and shear modulus are 32 GPa and 1.7 GPa respectively, modulus of elasticity and
shear modulus in the thickness direction are 15 GPa and 1.7 GPa respectively.
Fig. 1. WFC Specimen with Delamination.
Table 1. Properties of SP-5H PW Transducer [27].
Properties
Piezoelectric coupling coefficients
Kp
K33
Piezoelectric charge constants
d33 × 10-12 C/N
d31 × 10-12 C/N
Piezoelectric voltage constants
g33 × 10-3 Vm/N
g31 × 10-3 Vm/N
Relative dielectric constant
kT3
Density
ρ kg/m3
Elastic constants
SE11 × 10-12 m2/N
SE33 × 10-12 m2/N
Values
0.63
0.73
550
-265
19
-9
3100
7500
21
15
A digital storage oscilloscope (Tektronix 1002B), an arbitrary function generator (Tektronix
3021B), a high speed bipolar amplifier (NF BA4825), and a computer constitute the
experimental set up as shown schematically in Fig. 2. A sine wave tone burst having 8.5
cycles and windowed by Gaussian function is supplied to the actuator to produce Lamb waves
in the WFC laminate. The Gaussian window ensures less spectral leakage and the 8.5 cycles
give better time and frequency domain resolution. To ensure actuation and sensing of Lamb
waves in the WFC laminate, a few experiments are carried out at different frequencies on a
pristine specimen. A comparison of the group velocities (Cg) obtained through the
experiments and mathematical equations is carried out. As shown in Fig. 3, a good agreement
is observed in this comparison which confirms the effective generation and reception of
Lamb waves in the experiments.
Fig. 2. Schematic of the Experimental Setup.
Fig. 3. Comparison of Group Velocities (Cg) of Lamb Waves in the WFC Laminate
(o: A0 Mode, +: S0 Mode, ♦: Experimental).
Further, the experiments are carried out with the laminate having delamination. A 300 Vpp
tone burst is applied to the PW actuator at a frequency of 76.5 kHz. The frequency domain
results obtained are shown in Fig. 4. Two higher harmonics can be seen distinctly in this
figure, whereas no higher harmonics are seen in the case of an intact specimen.
Instrumentation nonlinearity can be said to be absent in the experimental setup as response
obtained in the case of intact specimen does not show any higher harmonic. This confirms
that the observed higher harmonics in the Lamb wave response are only because of the
contact nonlinearity at the delamination. FE simulation is also carried out to study the
nonlinear interaction between Lamb waves and delamination in a WFC laminate and is
explained in the following subsection.
Fig. 4. Experimental Frequency Response.
2.2 Finite Element Simulation
Lamb wave propagation in a WFC laminate is simulated in the FE environment using
ANSYS©. PLANE183 (8-node quadrilateral) and PLANE223 (coupled-field) elements are
used for modeling the WFC plate and PW patches respectively. The size selected for the
elements is 0.5 mm such that the mesh size is able to capture the higher harmonics. Newmark
algorithm is used for the time integration, and the selection of time step is based on the tradeoff between the desired accuracy and computational efforts [28]. Accordingly, a time step
size of 2e-7 s. is selected for the simulation. In detail explanation about the selection of
element and time step size for carrying out the FE simulation of Lamb wave propagation in a
plate is given in [23]. Master nodes are formed on the upper and lower faces of the PW
patches by coupling nodes on the respective faces. These master nodes are used for applying
and sensing voltage in the cases of actuator and sensor respectively. To confirm actuation and
sensing of Lamb waves in the simulation, a couple of simulations are carried out at a range of
different frequencies. A comparison of the group velocities obtained through the simulations
and mathematical analysis is carried out. A close agreement is observed in this comparison as
shown in Fig. 5 which ensures the effective generation and sensing of Lamb waves in the
simulation.
The challenge faced in this FE analysis of Lamb wave propagation across the delamination is
modeling the breathing delamination such that it can closely approximate the physical
opening and closing of the delaminated faces as the wave passes. To model the contact
nonlinearity at the delamination, two delaminated faces are assigned contact elements which
are CONTA172 (3-node surface-to-surface contact) and TARGE169 (target segment)
elements respectively. The algorithm used here to solve the contact problem is the
Augmented Lagrangian (AL) method. The AL method is an iterative series of penalty updates
to find the contact tractions. This algorithm is advantageous as it gives better penetration
control, satisfies constraints with the finite penalties, and avoids ill conditioning of the
governing equations [29]. Friction is assumed to be absent in the present contact problem. In
such a frictionless condition, the contact pressure J is defined as [22]
𝐽=0
if 𝑢𝑚 > 0 and
(1)
𝐽 = 𝐾𝑚 𝑢𝑚 + 𝜆𝑖+1
if 𝑢𝑚 ≤ 0,
(2)
where,
𝜆𝑖+1 = 𝜆𝑖 + 𝐾𝑚 𝑢𝑚
if |𝑢𝑚 | > 𝑧 and
(3)
𝜆𝑖+1 = 𝜆𝑖
if |𝑢𝑚 | < 𝑧,
(4)
and 𝐾𝑚 is the normal contact stiffness, 𝑢𝑚 is the contact gap size, 𝜆𝑖 is the Lagrange
multiplier at the ith iteration, and z is the compatibility tolerance. The Lagrange multiplier
component 𝜆𝑖 is computed locally, i.e., for each element and iteratively. In AL treatment of
the frictionless contact of delaminated faces, the admissible deformation of prospective points
of the contact satisfies [22]
𝑟(𝑥) ≤ 0,
𝑡𝑁 = −𝑛(𝑥) ∙ 𝑄𝑛′ ≥ 0,
𝑡𝑁 (𝑥)𝑟(𝑥) = 0, and
𝑡𝑇 = 𝑄𝑛′ + 𝑡𝑁 𝑛 = 0,
(5)
(6)
(7)
(8)
where 𝑟 is a scalar-valued gage function, n is the outward normal in the current configuration,
𝑛′ is the outward normal in the reference configuration, and Q is the first Piola-Kirchoff
stress tensor. Eq. 5 represents the condition of impenetrability, Eq. 6 represents the constraint
that the normal component of surface traction (𝑡𝑁 ) be compressive when contact is made, and
Eq. 7 is a condition ensuring that 𝑡𝑁 may only be nonzero when r(x) = 0. Eqs. 5–7 are
therefore recognized as the Kuhn-Tucker conditions [22]. Eq. 8 affirms that no friction is
present as tangential component of the surface traction (𝑡𝑇 ) is imposed to be zero. Maximum
allowable penetration considered in the present study is 0.1 times the thickness of the
element. This is because, its large value makes the AL method to work like the penalty
method, which is not desired here.
Fig. 5. Comparison of Group Velocities (Cg) of Lamb Wave in the WFC Laminate
(o: A0 Mode, +: S0 Mode, ▲: FE Simulation).
The PW actuator is given the same voltage as that mentioned in the experiential study. The
frequency domain results are shown in Fig. 6. This figure shows the presence of two distinct
higher harmonics similar to that observed in the experiments. The observed higher harmonics
in the Lamb wave response are only because of the contact nonlinearity modeled at the
delamination as response obtained in the case of an intact specimen does not show any higher
harmonic as shown in Fig. 6. The new hybrid method developed to locate a delamination is
explained in the following section.
Fig. 6. FE Simulation Frequency Response.
3. Locating Delamination in a Composite Laminate
The hybrid method which uses both frequency and time domain data to locate a delamination
in composite laminates is explained in this section. In fact, this method can be used to locate
any damage which can induce contact nonlinearity in the propagating wave and thereby
produce higher harmonics. The procedure of the method is explained here through the
experiment. A special arrangement of the PW transducers is made on the WFC laminate as
shown in Fig. 7. Two PW actuators are mounted on the WFC laminate back to back in a
collocated way. In Fig. 7 only PW actuator 1 can be seen. However PW actuator 2 is below
the actuator 1 and cannot be seen in the figure. A PW sensor is mounted on the laminate at a
known distance x1 from the actuators. The delamination is located at a distance x2 from the
sensor and x from the actuators respectively. The dimension of interest here is x.
Fig. 7. WFC Laminate with PW Transducers and Delamination.
The PW actuators 1 and 2 are supplied with voltages of opposite sense to produce only A0
mode of Lamb waves. S0 mode has very high group velocity as shown in Fig. 3, and therefore
it undergoes multiple reflections. This makes the response signal cluttered. Thus S0 mode is
avoided here. Also, Fig. 3 shows that the velocity dispersion plot of S0 mode is not flat in the
operational frequency range 0.1 MHz mm to 1 MHz mm. As a result of this, the fundamental
and higher harmonics of S0 mode travel with different velocities. However the dispersion
curve of A0 mode is flat from 0.1 MHz mm to 1 MHz mm as shown in Fig. 3, and therefore
the fundamental and higher harmonics of A0 mode have the same group velocity. This is a
necessary requirement of this method and will be discussed in the subsequent explanation.
Lamb wave A0 mode produced by the actuators passes through the sensor. Further, it interacts
with the delamination and produces higher harmonics as a result of contact nonlinearity.
These higher harmonics propagate in both forward and backward direction. The harmonics
which travel backward are again picked up by the sensor. The time domain data received at
the sensor is shown in Fig. 8. It contains the input pulse of A0 mode which is seen first, some
reflections, and the higher harmonics propagated backward from the delamination. Let us say
A0 mode appears at time t1 in this time data. Among all the higher harmonics the second
harmonic is of the considerable amplitude. Thus, a filter is designed to extract only second
harmonic from the total time domain data. It can be seen in Fig. 9 that the time domain data
of the second harmonic shows many packets because the wave-delamination nonlinear
interaction occurs over the entire width of the delamination. Let us say the first packet
appears at time t2. The time scale of both the total and second harmonics data is the same.
Fig. 8. Lamb Wave Signal Picked up at the Sensor (Experimental).
Another thing to be noted here is that the second harmonic is produced when A0 mode
reaches the delamination. Thus, the total time required for A0 mode to reach the delamination
from the sensor and that required for the second harmonic to reach the sensor from the
delamination can be given by Δt = t2 - t1. The group velocities of the fundamental and second
harmonic are same as shown in Fig. 3, and let us call it Cg. The product CgΔt gives the
distance 2x2. The delamination location then can be given by x = x1 + x2. The delamination
location is found to be 74.65 mm in the experiments and 73.8 mm in the FE simulation. The
delamination locations obtained through both the experiments and FE simulation fall well in
the span of delamination width. This method is simple and robust to apply.
Fig. 9. Lamb Wave Second Harmonics Picked up at the Sensor (Experimental).
4. Conclusions
Higher harmonics are seen in the Lamb wave response of a woven fiber composite laminate
with a delamination. However the Lamb wave response obtained in the case of an intact
laminate does not show any higher harmonic which obviates the possibility of presence of
instrumentation nonlinearity in the experimental set up. Therefore it can be concluded that the
higher harmonics observed in the response are solely because of contact nonlinearity at the
delamination. Further, the delamination is located using the new hybrid method which makes
use of both the frequency and time domain information. This method gives the delamination
location which is close enough to the actual one. The present method is capable of locating a
delamination in one dimension because it uses only one pair of collocated actuators and
sensor. However, with the help of such multiple pairs of transducers, it is possible to locate a
single (or multiple) delamination(s) in two dimensions.
Acknowledgment
The authors gratefully acknowledge the financial support by the Aeronautics Research and
Development Board, Govt. of India for this research work (Project No.1642).
References
1.
2.
3.
4.
5.
Z. Su, L. Ye, and Y. Lu, “Guided Lamb waves for identification of damage in composite
structures: A review”, J. Sound Vib., vol. 295, nos. 3-5, pp. 753–780, 2006.
K. Y. Jhang, “Nonlinear ultrasonic techniques for non-destructive assessment of micro
damage in material: A review”, Int. J. Prec. Eng. Man., vol. 10, no. 1, pp. 123–135, 2009.
P. Cawley and D. Alleyne, “The use of Lamb waves for the long range inspection of
large structures”, Ultrasonics, vol. 34, nos. 2-5, pp. 287–290, 1996.
F. Levent Degertekin and B. T. Khuri-Yakub, “Lamb wave excitation by hertzian
contacts with applications in NDE”, IEEE T. Ultrason. Ferr., vol. 44, no. 4, pp. 769–779,
1997.
S. S. Kessler, S. M. Spearing, and C. Soutis, “Damage detection in composite materials
using Lamb wave methods”, Smart Mater. Struct., vol. 11, no. 2, pp. 269–278, 2002.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
Z. Su and L. Ye, “Fundamental Lamb mode-based delamination detection for CF/EP
composite laminates using distributed piezoelectrics”, Struct. Health Monit., vol. 3, no.
1, pp. 43–68, 2004.
C. Soutis and K. Diamanti, “Active sensing of impact damage in composite sandwich
panels by low frequency Lamb waves”, Aeronaut. J., vol. 112, no. 1131, pp. 279–283,
2008.
S. H. Daz Valds and C. Soutis, “Real-time non- destructive evaluation of fiber composite
laminates using low-frequency Lamb waves”, J. Acoust. Soc. Am., vol. 111, no. 5, pp.
2026–2033, 2002.
W. Lin, L. S. Goh, and B. S. Wong, “A new Lamb-wave based NDT system for
detection and identification of defects in composites”, Proc. Singapore International
NDT Conference and Exhibition, Singapore, 2013.
R. Gangadharan, C. R. L. Murthy, S. Gopalakrishnan, M. R. Bhat, and D. Roy
Mahapatra, “Characterization of cracks and delaminations using PWAS ad Lamb wave
based time-frequency methods”, Int. J. Smart Sens. Intell. Syst., vol. 3, no. 4, pp. 703–
735, 2010.
Y. Okabe, K. Fujibayashi, M. Shimazaki, H. Soejima, and T. Ogisu, “Delamination
detection in composite laminates using dispersion change based on mode conversion of
Lamb waves”, Smart Mater. Struct., vol. 19, no. 11, 2010.
C. Ramadas, K. Balasubramaniam, M. Joshi, and C. V. Krishnamurthy, “Interaction of
guided Lamb waves with an asymmetrically located delamination in a laminated
composite plate”, Smart Mater. Struct., vol. 19, no. 6, 2010.
H. Sohn, H. W. Park, K. H. Law, and C. R. Farrar, “Damage detection in composite
plates by using an enhanced time reversal method”, J. Aerospace Eng., vol. 20, no. 3, pp.
141–151, 2007.
B. Poddar, C. R. Bijudas, M. Mitra, and P. M. Mujumdar, “Damage detection in a
woven-fabric composite laminate using time-reversed Lamb wave”, Struct. Health
Monit., vol. 11, no. 5, pp. 602–612, 2012.
P. Gudimetla, A. Kharidi, and P. K. D. V. Yarlagadda, “Simulation of delaminations in
composite laminates”, Proc. 6th International Conference on Precision, Meso, Micro and
Nano Engineering, Coimbatore, India, 2009.
D. Broda, W. J. Staszewski, A. Martowicz, T. Uhl, and V. V. Silberschmidt, “Modelling
of nonlinear crack-wave interactions for damage detection based on ultrasound - a
review”, J. Sound Vib., vol. 333, no. 4, pp. 1097–1118, 2014.
C. Bermes, J. Y. Kim, J. Qu, and L. J. Jacobs, “Nonlinear Lamb waves for the detection
of material nonlinearity” Mech. Syst. Signal Pr., vol. 22, no. 3, pp. 638–646, 2008.
M. Deng and J. Yang, “Characterization of elastic anisotropy of a solid plate using
nonlinear Lamb wave approach” J. Sound Vib., vol. 308, nos. 1-2, pp. 201–211, 2007.
C. Pruell, J. Y. Kim, J. Qu, and L. J. Jacobs, “Evaluation of plasticity driven material
damage using Lamb waves” Appl. Phys. Lett., vol. 91, no. 23, 2007.
C. Pruell, J. Y. Kim, J. Qu, and L. J. Jacobs, “Evaluation of fatigue damage using
nonlinear guided waves”, Smart Mater. Struct., vol. 18, no. 3, 2009.
M. Deng and J. Pei, “Assessment of accumulated fatigue damage in solid plates using
nonlinear Lamb wave approach”, Appl. Phys. Lett., vol. 90, no. 12, 2007.
22. N. P. Yelve, M. Mitra, and P. M. Mujumdar, “Spectral damage index for estimation of
breathing crack depth in an aluminum plate using nonlinear Lamb wave”, Struct. Control
Health Monit., vol. 21, no. 5, pp. 833–846, 2014
23. N. P. Yelve, M. Mitra, and P. M. Mujumdar, “Higher harmonics induced in Lamb wave
due to partial debonding of piezoelectric wafer transducers”, NDT&E Int., vol. 63, pp.
21–27, 2014.
24. G. Shkerdin and C. Glorieux, “Nonlinear modulation of Lamb modes by clapping
delamination”, J. Acoust. Soc. Am., vol. 124, no. 6, pp. 3397–3409, 2009.
25. B. Sarens, B. Verstraeten, C. Glorieux, G. Kalogiannakis, and D. Hemelrijck,
“Investigation of contact acoustic nonlinearity in delaminations by shearographic
imaging, laser doppler vibrometric scanning and finite difference modelling”, IEEE T.
Ultrason. Ferr., vol. 57, no. 6, pp. 1383–1395, 2010.
26. N. K. Naik, Y. C. Sekher, and S. Meduri, “Polymer matrix woven fabric composites
subjected to low velocity impact: Part I - damage initiation studies”, J. Reinf. Plast.
Comp., vol. 19, no. 12, pp. 912–954, 2000.
27. http://www.sparklerceramics.com/piezoelectricproperties.html.
28. K. J. Bathe, Finite Element Procedures in Engineering Analysis, Prentice-Hall Inc., New
Jersey, 1982.
29. J. C. Simo and T. A. Laursen, “An Augmented Lagrangian treatment of contact problems
involving friction”, Comput. Struct., vol. 42, no. 1, pp. 97–116, 1992.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Investigation of anisotropic propagation of shear horizontal modes in
composite laminates using fiber Bragg grating sensors
Pabitro Ray 1,2, Prabhu Rajagopal 2, Balaji Srinivasan 1 , Krishnan Balasubramaniam 2
2
Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai 600 036, India
Centre for Non-Destructive Evaluation, Department of Mechanical Engineering, Indian Institute of Technology
Madras, Chennai 600 036, India
1
Email: balajis@ee.iitm.ac.in
Abstract
We investigated the effect of carbon fiber orientation in a composite laminate on the propagation of fundamental
shear horizontal guided wave mode, using fiber Bragg grating sensors. Our experimental results are in agreement
with the results reported in the literature.
Keywords: Fiber Bragg grating, Ultrasonic guided mode, Shear Horizontal mode, Composite laminates
1. Introduction
Health monitoring of reinforced composite laminates is a critical requirement in aerospace
vehicles [1]. Ultrasonic guided waves are widely used for such applications because of their
ability to propagate over relatively long distances and detect both internal and surface defects
[2]. However, the anisotropic properties of each lamina makes conventional guided-wave
based inspection of composites challenging. Specifically, it has been shown that the waves are
guided preferentially along the direction of the reinforcement fiber [3]. We have previously
demonstrated the ability of fiber Bragg grating (FBG) sensors to detect specific guided waves
with different orientations in aluminum plates due to their highly directional response [4].
FBGs have the additional advantage that they can be embedded to the composite structures
during their manufacturing process, making them quite attractive for in-situ monitoring
applications.
In this paper, we extend the above work to the detection of fundamental shear
horizontal (SH0) waves in carbon fiber reinforced polymer (CFRP) laminates and demonstrate
the influence of fiber orientation on SH mode energy propagation. Our experimental results
show that the attenuation of SH mode is more for waves propagating across the carbon fiber
orientation as compared to modes propagating along the carbon fiber orientation.
The paper is organised as follows: we describe our approach for designing the
experimental setup, followed by results, discussion, overall summary of the work and
directions for future work.
2. Approach
Initial experiments were conducted on a 4-ply unidirectional CFRP laminate with stacking
thickness of 2 mm. The experiments are conducted in two parts, where a shear excitation is
provided at different locations with SH mode energy propagating along and across the CFRP
fiber orientation. A schematic of the experimental set up is shown in the Fig. 1.
Circulator
TLS
Analog Circuit Board
Band Pass filter
5 kHz – 1.25 MHz
APD - TIA
Receiver
FBG 2
90̊
0̊
FBG 1 To Shear Probe
T3 T2 T1
Shear Probe
RITEC PulserReceiver
Sync
DSO
CFRP 0/0/0/0
Fig1. Schematic of Experimental Setup
Two FBG sensors (FBG 1 and FBG 2) are pasted on to the CFRP laminate for
capturing SH0 modes propagating along θ = 0̊ and θ = 90̊ respectively. Shear excitation is
provided at different locations each separated by 5 cm from the FBG. Fig. 1 shows the
transverse orientation of FBG 1 such that it preferentially picks up shear modes of transducers
T1, T2 and T3. Similar approach is followed for the other case with waves propagating across
the fiber orientation (θ = 90̊) and picked up using FBG 2. Shear waves are generated using a
commercial shear transducer by exciting it with a 5-cycle Hanning windowed tone-burst
centred at a frequency of 200 kHz using a RITEC (RPR 4000, Ritec Inc., USA) pulser
receiver
In this work, tunable laser source (TLS) – FBG based interrogation technique is used
since it offers very high sensitivity [5]. To achieve linear response, the TLS wavelength is
tuned to the linear region of the slope of the FBG reflection spectrum. The sensor undergoes
changes in its Bragg wavelength as a result of the dynamic strain. Since the laser is tuned to
the slope of FBG reflection spectrum, the changes in the Bragg wavelength get converted to
corresponding optical intensity variations, which is then detected using an avalanchephotodiode – trans-impedance amplifier module (APD-TIA-12/2 oemarket.com). The
electrical output from the analog circuit is then observed and recorded on a digital storage
oscilloscope (DSO, Keysight DSO-X 2024A) which operates at a sampling rate of 2 GHz.
3. Results and Discussion
As discussed above, we aim to study the propagation of shear horizontal (SH) along and
perpendicular to the composite fiber direction using fiber Bragg grating sensors. The time
trace captured by FBG 1 for shear excitation of 200 kHz, placed 5, 10 and 15 cm away from
the sensor is shown in Fig 2.
0.5
A m plitude (V )
A m plitude (V )
0.5
0
-0.5
0
10
20
30
40
50
60
Time (s)
70
80
90
0
-0.5
100
0
10
20
30
40
(a)
50
60
Time (s)
70
80
90
100
(b)
A mplitude (V )
0.5
0
-0.5
0
10
20
30
40
50
60
Time (s)
70
80
90
100
(c)
Fig 2. Time trace captured using FBG for Shear excitation provided at (a) 5 cm (b) 10 cm, and (c) 15
cm respectively from the FBG 1 location
From the above Figure, we observe that the mode amplitude decays rapidly with increase in
distance thereby showing the high attenuation of SH0 modes in CFRP laminates. Further set
of experiments were performed to see the effect of anisotropy on SH 0 mode attenuation. In
this case, the shear waves were captured using FBG 2 and are shown in Fig 3 below.
(a)
(b)
(c)
Fig 2. Time trace captured using FBG for Shear excitation provided at (a) 5 cm (b) 10 cm, and (c) 15
cm respectively from the FBG 2 location
Fig. 3 shows that the decay of mode amplitude to be higher as compared to the previous case,
From Fig. 2 and Fig 3, the wave amplitude along the CFRP fiber orientation is found to be 3
times stronger than that along the perpendicular direction. This observation is in agreement to
results reported in the literature [3].
4. Conclusion and Future Work
In this work we demonstrated the use of fiber Bragg grating sensors to study the influence of
composite fiber orientation on the shear horizontal mode energy propagation. Experiments
confirm that the mode energy attenuation is more when it propagates across the fiber
orientation as compared to its propagation along the fiber orientation. Ongoing work is
focused on extending the use of FBGs for health monitoring of multi-layer composites and
de-lamination.
References
[1] J Krautkramer and H Krautkramer, [Ultrasonic Testing of Materials], Springer Verlag, Berlin,
(1969)
[2] Zhongqing, Su and Lin, Ye, Identification of Damage Using Lamb Waves, Springer London,
London (2009)
[3] X Yu, M Ratassepp, P Rajagopal and Z Fan, “Anisotropic effects on ultrasonic guided waves
propagation in composite bends”, Ultrasonics, Volume 72, Pages 95-105, December 2016, ISSN
0041-624X, http://dx.doi.org/10.1016/j.ultras.2016.07.016. (2016)
[4] AV Harish, P Ray, P Rajagopal, K Balasubramaniam and B Srinivasan, “Detection of fundamental
shear horizontal mode in plates using fibre Bragg gratings”, Journal of Intelligent Material Systems
and Structures, Volume 27, Issue 16, Pages 2229-2236 (2016)
[5] B Lissak, A Arie and M Tur, “Highly sensitive dynamic strain measurements by locking lasers to
fiber Bragg gratings,” Opt. Lett., 23 (24), 1930-1932 (1998)
Damage Detection using Nonlinear Interaction of Guided Wave with Breathing
Crack in 1-D Beam
Mira Mitra
Department of Aerospace Engineering,
Indian Institute of Technology Bombay, Mumbai, India
Contact: mira@aero.iitb.ac.in
In the last two decades guided wave propagation has been established as an efficient technique for
structural health monitoring (SHM) of aerospace structures. Substantial work has been reported on
various aspects of wave based SHM, however, they are primarily focused on exploring the linear
behavior of wave propagation and their interaction with damage. This includes change in velocity,
amplitude, mode conversion, pulse-echo and others. Recently, interests are been directed towards
studying non-linear interaction of wave with damages and its application to damage detection. Such
interaction gives rise to generation of higher/sub-harmonics, frequency mixing, frequency shift and
others.
This work presents a semi-analytical study to simulate wave propagation in beam with breathing
crack. The modeling is based on spectral finite element method and a piece-wise contact non-linearity
of the breathing crack. First, a forward problem is studied to understand the nonlinear interactions
generating higher harmonics and frequency mixing. A parametric study is presented to bring out the
influence of crack-severity and -location on the extent of harmonic separation and on the relative
strength of higher order harmonic. Next, the developed model is used to tackle the inverse problem of
crack identification from the non-linear parameters of the simulated response. The results corroborate
the potential of non-linear guided wave techniques for structural health monitoring of aircraft
structures.
Corrosion Detection on Aerospace Structural Materials by Lamb Wave
Visualization Method – Advantages & Challenges
Aparna Aradhye, S Kalyana Sundaram
Structural Technologies Division,
CSIR-National Aerospace Laboratories,
Bangalore, India
Contact: kalyanasunder@gmail.com
Aircraft structural components fabricated with metallic materials frequently encounter corrosion
damages due to their exposure to saline and chemically aggressive environment during routine
operation. Detection of corrosion in the structure is a continuously increasing demand in aircraft
industry. Corrosion detection with conventional NDE method always poses challenges due to uneven
spread of hidden and subsurface corrosion sites on airframe. Hence, exploring new NDE methods to
detect corrosion is a continuous research process worldwide. Investigation presented here is an
experimental work done with Lamb wave visualization technique for detection of subsurface hidden
corrosion elucidating its advantages, challenges & limitations.
Study was carried out on the specimens fabricated from aluminum alloy 2024 grade sheets of
thickness 1.6 mm. Corrosion sites of different size, shape and depth were simulated on the specimens
by chemical etching with hydrochloric acid. Lamb wave based experiments were carried out with
Scanning Laser Doppler Vibrometer of model Polytec – 400 having integral signal generator
synchronized with laser detector. Waves were generated on the specimen using bonded piezo actuator
excited with tone burst of 5.5 cycle Hanning windowed signal. Visualization of the Lamb wave
propagation on the specimens has been done with time synchronized display of acquired data from
individual grid points. Observation of mode conversion at the corrosion sites, diffraction and
scattering are discussed in this paper along with the advantages, limitations and difficulties in
detection. Evaluation was carried out to determine the detection thresholds on frequency of Lamb
waves, corrosion depth, shape, size.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Laser based Non-Contact method for Nonlinear Mixing of
Rayleigh waves
Chaitanya BAKRE 1, Prabhu RAJAGOPAL, Krishnan BALASUBRAMANIAM
Centre for Non-Destructive Evaluation & Department of Mechanical Engineering,
Indian Institute of Technology Madras, Chennai, India. 600036.
1
Phone: +91 9790464440; e-mail: chitbak@gmail.com
Abstract
This study presents the nonlinear mixing technique of two co-directionally travelling Rayleigh surface waves
generated and detected using laser ultrasonics. The optical generation of Rayleigh waves on the specimen is
obtained by shadow mask method. In conventional nonlinear measurements, the inherently small higher
harmonics are greatly influenced by the nonlinearites caused by coupling variabilities and surface roughness
between the transducer and specimen interface. The proposed technique is completely contactless and it should
be possible to eliminate this problem. Moreover, the nonlinear mixing phenomenon yields not only the second
harmonics, but also the sum and difference frequency components, which can be used to measure the acoustic
nonlinearity of the specimen. In this paper, we will be addressing the experimental configurations for this
technique and the finite element analysis to characterize the acoustic nonlinearity of the Aluminum 7075 alloy
specimen.
Keywords: Acoustic nonlinearity, Laser Utrasonics, Rayleigh waves
1. Introduction
In recent years, nonlinear ultrasonics has been an area of interest for researchers mainly
because of its ability to detect damage at the microstructural level as opposed to linear
ultrasonics, which is limited to detection of large scale damages like macro cracks [1]. In the
early state of damage, the dislocation density is affected, resulting in an increase in the
material nonlinearity, which has negligible influence on the basic elastic properties of the
material but there are significant changes in the third order elastic properties. In nonlinear
acoustic measurements, higher harmonics are generated because of the inherently present
material nonlinearity as well as the dislocations and precipitates present at the microstructural
level which cause spectral energy transfer due to the deformation of the waveform [2]. The
generation of higher harmonics can be related to the microstructural evolution [3].
Although, nonlinear ultrasonics presents a great potential in detecting the damages at an early
stage, it suffers from many challenges due to its high sensitivity. A typical method for the
generation and detection of an ultrasonic wave is by the use of piezoelectric transducers.
Electrical components such as amplifiers, cables etc. incorporate nonlinearity in addition to
the material nonlinearity. Moreover, additional nonlinearity is also incorporated in the
measurements due to the variability in the coupling and the surface roughness between
transducer-wedge and wedge-specimen interfaces. These factors influence the accurate
measurements of the inherently small higher harmonic amplitudes, jeopardizing the
repeatability of the experimental output [4] [5]. The background nonlinearities because of
electrical components can be reasonably filtered out by signal processing techniques as they
remain constant with the input power setting, but the coupling nonlinearities are variable in
nature. Thus more reliable techniques need to be developed in order to prudently deal with the
background nonlinearities. Non-contact methods presents potential to deal with such
problems.
In the last two decades, noncontact ultrasonic methods such a Laser ultrasonics, EMAT’s, aircoupled ultrasonics have found growing applications in NDE. Noncontact generation and
detection methods demonstrate potential to disregard these influences and provide advantages
in the applicability for in-situ testing and online monitoring in hazardous and hostile
environments [6]. Perhaps, until now their application in nonlinear acoustic measurements is
not explored to a great extent. Laser generation and detection of ultrasonic waves demonstrate
many advantages over EMAT’s which are limited to conducting specimens; and air coupled
transducers which requires to be placed in the proximity to the surface of the specimen and
their frequencies are limited up to 4 MHz. Moreover, Lasers do not require a plane or a
curved surface unlike wedge-transducers. Thus irregular contours can also be investigated by
lasers.
A more sophisticated method, known as nonlinear wave-mixing technique prevails in which
two initially monochromatic waves of significant amplitude interact with each other resulting
in generation of not only the second harmonics but also the sum and the difference component
frequency waves [7]. Contactless broadband receivers such as Laser Vibrometer are a good
choice in this case of wide frequency range. This mixing phenomenon occurs on the account
of the material nonlinearity and the wide range of the generated frequencies can provide more
physical information as compared to a single frequency to determine the state of the material.
In this paper, we demonstrate a new technique for nonlinear mixing of Rayleigh waves and
the paper is organized in the following manner. First, a 3D elastodynamic FE model is
implemented to study the physics of nonlinear wave mixing of Rayleigh waves. Secondly, an
experimental methodology for the generation of two co-directional Rayleigh waves using the
slit-mask method is proposed. This is followed by the concluding remarks and the potential
application of this technique for online monitoring is discussed.
2. Finite element analysis
A numerical analysis is conducted to affirm the physics of generation of nonlinear mixed
Rayleigh waves owing to the material nonlinearity. For this purpose, a 3D elastodynamic
model is developed using COMSOLTM Multiphysics 5.2. To reduce the computation time, the
model dimensions are reduced to 10(mm) x 40(mm) x 20(mm). Hyperelastic material module
is chosen and the literature values of Murunaghan’s third order elastic constants for AA 7075
are used to incorporate material nonlinearity in the model, refer Table 1. Two Hanning
windowed inputs of 2 MHz and 3 MHz frequencies, 200 µm apart, are used for the point
excitation of Rayleigh waves at its critical angle. PARADISO solver is used to obtain the
solution. A 3D visualization of the surface displacement field in the out-of-plane direction is
shown in Figure 3 (a). Out of plane displacements are recorded on the surface at 3 mm
distance from the source. The time domain signal is shown in Figure 3 (b), frequency domain
analysis of which shows generation of mixed nonlinear Rayleigh waves, 1 MHz (Difference),
2 MHz (Fundamental #1), 3 MHz (Fundamental #2), 4 MHz (Second harmonic #1), 5 MHz
(Sum) and 6 MHz (Second harmonic #2).
FIGURE. 1. (a) A 3D visualization of elastodynamic model of out of plane displacement components of mixed
nonlinear Rayleigh waves. (b) A time domain plot in the far field region representing out of plane displacements
of the mixed waves on the surface. (c) A frequency domain plot of the received time domain signal (simulation)
showing fundamental, second harmonics, sum and difference wave components.
TABLE 1. Elastic Parameters of Aluminum 7075 [11].
Values
E
7.08729.
σ
0.337224
ρ
2700
Description
Young’s modulus
Poisson’s ratio
mass density
A
-3.512 .
Landau’s constant
B
-1.494 .
Landau’s constant
C
-1.028 .
Landau’s constant
3. Experimental methodologies
A schematic of the experimental setup for nonlinear mixing of Rayleigh waves by laser
ultrasonic method is shown in Figure 1. A 1064 nm Nd-YAG Litron laser with maximum
energy 600 mJ and 6 mm beam diameter was used for generation. A special combinational
dual frequency slit-mask, made from a 0.5 mm thin copper sheet, was designed to co-generate
narrowband Rayleigh waves of 3 MHz and 5 MHz frequencies that travels along the surface
of the specimen in the same direction. The two co-directionally traveling Rayleigh waves
interact with each other owing to the material nonlinearity and the generated mixed waves are
picked up by a Polytec OFV 505 broadband Laser Vibrometer. The time domain signal is
obtained on an Oscilloscope (Keysight Technologies, maximum sampling rate 5 GHz) and
recorded for further signal processing in MATLAB®. The experiments are performed on an
undamaged Aluminum 7075 alloy rectangular block of dimensions 300 mm × 150 mm × 20
mm.
FIGURE 2. Schematic of the experimental setup.
The pulse width of the generated signal depends on the slit width and the laser spot size. In
order to excite a narrowband wave in the frequency domain a long signal with higher number
of cycles is desired. For this purpose, the laser beam is expanded using a cylindrical lens from
Edmund Optics Inc., Part #NT48-360 to obtain the desired spatial size at the cost of energy
amplitude. The expanded laser beam is allowed to impinge on the combinational slit mask
creating line-arrayed sources for the 3 MHz and 5 MHz frequencies on the surface of the
specimen, spatially separated by 200 µm.
FIGURE 3. Examples of various possible sit-mask designs.
The proposed technique of nonlinear mixing of waves using lasers can be further extended to
employ the nonlinear mixing phenomenon to Lamb waves, Guided waves etc. by modifying
the slit-mask designs. Figure 3 shows various examples of slit-mask designs. Depending on
the material under inspection and its geometry, the orientation of the slit patterns can be
modified to achieve nonlinear mixing of the choice of ultrasonic waves. In this paper, we have
demonstrated the primitive (first) slit-mask design which allows co-generation of dual
frequency Rayleigh waves travelling in the same direction.
4. Conclusion
This work demonstrates a noncontact laser based technique for nonlinear mixing of Rayleigh
waves. The technique presents the following advantages;



Noncontact technique; enabling inspection of irregular contours.
Ability of Rayleigh waves to retain finite displacement amplitudes over long
propagation distances.
Nonlinear wave mixing; generation of higher harmonics which can detect
microstructural damage at early stages, unlike conventional ultrasonic testing.
Acknowledgements
The authors wish to thank the Board of Research in Nuclear Sciences (BRNS), Mumbai, India
for their financial support towards this work. Authors also gratefully acknowledge Prof.
Nilesh J. Vasa, Department of Engineering Design, IIT- Madras, for his immense support,
access to facilities and guidance during the course of the work.
References
1. J. Achenbach, Wave Propagation In Elastic Solids (North-Holland Pub. Co.,
Amsterdam, 1973).
2. L. Zarembo and V. Krasil'nikov, Soviet Physics Uspekhi 13, (1971).
3. S. Liu, S. Best, S. Neild, A. Croxford and Z. Zhou, NDT & E International 48, (2012).
4. C. Scruby and L. Drain, Laser Ultrasonics (A. Hilger, Bristol, England, 1990).
5. T. Tanaka and Y. Izawa, Jpn. J. Appl. Phys. 40, (2001).
6. T. Graf, Laser (Springer Vieweg, Wiesbaden, 2015).
7. V. Jaya Rao, E. Kannan, R. Prakash and K. Balasubramaniam, J. Appl. Phys. 104,
(2008).
8. M. Morlock, J. Kim, L. Jacobs and J. Qu, The Journal Of The Acoustical Society Of
America 137, (2015).
9. S. Kenderian, B. Djordjevic and R. Green, The Journal Of The Acoustical Society Of
America 113, (2003).
10. A. Lomonosov, V. Mikhalevich, P. Hess, E. Knight, M. Hamilton and E.
Zabolotskaya, The Journal Of The Acoustical Society Of America 105, (1999).
11. G. Jones, The Journal Of The Acoustical Society Of America 35, (1963).
12. N. Kalayanasundaram, International Journal Of Engineering Science 19, (1981).
13. Optical Generation Of Tone-Burst Rayleigh Surface Waves For Nonlinear Ultrasonic
Measurements, Graduate, Georgia Institute of Technology, 2012.
14. J. Kyung-Young, N. Taehyung, C. Sungho, L. Taehun and K. Chung Seok, J. Korean
Phys. Soc. 57, (2010).
15. F. Di Scalea, T. Berndt, J. Spicer and B. Djordjevic, IEEE Transactions On
Ultrasonics, Ferroelectrics And Frequency Control 46, (1999).
16. in Quantitative Nondestructive Evaluation (Plenum Press, New York, 1995), pp. 593600.
17. E. Zabolotskaya, The Journal Of The Acoustical Society Of America 91, (1992).
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Localization of a breathing crack in a stepped beam using a method of
harmonic separation
Dhanashri M. JOGLEKAR
Department of Mechanical and Industrial Engineering, Indian Institute of Technology Roorkee, India;
Phone: +91 1332 284410; e-mail: dhanashri.joglekar.fme@iitr.ac.in
Abstract
The present study demonstrates a method of crack localization employing the time domain manifestation of the
nonlinear acoustics generated by guided wave - breathing crack interactions in a stepped beam. A method based
on wavelet spectral finite elements is used to simulate the propagation of flexural wave through a stepped beam
with a breathing crack. The separation of first and second harmonic at the sensor location, is related to the
location of crack with respect to sensor. The marginal errors in localizing the crack ascertain the applicability of
the method. As the developed technique uses the nonlinear acoustics generated by the cracks, the results are
specific to the crack and therefore can be positively used in presence of the other geometric discontinuities.
Keywords: Breathing crack, second harmonic, crack localization, stepped beam, wavelet spectral finite element
1. Introduction
The nonlinear interactions of propagating guided waves with damages generate peculiar
response characteristics, referred to as nonlinear acoustics[1]. Owing to their higher
sensitivity to the damages and the robustness, the usage of these nonlinear acoustics in health
monitoring techniques has attracted the focus of the present research in this domain.
Moreover, unlike their linear counterpart, these nonlinear acoustics prove to be beneficial for
the waveguides with other geometrical and material discontinuities. In this regards, the
present article studies the applicability of the nonlinear response characteristic in localizing a
crack in a stepped waveguide.
Although the guided waves are preferred over the bulk waves, owing to their ability to
scan larger area while maintaining the sensitivity to the defect size [2], the nonlinear
interactions of guided wave with crack are comparatively less explored, either experimentally
[3, 4] or theoretically [5–8]. Furthermore, almost all the aforementioned examples monitor the
frequency domain manifestation of nonlinear crack-wave interactions.
A few recent studies have demonstrated applicability of the time domain response
containing nonlinear acoustics for damage detection [6,9]. Joglekar and Mitra [6] have used
the separation of higher harmonics observed in time response for localizing the breathing
damage in a prismatic 1D beam structure. Boccardi et al. [9] used the arrival of second
harmonic signal to localize the defect in a composite laminate. Both these studies have
considered the waveguides to be free of other material and geometric discontinuities. The
present study extends the work of Joglekar and Mitra [6] to demonstrates the applicability of
the harmonic separation technique in localizing the breathing crack in stepped beams.
The remainder of the article is organized in three sections. Section 2 explains the
formulation and method of solutions. The results obtained are presented and discussed in the
Section 3. The damage localization method is explained in the latter part of this Section 3.
The following Section 4 concludes the article by stating the salient inferences drawn from the
present study.
2. Formulation and method of solution
The schematic of a slender stepped cantilever beam is shown in the Fig. 1. A breathing crack
of depth a is present in the beam at a length Lc from the free end. The beam changes its depth
at a location Ls units from left end. The excitation force f, in form a tone burst signal is
applied at a location La units away from the free end of the beam. The distance of the signal
receiver from the same end is Lr.
Lr
Ls
Lc
La
signal
receiver
f
Figure 1. Schematic of a stepped beam with a crack
To arrive at the solution of the displacement and velocity response at receiver end, firstly, the
breathing behaviour of the crack is approximated by an equivalent mathematical model as
explained in the following.
2.1 Equivalent crack model
The behaviour of the beam is assumed to be same as that of an intact beam, when the crack is
in a closed condition. In contrast, when the crack is in an open state, the localized reduction in
stiffness at the crack-location is modeled by a torsional spring acting between the two sections
adjacent to the crack surfaces. The stiffness Kr of this rotational spring is obtained as [10]
Ebh 2
(1)
Kr 
,
72 f a
h
where E represents the modulus of elasticity and f a
is a crack function. Thus the
h
problem of wave propagation analysis through the stepped beam with a breathing crack is
changed to analyzing a propagating wave through the beam with switches between aforesaid
two models depending on open or closed condition of crack faces. The solution is sought by
employing a WSFE method. Formulation of governing equations and the solution method is
described in the following section.
 
 
2.2 Formulation of finite element equations in transform domain
The temporal dependence of the transverse displacement variable w is approximated by a
series of compactly supported Daubechies wavelet scaling functions ϕ of the order N as the
following:
w   wˆ k    k 
kZ
(2)
k
Using this approximation and the orthogonal properties of the scaling functions, the
governing equation of motion for an intact free beam, following the Euler-Bernoulli
hypothesis, reduces to
d 4 wˆ j
dx 4
 A  1
 A 2

 wˆ j  
 wˆ j  0.
EI


 EI 
(3)
In this Eq. 3, I and A are the area moment of inertia and area of the cross section of the beam.
The term ρ represents the density of the beam material. 1Γ and 2Γ are the connection
coefficient matrices. Variable j takes the values from 0 to n-1, n representing the number of
translates of the scaling of functions.
Mapping the wavelet transform coefficients wˆ j on to the eigen-spaces of the connection
coefficient matrices, Eq. 3 is decoupled as the following:
d 4 w j
dx
4
  A 
 A 2
 
 i j  
  j  w j
 EI  
 EI 
(4)
where w is the displacement transform coefficient mapped on to the eigen-space of the
connection coefficient matrix, and  j is the eigenvalue of the same. Solutions of Eq. 4 are
used in formulating the dynamic stiffness matrix of WSFEs. The expression of the transverse
displacement in transform domain satisfying Eq. 4 is as the following:
w  C1e ik1x  C2 e ik2 x  C3 eik3x C4 e ik4 x (5)
where Ci’s are arbitrary constants and ki’s are roots of characteristic equation stated in Eq. 4.
The damaged beam structure is then divided into four WSFEs as shown in Fig 2. In which,
there are three healthy beam elements and one crack beam element . The first and third
elements are sandwiching the crack beam element. The last healthy element models the step
of the beam from length Ls to the right most end.
Figure 3. WSFE architecture of beam
In deriving the dynamic stiffness matrix of the healthy beam WSFE, the expression of
traverse displacement stated in Eq. 5 is used to arrive at interpolate functions for the nodal
dofs. The nodal dof vector consists oftransform coefficients of transverse displacement and
slopes. Thereafter, the essential and the natural boundary conditions are evaluated in the
transform domain. These two conditions take the following form:
u e  R1h a,
f e  R2 h a
(6)
e
e
Terms u and f are the nodal dof and force vectors, respectively. The superscript e is used to
represent the element level formulation, and h indicates the healthy (defect-free)
configuration. a is the vector of arbitrary coefficients. It is now possible to eliminate the vector
of arbitrary constants and relate the nodal dofs to nodal forces through the element dynamic
stiffness matrix Ke.
The second element in the assembly is a crack beam WSFE. To account for a local
reduction in stiffness, this element contains a massless rotational spring, connecting two
healthy beam segments as shown in Fig. 3.
Figure 3. Schematic of a crack beam element
The approximation functions for transform coefficients of the transverse displacements in
these two healthy beam segments have the same form as that of the solution of Eq. 4. Similar
to the case of a finite-length healthy element, the next step involves evaluation of natural and
essential boundary conditions at two ends of the crack element. However, since the element
carries a rotational spring in it, the continuity conditions at this location are also evaluated
along with the other two sets of conditions. The elimination of the vector of arbitrary
constants form these three sets leads to the dynamic stiffness matrix for the cracked beam
configuration.
The element level matrices for three elements are then assembled together to obtain the
global stiffness matrix corresponding to a single eigen space. Considering all such eigen
spaces simultaneously, the set of nonlinear algebraic equations for the cracked beam is written
as
K c u  fa  fc
(7)
where K c is the assembled dynamic stiffness matrix for the beam with an open crack, and u
is the vector of nodal dofs. fa represents the nodal force vector resulting from the externally
applied forces, whereas f represents the bilinear force in the transform domain arising from
c
the intermittent contact between the two crack surfaces.
2.2 Method of solution
The iterative procedure devised to solve Eq. 7 involves the following steps:
1.
A linear solution of Eq. 7, obtained by setting the term fc equal to zero is used as a
trial solution.
2.
The difference between the respective dofs of two successive nodes in time domain is
then determined.
3.
A state of the crack (open or closed) is then determined by multiplying the resulting
vector by the Heaviside function. This step yields a decision vector.
4.
The trial solution in transform domain is now multiplied by the difference between the
global stiffness matrix of the cracked configuration and that corresponding to the intact
configuration.
5.
Resulting vector is then transformed to the time domain and then multiplied by the
decision vector. This step gives the contact force vector in time domain, which is then mapped
to transform domain.
6.
The residue in satisfying the Eq. 7 is then obtained and the trail solution is modified to
minimize this error.
The results obtained by employing the proposed method of solution are discussed in the
following section.
3. Results and discussion
A exemplary 10 m long cantilever beam is considered to change its thickness from 2 cm to
2.5 cm at a location 5 m away from the free end. Width of the beam is 2.5 cm throughout the
length. The beam under consideration is assumed to be made up of a material for which the
density and the Young's modulus are equal to 2800 kg/m3 and 70 Gpa, respectively. A
diagnostic signal in form a modulated sinusoid with the central frequency of 25 kHz is
interrogated in the beam at 4 m away from free end. Figs. 4a and 4b show the frequency and
time domain representations of the excitation force.
(a)
(b)
Figure 4. Interrogated signal in (a) frequency and (b) time domain.
At the receiver location, 6 m away from free end, the response of the beam is determined by
employing the WSFE based method explained in the previous section. Daubechies wavelet
scaling functions of order 22 are used for temporal approximation. The time window of 1.6
ms is divided into 512 equal segments. The resulting velocity response in frequency as well
as time domain is shown in Figs. 5a and 5b respectively.
(a)
(b)
Figure 5. Velocity response at sensor location in (a) frequency and (b) time domain.
The existence of higher harmonics observed in Fig. 5a substantiates the presence of nonlinear
acoustics in the response. The time domain response in Fig. 5b however, conveys very little
information about nonlinear acoustics. Therefore, it is further processed to separate out
individual harmonics. The first and second harmonics filtered from this response are depicted
in Fig. 5.
Figure 6. Separation of first and second harmonics observed in the filtered response at sensor location.
The waveguide under consideration is idealized by the Euler Bernoulli beam theory, which
imparts dispersive nature to the waveguide. as a result, individual frequency components
travel with different velocities and are therefore received at different time at the same space
location. This is evident in the temporal separation of the first and second harmonics at the
receiver location. The correlation of this separation with crack location is explained in the
following.
3.1 Localization of a breathing crack
For the prismatic Euler Bernoulli beam, the group speed cg in terms of the frequency f is
written as,
EI 4 2 f 2
.
(8)
A
However, for the stepped beam, the group speed for the same frequency component will be
different in different parts of the beam. Therefore, for the single step beam under
consideration, the separation Δt can be written in terms of the length parameters and group
speeds as
cg  2 4
t 
 1
1
2
cg 2
 cg1
 Lr  Lc   2



 1
 1
1 
1
 Ls  Lc   1  1    Lr  Ls   2  2
cg 2 
cg 2
 c g1
 c g1
... LC  LS

 ... LC  LS

(9)
The terms 1cg1 and 1cg2, represent the group speeds of first and second harmonics in the first
step on left, whereas 2cg1 and 2cg2 are the group speeds of these frequency components in the
second step on right. By employing the group speeds of the first and second harmonics,
together with the observed separation in harmonics in Eq. 9, the crack location Lc can be
estimated. The prediction of crack location for the aforementioned case together with a
couple of other cases, is included in the following table. The excellent match shows the
potential of the harmonic separation method in damage localization.
Table 1.Case studies for crack localization
Cases
1
2
3
La
6
6
5.5
Ls
5
5
5
Actual Lc
4
4.5
4.5
Predicted Lc
4.0209
4.5086
4.5836
Percentage error
-0.5
-0.19
-1.86
Conclusion
The effect of nonlinear guided wave - crack interactions, in terms of generation of higher
harmonics is studied in time domain for a stwpped slender beam. An iterative method
employing wavelet spectral finite elements is used to simulate the propagation of flexural
wave through the stepped beam carrying a breathing crack in it.
An exemplary case of a stepped cantilever beam with a breathing crack shows the
existence of higher harmonics in the frequency domain representation of the velocity
response. The time domain representation of the same velocity response is analyzed using
signal filters, which reveal the separation of harmonics, owing to the dispersive nature of the
waveguide. The difference between the arrival time of the individual harmonics is used to
trace back the location of crack for the stepped beam structure. The marginal errors in
predicting the crack location with respect to sensor substantiate the applicability of method.
References
1.
2.
3.
4.
5.
D Broda, W J Staszewski, A Martowicz, T Uhl, and V V Silberschmidt, 'Modelling of
Nonlinear Crack-Wave Interactions for Damage Detection Based on Ultrasound - A
Review', Journal of Sound and Vibration, Vol. 333, No. 4, pp 1097–1118, 2014.
M Mitra and S Gopalakrishnan, 'Guided Wave Based Structural Health Monitoring: A
Review', Smart Materials and Structures, Vol. 25, No. 5, pp 053001, 2016.
M Ryles, F H Ngau, I McDonald, and W J Staszewski, 'Comparative Study of
Nonlinear Acoustic and Lamb Wave Techniques for Fatigue Crack Detection in
Metallic Structures', Fatigue & Fracture of Engineering Materials & Structures, Vol. 31
No. 8 pp 674–683, 2008.
N P Yelve, M Mitra, and P M Mujumdar, 'Spectral damage index for estimation of
breathing crack depth in an aluminum plate using nonlinear Lamb wave', Structural
Control and Health Monitoring, Vol. 21, No. 5, pp 833–846, 2014.
K Wang and Z Su, 'Analytical Modeling of Contact Acoustic Nonlinearity of Guided
Waves and its Application to Evaluating Severity of Fatigue Damage', SPIE
Proceedings of Smart Structures and Materials+ Nondestructive Evaluation and Health
Monitoring, pp. 98050L–98050L, April 2016.
6.
D M Joglekar and M Mitra, 'Analysis of Flexural Wave Propagation Through Beams
with a Breathing Crack Using Wavelet Spectral Finite Element Method', Mechanical
Systems and Signal Processing, Vol. 76, pp 576–591, 2016.
7. D M Joglekar and M Mitra, 'Nonlinear Analysis of Flexural Wave Propagation through
1D Waveguides with a Breathing Crack', Journal of Sound and Vibration, Vol. 344, pp
242–257, 2015.
8. D M Joglekar and M Mitra, 'Analysis of Nonlinear Frequency Mixing in 1D
Waveguides with a Breathing Crack using the Spectral Finite Element Method', Smart
Materials and Structures, Vol. 24, No. 11, pp 115004, 2015.
9. S Boccardi, D B Calla, G P M Fierro, F Ciampa, & M Meo, 'Nonlinear Damage
Detection and Localization using a Time Domain Approach', SPIE Proceedings of
Smart Structures and Materials+ Nondestructive Evaluation and Health Monitoring, pp.
98040T-98040T, April 2016.
10. J N Sundermeyer and R L Weaver, 'On Crack Identification and Characterization in a
Beam by Nonlinear Vibration Analysis', Journal of Sound and Vibration, Vol. 183, No.
5, pp 857-871, 1995.
The Interfacial Stiffness Evaluation of Single Lap Joint Assemblies using the
Transmission of Lamb Waves
E Siryabe, M Renier, A Meziane, M Castaings
Institut de Mécanique et d’ingénierie, I2M – Département d’Acoustique Physique,
Université Bordeaux Bat A4, 351 cours de la Libération,
33405 Talence Cedex, France
Contact: mathieu.renier@u-bordeaux.fr
The increasing production of adhesively bonded joints in transport industries such as aeronautic,
aerospace or automotive requires the development of non-destructive evaluation methods. Adhesive
bonding offers many advantages such as a reduction of the structure weight and a good repartition of
stress over the entire bonded area. However, specific defects can occur in the bonded joint and hugely
influence the mechanical strength of the assembly. These defects are located in two critical zones: the
bond-line (cohesive zone) and the adhesive-substrate interphase (adhesion zone or interphase zone).
Consequently, the defects might be separated into two classes, “cohesive defects” and “adhesive
defects”. Some cohesive defects can result from an imperfect curing of the adhesive layer. Adhesive
defects are located at the interphase between the adhesive layer and the substrate. Apart from
disbonds, those defects can correspond to a weak adhesion induced by inappropriate surface
treatments of the substrates. For safety reasons, it is necessary to non-destructively detect and evaluate
the quality of bonds. Among non-destructive techniques, ultrasounds are promising as they
mechanically solicitate bonding joints.
This work aims to characterize interfacial mechanical properties from measurements of Lamb wave
transmission through single lap joints assemblies. The assemblies are made of aluminium substrates
bonded with an adhesive epoxy. Variable interfacial properties are obtained with the help of different
surface treatments. A Lamb mode is generated from one of the substrates propagates and is then
transmitted to the other substrate. Using capacitive air coupled transducers providing a broadband
frequency bandwidth and a very good modal selectivity, the generated or converted modes are
measured before and after the joint, and the transmission coefficient is computed. Numerical
simulations performed using a FE model are also carried out. They are based on the adhesive
rheological model, where the mechanical behaviors of the interphases are reproduced using two
surfacic springs distributions (with a longitudinal stiffness kL and a transverse stiffness kT). From the
comparison with experiments, kT and kL are evaluated on the different assemblies. Results are
compared to those obtained from mechanical tests performed on samples prepared using the same
surface treatments.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
ISSUES IN PIEZOELECTRIC ENERGY HARVESTING
FROM CIVIL STRUCTURES
Sumit Balguvhar and Suresh Bhalla
Department of Civil Engineering, Indian Institute of Technology Delhi, Hauz Khas, New Delhi
110016, Email: sbhalla@civil.iitd.ac.in
Abstract
This paper deliberates upon common issues haunting piezoelectric energy harvesting from civil
structures. A piezo-electric harvester (PEH) can work as a sustainable and green power source
for low power consuming devices, such as wireless sensors used in transportation infrastructure,
building automation systems, implanted medical devices, structural health monitoring and
manufacturing process monitoring, thereby replacing batteries, which not only suffer from a
finite lifespan but also pose environmental risks arising from the disposal. However, the
application of PEH on civil infrastructures has not been fully mature. The power output of
typical energy harvesting devices could be limited to few hundreds of micro-watts only. This
paper presents a basic field study conducted on a city flyover followed by laboratory based
parametric study to highlight the typical practical issues associated with PEH from civilinfrastructures.
Keywords
Energy harvesting, piezoelectric, bridge, rectifier circuit, lead zirconate titanate (PZT), structure
health monitoring (SHM)
1. INTRODUCTION
Wireless sensor nodes (WSN) have received considerable attention for structural health
monitoring (SHM) during the recent years. The sensor nodes provide surveillance, evaluation
and assessment for existing or newly built civil infrastructures. These sensor nodes warrant their
own power supply, typically a battery pack. However, frequent replacement of battery packs is
infeasible, especially at inaccessible locations and where sensor units embed into the structures.
To solve this issue, the idea of energy harvesting has gained importance during the last one and a
half decades [1, 2] aiming to ensure self-sustaining power supply throughout the life of the
sensor node.
Energy can be harvested from civil structural vibrations by employing one of the various
transduction mechanisms such as piezoelectric, electrostatic and electromagnetic. Among them,
piezoelectric energy harvesting has commanded maximum attention owing to simplest structure,
no moving parts, negligible maintenance, large power density and no requirement of additional
8th International Symposium on NDT in Aerospace, November 3-5, 2016
voltage source [3]. For a piezoelectric plate of thickness h, the voltage generated across the
terminals of the PZT patch can be expressed in terms of the strain S1 as
 d Y Eh 
S
V   31
 T  1
33


(1)
where d31 is the piezoelectric strain coefficient providing coupling between the mechanical strain
(along axis ‘1’) and electric field E3 (along axis 3), Y E  Y E 1  j  is the complex Young’s
T
T
modulus of elasticity of the PZT patch at constant electric field and  33
1  j  the
  33
complex electric permittivity (in direction ‘3’) of the PZT material at constant stress;
with j   1 , η and δ respectively denoting the mechanical loss and the dielectric loss factors of
the PZT material. Piezoelectric materials are generally used in two modes i.e. d31 and d33. In the
d31 mode, the stress is applied along the length direction to produce voltage along the thickness
direction. In the d33 mode, on the contrary, both the stress and the voltage act in the same
direction, that is the thickness direction.
While piezo-electric harvesters (PEH) provide a promising way to generate power, there are
several impediments which hinder their deployment on civil infrastructure. Kaur and Bhalla [4]
investigated the feasibility of combined SHM and EH for civil-structures. They proposed to
utilize the harvested energy for the SHM purposes, thereby rendering the system autonomous.
Their theoretical cum experimental research clearly provided the feasibility of combined SHM
and energy harvesting from the same PZT patch in the form of concrete vibration sensor (CVS)
for reinforced concrete (RC) structures. Typically, it is possible to generate power in micro-watt
range. Elvin et al. [5] studied practical issues related to PEH from civil-structures. A PEH in
cantilever configuration with resonant frequency in the range of 100–300 Hz was deployed
whereas the civil structures typically exhibit a fundamental frequency less than 5 Hz. Due to the
mismatch of frequency, very low power was extractable. To improve the efficiency and power
generation several researchers [6] have focused on the size and operating frequency reduction of
piezoelectric transducers through various piezoelectric configurations rather than through
circuitry and storage medium. Another impediment is the random nature of excitation signals
typically encountered in the civil-structures. Additionally, there is the mismatch between the
natural frequencies of harvesters and civil infrastructures.
This paper aims at exploring the possibility of PEH from bridge vibrations utilizing traditional
electronic rectifier circuit and dwells upon the problems to be addressed by multidisciplinary
research involving civil, mechanical, electronics and instrumentation engineers.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
2. Experimental Voltage Measurements from a City Flyover
In order to get first hand information of the frequency and amplitude of possible voltage
generation from typical city flyovers, an ICP accelerometer (PCB 352C34) was attached to a
flyover in front of IIT Delhi at the mid-point. The flyover, shown in elevation in Figure 1(a), is a
steel structure with a span of 28.67 m and a girder depth of 2 m. The vibration measurement was
done under a normal traffic condition in daylight for continuous two hours. The typical
acceleration data is shown in Figure 1 (b) in time domain and the same is transformed to
frequency domain using fast Fourier transform as shown in Figure 2 (c). The measurement
shows that the natural frequency of the structure was 4 Hz. Other details of experiment can be
found in [8]. If a PZT patch of 10×10×0.3 mm size were employed instead of accelerometers,
computations show [8] that it would generate a peak voltage of 0.736 V.
(a)
Acceleration
(m/s2)
4 Hz
Time
(b)
(s)()(s)
(c)
Figure 2: (a) General view of bridge
(b) Traffic induced vibration signal in time domain
(c) Same vibration signal in frequency domain
8th International Symposium on NDT in Aerospace, November 3-5, 2016
After real life bridge measurement, a further experiment was performed in the laboratory
environment in order to estimate the power that can be harvested under different set of
conditions. For this purpose, a cantilever based PEH was designed using an aluminium beam
with dimensions 200×40×0.3 mm, as shown in Figure 2. A PZT patch of grade PIC 151
manufactured by PI Ceramic was bonded to the cantilever with epoxy adhesive. A horizontal
type shake table (operable in 1 Hz to 25 Hz frequency) was used to apply controlled vibration
excitations to the harvester. The bridge rectifier circuit using general purpose silicon diodesIN4007 (Figure 2) was connected to the PZT patch and 1µF electrolytic capacitor was allowed to
charge to its full potential. When the PEH was excited at 4 Hz, the open circuit voltage 4 V peak
was generated, as shown in Figure 3. Figure 4 shows the steady state voltage across the capacitor
(Vss) versus charging time (Tc) curve. It can be observed from the figure that when the cantilever
was excited at 4 Hz, 1.2 V was achieved in 3.584 s. The maximum harvestable power (P) can be
determined using following set of equations [8]
Ec
P 
Tc
(3)
Oscilloscope
PZT patch
Capacitor
Diodes
Shake Table
Figure 2: Experimental setup for charging capacitor using rectifier circuit
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Eh 
1
C V 2ss  3600
2Tc
(2)
where, C is the capacitance value ( 1µF), Vss is the steady state voltage across the capacitor and
Tc is the charging time when the voltage reaches to its maximum value. Using above equations,
the harvestable power was estimated as 0.2 µW. A detailed parametric study was performed
experimentally to investigate the effect of frequency, voltage and capacitor types [8]. However,
it was very problematic to store energy when the voltage output across the PZT patch was
reduced below 4 V. Hence, in real-life scenarios, where the peak voltage of less than 1 V is
expected, storage will be the major issue to be looked upon by researchers.
Steady state Voltage across
capacitor (Vss)
Figure 3: Open circuit voltage (Voc) from PZT
patch
Time (s)
Figure 4: Charging of capacitor at 4 Hz at 4V
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Most researchers believe that the key challenge for the successful deployment of energy
harvesting technology is the extraction of higher power. However, at the same time, literature
also reveals that the power requirements of numerous modern electronic devices are in micro to
milli watts. For such devices, the main aspect is storage, which remains problematic in civilstructures owing to their low frequencies coupled by low voltage outputs. Another critical part of
the PEH that needs attention is the energy processing and power delivery circuits, which involve
AC to DC conversion that interface between the PEH device and the electrical load. Particularly
for civil infrastructures, a significant problem coming in the way of the application of PEH in the
field of is the low frequency of vibrations, usually 5 Hz or less.
3. Conclusions
This paper reports the measurement of bridge vibrations and practical problems encountered
while storing the electrical energy at low frequencies and small amplitude of vibrations in civil
structures. The power generated was from laboratory experimental simulation was estimated as
0.20 µW at 4 Hz and 4 V. One of the existing issues for low extraction of power is attributed to
erratic and broadband nature of the structural vibrations. The limitation is also in the electrical
interfacing circuit due to which large voltage has been dropped. The mismatch of impedance in
PEH and the interface circuit leads to power dissipation. To improve the performance and
obtaining the more harvestable power, there is a need for further research from both aspects such
as efficient circuit for extraction besides further better transducers configurations to generate
higher voltages. Currently, further experiments are being carried out in [9] to overcome these
issues and the results will be published elsewhere.
References
1. S. Kim, S. Pakzad, D. Culler, J. Demmel, G. Fenves, S. Glaser, and M. Turon, “Health
Monitoring of Civil Infrastructures Using Wireless Sensor Networks” Proceedings ofthe
6th International Conference on Information Processing in Sensor Networks, April25–27,
Cambridge, Massachusetts, USA, pp.254–263, 2007.
2. Y. Wang, K. J. Loh, J. P. Lynch, M. Fraser, K. Law, and A. Elgamal, “Vibration
monitoring of the Voigt Bridge using wired and wireless monitoring systems”,
Proceedings of the 4th China-Japan-US Symposium on Structural Control and
Monitoring, Oct. 16–17, 2006
3. S. Cheng, Y. Jin, Y. Rao, and D. Arnold, “An active voltage doubling AC/DC converter
for low voltage energy harvesting applications”, IEEE Transactions on Power
Electronics, Vol 26, No. 99, pp. 2258–2265, 2011.
4. N. Kaur and S. Bhalla, “Combined energy harvesting and structural health monitoring
potential of embedded piezo concrete vibration sensors”, Journal of Energy Engineering,
8th International Symposium on NDT in Aerospace, November 3-5, 2016
American Society of Civil Engineers (ASCE),
10.1061/(ASCE)EY.1943-7897.0000224, 2015.
published
online,
DOI:
5. N.G. Elvin, N. Lajnef and A. Elvin, “Feasibility of Structural Monitoring with Vibration
Powered Sensors”, Smart Material and Structures, Vol.15, pp. 977-986, 2006.
6. M. Zhu, and E. Worthington, “Design and testing of piezoelectric energy harvesting
devices for generation of higher electric power for wireless sensor networks,”
Proceedings of the IEEE Sensors Conference (SENSORS '09), pp. 699–702, 2009.
7. M. Peigney, and D. Siegert, “Piezoelectric energy harvesting from traffic induced bridge
vibrations”, Smart Materials and Structures, Vol. 22, No. 9, pp. 1-11, Article No.095019, 2013.
8. S. Balguvhar, “Piezoelectric energy harvesting from civil structures”, Comprehensive
Report, Indian Institute of Technology, Delhi, 2016.
9. S. Balguvhar and S. Bhalla,“Parametric studies on the piezoelectric energy harvesting”,
Journal of Civil Structural Health Monitoring (under preparation).
Modeling of Elastic Wave Scattering in Polycrystalline Materials
Abhishek Pandala1, S.Shivaprasad1, C.V.Krishnamurthy2, Krishnan Balasubramaniam1
1
Centre for Non Destructive Evaluation, Department of Mechanical Engineering,
Indian Institute of Technology, Madras, Chennai, India- 600036
2
Department of Physics, Indian Institute of Technology, Madras, Chennai, India- 600036
Contact: abhishek.7512@gmail.com
Predicting ultrasonic wave propagation in large 3D structures is demanding in terms of computer
memory and CPU time. Accounting for polycrystalline nature of materials makes the task
significantly challenging. While elastic wave propagation and scattering characteristics of
polycrystalline metals have been studied, the role of polycrystalline microstructure in ultrasonic nondestructive evaluation is not fully comprehended. We have been studying FEM[1] and ray trace based
models[2] to understand how polycrystalline microstructure influences ultrasonic waves. We now
report on the development of an FDTD model that is capable of reducing computation time
significantly. Elastic wave propagation in 3D polycrystalline media is modelled using the velocitystress formulation employing the FDTD method. A controlled Voronoi tessellation represents the
polycrystalline microstructure. The variation of elastic constants due to varying grain to grain
crystallographic orientations is also taken into account. The numerical scheme employs parallel
computation using General Purpose Computing on Graphical Processing Units (GPGPU)[3] which
has shown speed up of 70 times compared to CPU. The FDTD formulation with parallel processing
allows one to deal with larger structures with additional complexities in terms of microstructural
features or as external appendages to a large structure.
Potential for Synchrotron radiation based NDE
in Aerospace Industry
C.V.Krishnamurthy
Department of Physics
Indian Institute of Technology, Madras
Chennai 600036, Tamilnadu, India
Email:cvkm@iitm.ac.in
Abstract: Aerospace industry uses several techniques to ensure that the various components that go into the
aircraft, missiles, spacecraft as well as the whole vehicle are tested for safety and reliability. Among the various
non-destructive techniques routinely employed by aerospace industry, x-ray radiography has been playing a
significant role. Advances in x-ray optics and detector technology has led to increasing use of microfocal, 3D
CT imaging,, and color X-ray imaging options for improved assessment of components critical for industrial
applications including aerospace applications. Synchrotron sources are, on the one hand, pushing the
possibilities of assessment to greater levels due to higher photon flux, smaller beam divergences etc and on the
other, opening up newer nondestructive inspection options on materials that go into the making of various
components such as FTIR, X-ray absorption and photoemission etc. We provide an overview of the possibilities
offered by synchrotron sources in imaging on sub-micron scale with higher resolution and characterization of
light weight metal alloy and composite materials used for aerospace applications.
Keywords: synchrotron radiation, color x-ray, -CT, aerospace NDE
1. Introduction
Aerospace industry deals with light weight materials that include aluminum and titanium
alloys, metal based honeycomb structures as well as all-composite structures. The extended
structures/panels are multilayered for corrosion protection among other functional
requirements. Adhesively bonded components are increasingly used. Operational conditions
in terms of temperatures, humidity, air pressure, vary significantly. Further, these structures
need to withstand lightning strikes and impact from debris/birds. Integrity of aerospace
structures is critical and needs to be evaluated /monitored throughout their lifecycle. Several
nondestructive evaluation techniques, contact and non-contact, are available to inspect locally
as well as to carry out structural health monitoring on aerospace structures [1]. Radiography
based inspection is an important nondestructive evaluation technique that continues to serve
the aerospace industry due to many developments (e.g. digital radiography, -CT, 3D
imaging).
We know that X-Rays are electromagnetic radiation produced when electrons decelerate
(giving rise to a continuous spectrum, termed Bremmstrahlung) after bombarding a target
(e.g., Fe, Cu, and Mo). We also know that X-Rays produced this way include the
characteristic emission lines from the target atoms, of high intensity, embedded in the
continuous spectrum. The spectral intensity of the electromagnetic radiation is non-uniform
and the angular distribution of the radiation is large so much so that only a fraction of the
total radiation is made use of through the deployment of spatial filters and energy filters. The
polarization state of the radiation is random.
(a)
(b)
(c)
Figure 1: (a) Comparison of spectral brightness arising from various mechanisms by which electrons are
accelerated. The figure includes lab X-ray generators along with the characteristic lines from different targets.
The data for conventional x-ray tubes are estimates only, since brightness depends strongly on such parameters
as operating voltage. The indicated two-order-of-magnitude ranges show the approximate variation that can be
expected among stationary-anode tubes (lower end of range), rotating-anode tubes (middle), and rotating-anode
tubes with microfocusing (upper end of range). Synchrotron radiation can be seen to be at least five orders of
magnitude brighter than conventional X-ray sources. [from http://xdb.lbl.gov/ Section2/Sec_2-1.html] (b) The
beam is concentrated into a forward cone with half angle of typically 0.1 to 1 mrad depending on the energy of
the electron. [from http://photon-science.desy.de/ research/ studentsteaching/ primers/ synchrotron_ radiation/
index_eng.html] (c) Synchrotron radiation is polarized linearly in the plane of the orbit. Above and below the
orbital plane, the polarization is circular which has important applications for x-ray scattering from magnetic
materials. [from Grübel G. et al, Vorlesung zum Haupt- oder Masterstudiengang Physik, SoSe 2016]
Synchrotron radiation is electromagnetic radiation produced from accelerating electrons in a
controlled manner by confining electron trajectories in a specific geometry [2]. Synchrotron
radiation is characterized by uniformity of intensity over a large spectral band (“white”
source), very high intensities (or flux) over the spectral band, and small angular spread. The
emission is polarized. There are specific spectral windows (beam lines), available for
particular applications. Advances in research on materials have led to the development of
energy-sensitive detectors, and x-ray optical elements such as Fresnel zone plates, and
photonic crystal reflectors for manipulating the beams (to reflect, focus and to change
polarization).
In what follows, a few examples drawn from recent published work are described to indicate
the possibilities of using synchrotron radiation as an NDE tool in aerospace industry.
2. Strain Mapping
The mechanical behaviour of thermal barrier coatings under actual operating conditions
needs to be understood to deal with durability issues of jet engine turbine blades. Highenergy synchrotron X-rays have been used to measure in situ internal strains as a thermal
gradient and mechanical load are applied to the coating on tubular IN100 specimens [3]. The
inner and outer radii of 2 and 4mm, allowed for the heating of the outer coated surface and
cooling of the inner substrate surface to create a thermal gradient across the coating layers.
Figure 2 shows the results.
Figure 2: Strain effects due to variations in mechanical load. (a) Loading conditions displaying mechanical load
cycles conducted at each homogeneous temperature, (b) measurement locations near the interface for both the
bond coat and YSZ, (c) NiAl (111) e22 strain versus applied mechanical load at various temperatures, (d) YSZ
(111) e22 strain versus applied mechanical load at various temperatures. (from [3])
Figure 3 shows that hard X-rays alone can be used for effective triaxial strain measurements
and to map the full 3D stress tensor in a Metal Inert Gas (MIG) welded highly textured 7150W51(T6) aluminum plates [4]. Two 12.6mm thick 500mm long AA7150 plates were
Variable Polarity Plasma Arc (VPPA) welded with the weld orientation parallel to the RD of
the parent plates. A 280mm square test-piece was cut off and then reduced in thickness to
7mm by machining from both sides in increments of 0.5mm to simulate a likely aerospace
manufacturing process.
Figure 3: (left) Map of the longitudinal strain (𝜀11 ) for the VPPA welded 7150-W51 (T6) aluminium alloy plate.
Midline strain (a) and stress (b) distributions. The solid line in (a) represents the data used to plot the map shown
on the left, whilst the solid line with error bars in (b) is the ‘near-zero’ variation of the corrected 𝜎33 (for the
positive side of the weld centre-line). (from [4])
3. -CT and Color X-Ray imaging
The monochromatic nature of the radiation and the small angular spread allows computer
tomography to be carried out leading to enhanced imaging capabilities. Coupled with spectral
tunability, synchrotron radiation also allows for discriminating materials with different
atomic numbers when imaging multicomponent materials at high spatial resolution and
spectral resolution.
Figure 4 shows an example where micro-X-ray CT (XCT) performance is compared with
synchrotron radiation based CT (sXCT). The data for sXCT in the top panel of Figure 4 show
not only an absorption contrast but also a phase contrast resulting in a much better contrast
for the different metallic interfaces [5]. Pores and Fe-aluminides are detectable by all three
methods but Mg2 Si-phases are also recognizable within the sXCT datasets as shown in the
bottom panel of Figure 4.
Figure 4: (top panel) Comparison between XCT (left) and sXCT (right) of an AlSi12Ni1 sample; crosssectional CT-images are shown. The voxel sizes are 3.5 and 0.3 m, respectively. (bottom panel) Comparison
between XCT (left) and sXCT (right) of an AlMg5Si7 sample. The voxel sizes were 3.5 and 0.3 m,
respectively. (from [5])
The availability of energy-selective detectors has added a new dimension to X-Ray imaging –
color, as the following example taken from [6] illustrates. Although this example does not
involve a synchrotron source, it is clear that its radiation characteristics would only enhance
the image quality.
Figure 5 - (left) Hyperspectral XCT: A regular area detector in a commercially available micro-tomography
scanner is replaced with a spectroscopic imaging detector with a high energy resolution. (right) 3D distribution
of mineral phases in a mineralised ore sample from a gold-rich hydrothermal vein. (a) Grayscale tomographic
slice through the sample created by integrating over the full spectral range. (b) Voxel spectra showing Au and
Pb K-edges. (c) Voxel spectra from quartz, pyrite and chalcopyrite minerals. (d) Vertical slice through the
sample with Au (blue) and Pb (red) containing voxels segmented and coloured. (e) 3D visualisations of the
distribution of mineral phases in the sample. (from [6]).
Aerospace industry has fuelled the search for newer/engineered materials that may meet its
demanding requirements. Material characterization of such materials has been a challenge for
conventional radiographic methods. The various spectral windows (beam lines) at the
Synchrotron radiation facilities have shown tremendous capabilities to characterize materials
spanning from soft-X-rays to the infrared! The recent study of aluminum corrosion layers of
aerospace vehicles using XANES (XANES, for X-ray Absorption Near Edge Structure)
experiments at the Al K-edge, with high spatial resolution and high sensitivity, performed on
the LUCIA beamline at SOLEIL synchrotron is an example [7].
Figure 6 shows an example of a material synthesized using Direct Laser Sintering (DSL), a
technology enabling the production of dense metal components directly from 3D CAD data
[8]. It was used for the first time to produce a Metal Matrix Composite (MMCp) based on AlSi-Cu alloy in view of its application in different fields, in particular for aeronautics. The
porosity of the material was investigated using synchrotron based X-ray computed microtomography technique. -CT experiments were performed at ELETTRA (Trieste-Italy) on
beamline SYRMEP with a monochromatic beam energy of 28 keV and a sample-to-detector
distance of 5 cm. A two-dimensional (2D) detector recorded 900 projections of the sample at
different angular positions spanning an angular range of 180°. The exposure time was 18
seconds per projection. Images were recorded on a 2048 × 2024 CCD detector with the pixel
size set to 4.5 μm. The 3D structure was reconstructed from 900 projections using an
algorithm implemented at ELETTRA. Each voxel of the reconstructed image was cubic with
a 9 μm size.
Figure 6: (left) Direct Metal Laser Sintered tensile samples of Al6061 + 7%Si + 5.6%Cu2O. (Middle) 3D
reconstruction of sub-volume of the sample showing the microstructure. (right) Example of central slices (a, b
and c) in the orthogonal planes (xy, yz and xz) and 3D volume (d) of the local pore size. The different
thicknesses of pores are plotted with different colors. (from [8])
4. Phase Contrast Imaging
The coherent nature of synchrotron radiation enables phase-contrast imaging capabilities
enlarging the scope of applications. Synchrotron radiation computed tomography (SRCT),
and synchrotron radiation computed laminography (SRCL) have been successfully used to
study composite materials at voxel resolutions in the order of 1 micron and below [9]. In
comparison, laboratory µCT offers moderate resolutions, typically several micrometers and
above. These techniques have allowed key features such as micro-cracking, voids and fibre
breaks within the material’s structure to be assessed in considerable detail as illustrated in
Figure 7 below.
Figure 7 - Cross-sectional views of thin (1 mm) impacted coupons of CFRP laminate via: (a) µCT, (b) SRCT
and (c) SRCL. Images (a) and (b) are of the same sample at the same location, whilst (c) is of a similar damage
region of a different sample (from [9])
Figure 8 is an example that demonstrates phase-contrast imaging capabilities using
synchrotron radiation from the 6.5 GeV TRISTAN Accumulator Ring located in Tsukuba,
Japan [10]. X-ray beams of high intensity and high spatial and chromatic coherence have
been used to image weakly absorbing objects in the energy regime above 30 keV, where
normal radiographs of such objects, based on absorption contrast alone, are essentially
featureless. The brilliance was  5  1013 photons/s/mm2 /mrad2/0.1% bandwidth) and the
angular divergence in the plane of diffraction was 5 arcsec.
Figure 8: X-ray phase-contrast image of an aerospace material comprising two thin sheets of aluminium
separated by an epoxy film adhesive recorded at Accumulation Ring (AR), Japan. Voids in the adhesive with
sizes below 100 m are clearly visible, with the typical void size being of order 200 m. (from [10])
5. Compton backscatter
Compton imaging tomography has been emerging as a very promising alternative to regular
CT particularly for aerospace applications as is illustrated in Figure 9 below using
conventional x-ray sources [11]. Based on Compton scattering, the technique is capable of
providing one-sided inspection, with the data acquired layer by layer and stitched to
reconstruct in 3D the part under inspection.
Figure 9: Compton backscatter data reconstructed in 3D to provide corrosion and defect detection assessment in
hard-to-access locations of multilayer aerospace components (from [11])
With synchrotron radiation sources, Compton imaging tomography can become a powerful
tool particularly as it allows single-side inspection.
6. Compact sources
Synchrotron radiation facilities are usually very large and the beam lines spread out from the
central “storage ring” that keeps the electrons accelerating while maintaining their
trajectories. They are not portable. As the synchrotron radiation source possesses highly
desirable features, efforts are underway to develop compact sources [12]. Newer radiation
schemes are envisaged [13]. Several compact sources have become operational [14], [15].
Soon we would witness the advent of compact sources that provide coherent, highly
collimated and spectrally tunable electromagnetic beams much like the tunable lasers! While
advances in x-ray optics and x-ray detection have improved radiographic inspection
significantly, these advances may redefine radiographic NDE with the availability of compact
synchrotron sources in the near future.
References
[1] Norbert G.H. Meyendorf, Peter B. Nagy, Stanislav I. Rokhlin (Eds.), Nondestructive
Materials Characterization With Applications to Aerospace Materials, Springer-Verlag
Berlin Heidelberg 2004
[2] Settimio Mobilio, Federico Boscherini, Carlo Meneghini (Eds), Synchrotron Radiation Basics, Methods and Applications, Springer-Verlag Berlin Heidelberg 2015
[3] Kevin Knipe, Albert Manero II, Sanna F. Siddiqui, Carla Meid, Janine Wischek , John
Okasinski,Jonathan Almer, Anette M. Karlsson, Marion Bartsch and Seetha Raghavan, Strain
response of thermal barrier coatings captured under extreme engine environments through
synchrotron X-ray diffraction, Nat. Commun. 5:4559 (2014)
[4] V. Stelmukh, L. Edwards and S. Ganguly, Full Stress Tensor Determination in a Textured
Aerospace Aluminium Alloy Plate Using Synchrotron X-Ray Diffraction, Textures and
Microstructures 35, No. 3/4, (2003), 175–183
[5] Johann Kastner, Bernhard Harrer, Guillermo Requena, Oliver Brunke, A comparative
study of high resolution cone beam X-ray tomography and synchrotron tomography applied
to Fe- and Al-alloys, NDT&E International 43 (2010) 599–605
[6] C. K. Egan, S. D. M. Jacques, M. D. Wilson, M. C. Veale, P. Seller, A. M. Beale,
R. A. D. Pattrick, P. J. Withers and R. J. Cernik, 3D chemical imaging in the
laboratory by hyperspectral X-ray computed tomography, Sci. Rep. (2015) 5:15979, 1-8.
[7] F. Mirambet, D. Vantelon, S. Reguer and E. Rocca, Synchrotron radiation contribution to
the study of aluminium corrosion layers of air and space museum aircrafts for their
preservation, J. Anal. At. Spectrom., (2016), 31, 1631–1637
[8] Emmanuelle Girardin, Chiara Renghini, Jack Dyson, Vittorio Calbucci, Francesca
Moroncini, Gianni Albertini, Characterization of Porosity in a Laser Sintered MMCp Using
X-Ray Synchrotron Phase Contrast Microtomography, Materials Sciences and Applications,
(2011), 2, 1322-1330
[9] D.J. Bull, L. Helfen, I. Sinclair, S.M. Spearing, T. Baumbach, A Comparison of MultiScale 3D X-ray Tomographic Inspection Techniques for Assessing Carbon Fibre Composite
Impact Damage, http://eprints.soton.ac.uk/355778/1/
[10] A.W. Stevenson, T.E. Gureyev, D. Paganin, S.W. Wilkins, T. Weitkamp, A. Snigirev, C.
Rau, I. Snigireva, H.S. Yound, I.P. Dolbnya, W. Yun, B. Lai, R.F. Garrett, D.J. Cookson, K.
Hyodo, M. Ando, Phase-contrast X-ray imaging with synchrotron radiation for materials
science applications, Nuclear Instruments and Methods in Physics Research B 199 (2003)
427–435
[11] Victor Grubsky, Volodymyr Romanov, Keith Shoemaker, and Rodion Tikhoplav, Recent
Progress on 3D Backscatter X-Ray NDE, Physical Optics Corporation Torrance, CA (2014)
[12] Hironari Yamada, Noriko Mochizuki-Oda, Makoto Sasaki (Eds): Portable Synchrotron
Light Sources and Advanced Applications, Intl. Symposium Shiga Japan, 13-14 Jan 2004,
AIP Conference Proceedings 716 Melville, New York, 2004
[13] Elena Eggl, Simone Schleede, Martin Bech, Klaus Achterhold, Roderick Loewen,
Ronald D. Ruth, and Franz Pfeiffer, X-ray phase-contrast tomography with a compact laserdriven synchrotron source, PNAS, 2015, 112, no. 18,5567–5572
[14] http://lynceantech.com/technology/x-ray-synchrotron/the-compact-light-source.html
[15] http://www.photon-production.co.jp/en/index_e.htm
Effect of surface asperity on thin film adhesion using laser induced stress waves
Sarthak S. Singh1, R. Kitey2
1,2
Department of Aerospace Engineering
Indian Institute of Technology Kanpur (India) 208016
ABSTRACT
Laser induced stress waves are used to characterize the interface tensile strength between
epoxy film and aluminum substrate for a range of interface roughness. Failure at
aluminum/epoxy interface is initiated by employing laser spallation technique where the
samples are loaded at a strain rate of 107/s. Michelson's interferometry is used to measure the
out-of-plane displacement of the top free surface of the film. Hybrid experimental/numerical
method is developed to evaluate thin film interface strength. The adhesion strength is
observed to decrease with increasing interface asperity. The effect of interface characteristics
on failure mechanisms is explained by performing computational analysis.
1. Introduction
The reliability and the functionality of sensors and miniaturized devices highly depend upon the
characteristics of thin adhesive layer which is used to bond them with the structural surfaces of
interest. Adhesively bonded aluminum joints are extensively used in aerospace applications
because of their high strength to weight ratio. Also there are several aircraft components which
are smaller in sizes and/or have intricate geometries that cannot be fastened to another structure
by conventional riveting, bolting or welding techniques. Prior to bonding, several pretreatments
such as chemical, electrochemical or mechanical processes [1], are used to get the desired
surface morphology which in turn enhances the failure characteristics.
The adhesion strength and the fracture toughness of a bi-material interface depend upon the
strain rate. Albers et al. [2] investigated the effects of different surface treatments on
aluminium/epoxy interface strength under quasi-static tensile and shear loading. They reported
higher shear strength when compared to the tensile counterpart. This was attributed to the
dominant effect of mechanical interlocking in shear loading case. By applying face milling,
Uehara et al. [3] developed several surface roughness values on metallic surfaces like steel and
brass of the order of few microns and measured tensile, shear and peel strength of with different
adhesives like epoxy resin and cyanoacrylate. They concluded that, there exist an optimum
roughness at which maximum tensile and shear strengths were attained irrespective of the time of
curing. On the contrary no such correlation was seen in case of peel strength. Interface
characterisation of aluminium/epoxy and wood/epoxy were carried out by Budhe et al. [4]. They
pointed out that there was an optimum surface roughness developed by emery paper, at which
the maximum interface shear strength was achieved. From the post failure analyses they revealed
that both adhesive and cohesive failure had taken place. Among a very few studies conducted
under dynamic loading conditions, Syn and Chen [5] reported that the energy dissipation at
1
2
Graduate student , sarthak@iitk.ac.in
Associate Professor and corresponding author, Tel: +91 512 259 7060, Email: kitey@iitk.ac.in
aluminum/epoxy interface increased with increasing interface roughness. They had performed
four point bend tests by loading the test samples using split Hopkinson pressure bar (SHPB).
The laser spallation technique used in this investigation is a non-contact based technique where
the interface is loaded dynamically by employing laser generated stress waves. These pulses
have sharp temporal rise and fall times and subject the substrate-film interface to a strain rate of
the order of 107/sec. At such high strain rate, the inelastic deformation is suppressed. This
technique was first introduced by Vossen [6]. Gupta and coworkers [7-10] further modified this
method by the introduction of laser Doppler interferometer to measure the free surface
displacement which was used to measure the substrate stress. Using this method they measured
the dynamic tensile strength of metal/ceramic [8,11], ceramic/ceramic [11], ceramic/polymer
[11], ice/metal [12] and fiber/matrix [13] interfaces. The shock development due to non-linear
elastic properties of fused-silica was exploited by Wang et.al [14] to measure the interface
strength of sub microns film dimension. Kandula et al. [15] and Kitey et al. [16] used a hybrid
experimental/numerical methodology to retrieve the interface strength of a multi-layer system.
The literature review indicates that most of the experiments were carried out for few microns of
interface roughness under quasi-static loading conditions. The influence surface morphology on
interface strength under extreme dynamic loading condition has not been carried out. The
objective of this work is to measure the interface tensile strength of aluminum and epoxy for
various interface roughness values using laser induced stress wave. A hybrid
experimental/numerical scheme [15, 16] is adopted to extract the interface strength. The
mechanism driving the failure at the wavy interface is explained using finite element
computational analysis.
Epoxy film (h)
(h)
Aluminum substrate
(750 µm)
Water-glass layer
(10 µm)
Fig.1. Schematic representation of the sample geometry used in the laser spallation experiment
2. Experimental Details
2.1 Material Preparation:
Aluminum slabs of 6063-T6 grade supplied by HINDALCO were machined down to 0.75 mm
thickness, followed by cutting them to the square pieces of 25 mm x 25 mm. These samples were
polished using alumina abrasive slurry with grit sizes varying from 35 µm to 1 µm. Stylus based
profiler (Dektak-XT) was used to measure the surface roughness of the prepared substrates. The
RMS roughness values obtained by scanning the polished samples ranged from 35 nm to 175
nm. The sample received after milling is also used as it is to prepare the test samples which had
the roughness value about 530 nm. In current investigation total five surface roughness values,
35 nm, 57 nm, 107 nm, 175 nm and 530 nm, are considered. For calibration experiments, a thin
layer of aluminum (~100 nm) is deposited on the finely polished aluminum substrate to increase
its reflectivity.
The epoxy was prepared by mixing TETA (Tri-ethyl tetra amine) hardener with DGEBA
(Diglycidyl Ether of Bisphenol A) epoxy resin in a weight fraction of 1:10. The mixture was
stirred carefully in such a way that no air entrapment took place. After an un-interrupted stirring
for about 20 minutes, a homogenous solution was obtained. First the aluminum substrate was
firmly fixed to a flat surface. Few droplets of epoxy were then placed on the polished surface of
the aluminum substrate. A glass slab of 50 mm x 17 mm x 5 mm dimension, wrapped with a
cellophane tape was placed on these droplets. In order to get a film thickness of desired
dimension, a known weight which was calibrated a- priori, was placed on the top of the glass
slab. The entire arrangement was left for nearly 15 hours which ensured complete curing of
epoxy. After that, the glass slab was removed carefully which yielded an epoxy film of desired
thickness being deposited on to the aluminum substrate. The thickness of the epoxy film (h) used
in this study varied from 13 m to 30 m. Before performing the experiment, the back side of
the specimen was spun coat with a layer of water glass which acted as a confinement layer. Fig.1
shows the dimension of various layers on a test specimen.
The densities of epoxy and aluminum samples are measured by weighing cubical samples of
volume 1.0 cm3. A 10 MHz longitudinal wave transducer (M112-RM) and 5 MHz shear wave
transducer (V154-RM), acting as actuator and sensor along with OLYMPUS Parametrics 35DL
were used to find the longitudinal (CL) and shear wave (CS) speed in the aluminum and epoxy
samples. The dynamic Young’s Modulus (Ed) and dynamic Poisson’s ratio (νd) are calculated
using the following plane strain equations.
CL 
Ed (1  d )
 (1   d )(1  2 d )
CS 
,
Ed
2  (1   d )
(1)
Table-1 shows the measured density, longitudinal and shear wave speed, dynamic elastic
modulus and dynamic Poisson’s ratio in Al and epoxy samples.
Table 1
Measured material properties of aluminum and epoxy layers
Material
ρ
d
CL
CS
Ed
(kg/m )
(m/s)
(m/s)
(GPa)
Aluminum
2700
6420
3100
69.95
0.348
Epoxy
1195
2610
1220
4.848
0.36
3
2.2 Laser spallation setup
A 5 ns Gaussian pulse of variable energy content (0-300 mJ) from Q-switched Nd:YAG laser
(λ = 1064 nm) is focused onto the aluminum substrate using an infrared achromatic lens. The
localized melting of aluminum by irradiated laser energy leads to the formation of plasma that
further expands. The localized expansion of aluminum is constricted by a layer water glass
which develops a compressive stress pulse, propagating towards interface. Upon reaching the
free surface of the film, the mode converted stress wave loads the substrate/film interface in
tension. The intensity of laser irradiation is increased till the sign of first interface failure is
observed. At the critical interface stress the film spalls away from the substrate.
Fig.2. Schematic representation of the laser induced spallation experiment and Michelson’s
interferometry setup for measuring out-of-plane surface displacement of the film/substrate in
calibration experiment
2.3 Interferometric Measurements
In situ out-of-plane displacement measurement at the free surface of the calibration aluminum
sample is performed by employing Michelson’s interferometer as shown in Fig. 2. The optical
path difference between the coherent DPSS (Diode pumped solid state) laser beam (λ= 514 nm),
reflected from free surface of the sample, and the reference mirror develops interferometric
fringes. The variation of fringe intensity I(t), recorded using a photodiode and an ultra high
frequency oscilloscope is related to the fringe order N (t) by,
I (t ) 
I max  I min I max  I min

sin 2 N (t )
2
2
(2)
where, Imax and Imin are the maximum and the minimum intensities of the fringe in consideration,
respectively. The out-of-plane displacement history u(t), is calculated from the Doppler shift
[17],
u (t ) 

2
(3)
N (t )
where λ is the wavelength of the probe beam. Using a thin film assumption, the following
equation is used for evaluating the substrate stress history [14].
1
2
 sub (t )   ( Cd ) sub
u
t
(4)
where, ρ and Cd are the density and dilatational wave speed of the substrate respectively.
(a)
(b)
(c)
(d)
Fig.3. Interferometric measurements obtained from the calibration experiments; (a) variation of
fringe intensity, recorded by using photo-diode at 95 mJ/mm2 laser fluence incidence. (b)
Corresponding out-of-plane displacement of the free film surface. (c) Substrate stress history.(d)
Associated stress history at Al/epoxy interface evaluated using elasto-dynamic plane wave
propagation model
2.4 Calibration protocol
Typically, the free surface displacement of the film recorded by using interferometric technique
is used for evaluating the interface stress. Since the epoxy film used in the present study has low
reflectance; the hybrid experimental/numerical scheme [15, 16] is employed to evaluate interface
stress. First, the laser spallation tests were conducted and the laser fluence corresponding to the
onset of interfacial failures were noted down for various Al/epoxy samples. This is followed by
obtaining substrate stress histories through in-situ interferometric measurements performed on
the calibration sample. The transient stress wave associated to the onset of interfacial failure is
applied to an elasto-dynamic planar wave propagation model to retrieve the stress at the
interface. The model is simulated in ABAQUS/CAE-6.12 using plane strain condition with a
four node element CPE4R. A mesh size of 0.12 µm is chosen based on a convergence study prior
to extracting the stress history at the interface between the aluminum and epoxy. A
representative interferometric measurement from the calibration experiments at 95 mJ/mm2 of
laser fluence is shown in the Fig. 3.
(a)
(c)
200 µm
(b)
150 µm
Fig.4. (a and b) Micrographs showing the interface failure in Al/epoxy samples with 35 nm and
530 nm interface roughness, respectively; (c) Variation of interface strength with the interface
roughness.
3. Results and Discussion
3.1 Effect of interface roughness on adhesion strength
The laser spallation experiments were conducted on different interface roughness samples by
impinging the laser energy from the lowest to the highest. At each laser energy, the experiment is
repeated at least three times to ascertain the repeatability of the recoded data. The samples were
observed under the optical microscope to get the minimum laser fluence at which the interface
failure initiated. The process is repeated for the samples of each interface roughness value. The
representative spallation micrographs correspond to the lowest and the highest interface
roughness are shown in Fig. 4 (a) and (b). It was observed from the experiments, that the laser
fluence required to instigate failure at the interface increased with decreasing interface
roughness. Fig. 4 (c) shows the plot interface stress corresponding to various roughness values.
(a)
(b)
Fig.5. (a) Schematic representation of the Al/epoxy interface along with loading and boundary
conditions used in computational modeling. (b) Stress history at the crest of the sinusoidal
interface for λint = 10 µm and varying Aint.
3.2 Numerical Modeling
A wave propagation analysis is performed on aluminum/epoxy bi-material geometry with
sinusoidal interface. Two dimensional finite element simulations with dynamic explicit
integration scheme in plane strain condition were carried out as discussed before. The
mechanical properties of Al and epoxy used in the simulation are given in Table-1. Since the
diameter of the laser beam was focused down to 1 mm and the dimensions of the films used were
smaller than 1 mm, the stress waves can be approximated to be the planar waves. In order to
ensure planar wave propagation in the simulation, roller support was applied to the substrate and
the film at the top and bottom along the direction of wave propagation. A symmetric Gaussian
pulse of 7 ns duration with amplitude of 470 MPa was applied at the free surface of the
aluminum. The specimen geometry along with the boundary conditions is depicted in Fig. 5 (a).
The pulse duration of 7 ns was chosen because the temporal width of substrate stresses obtained
from the calibration experiments were nearly of 7 ns. The dimension of the epoxy film was
chosen to be at least twice the distance travelled by the Gaussian pulse in the epoxy layer. This
was deliberately chosen to avoid stress wave interaction in the epoxy film. The amplitude (Aint)
of the sinusoidal interface was modeled as 2.5 µm, 1.25 µm and 0.0 µm with the wavelength
(λint) was kept constant as 10 µm. From the simulations it was observed that the magnitude of
normal stress at the crest of the sinusoidal interface was found to be higher than that at the
trough. The normalized transient tensile stress at the crest of the interface with respect to the
incident substrate stress, obtained for these three cases are illustrated in the Fig. 5(b). It is evident
that the magnitude of the tensile interface stress increases with increasing interface amplitude.
4. Conclusion
The laser spallation technique was successfully applied to calculate the adhesion strength
between aluminum and epoxy under extreme dynamic loading conditions for various interface
roughness values. It was observed that the adhesion strength increased with decreasing interface
roughness. To understand the under lying mechanism of failure, a two dimensional finite element
plane wave propagation model with sinusoidal interface was simulated in ABAQUS. It was
observed that the magnitude of interface tensile stress increased with increasing interface
amplitude. This is attributed to the lower failure strength in case of rougher interface.
5. References
1. Minford, J.D., 1993. Handbook of aluminum bonding technology and data. CRC Press.
2. Albers, R.G. and White, S.R., 1996. Experimental investigation of aluminum/epoxy
interfacial properties in shear and tension. The Journal of adhesion, 55(3-4), pp.303-316.
3. Uehara, K. and Sakurai, M., 2002. Bonding strength of adhesives and surface roughness
of joined parts. Journal of materials processing technology, 127(2), pp.178-181.
4. Budhe, S., Ghumatkar, A., Birajdar, N. and Banea, M.D., 2015. Effect of surface
roughness using different adherend materials on the adhesive bond strength. Applied
Adhesion Science, 3(1), p.1.
5. Syn, C.J. and Chen, W.W., 2008. Surface morphology effects on high-rate fracture of an
aluminum/epoxy interface. Journal of composite materials,42(16), pp.1639-1658.
6. Vossen, J.L., 1978. Measurements of film-substrate bond strength by laser spallation.
In Adhesion Measurement of Thin Films, Thick Films, and Bulk Coatings. ASTM
International.
7. Gupta, V., Argon, A.S., Parks, D.M. and Cornie, J.A., 1992. Measurement of interface
strength by a laser spallation technique. Journal of the Mechanics and Physics of
Solids, 40(1), pp.141-180.
8. Yuan, J. and Gupta, V., 1993. Measurement of interface strength by the modified laser
spallation technique. I. Experiment and simulation of the spallation process. Journal of
Applied Physics, 74(4), pp.2388-2396.
9. Gupta, V. and Yuan, J., 1993. Measurement of interface strength by the modified laser
spallation technique. II. Applications to metal/ceramic interfaces. Journal of Applied
Physics, 74(4), pp.2397-2404.
10. Yuan, J., Gupta, V. and Pronin, A., 1993. Measurement of interface strength by the
modified laser spallation technique. III. Experimental optimization of the stress
pulse. Journal of Applied Physics, 74(4), pp.2405-2410.
11. Gupta, V., Yuan, J. and Pronin, A., 1994. Recent developments in the laser spallation
technique to measure the interface strength and its relationship to interface toughness
with applications to metal/ceramic, ceramic/ceramic and ceramic/polymer
interfaces. Journal of adhesion science and technology, 8(6), pp.713-747.
12. Archer, P. and Gupta, V., 1998. Measurement and control of ice adhesion to aluminum
6061 alloy. Journal of the Mechanics and Physics of Solids,46(10), pp.1745-1771.
13. Yu, A. and Gupta, V., 1998. Measurement of in situ fiber/matrix interface strength in
graphite/epoxy composites. Composites science and technology,58(11), pp.1827-1837.
14. Wang, J., Weaver, R.L. and Sottos, N.R., 2002. A parametric study of laser induced thin
film spallation. Experimental Mechanics, 42(1), pp.74-83.
15. Kandula, S.S.V., Hartfield, C.D., Geubelle, P.H. and Sottos, N.R., 2008. Adhesion
strength measurement of polymer dielectric interfaces using laser spallation
technique. Thin Solid Films, 516(21), pp.7627-7635.
16. Kitey, R., Sottos, N.R. and Geubelle, P.H., 2010. A hybrid experimental/numerical
approach to characterize interfacial adhesion in multilayer low-κ thin film
specimens. Thin Solid Films, 519(1), pp.337-344.
17. Barker, L.M., 1972. Laser interferometry in shock-wave research. Experimental
Mechanics, 12(5), pp.209-215.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Effect of Shock Induced Acoustic Emission and Shock Waves Impact on
Polyurethane Foam
V. Jayaram1 and G. Arvind Raj2
1
Shock Induced Materials Chemistry Lab, SSCU, Indian Institute of Science, Bangalore 560012, India.
Laboratory for Hypersonic and Shock Wave Research, Department of Aerospace Engineering, IISc, Bangalore 560012, India.
Phone: +91-80-2293 3306, Fax: +91-80-2360 1310, E-mail: jayaram@sscu.iisc.ernet.in, aarvindaero@gmail.com.
2
Abstract: In this paper we present the effect of acoustic emission (AE) and shock waves impact on a rigid
polyurethane (PU) foam having density of 320 kg/m3. Diaphragm less shock tube is used to produce shock
waves of about 1.3 Mach number. When this shock wave enters the open atmosphere produces large AE due to
sonic boom at the open end of the shock tube and continues to travel in the atmosphere. PU foam materials were
placed at different distance from the open end of the shock tube to study the attenuation of AE. The PCB
piezoelectric acoustic pressure sensor is used to measure the AE with and without PU foam. The results shows
the attenuation of the AE measured at different position varied between 72 to 65 % in the air medium.
Shock wave impact on PU foam was studied using material shock tube. PU foam was mounted inside the
end flange of the driven section of the shock tube and exposed to different shock strength (Mach number 2 to
3.1). The PCB piezoelectric pressure transducers are used to record both reflected and back-wall transmitted
shock pressure at the end of the shock tube. Result shows that the pressure amplification at the back-wall of the
PU foam is about 2 to 2.6 times the reflected shock pressure at the front-wall of PU foam. The details of the
experimental setup and the results obtained will be presented in this paper.
Keywords: shock tube, acoustic emission, attenuation, shock pressure amplification, polyurethane foam
1. Introduction
Dynamic loading of shock and blast waves on materials have attracted a lot of scientific
attention in recent years. Presently, many devices like light gas gun, ballistic shock tube,
explosive launchers and conventional shock tubes are used to study hypervelocity impact,
dynamic compressibility, strength characteristics and spallation phenomena on the materials
in the laboratory. Such devices are capable of producing strong shock waves and acoustic
emission (AE). Interaction of short duration strong shocks with material leading to the
formation of a diffuse distribution of micro cracks or voids in the interior of the material
body. Reduction of the energy content in AE during shock and blast waves plays an
important role in the protection of civilian and military structures. To overcome these
problems light damping materials are used as barriers.
Many NDT methods have been used to characterize on going processes such as the
generation, propagation of shock waves and their attenuation into elastic waves. NDT method
is a valuable tool to determine the attenuation characteristic of target material by measuring
AE before, during and after hypervelocity impact [1]. The porous foam materials with
different properties and geometries were used for the shock attenuation studies [2-3]. During
shock impact, the internal energy imparted to the porous material is much greater than that
imparted to the solid of the same material. Usage of lightly compacted material such as sand
and aqueous porous foam results in the attenuation of the peak over pressure in blast waves
which has been reported in the literature [4]. Studies were also performed to attenuate AE
produced during detonation using porous foams and wire mesh as damping materials [5].
Attenuation of weak shock waves propagation in atmosphere was investigated using urethane
and fiberglass material [6]. Polyurethane foam materials are widely used as shock absorbent
material because of its high porosity. Experiment to find the influence of attenuation of AE in
transparent thin films [7], crossply and quasi-isotropic panels made up of carbon fibre
1
reinforced polymer are reported in the literature [8]. Investigation of seven different types of
foams exposed to shock waves which was observed that the transmitted pressure is amplified
at the back-wall due to the transfer of gas momentum to the foam mass [9]. Shock impact on
a surface covered with a layer of flexible foam [10] and porous fabric materials [11] have
shown interesting phenomenon of pressure amplification.
In this paper, diaphragm less shock tube (DST) is used to study the attenuation of AE in
rigid PU foam material at the exit of the shock tube. Attenuation of AE is performed in open
atmosphere at 1.3 Mach. Material shock tube (MST1) is used to study amplification of
transmitted pressure at the back-wall of the PU foam at different shock strength. Experiments
were conducted at four different shock Mach number by mounting PU foam at the end of the
shock tube.
2. Material and Experimental Details
Rigid PU foam (FR-4520) of density 320 kg/m3 was procured from General Plastics Mfg.
Company. Effect of shock induced acoustic emission and shock wave impact experiments
were performed on this PU foam. Diaphragm less shock tube is used to study shock induced
acoustic emission and material shock tube (MST1) is used to study the effect of shock wave
impact on PU foam.
2.1. Acoustic emission experiments using diaphragm less shock tube
DST consisting of a driver section (one meter) and a driven section (4 m length) with 80 mm
inner diameter is used to produce shock waves. The driver section filled with compressed air
of about 8 bar pressure and an electrically operated solenoid valve is opened to produce
single pulse shock wave of about 1.3 Mach number. When this shock wave enters the open
atmosphere produces large AE due to sonic boom and further travels in the atmosphere. In
the present experiments, PU foam material was used to study the attenuation of AE at
different distance by placing them at 30˚ angle from the open end of the shock tube as shown
in Fig. 1(a). The experiments were performed on FR-4520 rigid polyurethane foam of 38 mm
thick to investigate the attenuation of AE. The PCB piezoelectric acoustic pressure sensors
(model: 103B02) are mounted to measure the AE with and without the PU foam at four
different locations 1.15 m, 1.73 m, 2.31 m and 2.88 m as shown in Fig. 1(a). The typical AE
signals acquired with and without PU foam using the Tektronix digital storage oscilloscope
are shown in Fig. 1(b).
Experiments were performed by keeping two sensors at an angle of 30° from the exit of
the open end of the DST. Four locations were marked at 1.15 m, 1.73 m, 2.31 m and 2.88 m
from the open end of the shock tube and provisions are made to mount the acoustic pressure
sensor at both sides of the shock tube at 30° angle. The emitted acoustic signals travel in the
open atmosphere and are detected by an acoustic transducer mounted without PU foam at a
distance of 1.15 m and another transducer mounted at back side of the PU foam to measure
the attenuation of AE signal at the same distance from the source. Similarly experiments were
repeated by changing the position of acoustic transducers to 1.73 m, 2.31 m and 2.88 m.
During each experiment, acoustic emission data acquired from both the transducers were
stored in the Tektronix digital oscilloscope.
2
Figure 1(a) Schematic diagram of a diaphragm less shock tube showing the position of the
acoustic pressure transducer from the open end of the shock tube, 1(b) Typical AE signals
recorded with and without PU foam using Tektronix digital storage oscilloscope.
2.2. Shock wave impact experiments using material shock tube
MST1 consists of two sections: the driver section and the driven section. The length of the
driver and driven sections are 2 m and 5 m long with 80 mm inner and 115 mm outer
diameter. The driver section is separated from the driven section (shock tube) by an
aluminium diaphragm. The thickness of the Al diaphragm dictates its bursting pressure which
will influence the shock Mach number; the higher the bursting pressure the stronger the
shock. The shock tube employed here is used to test the PU foam material at different shock
strength. The schematic diagram of MST1 along with the typical pressure signal acquire are
shown in Fig. 2. The arrangements were made to mount the PU foam material at the end of
the shock tube. The PU foam experiences the reflected shock pressure (P5), estimated
temperature (T5) at the front-wall and the transmitted shock pressure (Pt) at the back-wall as
shown in the Fig. 2(a).
The experimental procedure starts by placing the aluminum diaphragm in between the
driver and the driven section and the PU foam mounted at the end flange of the shock tube.
The driven section (shock tube) is filled with air at 1.0 bar pressure and the driver section is
filled with high-pressure helium gas until the diaphragm bursts and generates the shock wave.
Different thickness Al diaphragm is used to produce different Mach number ranging from 2
to 3.1. Shock impact on foam material occurs at different velocities ranging from 600 m/s to
1100 m/s at different reflected shock pressure. Pressure sensor P5 and Pt at the end of the
shock tube is used to record the front-wall reflected shock pressure and back-wall transmitted
shock pressures respectively are shown in Fig. 2(b). The PU foam experiences both P5 and Pt
pressure during shock tube experiments and are measured using dynamic pressure sensors
(Model 113B22, PCB - Piezotronics Ltd., USA). A similar procedure is followed to perform
experiment at different Mach number. The shock speed (Vs), P5 and Pt data were acquired
and stored in a Tektronix digital storage oscilloscope (TDS2014B). Time taken (Δt) for the
shock to travel the distance between the two pressure sensors (ΔL = 0.5 m) is calculated from
the acquired data. These experimental data are used to calculate the shock velocity (VS =
ΔL/Δt) and the shock Mach number (MS = VS/a1), where a1 is the speed of sound in the test
3
gas present in the driven section of the shock tube. The reflected shock temperature (T5) for a
given shock Mach number is estimated using the 1D normal shock equations [12].
Figure 2(a) Schematic diagram of MST1 showing the position of transducers to measure the
front-wall (P5) and back-wall transmitted pressure (Pt), 2(b) Typical pressure signals recorded
at the side wall (reflected shock pressure) and transmitted pressure at the end flange of the
shock tube with and without PU foam.
T5 {2  1M s2  3   }{3  1 M s2  2  1}

T1
  12 M s2
[1]
The estimated temperature (T5) of the test gas (air) behind the reflected shock is a function of
shock Mach number, specific heat ratio of gas γ and the ambient temperature T1 of the test
gas. Table 1 shows the experimental pressure data (P5 and Pt) and estimated reflected shock
temperature (T5).
Table 1: Experimental and estimated values of shock conditions.
Mach
Number
2.0
2.6
2.9
3.1
Estimated
Temperature
T5 (K)
690
1175
1400
1580
Polyurethane Foam
Reflected Shock
Pressure
P5 (bar)
10.48
32
45.51
58.62
Transmitted First
Peak Pressure
Pt (bar)
50
89
109
130
3. Results and Discussions
3.1. Acoustic emission experiments
Measurements of acoustic emission (AE) are considered as unique techniques comparing
with that of other non-destructive techniques. Many of the NDT are static measurements but
the AE have major advantage to record the dynamic data. A large or considerable amount of
acoustic energy released during micro-cracking, friction, shock waves induced acoustic
emission, etc. can be recorded by acoustic sensors mounted on the surface of the material.
4
Spontaneous release of localized strain energy from the stressed material during shock impact
is recorded as AE within few meters to millimeters length. We recorded the first arrival of the
elastic waves from the exit of the DST using acoustic sensors located along 30° angle at a
distance of 1.15m, 1.73 m, 2.31m and 2.88 m as shown in Fig. 1. The recorded AE data with
and without PU foam are analyzed to understand the percentage of acoustic attenuation.
The arrival of the original acoustic emission signals detected without PU foam at a
distance of 1.15m and 2.88 m are shown in the Fig. 3(a). The acoustic pressure signal
recorded show the first peak value of 2.53 psi and 0.88 psi for the corresponding distances of
1.15 m and 2.88 m respectively and subsequent decays of the AE signal is also recorded until
1.5 ms time scale. In presence of PU foam, attenuated AE signal from the same location
shows a broad peak where the intensity of AE signal decreased to 0.73 psi at 1.15m to 0.26
psi at 2.88m as shown in Fig. 3(a). In all the experiments oscillation of AE signal is
characterized by compressive and tensile stress experienced by the material due to the impact
of acoustic waves. Figure 3(b) shows the non-linear characteristic of the first arrival of the
peak AE signal acquired at four different locations without PU foam. We can clearly observe
that drastic attenuation of AE detected at the back-wall of the 38 mm thick, 320 kg/m3
density PU foam (FR-4520) as shown in Fig. 3(b). The percentage of attenuation of AE
signal from minimum distance to maximum distance varies from 72% to 65% respectively.
Figure 3(a) Data acquired from the open end of the DST; AE detected by PCB 103B02
acoustic sensor at a distance of 1.15 m and 2.88 m with and without PU foam material, 3(b)
Intensity plot of the first arrival peak of AE signal with and without PU foam at different
location and also shows the intensity of attenuation.
3.2. Shock wave impact experiments
Shock tube experiments were performed with and without PU foam at the end flange of the
shock tube at different shock Mach numbers. The recorded pressure signals at the end flange
of the shock tube without PU foam are shown in Fig. 4(a). The corresponding peak pressure
increases as a function of Mach number and as the time progresses its amplitude decays as
shown in Fig. 4(a). The experiments were performed by rigidly fixing the PU foam to the end
flange with a pressure sensor. At this condition (i.e. the foam is placed against a rigid solid
surface) a significant pressure amplification is produced at the back-wall of the PU foam.
However, if gap exists between the PU foam and the sensor the PU foam attenuates the
reflected shock pressure. The present experiments were done only to study the shock pressure
5
amplification. Figure 4(b) shows the transmitted pressure-profiles recorded at the back-wall
of PU foam at different Mach numbers. First peak pressure at the back-wall of the PU foam
shows amplification of about 2-2.6 times the front-wall pressure. The profiles of the pressure
signals shows damped oscillations and decays within 2 ms as shown in the Fig. 4(b). The plot
of shock Mach number versus transmitted shock pressure with and without PU foam is shown
in Fig. 5(a). A non-linear variation of pressure signals are found as a function of shock Mach
number. Figure 5(b) shows the plot of P5/front-wall pressure of PU foam with respect to
transmitted peak pressure which represents the pressure amplification due to the shock-PU
foam interaction.
Figure 4(a) Shock tube end pressure profiles showing pressure signal oscillation without PU
foam at different Mach number. First peak pressure signal value is indicated using the arrow
mark, 4(b) Transmitted pressure-profiles at the back-wall of PU foam showing damped
pressure oscillation during shock-PU foam interaction. First peak pressure signal shows
amplification of pressure signal at different Mach number as indicated using the arrow mark.
6
Figure 5(a) Plot of shock Mach number versus transmitted peak pressure signals with and
without PU foam material, 5(b) Plot of side-wall pressure data (reflected shock pressure)
versus transmitted peak pressure signal.
4. Conclusions
DST is used to produce AE signal which are measured at different location like 1.15m,
1.73m, 2.31m and 2.88m from the open end of the shock tube. In presence of rigid PU foam
the intensity of the AE signal drastically reduced due to the attenuation property of PU foam.
To investigate the amplification property of the PU foam shock tube experiments were
performed using MST1. Shock waves of different Mach number 2, 2.6, 2.9 and 3.1 were
produced and made to interact with the PU foam. Study shows that the transmitted pressure at
the back-wall of the PU foam amplifies comparing with that of front-wall pressure. It is
concluded that the pressure amplification varies with Mach number, reflected shock pressure
and properties of the PU foam. In future we are planning to perform both AE and transmitted
pressure amplification experiments with foam materials of various densities.
5. Acknowledgements
Financial supports for this study from the DRDO and ISRO-IISc Space Technology Cell,
Government of India are gratefully acknowledged. Thanks to Professors K P J Reddy and G
Jagadeesh of Aerospace Engineering department, IISc, Bangalore for their fruitful discussion
and encouragements.
6. References
1. Dorothee, M., et al., Acoustic emission analysis of experimental impact processes in
comparison to ultrasound measurements and numerical modeling, J. Acoustic
Emission, 2013. 31, p. 50-66.
2. Kitagawa, K., et al., Attenuation of shock waves by porous materials, Proceedings of
the 24th International Symposium on Shock Waves, 2004. p. 1247-1252.
3. Jeon, H., et al., Shock wave attenuation using foam obstacles: Does geometry matter?,
Aerospace, 2015. ISSN 2226-4310.
7
4. Golub, V. V., et al., Blast wave attenuation by lightly destructible granular materials,
Shock Waves, 2005. p. 989-994.
5. Teodorczyk, A., et al., Detonation attenuation by foams and wire meshes lining the
walls, Shock Waves, 1995. 4, p. 225-236.
6. Cloutier, M., et al., Reflections of weak shock waves from acoustic materials, The J. of
Acous. Soc. of Amer., 1971. 50, p. 1393-1396.
7. Emery, P., et al., Acoustic attenuation measurements in transparent materials in the
hypersonic range by picosecond ultrasonics, Appl. Phy. Lett., 2006. 89, p. 191904-1191904-3.
8. Kassahun, A., et al., Influence of attenuation on acoustic emission signals in carbon
fiber reinforced polymer panels, Ultrasonics, 2015. 59, p. 86-93.
9. Michael, W. S., et al., Effect of compressible foam properties on pressure amplification
during shock wave impact, Shock Waves, 2006. 15, p. 177-197.
10. Gibson, P. W., Amplification of air shock waves by textile materials, J. Text. Instit.,
1995. 86[1], p. 119-128.
11. Thom. C. G., et. al., Shock wave amplification by fabric materials, Shock Waves,
2009. 19, p. 39-48.
12. Gaydon, A. G., et al., The shock tube in high temperature chemical physics, Reinhold
Publishing Corporation, New York, 1963.
8
Manufacturing Aspects of Fabrication of
Composite Reference Standard for NDT Ultrasonic Inspection
Pranab Biswal 1, B.N.Srinivasa Reddy 2, Pratim. M. Baruah 3
1
2
3
Senior Manager (NDT) , Chief Manager (NDT) , Deputy Manager (QA)
1
2
pranab.biswal@hal-india.com , srinivasareddybn@gmail.com , pratim.m.baruah@gmail.com
Aerospace Composites Division, Hindustan Aeronautics Limited, Bangalore-560037
3
Abstract
This paper provides a brief review of the manufacturing aspects considered during development of a composite reference
standard used for system standardization during ultrasonic testing of composite components. Most of the calibration / test
blocks available in the market are in accordance with national and international standards but are metallic in nature (i.e
Aluminum, Steel etc.). These standards cater only to the inspection requirements of metallic materials. Whereas in the
composite manufacturing industry/environment, it is not appropriate to use metallic test blocks as the properties are
completely different. For validating NDT techniques, characterizing defects and standardizing test equipment parameters
with respect to composite components, a dedicated composite reference / calibration standard is required. The composite
reference standard should be representative of the design configuration and the dimensional complexity of the part and
should focus on parameters like skin thickness, material type, surface finish, bonding adhesive, internal structure and
fabrication process. The reference standard should also contain induced artificial discontinuities which would produce a
similar response to that produced by typical defects like voids, de-lamination, de-bonds and inclusions (FOD) in
composites. Documenting the manufacturing aspects will standardize the manufacturing process and provide reference
data during any new development, modification and periodic assessment..
Key Words: Composite, Defect, De-lamination, FOD, Manufacturing process, NDT, Reference standard, Ultrasonic
Testing,
1. Introduction
Increased use of composite in aerospace has increased the importance of NDT methods which are
capable of identifying flaws in composites. The composite manufacturing process makes the
composite parts susceptible to different types of defects which mainly occur due to: human error,
improper manufacturing process, improper materials etc. Hence there is a great need of establishing
NDT techniques capable of detecting manufacturing defects to consistently maintain the quality of
the products. NDT also plays a key role in gathering information about structural properties and
service life of aerospace components.
Ultrasonic testing is extensively used in composite inspection because of its ability to detect defects
such as: delaminations, disbonds, inclusion (FOD), porosity, and voids. In order to evaluate the
capability of defect detection and assessment of defect severity, it is essential to develop a test
method to standardize the ultrasonic system and it can be achieved by making use of reference
standards. Hence composite reference standards, as they are often called, are required to be
developed to ensure that the test equipment is able to correctly detect the signals for a particular
setup and procedure. Results obtained during reference standard testing shall give information on
choice of technique based on the following: inspected material, thickness range, object geometry,
required damage detection, the best sensitivity.
2. Typical defects in composite
(A) Resin rich: - Local resin content is higher than the average of laminate due to improper lay-up,
compaction or curing, (B) Resin starvation: - Local resin content is lower than the average of
laminate due to improper flow of resin, (C) Delamination: - Separation between two or more layers
in a laminate because of contamination or improper adhesion during processing, (D) Disbond: Separation between the bonded joints caused by either contamination or improper adhesion during
processing e.g disbond between laminate to core, sheets and core or sheet-to-sheet, (E) Void: Presence of air or gas trapped inside the laminate during curing of the component. It is of
measurable size, (F) Porosity: - Presence of small air or gas bubbles inside the laminate caused by
volatiles. It is of non-measurable size, (G) Inclusion (FOD): - Any undesirable materials, which are
inadvertently left in the bonding area of a composite structure.
This paper emphasis on defects like: Delamination, Disbonding & Inclusion (FOD)
3. Ultrasonic inspection of composite parts
Ultrasonic inspection is the most valuable technique for inspection of composite parts. Ultrasonic
operates on the principle of transmitted and reflected sound waves. As the ultrasonic beam passes
through the composite, it is attenuated or lost due to scattering, absorption, and beam spreading. This
loss or attenuation is usually expressed in decibels (dB). Thicker laminates will attenuate more
sound than thinner laminates. An ultrasonic wave traveling through a composite laminate that
encounters a defect such as delamination or porosity will reflect some of the energy at the interface
while the remainder of the energy passes through the defect. The more severe the defect, the greater
the energy reflected and lesser the energy transmitted.
Through transmission ultrasonic technique is one of the two most common methods used to inspect
composite laminates and assemblies. This method is based on two aligned transducers; one is a
transmitter and the other is a receiver and part is coupled between them. If the part contains a defect,
such as porosity or a delamination, some (or all) of the sound will be either absorbed or scattered, so
that some (or all) of the sound is not received by the receiving transducer. Through transmission
method is excellent at detecting porosity, disbonds, delaminations, and some types of inclusions
(FOD). However, this method cannot detect all types of foreign object inclusions and does not
provide data in terms of depth of defect.
Since through transmission method is not capable of detecting all types of foreign object inclusions
and the depth of defects, pulse echo ultrasonic inspection is frequently used in conjunction with
through transmission ultrasonic technique to inspect parts. Pulse echo method is based on
presentation of reflected sound from within the part or the back-wall surface using a single probe
acting as both transmitter and receiver. Sound is reflected from back-wall surface or from the
discontinuity due to impedance mismatch and presented as signal amplitude along the vertical axis at
the corresponding time of flight / distance along horizontal axis.
4. Composite reference standard salient features
Selection of a reference standard depends upon the technique of testing, type of material to be
inspected, form type of the discontinuity to be detected and other specification requirements.
Sometimes it is preferable or required to prepare a reference standard from a piece of the same
material as that of the component to be tested. The advantage of such a reference standard is that the
test object and the standard will have the same composition, manufacturing history, surface
condition and geometry. Here are specified some of the salient features required for a composite
reference standard:
4.1 The reference standard shall represent the design configuration and complexity of the
component to be inspected with respect to component type (laminate or sandwich), component
thicknesses, type of material, surface finish ( tool side & bag side or both tool side), secondary
bonding medium, internal structure (fittings), fabrication & curing processes.
4.2 The Reference standard(s) shall contain artificial discrepancies as per the minimum defect
dimensions stated in the Acceptance Criteria and which would produce a response for the normally
encountered defects like voids, delamination, disbond, inclusions (FOD), voids, and porosity. The
defect shape may be triangular or circular or square / rectangular. Records shall identify that
standards are free from natural material discontinuities that would make the reference standard
unsatisfactory for use.
4.3 The reference standard shall represent full range of part geometry (i.e minimum to maximum
thickness). Multiple standards shall be used to represent the full range of defect sizes, if not possible
to demonstrate in a single reference standard.
4.4 Ultrasonic reference standard shall have near surface and far surface reference defects to
establish near surface resolution and far surface resolution and shall contain defects to assure
sensitivity requirements (i.e minimum acceptable defect size) as per the acceptance criteria.
5. Manufacturing Aspects of Fabrication of Composite Reference Standard
The following points shall be considered during fabrication of composite reference standard:
5.1 Identify the Type of Reference Standard
First step in fabrication of any composite reference standard is to identify the type of reference
standard required and its applicability. The reference standard shall represent the design
configuration and complexity of the component to be inspected. The selection of the reference
standard type should consider the following points:
(i)
Its configuration should be Laminate composite or Sandwich composite
(ii)
Its configuration should either be a replica of the component type or standard panel type
(iii) Its applicable range with respect to part geometry (i.e. to coverage minimum to maximum
thickness)
(iv)
Its material type (exactly the same material as per part geometry or any other equivalent
material)
(v)
Its size for operator convenience (i.e. if it’s a replica of component: Full component or
partially component and if panel type: length, width and number of steps)
(vi)
Its surface finish required ( i.e. one tool surface & other bag surface or both tool surface)
Figure-1 specifies C-scan color plot of some typically reference standards type:
Component Type
Panel Type
Laminate Type
Figure-1
5.2 Identify the Type of Artificial Discontinuity
Sandwich Type
The Reference standard(s) should contain artificial inclusions which would be capable to produce a
response similar to that produced by typical defects like voids, delamination, disbond, inclusions
(FOD), voids, and porosity. The artificial discontinuity should create sufficient acoustic impedance
mismatch so that sound can be reflected from the discontinuity surface and can be detected. The
artificial discontinuity (i.e Inclusion or FOD) shall be selected such that it either creates a solid-air
interface to reflect large amount of sound or shall create a higher attenuation zone.
To identify the right type of artificial inclusions, a number of composite panels need to be fabricated
by simulating various FODs - of various thicknesses at different depth levels. Multiple iterations of
ultrasonic testing are then carried out to establish the relationship between ultrasonic attenuation
level and different types of inclusions.
The following steps can be followed for selecting the artificial discontinuity with respect to
ultrasonic attenuation level.
(i) Listing out some of the FOD items can be used as inclusion in composite reference standard as
shown in Figure-2
1. Release Film
2. Bagging Film
5. Brass sheet
6. Tooltech Tape
3. Prepreg Cutting Blade
4.Prepreg Protective Film-BD
7. Flash Breaker
8.Prepreg Protective Film-UD
Figure-2
(ii) Fabricating an NDT panel including various FOD items at different ply level, to simulate the
defect conditions as shown in Figure-3
A typical Carbon & Glass Panel with FODs nearer to Tool Side & Bag Side during fabrication
Figure-3
(iii) Establishing relation between ultrasonic attenuation level with various types FODs in Composite
by ultrasonic A-Scan & C-Scan as shown in Figure-4 & Figure-5
Ultrasonic A-Scan PE Method (5.0 MHz)
Ultrasonic A-Scan TTU Method (5.0 MHz)
Figure-4
Ultrasonic C-Scan TTU Method
(1.0 MHz)
Ultrasonic C-Scan TTU Method
(2.25 MHz)
Ultrasonic C-Scan TTU Method
(5.0 MHz)
Figure-5
5.3 Defining size and location of artificial discontinuity
The reference standard(s) must contain artificial defects / discontinuity of dimensions that meet the
requirement of Acceptance Criteria.
The reference Standard(s) shall:
(i) Have discontinuities at all possible locations covering the full range by having the required sizes
and types of inclusions placed in radii, at bond lines, at chamfers and in flat areas,
(ii) Incorporate inclusions at minimum 3 locations for laminate construction i.e. minimum, midrange
and maximum thickness of the representing parts and at 2 locations for sandwich construction, i.e
between core and top skin & between core and bottom skin of the representing parts,
(iii) Have reference defects to assure sensitivity requirements (i.e maximum acceptable defect size or
minimum unacceptable defect size dimension) as applicable to the acceptance criteria,
(iv) Contain defects of various shapes i.e triangular, circular, elliptical, square, and rectangular,
(v) Have near surface and far surface reference defects to accurately establish near surface resolution
and far surface resolution as applicable to the acceptance criteria or better.
5.4 Preparing detail drawings / sketch of the reference standard
A detailed drawing for the reference standard should be prepared on which following information
would be specified:
(i) Type of reference standard and its identification number, (ii) Applicable material list with cure
ply thickness of each ply & honeycomb core thickness, (iii) Complete dimensions in terms of length,
width & thickness, (iv) Number of ply for each thickness configuration and applicable ply number,
(v) Applicable Lay-up plies orientation, (vi) Applicable manufacturing curing cycle, (vii)
Information regarding types, sizes & depth location of artificial defect to be incorporated, (viii)
Dimensions between two artificial defects and dimension with respect to edge of panel, (ix)
Applicable Tool side & Bag side of the panel, (x) Excess material in all the sides for final trimming.
Figure-6 specifies typically Drawings of a reference standard
Figure-6
5.5 Preparing the detailed manufacturing process plan / lay-up scheme
A detailed manufacturing process plan should be prepared with respect to the drawing of the
reference standard and it would include the following:
(i)
A detailed route sheet / scheduling sheet, which shall define various work centers and various
manufacturing & inspection operations as applicable like: issue of raw material, tool
preparation, material cutting and ply identification, ply lay-up, intermediate compaction,
vacuum bagging, curing, demoulding, final trimming / machining, visual & dimensional check,
NDT checks and final tagging / accepting.
(ii) A detailed lay-up scheme specifying ply number, ply orientation, ply dimension, ply material,
defect number, defect material, defect dimension, compaction schedule etc. in addition to the
route / scheduling sheet.
(iii) A detailed manufacturing curing cycle specifying: Heating rate, Cooling rate, Dwell
temperature, Dwell Time, applicable vacuum, applicable pressure etc. & all other supporting
documents like: various process work instructions, copy of drawings, Copies of Do’s & Don’ts
etc.
5.6 Preparing detail defect dimension report before lay-up
As it is quite a challenge to create artificial discontinuities to the exact size as per drawing
requirements, it is advisable that the before manufacturing of reference standard all the artificial
discrepancy dimensions shall be measured and recorded, so that during ultrasonic scanning of the
reference standard the dimensions shall be verified for its accuracy and reference standard shall be
certified or validated. It is also necessary to verify the defect / discrepancy dimensions during system
standardization of the equipment. The defect dimensions report shall reflect all defect dimensions in
00 Direction & 900 directions along with the thickness of the defect.
Below Figure-7 (Table-1) specifies a typically lay-up scheme & Figure-7 (Table-2) specifies a
typically defect dimensions report:
Table-1
Table-2
Figure-7
5.7 Documenting & certifying the detail manufacturing process of reference standard
During manufacturing phase of the reference standard, it is the responsibility of NDT personnel &
QA personnel to ensure all the process documents are verified and certified for its compliance. The
following should be documented and certified during fabrication stage:
(i) Material Details: Lot Number, Sub-lot / roll Number, Shelf life Expiry, Out life as applicable, (ii)
Cutting of all the plies as per lay-up scheme or drawings and identification of the same with ply
number, (iii) Tool / Template condition, its identification and tool preparation for lay-up, (iv) Layout
marking (orientation, reference points, dimension scaling in X & Y direction), (v) Lay-up of the
plies and artificial discontinuity as per lay-up scheme, (vi) Intermediate compaction of the plies with
sufficient vacuum if applicable, (vii) Final vacuum bagging and leak check, (viii) Curing details
from cure chart (i.e Temp, Dwell time, Heating & cooling rate, pressure, vacuum etc.), (ix)
Demoulding of the part after curing, (x) Final trimming and edge sealing as applicable, (xi)
Destructive testing of coupons to certify curing process if applicable.
5.8 Preparing the detailed Visual & Ultrasonic scan report
One of the important phases in the fabrication of an ultrasonic reference standard is to prepare a
detailed Non-destructive inspection report for the standard. The NDT report shall contain: Visual,
Ultrasonic A-Scan & Ultrasonic C-Scan as applicable.
Visual inspection should be carried out on the reference standard to verify the standard is free from
edge delamination, surface irregularity, marks (or dents), wrinkles, resin rich, resin starvation etc
which might affect the inspection result.
Ultrasonic A-Scan shall be carried out to confirm the ultrasonic response from the artificial
discontinuity with respect to Defect size, Defect depth location and delectability. Ultrasonic A-Scan
pulse echo method shall be carried out from tool surface to confirm that the standard complies with
the drawing requirements. For near surface resolution Lucite shoe or delay line transducer can be
used if required. Attenuation in laminate reference standard shall not vary more than  2 to 3 dB
across regions of same thickness.
Based on the availability of facility and complexity of the reference standard, Ultrasonic C-Scan
shall be carried out to verify the ultrasonic response from the artificial discontinuity with respect to
defect size and the ability of being detected. Whenever there is any difficulty in interpreting the CScan color plot due to intermixing of colors, C-Scan grey plot shall be preferred to confirm the
discontinuity.
Figure-8 specifies typically A-Scan & C-Scan report of a reference standard:
Reference
Standard
Photo
Ultrasonic A-Scan reporting of artificial
discontinuity with detail dimension &
location of defect
C-Scan Grey Plot
C-Scan Color Plot
A-Scan echo response
Figure-8
5.9 Preparing the detailed dimensional report & identification of the reference standard
The detailed dimensional report of the reference standard should be prepared as a part of the
inspection report and should be maintained for the reference standard. The detailed dimensional
should be prepared after the final trimming operation so that the reference standard is of the exact
size. The dimensional report should reflect the reference standard dimension as well as defect
dimension and its position from reference points.
If the reference standard is not identified with a suitable identification number during fabrication,
then the identification shall be done either by painting on the surface, a hanging metal tag by drilling
a hole on it or sticking any other suitable name plate by wet lay-up method. The identification shall
be placed only at locations where no artificial discontinuity place.
Figure-9 specifies typically dimension report of a sandwich reference standard:
Figure-9
5.10
Validating the reference standard for use
Finally the reference standard should be validated and certified by NDT Level-3 personnel.
Validation of reference standard can be performed in various ways to verify sensitivity, resolution
(Near surface & Far surface), minimum unacceptable defect size as per quality grades and
compliance of the reference standard with drawing requirements. The validation can be performed
with one or more NDT techniques as specified below
(i)
Ultrasonic A-Scan Pulse-Echo Method to validate Ultrasonic A-Scan ThroughTransmission method or vice versa
(ii)
Ultrasonic C-Scan Method to validate Ultrasonic A-Scan Method
(iii)
Radiographic CT-Scan or any other NDT Method to validate Ultrasonic C-Scan Method
Figure-10 & Figure-11 specifies typically method of validating a reference standard:
Verification of Sensitivity & Resolution on a reference standard by Ultrasonic C-Scan
Verification of Near Surface & Far Surface Resolution on a reference standard by Ultrasonic C-Scan
Verification of a Stepped Laminate Thickness calibration
reference standard by Ultrasonic C-Scan
Figure-10
Scout View
Typical CT-Scan Result of defect area
Higher attenuation indication at
De-lamination Zone
Typical C-Scan Result of defect area
Validation of Ultrasonic C-Scan Results By Radiographic CT-Scan
Figure-11
6. Conclusion
The responsibility of the fabrication of composite reference standards solely lies with the
manufacturer and NDT personnel involved in the process. During fabrication of composite reference
standards, several factors shall be considered starting from design, material selection, manufacturing
processes like lay-up and curing, NDT inspection methods, minimum defect size, defect location at
various depths, defect types etc. to achieve a good NDT reference standard. All the necessary data
with respect to the manufacturing process needs to be generated by the composite manufacturing
organization by introducing various artificial inclusions of various sizes at various depths to simulate
the actual defect occurrence during manufacturing. Documenting all the manufacturing aspects will
standardize the reference standards and validate the composite manufacturing process during any
new development, modification and periodic evaluation.
7. Reference
1. AMS CACRC Commercial Aircraft Composite Repair Committee, “Composite Honeycomb NDI
Reference Standards”, SAE Aerospace Recommended Practice ARP 5606, Dec 2011
2. AMS CACRC Commercial Aircraft Composite Repair Committee, Composite Repair NDT/NDI
Handbook," SAE Aerospace Recommended Practice ARP 5089, Nov. 2011
3. Yolken H. Thomas and Matzkanin George A., “Nondestructive Evaluation of Advanced Fiber
Reinforced Polymer Matrix Composites”, February 2009
4. Roach Dennis & Dorrell Larry, “Development of Composite Honeycomb and Solid Laminate
Reference Standards to Aid Aircraft Inspections”, NDT.net - March 1999, Vol. 4 No. 3,
5. Galella Dave, “FAA Inspection Administration Research Activities for Composite Materials”,
Composite Damage Tolerance & Maintenance Workshop, ATO-P, July 2006
6. Oster Reinhold, “Non-destructive testing methodologies on helicopter fiber composite
components challenges today and in the future”, World Conference on Nondestructive Testing,
April 2012
7. Martin R, “Ageing of Composites", Woodhead Publishing Ltd, 2008
8. HSU David K., “Nondestructive Inspection of Composite Structures: Methods and Practice”,
World Conference on Nondestructive Testing, Oct 2008
9. Gower, M., Sims, G., Lee, R., Frost, S. and Wall, M., Measurement Good Practice Guide No. 78
“Assessment and Criticality of Defects and Damage In Material Systems," National Physical
Laboratory, Teddington, Middlesex, United Kingdom, June 2005.
Wave Propagation in a Delaminated Thin Anisotropic Strip
Punith P, M. Mitra and P J Guruprasad
Department of Aerospace Engineering
Indian Institute of Technology Bombay
Mumbai 400076, India
Contact: punithatm.prakash@gmail.com
Thin pre-twisted beams, made of laminated composites, are increasingly being used in helicopter
rotor blades, wind turbine blades and propellers. In the case of helicopter rotor blades, to
accommodate the centrifugal force along with flapping, lead-lag, and torsional motion, flexible
structural component known as flex-beams are being used. This allows for the design of a rotor
system that is independent of bearings – making the design bearing-less. However, the design of
flex-beams and flex-beam like structures found in other applications is challenging owing to the
nonlinearity that arise because of the large displacements and moderate rotations. This
geometrical nonlinearity problem coupled with the anisotropy at the material scale adds to the
complexity of the design. This challenge is further compounded when the design has to account
for damages typical of laminated composites, ex: delaminations.
In this work, the delamination in a composite strip is modeled through variational asymptotic
method to obtain the governing equations. The equations are applied over to a delaminated pretwisted composite strip and solved using spectral finite element (SFE) method, wherein the
structure would be modeled as a wave guide and the dynamic stiffness matrix is derived in
frequency domain. The wave responses of the delaminated composite structure for different loads
and conditions are computed. Further, the model is validated by comparing the result obtained
from SFE method with the results obtained from a commercial Finite element (FE) package and
available literature.
The use of SFE is expected to yield a faster result and lesser computation than the FE method for
analyzing and monitoring the delaminated strip exposed to high frequency excitations.
Additionally, the SFE based modeling scheme are suitable for the inverse problem of damage
detection.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Rapid ultrasonic inspection of stiffened composite ailerons
Rabi S. PANDA 1, Prabhu RAJAGOPAL 1, Krishnan BALASUBRAMANIAM 1
1
Centre for Nondestructive Evaluation and Department of Mechanical Engineering,
Indian Institute of Technology Madras, Chennai-600036, Tamil Nadu, India
Phone: +91 44 2257 5688, Fax: +91 44 2257 0545; e-mail: balas@iitm.ac.in
Abstract
This paper demonstrates an approach for rapid non-contact air-coupled ultrasonic inspection of composite
ailerons with complex cross-sectional profile including thickness changes, curvature and the presence of a
number of stiffeners. A combination of plate guided ultrasound and air-coupled ultrasonics are used to determine
the propagation characteristics of the waves with different levels of complexities in the structure. Experimental
B-scans are generated with pitch-catch arrangement of probes for different regions of the test sample, capturing
the feature-rich sample profile. Guided wave propagation and interaction with the interface defects such as
disbonds between skin and stiffeners is also studied
Keywords: Composite aileron, air-coupled ultrasound, disbond
1. Introduction
Composite stiffened panels are widely used for main structural components of aircraft such as
wing, fuselage and empennage, and are required to be lightweight. These panels consist of a
thin skin reinforced with T-, I-, or hat shaped stiffeners. The stiffeners placed in between the
top and bottom skins provide remarkable bending rigidity, thereby reducing the overall panel
weight. However, these stiffeners are often co-cured or co-bonded to the skins using
adhesives instead of bolts or fasteners for additional weight reduction; also it provides more
uniform stress distribution. The skin-stiffener interface is prone to disbond due to
manufacturing flaws or impacts under service environments. Also, low velocity impacts, such
as tool drops can cause delaminations within the skin laminates/stiffener body. These defects
are often in the hidden/inaccessible regions of the skin-stiffener assembly and may not be
detected from the skin side [1]. Nondestructive testing (NDT) techniques [1-3] may be used to
identify skin-stiffener disbonds or delaminations in the skin/stiffener body. Non-contact
ultrasonic guided wave technique [3-5] is promising for structural health monitoring (SHM)
due to their desirable properties such as their relatively high range and sensitivity to small
damage.
The present work focuses on the results of experimental investigation of the
interaction of Lamb waves with composite structure using fully non-contact air-coupled [5-7]
ultrasounds. The test sample considered in this study is a stiffened composite panel, called as
"aileron" of the wing assembly of an aircraft, shown in Figure 1. It has various regions such as
skins, stiffeners and bend region. Structural damage is proposed to be detected acquiring a Bscan map [3,8,9] in pitch-catch arrangement, since they require less time than C-scans and
allow for rapid screening of structural components. Ultrasonic measurements are taken with
different arrangements of transmitter-receiver pair in order to inspect the skins, stiffeners and
bend region of the sample as explained in Section 3. The fundamental anti-symmetric (A0)
Lamb wave mode [2,8] is used for damage detection in inaccessible regions. The results are
presented in terms of B-scan images. These results are then compared with the results
obtained by the conventional water immersion testing to check the effectiveness of the present
technique.
Figure 1. Photograph showing various regions of the composite aileron sample
The paper is organised as follows. Section 2 gives a brief background to the choice of
the excitation frequency and Lamb wave mode selection along with the problem defination.
Then details of the experiment and different arrangements of the transmitter-receiver pair to
inspect the various regions of the aileron sample are given in Section 3. Experimental results
are presented with discussions in Section 4. Section 5 concludes the study with directions for
future work.
2. Background
Interaction of Lamb waves with various types of defects in composite structures are
investigated by many researchers [1-13]. The choice of the excitation frequency, mode
selection [2] is an important consideration in defect detection. Low-frequency excitation of
100 kHz was considered in this study to minimise the number of modes in the received signal
to limit the complexity in post-processing. Also the modes are relatively non-dispersive in
nature. At this frequency range, only fundamental symmetric (S0) and antisymmetric (A0)
modes exist. The criterion for mode selection was based on wavelength, which should be
equal to or smaller than the size of the damage to be detected. Here, we used A0 mode for
damage detection in view of its shorter wavelength, to be able to detect small damages. Aircoupled transducers are used to generate and receive A0 mode Lamb waves in the test sample.
Optimal incident and receiving angles are calculated from the dispersion [14] curves and
Snell’s law in order to generate and receive A0 mode, and it was found to be approximately
170 at the center frequency of 100 kHz.
The main objective of this study is to inspect the stiffened composite aileron to detect
and characterize the presence of damages. Air-coupled ultrasonic transducers are used in
pitch-catch arrangement for damage identification. The presence of damage can be detected
by comparing the experimental signals with the corresponding baselines obtained on the
healthy regions of the test sample. Experimental results in time domain demonstrate the
integrity of the whole aileron structure with approximate damage position in skin-stiffener
bonding, bend region, and body of the stiffeners.
3. Experiments
The experiments were performed in a custom made C-scan like setup, shown in Figure 2. The
experimental setup consists of a custom-built 3-axes scanner, a pulser-receiver, a 50-MHz
A/D card, a motion controller and a personal computer. The aileron structure is made from
carbon/epoxy material. It has upper and lower skins, supported by seven I-shaped stiffeners.
Guided waves were actuated and received by Ultran NCG-100 D25 air-coupled transducers
(http://www.ultrangroup.com/) with the center frequency of 100 kHz and active diameter of
25 mm. The A0 mode was generated with transmitter-receiver pair in pitch-catch
configuration. The air gap between each transducer and the sample was approximately 10
mm, corresponding to the best signal amplitude of the received A0 wave mode.
Figure 2. Photograph of the experimental setup
The broadband excitation pulses with the peak amplitude of 900 V were applied to the
transmitter at a burst period of 20 ms using Panametrics 5058 PR generator. Mechanical
scanners and the setup for ultrasonic data collection were synchronized with the help of the
LabView program. The experimentally obtained A-scans were averaged 16 times to increase
the signal-to-noise ratio. These signals were saved and processed using a MATLAB program
to generate B-scan images. The variation in the B-scan image gives the approximate position
of the damage/discontinuity in the structure. The configurations of the transmitter-receiver
pair for inspections of various regions of the aileron are shown below.
3.1 Inspection of skin and skin-stiffener bonding
Figure 3 shows the setup for inspection of skin and adhesive bonding between skin and
stiffeners. The transmitter and receiver were used in pitch-catch mode along the length of the
skin. The positions of stiffeners and other structural features are marked by white lines in the
sample. The transmitter is marked as 'T' and receiver as 'R' in the setup.
(a)
(b)
Figure 3. Photograph showing: (a) inspection of the skin with air-coupled ultrasonic transducers; (b) inspection
of bonding quality by sending guided waves across the bondline
3.2 Inspection of bend region
The setup for inspection of bend region of the aileron is shown in Figure 4. The transmitter
and receiver were used in pitch-catch mode along the length of the bend. The wave path is
across the bend as shown below.
(a)
(b)
Figure 4. Photograph showing: (a) wave path; (b) scan direction during scanning
3.3 Inspection of body of the stiffeners
Figure 5 shows the setup to inspect the body of stiffeners using air-coupled transducers. In
this case, the transmitter and receiver were placed in both side of the sample in pitch-catch
mode, inclined at an angle of 180 to the normal to the surface of the skin for exiting and
receiving A0 mode. The scanning was performed along the length of each stiffener.
(a)
(b)
Figure 5. Photograph of the setup for inspection of bondline and stiffener body: (a) wave path; (b) scan direction
4. Results and discussion
The experimental results were obtained for different regions such as skins, skin-stiffener
bonding, bend and stiffeners of the aileron sample with various arrangements of the
transmitter-receiver pair as shown in Section 3. The inaccessibility of the stiffeners from
outside poses difficulty in inspection. Here we have presented the inspection results of the
stiffeners along with its bonding with skin.
4.1 Air-coupled ultrasonic inspection
The photograph of the aileron test sample, from skin side is shown in Figure 6a. The stiffener
positions are marked by white lines and also numbered as 1 to 7. Each stiffener was scanned
separately along its length with transmitter-receiver pair on top and bottom skin in pitch-catch
configuration (see Figure 5a). The position of the transmitter is marked as 'T' and receiver as
'R'. Guided waves were generated on the upper skin and propagated to the lower skin through
stiffener. The B-scan results showing the acoustic responses of the healthy stiffener and
disbonded stiffeners are presented in Figure 6. In contrast to stiffener 4, the stiffeners 1 and 2
produce lower response at some regions, which indicates that the damage is present in the
wave path between the transducers. The damage could be the disbond between the skin and
stiffener or the delamination in the body of the stiffener. Since the amplitude of the signal is
nearly lost in some regions (Figure 6c & 6d, encircled region) implies that very less amount
of energy is going into the stiffener, which confirms the presence of disbond between skin and
stiffener. The time of arrival of the A0 mode is changing when the transducers move from the
bend region towards the edge due to the increasing length of the stiffener body inside the
sample and also change of the air gap between the transducers and the skins. In view of the
time domain response, the approximate position of disbond is marked in Figure 6c and 6d.
(a)
(b)
(c)
(d)
Figure 6. Figure showing the B-scan results along a stiffener: (a) Photograph of the skin with position of
transmitter and receiver, scan length in red arrow; (b) stiffener 4 with no disbond; (c) stiffener 2 with small
disbond; (d) stiffener 1 with large disbond
4.2 Conventional ultrasonic bulk wave inspection
The aileron test sample was also inspected using ultrasonic C-scan technique through water
immersion. Bulk waves were excited using the Panametrics transducer with 15 MHz center
frequency and collected in pulse-echo configuration with Olympus Focus LT module. Figure
7 shows the photograph of the top skin with immersion C-scan results. The dark red in the
colour scale corresponds to the maximum energy level. It was observed that, the stiffeners 1
and 2 (Figure 7) have maximum energy levels at some regions. Since the scan was carried out
with pulse-echo mode, the maximum energy level at those regions indicates that most of the
waves were reflected from the interface between skin and stiffener. This confirms the
presence of disbond between the skin and stiffener. This may be due to bad curing of the
adhesive or improper surface preparation.
(a)
(b)
Figure 7. (a) Photograph of the top skin of composite aileron; (b) C-scan of the corresponding region
Also, the aileron had variations of skin thickness in some regions, which can be seen
from the colour variation in the C-scan image. The C-scan result was also compared with the
air-coupled ultrasonic inspection results, which confirms the position of disbonds.
5. Conclusions
This study addresses the experimental investigation of a stiffened composite structure using
non-contact air-coupled ultrasonic technique. Different configurations of air-coupled
ultrasonic transducers in pitch-catch configuration were used to inspect skins, skin-stiffener
bonding and stiffeners in the test sample. The fundamental A0 mode was selectively excited
and found to be very effective in detecting disbonding in the structure. B-scan images were
generated from the air-coupled ultrasonic measurements of different regions of the structure
for rapid inspection. The time domain amplitude variations in the B-scan images give
approximate position of the damage in the structure. Results demonstrate that critical defects,
such a large disbonds between the skins and stiffeners can be detected using the A0 mode
based air-coupled scanning technique. Also the results obtained from the guided wave aircoupled ultrasonic technique were compared with the results from the water immersion bulk
wave inspection. The results in both cases are in good agreement. It was concluded that, the
guided wave air-coupled inspection is effective for damage detection in inaccessible regions
of a composite structure. Future study will more focus on the development of signal
processing algorithms capable of producing better images of defects, as well as on finite
element simulations that will help to better explain the physics of some of the obtained guided
wave signals.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
M J Padiyar and K Balasubramaniam, 'Lamb-Wave-Based Structural Health Monitoring
Technique for Inaccessible Regions in Complex Composite Structures', Structural
Control and Health Monitoring, Vol 21, No 5, pp 817-832, 2014.
F Ricci, E Monaco, L Maio, et al., 'Guided Waves in a Stiffened Composite Laminate
With a Delamination', Structural Health Monitoring, Vol 15, pp 351-358, 2016.
M J Padiyar and K Balasubramaniam, 'Quantitative Characterization of Interface
Delamination in Composite T-Joint Using Couplant-Free Lamb Wave Methods', Journal
of Reinforced Plastics and Composites, Vol 35, No 4, pp 345-361, 2016.
G K Geetha, D R Mahapatra, S Gopalakrishnan, et al., 'Laser Doppler Imaging of
Delamination in a Composite T-Joint With Remotely Located Ultrasonic Actuators',
Composite Structures, Vol 147, pp 197-210, 2016.
R S Panda, P Rajagopal and K Balasubramaniam, 'Characterization of DelaminationType Damages in Composite Laminates Using Guided Wave Visualization and AirCoupled Ultrasound', Structural Health Monitoring, September 7, 2016.
(doi:10.1177/1475921716666411)
Z Liu, H Yu, C He, et al., 'Delamination Detection in Composite Beams Using Pure
Lamb Mode Generated by Air-Coupled Ultrasonic Transducer', Journal of Intelligent
Material Systems and Structures, Vol 25, pp 541-550, 2014.
M Castaings and P Cawley, 'The Generation, Propagation, and Detection of Lamb
Waves in Plates Using Air-Coupled Ultrasonic Transducers', The Journal of the
Acoustical Society of America, Vol 100, pp 3070, 1996.
C Ramadas, J Padiyar, K Balasubramaniam, et al., 'Lamb Wave Based Ultrasonic
Imaging of Interface Delamination in a Composite T-Joint', NDT & E International, Vol
44, pp 523-530, 2011.
R Kazys, A Demcenko, L Mazeika, et al., 'Air-Coupled Ultrasonic Non-Destructive
Testing of Aerospace Components', Insight - Non-Destructive Testing and Condition
Monitoring, Vol 49, pp 195-199, 2007.
10.
11.
12.
13.
14.
Z Su, L Ye and Y Lu, 'Guided Lamb Waves for Identification of Damage in Composite
Structures: A Review', Journal of Sound and Vibration, Vol 295, pp 753-780, 2006.
M Mitra and S Gopalakrishnan, 'Guided Wave Based Structural Health Monitoring: A
Review', Smart Materials and Structures, Vol 25, pp 53001, 2016.
A Raghavan and C E S Cesnik, 'Review of Guided-Wave Structural Health Monitoring',
The Shock and Vibration Digest, Vol 39, pp 91-114, 2007.
E Monaco, N D Boffa, V Memmolo, et al., 'Detecting Delaminations and Disbondings
on Full-Scale Wing Composite Panel by Guided Waves Based SHM System', SPIE
Proceedings-9805, Health Monitoring of Structural and Biological Systems, March 20,
2016.
Disperse User's Manual, Version 2.0.11, June 2001.
EVALUATION OF COMPOSITE STRUCTURES – USING COMPUTED
TOMOGRAPHY (CT) EMERGING NDE METHODOLOGY
1
1
2
3
Ramesh Babu.G , Pranab Biswal , B.N.Srinivasa Reddy , P.Mukhopadhyay
1
2
3
Senior Manager (NDT) , Chief Manager (NDT) , Deputy General Manager (QA)
1
1
drgrameshbabu@yahoo.co.in ,pranab.biswal@hal-india.com , srinivasareddybn@gmail.com
Aerospace Composites Division, Hindustan Aeronautics Limited, Bangalore-560037
2
Abstract
Composite components are extensively used for various purposes and are effectively used in Aerospace structural
applications. Glass Fibre Reinforced Plastics (GFRP) & Carbon Fibre Reinforced Plastics (CFRP) yield less weight and
long life to enhance safety and strength of the structure. The objective of this work is to optimize compaction level to
achieve better Quality, reliability and enhanced product life using Computed Tomography (CT), a robust Non-Destructive
Evaluation (NDE) methodology.
Computed Tomography (CT) attenuation measurements serve as a very sensitive indicator of internal loss caused by
microstructures and micro-structural alterations in the test material. This also helps in Structural Health Monitoring (SHM)
related studies on damage in aerospace composites structures. X-Ray Computed Tomography is a robust NDE method to
characterize the defects and damage in composites for health monitoring of critical aircraft components and to extend the
life of components.
Inter laminar shear strength (ILSS) variations are considered while studying the sensitivity of compaction level on the
performance of helicopter blades during service. An appropriate density mapping through CT number & attenuation
distribution through dB variations are obtained from the Fatigue test, component test data to assess their safety and
structural reliability.
In the case of thick-wall, monolithic, self stiffened panels and sandwich composites, the conventional NDT-test procedures
fail. X-ray Computed Tomography (CT) as an imaging test procedure closes this gap and adds more confidence to evaluate
the part through study of internal structure giving detailed information through cross-sectional view presentation.
In this study, ultrasonic attenuation and CT Number are correlated to inter-laminar shear strength (ILSS), Fibre Volume
content, void content etc. Based on the correlation, the acceptable limits for CT Number is determined with respect to
ultrasonic attenuation, ILSS and fibre volume fraction. During curing of advanced Polymer Matrix composites (PMC),
adequate pressure needs to be applied. This consolidation pressure enhances compaction of lay-up, assists flow of resin to
fill all the gaps and controls the void growth. Improper distribution of pressure results in changes in fibre volume fraction,
voids and poor consolidation leading to poor quality which can be assessed during CT-Scan.
Experimental results were analyzed and successfully used to achieve good compaction at micro level for various complex
Composite structures. The enhancements in performance of the structures were further confirmed by carrying out Fatigue
test & Resonance test. Influence of compaction level on the structural reliability/performance of parts was verified.
Based on the performance of the part, part life can be extended in order to achieve the better Quality, reliability and
enhanced product life using process control approach.
Keywords: Fiber volume ratio, CT number, Compaction level, Computed Tomography (CT), Resin rich. Differential
Scanning Calorimeter (DSC), layer waviness, on set Temperature (deg C)
1. INTRODUCTION
In the field of aeronautical industry, CFRP (Carbon fibre Reinforced Plastics) & GFRP (Glass fibre
Reinforced Plastics) are widely used mainly due to their high specific mechanical properties. The main
focus of the aerospace industry research on multi-functional materials is to reduce their own weight,
enhance their mechanical properties and sensing capability.NDT based Structural Health Monitoring
(SHM) system informs the damage development and location of the damage to the engineer adequately.
There are different types of NDT Techniques namely, Ultrasonic Testing (UT), Radiographic Testing
(RT), Shearography, Thermography and so on [2-4].All these techniques are used for monitoring the
composite structure. But each technique has its own merits and demerits. By using CT in composite
structures to detect damages, images are generated and analyzed. This CT image is generated in the
composite structures as they need strong penetration of x-rays to detect even a small damage on the
structure internally.
Damage can also be detected by different NDT methods like UT,RT, Acoustic Emission Testing;
However, such NDT technique will not give clarity against fibre orientation, matrix percentage during
process cycle. Even though UT is used for damage detection, loss of transmission signal (attenuation)
might occur due to surface irregularity developed in composite structures which will directly affect the
quality of detection. If the damage is in the small size (micro), this UT technique will not be detected
and this forms the main drawback of this technique. To overcome of this drawback, we will go for CT
and this technique gives detailed internal information in cross-sectional image pictorial representation
to detect micro level damages caused in the composite structures. Through CT technique, damage can
be checked precisely and pinpoint the defect location, it plays a key role in collection of information
about composite structural properties and service life of aerospace components.
2. Computed Tomography (CT) inspection of composite parts
X-ray computed tomography (CT) can play an important role in the evaluation of composite materials
and structures. Thick graphite- or plastic-fiber composites, multilayer bonded structure, honeycomb,
and ceramic and metallic composites can exhibit difficult-to-interpret indications from standard NDE
methods. Computed tomography can be used to detect and measure features, such as delamination,
porosity/voiding, resin rich/resin-poor, bond line fill, cracking, dimensions, and so on, providing useful
information during the composite structure development or the evaluation of nonconforming articles.
While conventional radiography creates a shadow graph containing superposition of information, x-ray
CT uses measurements of x-ray transmission from many angles around a component to compute the
attenuation coefficient of small volume elements. This data is presented as a cross-sectional image map
of the object. The clear images of interior planes of an object are achieved without the confusion of
superposition of features. Computed tomography results are easy to interpret for feature detection and
placement. Computed tomography can provide quantitative information about the density/ constituents
and dimensions of the features imaged. Figure 1.1 demonstrates the application of CT to a composite
Main Rotor Blade (MRB). The CT slice gives the internal configuration, dimensions, and material
variation without the confusion of superposition of information in the radiograph. Multiple CT slices at
sequential locations along the vertical axis are required to image the entire blade with CT.
There is a significant point about the application of CT, because very often relatively large image
voxels (compared to very fine defects) may be used, but small features are still detected, although they
are not necessarily resolved. The contrast sensitivity typically provided by CT is in the range of 0.1 to
1%.Computed tomography data allows accurate evaluation of dimensions and locations in threedimensional object space or material density (as related to x-ray linear attenuation coefficient) to be
performed in any orientation throughout the volume of an object that has been scanned with the CT
system. Tables 1.0 and 2.0 list the benefits and limitations of CT for composite materials and structures
that result from the capability and complexity of the CT measurements.
Table 1.0 -Capabilities of computed tomography for NDE of Composites
Table 2.0 - Limitations of computed tomography for NDE of composites
Composite materials, because of their non-homogeneous, anisotropic characteristics, pose significant
challenges for defect detection and materials property characterization. Throughout their life cycle,
composites are susceptible to the formation of many possible defects, primarily due to their multiplestep production process, non homogeneous nature and brittle matrix. These defects include
delaminations, matrix cracking, fiber fracture, fiber pull out, impact damage, ply gap, ply waviness,
porosity that may appear in composite laminates and their effect on the structural performance can also
have critical effect on the host structures. The use of composites in aerospace applications especially in
dynamic rotor systems is largely due to their exceptional performances and high property- to-weight
ratios. These materials are distinguished by their high strength and rigidity, low density, excellent
damping properties and high resistances to impacting and corrosion combining with modifiable
thermal expansion to complement complex characteristics profile. Because of their excellent
mechanical properties, CFRP and GFRP materials have been widely used for critical dynamic
components and structures, such as aircraft fuselage and wing structures, helicopter rotors and
windmill blades, road and marine vehicle body structures, and bridges and large civil infrastructures.
2.1 Computed Tomography (CT) attenuation measurements serve as a very sensitive indicator
mechanisms of internal loss caused by microstructures and micro-structural alterations in the test
material. This sensitivity stems from the ability of X-Ray Computed Tomography(CT) (of an
appropriate kV rating) to interact with a variety of discontinuities including delamination, inclusions,
FOD, grain boundaries, voids , fibre dislocations etc.observed during manufacturing and service. This
also helps to improve compaction level between layers and in Structural Health Monitoring (SHM)
related studies[15] on damage in aerospace composites structures. X-Ray Computed Tomography [3]
is a robust NDE to characterize the defects and damage in composites for health monitoring of critical
aircraft dynamic components in Helicopters and to extend the life of components.
In the case of thick-wall, monolithic, self stiffened panels and sandwich composites, the conventional
NDT-test procedures fail. The X-ray Computed Tomography (CT) as an imaging test procedure closes
this gap and add more confident to evaluate the part through internal structure giving detailed
information by sectional view presentation. CT is especially suitable for applications involving spatial
analysis [1], differentiation, material identification, flaw analysis and structural quantification. This is
made possible by the good spatial resolution and the very good resolution of density of the CTmeasurements.
In this study, CT Number is correlated to percentage of resin rich, depth of ply waviness(amplitude),
Fibre volume content, void content etc. Based on the correlation, the acceptable limits for CT Number
is determined with respect to Resin percentage, ply waviness depth and fibre volume fraction. During
curing of advanced Polymer Matrix composites, adequate pressure need to be applied, this
consolidation pressure enhances compaction of lay-up, assist flow of resin to fill all the gaps and
controls the void growth. Improper distribution of pressure results in changes in fibre volume fraction,
voids and poor consolidation leading to poor quality which can be assessed during CT-Scan.
Fig 1.1 TYPICAL CT-SCAN MACHINE AT ACD-HAL
Fig 1.2 DSC MACHINE LOCATED AT ACD-HAL
2.2 DIFFERENTIAL SCANNING CALORIMETRY (DSC)
Differential scanning calorimetry (DSC) measures enthalpy changes in samples due to changes in their
physical and chemical properties as a function of temperature or time.
The Figure1.2 DSC[12,13] is used to measure melting temperature, heat of fusion, latent heat of
melting, reaction energy and temperature, glass transition temperature, crystalline phase transition
temperature and energy, precipitation energy and temperature, denaturization temperatures, oxidation
induction times, and specific heat or heat capacity.DSC [11] measures the amount of energy absorbed
or released by a sample when it is heated or cooled, providing quantitative and qualitative data on
endothermic (heat absorption) and exothermic (heat evolution) processes.
Temperature can range from -120°C to 725°C, though an inert atmosphere is required above 600°C.
The temperature is measured with a repeatability of ±0.1°C.
Sample size: from 0.5mg to 100mg.
Samples can be encapsulated in aluminum pans using a pan press. DSC [14] used to determine the
thermal properties of plastics, adhesives, sealants, metal alloys, pharmaceutical materials, waxes, foods,
lubricants, oils, catalysts, and fertilizers.
1.3 APPLICATIONS OF DIFFERENTIAL SCANNING CALORIMETRY THERMAL ANALYSIS
 Intermetallic phase formation temperatures and exothermal energies.
 Oxidation temperature and oxidation energy.
 Exothermal energy of polymer cure (as in epoxy adhesives), allows determination of the degree
and rate of cure.
 Determine the melting behavior of complex organic materials, both temperatures and enthalpies of
melting can be used to determine purity of a material.
3. PROCESS / TECHNICAL APPROACH
With advancing technology, good design, manufacturing and quality control are major concern for
Aerospace industry. A check for this quality comes in the form of defects or damages. These defects
are prone to occur either during manufacture or during service operations. In case of composites,
manufacturing defects can be during resin injection in fabrication, lay-up of “prepregs”
(preimpregnated unidirectional fibers or woven fabrics with a resin matrix to produce a uniform lamina
structure), or during autoclave processing of the laminates. Commonly occurring defects are out-ofround holes, delamination due to deposition of foreign materials; debond due to moisture absorption
and thus lack of binding between fiber and matrix, voids or porosity.
The literature has standards readily available for the analysis of the effects of delamination defects.
Aerosopace Composites Division of HAL practices the procedure of analyzing the effect of
delamination on mechanical properties of components using these standards. Similar set of standards
which relate the effect of resin rich, voids/porosity on the attenuation coefficients characteristics or
mechanical strength reduction of the composite component is not available. This set of standards
should relate the % of resin content / porosity in the component to the corresponding decrease in CT
Number (HU) through Computed Tomography (CT) testing. Also it should give a relation
between %porosity and the corresponding reduction in mechanical strength. This paper thus deals with
layer waviness, resin rich, voids/porosity defects and aims at helping the Quality Assurance, NDT
group in terms of defect analysis.
3.1 Porosity and voids:
Porosity is termed as the presence of a large number of microscopic air voids [17,18], typically
in the range of 10μm. It may be caused by volatiles and entrained gases (air and water vapor) inside the
resin/adhesive. While porosity can be considered a normal material property, it serves as a depository
for diffused moisture. Similar to porosity at the composite interface as shown in Fig.3.1.1& 3.1.2, this
flaw can occur inside the composite material. Initiating factors include volatiles given off during resin
cure. In certain areas, when these micro voids combine together, they form a large void.
Fig .3.1.1 CT-image shows Porosity noticed between layers
Fig .3.1.2 Free from Porosity (CT cross-sectional view)
3.2 Autoclave processing:
The autoclave as shown in the Fig.3.2.1 is a device that can generate a controlled pressure and
temperature environment. While several autoclave types are available, all consist primarily of three
units: a pressure vessel, a heating/cooling system, and a control unit.
When prepreg based composites are processed using autoclaves, they undergo several steps
prior to the actual autoclaving. The prepreg materials are stacked or laid-up to yield a laminate of the
required dimensions. This is done either manually or by lay-up machines which place the individual
plies of prepreg directly onto the treated tool surface.
The entire lay-up including prepreg plies, release film, and breather cloth is then covered with a
vacuum bag. During the lay-up itself, periodical debulking is done using this vacuum bag, to remove
the trapped air and for proper consolidation. After lay-up of all required prepreg layers, final vacuum
bagging is done before. The entire vacuum bag/tool assembly is connected to vacuum lines prior to
loading in the autoclave so that the laminates retain shape and are isolated from ambient moisture
while they are waiting for autoclaving.
200
7
175
6
5 bars
150
5
125
1350C
1 hr
Temp ( oC)
30C/min
100
4
Pressure
(bars)
3
45min
75
2
50
1
20C/min
25
0
50
100
150
200
250
0
Time ( mins)
Fig 3.2.1. Autoclave installed at ACD-HAL
Fig 3.2.2 Cure cycle for carbon/epoxy 135o cured system
Vacuum bagging has significant effects on the quality of the part produced. The bleeder used in the
sequence acts to absorb excess resin leaving the prepreg during autoclaving. This is very important in
reducing the void content in the cured laminate. The key processing variables in autoclaving are time,
temperature, heat-up rate, pressure, vacuum, and cool-down rate. Proper control of these parameters,
that is, sufficient high pressure (greater than Pmin), appropriate curing temperature based on the prepreg
characteristics, and good functioning of vacuum helps in suppressing void formation to the greatest
extent. Thus to know these parameters, a cure cycle is developed for each system and processing is
done based on that.
The cure cycle as shown in Fig.3.2.2. for carbon/epoxy 135o cured is used in manufacturing
Advanced Light Helicopter (ALH) components. The Tg is around 150 0C; no post curing is required
because the Tg and service temperatures are low.
3.3 MRB SPECIMEN MANUFACTURING:
Main Rotor Blade (MRB) is fabricated by closed mould technology utilizing several non-metallic
materials through bonding process. Where-in layup is carried out on top and bottom moulds and foam
core is positioned with other details in bottom mould. Then Top mould is positioned on the bottom
mould and temperature and pressure is applied by Heating and pressure system integrated with the
mould and layup is cured at 135 deg C.
There are following stages in manufacture of raw MRB Blade manufacturing:
a) Skin layup on top and bottom moulds
b) Positioning of pre-compacted Spar pack 1 & 2 (Autoclave cured detail parts) on top and
bottom moulds
c) Core preparation
d) Positioning of the core on the bottom mould
e) Collar core & tip core preparation and positioning
f) Nose block positioning and suiting
g) Mould closing
h) Curing operation
i) De-moulding of fatigue specimen
3.4 QUANTIFICATION OF FIBER COMPOSITE STRUCTURES
By the representation of the material cross-sections by the attenuation coefficients the material crosssection is available as a data record. So the material qualities become quantifiable. The medical
scanners do not use directly the attenuation coefficients but the so-called CT-number. This number was
established by Hounsfield. The Hounsfield scale standardises the reconstructed attenuation coefficients
(µ) to the linear attenuation coefficient of water at a photon energy of 73 keV.
Therefore water has by definition the CT-number zero. Cured epoxy resin is <50 HU& Air has -1000
HU, GFRP at 1200 up to 1400 HU and CFRP approx. 200 up to 500 HU. Computed Tomography
mainly used for ensuring high reliability processes during manufacturing and it plays major role in
Damage/defect detection, Structural Health Monitoring [16] and Material characterization.
4. DEFECT DETECTION
The normalization of the attenuation coefficients by the CT-number has the advantage, that materials
of fiber composites become comparable. The CT-numbers are represented at the screen as grey value
distribution. One grey value has always the same CT-number, independent of the maximum
attenuation differences of the component. Therefore visually and quantitatively comparable CT-images
are obtained. This is a quite essential prerequisite to analyze the CT-images automatically by image
processing analysis. In the Computed Tomography [3] will be distinguished between the spatial
(geometrical) and the density (contrast) resolution. In this paper the practical meaning of the resolution
for the CT-check [19] on fiber composites is presented.
Variation in compaction, less fibre content [2, 4] and inadequate pressure during manufacturing.
Optimum range of fiber volume ratio between 58-60% was not maintained. Due to variation in
compaction and fiber volume ratio, delaminations /dark line indication observed in few fatigue
specimens. Various composites defects are mentioned below:
4.1 DETAILS OF DEFECTS IN MRB SPECIMENS
Following types of defects are noticed:
Resin Rich: Localized area filled with resin and lacking reinforcing material. An apparent
accumulation of excess resin in a small, localized section visible on cut edges of molded surfaces, or
internal to the structure and nonvisible.
Delamination: Physical separation or loss of bond between laminate plies.
Porosity: A condition of trapped pockets of air, gas, or vacuum within a solid material. Porosity is an
aggregation of micro voids.
Voids: Air or gas that has been trapped and cured into a laminate usually within the matrix.Voids are
essentially incapable of transmitting structural stresses or nonradiative energy fields.
Wrinkle: A surface imperfection in laminated plastics that has the appearance of a crease or fold in
one or more outer sheets of the fabric or other base, which has been pressed in.
Inclusion (FOD): - Any undesirable materials, which are inadvertently left in the bonding area of a
composite structure.
CT Images with various composite defects:
Fig.4.1 : waviness noticed at surface level about 0.5mm depth Fig .4.2 waviness noticed at mid-plane about 1.0mm depth
Fig .4.3 waviness noticed at spar pack area about 2.5-3mm depth
Fig .4.5 waviness noticed at Mid-plane area about 1.5mm depth
Fig .4.4 waviness noticed about 4.0mm depth
Fig .4.6 Free from waviness
Fig .4.7 Resin rich observed between layers
Fig .4.8 Resin rich observed between layers at mid-plane
Fig .4.9 waviness noticed at mid-plane about 2.5mm depth
Fig .4.10 waviness noticed about 1.5-2.0mm depth
5. MATERIAL CHARACTERISATION
Material Characterization study can be experimented with the help of using Computed Tomography to evaluate the
internal structure and properties of a material. Characterization can take the form of evaluating and correlating CT Number
with process key compaction parameters of Fiber volume ratio, Temperature (on set temperature) and resin content, void
content & performance of specimen (Maximum hours flown). Visual observation like resin starvation, discoloration
analysis techniques are used simply to magnify the specimen, to visualize its internal structure and to gain knowledge to the
distribution of compaction level within the specimen and their interactions. The figure below 5.1 shows DSC curves of a
two-part room-temperature cured, low-shrinkage epoxy, with different resin to hardener ratios. The DSC Vs CT number
curve shows the result for epoxy resin onset temperature to Density value (CTn) at that event. i.e pre-compaction stage
condition. The endothermic onset temperature of the prescribed mixture is 124-131°C and that of the CT number value is
lesser at more than 129°C and material also become more soft. CT number value is higher once On set temperature is less
than 124 deg C and material also become too hard. The total endothermic energies are also different. Therefore, DSC can
be used as a tool for quality control of epoxy mixture ratios.
The higher DSC onset temperature shows that less compaction (under curing) and lower DSC onset temperature [13] shows
that abnormal compaction (over curing) at the time of precompaction stage. In order to achieve good precompaction we
need to optimize the onset temperature limit between 126-129 deg C. This optimized limit will give good compaction as
well as consistency in process also. DSC plot in the figure below shows the endotherm which results when the sample was
heated at a rate of 5°C/min. from 75°C to 150°C in nitrogen gas flowing at a rate of 25ml/min [12].The figure 5.2 shows
correlation curve drawn between CT Number and DSC Onset temperature. Due to poor compaction CT Number reduces
accordingly Onset temperature increases it shows that inversely proportional. If compaction is over cured and it shows that
CT Number is increases ant Onset temperature decreases.
5.1
Relationship between % of compaction level (FVR) & Onset
temperature (deg C)
FIG.5.2 Relationship between CT Number & Onset temperature (deg C)
Due to improper and varying compaction level the following process problems are encounter they are: Low mechanical
strength, void content, less fibre volume ratio and dark line indication. CT Measurements with GFRP and CFRP-fiber
samples with various resin content (% of resin), different depth of layer waviness, void content [2]. The figure 5.7 shown
that there is a linear correlation between CT Number & fiber volume ratio. The figure 5.2, 5.4, 5.5, 5.7- CT Number &
Maximum hours flown is inversely proportional to % of Resin content. The figure 5.3, 5.6 -Depth of layer waviness [5] is
decreasing and CT Number is increasing and Maximum hours of blades flown without delamination also more.
Fig.5.3 Relationship between Maximum depth of waviness & Maximum
hours flown without delamination
Fig.5.4 Relationship between % of Resin content & maximum hours
flown without delamination
FIG: 5.5. Relationship between % of Resin content & CT number (HU)
FIG: 5.6. Relationship between Maximum depth of waviness & CT
number (HU)
FIG :5.7 Relationship between fiber volume content & CT Number
Fig.5.8 Relationship between CT Number & maximum hours flown
without delamination
The Fiber volume ratio is plotted and shows that a better correlation between fiber volume ratio with increase in service
hours (Maximum hours flown) of blades, CT Number[20] and decrease in depth of Layer waviness [5] (superficially) &
dark line indication with respect to proper compaction.
It is found that, % of Resin content inversely proportional to service hours (Maximum hours flown) of blades & CT
Number. Depth of layer waviness is inversely proportional to CT Number & service hours (Maximum hours flown) of
blades and dark line indication (delamination). After improving fiber volume ratio as per the required range of 58-60 %,
Computed Tomography (CT) carried out after test and clearly shows that there is no dark line indication (delamination)
found and superficially layer waviness observed at top skin layers to ensure the better compaction level [8].
SPECIMEN with Resin rich & poor compaction of higher Onset
temperature & low CT number
(FVR: 52-54%)
SPECIMEN with LESS waviness noticed at surface level about 0.3-0.5mm
depth (FVR: 57-59%)
BEFORE & AFTER SERVICE (739:30Hrs flown) with minimum layer
waviness at surface level(0.3mm depth) optimized onset temperature of
126-129 deg C & CT Number of 1000-1200 HU (FVR-58%) –NO
DEFECT NOTICED
AFTER SERVICE (1747:46Hrs flown) with No layer waviness at surface
level - optimized onset temperature of 126-129 deg C & CT Number of
1060-1250 HU (FVR-59.2%) –NO DEFECT NOTICED
AFTER SERVICE (1747:46Hrs flown) with layer waviness of 3.0mm
depth - onset temperature of 122-124 deg C and 130-132 deg C & CT
Number of 820-860HU (FVR:55-56.5%) (DELAMINATION
NOTICED)
AFTER SERVICE (1206.41Hrs flown) with layer waviness of 4.0mm
depth - onset temperature of 122-124 deg C and 130-132 deg C & CT
Number of 790-820HU (FVR:54-56%) (DELAMINATION NOTICED)
AFTER SERVICE (1246:25Hrs flown) with no layer waviness optimized
onset temperature of 126-129 deg C & CT Number of 1060-1210 HU
(FVR-58.5%) –NO delamination observed.
AFTER SERVICE (1654:35Hrs flown) with no layer waviness & resin
rich & CT Number of 1080-1245 HU (FVR-58-59%) – NO delamination
observed.
It is concluded that, proper fiber volume ratio and optimized onset temperature of DSC value increases the service life of
the Blades [10] and reduces delamination / damage during flight. Optimization of dynamic systems need this process
control approach through CT & DSC observation and correct the process then and there to improve the part as well as
process quality for better performance of the dynamic systems. By this way we can extend the life of the dynamic systems.
Computed Tomography carried out under various fatigue cycles & service schedule and formulate the data before and after
service conditions, the ability to normalize the data becomes very important to the damage identification process. As it
applies to SHM, data normalization is the process of separating changes in CT Number reading caused by damage from
those caused by varying operational and environmental conditions.
6. CONCLUSION
This study experiments the fiber volume ratio, resin content, dark line indication (delamination) and layer waviness
characteristics by the use of Computed Tomography (CT).Fatigue specimens (blades) of laminated composites, resin
content with varying fiber volume ratio and layer waviness, are subjected to fatigue loading at various service hours.
Computed Tomography is used to observe matrix cracks, performance of blades and overall delamination propagation;
whilst Computed Tomography is exploited to examine cross-sectional views showing detailed internal information about
fiber volume ratio and through-thickness matrix cracks distribution and 3D delamination damage pattern.
Computed Tomography [6] reveals that low fiber volume ratio, resin-rich greatly reduces service life of blades as well as
and increases delamination growth. Specimens with optimum fiber volume ratio and appropriate CT number (density)
proper compaction level are more capable of impeding dark line indication (delamination) growth by effectively bridging
delamination initiation and arrest matrix crack propagation. It is also found that low fiber volume ratio act as crack
initiation sites, due to the presence of resin-rich pockets, layer waviness thus resulting in densely resin pockets having more
matrix cracks upon fatigue loading.
CT continuously used in HAL for evaluating blades life cycle from the rotor blade development, series production and
maintenance and enhancing the dynamic component life based on the CT results. During the development of the helicopters
programme, quality and the life time of the rotor blades became verified decisively by use of the CT. In the series
production the quality of the rotor blades of the helicopter programme is checked by means of CT and improved step by
step based on the fatigue results with optimized fiber volume ratio. If CT is used consistently throughout the development,
production and maintenance robust CT NDE methodology can help increasing the quality and fatigue behaviour of
dynamically loaded helicopter components made of carbon [7] composite materials and monitor integral structural of the
component. Computed Tomography(CT) imaging for considerable use in engineering development, problem solving, and
production of composite components (particularly carbon and glass composite components) & the ability to track the
integrity of individual components by greatly simplifying the process analysis.
7. ACKNOWLEDGEMENT
This work was carried out at Aerospace Composites Division, HAL. We, the authors would like to thank The General
Manager (ACD) Division, HAL providing all help, encouragement and permission to publish this technical paper in 8th
International symposium on NDT in Aerospace, IISc-Bangalore, November 2016.The authors also like to thank all
Composite Shop, Quality Methods, Design department those who have involved in this work directly and indirectly.
8. REFERENCES:
[1] Non-Destructive Evaluation and Quality Control was published in 2008 as Volume 17 of the 9th Edition ASM Metals
Handbook. & ASM HAND BOOK VOL.21 - COMPOSITES
[2] Journal of composite materials, vol26, 10-12/1992,”Void effects on ILSS of UD graphite FR composites”.
[3] Standard practice for Computed Tomography (CT) Imaging Designation: E 1570 & E-1441 American Society for
Testing and Materials.
[4] Composites Materials: Testing and design (3rd conference ASTM), “Methods of fiber and void measurement in
graphite/epoxy composites”.
[5] Journal of Composites: Part B 42(2011) 62-70 The effect of inclusion waviness and waviness distribution on elastic
properties of fiber-reinforced composites”.
[6] Schell JSU, Renggli M, Lenthe GH, Muller R, Ermanni P. Micro-Computed Tomography determination of glass fibre
reinforced polymer meso-structure. Composite Science Technology 2006; 66:2016-22.
[7] Abrate S. Impact on laminated composites: recent advance. Appl Mech Rev 1994; 44(4):517-44.
[8] Davies GAO,Zhang X. Impact damage prediction in carbon composite structures. Int J Impact Engg 1995:16(1):149-70.
[9] Bayraktar E, Antolonovich S,Bathias C. Multi-scale study of fatigue behavior of composite materials by X-Rays
Computed tomography. Int J Fatigue 2006;28:1322-33.
[10] Symons DD, Davis G, Fatigue testing of impact-damaged T300/914 Carbon fiber reinforced plastic. Compos Sci
Techno 2000; 60:379-89.
[11] ASTM E1356 – Standard Test Method for Glass Transition Temperatures by DSC.
[12] ASTM E793 – Standard Test Method for Enthalpies of Fusion and Crystallization by DSC.
[13] ASTM E 473-08 Standard Terminology Relating to Thermal analysis and Rheology
[14] ASTM E 2161-08 Standard Terminology Relating to Performance Validation in Thermal analysis.
[15] C.Boller,F.K.Chang and Y.Fujino.2009.Encyclopedia of Structural Health Monitoring (SHM),John Wiley & Sons
Limited.
[16] D.Balageas,C.Fritzen and A.Gnemes.2006. Structural Health Monitoring (SHM), London Newport Beach.
[17] Composites Materials: Testing and design (3rd conference ASTM), “Methods of fiber and void measurement in
graphite/epoxy composites”.
[18] Journal of composite materials, vol27,4-6/1987, “Effects of cure pressure on resin flow, voids and mechanical
properties”.
[19] K.D. Friddell, A.R. Lowrey, and B.M. Lemprier, Application of Medical Computed Tomography (CT) Scanners to
Advanced Aerospace Composites, Vol 4, Review of Progress in Quantitative Nondestructive Evaluation, D.O. Thompson
and D.E. Chimenti, Ed., Plenum Press, New York, 1985
[20] R.H. Bossi and G.E. Georgeson, Composite Structure Development Decisions Using X-Ray CT
Measurements, Mater. Eval., 1995
Non Destructive Evaluation of In-situ Skirt Joint of Filament Wound
Composite Pressure Vessel
Sanjeev Kumar and P J Thakar
Advanced Systems Laboratory
Directorate of CPDC,
Hyderabad, India
Contact: sanjeev.hyd@gmail.com
Carbon Epoxy composite Pressure vessels are designed to withstand high internal pressure and
structural load during the flight. In-situ skirts are fabricated along with composite pressure vessel in
single cure cycle. Skirt & Skirt Joint need to be critically evaluated as they transfer all the loads to
subsequent stages and not structurally load tested in all cases. This paper brings out problems and
challenges in non-destructive evaluation of In-situ Skirt & skirt joint of filament wound composite
pressure vessels.
Feature Guided Waves: New Paradigm for Inspection and Health Monitoring of
Aerospace Composite Structures and Components
Prabhu Rajagopal
Centre for Non-destructive Evaluation & Department of Mechanical Engineering
IIT Madras, Chennai, Tamil Nadu, India
Contact: prajagopal@iitm.ac.in
Composites find increasing application for light-weighting in various industries such as automotive
and aerospace and fast inspection and on-line monitoring of such structures is of much industrial
interest. Ultrasonic guided waves have long been considered attractive for rapid long-range health
assessment of extended structures from a single sensor location. However, aerospace composite
structures, with complex cross-sectional features such as bends, cross-section changes, ribs and
stiffeners, are difficult to inspect and monitor using such strategies. This paper proposes a novel and
counter-intuitive approach to health assessment of such complex structures, by making use of waves
confined to these very features. Recent research, a lot of it at our research group, has revealed a series
of new guided wave modes confined to features such as bends, cross-section changes and joints in
plates and tubes. Combining knowledge of these modes with sensor approaches such as fiber Bragg
gratings (FBGs) can offer new and powerful ways of monitoring the health of complex composite
structures. The paper will provide an overview of key developments in feature-guided waves over the
last few years, and also outline some strategies for NDT and SHM of such structures using example
cases.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Multiphysics Simulation of Guided Wave Propagation under Load
Condition
Lei QIU1,2, Ramanan SRIDARAN VENKAT2, Christian BOLLER2, Shenfang YUAN1
1
Research Center of Structural Health Monitoring and Prognosis, State Key Lab of Mechanics and Control of
Mechanical Structures, Nanjing University of Aeronautics and Astronautics; Nanjing, China
E-mail: lei.qiu@nuaa.edu.cn, ysf@nuaa.edu.cn
2 Chair of Non-Destructive Testing and Quality Assurance (LZfPQ), Saarland University; Saarbrücken,
Germany; ramanan.sridaran@uni-saarland.de, c.boller@mx.uni-saarland.de
Abstract
A multiphysics simulation method of Guided Wave (GW) propagation under load condition is proposed. With
this method, two key mechanisms of load influence on GW propagation are considered and coupled with each
other. The first key mechanism is the acoustoelastic effect which is the main reason of GW phase change. The
second key mechanism is the load influence on piezoelectric materials, which results in a change of the GW
amplitude. Based on COMSOL multiphysics, a finite element model of GW propagation on an aluminium plate
under load condition has been established. The simulation model includes two physical phenomena to be
considered represented by simulation modules. The first module is called solid mechanics, which is used to
simulate the acoustoelastic effect being combined with the hyperelastic material property. The second module is
called electrostatics, which considers the simulation of the piezoelectric effect for GW excitation and response.
To simulate the load influence on piezoelectric materials, a non-linear numerical model of the relationship
between load and piezoelectric constant d31 is built. The simulation results under uniaxial load are obtained and
they are compared with the data obtained from an experiment of load influence on GW. It shows that the
variations of phase and amplitude of GW obtained from the simulation match the experimental results well.
Keywords: Structural health monitoring, time-varying condition, guided wave, multiphysics simulation,
acoustoelastic effect
1. Introduction
Real aircraft structures serve under uncertain time-varying conditions such as environmental
conditions, load conditions and structural boundary conditions etc. Almost all the damage
monitoring features can be directly affected by the time-varying conditions, which leads to
low damage monitoring reliability. Among Structural Health Monitoring (SHM) methods,
Guided Wave (GW) and piezoelectric sensor based method is a promising one because it is a
regional monitoring method and is sensitive to small damage [1]. To deal with the timevarying problem [2], several methods [3-7] such as the environmental compensation method,
baseline free method, data normalization method and mixture probability method etc. have
been proposed but limitations remain. Thus, for real applications, the problem of reliable
damage monitoring under time-varying conditions must be fully studied.
Considering that the time-varying conditions are often complicated, and the
corresponding experiments are highly costly and time consuming, GW simulation under timevarying conditions is an effective way to study the time-varying problem. Based on the
simulation, the GW propagation on complex structure under complicated time-varying
conditions can be studied easily and the simulation data can be also used to validate the
methods which are aimed to deal with the time-varying problem.
Although the simulation of GW propagation has been widely studied [8-11], the GW
simulation under time-varying conditions is still rarely reported [12], especially for a
simulation method which fully considers the influence of time-varying conditions recorded by
piezoelectric sensors adhered on a structure and this under close to real conditions.
1
Among a lot of time-varying factors, the changing load condition is a main factor,
which can introduce large variations to the phase and amplitude of a GW signal recorded. In
this paper, an efficient method of multiphysics simulation of GW propagation under load
condition is proposed. The two key mechanisms of load influence on GW are the
acoustoelatstic effect which is the main reason of GW phase change and the load induced
influence on piezoelectric materials which results in a change of the GW amplitude [12].
Thus, the two mechanisms are considered and coupled together in this method. Based on
COMSOL multiphysics, a finite element model of GW propagation on an aluminum plate
under load condition is established. An experiment of load influence on GW is performed to
validate the proposed multiphysics simulation method.
2. The Experiment of Load Influence on Guided Wave
2.1 Experimental Setup
The experimental system is shown in Figure 1. The structure is 2024 aluminium alloy and its
dimension is 400mm×200mm×2mm (length×width×thickness). Two piezoelectric sensors
(PZT-5A) are placed on the structure which are numbered as PZT 1 and PZT 2. PZT 1 is used
to excite GW and PZT 2 is used to be a GW receiver. The distance between PZT 1 and PZT 2
is 200 mm. The structure is fixed on a static tensile machine which is used to apply a uniaxial
tension load to the structure. The eleven levels of load (from 0 MPa to 100 MPa with 10 MPa
interval) are applied to the structure. For each load level, the GW excitation and response of
the two sensors are performed and controlled by a GW based SHM system which is
developed by the authors [13]. The excitation signal is a five-cycle sine burst modulated by
Hanning window. The central frequency and amplitude of the excitation signal are 200 kHz
and ±70V respectively. The sampling rate of GW signal is 10 Msamples/s.
Fig.1. The experiment system of load influence on guided wave.
2.2 Experimental Results
The acquired GW signals at all load levels are de-noised by a method [14] based on complex
continuous Shannon wavelet transform first. The de-noised GW signals are displaced in
Figure 2 (a). The amplitude and phase variation of S0 mode of the GW signals are enlarged to
be a better observation as well. There should be noted that in this paper, only S0 mode is
considered. The reason will be explained in section 4.
Figure 2 (b) gives the quantitative variation of the amplitude and phase. It can be
noted that the phase velocity decreases linearly and the amplitude increases non-linearly
accompanying with the increasing of the load. For measuring the change in phase velocity,
2
equation (1) is used [14]. The slope of phase velocity change is -0.576m·s-1·MPa-1. For
measuring the change in amplitude, equation (2) is used. The change in amplitude is fitted by
a mixed exponential equation shown in equation (3) and the parameters are obtained as a =
0.1328, b = 0.0022, c = -0.1328 and d = -0.0476.
0MPa
10MPa
20MPa
30MPa
40MPa
50MPa
60MPa
70MPa
80MPa
90MPa
100MPa
Relative amplitude
5
4
3
2
1
1.8
1.85
1.9
1.95
2
2.05
Time(s)
2.1
2.15
2.2
2.25
2.3
-5
x 10
15
Phase velocity changes(m/s)
0
6
Experimental result
Linear fit
-10
-20
-30
-40
-50
Relative amplitude
10
-60
5
0
20
0
40
60
External load(MPa)
80
100
20
-5
-15
0
1
2
3
4
Time(s)
5
6
7
0MPa
10MPa
20MPa
30MPa
40MPa
50MPa
60MPa
70MPa
80MPa
90MPa
100MPa
0.1
Relative amplitude
8
-5
x 10
0.05
0
-0.05
-0.1
2.1
2.11
2.12
2.13
2.14
Time(s)
2.15
2.16
2.17
Amplitude changes(%)
-10
15
10
5
Experimental result
Exponential fit
0
2.18
0
20
-5
x 10
(a) GW signals under all load levels
40
60
External load(MPa)
80
100
(b) Phase and amplitude variation of GW signals
Fig.2. The experimental results of load influence on GW signals.
Vp  
Amp 
VP2
t
l
Amplevel  Amp0
100%
Amp0
Amp fit  a  ebload  c  ed load
(1)
(2)
(3)
Where, Vp is the phase velocity of S0 mode and l is the GW propagation distance. Δt is the
time shit of constant phase of GW signal. Amplevel is the GW amplitude at the corresponding
load level and Amp0 is the amplitude at level 0. load is the load induced stress (Unit: MPa).
3. The Mechanism of Load Influence on Guided Wave
3.1 Acoustoelastic Effect
Acoustoelastic effect refers to the stress-dependence of acoustic bulk wave velocity in solid
media [15]. When a structure is stress-free, the longitudinal wave velocity and transverse
wave velocity of a non-dispersion elastic wave propagating in a solid structure can be
expressed as equation (4) and (5) by using the second order Lame constants λ and μ. When the
structure is in a stressed condition because of external load, the above two velocities can be
expressed as equation (6) and (7) by combining with the third order Murnaghan constants l, m
and n. In these two equations, T denotes the external load and K is the bulk modulus.
3
0CL2    2 
0CL2    2
(4)
0CT2  
(5)
T
3K



 2l      4m  4  10 


0CL2   

T 
n
m
   2 
3K 
4

(6)
(7)
It can be known that the two velocities become to be stress-independence because of
the external load, and the relationship between the velocity change and the stress is linear.
Although GW is a dispersion and multi-mode wave, the abovementioned equation can be also
applicable to describe the acoustoelastic effect of GW because the GW is composed by the
two wave components of longitudinal wave and transverse wave [16]. Thus, it is clear that if
the load influence on GW phase change needs to be simulated, the acoustoelastic effect
should be considered combined with the third order elastic constants. Specifically, the third
order Murnaghan constants l, m and n should be included in the simulation model.
3.2 Load Influence on Piezoelectric Material
Piezoelectric material works based on the piezoelectric effect which is controlled by the
piezoelectric constitutive equation. However, according to the studies of Hall [17] and Lynch
[18] et al., the linear behaviour of piezoelectric constitutive relationships is generally confined
to relatively low levels of applied electric field and stress, under which the dielectric, elastic
and piezoelectric constant keep unchanged. When the piezoelectric material is loaded, the
external load can change the polarization state of the piezoelectric material so as to change the
piezoelectric constitutive relationships. According to the experimental results given by Lynch
[18], the change of polarization state happens when the stress of piezoelectric material PZT5A is only 5MPa or larger. Besides, the experimental results also show that the dielectric,
elastic constant and piezoelectric constant d33 are non-linear stress-dependence. Kang et al.
[19] gave that the compression in the longitudinal direction of PZT-5A causes the non-linear
decreasing of the piezoelectric constant d31 but does not affect the elastic constant.
Thus based on the abovementioned research, it is clear that if the load influence on
GW amplitude change needs to be simulated, the load influence on piezoelectric materials
should be considered. However, the main obstacle for the simulation is that the numerical
models which can be used to describe load influence on different piezoelectric parameters in
the simulation model are unknown.
Actually speaking, the GW signal waveform get from simulation model should match
the real GW signal waveform get from real world. However, for the research on time-varying
problem, it is enough if the variation trend of the simulated GW features can match that of the
real GW features. Considering this point, some simplications are made as follows. (1) The
load influence on the piezoelectric constant d31 is only considered because GW excitation and
response by adhered piezoelectric sensors are mainly controlled by this parameter. (2)
Considering that the load influence on piezoelectric material is the main factor that changes
the GW amplitude, a numerical model is established to be equation (8) based on the amplitude
change obtained from the experiment results in section 2.
d31  load   d31  0   d31  0    a  ebload  c  ed load 
(8)
Where, a = 0.1328, b = 0.0022, c = -0.1328 and d = -0.0476, and load is the actual load
induced stress (Unit: MPa).
4
4. The Multiphysics Simulation Model
4.1 The Simulation Physics
The physics of simulating GW under load condition are simplified shown in Figure 3. An
aluminum plate is loaded at one end and is fixed on the other end. Two piezoelectric sensors
(PZT 1 and PZT 2) are placed on the plate surface and are coupled with the aluminum plate
directly. The adhesive layer is ignored because the load introduce little influence on it [12].
Thus, the acoustoelastic effect, piezoelectric effect and the load influence on piezoelectric
sensors should be integrated into one simulation model. In addition, the load may change with
time. However, in terms of short duration of GW, the load can be considered as static load.
Thus, the process of GW excitation and response needs to be simulated under static load
condition in the simulation model. In this paper, the multiphysics simulation is realized by
using COMSOL Multiphysics 5.0. Considering the above discussed physics, the physic
module called Piezoelectric Devices is adopted. It is a kind of multiphysics model shown in
Figure 4 which includes the physic modules of Solid Mechanics and Electrostatics.
Fig.3. The physics diagram of GW propagation under load condition.
Fig.4. The simulation physics in COMSOL of GW propagation under load condition
The Solid Mechanics is used to simulate mechanics feature of the aluminum plate and
the piezoelectric sensors. The Electrostatics is used to simulate the electric feature of the
piezoelectric sensors. The two physics are coupled by the Multiphyiscs-Piezoelectric Effect
and they are described as follows,
1) In Solid Mechanics, Fixed Constraint and Boundary Load are used to simulate the
fixed end and the external load of the aluminum plate respectively. The external load is
applied to the plate from 0 MPa to 100 MPa with 20 MPa interval (Six load levels totally).
2) Hyperelastic material is used to realize the simulation of acoustoelastic effect. In
the material model of Hyperelastic material, the Murnaghan model is adopted and the values
of the third order Murnaghan constants will be given out in section 4.2.
5
3) Piezoelectric material combinded with Electrostatics is used to realize the
simulation of the piezoelectric sensors. The property of Mechanical Damping is adopted, in
which the Rayleigh damping is used and the parameters are set to be α = 0 and β = 2.2×10-8.
The material property of the piezoelectric sensors will be given out in section 4.2.
4) Low-Reflecting Boundary is used to reduce the GW boundary reflection. The
damping type is set to be ‘P and S waves’.
5) In Electrostatics, Ground is the electric ground of the piezoelectric sensors and it is
set to their lower surface.
6) Electric Potential 1 and Electric Potential 2 are a zero potential and a voltage
waveform of the GW excitation respectively. They are set to the upper surface of PZT 1 but
they are mutually exclusive. This point will be explained in section 4.3.
4.2 The Simulation model
The simulation model is decribed as follows based on the modeling process of COMSOL.
1) Geometry: 3D model is used in this paper including the aluminum plate and the
pieozelectric sensors. The dimension of the aluminum plate is 400mm×200mm×2mm
(length×width×thickness). For the piezoelectric sensor, the diameter is 8mm and the thickness
is 0.48mm. These diemnsions are the same with those of experiment.
2) Definitions: The definitions include two parts. The first part is the GW excitation
waveform which is expressed as equation (9). The corresponding parameters are set to be A =
35 (Unit: V), f = 200 (Unit: kHz) and N = 5. The second part is the GW observation probe. It
is set to the upper surface of PZT 2. The type of the probe is voltage and the average voltage
of the whole surface will be output.
E1  A  1  cos  2 ft N   sin  2 ft   t   N f 
(9)
3) Material: The material of the aluminum plate is shown in table 1 which contains the
third order Murnaghan constants for the simualtion of acoustoelastic effect. The material of
the piezoelectric sensors is shown in table 2. It is PZT-5A and the d31 parameter is set to be
eqaution (9) for the simulation of load influence on the pieozelectric sensors.
Table 1. Material property of the aluminum plate (2024 aluminum alloy).
Parameter
Density (kg/m3)
Lame constant (GPa)
Murnaghan constant (GPa)
Value
2700
26
51
-250
-330
-350
ρ
μ
λ
l
m
n
Table 2. Material property of piezoelectric sensor (PZT-5A).
Parameter
Relative permittivity
Piezoelectric constant
(×10-10 C/N)
Compliance coefficient
(×10-12 m2/N)
Value
ε11
ε33
d31
d33
d15
sE11
sE12
sE13
sE33
sE55
sE66
1730
1700
-1.71-1.71×(a·eb·load + c·ed·load )
3.74
5.84
16.4
-5.74
-7.22
18.8
4.75
4.43
6
4) Physics: The physics have been given out in section 4.1.
5) Mesh: The mesh type is Free Tetrahedral. For the plate, the mesh size depends on
GW wavelength. To the GW of 200 kHz on the plate of 2mm, the wavelength of S0 mode is
nearly 27mm. According to some research on GW simulation [9-11], the largest mesh length
is recommended to be smaller than 1/6 of the wavelength. Thus, the largest mesh and smallest
mesh size are set to be 3mm and 2mm respectively. The wavelength of A0 mode is only
8.6mm and the mesh size should be less than 1.4mm. It will lead to a huge amount of
computation and a normal computer with 16 GB RAM cannot support such computation. That
is reason why this paper only considers S0 mode. For the piezoelectric sensors, the largest and
smallest mesh size are set to be 2 mm. The complete mesh consists of 83373 domain elements,
55246 boundary elements and 996 edge elements. There are 501876 degrees of freedom.
4.3 The solver of the Simulation Model
For solving the multiphysics simulation model given above, two study steps are adopted to
construct the Study of the simulation model, as shown in Figure 5. The first step is Stationary
which is used to perform the static load. After that, Time Dependent is performed to simulate
the process of GW excitation-propagation-response under the static load condition. The
results of Step 1 are used to be the initial values of Step 2.
Fig.5. The solver settings of the Multiphysics simulation model
For each step, the physics settings are also shown in Figure 5. In Step 1, the Electric
Potential 1 of zero potential is enabled to disable the PZT 1 but the Electric Potential 2 of GW
excitation is disabled. This is because of two reasons. First, if the PZT 1 is not disabled in
Step 1, the static load would introduce stress to PZT 1 and makes it generate a DC bias
voltage. In Step 2, the DC bias voltage would act as a step excitation to introduce wideband
GW propgating on the aluminium plate so as to lead a false GW response signal output from
PZT 2. The second reason is that Step 1 is stationary study, there is no need to excite GW. In
Step 2, Electric Potential 1 is disabled and Electric Potential 2 is enabled for GW excitation.
The solver run one time for each load level mentioned in section 4.1. Thus, the GW
signals under all six load levels can be obtained. The time step is set to be 1×10-7s in Step 2.
7
5. Simulation Results of Guided Wave Propagation under Load Condition
In the simulation model, there is no charge amplifier but the experimental system has one. So
under this situation, the simulation signal under load 0 MPa and the experimental signal under
0 MPa are compared with each other as shown in Figure 6 (a). For a better comparsion, the
two signals, which are normalized based on the amplitude of S0 mode, are given in Figure 6
(b). As it can be seen that the S0 mode of the two signals are matched well but there is a large
error in A0 mode. This is because of the large mesh size.
The simulation GW signals under all load conditions are displayed in Figure 7 (a). The
amplitude and phase variation of S0 mode of the GW signals are enlarged to a better
observation as well. Figure 7 (b) gives the quantitative variation of the amplitude and phase.
Simulation GW response signal
0.4
1.5
0
-0.2
-0.4
0
100
200
300
400
500
600
700
Signal dots
Experiment GW response signal
800
900
1000
2
Voltage(V)
1
Normalized amplitude
Voltage(V)
2
0.2
1
0.5
0
-0.5
-1
0
-2
Experiment GW response signal
Simulation GW response signal
-1.5
-1
0
100
200
300
400
500
600
Signal dots
700
800
900
-2
300
1000
(a) Original simulation and experimental signals
400
500
600
700
Signal dots
800
900
1000
(b) Normalized comparison (S0 mode) of the two signals
Fig.6. The GW signals comparsion between simulation and experiment.
0.26
0.24
0.22
0.2
5.05
5.1
5.15
Time(s)
5.2
5.25
-5
x 10
Normalized amplitude
0.4
0
-0.2
-0.4
0
0.1
0.2
Normalized amplitude
-40
0
20
40
60
External load(MPa)
80
100
20
0.3
0.4
-3
0
-30
-60
x 10
5
-20
-50
0MPa
20MPa
40MPa
60MPa
80MPa
100MPa
0.2
Simulation result
Curve fit
-10
0.5
Time(s)
0.6
0.7
0.8
0.9
1
-4
x 10
0MPa
20MPa
40MPa
60MPa
80MPa
100MPa
Amplitude changes(%)
Normalized amplitude
0MPa
20MPa
40MPa
60MPa
80MPa
100MPa
Phase velocity changes(m/s)
0
0.28
15
10
5
Simulation result
Exponential fit
-5
5.25
5.26
5.27
Time(s)
5.28
5.29
0
5.3
-5
x 10
(a) Simulation GW signals under all load levels
0
20
40
60
External load(MPa)
80
100
(b) Phase and amplitude variation of signal GW signals
Fig.7. The simulation results of load influence on GW signals.
It can be noted from Figure 7 that the phase velocity decreases linearly and the
amplitude increases non-linearly accompanying with the increasing of the load. The slope of
phase velocity change is -0.467m·s-1·MPa-1. It is a little lower than the experimental result of
8
-0.576m·s-1·MPa-1. This error may be due to the material difference of the aluminium plate
between experiment and simulation. Based on the numerical model shown in equation (9), the
simulation of load influence on piezoelectric sensor is realized and the simulation results
match the experimental results well.
6. Conclusion
This paper proposes a simple but efficient method for multiphysics simulation of GW
propagation under load condition based on COMSOL Multiphysics. The acoustoelastic effect
and the load influence on piezoelectric sensor are integrated into one multiphysics simulation
model. The simuation of acoustoelastic effect is realized by using Hyperelastic material
model which contains the third order Murnaghan constants. The simuation of load influence
on piezoelectric sensor is realized by building a d31 numerical model of stress-dependence
based on the experimental data. The model solver is constructed by combing the stationary
analysis and time-dependent analysis together. The whole GW propagation simulation under
load condition is fullfiled in one software platform and in one model solver. The simulation
results under uniaxial load from 0 MPa to 100 MPa are obtained and they are compared with
experimental data from the two aspects of amplitude change and phase change. It shows that
the results obtained from the simulation match the experimental results well which indicates
the correctness of the proposed method.
According to the simulation results, a preliminary conclusion can be made that a
numerical model of load influence on piezoelectric material, which is constructed by a
calibration experiment on simple structure and load condtion, can be applied to more complex
structure and load condition because the load influence on piezoelectric sensor is structural
independent and is only stress-dependence.
In the near future, the multiphysics simulation of GW propagation under changing
temperature condition will be studied. It will be combined with the method proposed by this
paper to achieve a comprehensive multiphysics simulation of GW propagation under timevarying conditions.
7. Acknowledgements
Lei QIU would like to acknowledge the Alexander von Humboldt Research Foundation for its
support to undertake scientific collaborations in Germany via a Humboldt Research Award.
He also expresses his gratitude to the Chair of Non-Destructive Testing and Quality
Assurance of Saarland University for its hospitality in hosting him.
This work is supported by National Science Fund for Distinguished Young Scholars of
China (Grant No.51225502), Key Program of Natural Science Foundation of China (Grant
No. 51635008), National Natural Science Foundation of China (Grant No. 51575263), Qing
Lan Project and Young Elite Scientist Sponsorship Program by CAST of China.
References
1.
2.
C Boller, F K Chang, and Y Fujino, ‘Encyclopedia of structural health monitoring’,
John Wiley, January 2009.
I Lopez and N Sarigul-Klijn, ‘A review of uncertainty in flight vehicle structural
damage monitoring, diagnosis and control: Challenges and opportunities’, Progress in
Aerospace Sciences, Vol 46, No 7, pp 247-273, October 2010.
9
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
A J Croxford, J Moll, P D Wilcox and J E Michaels, ‘Efficient temperature
compensation strategies for guided wave structural health monitoring’, Ultrasonics, Vol
50, No 4, pp 517-528, April 2010.
H J Lim, H Sohn, C M Yeum, and J M Kim, ‘Reference-free damage detection,
localization, and quantification in composites’, The Journal of the Acoustical Society of
America, Vol 133, No 6, 3838-3845, June 2013.
P B Dao and W J Staszewski, ‘Cointegration approach for temperature effect
compensation in Lamb-wave-based damage detection’, Smart Materials and Structures,
Vol 22, No 9, 095002, June 2013.
C Zhou, M Hong, Z Su, Q Wang and L Cheng, ‘Evaluation of fatigue cracks using
nonlinearities of acousto-ultrasonic waves acquired by an active sensor network’, Smart
Materials and Structures, Vol 22, No 1, 015018, December 2012.
L Qiu, S Yuan, F K Chang, Q Bao and H Mei, ‘On-line updating Gaussian mixture
model for aircraft wing spar damage evaluation under time-varying boundary
condition’, Smart Materials and Structures, Vol 23, No 12, 125001, October 2014.
J H Nienwenhui, J J Neumann, D W Greve and I J Oppenheim, ‘Generation and
detection of guided waves using PZT wafer transducers’, IEEE transactions on
ultrasonics, ferroelectrics, and frequency control, Vol 52, No 11, pp 2103-2111,
December 2005.
C Yang, L Ye, Z Su and M Bannister, ‘Some aspects of numerical simulation for Lamb
wave propagation in composite laminates’, Composite structures, Vol 75, No 1, pp 267275, September 2006.
Y Lu, L Ye, Z Su and C Yang, ‘Quantitative assessment of through-thickness crack size
based on Lamb wave scattering in aluminium plates’, NDT & E International, Vol 41,
No 1, pp 59-68, January 2008.
G Giridhara, V T Rathod, S Naik, D R Mahapatra and S Gopalakrishnan, ‘Rapid
localization of damage using a circular sensor array and Lamb wave based
triangulation’, Mechanical Systems and Signal Processing, Vol 24, No 8, pp 2929-2946,
November 2010.
K Lonkar, ‘Modeling of piezo-induced ultrasonic wave propagation for structural health
monitoring Stanford’, Doctoral dissertation, Stanford University 2013.
L Qiu and S Yuan, ‘On development of a multi-channel PZT array scanning system and
its evaluating application on UAV wing box’, Sensors and Actuators A: physical, Vol
151, No 2, pp 220-230, April 2009.
L Qiu, M Liu, X Qing, and S Yuan, ‘A quantitative multi-damage monitoring method
for large-scale complex composite’, Structural Health Monitoring, Vol 12, No 3, pp
183-196, May 2013.
D S Hughes and J L Kelly, ‘Second-order elastic deformation of solids’, Physical
review, Vol 92, No 5, pp 1145, December 1953.
V Giurgiutiu, ‘Structural Health Monitoring with Piezoelectric Wafer Active Sensors’,
Academic Press: San Diego, CA, USA, June 2014.
D A Hall, ‘Review nonlinearity in piezoelectric ceramics’, Journal of materials
science, Vol 36, No 19, pp 4575-4601, October 2001.
C S Lynch, ‘The effect of uniaxial stress on the electro-mechanical response of 8/65/35
PLZT’, Acta materialia, Vol 44, No 10, pp 4137-4148, October 1996.
L H Kang, D O Lee and J H Han, ‘A measurement method for piezoelectric material
properties under longitudinal compressive stress – a compression test method for thin
piezoelectric materials’, Measurement Science and Technology, Vol 22, No 6, 065701,
April 2011.
10
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Lamb Wave Based Damage Detection using Orthogonal Matching Pursuit
and Artificial Neural Network
Navjeet Kumar, Mira Mitra
Department of Aerospace Engineering, Indian Institute of Technology, Bombay; Mumbai, India
Phone: +91 9769474376
E-mail: navjeet_kumar@iitb.ac.in,navjeet.kumar08@gmail.com, mira@aero.iitb.ac.in
Abstract
This work presents a damage detection technique which involves three steps, namely, data generation, signal
processing or sparse signal approximation, and classification using machine learning algorithm i.e. Artificial
Neural Network (ANN). Lamb-waves, which are elastic waves that travel in traction free thin-walled solid
structures, are considered highly suitable for damage detection. The database of Lamb wave response for
damaged and undamaged structures are generated using a finite element (FE) method. The data generation step
consists of two sub-steps; firstly, the wave propagation response is recorded for a healthy plate. Next, in the
simulation, a defective specimen is modelled by creating an open rectangular crack in healthy specimen. To
reduce the complexity of the Lamb wave response, asymmetric mode is selectively generated and sensed. The
time domain results obtained for defective specimen are compared with the baseline wave signals (healthy
specimen) to determine the presence of damage. In this study, Orthogonal Matching Pursuit (OMP) has been
tested with machine learning algorithm, ANN to automate the process of damage detection. Signal processing
step involves increasing the sparsity of the signal using OMP algorithm with the aim to reduce the number of
nonzero data-points to improve machine learning classification. The OMP algorithm also succeeds in extracting
meaningful pulses from simulated noisy signals. The output wave response database from the signal processing
is passed as input data for ANN. The Feed Forward Neural Network (FFNN) is selected for this study in which
no connections exist between nodes that are not in adjacent layers. 70% of the input database is used for training,
15% for validation and 15% for testing of the ANN. This study demonstrates that the OMP reduces the non-zero
data points approximately to one fourth and the ANN is a robust classifier and can estimate the location of
damage with an error of 3-4 %.
Keywords: Damage Detection, Lamb Wave, OMP, ANN
1. Introduction
Damage can be defined as changes introduced into a system that adversely affect its current or
future performance. Implementation of a damage identification strategy for aerospace, civil
and mechanical engineering infrastructure is referred to as structural health monitoring
(SHM). This process involves the observation of a structure or mechanical system over time
using periodically spaced measurements, the extraction of damage-sensitive features from
these measurements and the statistical analysis of these features to determine the current state
of system health. The schemes available for SHM can be broadly classified as active or
passive depending on whether or not they involve the use of actuators, respectively. Passive
schemes which have been demonstrated with some success are Acoustic emission (AE) and
strain loading monitoring. However, these passive techniques require higher sensor densities
on the structure. Unlike passive methods, active schemes are capable of exciting the structure
and, in a prescribed manner, they can examine it for damage within seconds, where and when
required. Guided lamb wave testing has emerged as a very prominent option among active
schemes. It can offer an effective method to estimate the location, severity and type of
damage, and it is a well-established practice in the non-destructive evaluation and testing
(NDE/NDT) industry. Lamb waves are guided waves which can exist in plate-like thin bodies
with parallel free boundaries. The actuator sensor pair in Guided Wave testing has a large
coverage area, resulting in fewer units distributed over the structure. The present study
concerns with ultrasonic guided wave based damage detection techniques employed for twodimensional metallic plate structure. The complexity of the lamb wave can be reduced via
some signal sparse representation algorithm and the detection can be automated by some
machine learning techniques. Structural damage, defect or crack in structures used in
aerospace, civil and infrastructure applications during service life can result in catastrophic
failure. It is essential to develop health monitoring techniques which would lead to cutting
down the time for which structures are offline when they are damaged or defected. These
techniques will result in cost saving and reducing labour requirements. These cases are more
prominent in the ageing aircraft structure which is repeatedly subjected to cyclic loads. Even
the confidence level in operating in such structures would increase as condition based
monitoring would provide new safeguards against unpredictable structural degradation of
aging structures. Propagation of Lamb waves is highly complicated due to dispersion, wave
splitting into multiple modes, and waves reflected from boundaries and discontinuities. As the
Lamb wave responses are complicated, some signal processing technique is required for the
sparse signal representation. Data driven based approach doesn’t require a prior knowledge of
the structure’s parameters unlike the model based approach. The aim of this work is to
develop technique to detect defect/damage in a structure and estimate its location using some
signal processing technique and machine learning.
2. Finite Element (FE) simulations
2.1 Simulation Parameters
The Finite element simulation for damage detection is carried out in commercially available
software package, ANSYS. All the sensor data is plotted in MATLAB. The plate is modelled
in the thickness plane with plane strain assumption. The plane strain approximation saves
computational time without affecting the trend of results. The test specimen is modelled using
a PLANE-183 (8-node quadrilateral) elements. Time step size and element size used for the
simulation greatly affects the time and accuracy of the simulation. Selection of element size
and time step is crucial in FE simulation of wave propagation. The maximum element size
and time step to ensure accuracy is selected based on the expressions:
Element size
Time Step
Mesh size of 0.5 mm and a time step of 0.2 μs is used to ensure the accuracy of the
simulations. The specimen was densely meshed, with much smaller elements to accommodate
the complicated mechanical response. An 8.5 cycle sinusoidal modulated tone burst of
frequency 100 kHz has been used for excitation of signal as shown in Figure 1.
Figure 2: Schematic of Undamaged Plate
Figure 1: Tone burst excitation signal
2.2 Modelling and Simulation of plate
2.2.1 Finite element modelling of undamaged plate
A two-dimensional thin aluminium plate of length 1 m and thickness 1.2 mm has been used
and only the thickness plane in the direction of wave propagation is modelled for generation
of Lamb wave in the specimen, using plane strain assumption. The material properties of
aluminium are considered and is as follows: density 2712 kg/m3, Poisson ratio 0.3 and
Young’s modulus 70 GPa. The schematic of the undamaged plate used in this case for
simulation is shown in Figure 2.
2.2.2 Velocity response of lamb wave (undamaged)
As shown in the schematic, velocity response is recorded using the sensor (node) placed
exactly at the middle of the plate. Response from the sensor is shown in Figure 3, showing
both the showing both A0 and S0 modes.
Figure 3: Lamb wave response (Particle velocity in x vs time) for the undamaged Plate
2.2.3 Validation of velocity response
Figure 4: Group Velocity Dispersion Curve
Dispersion curve is plotted for the given specifications and material of the plate. A MATLAB
based interface, Wavescope developed by University of South Carolina is used to plot the
dispersion curve [1]. Velocity for A0 from the dispersion plot comes out to be around 1939
m/s and S0 for around 5403 m/s at frequency of 100 kHz. Arrival time of the A0 and S0 wave
packet is subtracted from the peak time of the incident signal to calculate the velocity for A0
and S0 from the velocity response. The table below shows that the velocity response from the
simulation of lamb wave is less erroneous and this configuration of actuator and sensor can be
used to simulate the plate with the crack.
Mode
A0
S0
Dispersion curve
(m/s)
1939
5403
Velocity response
(m/s)
1866
5359
Table 1: Velocity Comparison
Figure 5: Schematic of plate with notch
2.2.4 Finite Element modelling of damaged plate
An open crack is introduced in plate to model the damage. A rectangular notch of 1cm x 1.2
mm is made at the 1/4th of the length of the plate, as shown in Figure 5.
2.2.5 Velocity response of lamb wave (damaged)
As shown in the schematic, velocity response is recorded using the sensor (node) placed
exactly at the middle of the plate. Response from the sensor is shown in Figure 6.
Figure 6: Lamb wave response (Particle velocity in x vs time) for the damaged Plate
A mode conversion is observed when the lamb modes interact with the discontinuity or
damage. An A0 mode converts into both A0 and S0 modes referred as A0A0 and A0S0
representing A0 mode generated from an A0 mode and S0 mode generated from an A0 mode
respectively [5]. The same pattern of nomenclature follows for the mode conversion in the
case of S0 mode. The circled wave packets in the velocity response shows the new wave
packets which are formed due to the mode conversion at crack and their reflections from the
edges. It is difficult to process this data for feature extraction and pattern recognition because
of the complexity of the wave packets in the response. Hence, A selective generation of the
mode is preferred for simulation.
2.3 Selective Mode Generation
Dispersive nature of Lamb waves makes the analysis of data a challenging task. Complexity is
further increased due to presence of multiple modes and reflections from the edges and
discontinuities. Complexity in simulation can be reduced by selectively generating the A0 and
S0 modes. Selective generation of lamb wave modes are achieved by placing the two opposite
poled PZT patches exactly one below another [2]. Generally speaking, both the S0 and A0
mode are sensitive to structural damage, and both can be used for identifying damage, though
the A0 mode exhibits higher sensitivity to damage in the structural thickness and delamination
in particular whereas the A0 mode outperforms the S0 mode with higher sensitivity to surface
damage such as surface cracks, corrosion or surface crack growth. In this study, A0 mode is
selectively generated. It’s merit over S0 mode includes shorter wavelength at a given
excitation frequency and larger signal magnitude.
2.3.1 Finite element modelling of undamaged plate
For selective generation of the A0 mode, two point forces 5 mm apart are applied on top and
the bottom surface in the opposite direction. The forces are applied almost at the edge of the
plate and sensor is located at the middle of the plate.
Figure 7: Schematic of the undamaged Plate;
Selective generation of A0 mode
Figure 8: Lamb wave response;
Selective generation of A0 mode
2.3.2 Velocity response of lamb wave (Undamaged)
As shown in the schematic, velocity response is recorded using the sensor placed exactly at
the middle of the plate. Response from the sensor located on the top of the plate is shown in
Figure 8, showing only the A0 mode.
2.3.3 Finite element modelling of damaged plate
An open crack is introduced in plate to model the damage. A rectangular notch of 1cm
*0.6mm is made at the 1/4th of the length of the plate as shown in Figure 9.
Figure 9: Schematic of the Plate with notch; Selective generation of A0 mode
Figure 10: Lamb wave response; (a) Sensor 1; (b) Sensor 2; (c) Subtraction plot
2.3.4 Velocity response of lamb wave (damaged)
Symmetric in-plane signals i.e. A0S0 are eliminated by subtraction of response of the top
sensor (1) from the bottom sensor (2). Hence, A0A0 and similar signals are the major residuals
modes after subtraction. Response from both sensors and subtraction plots is shown in Figure
10.
3. Signal Processing: Orthogonal Matching Pursuit
3.1 Orthogonal Matching Pursuit
The Matching Pursuit (MP) is a well-known technique for sparse signal representation. MP
approach is basically a 'greedy' algorithm which iteratively projects signal onto a large and
redundant dictionary of waveforms and chooses a waveform from that dictionary that is best
adapted to approximate part of the waveform. The signal is decomposed into waveforms
selected among a dictionary of time-frequency atoms that are dilations, translations or
modulations of a single window function. The matching pursuit algorithm is very simple. But
because of the sub-optimality, it suffers from slow convergence and poor sparsity result. The
orthogonal matching pursuit (OMP) removes this drawback by projecting the signal vector to
the subspace spanned by the selected atoms [3]. The atom selection method in OMP remains
the same as in MP. Because of the orthogonalization, once an atom is selected, it is never
selected again in subsequent iterations. Each iteration of the algorithm consists of two steps:
an atom selection step and a residual update step. The atom selection step finds the atom
which has the highest correlation with the current residual error.
The dictionary selected for OMP in the analysis is the Daubechies' least asymmetric wavelet
[2]. The symlet wavelet dictionary is selected as it retains high energy in fairly less number of
iterations. Symlet wave of level 4 with 5 vanishing moments is selected for this study. The
automated damage detection technique consists of two major steps. First step involves
increasing the sparsity of the signal using OMP with the aim to reduce the number of nonzero
data-points to improve ML classification. The second step is the ML classification where
ANN have been tested and used. Orthogonal Matching pursuit is used to reduce the non-zero
data points. This signal processing technique is implemented in MATLAB. After using the
OMP, the idea of reflections from the edge or the damage are much clearer. The magnitude of
reflection from the end or the boundary are larger than the reflection from the damage. The
stopping criteria used in this study is the no of iterations. The no of iterations is decided by
running the simulations for different no of iterations and iteration at which the only the three
different modes (1 from the reflection from damage and 2 different modes from the reflection
from the edges) are clearly represented in signal is used in this study. OMP is used on all the
velocity- time response data with the different damage locations to reduce the number of nonzero data points. Comparison of the velocity response after and before using OMP for one of
the damage location is shown in Figure 11 and Figure 12. Out of 5000 total data-points after
using OMP, only around 1100-1200 non zero data points remain. All the unnecessary features
like noise are discarded from the signal. Examples with different damage locations have been
considered to emphasize the efficiency of OMP in achieving sparsity. Figure 13 shows the
comparison of the non-zero data points before and after OMP for one of the damage location.
The comparison plot is for three different locations of damage. It is observed that the number
of non-zero data-points almost reduced to one-fourth.
Figure 11: Velocity- time response: Before OMP
Damage location (25 cm)
Figure 12: Velocity – time response: After OMP
Damage location (25 cm)
Figure 13: No of Non-zero data points
before and after OMP
4. Machine Learning: Artificial Neural Network
4.1 Training of Neural Network
In this study, for training of ANN network a database consisting of velocity responses
processed by OMP is used. The input database has around 20,000 rows of data. It consists of
two columns with one column having the velocity response and the other column having time.
Edges are estimated based on time computed using velocity and length of the plate. The
output / target dataset consists of binary values with 1 for reflection from damage and 0 for
the remaining points.
4.2 Artificial Neural Network Scheme
In this study, the ANN scheme is implemented using MATLAB Neural Network Toolbox [6].
A two-layer feed-forward network space, with sigmoid hidden and output neurons, has been
used. The network is trained with scaled conjugate gradient back propagation. Here, 10
hidden layers have been considered as shown in Figure 14 which depicts the implemented
ANN schematic. In Artificial Neural Network the complete input dataset is divided into three
subset of data. (a)Training set – serves to train the model – 70%, (b)Validation set – serves to
select the hyper-parameters – 15 % and (c) Test set – serves to estimate the generalization
performance (error) – 15 %.
Figure 14: ANN scheme implemented
4.3 Confusion Matrix and Performance - ANN
The Confusion plot shows the confusion
matrices for training, testing, and
validation, and the three kinds of data
combined. The network outputs are very
accurate, as you can see by the high
numbers of correct responses in the green
squares and the low numbers of incorrect
responses in the red squares. The lower
right blue squares illustrate the overall
accuracies.
The number of epochs is selected
by stopping the training when validation set
error increases (with some look ahead). In
this case, the no of epochs has been tried up
to 63 and 57 is found to be the best as after
57 the Validation set error increases. Figure
17 shows the above concept [4]. The
training error always decrease with the
increase in no of epochs because we are
optimizing the training error through the
stochastic gradient descent algorithm. The yellow
Figure 15: Confusion Matrix
marker indicates the no of epochs after which the validation set error increases.
Figure 16: Performance Plot
Figure 17: No of Epochs Calculation
5. Damage Detection
5.1 Damage Detection Technique
For a single damage in the structure, the aim is to firstly select the right mode for determining
the damage location. In general, the magnitude of reflection from a damage is small as
compared to the edge reflection. The velocity of a wave packet can be calculated by
subtracting the peak time of the incident signal from the arrival time of the wave packet to
calculate the velocity from the velocity response and the distance between the sensor and
actuator and damage location is also known in the dataset. The velocity response from lamb
wave after subtraction and before OMP shows mainly three wave packets (𝐴0 𝑚𝑜𝑑𝑒𝑠,
𝑠𝑒𝑙𝑒𝑐𝑡𝑖𝑣𝑒𝑙𝑦 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑). Time of arrival of these wave packets at sensor is used to detect
their origin, reflection from different edges and discontinuities. The wave packet 1
corresponds to the wave packet transmitted from the damage. Wave packet 2 corresponds to
the one reflected from the damage first and then from the edge at the actuator end. Wave
packet 3 corresponds to the packet reflected from the other end of the of the plate. It can be
concluded that the wave packet 1 and 3 corresponds to the reflection from the edges and wave
packet 2 corresponds to the reflection from the damage.
Figure 18: Wavepackets
Figure 19: Arrival of wavepackets (1,2 & 3)at sensor
As the velocity of the wave packet 2 is known the arrival time of wave packet 2 can be used
to find out the location of damage. After the Artificial Neural Network is trained, the network
can be used to classify damage in the case of new dataset and thus the location of damage can
be calculated.
5.2 Damage Location Estimation
The trained Neural Network is used to classify damage in the case of new damage dataset.
Two dataset with the different damage location is used to test the efficiency of the Neural
Network for the damage classification and location estimation. Damage location estimation
for two new datasets (A &B) is shown
where output of the network is plotted
Total distance travelled = (2x + 0.5) m = t*v
along with the wave velocity response. The
area of the response spanned by the green
line denotes the probability of damage. It
can be observed that ANN is very effective
to classify the damage. The arrival time for
the wave packet (spanned) can be calculated by taking the average of the arrival time of the
first and last peak point of the wave packet spanned. The wave packet 2 first get reflected
from the damage (located at x) and then from the edge at the actuator end.
Figure 20: Damage location calculated from ANN & exact location
Figure 21: ANN response for Data set "A"
Figure 22: ANN response for Data set "B"
6. Conclusions
Artificial neural network and OMP is used to automate the damage detection and estimating
the location of damage. The error is less than 3-4 %. This study attempts to develop a data
driven damage detection technique in metallic plate using Lamb wave response. The main
motivation behind this study was to make the technique computationally efficient so that
online health monitoring can be performed using the limited processing power available onboard. The orthogonal matching pursuit algorithm successfully extracts meaningful pulses
from simulated noisy signals. Hence, the technique possess potential for SHM of real-life
aerospace structures, which would be explored in the future work. Also, the proposed
technique presently works for detection of a single damage, further investigation and study is
required on dispersion of Lamb waves to make this technique implementable for multiple
damages. Other Configuration of sensor and actuators should be studied to improve the
efficiency of the Damage detection.
References
1.
2.
3.
4.
5.
6.
WAVESCOPE, LMSS Products, University of South Carolina
S. Agarwal and M. Mitra, ‘Lamb wave based automatic damage detection using
matching pursuit and machine learning’, Smart Mater. Struct. 23 (2014) 085012 (11pp)
G. Rath and A. Sahoo ‘A comparative study of some greedy pursuit algorithms for
sparse approximation’.
Hugo Larochelle, ‘Neural Networks – Course Content’, Université de Sherbrooke.
C. Ramadas, K. Balasubramaniam, M. Joshi, and C. Krishnamurthy, ‘Interaction of
lamb mode A0 with structural discontinuity and generation of turning modes in a Tjoint’, Ultrasonics, vol. 51, no. 5, pp. 586–595, 2011
Artificial Neural Network Toolbox, MATLAB 2014a
8th International Symposium on NDT in Aerospace, November 3-5, 2016
ATTENUATION OF FUNDAMENTAL ANTI-SYMMETRIC LAMB
MODE (Ao) IN ISOTROPIC PLATES
*C. Ramadas, Irfan Khan and Makarand Joshi
R&DE(E), DRDO, Alandi road, Dighi, Pune-411 015, India
Phone: +91-20-2704 4875, Fax: +91-20-2704 4861, *email: rd_mech@yahoo.co.in
Abstract
When a Lamb mode propagates in a defect free medium, amplitude of the mode reduces with the distance of
propagation. The reduction in amplitude is attributed to attenuation of the mode on account of material and
geometry assuming negligible dispersion and leakage. Experiments were performed, employing air-coupled
transducers, on quantifying attenuation coefficient of fundamental anti-symmetric Lamb mode (Ao) in isotropic
plates. Variation in attenuation coefficient with respect to excitation frequency has also been studied. Moreover,
some numerical modelling aspects of attenuation of the Lamb mode are also discussed.
Keywords: Ao mode, attenuation, isotropic plate, air-coupled transducers
1. Introduction
Lamb waves are ultrasonic guided waves which propagate in thin plate-like structures [1].
These waves have the potential to interact with defects in materials and the interaction is
manifested through mode conversion, amplitude [2], arrival time [3], energy [4], group
velocity [5] and occasionally frequency etc. Change in amplitude is one of the key features
used for damage detection while employing Lamb waves [6]. When a Lamb wave propagates
in a waveguide, reduction in amplitude with increase in distance of propagation depends on
various factors [7] – material attenuation (damping), geometry, dispersion and dissipation (or
leakage) into the surrounding medium. Among these four factors, attenuation of Lamb wave
due to material is very important. This has to be evaluated experimentally. However, because
of non-availability of data on attenuation of Lamb waves due to material, the effect of
material attenuation on Lamb wave propagation is not considered in numerical modelling,
especially when there is a defect and amplitude/energy is the criterion for damage detection.
In this paper, systematic experimental measurements on effective attenuation of Ao mode in
aluminium plates using air-coupled transducers [8] are presented.
2. Measurement of Effective Attenuation
Effective attenuation of Ao mode was measured in 1 mm and 2 mm thick aluminium plates at
three different frequencies – 100 kHz, 200 kHz and 500 kHz. Table 1 shows mechanical
properties of the material. Size of each plate was 500 mm in length and 300 mm in width. Aircoupled transducers were employed in order to transmit and receive single Lamb mode, Ao, in
the plates.
Table 1: Mechanical properties of aluminium
Young’s modulus
Poisson’s ratio
Density
70 GPa
0.30
2750
Angle () of the air-coupled transmitter and receiver probes to be oriented with respect to
normal to the plate is computed using Snell’s law, as given below.
Sin  = Vair/Vph
(1)
Here, Vair is velocity of acoustic wave in air, which is 330 m/s, and Vph is phase velocity of
Lamb mode to be transmitted in the plate. Once, plate thickness and excitation frequency are
given, phase velocity of the mode can be found from dispersion relation [1] or DISPERSE [9]
can be used to compute the phase velocity. Now, using the phase velocity, orientation of the
probe is calculated using equation (1). Table 2 shows the orientations of the probe, from
DISPERSE and those set in experiments, to capture Ao mode at different frequencies in 1 mm
and 2 mm thick aluminium plates.
Table 2: Probe angles from DISPERSE and experiments.
Frequency in kHz
Thickness in mm
100
1
2
1
2
1
2
200
500
Angle in degree
DISPERSE
Experimental
22.0
21
14.8
14
14.8
15
11.1
12
10.2
11
8.2
9
2.1 Experimental setup
Figure 1 shows experimental setup for transmitting and receiving Lamb waves. The setup
consists of pulse-receiver, digitizer and a desktop computer. Gas Matrix Piezoelectric (GMP)
based circular ( 25 mm) air-coupled probes, from Ultran Group, with central frequencies 100
kHz, 200 kHz and 500 kHz were used in experiments for generation and reception of the Ao
mode. Sampling frequency was set at 10 MHz.
Figure 1: Experimental setup for transmission and reception of Lamb waves.
2.2 Lamb wave signals
The following describes the procedure followed to capture the Ao mode at 100 kHz in 1 mm
thick aluminium plate. Both transmitter and receiver were oriented at 21o (refer Table 2) and
initial distance of separation between the transmitter and receiver was 130 mm. Figure 2(a)
shows Ao mode captured in this configuration. Keeping the position of the transmitter fixed,
the receiver was moved by 10 mm. Now the separation between the probes became 140 mm.
One more A-scan (Ao mode) was captured in this configuration. This process was continued
till the distance of separation was 300 mm. Figure 2(b) shows Ao mode captured at 200 mm
distance of separation.
Figure 2: Ao mode captured at (a) 130 mm and (b) 200 mm from transmitter
at 100 kHz in 1 mm thick aluminium plate.
Similar measurements were carried out at 200 kHz and 500 kHz frequencies as well. This
completes one set of experiments on 1 mm thick plate. One more set of experiments were
performed on 2 mm thick plate at 100 kHz, 200 kHz and 500 kHz frequencies. Each set of
experiments was repeated thrice. As a whole eighteen experiments were conducted.
3. Analysis of Experimental data
3.1 Group velocity
Initially, average group velocity of Lamb mode was calculated and compared to that obtained
from DISPERSE. This ensures that the captured mode is the required Lamb mode. Since A o
mode was captured for every 10 mm movement of receiving probe over a distance of 170
mm, group velocity of Ao mode was calculated taking arrival times at various distances.
Table 3 shows experimentally obtained average group velocities of Ao mode at 100 kHz, 200
kHz & 500 kHz in 1mm and 2mm thick aluminium plates. Since the experimentally measured
group velocities and probe angles set in experiments are in compliance with those obtained
from DISPERSE as listed in Tables 2 and 3, it was concluded that the mode captured was Ao.
Table 3: Group velocities of Ao mode
Frequency in kHz
100
200
500
Thickness in mm
1
2
1
2
1
2
DISPERSE, m/s
1771.1
2279
2290.3
2773
2904.3
3143
Experimental, m/s
1863
2194.5
2391
2876.4
2909.1
3125
3.2 Wave amplitude
As discussed earlier, when a Lamb wave propagates in a defect free plate, amplitude reduces
with distance of propagation. It is required to determine representative value of amplitude of
the wave group at excitation frequency. Therefore, Fast Fourier Transform (FFT) has been
performed on all the signals, thus converting time signal domain to frequency domain. In the
frequency domain, it was found that peak amplitude occurs nearly at the excitation frequency.
Figure 3 shows FFT of the signal at 130 mm in 1 mm thick aluminium plate at 100 kHz
excitation frequency.
Figure 3: FFT of Ao mode recorded at 130 mm in 1 mm plate at 100 kHz
Variation in amplitude with distance of propagation in 1 mm and 2 mm thick plates at three
different frequencies – 100 kHz, 200 kHz and 500 kHz is shown in Figure 4. Exponential
decay curve, with the following characteristic equation, has been fitted to the data.
A(x) = Ap exp(-ki x)
(2)
Where, A is amplitude at distance x from source or transmitter, Ap is amplitude at the source
or transmitter (x = 0) and ki is effective attenuation coefficient, expressed in Np/m. Here, ki
characterizes effective attenuation of the Ao mode in a given plate at a given frequency. Since
ki carries both effects of material and geometry, it is termed as ‘effective attenuation
coefficient’. Table 4 lists effective attenuation coefficient of the Ao mode for various cases.
Table 4: Attenuation coefficients of Ao mode
Frequency in
kHz
Thickness in mm
Effective attenuation
coefficient, Np/m
1
2.553
2
3.794
1
2.869
2
3.973
1
3.142
2
4.015
100
200
500
Figure 4: Variation in amplitude of Ao mode at 100 kHz in (a) 1 mm and (b) 2 mm thick plate;
At 200 kHz in (c) 1 mm and (d) 2 mm thick plate; At 500 kHz in (d) 1 mm and (e) 2 mm thick plate.
4. Discussion and Conclusions
Experiments were conducted to evaluate effective attenuation coefficient of Ao Lamb mode at
100 kHz, 200 kHz and 500 kHz frequencies in 1 mm and 2 mm thick aluminium plates. The
frequencies of excitation and thicknesses of the plate were chosen in such a way that the A o
mode falls in non-dispersive region. Air-coupled probes, which enable transmission and
reception of selective Lamb mode, were employed for transmission and reception of the Ao
mode. Orientations of the probes were computed using phase velocity obtained from
DISPERSE. Group velocity of the Ao mode was calculated from experimental measurements
to ensure that the desired mode is transmitted and received. To avoid any edge reflections, all
the measurements were carried out at the center of the plate. Leakage of Ao mode into the
surrounding medium, i.e. air was ignored, because, impedance mismatch between the
waveguide (aluminium plate) and air is very high. Therefore, in this work, reduction in
amplitude of the wave as it propagates through the medium was purely due to material and
geometry only. Variation in effective attenuation coefficient of Ao mode with frequency in 1
mm and 2 mm thick aluminium plates is shown in Figure 5.
Figure 5: Effective attenuation of Ao mode at various frequencies.
It is inferred from the Figure that effective attenuation of Ao mode increases with increase in
excitation frequency and it is more in a thicker plate when compared to a thinner one.
When effective attenuation characteristics of Ao mode in metals is compared with that in
composites [10], it is found that composites, both GFRP (Glass Fiber Reinforced Plastic) and
CFRP (Carbon Fiber Reinforced Plastic) of nearly same thickness as metal, cause more
attenuation. For example, at 500 kHz, effective attenuation coefficient of Ao mode in GFRP
and CFRP plates is 22.5 Np/m and 17.5 Np/m, respectively, whereas in the aluminium plate it
is 4.015 Np/m. This is because of the fact that, a composite laminate is a layered medium with
anisotropy. This results in higher attenuation of the mode. However, at low frequency, the
difference in effective attenuation coefficient between metal and composite is less. For
instance at 100 kHz, effective attenuation coefficient of Ao mode in GFRP and CFRP plates is
5 Np/m and 4.8 Np/m, respectively, whereas in the aluminium plate it is 3.79 Np/m.
Effective attenuation coefficient obtained from the curve fitting considers two aspects –
material and geometric attenuation. In order to capture attenuation of the wave in numerical
simulations, Rayleigh (proportional) damping is used [11-13]. Lamb wave attenuation
constants (LACs), which are nothing but mass and stiffness proportional constants in
Rayleigh damping frame work, are expressed in terms of group velocity (Cg), attenuation
coefficient (ki) and circular frequency (). Since Rayleigh damping model is a global model,
which means it assigns the same damping characteristics for entire model, it works well when
the damping is same (uniform) everywhere. When this model is adapted to capture attenuation
of Lamb waves, the model assumes that every waveguide has same attenuation coefficient.
Furthermore, it also assumes that all Lamb waves have the same attenuation characteristics.
For instance, consider propagation of Ao mode at 100 kHz from a 2 mm thick to a 1 mm thick
plate. Albeit attenuation coefficient of the mode is different in the 2 mm and 1 mm thick
plates (refer Table 4), it is not feasible to assign two different values of LACs in numerical
model. This is because of the fact that the LACs are defined globally for the entire model.
However, the model works well for simulating attenuation of Lamb wave in waveguides,
which result in uniform attenuation of the wave.
References
1. Rose J L, Ultrasonic Waves in Solid Media, United Kingdom, Cambridge University
Press, 1999
2. Staszewski W J, Mahzana S and Traynor R (2009) Health monitoring of aerospace
composite structures – Active and passive approach. Composites Science and
Technology, 69, 1678 – 1685
3. Diamanti K, Soutis C and Hodgkinson J M (2005) Non-destructive inspection of
sandwich and repaired composite laminated structures. Composites Science and
Technology, 65, 2059 – 2067
4. Ye Lu, Lin Ye, Zhongqing Su, Chunhui Yang (2008) Quantitative assessment of
through-thickness crack size based on Lamb wave scattering in aluminium plates,
NDT&E International, 41, 59-68.
5. Rosalie S C, Vaughan M, Bremner A and Chiu W. K (2004) Variation in the group
velocity of Lamb waves as a tool for the detection of delamination in GLARE
aluminium plate-like structures, Composite Structures, 66, 77 – 86
6. Lee C M , Rose J L, Cho Y (2009) A guided wave approach to defect detection under
shelling in rail, NDT&E International, 42, 174-180.
7. Schubert J K, Herrmann A S (2011) On attenuation and measurement of Lamb waves
in viscoelastic composites, Composite Structures, 94, 177-185
8. Bhardwaj M C, Non-Contact Ultrasound – The final frontier in non-destructive
analysis, Encyclopedia of Smart Materials, Edited by A. Bide man, John Wiley &
Sons, New York, 2001
9. Lowe M J S. Disperse software, version 2.0.16b, 2003
10. Sreekumar P, Ramadas C, Anoop Anand, Makarand Joshi (2015) Attenuation of Ao
Lamb mode in hybrid structural composites with nanofillers, Composite Structures,
132, 198-204
11. Ramadas C, Avinash Hood, Krishnan Balasubramaniam, Makarand Joshi and
Krishnamurthy C V (2011) Modeling of attenuation of Lamb waves using Rayleigh
damping: Numerical and experimental studies, Composite Structures, Vol. 93(8),
2020-25.
12. Ramadas C (2014) 3D modeling of Lamb wave attenuation due to material and
geometry in composite laminates. Journal of Reinforced Plastics and Composites,
33(9), 28-29.
13. Gresil M, Giurgiutiu V; Prediction of attenuated guided wave propagation in carbon
fiber composites, International Conference on Composite Materials 2013 (ICCM-19),
28 July – 02 August 2013, Montreal, Quebec, Canada.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Interaction of Fundamental Symmetric S0 Lamb Mode with Delaminations
in Composite Plate Structures
Saurabh GUPTA and Prabhu RAJAGOPAL*
Center for Non-Destructive Evaluation and Department of Mechanical Engineering, IIT Madras, Chennai
600036, T.N., India
*
Corresponding author: prajagopal@iitm.ac.in
Abstract
This paper addresses a gap in the literature on the 3-dimensional scattering of the fundamental symmetric Lamb
mode S0 from delimitations in composite plates. We study the scattering of low-frequency S0 Lamb mode from a
delamination in a stiffened 4-ply CFRP composite plate with 0/0/0/0 ply orientation. This work uses three
dimensional finite element simulations. The FE simulated in-plane displacement contour obtained from the
simulations represents wave propagation in the unidirectional composite laminate and includes complex wave
interaction at the delamination region. Far field scattering coefficients for the S0 Lamb mode are plotted as a
function of circumferential position around the delamination. Results show that the delamination size has less
influence on S0 Lamb wave scattering in the low-frequency regime where the S0 mode is non-dispersive. Further
analysis was done using two-dimensional FE simulation for different ply-layup orientations with S0 Lamb mode.
This study shows that ply-layup orientation and through-thickness delamination location in fiber composite
laminate have a significant influence on S0 Lamb mode interaction. This work will be useful for practical Lamb
wave based inspection of composite plate structures.
Keywords: Composites, Finite element analysis, S0 Guided Lamb wave.
1. Introduction
Recent investigations of aircraft and space construction techniques have explored the use
of composite materials because of their high strength to weight ratio and thermal stability.
Composites are made of fibres of different materials in order to increase or decrease the
strength of the composite laminate in the required direction. The damages in composites are
more critical than metals and their detection is difficult due to the anisotropic nature of the
composite laminates. Delamination is one of them and it is the most common defect in
composite laminates. In delamination, adjacent surfaces called laminae separate from each
other without any obvious visual evidence on the surface. This results in significant loss of
strength or stiffness of composite laminate. Therefore, analyzing the effect of delamination
and detecting it non-destructively has become a subject of considerable interest.
Ultrasonic Lamb waves offer a convenient approach to evaluate composite laminates
because they can propagate over a long distance even in materials with a high attenuation
ratio and thus a broad area can be quickly examined [1]. Wave propagation in layered
anisotropic media is more complex than isotropic materials [1-3]. This makes the detection of
defects and scattering analysis difficult in anisotropic materials. Due to the complex wave
interaction that occurs when hidden delamination damage is present, extensive research has
been carried out on guided wave delamination detection methodologies [3-6]. Authors
reported that S0 guided Lamb mode shows less attenuation compare to other modes when it
propagates through the laminated composite structures [3,4]. Quantitatively, delamination can
be detected in cross ply composite laminates using guided Lamb (S0) wave velocity [4]. It is
important to analyze the scattered energy in arbitrary directions for characterizing the
delaminations. In this context, a scattering study was done using SDP technique (Scattering
directivity pattern) with guided Lamb waves in CFRP quasi-isotropic
isotropic composite laminate [5
[5].
The results obtained in this study show that scattering amplitude depends
depends upon size as well as
through-thickness
thickness location of delamination. When guided waves propagate in a delaminated
composite structure, multiple reflections can occur within and around the delamination [6
[6-8].
A similar kind of study was done using fundamental
fundam
symmetric S0 Lamb mode to detect a
delamination at various ply-interfaces
interfaces in a quasi-isotropic
isotropic composite laminate [9
[9], and a linear
relationship was observed between the attenuation of the S0 Lamb mode and the degree of the
impact damage.
The previously mentioned studies provided an insight to the fundamental physical
phenomena of guided Lamb wave interaction at a delamination.
delamination They show that S0 guided
Lamb mode can successfully used for damage detection and characterization in composite
laminates. In this study, we analyzed th
the scattering coefficients, for 4-ply
ply unidirectional
CFRP composite laminate, at the monitoring points taken around the delamination by using
fundamental symmetric S0 Lamb mode. It shows that delamination size does not influence on
scattering significantly. Further analysis was conducted for 4-ply
4 ply symmetric and non
nonsymmetric composite laminates using two-dimension
two dimension FE simulation with different ply
ply-layup
orientations [(0/90)]S, [(0/45)]S, [(0/90)]
[
2, [(0/45/-45/0)], [(0/90/45/-45)] using S0 Lamb mode.
These studies show that S0 lamb mode has less influence on delamination size but it strongly
depends upon ply-layup
layup orientations and through-thickness
through hickness delamination location. Reflection
coefficients were calculated for all cases studied.
studied. This research will be useful for practical
Lamb wave based inspection of composite plate structures.
2. 3D FE Simulations of Delaminated Composite Laminates
A 3D FE method was used to simulate a 44 ply (0/0/0/0) orientation quasi
quasi-isotropic
composite laminatee with square delamination between ply 2&3 as shown in Fig
Fig. 1. A
commercial finite element
lement package [10]
[10] was used to generate the geometry and perform the
meshing operation. A schematic diagram of the configuration used in the FE simulations is
shown in Fig 1. Each lamina is modelled using eight-node
eight node 3D reduced integration solid brick
elements with hourglass control in which each node has three degrees of freedom. The
property of each lamina is given in Table1. Square shaped delaminations in terms of the
wavelength (λS0 ) was modeled at the centre of the composite plate.
TABLE 1. Elastic properties of CFRP composite lamina
(
147
)
(
8.17
)
(
8.17
)
(
2.42
)
(
2.42
)
(
3.1
FIGURE.1. Schematic diagram of the configuration used in FE simulations
ρ(kg/m3)
)
0.317
0.317
0.317
1550
Figure 2 shows the theoretical phase velocity dispersion curve for a (0/0/0/0) degree
orientation CFRP composite laminate in the 0o propagation direction
on calculated using
Disperse [11]. The dispersion results for unidirectional laminate show that fundamental
symmetric Lamb mode S0 and the anti-symmetric
anti
Lamb mode A0 exist at the low frequency
frequencythickness regime.
S0
Working
regime
How
to
Use
this
Template
(Second
Level
Heading)
(Use the Microsoft Word template style: Heading 2)
A0
FIGURE 2. Phase velocity dispersion
spersion curves for the [0/0]
[0/0 S quasi-isotropic
isotropic composite laminate (CFRP) at 0o
propagation direction
Figure 3 presents typical contour snapshots of FE simulated in-plane
in plane displacement
magnitude of the [0/0/0/0] degree fiber orientation of CFRP composite laminate. Figure 3(a)
shows an instant soon (42µs) after
afte the excitation in which the S0 Lamb wave is generated.
Figure 3(b) shows waves scattered by the square shaped delamination.
delamination
FIGURE.3. Typical snapshots contour of total in
in-plane
plane displacement for the [0/0/0/0] degree fiber orientations
composite laminate at the different time instances. (a) soon after excitation
excitation and (b) shortly after S0 Lamb wave
interaction with a 0.5λ S0 square delamination
delamin
located between second and third lamina
3. 2D FE Simulations of Delaminated Composite Laminates
A 2D FE method was used to simulate a 44 ply [(0/90)]S, [(0/45)]S, [(0/90/0/90)], [(0/45/
[(0/45/45/0)], [(0/90/45/-45)]
45)] orientation quasi-isotropic
quasi
composite laminate.
inate. A delamination was
introduced between the first and second lamina (for remaining cases 22-3 interface
delamination and 3-4
4 interface delamination) with an axis-span
axis
of d/λS0, and distance of L
from its left tip to the left beam end as shown in Fig. 4.
4 Monitoring points were ttaken in far
field (more than 5λS0) for all the cases studied. A commercial
commercial finite element package [9
[9] was
used to generate the geometry and
an perform the meshing operation. The dispersion results
results,
presented in Fig. 5, for cross-ply
ply composite laminates shows that either the fundamental
symmetric Lamb mode S0 or the anti-symmetric
anti
Lamb mode A0 exists at the working regime.
E
FIGURE.4. Schematic diagram of the configuration used in 2D
2D-FE
Working
regime
S0
S0
Working
regime
A0
A0
(a)
(b)
S0
S0
Working
Working
regime
regime
Higher order
modes
(c)
(d)
FIGURE.5. Dispersion curves for various ply-layups
ply
(a) [(0/90)]S (b) [(0/90)]2 (c) [(0/45)]S (d) [(0/45/
[(0/45/-45/0)]
4. Results & Discussion
Figure 6 represents
presents typical contour snapshots of FE simulated in-plane
in plane displacement of the
[0/0/0/0] degree orientation quasi-isotropic
quasi
composite laminate.
ate. Various stages of the wave
defect interaction can be seen, including effects such as mode trapping in the delamination. In
order to quantify the phenomenon, a scattering coefficient was obtained as the spectral ratio of
scattered to incident
ident signal amplitude. Figure 7 shows the scattering ratio for the different
monitoring positions and for the various delamination sizes in an overlay plot.
4.1 3D FE Simulation Results
Energy trapped within
delamination
FIGURE.6. Snapshots of the contour of total displacement mag
magnitude from 3D FE simulation of S0 Lamb wave
interaction with a square delamination 4-ply
4
composite laminate of size (a) λS0 (b) 0.75 λS0 (c) 0.5 λS0 (d) 0.25
λS0
FIGURE.7.Scattering coefficient of S0 Lamb incident at square delamination of 4 different dimensions (0.25λS0,
0.5λS0, 0.75λS0, λS0) located between second and third lamina in a 4-ply
4
composite laminate
Overall, the results show that very little wave energy is reflected from the delami
delamination, for
all sizes studied. The 3D wave scattering results show that the delamination size does not
influence scattering. Hence, subsequent models are 2D. The calculation
on cases an
and 2D FE
results shown in Figs. 8, 9 and 10 are listed in Table 2.
TABLE 2. Calculation cases and results in this study
Cases
Reflected wave
Delamination location
1-2 interface
2-3 interface
3-4 interface
Large
None
Large
Large
None
Multiple
reflections
Large
None
Large
Incident
ncident wave
[(0/90)]S
[(0/90)]2
S0
S0
[(0/45)]S
S0
[(0/45/-45/0)]
S0
Small
None
Small
[(0/90/45/-45)]
S0
Medium,
dispersive
Multiple
Small, more
dispersive
Figure 8 shows the A scans of in-plane
in
displacement of [(0/90)S] & [(0/90)]]2 ply lay-ups.
Figure 9(a), (b) & Fig. 10 shows reflection coefficients for S0 Lamb mode incidence
incidence. In-plane
S0 lamb wave mode was excited at the source point x=zero, by exciting all the nodes through
the thickness, in +x direction as shown in Fig. 4. Large
arge reflection coefficients were observed
when the delamination was located
locate in the 1-2 and 3-4 interfaces Fig. 9(a), (b)
b) for ply lay-ups
[(0/90)]S, [(0/90)]2, [(0/45)]S, [(0/45/
(0/45/-45/0)] vanishing when the delamination is at the mid
plane.
4.2 2D FE Simulation Results
,
,
,
(a)
(b)
,
,
,
(c)
,
(d)
,
,
,
Observe multiple
reflections
(f)
(e)
FIGURE.8. In-plane
plane displacements with through-thickness
through thickness location of delaminations between layers 1&2, 2&3,
and 3&4 for CFRP composite laminate (a), (b), (c) for ply lay-up [(0/90)S] respectively & (d), (e)), (f) for ply layup [(0/90)2] respectively
(a)
(b)
FIGURE.9. (a), (b) Reflection coefficients
oefficients for ply lay-ups
lay
[(0/90)]S, [(0/90)]2, [(0/45)]S, [(0/45/-45/0)]
45/0)] upon
incident of S0 Lamb mode.
Large reflections at the delamination (mid-plane
(mid plane interface) can be observed for the fully
non-symmetric
symmetric composite laminate [(0/90/45/-45)]
[(0/90/45/ 45)] as shown in Fig. 10. Multiple reflections
were observed within and around the delamination due to discontinuity in S0 Lamb mode
propagation, at the entry and exit of the delamination. The detailed physics of this phenomena
is studied in more detail.
Reflection at mid plane
FIGURE.10. Reflection coefficient for ply laylay up [(0/90/45/-45)] upon incident of S0 Lamb mode
5. Conclusions
FE simulations were conducted for 4-ply unidirectional and cross ply CFRP composite
laminates. The 3-dimensional scattering describes that there is less influence of delamination
size upon S0 Lamb mode interaction. For all cases studied, 2D FE simulation results show that
the reflection coefficient strongly depends upon ply-layup orientation and through-thickness
delamination location. This study shows that it is difficult to detect the delamination when it
is present at the mid plane of the composite laminate for all the cases except [(0/90/45/-45)]
ply-layup orientation with incident of S0 Lamb mode. Multiple reflections were observed at
non-symmetric delamination location site in the received signal with S0 Lamb mode. The
outcome of this study will be helpful for study of interaction of S0 guided Lamb modes with
delaminations in laminated composites with varying ply-layups.
References
1.
2.
3.
4.
A. Raghavan and C. E. S. Cesnik, Shock Vib. Digest, 39, 91–114 (2007).
P. Rajagopal and M. J. S. Lowe, J. Acoust. Soc. Am, 124, 2895–2904 (2008).
N. Guo and P. Cawley, J. Acoust. Soc. Am, 94, 2240 (1993).
N. Toyama, J. Noda and T. Okabe Composite Science and Technology, 63, 1473-1479
(2003).
5. C. Ng and M. Vedit, J. Acoust. Soc. Am, 129, 1288–1296 (2010).
6. T. Hayashi and K. Kawashima, Ultrasonics, 40, 193–197 (2002).
7. R. S Panda, P. Rajagopal and K. Balasubramaniam, “An Approach for Defect
Visualization and Identification in Composite Plate Structures Using Air-Coupled Guided
Ultrasound”, in Review of Progress in Quantitative Nondestructive Evaluation,
(American Institute of Physics 1650, Melville, NY, 2015), pp. 1299-1306.
8. Z. Su, L. Ye and Y. Lu, J. Sound Vib, 295, 753–780 (2006).
9. Birt. E. A, NDT and Condition Monitoring, 40, 335-339 (1998).
10. See http://www.3ds.com/products/simulia/portfolio/abaqus/abaqus-portfolio for
ABAQUS Analysis User’s Manual, Version 6.10-1; accessed 28 July 2015.
11. DISPERSE user's manual, version 2.0.11 (2001).
Detecting Onset of Combustion Instability in Gas Turbines Through Flame
Visualisation using Fiber Optic Bundle
Suma H. 1, Joel Vasanth 2, B. Srinivasan*1 and S. R. Chakravarthy2
1: Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai 600036, India
2: Department of Aerospace Engineering, Indian Institute of Technology Madras, Chennai 600036, India
*balajis@ee.iitm.ac.in
Abstract
In this paper, we propose and demonstrate a precursor for predicting the onset of combustion instability in gas
turbines through visualising the flame with the help of a fiber optic bundle and photomultiplier tube (PMT). The
performance of such a flame visualization system is correlated with that of a conventional pressure transducer
deployed in the combustion chamber.
Keywords: Combustion instability, Fiber optic bundle, flame visualization, precursors
Introduction
Combustion instability poses a negative impact on the performance and structural durability
of both land-based and aircraft gas turbine engines [1]. As such, detection of the onset of
combustion instability is not only critical for performance monitoring and fault diagnosis, but
also for initiating efficient decision & control of such engines.
Different techniques have been developed to monitor and study the flame dynamics in a
turbine. Among them, recent state-of-the-art techniques employ fast, on-the-go data-driven
detection of precursors which senses features that help to detect and characterize combustion
instability. Precursors such as decay of Hurst exponent [2] , loss of chaotic behaviour [3],
dawn of ‘lock on’[4] have been reported. Another promising method employing time series
analysis has also been investigated previously [5], [6]. Specifically, Sarkar et al.[7], proposed
a fast symbolic time series analysis (STSA) approach built upon the generalized D-Markov
machine to construct a complexity-based measure for detecting an early onset of thermoacoustic instability in swirl-stabilized combustors. This approach models spatio-temporal codependence among time series from heterogeneous (e.g. pressure and chemiluminescence)
sensors to generate a data-driven precursor which is uniformly applicable across multiple
experiment protocols with various premixing levels. This method exhibits robustness to
varying levels of sensor noise when compared to other existing methods. Based on such
work, we are proposing a method where the chemiluminescence sensor may be replaced by
fiber optic bundle which offers advantages like immunity to electromagnetic interference,
small size and multiplexing capability. In fact, a similar fiber based mechanism has been
reported previously by Muruganandam et al.[8], [9] for the prediction of lean blow out in
combustion engines. In this paper, we demonstrate the response of fiber optic bundle to the
turbulent flame for replacing the chemiluminescence sensor.
Background
In previous work [7], information from heterogeneous sensors (chemiluminescence and
pressure) is used to obtain precursors to the problems of combustion instability. A statistical
process known as the D-Markov Machine is used to produce a parameter known as the
information entropy rate, as a measure of the content of information in the detected time
signal from both the sensors. The time-series data is reduced into a string of symbols
(
) based on partitioning of the time series into different cells, each cell
being assigned a symbol. Using the concept of probabilistic finite state automata, a finite set
of states is generated with transitions between these states corresponding to a symbol. The
states are then split based on the criterion of minimization of entropy and then states that have
similar statistical behavior are merged. Here, entropy refers to a measure of predictability of
information generated by a particular process. The more the predictability of a particular
outcome, the lower the entropy of the information contained in the process generating that
outcome. When this measure is computed by the D-Markov machine, it is called the DMarkov entropy rate. More formal definitions for these tools are provided in [7] .
Combustion instability is characterized by self-sustained high amplitude periodic oscillations.
The dynamic complexity of such organized oscillations are low. So this means the waveforms
become less noisy or less random, and this would correspond to a low value of the D-Markov
entropy rate in the obtained time signal during instability. Transition to instability is marked
by a decline of entropy rate. Thus, as soon as this drop in entropy rate is noted, a suitable
control action can be applied to suppress the impending instability.
In order to demonstrate the feasibility of using a fiber bundle with PMT as a replacement for
the chemiluminescence sensor, we demonstrated the response of fiber optic bundle to a
bunsen burner as an initial controlled experiment. Bunsen burner was chosen since it has a
flame covering the blue region of the visible light spectrum which resembles more to the
flame in the combustor. Experiments were conducted using commercial optical fiber bundle
of active diameter 4.8mm, made of Schott Puravis glass fiber with a transmission of more
than 60% at 546 nm and a numerical aperture of 0.57. Photo Multiplier Tube (PMT)
(Hamamatsu R375) is used as the detector with a negative bias of 900V. One end of the fiber
bundle is exposed to the bunsen burner flame and the other end is connected to the PMT.
Fiber bundle is kept approx. 20cm away from the burner flame. The frequency response of
the receiver was tested by exciting the burner flame with a loud speaker held transversely to
the burner flame. Results are as shown in Figure 1.
x 10
-3
6
Magnitude (a.u.)
Amplitude (V)
2
1
0
-1
-2
-0.1
0
Time (s)
4
2
0
0
0.1
1 a)
250
2
Magnitude (a.u.)
Magnitude (a.u.)
100 150 200
Frequency (Hz)
1 b)
2
1.5
1
0.5
0
0
50
100
200
Frequency (Hz)
1 c)
1.5
1
0.5
0
0
100
200
Frequency (Hz)
1 d)
Fig. 1 : Response of the fiber optic system to the Bunsen burner flame a) output of the PMT when the flame is
excited with 50Hz using loud speaker. FFT of the sensor output when the flame is excited with b) 50Hz, c)
75Hz and d) 100Hz
Even though the results exhibit significant peaks at 50 Hz and 150 Hz in all the above data, it
is clear that the fibre optic system is able to sense the frequencies associated with the
flickering of the flame. The 50 Hz and 150 Hz peaks are attributed to power supply leakage
noise associated with the PMT. The successful detection of the pressure excitation waves is a
‘proof of concept’ and it motivated to extend the experiment to the laboratory scale
combustor.
Experimental Results and Discussion
The experimental setup used to carry out the combustion instability studies is shown in Fig.
2. The setup consists of a 1440 mm long combustor stabilized by a circular bluff body of
diameter 30 mm and thickness 15 mm placed about 1 cm downstream of the duct from the
inlet end. Two piezoelectric transducers (PZT) are placed near the base of the flame to
monitor pressure variations in that section. The multimode fiber optic bundle is placed at 20
mm from the optical access window and is attached with a photomultiplier tube (Hamamatsu
R375) for monitoring the flame.
2 a)
2 b)
Fig 2. a) Schematic diagram of the laboratory scale combustor apparatus [7] b) Experimental setup
The longitudinal eigenmodes of the combustor are self-excited by varying the levels of airfuel premixing. In our experiment, the combustion mixture consists of 300-500 lpm of air
and 30-40 lpm of methane. The acoustic boundary conditions can be considered as closedopen duct with closed (inlet) end only being an approximation of swirler with 0o vane angle.
The first longitudinal eigenmode of the combustor is obtained at a self-excitation frequency
of 161 Hz. This is in close approximation to a standing-wave of f = c/4L, corresponding to
the closed-open geometry, where the veolocity c ~ 920 m/s. Comparison of the fiber optic
bundle based measurements and PZT measurements at the instability state are plotted in time
series and the corresponding FFTs are also plotted in Fig.3.
Output Data of Sensors
1.5
PZT
PMT
Normalized Amplitude
1
0.5
0
-0.5
-1
-1.5
0
0.05
0.1
Time (s)
0.15
0.2
3 a)
FFT spectrum of sensors
PZT
PMT
Normalized Magnitude
1
0.8
0.6
0.4
0.2
0
0
50
100
150
200
250
Frequency (Hz)
3 b)
300
350
400
Fig. 3a) output data of both sensors processed with moving average filter and b) FFT spectrum at instability.
As seen in Fig. 3b, the dominant frequency (~150 Hz) corresponding to the longitudinal
mode as well as its harmonic match well for the fiber bundle based measurement and the PZT
measurement. From Fig. 3a, we can also gather that fiber optic bundle measurement and PZT
measurement are in phase implying that the heat release and pressure fluctuations are in
phase as well. This is in accordance with the Rayleigh criterion that the pressure and heat
release fluctuations are to be in phase for positive feedback, thereby causing thermo-acoustic
instability.
Response of the system is also investigated in the intermittency regime. Intermittency is an
intermediate regime of flow conditions observed during transition from stable to unstable
operation. It is characterized by intermittent bursts of high-amplitude tonal sound, alternated
by low-amplitude aperiodic fluctuations at random intervals to each other. Therefore, both
signals are noisier than the case of full-blown instability, but still some distinct periodicity
exists. This was observed for the case of 300 lpm and 30 lpm of air and fuel respectively, as
observed in Fig 4.
We find that the output of the PMT is noisier than the pressure transducer in this
intermittency regime in which we encounter a turbulent flame. As such, the SNR of PMT
data is much lesser than that of PZT. However, the spectral features match well between the
two measurements (with a peak at 184 Hz) and the relevant information for detecting the
intermittency can still be extracted from the PMT output. Since combustion instability is
typically preceded by intermittency, results obtained during intermittency period can be
developed for predicting the onset of thermo- acoustic instability.
Output Data of PZT
Output Data of PMT
1
Normalized Amplitude
Normalized Amplitude
1
0.5
0
-0.5
-1
0
0.05
0.1
0.15
Time (s)
0.2
0.5
0
-0.5
-1
0
0.05
0.1
0.15
Time (s)
0.2
4 a)
FFT spectrum of sensors
PZT
PMT
Normalized Magnitude
1
0.8
0.6
0.4
0.2
0
0
50
100
150
200
250
Frequency (Hz)
300
350
400
4 b)
Fig.4 a) Output data of fiber bundle-based measurements and PZT measurements at intermittency upon
processing using moving average filter and b) the corresponding FFT spectrum.
Challenges
One of the key issues involved in optimizing the fibre optic system performance is improving
SNR since the turbulent flame causes a high noise level. There is scope for using better signal
processing algorithms for filtering out the noise. Another challenge is the design of a suitable
packaging for the optical probe such that it prevents dust particles from getting deposited on
the probe face and also isolating the probe from the high temperature environment. The
optical probe that we are using can withstand a maximum temperature of 350°C, whereas the
inside temperature of the combustor may reach a temperature beyond 500 °C . We may have
address this issue by placing it in cooler regions of the combustor.
Conclusion
In this paper, we have presented a precursor for predicting the onsets of the thermo-acoustic
instability by monitoring the chemiuminescence in the combustor using a fiber optic bundle.
The fiber bundle with PMT system offers a reliable and compact solution for the
chemiuminescence sensors. Ongoing work is designing a packaging for the optical fiber
probe and implementing suitable noise reduction techniques.
Acknowledgment
The authors would like to thank Schott AG for providing the fiber bundle used in our
measurements.
References
[1] S. B. Lippman, Gas Turbine Combustion: Alternative Fuels and Emissions, Third
Edition. 2013.
[2] V. Nair and R. I. Sujith, “Multifractality in combustion noise: predicting an impending
combustion instability,” J. Fluid Mech., vol. 747, pp. 635–655, 2014.
[3] V. Nair, G. Thampi, S. Karuppusamy, S. Gopalan, and R. I. Sujith, “Loss of chaos in
combustion noise as a precursor of impending combustion instability,” Int. J. Spray
Combust. Dyn., vol. 5, no. 4, pp. 273–290, 2013.
[4] S. R. Chakravarthy, R. Sivakumar, and O. J. Shreenivasan, “Vortex-acoustic lock-on
in bluff-body and backward-facing step combustors,” Sadhana - Acad. Proc. Eng. Sci.,
vol. 32, no. 1–2, pp. 145–154, 2007.
[5] C. S. Daw and C. E. a. Finney, “A review of symbolic analysis of experimental data,”
Rev. Sci. Instrum., vol. 74, no. 2, pp. 915–930, 2003.
[6] V. R. Unni, A. Mukhopadhyay, and R. I. Sujith, “Online detection of impending
instability in a combustion system using tools from,” Int. J. spray Combust. Dyn., vol.
7, no. 3, pp. 243–256, 2015.
[7] S. Sarkar, S. R. Chakravarthy, V. Ramanan, and A. Ray, “Dynamic data-driven
prediction of instability in a swirl-stabilized combustor,” Int. J. Spray Combust. Dyn.,
pp. 1–19, 2016.
[8] T. M. Muruganandam, S. Nair, and Y. Neumeier, “Optical and Acoustic Sensing of
Lean Blowout Precursors,” in 38th AIAA/ASME/SAE/ASEE Joint Propulsion
Conference & Exhibit, 2002, July, pp. 1–10.
[9] T. M. Muruganandam and J. M. Seitzman, “Optical Sensing of Lean Blowout
Precursors in a,” Proc. ASME/IGTI Turbo Expo 2003, January 2003, pp. 1–9, 2003.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Practical Experiences in POD Determination for Airframe ET Inspection
Virkkunen, I.1 and Ylitalo, M.2
1
Trueflaw Ltd., e-mail: iikka@trueflaw.com
2
Patria Aviation Oy
Abstract
Evaluation of NDT reliability has received increasing emphasis in recent times. In particular, quantifying the
probability of detection (POD) attained in routine inspections have become more widespread. Although there are
good guidelines and standards for POD determination, the process is still far from trivial. Various choices made
during the experimental set-up may have significant effect on the results. Also, the cracked samples used are often
limited necessitating various compromises in the analysis.
Patria performed a set of POD studies for eddy-current inspections performed on various parts of typical metal
airframe. The project included manufacturing of cracked samples, organizing the inspection of these samples and
final analysis of the results. Several inspectors from different organizations took part in the exercise. The project
was done in collaboration with Finnish and international partners.
The data showed various unlikely events (small hits, big misses and poor separation), which necessitated
adjustment for the standard methodology. Contrary to expectation, the false call rate did not show significant
correlation with the inspection performance. When the â vs. a and hit/miss analyses could be directly compared,
they showed surprisingly poor correlation and caution is advised in using â vs. a analysis for manual inspections
such as the ones shown here.
Keywords: Probability of detection (POD), Eddy current inspection (ET)
1. Introduction
The best practices of estimating probability of detection (POD) in non-destructive evaluation
(NDE) are now well established. The venerable MIL-HDBK-1823A (most recent release from
2009) [1] is used extensively in the aerospace industry and is now finding increasing use also
in other areas, like the rail industry and even nuclear industry. The methods have recently been
standardized by ASTM (ASTM-E2862 [2] and ASTM-E3023 [3]) and these standards are
congruent with the current MIL-HDBK methodology.
Despite the now standardized methodology and significant tradition in POD determination, the
process still offers some practical challenges. The requirements for cracked test pieces are
sometimes difficult or costly to fulfill, the statistical analysis may prove demanding and,
perhaps most importantly, justifying that the various assumptions behind the methodology are
fulfilled to sufficient extend may prove challenging.
The standard practice offers two variant of POD curve estimation, the â vs. a approach and the
hit/miss approach. The â vs. a approach models, in simple terms, the NDE reliability as kind of
measurement system problem, where the quantity to be measured (crack size a) give rise to
measured signal (â) proportional to the measured quantity and the task is to determine the
possible existence of the signal with decreasing a (and thus decreasing â). The system has noise
both related to the signal and independent of the signal. That is, â varies due to factors other
than a, like crack orientation and tortuosity, which results in noisy â vs. a relation. In addition,
there's noise, that is independent of a, e.g. electric noise on the signal path. Thus, the task is to
find a decision threshold (â value), that minimizes false calls from the noise and, in parallel,
maximizes the number of cracks found (i.e. cracks with â above the threshold), given the
variation in the â-vs-a relation. This is done by fitting a linear function through the â-vs-a data,
computing prediction intervals to take the noise and statistical uncertainty into account. The
resulting best-fit and confidence limit lines are then compared to the set detection threshold and
the corresponding POD curves computed. For input, the â vs. a analysis requires a set of
representative flaws (at least 40) and measurements of signal strength â and corresponding
crack size a. In addition, noise independent of crack size needs to be evaluated either with
additional measurements of crack-free samples or in connection with the same sample set
measurement.
The hit/miss approach, in contrast, does not deal with signal values, but estimates the POD
curve based on binary results, that is hits (correctly found cracks) and misses (cracks not found
in the inspection). Because the data contains less information (regarding the correlation between
crack size and signal strength or "ease of detection") more samples are needed for reliable POD
determination. The POD curve is solved using generalized linear model and a chosen link
function (typically logit), that gives the shape of the POD curve using maximum likelihood fit
to the data. The corresponding confidence limits are then obtained by the likelihood-ratio
method, where a likelihood surface near the maximum likelihood value is interrogated, POD
curves with likelihoods corresponding to the chosen confidence interval computed and the
lower (and upper) limit curves resolved. For input, the hit/miss analysis requires a set of
representative flaws (at least 60) and hit/miss results for each crack. In addition, the hit/miss
results should exhibit a range with "unlikely to find" cracks, a range with "likely to find" cracks
and transition in between. Otherwise, the model does not describe the data and, while a fit may
in some cases be obtained, it does not describe the underlying probability of detection.
In both cases, the basic assumptions underlying both POD models should be fulfilled: the POD
should be an increasing function of the crack size and should reach 100% with sufficient crack
size. If the data contains signs of violation of these assumptions (e.g. a miss with big crack
length indicating that the POD does not reach 100% even with large crack size), the standard
models are not applicable and alternate model must be sought.
Patria performed a set of POD studies for eddy-current inspections performed on various parts
of typical metal airframe. The project included manufacturing of cracked samples, organizing
the inspection of these samples and final analysis of the results. Some of the cracked samples
were provided by collaborating organizations. Several inspectors took part in the exercise from
different organizations. The project was done in collaboration with Finnish and international
partners. This study provided an opportunity to study various practical aspects of the POD
determination process and to compare POD results obtained in different settings.
2. Materials and methods
The study was divided in three cases, as described in Table 1. Each case had different set of
cracked samples and was completed in one "go". The analyses were completed with the openly
available mh1823 software package [4].
Table 1. Summary of studied inspection cases.
Case
Description
Cracks
A
Typical fillet
49
B
Rivet hole
58
C
Rivet hole
68
Inspectors
7
11
7
3. Results and discussion
For each case, the practical difficulties obtained were somewhat different and are analyzed on
case-by-case basis below.
3.1 Case A
For case A a hit/miss analysis was completed. The sample set contained somewhat smaller
number of cracks (49) than required by the MIL-HDBK [1]. However, the crack sizes were well
distributed in terms of hits and misses and showed no adverse behavior in the statistical analysis.
Thus, hit/miss analysis was deemed appropriate for the data. The lack of sufficient cracked
samples increases the uncertainty of the analysis and thus the reported a90/95 values are expected
to be greater than what would be obtained with additional number of samples. The cracks used
were produced using mechanical fatigue. Produced cracks were inspected using automated ET
and selected cracks were destructively examined.
A typical obtained POD curve is shown in Figure 1. The data shows good separation between
crack sizes likely to be missed, a transition zone and crack sizes likely to be found. On several
cases, the inspectors also found some very small cracks. This unlikely hit significantly changed
the confidence bounds and, paradoxically, increased the size of the computed a90/95. Such curves
were re-analyzed with the small hit changed to miss (thus "worsening" the inspection behavior)
and smaller a90/95 values were obtained.
Figure 1. Typical POD curve for case A. The curve includes an unlikely small hit, which widens
the confidence bounds and paradoxically decreases the measured performance. With this hit
changed to miss, the a90/95 value decreased by 14%.
The inspectors also showed strong variation in false call rates. Interestingly, the false call rate
was not correlated with inspection performance (as measured by the a90/95). Figure 2. shows the
obtained a90/95 values in comparison to the false call rate of the inspectors.
False calls
a90/95
Figure 2. False call rate as a function of obtained a90/95 values. Negative correlation would be
expected, but there's no clear correlation observed.
The overall performance was not quite as good as was hoped. This was attributed partly to
significant time pressure during the inspection.
3.2 Case B
For case B a hit/miss analysis was completed. The sample set contained somewhat smaller
number of cracks (58) than required by the MIL-HDBK [1]. However, the crack sizes were well
distributed in terms of hits and misses and showed no adverse behavior in the statistical analysis.
Thus, hit/miss analysis was deemed appropriate for the data. A typical obtained POD curve is
shown in Figure 3. As in the case A, the data shows good separation between crack sizes likely
to be missed, a transition zone and crack sizes likely to be found. The overall results were
significantly better than for Case A.
Figure 3. Typical POD curve for case B.
In case of one inspector, a single large (or larger than other) crack was missed. This unlikely
event significantly changed the maximum likelihood curve and widened the confidence bounds.
The resulting a90/95 value was quite large for this data set, but with the single big miss changed
to hit, decreased by 52%. Furthermore, with the one outlier, the confidence bounds do not seem
to sufficiently cover the variability: the biggest miss is still significantly over the computed
a90/95 value and would be highly unlikely, according to the computed POD curve. Thus, the
single big miss effectively calls to question the applicability of the POD model used. A
comparison is shown in Figure 4.
Figure 4. Atypical POD curve, where a single big miss has significantly altered the obtained
POD curve. For comparison, curves computed with the same data except the biggest miss
changed to hit are shown in light-blue. With the modified data, the computed a90/95 decreased
by 52%.
False calls
Again, the inspectors also showed strong variation in false call rates and the false call rate was
not correlated with inspection performance (as measured by the a90/95). Figure 5. shows the
obtained a90/95 values in comparison to the false call rate of the inspectors.
a90/95
Figure 5. False call rate as a function of obtained a90/95 values. Negative correlation would be
expected, but there's no clear correlation observed.
3.3 Case C
For case C both â vs a and a hit/miss analysis were completed. The sample set contained 68
cracks. However, the inspection performance in this case was better than expected and most
inpsectors found most of the cracks and some inspectors found all of the cracks. Paradoxically,
this good performance caused numerical difficulties with the POD curve determination (for the
hit/miss analysis) since now the crack sizes were not well distributed in terms of hits and misses,
despite large population of small crack sizes. To obtain maximum likelihood estimates for the
hit/miss analysis, a single small miss was added to the data. (Had there been such a small crack
in the samples, it would have likely been missed by both the inspection under investigation and
the previous after-manufacturing inspection. Thus assuming such a miss is not unrealistic.) This
single miss allowed the maximum likelihood estimate to converge and provided a90/95 results.
In practice, the induced single miss forced the POD curve to steep curve between the artificial
miss and the smallest found crack and the lower-limit estimate near the second-smallest crack
missed. Thus the shape of the POD curve does not carry much information, but the obtained
a90/95 results are justifiable. A typical POD curve is shown in Figure 6.
Figure 6. Typical hit/miss POD curve for case C.
For case C, results enabled both â vs. a and hit/miss analyses to be completed and allowed
direct comparison between the two methodologies. Figure 7. shows typical POD curve obtained
from â vs. a analysis.
Figure 7. Typical â vs. a POD curve for case C.(Same inspector as for Figure 6 for direct
comparability.)
To compare the POD values obtained from â vs. a and hit/miss analysis, the values were
compared inspector-by-inspector. The results are shown in Figure 8. Although both results were
obtained with standard methodology, they present significant variation and the overall
correlation is not very good. For the very small a90/95 sizes (where inspectors found all or almost
all the cracks), the hit/miss analysis shows smaller (better) a90/95 values indicating, that the
inspectors included factors other than the signal strength â for their judgement (e.g. signal
stability in repeated measurements were sited). Conversely, for the larger a90/95 values, the â vs.
a results show smaller a90/95 values. This can be attributed to the â vs. a methodology failing to
account sufficiently to the larger missed cracks in the data. Furthermore, the â vs. a is sensitive
to variation in the â vs. a relation, which in this case was also affected by inspector reporting
practices. Values of â were read from the equipment screen and there may have been differences
of accuracy between inspectors in this respect. This accuracy did not affect the inspector
performance (as shown in hit/miss), but it did affect the confidence bounds obtained from â vs.
a and thus measured performance. On one occasion, the reported â vs. a relation showed
significant non-linearity and the reliability of the â vs. a was questionable (despite this nonlinearity having no effect on actual inspector performance). In conclusion, the hit/miss method
seems to better describe the present manual inspection case and caution is advised if using â vs.
a for such cases.
a vs. â a90/95
hit/miss a90/95
Figure 8. Comparison of â vs. a and hit/miss a90/95 results.
Finally, case C was, on the surface, very similar in inspection arrangement with case A.
However, inspectors showed significantly better performance in Case C. This can be attributed
to smaller time-pressure during case C, possible learning from earlier cases and differences in
samples. This indicates, that rather small changes in the set-up or inspector feedback can have
significant impact to the obtained POD performance.
4. Conclusions
Three separate POD exercises were completed, each showing separate experimental challenges
and solutions. The following conclusions may be drawn from this study:
o Contrary to expectation, the false call rate did not show significant correlation
with the inspection performance.
o The data showed various unlikely events (small hits, big misses and poor
separation), which necessitated adjustment for the standard methodology. This
shows, that the standard POD methodology can not be used as a "black box",
and must be accompanied by careful analysis of the underlying technical and
statistical factors leading to the obtained POD values.
o â vs. a and hit/miss analyses showed surprisingly poor correlation and caution is
adviced in using â vs. a analysis for manual inspections such as the ones shown
here.
o Small changes in the inspection set-up (e.g. time pressure) can have significant
impact to the obtained POD performance.
Acknowledgements
This work was supported by several partners who provided test pieces and inspection results
for the study. Their support is gratefully acknowledged. In particular, the authors wish to thank
RUAG aviation, Switzerland for providing some of the test samples for this study.
References
1.
2.
3.
4.
Anon. 2009. Nondestructive Evaluation System Reliability Assessment. Department of
Defense Handbook. MIL-HDBK-1823A. 171 p.
Anon. 2012. Standard Practice for Probability of Detection analysis for Hit/Miss Data.
American Society for Testing and Materials, ASTM E2862-1
Anon. 2015. Standard Practice for Probability of Detection Analysis for â Versus a Data.
American Society for Testing and Materials, ASTM-E3023
Annis, C., 2015. mh1823 R software package, version 4.3.2. Available online:
http://statisticalengineering.com/mh1823/mh1823-algorithms.html
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Post impact damage evaluation of high velocity impacted E-glass composites
using Immersion Ultrasonic C-scan technique
M. Srinivasa Raoa, T. Sreekantha Reddyb, P. Rama Subba Reddyb*,
V. Madhub, B. V. S. R. Murthya
a
Advanced Systems Laboratory, Hyderabad – 500 058
Defence Metallurgical Research Laboratory, Kanchanbagh, Hyderabad, India - 500 058
b
***************************************************************************
***
Abstract
E-glass composites are versatile materials for aerospace and armour applications due to their high
specific strength, better energy absorption and low cost. The present study has been attempted for quantitative
determination of damage area through thickness in E-glass composite laminates using immersion type ultrasonic
C-scan technique. Two different laminates namely E-glass/phenolic and E-glass/epoxy were prepared through
hot pressing method with thickness of 10mm and 25mm. The laminates were subjected to ballistic impact
against 7.62 x 39 mm projectile with strike velocity of 720±10m/s and their energy absorption is determined.
Extent of damage due to ballistic impact is determined quantitatively by measuring the ultrasonic time of flight
of the defect echo through the thickness. It is observed that the extent of damage increases from entry point of
projectile to exit point along its path of the projectile. It is also observed that E-glass/epoxy shows lesser damage
area than E-glass/phenolic laminates for both the thicknesses. Laminates which has undergone more damage
area due to ballistic impact has shown higher energy absorption. The method described in the present paper is
highly useful for understanding the failure behavior of composites during high velocity impact and the data can
be used for designing the improved composites for aerospace and ballistic applications.
Key words: E-glass composites, ballistic impact, energy absorption, ultrasonic C-scan
analysis, damage area.
1. Introduction
Fiber reinforced laminated composite materials are widely used in armoured combat
systems due to their high specific strength and better energy absorption. Composite armours
are usually made up of continuous fibre reinforced either in thermoset or thermoplastic
matrix systems. During service, these composite armors are subjected high velocity
projectile or splinters impact. Hence it is expected that the composites should have an
adequate strength and toughness to withstand and absorb high velocity impact energies.
However some defects in composites laminates are expected during fabrication and also
usage. The defects like delaminations are formed through impact damage. Delaminations
occur due low transverse and interlaminar shear strength of fibre reinforced composite
laminates. The depth which the delaminations are produced varies and it is important to
determine this depth by using non-destructive test (NDT). An aspect of particular concern
for NDT community is the detection and sizing of impact damage in composites through
thickness of the laminates.
____________________________________________________________________________
*Corresponding author, E- mail: rsreddy@dmrl.drdo.in Ph: +91-40-24588011
Various NDT techniques were reported in literature to characterize the delamination
and damage areas in composite laminates. Among the different techniques Ultrasonic C-scan
analysis is most promising technique for detection and determine the damage through
thickness immersion type testing is time of flight and amplitude based imaging technique.
Many researchers have reported on post impact damage of laminate composites using
various NDT techniques. Nayak et al. [2] were used immersion type ultrasonic C-scan
technique to estimate the internal damage of ballistically impacted thermoplastic and
thermoset based aramid composites. They observed that in the similar range of impact
velocity the damage area is higher in composites made from thermoplastic resin as compared
to thermoset resin. Samant et al. [3] also investigated the core damage area of projectile
impacted Kevlar-polypropylene composites using the immersion type ultrasonic C-scan
method. They found that core damage area depends on the striking velocity of the projectile.
Hosur et al. [5] studied the effect of impact energy (3-30J) on delamination area of
carbon/epoxy composites through ultrasonic imaging. They concluded that in addition to
impact energy, mass and velocity of impactor, different lay-ups and thickness of composite
also affect the delamination area. However all these studies were highlighted on determining
the damage area on surface of the laminates. It is very important to determine the damage
area of the laminates across the thickness. Therefore the objective of the present paper is to
find out the post impact delamination area of the laminates across the thickness. Two
important ballistic grade composite laminates viz. E-glass/epoxy and E-glass/phenolic having
two different thicknesses of 10mm and 25 mm. These laminates were initially impacted by
7.62 mm mild steel projectile at 720±10 m/sec velocity to induce the damage.
2. Experimental details
2.1 Materials & fabrication of laminates
Diglycidyl ether of bisphenol-A (DGEBA) epoxy resin (LY556) with hardener Diethyle
toluene diamine (DETDA) (HY5200) supplied by M/s. Huntsman Chemicals were used in
present studies. Commercially available phenolic resin (Resole grade) and E-glass woven
roving having 0.25mm thickness and 360 GSM with warp and weft of 55x50 per 10 cm width
was used as reinforcement. E-glass/epoxy and E-glass/phenolic composite laminates of sizes
350 mm x 350 mm were made through hand layup technique followed by hot pressing under
hydraulic pressure details of the laminate curing is given in elsewhere [1]. Thickness of
fabricated composite laminates was controlled at 10±0.2mm and 25±0.2mm. Specimens
were cut in to the dimensions of 300x300 mm for impact tests.
2.2 Ballistic impact test
Ballistic impact tests were carried out using 7.62 x 39 mm mild steel core service
ammunition. The projectile was fired from AK-47 rifle at a distance of 10 m from target at
normal impact angle. The strike velocity of the projectile was 720±10 m s-1. Targets in size of
300 x 300 mm were cut using diamond wheel cutting machine. Minimum three specimens
were prepared for each thickness and velocity. Striking and residual velocity of the projectile
was measured and absorbed energy of laminate was calculated using equation (1) given
below.
Where,
Eabs- Energy absorbed by the laminate (J)
Vi- Striking velocity (m s-1)
Vr- Residual velocity (m s-1)
m - Mass of the projectile (g)
2.3 Post impact damage evaluation using Ultrasonic C-Scan analysis
Fig. 1: Ultrasonic Immersion Type Test Facility
Post impact damage analysis of the laminates was carried out using Ultrasonic Cscan evaluation technique. Fig.1. Shows the immersion type ultrasonic C-scanning setup
which is used for automated data acquisition and imaging. The setup consists of a three-axis
fixture, an immersion tank, add-on cards, software and transducers. The transducer is
connected to an ultrasonic board that acts as the pulser, receiver and digitizer of the ultrasonic
waveform. Add-on cards control the mechanical motion and pulser/receiver parameters. The
software Acq-scan is used to program the scan cycle acquire/display data, carryout data
processing and produce a colour-coded display. The transducers used are of the immersion
type and the frequency used was 1MHz. A perspex tank of size 1m × 1m is used for keeping
the laminate in water couplent. A square region of 200 mm x 200 mm around the impact
point on the laminate was selected for testing which covers the total damage area of the
laminates. The pulse parameters used in the present tests are pulse width; 3.1µs, pulse
amplitude; 200V, filter; 0.5 - 6MHz, Sampling rate; 100MS/s and record length; 5000
samples. For each scan an optimum gain was used which varied from 10dB to 15dB.
Determination of damage area across the thickness is described below.
Transducer emits an ultrasonic wave which is then reflected by the interfaces of the
material and by defects. At each reflection an echo is observed, the position in the material
being a function of time. Selecting a time window is therefore equivalent to selecting a slice
of the medium. Hence it is possible to observe the material section by section by selecting the
appropriate time window. Depth wise information can be obtained by recording the position
of defect echoes in time on the ultrasonic A-Scan and using this time-of-flight information
rather than the back-wall echo amplitude to construct the C-Scan. This method is to record
the time-of-flight to the first return echo in the A-Scan as a function of position on the
sample. This first return echo will normally represent the back wall in the case of a perfect
laminate, or the first delamination if the panel has been damaged.
The scan controller moves the ultrasonic probe over the laminate in a regular fashion,
and simultaneously the ultrasonic unit provides an analogue output proportional to the timeof-flight to the first reflection in the A-Scan. This analogue output is digitized by a
microcomputer. The digitized scan lines collected in this manner are stored on hard disc and
are further processed and produce false-colour image of the component suitably scaled with
the different thickness ranges being assigned different colours. The image is effectively a
plain view of the component, with depth mapping of any internal features. A two-dimentional
ultrasonic C-scan performed section by section in pulse-echo-mode, and at different positions
of 10 mm and 25 mm thick E-glass composite laminate relative to the impact point. The
ensuing delamination damage was determined by ultrasonic C-Scans using the pulse-echo
immersion method for both the laminates in layer wise distribution. By setting a gate on the
backwall echo, information about the attenuation of the initial pulse after it has passed
through the thickness of the laminate is collected thereby giving the image of the projected
damage of the laminate. Layer wise distribution was obtained by successive time delay from
the front wall to the back wall echo covering each interface.
3. Results & Discussion
Energy absorbed during ballistic impact of E-glass/phenolic and E-glass/epoxy composites is
calculated. Fig. 2 shows absorbed energy against thickness for both composites. It is observed
that at 10 mm thickness, both E-glass/phenolic and E-glass/epoxy composite laminates
behaved similarly. However, at higher thickness i.e. 25 mm, E-glass/phenolic absorbed
higher energy than E-glass/epoxy laminate.
Fig.2: Absorbed energy of both composites as a function of thickness
The Fig.3 shows the visual impression of the laminate and the corresponding A-scan & CScan image. Fig. 3 b1&b2 indicate A-scan plots for undamaged and damaged portion of the
laminate respectively. From Fig.3b2 it is observed multiple defect echos which indicate
multiple delaminations through the thickness. In the present study it has been possible to
adopt layerwise scanning by locating the depth using time delay to determine the damage at
required depth from the top surface of the laminate to assess the damage area across the
thickness. The C-Scan image (Fig.3c) is consist of distinct colour regions indicating varying
order of severity of damage. The central white area represents the most severely damaged
region and is reported as completely damage area. The blue region represents the least
damage/unaffected area and the orange is in between in terms of severity of damage.
Fig. 3: Ballistically impacted composite(a) it’s typical A-scan(b1&b2) and C-scan image(c)
Fig.4 shows C-scan images of 10 mm thickness E-glass epoxy and E-glass
phenolic laminates at different depth from the top surface of the laminate. From the figure it
is clear that damage area is increased progressively with the increase in depth of laminate
upto 5mm thickness and became constant at rear layers. The trend is found to be same for Eglass phenolic laminate also. However, the damage area of E-glass phenolic laminate is
higher as compare to E-glass epoxy laminate at all depth levels. Damage area of both the
laminates against depth is plotted in Fig.5
Fig.4: Layer wise C-scan images of E-glass epoxy & E-glass phenolic laminates of 10 mm thickness at different
depth (damage areas are given in the parenthesis)
Fig.5: Damage areas of 10 mm thickness E-glass epoxy and E-glass phenolic laminates at different depth
Fig.6 shows C-scan images of 25 mm thickness E-glass epoxy and E-glass phenolic
laminates at different depth from the impact surface of the laminate. Damage area is found to
be similar for both the laminates at initial layers and later it increased exponentially from
~100 mm2 to 360 mm2 in case of E-glass epoxy laminate and upto 500 mm2 for E-glass
phenolic laminate as progressed to the rear layers. In this instance also damage area of Eglass phenolic laminate is determined to be more as that of E-glass epoxy laminate. This can
be attributed to the weak inter laminar stregth of the E-glass phenolic laminate which allowed
to radial delamination of layers and in turn absorbed more energy during impact. Comparison
of damage areas of both the laminates of 25 mm thickness at different depth levels is given in
Fig.7
Fig.6: Layer wise C-scan images of E-glass epoxy & E-glass phenolic laminates of 25mm thickness at different
depth levels (damage areas are given in the brackets)
Fig.7: Damage areas of 25 mm thickness E-glass/epoxy and E-glass/phenolic laminates at different depth
CONCLUSIONS
The time of flight and amplitude based ultrasonic immersion C-scan technique was used to
detect and determine the damage area on post ballistic impacted E-glass composite laminates.
Some impartant inferences that can be drawn from the study are:
1. The present technique is capable to determine the depth wise damage of post impacted
laminates which is useful input for the laminate designers.
2. Damage area of E-glass phenolic laminate is found to be higher as compare to E-glass
epoxy laminate for both thicknesses.
3. The delaimanation area increases with increase of laminate thickness.
Acknowledgements
Authors gratefully acknowledge Dr. Tessy Thomos, Director, Advanced Systems Laboratory (ASL), Hyderabad
for her encouragement to publish this work.
REFERENCES
1.
2.
3.
4.
5.
6.
P Rama Subba Reddy, T Sreekantha Reddy, V Madhu, A K Gogia and K Venkateswara Rao,
‘Behaviour of E-glass composite laminates under ballistic impact’, Materials and design 84, pp 79-86,
2015.
N Nayak, A Banerjee, and D Datta, ‘Ultrasonic Assessment of Bullet Inflicted Damage in Aramid
Laminated Composites’, Defence Science Journal, Vol 62, No.3, pp 153-158, 2012.
S S Samant, A Joshi and K K S Mer, ‘Ultrasonic Imaging of Ballistically Impacted Composite
Armour’, Int. J. on Theo. and Appl. Res. in Mech. Engg., Vol 2 No.4, pp79-84, 2013.
D Datta, S Samanta and P Maity, ‘Automated imaging of composite specimens’, Journal of Non
destructive testing and evaluation, Vol 3, pp 21-27, 2004.
M V Hosur, C R L Murthy, T S Ramamurthy and Anita Shet, ‘Estimation of impact-induced damage
in CFRP laminates through ultrasonic imaging’, NDT&E International, Vol.31, No.5, pp 359-374,
1998.
C J Jih, and C T Sun, ‘ Prediction of delamination in composite laminates subjected to low velocity
impact’, Journal of Composite Materials, 27 No.7, pp 684-701, 1993.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Model Assisted Probability of Detection for Lognormally Distributed
Defects
Vamsi Krishna Rentala 1, Phani Mylavarapu 2, K.Gopinath 2, J.P. Gautam1, Vikas Kumar2
1
School of Engineering Science and Technology, University of Hyderabad, India
Phone: +91 9440598223; e-mail: vamsikrishnarentala@uohyd.ac.in, jaiprakashgautam@gmail.com
2
Defence Metallurgical Research Lab, Hyderabad, India; E-mail: phanimylavarapu@dmrl.drdo.in
Abstract
Fatigue cracks originating from in-service aero-engine turbine discs are usually known to follow log-normal
distribution. Non-destructive testing techniques used for detecting fatigue cracks produce response either as HIT
(detected)/ MISS (undetected) or “a” (crack size) Vs. “â” (crack response) depending on the type of the NDT
techniques used. Under the fracture mechanics based damage tolerance methodology widely used in aero-engine
industry, thorough understanding of the reliability of NDT techniques is as equally important as identifying
cracks. In general, the reliability of an NDT technique is usually estimated by plotting Probability of Detection
(POD) curves. POD is a function of crack parameters such as size, shape and orientation along with type of
material.POD of any NDT technique can be estimated by using the standard test procedure and methods
mentioned in MIL-HDBK 1823A. However, as the experimental estimation of POD involves more laborious and
time consuming process, model assisted POD (MAPOD) approaches are currently in practice. In this study,
MAPOD approaches were demonstrated for volumetric cracks using ultrasonic testing. A commercial grade
Titanium alloy Ti-6Al-4V cylindrical block (50 mm x15 mm) with a cylindrical defect (0.5mm x 5 mm) at the
centre was initially inspected with ultrasonic testing in A-scan mode and the corresponding amplitude vs. time
data of the block was analyzed. Ultrasonic wave interaction with a cylindrical defect was simulated as a 2-D
axisymmetric model using COMSOL Multiphysics software, which was further validated with the help of
experimental data. As fatigue cracks usually follow log-normal distribution, distribution of crack sizes for POD
curve generation in MAPOD approach was also assumed to be log-normal in nature. Hence, in this study, both
‘a’ and ‘â’ follow log normal distribution. However, ‘log (a)’ or log (â)’ follow normal distribution. Further, loglog linear regression with normal distribution was performed and correspondingly mean and standard deviation
of the distribution was obtained. Further, these mean and standard deviation were log-transformed for obtaining
scale and location parameters of log normal distribution. Furthermore, using these scale and location parameters,
CDF of log-normal distribution was plotted resulting in a POD curve. Moreover, 95% confidence bounds of the
POD curve were also plotted and a flaw size with 90 % probability and 95 % confidence limit (a 90/95) value was
obtained.
Keywords: Ultrasonic testing, Probability of Detection, Flat Bottom Hole, MAPOD, bootstrap confidence
intervals.
1. Introduction
Life revision program for aero-engines are usually classified as either based on safe life
approach or on damage tolerance methodology. While safe life approach is concerned with
defect detection using non-destructive techniques (NDT), in damage tolerance (DT), NDT is
used to monitor the crack propagation in the component. Therefore, knowledge of initial crack
size detected using NDT techniques is of significant importance in DT approach. However, as
estimation of initial crack size using NDT techniques is probabilistic rather than deterministic
in nature, POD of NDT techniques is very vital information. Experimental estimation of POD
data is usually generated as per MIL-HDBK 1823A (2009) standard [1] which describes the
procedure to be adopted, sample shape, size and defect classification to be considered. In
addition, POD is also a factor of material, crack type, location, etc. Also, POD data generation
involves generating sensitivity [2] and reliability data. These kind of requirements put
enormous limitations on the feasibility for performing POD. In order to reduce the complexity
in experimental generation of POD, alternative source of POD generation using numerical
models also called as model assisted POD
(MAPOD) [3-7] are also in vogue today.
Figure 1 shows the flowchart of MAPOD. In
the current study, MAPOD has been adopted
for estimation of reliability in detection of
bulk defects using ultrasonic testing.
Different statistical functions have been
suggested for calculating POD and confidence
curves depending on the type of the inspection
data [8]. Two models are often used for the
POD analysis of inspection data depending on
whether the outcome of the system is
“hit/miss” or “â vs. a”. Berens and Hovey
examined various methods of modelling NDT
data to determine POD curves [9]. They
concluded that the log-logistic (log-odds)
model and log-normal model are the most
consistent distributions for determining a
POD curve as a function of defect size, ai.
However, in the current study, only lognormal distribution was adapted for obtaining Figure 1: Flowchart of Model assisted POD
the POD curve as the NDT response is â vs. a
type.
The cumulative log-normal distribution is expressed as:
Pi=1-Q(zi)
(1)
For
Zi=
ln(𝑎𝑖 )−𝜇
𝜎
where, μ =
(2)
ln(𝑦𝑡ℎ )−β0
β1
𝛿
and σ = β1
(3)
where, Q(z) is the standard normal survivor function, zi is the standard normal variant, and μ
and σ are the mean and standard deviation of the normal distribution function. yth is the value
of the signal â at the decision threshold (adec). Even though two methods are available for the
estimation of location and scale parameters, the method of Maximum Likelihood Estimators
(MLE) has gained significant importance and hence used in the current study.
2. Experimental Ultrasonic Testing
A Ti-6Al-4V cylindrical block with a diameter of 50 mm and
height of 15 mm was considered for POD generation. A flat bottom
hole (FBH) of diameter 0.5 mm and a height of 5 mm present in
the Ti-6Al-4V block as shown in Figure 2 was used for bulk defect
representation while generating ultrasonic signals. Ultrasonic
signals from the FBH were captured using a V110 (M/s
Panametrics Inc, USA) 5 MHz ultrasonic longitudinal transducer.
Signals were captured using an Olympus make 5900 PulserReceiver at a sampling frequency of 50 MHz. Experimental Ascans from the block are captured. In addition to the front and back
wall echo, echo from the FBH was observed at a time of 3 µs. Figure 2:Representative3From Table 1, it is observed that the defect echo as a % of full D model of Ti-6Al-4V
block with FBH
screen height was calculated to be 2.08%.
3. Mathematical Model
3.1 Development and validation of Mathematical model
As the current study involves with a flat
bottom hole present in a cylindrical block, a 2D axisymmetric model of wave propagation
was developed for simulations. Transducer is
assumed to be a line source in the 2-D model
axisymmetry model. Fundamental details
about the basis of selecting analytical model of
signal, its frequency content, selection of time
step, selection of mesh size and the study
selected for performing simulation was
discussed by Phani et al [10].
A 2D-axisymmetric model of Ti-6Al-4V
cylindrical block was created with dimensions Figure 3: (a) Input waveform for 5MHz
of the block as well as FBH similar to that of (b) FFT of input waveform.
used
during
experimental
ultrasonic
inspection. FEM modelling and simulation has been performed using COMSOL Multiphysics
Package ver.4.3a using structural mechanics module. As the 5 MHz transducer probe diameter
is 0.25 inch, the line source in the model was also created for a length of 6.35 mm. Frequency
content for 5MHz probe was incorporated with an interpolation function from the 6.35 mm
line source. Input signal and its frequency content are as shown in Figure 3(a) and 3(b).The
model was meshed with an element size of λ/8 (i.e. 0.15 mm). Time step for convergence was
maintained at 3.6e-9 s as per CFL criterion. Further, the FEM model was solved for a total
duration of 7 µs with a step size of 20 ns. The response signal from the FBH was observed at
3µs, similar to that of experimental A-scan signal. Comparing the% full screen height of
defect echo from the experiment as well as numerical model matched, it was observed the
results match at around 2% as shown in Table 1. Hence, the 2D-axisymmetry model was
validated and is considered for further simulations for the estimation of POD curves.
Table 1: % full screen height of defect echo amplitude for both experiment and 2-D
axisymmetric model
Model
Experiment
Numerical
Model
Line source width,
mm
Front wall echo
amplitude
Defect echo
amplitude
6.35
0.90
0.02
% full screen
height of defect
echo amplitude
2.08
6.35
0.7
0.01
2.27
4. Results and Discussion
4.1. Generation of Random Crack sizes
Unlike in the case of experimental POD curve generation, distribution of crack sizes has to be
assumed for MAPOD and this is usually dependent on the distribution of service induced
fatigue cracks. In general, fatigue cracks originating from in-service aero-engine turbine disks
follow log-normal distribution [2,11].Studies performed by the current authors on distribution
of naturally initiated fatigue cracks also reveal a lognormal distribution [2].Figure 4 shows the
histogram of crack sizes generated from lognormal distribution along with its probability
density function for the study mentioned in [2]. Hence, in the current study, all the crack sizes
are generated in lognormal distribution and the POD of all these cracks sizes was estimated
for Ultrasonic testing.
4.1.1. Random crack sizes by lognormal distribution
The probability density
function of the lognormal
distribution is: f(xǀμ,σ)
=
𝟏
√(𝟐𝝅)𝝈𝒙
𝒆
−(𝒍𝒏⁡(𝒙)−𝝁)𝟐
𝟐𝝈𝟐
(4)Where, μ is the location
parameter
of
the
distribution and σ is scale
parameter. As the material
selected for ultrasonic
inspection with a 5 MHz
transducer probe is Ti-6AlFigure 4: Fatigue cracks in Ni based superalloy following
4V,
the
longitudinal
lognormal distribution [2]
velocity of the ultrasonic
wave is approx.6000 m/s
and the corresponding wavelength is λ=V/f=1.2mm.Therefore, approximately any cracks
above λ/2=0.6mm would be detectable using this technique. However, every crack either
above or below λ/2 possesses a probability to be detected. Hence, in this study MAPOD is
performed by considering crack sizes less than 0.6 mm and more than 0.6 mm. For example, a
crack size of less than 0.6 mm should have lesser probability of hit compared to a crack size
of greater than 0.6 mm. In order to generate this range of crack sizes following lognormal
distribution, lognormal parameters such as location is assumed to be -0.2, scale is assumed to
be 0.2and a threshold value of 0 is assumed. This has resulted in generating 20 crack sizes in
the range of 0.4-1.34 mm following lognormal distribution. Figure 8 shows the histogram of
crack sizes generated from lognormal distribution along with its probability density function.
4.2. POD from lognormally distributed crack sizes
The validated FEM model for ultrasonic inspection has been parameterized with random
crack sizes generated using lognormal distribution. From the numerically obtained A-scans
automatic defect echo identification has been carried out by applying a Hilbert transform to
the signal in Matlab program. The actual FBH sizes are considered as 'a' whereas the
ultrasonic signal amplitude from the defect echo is considered as 'â'. Statistical analysis of
data for POD curve generation involves in estimation of best linear fit between a vs. â. From
Figure 5(a-d), it is observed that log-log scale of a vs. â data achieved the best linear fit as the
R2 value of log(â)vs. log(a) is 99.6% whereas for the â vs. a, â vs. log (a) and log(â) vs. a plots
the R2 values are 96.1%, 86% and 98.1%, respectively. Hence, 'log (â) vs. log(a)' was
considered for regression analysis. Regression of data is usually carried out using an ordinary
least square (OLS) method. Annis et al [12] have commented about the appropriateness of the
POD curve using regression parameters estimated by OLS method. In their study, they have
concluded that the shape of POD curve changes with or without the influence of censoring
data. However, as POD curve is not supposed to be dependent on the censoring process,
maximum likelihood estimation (MLE) method has been recommended. Using MLE, it was
observed by Annis et al that the POD curve shape stayed the same either by considering
censoring or without. However, in this study, MLE is considered with censored data to
perform regression analysis .Hence censoring of the data was performed in Minitab statistical
software using "Regression with Life Data" condition. Further, censoring of the data was
performed for crack sizes of 0.5 mm and below. Table 2 shows a vs. â data for 11 out of 20
cracks for illustration purpose. From Table 2, it can be observed that the crack size 0.4 mm
was censored as the crack size is below 0.5 mm. ‘C’ is indicated for the censored data and ‘S’
is indicated for uncensored data
Table 1: Censored data indicating 'C' for censored and 'S' for uncensored
a, mm 0.82 0.57
â
censor
7.12 3.97
S
S
0.73
0.58
0.70
0.84
0.57
0.97
0.84
0.70
0.40
6.04
4.16
5.44
7.59
3.97
9.71
7.59
5.44
2.34
S
S
S
S
S
S
S
S
C
(a)
(b)
(c)
(d)
Figure 5: Linear curve fitting of (a) â vs. a (b) log(â) vs. log(a) (c) â vs. log(a) and
(d) log(â) vs. a
4.2.1. Regression with censoring data
Regression of the log (a) vs. log (â) data has to be performed with censoring option in
order to calculate the mean and standard deviation and further to generate the POD curve.
This regression of the censoring data can be performed using “Regression with life data”
option available in Minitab 17 statistical software. Figure 6 shows result of the regression
performed between log (â) and log(a). From Figure 6, it can be observed that 1 censored value
which is below 0.5 mm exists in the 20 different crack sizes. Also from Figure 6, the
coefficient of intercept also called as intercept and coefficient of log (a) value (slope) of the
regression line were obtained. In addition, error of log (a) was also calculated. Using these
regression parameters (intercept and slope), mean and standard deviation were calculated as
shown below.
𝑙𝑜𝑔â𝑑𝑒𝑐 −⁡𝛽0
Mean, m =
(5)
𝛽
1
𝛿
Standard Deviation, σ = 𝛽
(6)
1
where, â𝑑𝑒𝑐 = Signal Response decision Threshold value for 𝑎𝑑𝑒𝑐 = 0.5 mm
𝛽0= Intercept of the regression line
𝛽1 = Slope of the regression line
Ʃ(log(𝑎)−𝑚)2
𝛿 = Error of log(a) = √
(7)
𝑛
n= number of cracks or FBH sizes
With these mean and standard deviations, the cumulative distribution function (CDF) of
log (a) was plotted. Further, the 95 % confidence curve was also plotted by estimating the 95
% confidence limits of the mean value using the below equation
𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑⁡𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛
Lower Confidence Limit (LCL) of mean = mean – (1.96⁡ ∗ ⁡ (
√𝑛
))
(8)
Figure 7 shows the
CDF of log (a), which is
actually the POD vs.
log(a) curve. From Figure
7, the log(a)90/95 value can
be observed as -0.216
mm, and the exponential
of this
log(a)90/95 results in a90/95
value as 0.806 mm.
Moreover, as the mean
and standard deviation of
the normal distribution is
exactly equal to the
location and scale of the
lognormal
distribution
[13], the CDF of (a) was
also plotted with the same
mean
and
standard
deviation values of log(a). Figure 6: Regression Table
Figure 8 shows the POD
vs. a curve of 20 lognormally distributed FBH sizes. From Figure 8, it is observed that, the
a90/95 value (0.806 mm) was exactly equal to the a90/95 value obtained from the POD vs. log(a)
curve. Apart from a90/95 value, the a50 and a90 values were also found to be 0.5 mm and 0.707
mm, respectively. In addition, it can also be observed that the a50 value which is equal to 0.5
mm is actually the same as assumed decision threshold value. Fahr et al have commented that
the validity of the POD curve plotted can be ascertained when the observed a50 value equals to
the decision threshold value selected for censoring. As the two values i.e., decision threshold
and a50 value match in this study, it can be concluded that the POD curve is plotted by
following the exact statistical procedures adopted by western researchers.
Figure 7: POD vs. log(a) for 20 lognormally distributed crack sizes indicating
a90/95 = 0.806 mm
Figure 8: POD vs. flaw size (a) for 20 lognormally distributed crack sizes
indicating a50, a90 and a90/95 as 0.5 mm,0.707 mm and 0.806 mm respectively.
5. Conclusions
Due to the importance of knowledge of smallest crack size detected using NDT techniques
in damage tolerant methodology of aero-engines, POD has gained significant importance.
Challenges in experimental determination of POD have led several researchers to explore the
possibility of MAPOD. In this study, a MAPOD approach has been developed to estimate the
POD curves of volumetric defects such as flat bottom holes (calibration reflector) using
ultrasonic testing. Even though elliptical cracks are more possible type of volumetric defects,
modelling approach would be the same irrespective of whether it is a flat bottom hole or an
elliptical crack. Hence, the methodology to carry out MAPOD of ultrasonic testing for
volumetric defects has been demonstrated. The following are some of the noteworthy
conclusions from the work.
1. Ultrasonic signal response from a symmetrical defect can be successfully estimated using
a 2-D axisymmetric FEM model. Both the % defect echo amplitudes as observed from
model as well as experiment matched around 2 %.
2. Log (â) vs. Log(a) was found to be the best linear regression fit.
3. Regression parameters from the censored data can be estimated using Maximum
Likelihood Estimators (MLE) method.
4. a50 value is exactly equal to the assumed decision threshold adec value in the case of
regression between natural logarithms of â vs. a.
a90/95 of POD curves obtained 20 lognormally distributed crack sizes is 0.806 mm.
6. Acknowledgements
The authors express their gratitude to Dr. S. V. Kamat, Director, DMRL for the
encouragement provided to publish this work. The funding provided by Defence Research and
Development Organization (DRDO) to carry out the work is acknowledged. Special thanks
are due to Prof. V.V.Haragopal, Osmania University and Dr. Hina Gokhale, Director-DHRD
for their stimulating comments during statistical analysis.
References
1.
Department of Defense handbook, 'Non-destructive evaluation system reliability
assessment', MIL-HDBK-1823A (USA), April 2009.
2.
PhaniMylavarapu, Vamsi Krishna Rentala, M.Sundararaman, Vikas Kumar and
HinaGokhale, 'Sensitivity evaluation of NDT techniques on naturally initiating fatigue
cracks-An experimental approach for a POD framework', DMRL Technical
Report,DRDO-DMRL-NDTG-083,March 2015.
3.
T. A. Gray, F. Amin, and R. B. Thompson, 'Application of Ultrasonic Pod Models',
Technical Paper, Iowa State University, pp:1737-1744.
4.
Jeremy.S.Knopp, Eric Lindgren, Enrique Medina, and Mark P. Blodgett, 'Computational
Methods in Non-destructive Evaluation: A Revolution in Maintenance', Technical
Paper, Air Force Research Laboratory, October 2011.
5.
F. Jenson, E. Iakovleva and C. Reboud, 'Evaluation of Pod Curves Based on Simulation
Results', Technical Paper, French National Research Agency, 2008.
6.
Aacarvalho, R R Silva, L V S Sagrilo and J M Arebello, 'Reliability of the Ultrasonic
Inspection by Numerical Simulation Technique', Insight Vol 53, pp:603-609, November
2011.
7.
J.A. Ogilvy, 'Model for Predicting Ultrasonic Pulse-Echo Probability of Detection',
Journal Paper, NDT& E, Vol 26,pp:19-29, November 1993.
8.
Abbas Fahr, 'Aeronautical Applications of Non-Destructive Testing', Text Book,
pp:403-405, 2014.
9.
A.Fahr, D.S.Forsyth and M.Bullock, 'NDI Techniques for Damage Tolerance-Based
Life Prediction of Aero-Engine Turbine Disks', NRC Report:LTR-ST-1962,February
1994.
10.
Phani Surya KiranMylavarapu and Srivathsa Boddapati, 'A Numerical study on particle
scattering of ultrasonic waves in syntactic foams-I', DMRL Technical Report, DRDODMRL-NDTG-043-2013, February 2013.
11.
Hina Gokhale, Shahnawaz, Raghu and Vikaskumar, 'Probabilistic fatigue life
assessment of gun barrels', DMRL Technical Report, DRDO-DMRL-034-December
2012.
12.
Charles Annis& Luca Gandossi, 'Influence of Sample Size and Other Factors on
Hit/Miss Probability of Detection Curves', ENIQ TGR technical document, ENIQ
Report nr. 47, January 2012.
13.
William Q.Meeker and L.A.Escobar, 'Statistical methods for reliability data', Wiley,
Hoboken, New Jersey, 1998.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
A Scanning Spatial-Wavenumber Filter based On-Line Damage Imaging
Method of Composite Structure
Lei QIU, Shenfang YUAN, Yuanqiang Ren
Research Center of Structural Health Monitoring and Prognosis, State Key Lab of Mechanics and Control of
Mechanical Structures, Nanjing University of Aeronautics and Astronautics; Nanjing, China
E-mail: lei.qiu@nuaa.edu.cn, ysf@nuaa.edu.cn, renyuanqiang@nuaa.edu.cn
Abstract
Lamb wave Spatial-Wavenumber Filter (SWF) is gradually applied to Nondestructive Testing (NDT) of
composite structures in recent years because it is an effective approach to distinguish wave propagating direction
and mode. Nevertheless, to realize on-line damage imaging of composite structures by using SWF, the problem
is on how to realize on-line spatial-wavenumber filtering of Lamb waves when the wavenumber response cannot
be measured or modelled. This paper proposes an on-line damage imaging method based on a new Scanning
Spatial-Wavenumber Filter (SSWF) and PieZoelectric Transducer (PZT) 2-D cruciform array for composite
structures. With this method, a 2-D cruciform array constructed by two linear PZT arrays is placed on the
composite structure permanently for on-line continuously spatial sampling of the Lamb wave damage scattering
signal. For each linear PZT array, a SSWF is designed, which does not rely on any modelled or measured
wavenumber response but scans and filters the wavenumber of the damage scattering signal at a designed
wavenumber bandwidth to obtain a wavenumber-time image. A 2-D cruciform array based damage localization
method is proposed to get the final angle-distance image of the damage which can be localized without blind
angle. The method is validated on an aircraft composite fuel tank of variable panel thickness. The validation
results show that the damage direction estimation error is less than 2° and the damage distance estimation error is
around 20mm in the monitoring area of nearly 600 mm × 300 mm. It indicates an acceptable performance of the
on-line damage imaging method based on the SSWF for a complex composite structure
Keywords: Structural health monitoring, composite structure, damage imaging, Lamb wave, scanning spatialwavenumber filter, 2-D cruciform array
1. Introduction
In recent decades, the damage imaging methodology has been widely studied [1-2] such as
delay-and-sum imaging method, time reversal focusing method, ultrasonic phased array It
utilized a large number of actuator-sensor channels from a network of PieZoelectric
Transducer (PZT) to map the structure that was interrogated based on the measurement of
damage reflections and induced features differences, producing a visual indication of damage
location and size. It has advantages of high signal-to-noise ratio, high damage sensitivity and
large scale structure monitoring. This paper studied a damage imaging method of composite
structures based on a cruciform array and Spatial Wavenumber-Filtering (SWF) technique.
SWF for damage estimation has been gradually studied in recent years. Compared
with the time domain and frequency domain analysis of Lamb wave signals, the wavenumber
domain analysis is an effective approach to distinguish wave propagating direction and
various wave modes. Michaels et al.[3] used a frequency-wavenumber filtering method to
analyze the direction and wave modes of damage-scattering signal on an aluminium plate.
Sohn et al. [4] created a standing wave filter which was a signal processing technique in
frequency-wavenumber domain to isolate only the standing wave components on a composite
plate for delamination or disbond evaluation. Rogge and Leckey [5] built the relationship
between wavenumber and wave spatial location using SWF and local wavenumber domain
analysis to estimate the delamination size and depth of a composite laminate. Yu et al.[6]
presented short time-spatial-frequency-wavenumber analysis for obtaining a wavenumber
spectrum at a selected frequency and time to decompose Lamb wave modes by SWF. Purekar
1
et al. [7] studied a model-dependent SWF based linear PZTs phased array imaging method.
They also studied the method further on a composite plate based on finite element modelling
[8]. In the above research, the wavenumber of Lamb wave signal was measured by a scanning
laser Doppler vibrometer of high spatial resolution or obtained by modelling. Thus, these
method are difficult to be applied to on-line damage monitoring.
Considering on-line damage imaging by taking the advantage of SWF, a new damage
imaging method of composite structure based on a Scanning Spatial-Wavenumber Filter
(SSWF) and PZT 2-D cruciform array is proposed in this paper. The scanning spatialwavenumber filter which does not rely on only modelled or measured wavenumber response
is proposed first. And then, a damage imaging method based on the filter is proposed, and the
corresponding damage localization method of no blind angle is given as well. Finally, the
damage monitoring performance of the method for complex composite structure is validated
on an aircraft composite fuel tank of variable thickness.
2. The Scanning Spatial-Wavenumber Filter
2.1 Lamb Wave Time Domain Sampling and Spatial Sampling
There is a linear PZT array placed on a structure as shown in Figure 1. It consists of M PZTs
and the distance between the centres of each two adjacent PZTs is Δx. The PZTs are
numbered as m=1,2,…,M. A Cartesian coordinate is built on the array. The centre point of the
array is set to be the original point.
There is a damage located at (xa, ya). The direction (angle) and distance of the damage
relative to the linear PZT array are supposed to be θa and la respectively. To obtain the
damage scattering signal, a frequency narrowband excitation signal of central frequency ω is
input to the PZT at the original point to excite Lamb wave of frequency narrowband. When
the Lamb wave propagates to the damage, the damage scattering signal is generated and it can
be acquired by the array. According to some previous studies [9-11], the amplitude of Lamb
wave A0 mode is dominant at low excitation frequency. Thus, the damage scattering signal
can be approximated to be single-mode signal when the excitation frequency is low. The
wavenumber of the damage scattering signal is denoted as ka. It is wavenumber narrowband
and it can be considered to be constructed by two components. The first component is the
wavenumber projecting at the array direction (X-axis projection wavenumber) kx=kacosθa and
the second component is the Y-axis projection wavenumber ky=kasinθa. Figure 2 gives an
example of the damage scattering signal acquired by the linear PZT array. It is shown as a
waterfall plot. The horizontal coordinate of the waterfall plot is sampling time and the
longitudinal coordinate is sampling distance which is corresponding to the location of the
PZTs. Ordinarily speaking, the linear PZT array is seen to be a time domain sampling device.
Each PZT can output a time domain sampling signal as shown in Figure 2 when looking the
waterfall plot from the time direction (blue line). The sampling dots (length) and the sampling
rate are denoted as L and fs respectively.
The linear PZT array can be also regarded as a spatial sampling device to acquire
spatial response of the damage scattering signal at the area covered by the array when looking
the waterfall plot from the distance direction (red line). The spatial sampling rate is 2π/Δx.
The spatial response acquired by the linear PZT array at time tr can be represented as equation
(1), in which, f(xm, tr) is the damage scattering signal acquired by the PZT located at (xm, 0) at
time tr, xm=((2m-1)-M)Δx/2. tr=r/fs and r=1,…,L. f(x, tr) is defined as one spatial sampling of
the damage scattering signal at time tr . Thus, the waterfall plot of figure 2 can be regarded to
be L times of spatial sampling of the damage scattering signal. Based on this point, the
theoretic fundamental of SSWF is given in the next section.
2
f  x, tr  =  f  x1 , tr  , f  x2 , tr  ,..., f  xm , tr  ,..., f  xM , tr 
Time domain
sampling
Waterfall plot of damage scattering signal
f(xm, t) 90
f(x
Y
M,
0.5
-0.5
-1
45
0
0.2
0.4
0.6
0.8
1
1
1 2 3
θa
…
Δx
Normalized amplitude
-45
0
-0.5
-1
-90
M
f(xm, t)
0
0.5
…
Linear PZTs array
X (mm)
Time (ms)
Wavenumber
t)
0
…
Normalized amplitude
1
Acousitc source
(xa, ya)
(1)
-45
0
45
90
Spatial sampling
f(x, t ) -90
r
x (mm)
X
f(x1, t)
0
0.2
0.4
0.6
0.8
1
Time (ms)
o
Time tr
Fig.1. Illustration of Lamb wave wavenumber.
Fig.2. Lamb wave time domain sampling and spatial sampling.
2.2 The Design of Scanning Spatial-Wavenumber Filter
For one time spatial sampling of the damage scattering signal of wavenumber narrowband at
low excitation frequency, f(xm, tr) can be represented as equation (2), in which u(tr) is the
normalized amplitude of the damage scattering signal. La and X m represent the distance
vector of la and xm respectively. Based on Fraunhofer approximation [12], equation (2) can be
approximated to be equation (3), in which, Lˆ a denotes the unit direction vector of the La . In
far-field situation, the damage scattering signal can be regarded to be a planar wave acquired
by the linear PZT array [12]. Thus, Eq. (3) can be approximated to be equation (4), in which,
A(tr) denotes the amplitude term. By using Fourier Transform, the spatial response as equation
(1) can be transformed to wavenumber response as equation (5). δ is the Dirac function.
f  xm , tr   u(tr )e 
i tr ka La  X m 
f  xm , tr   u (tr )e
itr
e
 ika la ika Lˆ a X m
e
e
 ika
xm 2 ( Lˆ a X m )2
2la
(2)
(3)
f  xm , tr  =A(tr )eika La X m
(4)
F  k , tr   2πA(tr )   (k  ka cosa )
(5)
ˆ
Thus, a SWF can be designed to be equation (6) based on the wavenumber ka [24-25].
In equation (6), θ is a searching direction which can be changed from 0° to 180°. The
wavenumber response of the spatial wavenumber filter can be also obtained by using Fourier
Transform, as shown in equation (7). It indicates that the SWF has the purpose of selectively
passing through the signal of wavenumber k=kacosθ, while rejecting the signal of the other
wavenumbers k≠kacosθ. By applying the SWF to the spatial sampling damage scattering
signal, the filtered wavenumber response can be expressed as equation (8). The symbol ‘’
denotes the convolution operation. It can be noted that if θ=θa, which means that the searching
direction of the SWF is equal to the damage direction, the amplitude of the filtered response
will reach to the maximum value. The damage direction can be estimated correspondingly.
This is the original SWF [7-8].
  x   eik
a
cos x1
, eika cos x2 ,..., eika cos xm ,..., eika cos xM 
3
(6)
xM
Φ  k    eika cos x eikx  2π  k  ka cos 
(7)
H  k , tr    f  x, tr     x    4π2  A(tr )   (k  ka cosa )   (k  ka cos )
(8)
x  x1
However, the wavenumber ka must be obtained by modelling or measuring beforehand
when using the original SWF. It is difficult to be applied to composite structure at current
stage. Therefore to promote the SWF technique to be applied to composite structure, a new
SSWF is proposed.
It can be seen from equation (6) and (7) that the original SWF is designed to search
the damage direction directly based on a linear PZT array. Hence, the wavenumber ka of the
damage scattering signal is needed to design the wavenumber of the SWF, but as mentioned
above, the wavenumber ka can be considered to be two components including the X-axis
projection component and the Y-axis projection component. If the two axis projection
wavenumbers can be both obtained, the damage direction can still be estimated. Thus, a
SSWF can be designed based on a linear PZT array to search the axis projection wavenumber
of the damage scattering not to search the damage direction directly. By using a PZT 2-D
cruciform array, damage direction estimation can be achieved. Based on this idea, the SSWF
is designed to be equation (9), in which, kn is the wavenumber of the SSWF. The wavenumber
response of SSWF is expressed as equation (10).
k  x   eik x , eik x ,..., eik x ,..., eik x 
n 1
n 2
n m
n M
n
(9)
xM
Φkn  k    eikn x eikx  2π  k  kn 
x  x1
(10)
Considering the damage scattering signal is wavenumber narrowband, kn is set to be
scanned in a wavenumber narrowband from k1 to kN, where n=1,…,N. The wavenumber
scanning range is set to be wider than that of the damage scattering signal but it is also limited
by the spatial sampling rate. In this paper, kn=-kmax+(n-1)Δk. It is scanned from -kmax to +kmax.
kmax is the maximum cut off wavenumber and it is π/Δx. Δx is the distance between the centers
of each two adjacent PZTs. The wavenumber scanning interval is denoted as Δk. n is also
denoted as the scanning step. The maximum scanning step is N=(2kmax/Δk)+1.
For a given kn, the amplitude of the filtered response can be calculated by equation
(11). If kn=kacosθa, which means that the wavenumber of the SSWF is equal to the X-axis
projection wavenumber of the damage scattering signal, the amplitude of the filtered response
will reach to the maximum value. Thus, a scanning filtered response vector can be obtained
shown in equation (12). Figure 3 shows an example. The spatial sampling signal shown in
Figure 3(a) comes from the waterfall plot shown in Figure 2 at the time 0.5ms. The
corresponding scanning filtered response is shown in Figure 3(b) which consists of the
filtered response from k1 to kN. The wavenumber corresponding to the maximum value can be
considered to be the X-axis projection wavenumber of the damage scattering signal.
H kn  k , tr    f  x, tr   kn  x    4π2  A(tr )   (k  ka cosa )   (k  kn )
(11)
H  tr   [ H k1 ,..., H kn ,..., H kN ]
(12)
It should be noted that the SSWF is realized numerically as shown in equation (9) and
(11). There is no need to model or measure the wavenumber response of the damage
scattering signal. Thus, it can be on-line applied to composite structure easily.
4
1
Normalized
幅度
归 一 化amplitude
Normalized amplitude
1
0.5
0
-0.5
-1
-90
-45
0
45
0.8
0.6
0.4
0.2
0
-349
90
-175
0
175
349
波 数 (rad/m)
Wavenumber
(rad/s)
(b) Scanning filtered wavenumber response
x (mm)
(a) Spatial sampling signal
Fig.3. Example of the scanning spatial-wavenumber filtering result.
3. The Damage imaging and localization method
3.1 Damage imaging method based on the scanning spatial-wavenumber filter
As discussed in Section 2.2, a SSWF can be designed to filter the damage scattering signal of
one time spatial sampling at time tr. Thus, it can be also designed to filter the damage
scattering signal of L times spatial sampling. This process can be regarded to be a time
scanning process performed from t1 to tL, as shown in Figure 4(a). Based on the time scanning
process, a matrix of the scanning spatial-wavenumber filtering result can be obtained as
represented in equation (13). Each row of the matrix represents the scanning spatialwavenumber filtering result at different wavenumber kn at one time spatial sampling and each
column of the matrix represents the filtering result at different times.
0.6
Time (ms)
(ms)
Time
0.8
0.4
0.6
Time (ms)
Absolute arrival time, tR
0.8
1
0.8
0.6
0.4
0.4
0.2
0.2
0.2
0
PZT1-1
PZT1-2
PZT1-3
PZT1-4
PZT1-5
PZT1-6
PZT1-7
0
(a) Demonstration of damage imaging process
-200
0
200
Wavenumber (rad/m)
Wavenumber
(rad/m)
0
Projection
Wavenumber, kn
(b) Example of damage imaging result
Fig.4. Damage imaging process and an example of wave-number time image.
H L N
 H  t1  


 ... 
  H  tr  


 ... 


 H  t L  
(13)
Finally, a wavenumber-time image can be generated by imaging the matrix H, as
shown in Figure 4(b). In the wavenumber-time image, the wavenumber and the time
corresponding to the point of the highest pixel value can be estimated to be the X-axis
projection wavenumber kn=kacosθa and the absolute arrival time tR of the damage scattering
signal respectively. It should be noted that the absolute arrival time is not the actual time-offlight of the damage scattering signal. If a 2-D cruciform array constructed by two linear PZT
5
arrays is adopted, the X-axis and Y-axis projection wavenumbers can be obtained. Combining
with the two axis projection wavenumbers and the absolute arrival time, the damage
localization of no blind angle can be achieved as given in the next section.
3.2 Damage Localization based on PZT 2-D Cruciform Array
The PZT 2-D cruciform array is constructed by two linear PZT arrays which are numbered as
No.1 and No.2 respectively. They are shown in Figure 5. The centre point of the PZT 2-D
cruciform array is set to be the origin point. X-axis and Y-axis are set to be along with No.1
PZT array and No.2 PZT array respectively. The Lamb wave is excited at the origin point at
time te. When the damage scattering signal is acquired by the PZT 2-D cruciform array, the
axis projection wavenumber and the absolute arrival time (kn1, tR1) of No.1 PZT array and the
(kn2, tR2) of No.2 PZT array can be obtained by using the damage imaging method. Then, the
damage can be localized based on the following method.
Damage
(xa, ya)
Y
Excitation
ka
kn2
la=Cg·ta/2
tR2≈te+ta
θa
o
No.1 PZT array
Damage
(xa, ya)
Y
kn1
X
o
No.2 PZT array
No.1 PZT array
(a) Damage direction estimation
Damage scattering
tR1≈te+ta
X
No.2 PZT array
(b) Damage distance estimation
Fig.5. Schematic diagram of damage localization based on PZT 2-D cruciform array.

kn 2
arctan( ),
kn1

90 ,

k

 a  180  arctan( n 2 ),
kn1

270 ,


kn 2
360 + arctan( k ),

n1
( kn1  0, kn 2  0)
( kn1  0, kn 2  0)
( kn1  0)
(14)
( kn1  0, kn 2  0)
( kn1  0, kn 2  0)

t R1  t R 2
t R 
2

t

t

 a R te

ct
l a  g a

2
(15)
For damage direction estimation as shown in Figure 5(a), the wavenumber kn1 can be
regarded to be the X-axis projection wavenumber of ka and the wavenumber kn2 can be
regarded to be the Y-axis projection wavenumber. They are expressed as kn1=kacosθa and
kn2=kasinθa Based on it, the direction θa of the damage direction relative to the centre point of
the PZT 2-D cruciform array can be calculated by equation (14). It shows that θa can be
calculated from 0° to 360° without any blind angle. For damage distance estimation as shown
in Figure 5(b), the absolute arrival time tR obtained from the wavenumber-time image
6
contains two parts. The first part is the excitation time te which can be determined by the
excitation signal. The second part is the actual time-of-flight ta including the time of Lamb
wave signal propagating from the excitation position to the damage position and the time of
the damage scattering signal propagating back to the PZT 2-D cruciform array. The damage
distance la relative to the centre point of the PZT 2-D cruciform array can be calculated by
equation (15) combing with Lamb wave group velocity cg. The final damage position can be
expressed as xa=lacosθa and ya=lasinθa.
4. Validation on an aircraft composite fuel tank
4.1 Validation Setup
To validate the performance of the damage imaging method on complex composite structure,
an aircraft composite fuel tank is adopted as shown in Figure 6(a). The dimension of the fuel
tank is 600mm×300mm×240mm (length×width×height). The composite panel of the fuel tank
is made of T300/QY8911 carbon fibre and the thickness is variable. The central part of the
composite panel is the thickest part and it consists of 58 stacked layers. The thickness of each
layer is 0.125 mm and the total thickness is 7.25 mm. The thinnest parts are at the two ends
and the thicknesses are both 4.5 mm.
(a) Validation system
(b) The PZT 2-D cruciform array
50mm
B
Damage
No.2 PZT
array 1-1
O
2-1
(-80, 100) E
Ref 1 C
(-40, 150)
F Ref 2
I
(-40, 300)
100mm
Y
2-7 (0, 100) (0, 150) G
H Ref 3
No.1 PZT 1-7
D
(40, 200) (40, 250)
A
array
(80, 0)
(80, 100)
X
100mm
440mm
50mm
50mm
50mm
60mm
(c) The damage positions and the placement of the PZT 2-D cruciform array
Fig.6. Illustration of the validation system performed on the aircraft composite fuel tank.
A PZT 2-D cruciform array is placed on the composite panel as shown in Figure 6(b).
The distance between the centres of each two adjacent PZTs is Δx=9.0 mm. Thus, the
maximum cut off wavenumber of the linear PZT array is kmax=349 rad/m. Another 3 reference
PZTs are used to measure the Lamb wave group velocity. Their positions are labelled as Ref
1, Ref 2 and Ref 3 as shown in Figure 6(c). The 9 damages labelled as A to I are simulated on
the structure and their positions are shown in Figure 6(c). The damage direction is defined
according to counter clockwise direction relative to the positive direction of X-axis. The Lamb
7
wave based SHM system [13] shown in Figure 6(a) is used to excite and receive Lamb wave
signals. The excitation signal is a five-cycle sine burst modulated by Hanning window [14].
The center frequency of the excitation signal is 55 kHz and the excitation amplitude is ±70
volts. The sampling rate is 10 MS/s and the sampling length is 8000 samples including 1000
pre-samples. PZT 2-5 is used to be the Lamb wave actuator for No.1 PZT, and PZT 1-5 is
used to be the actuator for No.2 PZT. The method based on the continuous complex Shannon
wavelet transform is used to measure the group velocity [14]. PZT1-4 is used to excite Lamb
wave. The measuring results by using the 3 reference PZTs are cg-Ref1=1695.5m/s, cgRef2=1813.8m/s and cg-Ref3=1928.2m/s respectively. Finally, the average group velocity
cg=1812.5m/s is obtained and it is applied to the following damage localization.
4.2 Damage imaging validation
The damage E is selected to be an example to show the damage imaging and localization
process. The damage imaging method is applied to the damage scattering signal. The
wavenumber scanning interval is set to be Δk=1 rad/m and the wavenumber scanning range is
from -kmax=-349 rad/m to kmax=349 rad/m. Figure 7 shows the two wavenumber-time images
of the damage E. For damage direction estimation, the axis projection wavenumbers kn1=-87
rad/m and kn2=299 rad/m are obtained from the two wavenumber-time images. The damage
direction is θa=106.3°. For damage distance estimation, the Lamb wave excitation time must
be obtained first. The excitation signal is also acquired by the SHM system. By using the
continuous complex Shannon wavelet transform [14], the envelope of the excitation signal
can be obtained. The time which is corresponding to the maximum value of the envelope is
judged to be the excitation time te=0.1031ms. The absolute arrival time tR1=0.2964ms and
tR2=0.2917ms are obtained based on the damage imaging results shown in Figure 7.
According to equation (15) and the average group velocity cg=1812.5 m/s, the damage
distance is la=173.0 mm. Finally, the damage position is obtained to be (-48.5 mm, 166.1
mm), and the damage localization error is Δl=18.2 mm.
0.8
1
0.8
kn1=-87 rad/m, tR1=0.2964 ms
0.8
0.6
0.4
0.4
0.2
0
-349
0
175
349
0.6
0.4
0.4
0.2
0.2
-175
0.8
0.6
Time (ms)
Time (ms)
0.6
1
kn2=299 rad/m, tR2=0.2917 ms
0
0
-349
Wavenumber (rad/m)
(a) No.1 linear PZT array
0.2
-175
0
175
349
0
Wavenumber (rad/m)
(b) No.2 linear PZT array
Fig7. Damage imaging results of the damage E.
According to the damage imaging and location process discussed above, the damage
localization results of the 9 damages are listed in Table 1. It indicates that the damage
localization results are in accordance with the actual damage. For the damage A, it is in the
blind angle area of No.1 PZT array. For the damage C and F, they are in the blind angle area
of No.2 PZT array but all of them can be localized correctly. For the damage I, the Lamb
wave excited at the array must pass through the thickness variable area of the composite panel
and the corresponding damage scattering signal also must pass through the thickness variable
area. After that, it can be acquired by the PZT 2-D cruciform array. Thus, the distance
estimation error of the damage I is relative larger than the other damages. Totally speaking,
8
the damage direction estimation error is less than 2° and the damage distance estimation error
is around 20mm. Considering that the monitoring area is nearly 600mm×300mm and the
structure is of variable thickness, the damage localization error can be accepted.
Table 1. Damage localization results of the 9 damages on the composite panel of the fuel tank.
Δl
(mm)
(0.0, 80.0)
Direction and
distance error
(°, mm)
(0.1, 14.3)
(127.1, 138.6)
(128.7, 128.1)
(-1.6, 10.5)
11.2
(90.1, 112.4)
(90.0, 100.0)
(0.1, 12.4)
12.4
(51.3, 128.1)
(1.9, 9.8)
10.7
(104.9, 155.2)
(1.3, 17.8)
18.2
Damage
label
(kn1, tR1)
(rad/m, ms)
(kn2, tR2)
(rad/m, ms)
Localized position
(°, mm)
Actual position
(°, mm)
A
(309, 0.2085)
(1, 0.2059)
(0.1, 94.3)
B
(-179, 0.3635)
(238, 0.3590)
C
(-1, 0.3227)
(302, 0.3230)
D
(180, 0.3466)
(241, 0.3428)
(53.2, 137.8)
E
(-87, 0.3474)
(299, 0.3417)
(106.3, 173.0)
14.3
F
(-1, 0.3304)
(305, 0.3293)
(90.1, 158.8)
(90.0, 150.0)
(0.1, 8.8)
8.8
G
(51, 0.3337)
(299, 0.3323)
(80.4, 208.8)
(78.7, 204.0)
(1.7, 4.4)
7.5
H
(38, 0.3340)
(299, 0.3323)
(82.7, 237.7)
(80.9, 253.2)
(1.8, -15.5)
17.4
I
(-40, 0.3389)
(300, 0.3360)
(97.7, 282.4)
(97.6, 302.7)
(0.1, -20.3)
20.3
5. Conclusion
To promote spatial-wavenumber filtering technique to be applied to on-line damage
monitoring of composite structure, this paper proposes a new on-line damage imaging method
based on a SSWF and PZT 2-D cruciform array. The SSWF can be designed only according
to the maximum cut off wavenumber of the linear PZT array to scan the actual wavenumber
of the damage scattering signal to give out a wavenumber-time image. By using the PZT 2-D
cruciform array, the axis projection wavenumbers of the damage scattering signal can be
obtained from the two wavenumber-time images and the damage direction can estimated
correspondingly without any blind angle. According to the absolute arrival time obtained
from the wavenumber-time images and combing with the Lamb wave group velocity, the
damage distance can be estimated. The method is validated on the aircraft composite fuel tank
of variable thickness and the validation results show that the damage imaging method is
feasible to be applied to complex composite structure.
6. Acknowledgements
This work is supported by National Science Fund for Distinguished Young Scholars of China
(Grant No.51225502), key program of Natural Science Foundation of China (Grant No.
51635008), National Natural Science Foundation of China (Grant No. 51575263), Aviation
Foundation of China (Grant No. 20140952010), Priority Academic Program Development of
Jiangsu Higher Education Institutions of China, Qing Lan Project of China and Young Elite
Scientist Sponsorship Program by CAST of China.
References
1.
2.
3.
C Boller, F K Chang, and Y Fujino, ‘Encyclopedia of structural health monitoring’,
John Wiley, January 2009.
Z Su and L Ye, ‘Identification of damage using lamb waves: from fundamentals to
applications’, Springer, January 2009.
T E Michaels, J E Michaels and M Ruzzene, ‘Frequency-wavenumber domain analysis
of guided wavefields’, Ultrasonics, Vol 51, No 4, pp 452-466, May 2011.
9
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
H Sohn, D Dutta, J Y Yang, M DeSimio, S Olson and E Swenson, ‘Automated detection
of delamination and disbond from wavefield images obtained using a scanning laser
vibrometer’, Smart Materials and Structures, Vol 20, No 4, 045017, March 2011.
M D Rogge, C A Leckey, ‘Characterization of impact damage in composite laminates
using guided wavefield imaging and local wavenumber domain analysis’, Ultrasonics,
Vol 53, No 7, pp 1217-1226, March 2013.
L Yu, C A Leckey and Z Tian, ‘Study on crack scattering in aluminum plates with
Lamb wave frequency-wavenumber analysis’, Smart Materials and Structures, Vol 22,
No 6, 065019, May 2013.
A S Purekar, D J Pines, S Sundararaman and D E Adams, ‘Directional piezoelectric
phased array filters for detecting damage in isotropic plates’, Smart Materials and
Structures, Vol 13, No 4, pp 838, June 2004.
A S Purekar and D J Pines, ‘Damage detection in thin composite laminates using
piezoelectric phased sensor arrays and guided lamb wave interrogation’, Journal of
Intelligent Material Systems and Structures, Vol 21, No 10, pp 995-1010, June 2010.
B Xu and V Giurgiutiu, ‘Single mode tuning effects on Lamb wave time reversal with
piezoelectric wafer active sensors for structural health monitoring’, Journal of
Nondestructive Evaluation, Vol 26, No 2, pp. 123-134, November 2007.
H W Park, S B Kim and H Sohn, ‘Understanding a time reversal process in Lamb wave
propagation’, Wave Motion, Vol 46, No 7, pp 451-467, November 2009.
L Qiu, S Yuan, X Zhang and Y Wang, ‘A time reversal focusing based impact imaging
method and its evaluation on complex composite structures’, Smart Materials and
Structures, Vol 20, No 10, 105014, August 2011.
C A Balanis, ‘Antenna theory analysis and design’, John Wiley, April 2005.
L Qiu and S Yuan, ‘On development of a multi-channel PZT array scanning system and
its evaluating application on UAV wing box’, Sensors and Actuators A: physical, Vol
151, No 2, pp 220-230, April 2009.
L Qiu, M Liu, X Qing, and S Yuan, ‘A quantitative multi-damage monitoring method
for large-scale complex composite’, Structural Health Monitoring, Vol 12, No 3, pp
183-196, May 2013.
10
Non-destructive Evaluation of Wrinkles
+
Sivaramanivas R+, Megha Navalgund+, Debasish Mishra+ and Richard Klaassen*
GE Global Research, GE India Technology Centre, 122 EPIP, Whitefield Road, Bangalore 560066,
India
*GE Aviation, 1 Neumann Way, Quality Technology Centre, Evendale, Ohio 45215-1988, US
A study to develop nondestructive characterization methods for wrinkles in polymer matrix
composites (PMC) parts is presented. Wrinkles are small imperfections produced during the
manufacturing process that involve lay-up over a double curved geometry. The advantage that PMC
parts hold owing to the ability to design and produce required mechanical properties in specific
directions while keeping the weight of component low, is compromised by the presence of
manufacturing defects. Sensitivity of the NDE techniques has been studied using control wrinkle
samples and tensile strength tests. Control samples with varying wrinkle angles were produced using
CFRP laminates, six-layer laminate using TenCate E720, Epoxy RM135 inserts. Tensile tests were
used to identify the critical wrinkles and NDE technique capability was benchmarked against the
worst-case wrinkles.
We present test results from ultrasonic through transmission data and Computed Tomography on
control samples with varying wrinkle severity.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
A Study on Highly Porous Carbon-Carbon Aircraft Brake Disc
using Air Coupled Ultrasonic
Suresh Chand Jangir1, Srinivasa.V1, Ramesh Kumar. M1,
Ramesh Sundaram1, GururajaRao.J2
1
Advanced Composites Division, CSIR-National Aerospace Laboratories,
P.B.No.1779, HAL Airport Road, Bangalore – 560017
Phone: +91 80 25086402, Fax: +91 80 25267352; e-mail: srinivasa_v@nal.res.in, rameshrk@nal.res.in
2
Advanced System Laboratory, DRDO, Hyderabad; E-mail: rao214gm@gmail.com
Abstract
Carbon-Carbon (C-C) composites is a special class of ceramic matrix composites used in manufacturing of high
temperature applications has excellent properties like high specific heat and thermal conductivity, lower in
thermal expansion, good wear resistance and retention of mechanical properties. Inspection of these category of
composite is a challenge as it possesses more of deceptive and attenuate. Air Coupled Ultrasonic (ACU) is used
to over these inspection footraces. Evaluation of these Non Destructive Testing (NDT) results is a huge task
involving expertise in manufacturing and NDT inspection. As Carbon-Carbon (C-C) composites is more
attenuative and developing a standard / reference was more difficult and further evaluation on the actual aircraft
brake disc becomes a mammoth task. Air Coupled Ultrasonic (ACU) inspection parameters were frozen and the
brake disc are inspected and evaluated to the requirements of the approving authorities.
Keywords: Carbon-Carbon (C-C) composites, Air Coupled Ultrasonic (ACU), non-contact ultrasonic, aircraft
carbon brake disc
1. Introduction
Composites of different types; ceramic, polymer and metal matrix are being used in
aircraft industries for structures. These range of composites are being manufactured using
chemical vapour infiltration, liquid metal infiltration, carbon-carbon composites, VARTM,
autoclave moulding process. One of the important applications of C-C composites is the
Carbon-Carbon aircraft brakes. This is in the form of multiple discs and are used in
various civilian commercial aircrafts due to its excellent combination of properties like
high specific heat, good thermal conductivity, very less co-efficient of thermal expansion,
retention of mechanical properties at high operating temperatures and excellent wear
resistant. Manufacturing of C-C brake disc is a mammoth task involving lot of time during
development as the process of infiltration is very slow.
One of the challenges of Non Destructive Testing (NDT) is that these types of
composites are generally more attenuated. Furthermore, ceramic composites have
inherent porosity with multiple phases and degrees of non-uniformity. The porosity in
Carbon-Carbon composites is in the range of 5 to 15%. Conventional water coupled
ultrasonic technique is not feasible for the Carbon-Carbon composites due to the
inherently high porosity levels. This paper describes the challenges faced in the Non
Destructive Evaluation (NDE) [1,2] followed for inspection of these C-C disc.
2. Manufacturing Process
Manufacturing of C-C brake disc is a cumbersome and time consumption process.
Chemical Vapour Infiltration (CVI) and Liquid pitch/resin Impregnation methods [3,4,5]
are two important methods for the manufacturing of C-C composites. The CVI process is
gaseous process in which the hydrocarbon gases diffuse [6] into the carbon fibre preform
and cracked high temperature and deposits the solid carbon on the fibre surface forming
the carbon matrix. This a very also process, which takes about 3 months or more to get the
C-C composite of required density. The other process involves a repeated impregnation
and carbonation of carbon fibres preforms with carbon rich materials like pitch or resin.
Several cycles of impregnation and pyrolysis is required to get the required densities.
These process are the densification processes, and there always a residual porosity at the
end of the process, the amount which depends on the final density of the product. This
porosity makes the C-C products more attenuating and challenge to carry out the NDT
especially the ultrasonic based ones [7,8].
3. Results and Evaluation
Non Destructive Evaluation process is being followed using Air Coupled Ultrasonic
(ACU) technique have been used with very high power and low frequency ultrasonic
transducer [5,9] for specific inspection requirements. Evaluation of the NDT results
requires knowledge in the manufacturing process as well as analytical capability of NonDestructive Evaluation. The C-C disc is having minimum thicknesses of 1.0” with/without
cut-outs. Since the C-C disc is highly porous and the thicknesses are higher, the most
feasible method of inspection is through transmission Air Coupled Ultrasonic (ACU)
technique.
3.1 ACU Inspection: High transduction Non-Contact Ultrasound Second WaveTM, Ultran
system [8,9,10] is being used for inspection of the C-C brake disc. This system is built to
suit the inspection of highly porous composite components [11,12,13].
Transducer
Ref laminate
Figure 1 Test setup of ACU on a reference panel
ACU techniques have been used for inspection of C-C disc with the following parameters
based on the iteration conducted during experiments:
Frequency – 500 KHz
Voltage – 325 volt
Burst – 2
Gain – 75 dB
PRF – 100 Hz
LPF – 800 KHz
HPF – 80 KHz
Scan resolution – 1 x 1 mm
Scan speed – 50 mm/sec
Index speed – 20mm/sec
Ambient air path between transducer faces – 85mm.
The parameters followed for inspection produce lot of noise resulting in difficulties in
receiving the flawless signal for analysis. Signals received on a good healthy C-C disc [14]
have less amplitude due to inherent property of the material; it is difficult to differentiate
the signal on C-C with porosity on a good region and porous region. By ACU, different
category discs were inspected and evaluated.
The NDE results have given vital
information for the improvisation of the process to achieve the desired final quality.
During this experiments many C-C brake disc have been inspected and evaluated.
3.2 C-C brake disc analysis:
C-C brake discs were processed through Resin/Pitch impregnation and pyrolysis route. The
disc densities were increased by subjected to densification cycles which involves repeated
impregnation and pyrolysis cycles. The discs of different densities are obtained by carrying
out different No. densification cycles. It has been observed that, when the disc is densified
beyond certain extent, the micro cracking has occurred within the discs [15]. The
occurrence of micro cracking within the disc is actually revealed in the NDT. This gives
an excellent input that up to what level of densification can be achieved without damaging
the discs.
Region
in Air
Porous
Region
Figure 2 Good (LHS) and porosity accumulated (RHS) micro
cracked (due to over densification) C-C rotor disc
Many number of C-C rotor brake disc with cut out was inspected and evaluated. The brake
disc which has very less porosity considered as good sample. Based on the qualitative
approach methods the other disc was evaluated and qualified. The C-C disc having highly
porous when compared with good disc is shown in figure 2. In the highly porous disc the
micro porous are accumulated during infiltration process of manufacturing led to weaken
the disc when compared with good one. Based on this exercise confidence level was buildup to inspection any level of porous with different thickness can be inspected and
evaluated. The stator disc with cut out in the ID of the C-C brake disc. Few discs were
without cut out too. In the stator disc shown has very high level of porous when compared
with the good stator disc in figure 3. It is observed that the peak to peak scale of the ACU
image the highly porous disc in the range of 31 to 35 dB, but the good disc has peak to
peak attenuation levels of 16 dB. This wide range of dB levels brings out the differences
between the good and porous disc which is predictable phenomena.
Frame
Figure 3 Good (LHS) and highly porous (RHS) C-C stator disc
4. Conclusion
The C-C disc manufacturing process was stabilized to attain very less porous disc of rotor
and stator for it intended application. NDE is the back bone of this process improvement
towards production standards. During this process the NDE procedure and parameters were
also iterated in identification of porosity and other defects. NDE analysis play a crucial role
in the acceptance of the CVI manufacturing process of C-C brake disc for aircraft
applications. Based on the criticality of the disc i.e very high porous disc, the disc was
reprocessed for further improvement to reduce the porosity level. This was carried out based
on the NDE analysis report. Due to this exercise the rejection rate was reduced and led to
cost saving and production time of the programme.
References
1. Bhardwaj, M.C., “Non-Destructive Evaluation: Introduction to Non-Contact
Ultrasound,” Encyclopedia of Smart Materials, ed. M. Schwartz, John Wiley & Sons,
New York, NY (2002)
2. Vun, R., Eischeid, T., and Bhardwaj, M., Quantitative Non-Contact Ultrasound
Testing and Analysis of Materials for Process and Quality Control,” Proc. European
Conference on Non-Destructive Testing, Berlin, Germany (2006.)
3. Bhardwaj, M.C., “Non-Contact Ultrasonic Testing and Analysis of Materials,” Smart
Materials, Ed. Mel Schwartz, Taylor Francis Group, CRC Press (2009)
4. Bhardwaj, M.C., “Non-Contact Ultrasonic Characterization of Ceramics and
Composites,” Proceedings Am.Cer.Soc., V 89 (1998).
5. T. Carneim, D.J. Green & M.C. Bhardwaj, “Non-Contact Ultrasonic Characterization
of Green Bodies,” Cer. Bull., April 1999.
6. Jones, J.P., Lee, D., Bhardwaj, M., Vanderkam, V., and Achauer, B., “Non-Contact
Ultrasonic Imaging for the Evaluation of Burn-Depth and for Other Biomedical
Applications,” Acoust. Imaging, V. 23 (1997).
7. Wu, Q., Vun, R., Bhardwaj, M.C., and Stead, G., “Through-Thickness Ultrasonic
Transmission Properties of Oriented Strandboard,” Proceedings, 12th International
Symposium on Nondestructive Testing of Wood, University of Western Hungary,
Soporon, Hungary, Sept. 13-15: 77-86 (2000.)
8. Hoover, K., Bhardwaj, M.C., Ostiguy, N., and Thompson, O., “Destruction of
Bacterial Spores by High Power Non-Contact Ultrasound,” Mat. Res. Innovat. 6:291295 (2002).
9. Bhardwaj. M.C., “High Effieicncy Non-Contact Transducers and a Very High
Coupling Piezoelectric Composite,” an invited paper, World Conference on NonDestructive Testing, August-September 2004, Montreal, Canada.
10. Ganezer, K., Hurst, K., Shukla, S., Bhardwaj, M., “Initial Studies of Non-Contact
Ultrasound for Osteoporosis and Bone Imaging,” 44th Meeting American Association
of Physicists in Medicine, Montreal, QC, Canada, July 14-18, 2002.
11. Bhardwaj, M.C., “Evolution of Piezoelectric Transducers to Full Scale Non-Contact
Ultrasonic Analysis Mode,” a special symposium panel discussion, World Conference
on Non-Destructive Testing, August-September 2004, Montreal, Canada
12. Bhardwaj, M.C. Mizuta, M and Yaoita, T., “Non-Contact Ultrasound: The Last
Frontier in Non-Destructive Analysis,” Proc. The 6th Far-East Conference on NonDestructive Testing, October 21-24, 2002, Tokyo, Japan
13. Brunk, J.A., Valenza, C.J., and Bhardwaj, M.C., "Applications and Advantages of Dry
Coupling Ultrasonic Transducers for Materials Characterization and Inspection," in
Acousto-Ultrasonics, Theory and Applications, John C. Duke, Jr., Editor, Plenum
Press, New York (1988).
14. Kulkarni, N., Moudgil, B. and Bhardwaj,M., “Ultrasonic Characterization of Green
and Sintered Ceramics: II, Frequency Domain,” Am. Cer. Soc., Cer. Bull, Vol. 73, No.
7, (1994).
15. Bhardwaj, M.C., “Innovation in Non-Contact Ultrasonic Analysis: Applications for
Hidden Objects Detection,” Mat. Res. Innovat. (1997) 1:188-196.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Evaluation of Thermography and Ultrasonic NDT Techniques for Detecting
Resin Rich and Resin Starved Defects in Composites
Y.L.V.D. PRASAD, Dr. S.K. MAJEE, A.O. SIDDIQUI
Advanced Systems Laboratory, DRDO, Hyderabad, India
Phone: +91 40 24584403; e-mail: ahmedovaissiddiqui@asl.drdo.in
Abstract
Composites are the materials which have proven their superiority over other conventional materials for
applications in aerospace, defence and other advanced application areas. However, along with multiple
advantages associated with composites, one major area for concern associated with composites is the presence of
defects. These defects may significantly degrade the properties of composites leading to the catastrophic failure
of the component. Composites may have process induced defects like delamination, resin starved and the resin
rich zones etc. Sometimes it becomes very difficult to distinguish between resin rich and resin starved zones by
NDT inspection, and sometimes these defects may even go undetected. With the objective of assessing the
efficiency of Thermography and Pulse Echo Ultrasonic NDT (PET) techniques for distinguishing between resin
rich and resin starved zones, a 5mm thick carbon-epoxy laminate was fabricated with multiple types of defects
embedded in it. A novel fabrication technique was adopted to create clear air defects and defects to simulate the
presence of resin rich, resin starved and foreign object inclusions. Detailed Thermography and PET tests were
carried out and it was found that for clear air defects, the Thermography and PET are equally good; whereas, the
Thermography has advantage over PET in detecting resin rich zones. Conversely, the PET has advantage over
Thermography in detecting resin starved zones. Findings from this study need to be further probed for other
materials and defect configurations to derive final conclusions for distinguishing between resin rich and resin
starved zones.
Keywords: Composites, Delamination, PET, Resin Rich, Resin Starved, Thermography
1. Introduction
Composite materials are increasingly used in aerospace, naval, automotive and many other
industries due to their high specific strength and stiffness properties. Due to inherent
complexities of composite structure manufacturing, various types of defects such as voids,
inclusions, delaminations, disbonds, resin rich resin starved zones etc. may be present in the
final product. Proper assessment of these defects is essential for effective utilization of these
products. There are several NDT techniques like radiography, ultrasonic testing,
shearography, infrared thermography etc for detecting defects in composite. Although
traditional techniques such as ultrasonics, easily reveal the presence of flawed areas, however,
they are time consuming; Whereas, Infrared thermography is a faster and non-contact
technique, which does not require any coupling agent.
Infrared thermography is being applied for variety of applications ranging from detection of
delaminations and disbonds in layered structures, assessing hidden corrosion in metallic
components, detection of local wall thinning in metallic tubes, cracks in ceramics and metals,
voids, impact damages and inclusions in composite materials, investigation of adhesion
integrity of kissing bond region in plastic welded joints etc. [1-3]. It provides faster, safe, noncontact and nondestructive tool for evaluation of sub-surface defects in materials. Active
Infrared (IR) thermography NDT is well known technique for assessment of sub-surface
defects in composite structures [4-5]. This technique does not affect the material or structure’s
future usefulness and at the same time provides an excellent balance between quality control
and cost-effectiveness. In active infrared thermography, the object is heated briefly by a heat
source and an IR camera monitors the transient temperature behavior of the surface. The
presence of sub-surface defects in the test material disturbs the heat flow inside the test
specimen, and this disturbance caused by the defect gets manifested on the surface as a hot or
cold spot as seen by the IR camera.
The two types of defects which occur in composites are resin starved (Voids, Delamination)
and the resin rich defects. In current study an attempt has been made to compare the
applicability of Ultrasonic and Thermography techniques in detecting the resin rich and resin
starved zones in composites. Effect of the ultrasonic probe diameter on defect detectibility too
has been studied.
2. Experimentation
2.1 Fabrication of composite laminates with implanted defects
The material used in the current studies was Carbon-epoxy. One laminate of dimensions
300x300x5mm was fabricated by using carbon-epoxy UD prepreg. Depth of all the defects
was 2.5mm from one of the side. All the defects had same thickness. Prepreg layers were laid
in 0-90o pattern. Various types of defects were implanted in the laminate. Defect details are
given in Table 1 and Figure 1.
52.5mm 20mm 75mm
52.5mm
5mm
20mm
20mm
13
12
14
2
11
15
3
10
16
5mm
1
10mm
20mm
150mm
20mm
4
5
6
82.5mm 5mm
5mm
75m
75mm
7.5mm
9
9
17
8
18
7
19
20mm
5mm 82.5mm
Figure 1: Details of implanted defects
Table 1: Details of implanted defects
Defect No.
Defect type
1
Glass balloons
2
Blank groove
3
Foam Pieces
4, 5, 6
Same as 1, 2, 3
7
Liquid epoxy resin (later cured during curing of laminate)
8
Carbon fiber tow
9
Carbon fiber tow
10, 11
Same as 7 & 8
12, 13
Same as 9
14
Cured epoxy resin pieces
15
Metal piece
16
Blank groove covered with high temp. polyethylene sheet
2.2 Experimental details
NDT experiments were carried out by using Thermography and Pulse Echo Ultrasonic
Technique (PET). Experimental details are given as follows:
2.2.1 Thermography
Thermography experiments were conducted by using Flash Thermography setup. The flash
time was 5ms and the heating power was 9.6KJ. Frame capture rate was 50Hz. The raw data
collected from thermography experiments was curve fitted by using TSR technique [6]. Data
processing was carried out by using 1st derivative and 2nd derivative techniques. The
processed images were analyzed and the thermogram at the time instance with maximum
defect detectibility and clarity was used for interpretations and drawing conclusion.
2.2.2 PET ultrasonic
PET Ultrasonic experiments were carried out with 2.5MHz probe. The probe diameter was
10mm and the coupling agent used was water soluble gel.
3. Results and Discussion
Data obtained from Thermography and Ultrasonic NDT was analyzed and for clear analysis
the defects have been segregated in four groups. These are:
1. Defects with defect echo and no back-wall echo by PET
2. Defects with both defect echo and back-wall echo by PET
3. Defects detectible by Thermography but not by PET
4. Defects detectible by none of the technique
These four groups of defects are discussed in following sections.
3.1 Defects with defect echo and no back-wall echo by PET
Under this category the defects were detectible by both the techniques. These defects are
shown in Figure 2. The processed thermogram is given in Figure 3. The detectible defects
include defect no. 1, which constituted air balloons. Air balloons consist of finite volume, due
to which during cure the prepreg layers do not compress and result in finite volume defect.
Similarly defect No. 3, 5 & 15 constituted finite volume inserts resulting in either clear air
defects or low diffusivity insert. Defect no. 2 didn’t constitute any insert but was a blank
groove. During cure the prepreg layers might have slightly compressed resulting in distorted
and smaller sized defect. Defect no. 16 and 19 constituted empty grooves but were covered
with thin polyethylene sheet. Due to this sheet, neither the resin filled the groove nor the
prepreg layers got compressed. This has resulted in clear air defect with well defined shape.
From these observations it is inferred that the defects which were detectible by both the
techniques constituted clear air gap or the insert with low thermal diffusivity. Thermography
and PET ultrasonic were found equally good for detecting these types of defects; however,
thermography has added advantage over PET that it gives information about the defect size
too, which is very important information for Non–destructive inspection of composites.
1
Glass Balloon
Metal
2
Blank
15
3
Foam Pieces
16
Blank groove covered
with polyethene
6
Foam Pieces
(Normal defect
Peak)
19
Blank groove covered
with polyethene
Figure 2: Defects detectible both by Thermography and PET techniques
Figure 3: Processed thermogram
Carbon fiber
13
12
4
Glass balloons
8
Carbon Fiber
7
Liquid Resin (Later cured)
Figure 4: Defects with defect echo and back-wall echo by PET
3.3 Defects detectible by Thermography only
Defects which are detectible by only Thermography are shown in Figure 5. Defects under this
category include defect No. 10, 14 & 17, which are filled with resin. These defects simulate
resin rich zones in composites. These defects have been clearly detected by Thermography
but not detectible by PET. No detection of these defects by PET implies that there is no air
entrapment in these defects.
Defect no. 9 (Carbon fibre tow) has been detected by Thermography but not by PET.
Incidentally, it has been seen in previous section that the defect no. 8, which is also a carbon
fiber tow defect and of much larger size than the similar defect 9 has not been detected by
Thermography. This shows that the defect no. 9 has definitely got some air entrapment due to
which it is detectible by thermography. However, even with air entrapment it is not detectible
by PET. This could be due to the smaller size of the defect in comparison to the Ultrasonic
probe diameter. Similarly, there could be some air entrapment at defect 18 too, due to which it
is detectible by thermography but not by PET.
From these observations it is inferred that the defects which are detectible by thermography
but not by PET are either resin rich defects or are the defects with air entrapment and of
smaller size than the UT probe diameter.
Resin Pieces
14
Liquid Resin
10
Resin Pieces
9
17
Carbon fiber
18
Metal Pieces
(Normal-BW)
Figure 5: Defects detectible by Thermography only
3.4 Defects detectible by none of the technique
Defect no 9 & 11 are having carbon fiber tow embedded in it and have not been detected by
any of the methods. It could be due to the reason that these defects are much smaller in size
and the PET has difficulty in detecting the defects smaller than the probe diameter. Moreover,
thermography has difficulties in detecting resin starved defects. These defects are shown in
Figure 6.
Defect no. 5 too has not been detected by any of the method. This could be due to the reason
that the pepreg layers have been compressed during curing resulting in disappearance of the
defect. Due to which none of the technique could identify this defect.
Carbon Fiber
11
9
Carbon fiber
5
Blank
Figure 6: Defects detectible by none of the techniques
4. Conclusion
A carbon-epoxy laminate with artificially implanted defects was fabricated. A variety of
defects were implanted in the laminate to assess the feasibility of defect detection by
thermography and PET NDT techniques. It was observed that certain defects were detectible
by both the techniques and some went undetected by either of the technique. There were
certain defects detectible by only one of the technique. It has been found that for clear air
defects both the techniques are equally good. PET has been found advantageous over
Thermography in detecting resin starved zones; whereas, Thermography has an edge over
PET in detecting resin rich zones. For the defects detectible by none of the techniques it is
expected that the prepreg layers have compressed during cure resulting in disappearance of
the defect. However, for finding the exact reason why few defects went undetected, it is
planned to carry out the CT scan on the laminate. This will help in visualizing the internal
structure of the laminate layer-by-layer and act as a tool in corroborating the findings from
Thermography and PET techniques.
References
1. XPV Maldague, ‘Theory and practice of infrared technology for nondestructive
testing’, 1st ed., New York: Wiley-Interscience, pp. 453–525, 2001.
2. M Omar, M Haassan, K Donohue, et al., ‘Infrared thermography for inspecting the
adhesion integrity of plastic welded joints’, NDT and E Int, Vol 39, pp 1-7, 2006.
3. A Vageswar, K Balasubramaniam, CV Krishnamurthy et al., ‘Periscope infrared
thermography for local wall thinning in tubes’, NDT and E Int, Vol 42, pp 275-282,
2009.
4. XPV Maldague and S Marinetti, ‘Pulse Phase Infrared Thermography’, J.Appl.Phys,
Vol 79, pp 2694-2698, 1999.
5. BC Ray, ST Hasan and DW Clegg, ‘Evaluation of defects in FRP composites by NDT
techniques’, J. Reinf. Plast. Compos, Vol 26, pp 1187-1192, 2001.
6. SM Shepard, ‘Temporal noise reduction, compression and analysis of thermographic
image data sequences’, U.S. Patent No. 6,516,084, 2003.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Aging study of Coated fabric materials
Harshavardhana Nanjundegowda, Rajamanikandan Sivaraman, Bharath Marappan &
Venkatadiri Seshadiri
UTC AEROSPACE SYSTEMS, Bangalore, INDIA
Tel: +91-80-6737 0102,
Harshavardhana.Nanjundegowda@utas.utc.com, www.utcaerospacesystems.com
Abstract:
Aircraft inflatable assemblies are made from woven fabrics that are coated with polymeric materials. These
polymeric coatings form the basis of the joining and sealing of the inflatable structures. Polymeric coatings by
nature are susceptible to degradation over time, especially when exposed to environmental conditions such as ozone,
hot and cold cycling, humidity and moisture, or other fluid exposure and will show a degradation of mechanical
properties. The woven fabric provides the underlying structural properties of the system. The fabric is comprised
mostly of nylon fibers which will also degrade over time upon exposure to the same environments. If the
degradation of either the fabric or the coating is significant enough, the structural properties of the inflatable
assembly can become compromised and will result in a leaking, damaged. A study was conducted to acquire
mechanical property data on fabrics and adhesives that were in service for many years. The purpose of this study
was to understand how these materials deteriorate with respect to number of years in service and exposure to the
various field environments. The test results were analyzed to predict the failure rate of fabric materials over the
period of time using Binary Logistic Regression analysis. This particular effort was a comprehensive study meant to
assess the entire population of inflatables for different ages.
1. Background
The evacuation systems assembly is made from lightweight, age sensitive fabrics, and assembled
with adhesives. The in-service environment that the systems are exposed to can cause
deterioration and wear with age. The study was conducted to understand the effect of age on the
assemblies by collecting the data on mechanical properties on fabrics and adhesives from the
field service assemblies.
2. Introduction
Emergencies at take-off and landing often demand swift removal of the passengers from the
aircraft because of the potential for injuries from fire, explosion, or sinking in water. A
conventional method of fast evacuating a large number of passengers from an aircraft is to
provide multiple emergency exits, each of which is equipped with an evacuation slide.
Emergency evacuation slides are air holding escape devices that are put at each of the aircraft
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
emergency exit doors. The inflation system automatically actuates when the packed slide is
ejected out of the aircraft exit door.
The air holding fabric assembly which inflates to become the structure of the slide. It is primarily
made of polyurethane/neoprene coated nylon fabric panels bonded together using adhesives. The
base cloth is made woven fabric weaved in two direction. The fibers running along the length of
the fabric roll is referred as warp direction and the fibers running across the width is referred as
fill direction as shown in figure 1.
Figure 1 : Fiber direction in coated fabric
3. Sampling and Testing
Sampling required destructive testing of the assembly and slide lane fabric. Seam peel, tensile,
and tear samples were taken from the head, center/middle, and toe of assembly as well as the
slide lane fabric The samples were then grouped and tested by location for each slide using an
UTM.
Figure 2: Evacuation System Sample Locations
All test specimens are obtained from service slides, with ages between varying 6 and 17 years
old. As made production units were subject to testing to represent time equals zero (0) years. A
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
total of 32 in-service slides were evaluated. All test slides were tested to product requirements
for visual inspection, air retention, and proof pressure prior to physical destructive testing.
A histogram of the slide construction ages versus number of slides in this study is shown in
Figure 3.
Figure 3: Number of slides Vs. Age
3.1.
Material Level Tests
Material test requirements for each aircraft slide are summarized in Table 2.’
Table 1: Material test requirements for inflatable and slide fabrics
Fabric
Inflatable
Fabric
Slide
Test
Test Method
Seam Peel Adhesion
Coating Adhesion
FTMS 191A, Method 5960
FTMS 191A, Method 5970
Tensile Strength (Warp and Fill)
Tear Strength
(Warp and Fill)
Seam Peel Adhesion
Tensile
(Warp and Fill)
Tear Strength, (Warp and Fill)
FTMS 191A, Method 5100
FTMS 191, Method 5136
FTMS 191A, Method 5960
FTMS 191A, Method 5100
FTMS 191, Method 5136
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
4. Results:
Evacuation system inflatable test results for properties are show in below box plots. The boxes in
the boxplots represent the middle 50% (interquartile range) of the test values. The horizontal line
within the box represents the median of the test results. The whiskers extending from the boxes
indicate the highest and lowest values in the data set, excluding outliers. Outliers are represented
by asterisks in the boxplot figures. Outliers are considered to be values ≥ 1.5 times the
interquartile range. The boxplots were marked with the material specification requirements using
a red line.
4.1.
Box Plots for Fabric Material Testing:
Figure 4: Boxplot of tensile strength of inflatable fabric in the warp direction vs. slide age
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Figure 5: Boxplot of tensile strength of inflatable fabric in the fill direction vs. slide age
Figure 6: Boxplot of tensile strength of slide lane fabric in the warp direction vs. slide age
Figure 7: Boxplot of tensile strength of slide lane fabric in the fill direction vs. slide age
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Figure 8: Boxplot of tensile strength of slide lane fabric in the fill direction vs. slide age
Figure 9: Boxplot of coating adhesion strength of inflatable fabric vs. slide age
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Figure 9. Boxplot of seam peel strength of inflatable fabric vs. slide age
Figure 10: Boxplot of seam peel strength of slide lane fabric vs. slide age
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
5. Analysis
The box plots identify the tensile strength of the inflatable fabric and the tear strength of the slide
lane fabric as critical mechanical properties that show the most response to age. The test data
used for analysis came from multiple locations on the slide and contained one or more inflatable
assemblies of the same age from various carriers. As such, the data considers variation with age,
carrier, environment differences from evacuation system-to-evacuation system, and sample
location. The large variability in the mechanical properties is an expected outcome due to the
variation of environments that the slides are exposed to in service.
In order to take into account the mixed effects of these variables, a statistical analysis was
conducted and an analytical model was generated. The test data for all material mechanical
properties conducted was further analyzed using binary logistic regression. Binary logistic
regression was chosen as the predictive analysis technique due to the large volume of data and
the binary nature of the data (pass or fail to meet minimum requirements). The analysis takes
into account the entire data population for each inflatable assembly analyzed, and not just the
worst case retained strength or weakest point of the inflatable, and evaluates the true age effect.
We have presented the specific techniques utilized to analyze the data in this study below:
A binary logistic regression analysis was implemented on data collected four physical properties
as a function of age (time, t, in years). The following physical properties were analyzed:
Coating adhesion strength
 Seam peel strength
 Tear strength
 Tensile strength
These properties were evaluated for inflatable and slide lane fabrics.
For each property there is a minimum specification requirement. Each data point is compared to
its minimum requirements, and the binary variable is defined as “1” if the measurement meets
the requirement and “0” if not. The regression model is of the following form:
p
ey
1 e y
Where y is the linear function:
y   0  1t
The variable t is time in years. The response variable p is interpreted as the probability of failure
to meet the property requirement. The parameters β1 and β0 are determined from the variable
data and time variable t using the binary regression methodology. Once determined the model
returns the probability of failing to meet the requirements as a function of age of the material. In
addition, the analysis allows for the determination of confidence intervals for this probability at
each time. This is to interpret as uncertainty about the probability number being reported. It is
important to note that the probability p measures the propensity of the material to fail to meet the
requirement at the associated age. It does not measure the actual extent of the failure at the
associated age.
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Figure 10 combines the results of the binary logistic regression analyses for all physical
properties determined in this study. In all cases, except for the tensile strength of the inflatable
fabric and the tear strengths of the slide lane and inflatable fabrics, the probability of the physical
properties not meeting minimum requirements increases gradually during the life of the fabrics.
For the tensile strength of the inflatable fabric and the tear strength of the slide lane and
inflatable fabrics the probability of not meeting minimum requirements does not follow the
gradual increase in probability of not meeting the minimum requirements. However, the tensile
strength of the inflatable fabric and tear strength of the slide lane fabric quickly deviate from the
predicted gradual increasing probability of not meeting minimum requirements. Because the
probability of not meeting requirements of the tensile strength of the inflatable fabric and tear
strength of the slide lane fabric are significantly higher than the tear strength of the inflatable
fabric, these were focused on as the highest risk properties.
Figure 10: Regression analysis summary of all physical properties vs. slide age
6. Conclusion
The study shows a decrease in the tensile strength of the inflatable fabric and the tear strength of
the slide lane fabric of the inflatable assembly with age. While a direct correlation of the tensile
strength of the inflatable fabric and the tear strength of the slide lane fabric values with inflatable
assembly system failure has not been established, it is reasonable to conclude that the reduced
tensile strength of the inflatable fabric and the tear strength of the slide lane fabric as a function
of age is a strong predictor of inflatable failure. The increasing probability that the evacuation
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
system will fail to meet the minimum strength requirement after a certain age is a strong
indicator of an increased risk of structural degradation of the inflatable assembly.
Based on the above, the decrease in the physical properties of the fabrics over time is due to agerelated degradation of the slide lane and inflatable fabrics
UTC AEROSPACE SYSTEMS PROPRIETARY
This Document contains no technical data subject to the EAR or the ITAR
Online Structural Health Monitoring of Composites
Harsh Shah, Prabhu Rajagopal, Krishnan Balasubramaniam
Centre for Non-Destructive Evaluation,
Department of Mechanical Engineering,
Indian Institute of Technology Madras,
Chennai India
Contact: hjharsh@gmail.com, prajagopal@iitm.ac.in
Aerospace & Automotive industries make extensive use of composite structures, often with complex
geometries. Ultrasonic guided waves are attractive for long-range inspection of large-scale structures
and have developed into a very effective tool because they can inspect a relatively large area in a short
time. The effectiveness of guided waves in quantitative defect detection in composites is well
documented [1]. Present methods use contact surface transducers or air coupled transducers to
generate and propagate waves through the composites [2]. The proposed system utilizes a waveguide
sensor which is embedded into the inter-laminar region of composites. These waveguides confine the
wave transmission in one dimension and waves leak only through the opening provided, which
enhances the capability to inspect large composite structures with very low attenuation rate.
Inaccessible areas can be inspected and inter-laminar delamination detection can be achieved. Live
monitoring and assessment of discontinuities can be accomplished effectively by using this
mechanism.
References
[1] Mitra M., Gopalakrishnan S., Guided wave based structural health monitoring: A Review. Smart
Materials & Structures 25 (2016) 053001 (27pp)
[2] Zhao X., Gao H., Zhang G., Ayhan B., Yan F., Kwan C. and Rose J. L.. Active health monitoring
of an aircraft wing with embedded piezoelectric sensor/actuator network: I. Defect detection,
localization and growth monitoring. Smart Materials & Structures 16 (2007) 1208–1217
Integrated Approach to Demonstrate Optimum Sensor Positions in a Guided
Wave Based SHM System using Numerical Simulation
Ramanan Sridaran Venkat1, Christian Boller1, Lei Qiu3 Nitin Balajee Ravi2, Debiprosad Roy
Mahapatra2, Nibir Chakraborty2
1
Universität des Saarlandes, 66125 Saarbrücken, Germany
2
Department of Aerospace Engineering, Indian Institute of Science, Bangalore 560012, India
3
Key State Lab of Mech. & Control of Mech. Struct., Nanjing University of Aeronautics & Astronautics (NUAA), 210016 Nanjing, P.R. China
Contact: Ramanan.sridaran@uni-saarland.de, c.boller@mx.uni-saarland.de, nitinb@aero.iisc.ernet.in,
droymahapatra@aero.iisc.ernet.in, nibir.chakraborty@aero.iisc.ernet.in, lei.qiu@nuaa.edu.cn
In aeronautical applications a guided wave based structural health monitoring (SHM) system has the
advantage of covering comparatively larger distances with high sensitivity for monitoring structural
damages specifically when the structure has plate like shapes being common in various aeronautical
structures. According to the damage tolerant principle, the structure has been designed for an
allowable damage and the condition of the damage has been continuously monitored using
piezoelectric wafer acoustic sensors (PWAS) that record signal having been generated by a
piezoelectric actuator. The location of such actuators and sensors around the tolerable damage to be
monitored has been widely regarded as the important prerequisite for an efficient SHM system design.
Numerical simulations provide thorough understanding of wave mechanics within the structure. FEM
based COMSOL Multiphysics has significant advantages of coupling piezoelectric and structural
mechanics modules to model the guided wave propagation in a more efficient way. The advantage of
viewing wave patterns at various time intervals for pristine and damaged conditions would allow one
to compute differential signals/ differential images in order to identify the hot spot areas where the
actuators and sensors could be placed to reliably detect the tolerable damage. However, FEM based
simulation is time consuming when the structure to be simulated is large. In a modular approach, ray
tracing has been combined with FEM for simulating those large structures in a less computation time.
A plate with three holes and a crack on the center has been identified as the demonstrator for this
paper. The simulated results have been validated with the experiments. Besides the optimal sensor
placement, this paper attempts to introduce the coated piezoelectric material onto the plate based on
the differential wave pattern from the FEM simulation.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Wave scattering analysis in a delaminated cross-ply laminate due
to incident S0 wave
Rajendra Kumar Munian, G Kolappan Geetha, D Roy Mahapatra, S Gopalakrishnan
Department of Aerospace Engineering, Indian Institute of Science, Bangalore, India
e-mail: rkmunian@aero.iisc.ernet.in, ganesh.kolappan.geetha@gmail.com, droymahapatra@aero.iisc.ernet.in,
krishnan@aero.iisc.ernet.in
Abstract
This paper investigates the effect of delamination on wave scattering due to incident S0 mode. When composite
laminate is subjected to ultrasonic guided wave, due to the wave-delamination interaction forward and backward
wave appear which includes important cross-correlation terms that couple the propagating modes with the
evanescent waves, and the forward waves with the backward waves. These cross-correlation terms can be used
effectively to extract certain quantities of interest that give significant insight regarding the delamination
characteristics. Reliable identification of scattering mechanism in this way can make it possible to confirm the
presence and predict the location and geometric properties of delamination. In this paper various wave scattering
behaviour is presented from analytical model as well as time domain spectral finite element simulation. Wave
interaction with delamination in different interfaces is studied to understand the nature of wave near the damaged
region and predict the response characteristics for different delamination interface in a laminate. S0 wave
produces no reflection for unidirectional laminates but for cross-ply laminate there is reflection if the
delamination is not in mid-plane. It happens due to mismatch in the stiffness properties in sub-laminates which
leads to S0 wave to propagate in different velocities with phase differences, that causes reflection as well as
mode conversions.
Keywords: Guided wave, delamination, mode conversion, wave scattering
1. Introduction
Composite has wide application in various industries especially in aerospace because of
its high specific strength which offers scope to reduce the weight of the structure. Composites
are sensitive to various unwanted loading it experience during service life or maintenance.
One of the most common types of damage found in composite is delamination. Difficulty in
detection and characterization of delamination remains challenging tasks due to growth of
damage in interfacial plane and direction dependent and non-homogeneous properties of
composite laminate. There are various methods existing in non-destructive method which
includes through transmission, thermo-graphic, X-ray scan, tomographic testing etc. Among
all those method guided wave based damaged detection method becomes popular [1] because
it can propagate longer distance without much attenuation and can bring damage induced
signature from hidden location. Using guided wave based damage detection method damage
location is determined form time of flight which can be calculated from the responses whereas
size of the damage estimated from the wave packet characteristics. Information of
delamination depth can be predicted from the damage induced scattering and local
wavelength analysis [2,3]. Both A0 and S0 are widely used in damage detection structure. S0
wave is non-dispersive in nature and have low attenuation compared to A0 wave [4].
There are several mathematical techniques available to model wave propagation in
structure including finite difference, boundary element, finite element method etc. All those
methods have their unique advantage depending upon the complexity of the problem.
Simulation of ultrasonic wave propagation in structure using conventional method requires
fine spatial as well as temporal discretization [5, 6] which increase the computational size
enormously. Spectral methods are applied successfully in modelling wave propagation
structures to reduce the computation cost. There are two types of spectral method available in
literature, one is frequency domain spectral method [7] and other one is time domain spectral
method [8]. Time domain spectral method has the advantages over frequency domain
counterpart when dealing with complex 2-D or 3-D structure [9]. Time domain spectral finite
element method has been used in modelling wave propagation in rod, beam and shell structure
[9, 10] with superior accuracy and convergence.
In this study effect of S0 wave in cross-ply laminate with large delamination is studied
using analytical model based on Euler-Bernoulli beam approximation as well as simulation
using time domain spectral finite element method (TSFEM). Wave interaction with
delamination causes reflection and transmission. Reflection happens from the leading
delamination tip as well as trailing delamination tip. If the delamination length is small
reflection from both the delamination tip cannot be identified separately as responses
overlaps. To study the delamination tip effect on the reflection and mode conversion, very
large delamination is considered that is wave length is very small compared to delamination
length. Reflection characteristic as well as mode conversion is studied for various cases with
different interface positions of delamination for various frequencies.
2. Mathematical modelling for semi-infinite delamination
In this section, analytical model of wave propagation in cross ply laminate with semiinfinite delamination is derived based on the Euler-Bernoulli beam approximation to study the
delamination tip effect on the wave propagation. When incident wave reaches the
delamination tip, some of the energy is reflected and rest is split into two sub-laminates. In
order to identify the effect of the tip on wave interaction, semi-infinite delamination is
considered. In reflected wave content there can be symmetric as well as anti-symmetric wave
mode due to the mode conversion. Similarly in sub-laminates also both the wave modes can
be present. It can be considered that wave propagation takes place in three regions. In baselaminate (region-1) forward moving incident S0 wave, backward moving reflected S0 and A0
waves are present. In upper sub-laminate (region-2) and in lower sub-laminate (region-3)
forward moving S0 and A0 waves are present.
z
V2
P2
P2
M1
M2
x
(1)
P3
(2)
(3)
P3
M3
L1
V1
V3
Figure-1: Schematic diagram of tip of the delamination
First, displacement fields for a particular frequency are assumed, and all the displacement
fields are considered with respect to neutral axis of corresponding regions. Notations are as
follows, uˆ (i ) is axial wave corresponding to ith region, uˆ1(i ) and uˆ2(i ) are amplitudes of forward
and backward moving axial wave in region-i. Similarly, wˆ (i ) denotes the transverse wave in
region-i, wˆ1(i ) and wˆ 2(i ) are the amplitudes of propagating part and evanescent part of
transverse wave respectively in region-i. kai and kbi are the respective wave number for axial
and transverse wave in ith region. Displacement field due to axial and transverse waves at
neutral axis in base laminate are given respectively
(1)
uˆ (1) ( x,  )  uˆ1(1) eika1x  uˆ2(1) eika1x ( L1  x )
ˆ (1) ( x, )  wˆ1(1)eikb1 ( L1  x )  wˆ 2(1)e kb1 ( L1  x )
w
Displacement field of upper sub-laminate (region-2) is given by
uˆ (2) ( x,  )  uˆ1(2) eika 2 x
(2)
ˆ ( x, )  wˆ e
w
 wˆ e
Displacement field in lower sub-laminate (region-3) is given as
uˆ (3) ( x,  )  uˆ1(3) eika 2 ( x  L1 )
(4)
(2) ikb1 ( x  L1 )
1
(2)
(3)
(2)  kb1 ( x  L1 )
2
(5)
ˆ ( x, )  w
ˆ e
ˆ e
(6)
w
w
Following boundary conditions can be imposed to determine the wave amplitude in different
region in terms of amplitude of incident S0 wave. Continuity in axial displacements at the tip
of the delamination gives
uˆ (1)
 uˆ (2)
 uˆ (3)
(7)
(3) ikb 3 ( x  L1 )
1
(3)
x  L1
(3)  kb 3 ( x  L1 )
2
x  L1
x  L1
Continuity of transverse displacements at delamination tip provides
wˆ (1)
 wˆ (2)
 wˆ (3)
x  L1
x  L1
Continuity of bending slops at delamination tip
dwˆ (1)
dwˆ (2)

dx x  L
dx
1
(8)
x  L1

x  L1
dwˆ (3)
dx
(9)
x  L1
Continuity in axial force at the tip of the delamination
ˆ (3)
duˆ (1)
duˆ (2)
(3) (3) du
Eeff(1) A(1)
 Eeff(2) A(2)
 Eeff
A
dx x  L
dx x  L
dx
1
1
(10)
x  L1
where, E and A are the effective modulus of elasticity and cross section area of ith region
respectively. Effective modulus of elasticity is calculated from
 Ei hi
Eeff 
 hi
where, Ei and hi are the modulus of elasticity in axial direction and thickness of ith lamina
respectively. Equilibrium of shear force gives
(11)
V1 x  L  V2 x  L  V3 x  L
(i )
eff
(i )
1
1
1
where, D is flexural stiffness of beam corresponding to ith region. Equilibrium in bending
moment about neutral axis at x = L1
(12)
M1 x L  M 2 x L  M 3 x L  P2 zn 2  P3 zn3
(i )
11
1
1
1
where, P2 and P3 is the axial force exerted to the base laminate by sub-laminates (region-2 and
region-3 respectively). zn 2 and zn 3 are the distance of neutral axis of upper and lower sublaminates respectively from neutral axis of base laminate. From the boundary conditions we
get a system of equations as given in equation (7-12) to solve for the amplitude of waves in
different region in terms of amplitude of incident S0 wave in base laminate.
3. Wave scattering from tip of a large delamination
An eight layer laminated composite beam is considered for analysis. Material properties of
the laminate are as follows E11 = 144.8 GPa E22 = 9.65 GPa, G12 = 4.14 GPa, G23 = 3.45 GPa,
ν12 = 0.3, ν23 = 0.49, density ρ= 1389 Kg/m3. Cross ply laminate has the stacking sequence of
[(0/90)2]s. And a through thickness semi-infinite delamination is considered, tip of which lies
at L1 distance from free end. Wave amplitude in different region is calculated for various
frequencies and result is compared with TSFEM simulation. For TSFEM simulation beam of
1200 mm length is considered and a long delamination with tip at 600 mm from the free end
extended till the fixed end as shown in the figure-2. Actuators are mounted on both top and
bottom surface of the beam at the tip as shown in the figure-2. Each lamina is modelled using
one element in thickness direction and there are 2400 elements in axial direction. 3rd order
polynomial is used to approximate the solution. And input voltage is applied to both the
actuators in phase which creates the S0 wave. Although axial wave is not dispersive but due to
wave interaction with delamination may create A0 wave which is dispersive in nature. So the
single frequency wave is preferred for interrogation. Therefore five cycle tone-burst signal is
considered to generate narrow-band wave. Responses are captured at the point A and point B
to capture responses which are located at 500 mm and 700 mm away from the free end
respectively. Point A captures the incident wave and reflected wave from the delamination
using transducer A1 and A2, and responses for transmitted wave are captured at point B using
transducers B1 and B2 for region-2&3 respectively. Wave generated at the tip propagates
through the beam and interacts with the delamination. Wave scattering from delamination
depends on its interfacial position.
0
+ -
A1
B1
Delamination
+ Actuator
A2
B2
90
0
90
90
0
90
0
Figure-2: Composite beam with delamination modelled with actuators and stacking sequences
given at right side
Analysis is performed for cases with delamination located at mid-plane (between 4th and 5th
lamina) and away from mid-plane (3rd interface that is delamination located between 3rd and
4th lamina). Wave amplitude in terms incident S0 wave amplitude in different region for
various frequencies are plotted in figure-3to5 for the delamination located at 3rd interface.
From both the approaches it is observed that transverse wave also present in sub-laminates
and base laminate, which indicate the mode conversion happens due to the incident S0 wave.
Wave field in different time is visualised in terms strain distributions in figure-6. As we can
see that as wave reaches to delamination some of the part of wave energy is reflected rest
transmitted. In case of delamination is in mid plane there is no reflection as well as no mode
conversion and wave in both the sub-laminate propagates in same velocity as shown in figure6(a). But when delamination is in 3rd interface there is mismatch in stiffness properties of both
the sub-laminates, reflections and mode conversion happens. Waves in both the sub-laminates
propagate in different velocities. Reflected and transmitted wave contains both S0 as well as
A0 wave. As S0 wave has higher velocity therefore S0 wave propagates faster than the A0
wave as observed in figure-6(b).
(a)
(b)
Figure-3: (a) Reflection coefficient ( uˆ2(1) / uˆ1(1) ) of axial wave and (b) reflection coefficient
( wˆ1(1) / uˆ1(1) ) of A0 wave in region-1
(a)
(b)
Figure-4: (a) Transmission coefficient ( uˆ1(2) / uˆ1(1) ) of axial wave and (b) transmission
coefficient ( wˆ1(2) / uˆ1(1) ) of transverse wave in region-2
(a)
(b)
Figure-5: (a) Transmission coefficient ( uˆ1(3) / uˆ1(1) ) of axial wave and (b) transmission
coefficient ( wˆ1(3) / uˆ1(1) ) of transverse wave in region-3
(a)
Incident S0 wave
Incident S0 wave
(b)
A0 wave
Figure-6: Wave field visualized in terms of strain distribution (εxx) at different time as wave
interacts with delamination for delamination located between (a) 4th and 5th lamina (midplane) and (b) 3rd and 4th lamina for excitation frequency 200 KHz
4. Conclusions
S0 mode Lamb wave propagation in cross-ply with large delamination is modelled using
analytical model as well as time domain spectral finite element method. S0 wave does not
create any reflection in case of delamination is in mid-plane hence is not sensitive to midplane delamination for symmetric cross-ply laminate. Off axis delamination cause the
reflection as well as the mode conversion. Strength of the reflected wave and mode
conversions depends on the difference in effective modulus of elasticity in axial direction
between sub-laminates. If there is no difference in effective modulus of elasticity reflection is
very week.
Acknowledgements
Authors gratefully acknowledge financial support from the Boeing Company, USA to carry
out this research.
References
1.
Su ZQ, Ye L, Lu Y. Guided Lamb waves for identification of damage in composite
structures: A review. J Sound & Vibration 2006; 295: 753-80.
2.
A. Nag, D. Roy Mahapatra, S. Gopalakrishnan and T. S. Sankar. "A spectral finite
element with embedded delamination for modeling of wave scattering in composite
beams", Composites Science and Technology 63, 2187–2200, 2003.
3.
G. Kolappan Geetha, D. Roy Mahapatra, S. Gopalakrishnan, S. Hanagud, "Laser
Doppler imaging of delamination in a composite T-joint with remotely located
ultrasonic actuators ", Composite Structures 147 (2016) 197–210.
4.
Liu Y, Alamusi, Li J, Ning H, Wu L, Hu N, et al. "Locating delamination in composite
laminated beams using the zero-order mode of lamb waves". In: Composites and their
applications. InTech Publisher; 2012.
5.
F. Moser, J.J. Laurence, J. Qu, Modeling elastic wave propagation in waveguides with
finite element method, NDTE 32, 225–234, 1999.
6.
C. Willberg, S. Duczek, J.M. Vivar Perez, D. Schmicker, U. Gabbert,"Comparison of
different higher order finite element schemes for the simulation of Lamb waves",
Comput. Methods Appl. Mech. Engrg. 246–261, 2012.
7.
J.F. Doyle, A spectrally formulated finite element for longitudinal wave propagation,
International Journal of Analytical and Experimental Modal Analysis, 3, 1-5, 1988.
8.
Anthony T. Patera. "A spectral element method for fluid dynamics: laminar flow in a
channel expansion", Journal of computational physics 54, 468-488, 1984.
9.
A. Zak, M. Krawczuk, "Certain numerical issues of wave propagation modelling in rods
by the spectral finite element method", Finite element in analysis and design 47, pp.
1036-1046, 2011.
10.
A Zak, M Radzienski, M Krawczuk and W Ostachowic. "Damage detection strategies
based on propagation of guided elastic waves", Smart Mater. Struct. 21, 035024 (18pp),
2012.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
SHM - Prognostic Analysis of Crack Growth retardation in Fastener Joints
with Bush around the Pin
K. Bharath1, L. Chikmath1, B.V. Sravan2, and B. Dattaguru1
1
Department of Aerospace Engineering, Jain University, Kanakapura Road, Bengaluru® 562112
2
National Aerospace Laboratories, Kodihalli, Bengaluru-560017
Email: datgurb@gmail.com
Abstract
Fasteners are extensively used for connecting components in engineering structures. These fasteners become
ineluctable in numerous situations, particularly for repeated assembly and disassembly. These are identified as
critical locations in the modern damage tolerance based design concept to ensure structural safety and integrity
and Structural Health Monitoring (SHM) has become mandatory at these locations. The basic mechanism
involved in these joints is the pin-in-a-hole for load transfer through a fastener which could cause tensile stress
concentrations around the hole, which may consequently lead to the failure of the structure. Presence of nonlinearity in the form of varying contact/separation at the pin-lug interface due to the load reversals of the pin is a
prominent issue in the analysis of these joints. The current work focuses on fatigue and prognostic analysis of
the tapered lug joints with push fit-rigid pin with and without a metallic bush around pin-lug interface. The
numerical analysis concentrates on bringing out the beneficial effect of inserting a metallic bush, providing taper
to the lug and also evaluate possibility of arriving at a near optimum bush thickness.
Keywords: Lug joint, metallic bush, fatigue, crack initiation and crack growth
Nomenclatures
Hole diameter
Crack growth material constants
Critical crack length
Fatigue strength exponent
Material constants
Regions of contact/Separation
Mode-1 Strain energy release rate
Mode-1 Stress intensity factor
Mode-1 Fracture toughness
Stress concentration factor
τrθ
Length & width of the lug
Pin load
Stress & load ratios
Inner & outer radii of the lug
Lug thickness
Bush thickness
Taper angle
Tangential & Radial stresses
Bearing stress
Shear stress
1. Introduction
Large scale aerospace vehicles are made in parts and assembled using joints. In most
circumstances fasteners are preferred due to the ease of assembly and disassembly. The most
commonly used configurations for load transfer are lug joints and these joints are potential
sources of stress concentration and failure. Hence the implementation of damage tolerance
design concept at joints is mandatory and the modern Structural Health Monitoring (SHM)
system should continuously monitor these locations [1]. The major issue in these joint
problems is the contact conditions between the pin and the hole in the lug. The case of a rigid
push fit pin in lug joint with smooth interface is analysed and presented in this paper (rigid
pin nearly correspond to a steel pin in aluminium lug).
8th International Symposium on NDT in Aerospace, November 3-5, 2016
The joint is subjected to pull (away from the lug and is positive) and push (towards the lug
and is negative) loads. It is well known that in the case of push fit the contact/separation
regions are constant with load level as long as the loading is in either pull or push condition
only [2-3]. However when the fatigue loading has negative load ratio the contact conditions
change when the load changes from pull to push and vice-versa. The stress distributions are
no more linear throughout the loading. In the current paper, the analysis is conducted
independently for both pull and push loading conditions. The applied fatigue load cycle is
mapped on to the stress variation at the maximum tensile stress location(s) and the local
stress cycles are used to determine the life for fatigue crack initiation and life for crack
growth till failure. S-N life is used along with Basquin's relations and Miner’s rule for crack
initiation. Modified Virtual Crack Closure Integral (MVCCI) [4-5] and Paris law modified
with Elber Crack Closure [6] correction is used to find the crack growth life and the
remaining life at any instant of time. Numerical analysis is conducted for a typical tapered
attachment lug joint with a metallic bush bonded to the hole boundary. The configurations
with and without bush for the pull load case are shown in figs.1a and 1b. Metallic lug with
varying taper and bush with varying thickness is analysed to study the effect of bush
thickness and lug taper on fatigue crack initiation life and further crack growth life. Results
are presented to bring out the life enhancement due to lug taper and the inclusion of bush
which have also been investigated in different cases earlier [7-8].
Fig 1a: Push fit lug joint without bush
Fig 1b: Push fit lug joint with bush
2. Methodology
Push fit lug joints are analysed where the pin diameter is equal to lug-hole diameter. In this
joint there is partial loss of contact for an infinitesimally small pin load and these
contact/separation regions do not change with load level as long as the load is either in Pull or
Push loads only and does not change sign. This leads to linear elastic problem [2], but
different problems in Pull and Push load cases. Pin is assumed to be rigid and this nearly
corresponds to a steel pin in an aluminium lug. The regions of separation/contact change
when the fatigue loading is of Pull- Push type with Rp< 0. These lug joints are analysed with
and without bush around the pin for both Pull and Push loads. The interface between the pin
and lug/bush is assumed to be smooth (frictionless). The bush thickness is varied from 2.5- 5
mm. The variation of contact between the pin and the lug for both the cases is shown in figs.
2a and 2b. For the case of Pull load, the contact of pin to bush is 1800 and the region of
contact between the pin and the bush for this Pull load is -900: 00: 900. Similarly for the case
8th International Symposium on NDT in Aerospace, November 3-5, 2016
of Push load the region of contact between the pin and bush is -900: 1800: 900. These lug
joints are analysed for the fatigue load cycles of Rp (Pmin/Pmax) equal to 0, -1,-2 and -3.
Typical load cycles for which the above lug joints are analysed are shown in figs. 3a and 3b.
Fig 2a: Lug joint with bush under Pull load
Fig 3a: Pull-Pull fatigue load cycle (Rp = 0)
Fig 2b: Lug joint with bush under Push load
Fig 3b: Pull-Push fatigue load Cycle (Rp<0)
Critical locations of stress concentrations are identified from the finite element analysis using
MSC NASTRAN/PATRAN software package. The maximum stress concentration location
for the lug joint without bush is shown in fig 4a, where the crack is likely to be initiated. It is
seen that the crack initiates in the lug at the pin-lug interface when there is no bush. When
there is a metallic bonded bush around the pin-hole interface, the crack is prone to initiate at
the bush and lug interface (fig 4b). The crack initiation life is estimated with local stress
fatigue cycle in all the cases using S-N approach and Basquin's relation. Further the crack
growth study is also conducted at the same location for the same fatigue cycles using the
Paris law and MVCCI [4-5] to estimate the fracture parameters.
3. Configurations Analysed
The lug joint analysed for the current analysis is shown in fig. 5 with dimensions. Lug is of
Aluminium T2024 alloy and pin is assumed to be rigid which nearly corresponds to steel pin
in aluminium lug. Outer radius of the lug Ro is kept constant and the inner radius Ri is varied
and (Ro/Ri) ratio is varied as 2, 2.5 and 3. In one of earlier studies, tapered lug joints of
different taper angle were analysed for the improvement of fatigue life for crack initiation and
growth and it was found that the tapered lug joint with taper angle of 22.50 was found to be
the best option [3]. In the current work too lug with taper (β/2) of 22.50 is used for numerical
analysis. Metallic bush made up of precipitation hardened high alloy steel S17700 material is
included and its thickness is varied progressively as 2.5mm, 3mm, 4mm and 5mm [7].
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Fig 4: Crack initiation locations for Pull load with and without Bush
Fig 5: Configuration of the lug joint analysed
4. Boundary conditions
The load applied to the pin gets distributed over the region of contact between the pin and the
lug with the resultant in the direction of and equal to the load applied. In this problem the
load can be applied as a rigid body displacement of the rigid pin. Then estimate the load
transferred to the lug by summing up the reactions at the far end of the lug. This is further
simplified without any loss of generality by superposing a rigid body displacement on the
configuration so that the pin is held with zero displacement and the rigid body displacement
is applied at the far end of the lug. Again the load transferred is computed by summing up the
reactions at the far end of the lug.
Since the problem is symmetric about the x-axis, the symmetric lug portion with bush for Pull
and Push load case is shown in fig.6. A convergence study is carried out and it was decided to
run the lug joint for further analysis with 6012 elements and 6193 nodes. Finite element
model of tapered attachment lug without and with bush is shown in figs. 7a and 7b. In order
to maintain the 900 degree contact/separation between the pin and the hole boundary of the
lug, the nodes on the hole boundary should satisfy appropriate inequality constraints in
contact/separation region. The boundary conditions and inequality constraints in the region of
contact and separation are given in eqns.(1) and (2) respectively. Also the boundary
conditions at the far end after the superposition of rigid body displacement on the
8th International Symposium on NDT in Aerospace, November 3-5, 2016
configuration are given in eqns. (3) and (4). Further the load equilibrium condition is given in
eqn. 5.
Fig 6: Boundary conditions for Pull and Push loads
Fig 7: FEM models in half (symmetric) of the configuration
Region of contact
,
inequality constraints,
.............(1)
Region of
,
inequality constraints,
...........(2)
separation
Far end for the case of Pull load
.......................................(3)
Far end for the case of Push load
.......................................(4)
The load equilibrium condition in both lug and bush:
Pin load
∫
,
.........................(5)
5. Static analysis
5.1 Maximum stresses at pin-bush interface or lug-bush interface
Maximum tensile stresses at pin-bush interface and lug-bush interface for the applied rigid
pin displacement are shown in Table 1. It is seen that for the same load rigid pin-bush
interface experiences higher stresses in comparison to bush-lug interface. However, the
metallic bush is made up of high steel alloy and has higher tensile and ultimate strength limit
rather than the bush-lug interface, where the lug is made up of aluminium. As a result bushlug interface is more susceptible to stress concentrations leading to damage. Hence, further
analysis is conducted with crack initiating from bush-lug interface. Fig. 8 shows typical
maximum tensile stress variation at the bush-lug interface for Pull and Push load. It is also
seen that tensile σmax is relatively higher for Pull load in comparison with Push load. This is
8th International Symposium on NDT in Aerospace, November 3-5, 2016
primarily because the load flow from the load point to the far end goes around the pin in case
of Pull load whereas the load flow does not go around the hole in case of Push loading.
Table 1: Stresses at pin-bush and
bush-lug interface
Fig 8: Variation of tensile stress for Pull
and Push loads
5.2 Radial and tangential stress distributions
The stress distributions have been studied, but not presented in this paper for the sake of
brevity. Radial stresses are found to be highest for smaller bush thickness of 2.5mm
compared to other bush thickness of 3, 4 and 5mm which imply that they are the highest for
smaller thickness of the bush. Subsequently tensile tangential stresses are also computed at
bush-lug interface and are normalised with the bearing stresses It is observed that the
maximum peak points occur at two different locations (±900 degree). It is found that bush
thickness with 4mm and 5mm are relatively close enough, which implies a possibility of
choosing the best option bush thickness for this lug configuration.
5.3 Stress concentration factor (SCF)
The Stress Concentration Factor (SCF) is estimated for Pull load case and is normalized with
the bearing stress (σbr=P/2Rit, t=thickness of the lug). For Pull loading (fig. 9a), it seen that
with increasing Ro/Ri (keeping Ro constant and making the hole smaller) ratios the stress
concentration factor decreases due to increase in bearing stress as Ri decreases (σbr α 1/Ri).
From the figs. 9a and 9b, it is observed that lug joint without bush shows higher stress
concentration and this SCF is much less with the inclusion of bush. As a limiting case, SCF is
computed for straight lug and compared with the tapered lug with 22.50 with bush of 4mm
thickness. Straight lug has higher stress concentration factor (31% more) in comparison with
attachment lug with taper angle 22.50. It is so obvious that when the taper angle is increased
keeping Ro constant, the load flow will be smoother into a wider lug, and this causes lesser
SCF. Also inclusion of 4mm metallic bush for these lug joints minimizes the SCF by 34%
over the case of no bush.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Fig 9a: SCF for tapered lug with various bushes
Fig 9b: Effect of taper and bush on SCF
6. Fatigue analysis
6.1 Crack initiation
Fatigue crack initiation life analysis is carried out using the Basquin's equation and S-N
approach at the critical nodes identified from the stress analysis. Damage per each applied
cycle is estimated using Miner's rule [9-10]. The fatigue cycle is derived from the local stress
variation at the maximum tensile stress location. Respective local stress variation due to the
applied load cycles are qualitatively shown in the figs. 10a and 10b.
Fig 10a: local stress variation for
applied load ratio Rp = 0
Fig 10b: local stress variation for
applied load ratio Rp = -1
It is seen for both Pull and Push loadings, the tangential stress at the critical locations is
tensile, but of a smaller magnitude for Push loads. It is observed that for R= - 1 the local
maximum tensile stress at the critical nodes is too small for the Push load in comparison with
Pull load. Hence in case of Rp = - 1, the contribution of push load portion towards the fatigue
failure is not significant. The crack initiation life is shown for Rp = 0 (figs. 11a and 11b).
Initially for smaller loads, the tapered attachment lug joint with varying metallic bush possess
very high life and damage is negligible. With the progressively increasing load P max, there is
a drop in the fatigue life with increase in damage. Lug joint with higher bush thickness have
better crack initiation life. Further straight and tapered attachment lug joints are analysed for
without and with metallic bush (fig. 11b). It observed that providing taper of 22.50 for the lug
8th International Symposium on NDT in Aerospace, November 3-5, 2016
increases the maximum load taking capacity by10.2% for crack initiation compared to
straight lug. Also insertion of metallic bush of 4mm to the straight lug increases the load
taking capacity by 30.6% in comparison to straight lug without bush. Tapered lug with bush
shows better fatigue crack initiation life in comparison to other cases.
Fig 11a: Crack initiation life for tapered lug
with various bush thicknesses
Fig 11b: Effect of taper and bush on
crack initiation life for straight lug
Since the stresses are too low at the critical locations for the negative P (Push), its effect is
insignificant to the total fatigue crack initiation life. The effect of push load could be
significant at higher negative values Rp. Crack initiation life is compared between load ratio
Rp=0, and -2 for the case of tapered lug with 4mm metallic bush in Table 2. For Rp= -2, the
damage initiates at about 8% lesser maximum load when compared load to Rp=0. At the load
ratio of Rp= -3 the maximum stresses exceed the yield strength and not considered.
Table 2: Comparison of fatigue crack initiation life for different load ratios
Tapered lug
Max. load in
fatigue cycle, N
Rp=0
3525
3760
3930
3995
4230
5.54E+05
2.99 E+05
1.95 E+05
1.66 E+05
9.71 E+04*
0
(β/2=22.5 )
with 4mm
metallic bush
Life in cycles
Rp= -2
2.93E+05
1.58 E+05
9.67 E+04*
Rp= -3
Yielding of
lug at critical
locations
*Crack initiation
6.2 Crack growth
It is assumed that a crack length of 2mm is the initial crack size at the initiation point which
is quoted as smallest crack size that can be detected with 95% Probability Of Detection [11].
Further cycles to reach the critical crack length are estimated using Paris law considering
Elber crack closure modification [6]. Mode-I is found to be dominant from the direction of
8th International Symposium on NDT in Aerospace, November 3-5, 2016
crack growth. Fracture parameters at the crack tip are computed using MVCCI for
unsymmetrical case about crack path axis. MVCCI equations are well known [4-5] and the
material constants are given in Ref. [12].
It is also verified that crack growth is negligible during negative load ratios. Lug joint with
higher bush thickness results in higher crack growth life till failure in comparison with other
lesser bush thicknesses (fig. 12a). The crack length at final failure is obviously larger for
thicker bush. Also comparison is made between the straight and taper lug joints without and
with bush of 4mm in fig. 12b. It can be observed that providing taper to the straight lug
increases critical crack length by 21.4% and cycles to reach critical crack length by 94.6%. It
seen that the tapered lug with bush with taper angle of 22.50, fails at 28.6% more crack
length and subsequently at much larger life 153.7% compared to the case of without bush for
the same applied load cycles. Also adding the metallic bush of 4mm to the straight lug
extends the life by 25.5% (34%) and the life in cycles to reach this critical crack length by
223.9% over straight lug without bush. Further, tapered lug with bush extends the crack
length by 16.5% and cycles to reach critical crack length by 52.5% over straight lug with
bush. Hence tapered lug with metallic bush shows higher cycles to failure than other cases.
Fig 12a: Crack growth life for tapered lug
with different bush thickness
Fig 12b: Effect of taper and bush on
crack growth life for straight lug
7. Conclusions
Lug joints which are primarily associated with large scale structural components for load
transfer need a vigilant application of prognostic analysis. The major problem in these lug
joints is contact variation between pin and the lug due to application of pin load with load
reversals and it is addressed in this paper. Typical tapered attachment lug with and without
metallic bush is carried out to study the life enhancement effects. Maximum tensile stresses
at pin-bush interface and bush-lug interface are prudently monitored and it is found that lugbush interface is more prone to initiation of crack like damages during fatigue loading and the
8th International Symposium on NDT in Aerospace, November 3-5, 2016
cracks grow from there till failure. For the case of without bush the crack initiates and grows
from pin-lug interface. It is seen that lug taper and the presence of metallic bonded bush
significantly increase life of the joint configuration. The push load portion of the fatigue
cycle has insignificant effect on the fatigue life for the load cycle provided unless the
negative part of the cycle is large (Rp ≤ - 2). The results of the paper are of significance to
prognostic part of the Structural Health Monitoring of Joints.
References
1. S Gopalkrishnan, M Ruzzene and S Hanagud, 'Computational Techniques for
Structural Health Monitoring', Springer, London Dordrecht Heidelberg, New York,
2011.
2. A K Rao, 'Elastic Analysis of Pin Joints', Computers and Structures, Vol 9, pp 125144, 1978.
3. L Chikmath, 'Prognostic Analysis of Fasteners in Lug Joints', Ph.D. Thesis,
Department of Aerospace Engineering, Jain University, India, 2016.
4. E F Rybicki, and M F Kanninnen, 'A Finite Element Calculation of Stress Intensity
Factors by a Modified Crack Closure Integral, Journal of Engineering Fracture
Mechanics, Vol 9, pp 931-938, 1977.
5. T S Ramamurthy, T Krishnamurthy, K Badari Narayana, K Vijayakumar and B
Dattaguru, ‘Modified Crack Closure Integral Method with Quarter Point Elements,
Mech. Res. Communications, Vol 13, No 4, pp 179-186, 1986
6. W Elber, 'The Significance of Fatigue Crack Closure, In:Damage Tolerance in
Aircraft Structure, ASTM STR 486, pp 230-240, 1971.
7. J Y Mann, A S Machin, W F Lupson and R A Pell, 'The Use of Interference-fit Bolts
or Bushes and Hole Cold Expansion for Increasing the Fatigue Life of Thick-Section
Aluminium Alloy Bolted Joints', ARL-Structures-Note-490, Melbourne, 1983
8. K Bharath, B V Sravan, L Chikmath and B. Dattaguru, 'SHM- Prognostic Analysis of
Tapered Attachment Lugs under Fatigue Loading', Proceedings of SICE conference
organised by Indian Structural Integrity Society (InSIS), Bangalore, June 2016.
9. S Suresh, 'Fatigue of Materials', Cambridge University Press, UK, 1998.
10. T L Anderson, 'Fracture Mechanics: Fundamentals and Applications', CRC Press,
Taylor & Francis Group, Boca Raton, 2000
11. J Haapalainen, and E Leskela, 'Probability of Detection Simulations for Ultrasonic
Pulse-Echo Testing, 18th World Conference on Non-destructive Testing, Durban,
South Africa, pp 16-20, April 2012.
12. J Gillbert Kaufman, 'Properties of Aluminum Alloys- Fatigue Data and the Effects of
Temperature-Product Form and Processing, ASM International Data Handbook, 2008.
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Multiple NDE methods for crack characterization on spur gear
M.R.Vijaya Lakshmi*, A.K.Mondal, Shubhanjali, M.V.Subbaraju, V.Thangavelu
Sreelal Sreedhar
Gas Turbine Research Establishment, Bangalore – 93, India
Phone: +91-080-2504 0632, Fax: +91-080-2524 150, e-mail: vijayalakshmi_mr@gtre.drdo.in
Abstract
Spur gears are critical rotating components being developed using S132 material for transmitting drive to the
fuel pump in the fuel system of aero engine. During prototype manufacturing, circumferential cracks were
observed by the operator on 02 gears. These gears were initially subjected to Magnetic Particle Testing.
However, complex shape of the gear and difficulty in accessing the cracked region necessitated further testing to
ascertain the depth of the crack. Since Eddy Current Testing is the most suitable method to detect and size
surface breaking discontinuities in conductive materials, it was decided to adopt this method for further defect
characterization. Customized calibration block was developed with notches of different depths for amplitude
calibration and defect sizing. Experiments to differentiate signal from edge and defect were also carried out.
Further, the cracks on the gears were successfully quantified and provided vital information for repair and usage
of gear.
Keywords: Magnetic Particle Testing, Eddy Current Testing, Spur Gear, Aero Engine
1.
Introduction
Non-Destructive Testing methods play an essential role in determining the reliability and
integrity of parts throughout the product life cycle. One or more NDT methods are used
during various stages of manufacturing to ensure that defect free material is processed further
thus avoiding wastage of time, resources and expenditure. The criticality and complexity of
NDT is significantly different for aero engine parts owing to their design, methods of
manufacturing, operating environment, and safe life management issues. Typical faults and
defects targeted in NDT of gas turbine components include original defects and deviations
from manufacturing or repairs, and defects of coatings and base materials emerging and
growing during service[1]. Quantitative NDE is essential for defect characterization and
provides valuable information for further investigations [2].
The present paper discusses one such investigation and characterization of a
crack noticed on a spur gear during manufacturing. Spurs gears are the simplest and most
common type of gears. They are used to transfer motion between parallel shafts, and they
have teeth that are parallel to the shaft axes [3]. Figure 1 contains the photograph of the spur
gear discussed in this context.
Page 1 of 7
Fig 1: Photograph of spur gear
The spur gears (02 no.s) were manufactured from bar stock of BS S132 Nitriding Steel.
During the finishing operation, the operator noticed circumferential cracks on the surface on
both the gears. As this was the development phase of this product including pre-decided
timelines to understand their performance, it was essential to salvage the gears to put to
further use. So, there was a need to estimate the depth of the cracks for carrying out the
rework due to limited availability of grinding stock. Multiple NDE methods were resorted to
characterize and quantify the cracks. The following section describes the methodologies and
techniques.
2. NDE methods
2a. Magnetic Particle Testing
The gears were initially subjected to Magnetic Particle Testing on a stationery wet horizontal
magnetic crack detector. The parts were subjected to longitudinal magnetization (coil shot) as
per the principle of Figure 2. The circumferential crack with multiple branches was distinctly
observed in both the components and is shown in Figure 3.
Fig. 2: Principle of longitudinal magnetization [4]
Page 2 of 7
Fig. 3: Circumferential intermittent crack with branches
However, Magnetic Particle Testing is qualitative method and is has the inability to estimate
the depth of the crack. Hence, it was decided to inspect the gear by Eddy Current Testing to
characterize the crack.
2b. Eddy Current Testing
Eddy Current Testing works on the principle of Electromagnetic induction. The principle of
the method along with the behavior in the presence of anomalies in illustrated in Figure 4.
Fig. 4: Principle of eddy current testing (left) and distortion of eddy current due to crack,
edge effect, surface crack and sub-surface void (right) [5]
Calibration of eddy current measurement systems is an important factor for attaining the
accuracy and precision of measurement that quantitative non destructive evaluation requires
[6]. In principle it is possible to estimate crack depth by comparing the eddy-current signal
from an unknown crack against data from a calibration crack of the same surface length in the
Page 3 of 7
same component, presuming the depth of the calibration crack is already known and that
materials factors such as crack closure, crack branching and crack-face contact are
equivalent[7].Narrow notches produced with electron discharge machining (EDM) and saw
cuts are commonly used to represent cracks, and drilled holes are often used to simulate
corrosion pitting[8].Customized calibration block of S132 material was manufactured for
calibration purpose having notches of depth, 0.3, 0.5 and 1.0mm. The schematic of the block
is depicted in Figure 5. The actual block is shown in Figure 6.
Fig. 5: Schematic of reference standard
Fig. 6: Photograph of Calibration block
A 10 MHz absolute probe was used along with EddyMax Eddy Current Equipment
(Тest Мacshinen Tecknik, Germany). Further, system was calibrated using this block.
The response from the artificial notch of 1.0mm notch is depicted in figure 7.
Page 4 of 7
Fig.7: Eddy current signal response to artificial notch of depth 1.0mm
The spur gears were then inspected using the same settings. In eddy current testing, edge
effect of the material produces a large amplitude signal, masking the defects present at the
edge region [9]. Initial difficulty was faced to understand and distinguish the signals from
interface and defective regions. After assuring repeatability of measurements, the spur gears
were inspected in the defective region to measure the depth of the cracks. It was observed
that on Sl.No.1, the crack depth varied from 0.5 to 1mm (approx.) (Figure 8) and on Sl.No.2,
it varied from 0.5mm to 1.5mm (approx.) (Figure 9).
Fig. 8: Signal from crack in Sl.No.1
Page 5 of 7
Fig. Signal from crack in Sl.No.2
Fig. 9: Signal from crack in Sl.No.2
Based on the results, the gears were ground to remove the crack as around 2mm
grinding stock was available. The actual depth of the cracks was found to be in close
agreement with the measurement obtained by eddy current testing. Further, after rework,
Magnetic Particle Testing was repeated on both the gears. No surface cracks were observed
and were found satisfactory. It was proven that Eddy current testing can be successfully
complemented to other Non-destructive testing methods for quantifications of defects.
3.
Conclusions
Magnetic Particle Testing and Eddy Current Testing were utilized in conjunction to
characterize the surface cracks observed on spur gears. The gears were successfully reworked
and used for proving the concept of gear pump. Control of the inspection variables in eddy
current testing along with modeling and manufacture of calibration blocks to simulate the
actual defects will further reduce the uncertainties and improve the reliability of inspection.
4.
References
1.
M.R.Vijaya Lakshmi et.al., ”Overview of NDT methods applied on an aero engine
turbine rotor blade”, Insight, British Journal for Non-Destructive Testing, Vol 55, No
9, Sep 2013pp 482 - 486
M.R.Vijaya Lakshmi et.al., “Quantitative NDE of Aero Engine Turbine Rotor Blade
– A case study”, Proceedings of National Seminar & Exhibition on Non-Destructive
Evaluation (NDE 2011), 2011, Chennai, India, pp 301 - 304
Mechanical Engineer’s Handbook, 2001, pp 253
www.nde-ed.org/ EducationResources/ CommunityCollege/ MagParticle/ Physics/
Magnetization.htm
B.P.C.Rao, “Eddy Current Testing:Basics”, Journal of Non-Destructive Testing &
Evaluation, Volume 10 , Issue 3, pp7-16, 2011
J.C.Moulder et.al.,”Calibration methods for eddy current measurement systems”,
Contirbution of the National Bureau of Standards
S.K.Burke, “Crack depth measurement using Eddy Current-NDE”, 10th Asia-Pacific
Conference on Non-Destructive Testing, Brisbane, Australia, 2001
2.
3.
4.
5.
6.
7.
Page 6 of 7
8.
9.
www.nde-ed.org/EducationResources/CommunityCollege/EddyCurrents/ Procedures/
Reference Standards.htm
B.Sasi et.al, “Dual-Frequency Eddy Current Non-destructive detection of Fatigue
Cracks in Compressor Discs of Aero Engines”, Defence Sceince Journal, Vol.54,
No.4, October 2004, pp563-570
Page 7 of 7
8TH INTERNATIONAL SYMPOSIUM ON NDT IN AEROSPACE, NOVEMBER 3-5, 2016
ON HEALTH MONITORING OF INSERT JOINTS IN SPACECRAFT
STRUCTURE
P. Sathish Kumar, A. Ananthan, T.S.Srirangaa, Krushna Chandra Dakua, S.Shankar Narayan
a
Email:rangan@isac.gov.in
Indian Space Research Organization, Satellite centre, HAL-Airport road, Vimanapura P.O
Bangalore 560017
Abstract
Inserts are preferred choice to transfer concentrated loads into honeycomb sandwich panels. Such panels with
inserts are used widely in spacecraft structures due to their high bending stiffness to weight ratio. There are
numerous insert joints in a structure designed to integrate payload electronic packages, solar array and structural
decks to central thrust cylinder. Structural integrity of these inserts is essential for ensuring alignment of
payloads, proper deployment of solar arrays, relative position of antenna reflector and feed. Often during or after
assembly of the structure, occasions arise to assess the health of suspect inserts in comparison to adjacent inserts
or a pre established database. Essentially health of the insert joint is required to be assessed in-situ. The basic
tenet of the work proposed here is that frequency response of the suspect insert under dynamic excitation gets
altered in relation to its residual strength. This has been verified by tests on coupons with varying insert potting
process parameters and results are explained here. The outcome of this work is a methodology for health
diagnostics of insert joints at the spacecraft assembly level. A full-field, non-contact technique, Speckle Shear
Interferometry, is employed for measuring shift in frequency signature by measuring strain and displacements
under resonant condition. The test equipment being portable, this method can be used for similar applications in
aerospace field.
Keywords: Health Monitoring, Diagnostic technique, Insert joint, spacecraft structure, dynamic excitation,
resonance, Shearography.
1. INTRODUCTION
Honeycomb sandwich panels are extensively used in spacecraft structures due to their high
specific bending stiffness. Inherently such panels cannot take high concentrated loads. A
local reinforcement of the core, usually in the form of metallic inserts, is required where a
joint is to be made. Inserts are bonded in the core using an adhesive potting compound. There
are numerous such insert joints in a structure that are used to integrate payloads to the
structure. For example horizontal decks are connected to central cylinder through rings and
inserts, solar arrays have insert interfaces to vertical decks, spacecraft handling inserts, etc.
Often after assembly of the structure, occasions arise to assess the health of suspect inserts in
comparison to adjacent inserts. Suspect inserts may have been accidentally overloaded due to
wrong assembly or wrong handling.
In this work a non contact measurement technique has been applied to make such an
assessment. Basic tenet of the concept is that frequency response of the suspect insert gets
altered in direct relation to its residual strength. This has been verified by tests on coupons
with varying process parameters and results are explained here.
In order to study the sensitivity of the insert pull-out load and its resonant frequency, three
process parameters were identified to effectively create insert joints of varying strengths and
resonant frequency.
1.1. Methodology for Health Monitoring
The three parameters are namely the geometry of the potting compound close to the insert,
resin and hardener mix ratio in the adhesive and the micro-balloon content in the adhesive.
Micro-balloon is a low density material added to the adhesive in order to bring down the
weight of the adhesive. The correlation between insert resonant frequency and pull-out load is
determined experimentally and a database is created.
Shearographic technique was used to determine the frequency shift.
The objective of the work is to study the sensitivity of the insert joint resonant frequency and
insert pull-out load by varying one parameter at a time. This enabled availability of insert
joints of varying stiffness, hence of varying resonant frequency and of varying strength
capacity. Also, several insert joints of nearly same strength and stiffness were made under
normal tight process control. Measurements on this set of inserts provided frequency changes
due to statistical variation that anyway has to be expected among insert joints made with tight
process control. It is only the frequency shifts, over and above this variation, that are of
practical interest.
The outcome of this work is a methodology for health diagnostics of insert joints at the
spacecraft assembly level by a non contact technique capable of determining shift in
frequency signatures.
2. Test setup
Speckle Shearography is the technique employed in this work. This technique, depending on
optical setup, can pick up out- of- plane displacement gradients of sub micron range. This
setup is being used for detecting surface or sub surface defects in spacecraft propellant tanks
during pressurisation for static deformations [1]. For the current work, a setup is made to
measure the dynamic deformations. This
technique is insensitive to environmental
disturbances - which are leveraged in the current proposed work- It is not easy to have a
vibration free environment in an assembly shop of structure.
The working principle of shearography [2] is as follows: The test specimen containing the
insert joint in honeycomb sandwich is illuminated by laser light and observed by a CCD
camera. A shearing element introduces required shear. Consequently the camera sees the
usual and a sheared speckle image of the object. These two images interfere on the sensor area
and result in another speckle field, which carry the gradient in the shear direction. This is
useful information in experimental stress analysis since this is proportional to strain.
Piezo Actuator for
dynamic excitation
Laser source &
camera
Insert Joint in honeycomb
sandwich panel.
Fig.1: Optical Shearography test set up
2
Shearography requires the component to be stressed by one of many means such as: vacuum,
pressure, heat or mechanical or vibration [3]. In the present work dynamic excitation for
stressing is employed in keeping with the already stated objective of detecting frequency shift.
The test specimen is kept in a vibration free table which is shown in Fig.1 and a vacuum cup
with piezo electric actuator is connected to the rear of the test specimen. The vacuum cup is
used to hold the piezo electric actuator with the test specimen during the test. The frequency
input to the piezo electric actuator is controlled by ISYS software. The frequency sweep can
be setup to be in auto or manually controlled.
During the process of health monitoring, the deformations or motion of vibrating test
specimen is captured by CCD camera. Specifically w- deformation gradient w/x is captured
by the laser light reflected from the specimen and the same is super-imposed with the sheared
(x) speckle pattern. This manifests on the PC screen in the form of black and white fringes
as real time speckle shear correlation fringe pattern. The corresponding input frequency in
which white and black ‘fringes’ appears is the resonant frequency of the insert.
2.1. Details of Experiments
The M4 insert is put in an aluminium plate of dimensions 100*100*20mm with 12mm dia and
15mm deep hole in the center. By introducing the insert in aluminium plate and having 10%
of resin as micro-balloon in all the test specimens, the parameters geometry and micro-balloon
content are eliminated from the sensitivity study and the parameters are influenced only by the
mix ratio of the resin and hardener. The different mix ratios used in the test specimens are
listed in table 1 below.
The adhesive used for potting insert in sandwich or aluminium plate has the following
constituents.
Hardener
: CAT43
Resin
: STYCAST 1090SI
Micro-balloon: 10% of resin in all test specimens.
Similarly test coupons with insert joints of varying stiffness due to variation in the other two
parameters of mix ratio and micro balloon content are also fabricated and excited on the test
bench to determine their frequency [4].
2.2. Test results
During the test, the frequency is swept automatically up to 40 KHz with a sweep rate of 25Hz.
Once the input frequency given in the form of electrical signal to piezo electric actuator equals
the frequency of the plate or insert, white and black fringes will appear in the screen. Using
manual sweep option the excitation is fine tuned to 1Hz accuracy. The mode shapes of insert
and plate can be distinguished from the fringe pattern of the shearograms. Only the insert
mode shapes are captured and the corresponding frequencies are listed in table 1 below for
various test specimens fabricated under controlled process parameter changes.
2.3. Identification of natural frequency
Once the input frequency equals the natural frequency of the plate or insert, white and black
fringes will appear in the screen. This image is further processed by filtering and
demodulation [5]. The 3D animation of the mode shape was also studied. The shearography
pictures showing the mode shapes of the insert are given below in Fig. 2a & b.
3
Fig.2a. Shearograph of insert mode shape.
Fig. 2b. 3D Visualisation showing
insert under resonance.
Table 1: Resonant frequency and insert pull-out load of various test specimens. b Parameter #
1 : Aluminium coupon Mix ratio, c Parameter #2 : Sandwich Micro Balloon (MB) content
variation, dParameter #3: Sandwich mix ratio (MR).
Test Specimen
name
Parameter
number varied
Variation
In %
Insert resonant
frequency (In KHz)
Insert pull-out
load (In Kgf)
TS-1
TS-2
TS-3
TS-4
TS-5
TS-6
TS-7
TS-8
TS-9
TS-10
TS-11
TS-12
TS-13 to TS-38
(26 inserts)
#1 b
1
1
1
MB #2 c
2
2
2
2
MR#3 d
3
3
10
8
6
5
5
8
10
15
20
8
10
12
24.860
24.740
23.480
23.170
6.922
7.320
7.275
7.692
7.893
7.136
7.275
7.505
320
302
245
224
182
174
170
169
167
183
170
159
9.650-9.660
170 to 171
2.150
38
8.690-8.700
158 to 172
TS-39
TS-40
Normal process
Very poor
potting
compound
Parameter 1 &
2 combined
100:5:5
2.4. Sensitivity analysis
As seen from Table 1, this work involved fabrication of several test coupons of varying
strength and stiffness with controlled process parameters. Frequencies of each such insert
joint was determined by Shearography The Table-1 also shows, when the amount of
hardener per 100grams of resin is reduced by 5g from the nominal value, the resonant
frequency falls by 1.69 KHz and insert pull-out load falls by 96 Kgf from that of nominal
values. It is seen, as expected, dynamic and strength are sensitive to resin and hardener mix
4
ratio & other chosen parameters. The normal expected dispersion in measured frequencies of
many inserts of equal strength made under strict usual process control is also indicated and
small compared to the tested insert coupon dynamics and strength [5].
Fig.4 captures the correlation between parameters vis-a-vis frequency and pull-out strength of
the insert.
INSERT PULL OUT LOAD COMPARISON
INSERT RESONANT FREQUENCY COMPARISON
195
7.6
SW - MICRO
BALLOON
7.1
SW - MIX
RATIO
6.6
0
5
10
15
20
25
AMOUNT (grams) OF VARIED PARAMETER IN 100GRAMS OF RESIN
INSERT PULL OUT LOAD AT FAILURE (KGF)
INSERT RESONANT FREQUENCY (KHZ)
8.1
180
SW - MICRO
BALLOON
165
SW - MIX
RATIO
150
0
5
10
15
20
25
AMOUNT (grams) OF VARIED PARAMETER IN 100GRAMS OF RESIN
Fig. 4: Correlation of insert resonant frequencies and insert pull-out failure loads with
parameters micro balloon content and mix ratio
Fig.5: Processed image showing resonance of inserts of equally good strength.
Fig a) Raw image b) Image showing strain pattern.
Fig.5a & b shows the ability of Shearography to determine the frequencies of 4 adjacent
inserts of nearly same stiffness and strength.
3. CONCLUSION
A viable method of health monitoring of insert joints in spacecraft structural assembly has
been experimentally verified. Basic tenet of the work proposed here is that frequency response
of the suspect insert under dynamic excitation gets altered in direct relation to its residual
strength. This has been verified by tests on coupons with varying process parameters and
results are explained here. The outcome of this work is a methodology for health diagnostics
of insert joints at the spacecraft assembly level.
Acknowledgement
Authors wish to thank Dr. Anand Kumar Sharma, Deputy Director of Mechanical Systems
Area, ISRO Satellite centre and Dr. K Renji, Group Director for kind permission to present
this work. We also would like to thank H. A. Venkatesh Prasad, Govinda Raja.T.S, Harsh
Kumar, Krushna C D, Structures Group for their valuable suggestions and also Ms. Chitra S,
5
R.Rajesh, Lokesh A.H, ChaluvaRaj.C.M in carrying out strain gauging activities and the
operation of UTM.
References
1. Laser Based Interferometric and Instrumented Coin Tap Technique for NDT of
honeycomb sandwich panels, Workshop for ISAMPE at NAL Bangalore, Feb 1996. R
Samuel et.al
2. Measurement of slopes and structural deflections by speckle shear interferometry,
Experimental Mechanics, 14,281-285, 1974. Hung, Y.Y and C.E. Taylor.
3. Mechanical measurements by R.S.Sirohi & H.C. Radha Krishna, 3rd ed. Wiley Eastern.
4. Acoustic Emission studies on adhesively potted inserts of honeycomb sandwich panels,
Sriranga. T.S., J of Acoustic Emission, Vol.7, No.3, July-Sep, 1988, Kalpakkam,
IGCAR,Int. Natl.conf.on NDT.
5. ISRO Internal Technology Development Programme Report on Diagnostics of Insert
joints using Shearography, under preparation, A.Ananthan & Krushna C.D, 2016.
6
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Thermo Elastic Deformation Measurements on Spacecraft Components
using Non contact Target- less Image Based Technique
Raghunatha Behara, T S Srirangaa, CS Varghese, C Koteshwar Rao, Swapnil Pathak,
Pravesh Mathur, Krushna Chandra Dakua.
Email for correspondence: arangan@isac.gov.in
ISRO Satellite center, HAL-Airport Road, Vimanapura Post, Bangalore 560017
Abstract
Communication Antenna reflectors need to maintain their shape and be dimensionally stable under on-orbit
thermal and vacuum conditions in space. The RF performance of such antenna reflectors is crucial for data
transmission from communication spacecraft.
To verify the design and dimensional stability, KFRP reflectors are subject to specified thermal vacuum
conditions in a special chamber. Photogrammetry is used to measure the shape of reflectors under soak
conditions of hot and cold temperatures. The technique depends on physical retro reflective targets mounted on
the reflector surface and by suitable image processing the 3D coordinates and hence the shape in an RMS sense
is computed. In this work, new technique of contactless projection dots are used in lieu of physical targets and
shape is measured. A study of uncertainty in such measurements, effect on the calibration of cameras, is
presented here. Network of cameras used simulates the camera geometry obtained at the Thermal chamber. Invar
scale bars were used to validate the accuracy of measurement. Typically 50 micrometer accuracy is expected
from such shape measurements. The new method has the additional advantage of no compensation for thickness
of physical targets, increased density of targets, no risk of loss of targets due to air entrapment, reduced test
preparation time, and no adhesive residue on removal of targets.
Keywords: Data evaluation, Image Processing, Deformation measurement, Photogrammetry, Projection dots,
Thermal cycling.
1. Introduction
It is well recognized by Structural designers that Thermo Elastic and Hygroscopic Distortions
(TED) need to be minimized. This is especially true in structural applications requiring
dimensional stability such as payload support structures and communication antenna
reflectors made of composite materials. Generally this is achieved by proper choice of
materials (CFRP, KFRP), proper lay-up sequence of lamina, and other constructional aspects.
One example of design goal could be a matched Coefficient of Thermal Expansion at various
parts of the structure. Although TED is small, its knowledge is required and especially
important for communication satellite antenna reflectors.
The system includes: a) high resolution 8 Mega pixel digital camera (Fig 1) mounted in front
of a Quartz optical window of the 6 meter thermo vac chamber, b) Invar scale bars, c) rotating
and flipping fixture (MGSE) thermally decoupled from a test article mounted on it using
Isostatic mounts, d) Photogrammetry software to compute the 3 D positions of the retro
reflective targets mounted on the test article, e) Thermal control of the fixture (Fig 2). The
Invar scale bars are provided with heaters so that their temperatures are not too far away from
ambient and distortion in the scale bar itself is small and easily computed from CTE and
Thermo couple data. The camera has onboard processor to do image processing tasks such
computing number of coded targets detected in an image and estimation of camera intrinsic
1
and extrinsic parameters. The camera can store about 500 images in memory card. The
camera settings are fixed focus to infinity, exposure and strobe powers adjustable, with a ring
flash so that retro reflective targeting is effective. The rotating and flipping fixture enables
measurement on both front and rear shell of a dual grid reflector. The test article mounted on
the fixture is slowly but continuously rotating at 50 degrees per minute so chosen as not to
cause image blur due to motion of targets. Every 10 deg, a camera image is taken and after 36
images, the camera is rolled and another set of 36 images is acquired over another 360 deg
rotation of test article. This roll of the camera (CCD array) is essential for proper estimation of
the camera optics parameters and camera location (by a process called Self-Calibration).
Fig 1: Phogrammetry system
Fig 2: Reflector on MGSE with thermal control
system
2. Description of existing TED system
A digital camera based Photogrammetry system is now part of measurement facility installed
at ISITE 6 meter thermo vacuum chamber for making Thermo Elastic and Hygroscopic
distortion (TED) measurements on Composite structures. Photogrammetry is a whole field
non-contact type of measurement of shape, distortions of an object from multiple images. This
technique has gained lot of attention and interest in aerospace industry due to technological
advances in digital cameras, faster Image processing and automation of the Photogrammetry
measurement process. Initially the object under measurement is mounted on a special fixture
(Fig 2) and then retro-reflective targets of 6mm diameter are stuck on the surface of the object
(Fig 3).These targets are specially made such that their optical property does not deteriorate
during critical environments inside TVAC. Retro reflective targets are used to overexpose the
region of interest and remaining area of the reflector surface is underexposed to enhance the
quality of photography. This further improves the quality of image analysis and thereby the
overall measurement process. For a typical size of 2m reflector surface more than 2000
numbers of targets are used. These targets are stuck manually on the surface. Sticking these
targets is a tedious task and removal of these targets leaves adhesive residue on the surface of
the reflector.
Photography of these targets is carried out using a specially designed metric camera GSI.The
camera is placed outside the window to protect it from harsh environment of TVAC chamber
(Fig 4). A specially designed optical glass window made of quartz material is used which
helps in retaining optical quality of images captured by the camera.
It is desirable to use non-contact type of target projection system for such applications.
However, placement of such a projector inside the TVAC is restricted due to the environment.
This study is carried out to ascertain the feasibility of carrying out surface measurement using
2
projected targets. The study establishes a process measurement by comparing the quality of
measured point cloud on the reflector surface projected through optical window with that of
the targets projected without glass window.
2a. Measurement Accuracy
The camera accuracy has a fixed component of 5 microns plus 5 microns for every meter size
of the test article. The measurement uncertainty of Photogrammetry is known to vary linearly
as the size of the test object. Other factors affecting measurement uncertainty such as the
number of images to be used, camera resolution, rotation speed of the test article, orientation
of the test article w.r. t camera, micro vibrations in the test fixture, deformation of MGSE
fixture itself, Iso static supports, were addressed at the system configuration stage.
2b. Image processing
The Centroid ( 2D pixel coordinates) of each of many circular targets is extracted by
software. With the help of coded targets auto recognition of targets and identifying
corresponding targets across all images is achieved. The 2D pixel coordinates are converted to
3D spatial position by the Photogrammetry software (using Bundle adjustment.) Essentially
each 2D pixel coordinate defines a ray issuing from the CCD image plane and passing through
the camera optic center. An optimization algorithm using a least squares approach achieves
intersection of these rays in 3D space. Due to various noise, the rays in general do not
intersect. The algebraic distance between nearly converging rays is minimized. Once an
optimal solution is on hand, the rays are back projected starting from 3D position, passing
through the camera optic center on to the image plane. The deviation between the back
projected ray pixels and the initial image measurements gives the image residuals. Typically
0.18 micron is possible to achieve with this test set up including camera, optical window and
associated test set up.
Fig 3: Reflector with physical targets
Fig 4: Reflector inside chamber on MGSE
3. Shape Measurement Using Contactless Target Projection System:
The principle of the projection system is straightforward. It works much like an
ordinary slide projector. A light source illuminates a target slide. This illuminated pattern
passes through a series of lenses that magnify the slide and project it onto the object. It is
necessary to focus the lens so that the target slide is in focus on the object surface.
3
Figure 5: Projected targets using Prospot
Fig 5 shows the dots from the projector on a surface [1]. They are of high- contrast
and quality. They mimic conventional retro-reflective targets, but have no inherent target
thickness. The size of the projected dots grows as the projector gets further from the object.
Like conventional targets, the projected dots [4] must be bright and of high contrast for a good
measurement. Generally, the projector is set far enough back to cover the object with the dot
pattern. The strobe intensity is then adjusted so the dots are measurable. If the entire object
cannot be measured in a single set up, then multiple setups that collectively cover the entire
object can be used. The area that can be measured in a single setup of the projector depends
on several factors such as the color of the surface, its finish and its curvature.
The major difference between the retro reflective targets and dot projection system [2] that
the retrorefletive targets such that relative distortion of the point can be tracked w.r.t to
hardware where as projected targets can be used to generate / comparison of shape or surface
r.m.s, with respect to any two measurements.
3a. Case study to validate the new process of shape measurement
In the work presented here instead of mounting physical targets on the reflector a dot
projection [3, 4] system is used for measurements. Physical targets need to be mounted on a
piece of prefixed kapton sheet on Reflector which requires lot of physical effort for mounting
(2500 targets and 2500 kapton pieces). The main drawback of these physical targets is that
they can lose their adhesion and shape during thermal vacuum cycling .Another problem with
these targets can lead to undesirable contamination and damage to the Test article, which can
lead to quality assurance issues. Also these physical targets have shelf life. Another
disadvantage with physical targets is it cannot be used for higher temperature applications like
thermal cycling of +150 Deg to -150 deg.When repetitive cycles to be done on the hardware
the target removal will be difficult with adhesive residuals left out. Also after fabrication of
front shell and rear shell in components like DGR the shape measurement will be in a single
shot compared to CMM, Laser tracker and Physical targets a measurement which requires lot
of time. By employing projected dot technique [2] at each stage the shape can be captured and
compared with the nominal surface.
4
Port for
Dot projection
Inca 3a
Camera
Quartz glass
window
Dot projector
Figure 6: Set up for target Projection with
glass window outside TVAC
Figure 7: proposed schematic set up for
future target projection inside TVAC
First measurements were carried out using Quartz glass window in front of the dot projector
as shown in Fig 6. By using Pro spot the targets were projected on to the Reflector. Here the
camera and the controller for dot projector were synchronized such that simultaneous dot
projection and camera imaging will be done. The dot projection has an advantage that
depending on the size of the object and density of targets can be varied depending on user
requirement.
Secondly the measurements were carried out without using the glass window by projecting the
dots on to the reflector. In both the cases the measurements were good with minor variation in
plan quality factor and overall project rms in the order of 10 microns shown in Table 1
The point cloud data which was measured through dot projector was analyzed and compared
with the physical targets point cloud data. The best fit rms data of the surface is closely
matching with glass window and without glass window using dot projector which was shown
in Table 3.The point cloud data and camera network geometry is shown in Fig 8 and Fig 9.
Table 1 Output of Bundle adjustment parameters
Rejection limit
Residual rms
Rms X
Rms Y
Rms Z
Without glass window
With glass window
1.37 µm
1.49 µm
0.39 µm
0.43 µm
Accuracy in µm
12µm
17 µm
20 µm
16 µm
11 µm
11 µm
5
Figure 8: Shape Measurement using
Target Projection system with window
Figure 9: Network of camera geometry
4. Results & Discussion
The Bundle Adjustment software accounts for the presence of optical window by calibration
parameters. The effect of camera calibration parameters projected dots through the window is
brought out in Table.1&2. We see small perturbations in the parameters. The shape
parameters rms, peak and valley etc, indicate the quality of measurement. The best fit rms
using dot projection with window or without window is almost same. But the best fit rms
using physical targets is slight variation with respect to projected targets (Table 3) because of
test set up limitation and number of targets. Also the measurement time required for dot
projection system is much lesser than the physical targets. So in future it is proposed that in
side TVAC this dot projection system can be implemented (Fig 7) by augmenting the facility
for another quartz glass window and prospot dot projector.
Table 2. Camera calibration internal parameters
Camera parameters
C-camera focal length
XP principal point offset
YP principal point offset
AP1
AP2
K Total Radial Distortion
P1(Decentering lense distortion)
P2 (Decentering lense distortion)
Without Glass window
21.7550
0.0142
0.0894
0.0038
0.0017
0.5926
0.0024
-0.0109
With Glass window
21.7584
0.0156
0.0907
0.0039
0.0011
0.5952
0.0011
-0.0120
6
Table 3. Comparison of data using physical targets and projected targets
Surface best fit rms
Physical
targets
without glass window
Projected targets
without glass window
Projected Targets with
glass window
0.45 mm
0.39 mm
0.37 mm
Acknowledgement
Authors wish to thank Dr.Anand Kumar Sharma, Deputy Director of Mechanical systems
area, ISRO Satellite center, and Dr.K Renji, Group Director, STG, for kind permission to
present this work. We thank Mr. S Shankar Narayan, Head of Experimental Structures
division, for technical guidance and suggestions.
We also would like to thank, Ravi T, Rudra Gowda, Vadivel M, CheluvaRaju.C.M, R.Rajesh,
Yugendar, Lokesh A.H, Naveen kumar, Mubbashir Mohammed in carrying out the
experimental work.
References:
1.
Shape measurement using Close Range Photogrammetry (Target less)-IITP
Project.Raghunatha Behara (ISRO-ISAC-internal report), 2008.
2. “Close Range Photogrammetry: A tool for Shape Measurement & Reflector alignment” by
C.Koteshwar Rao in CMSC-2010.
3. R.S. Pappa et al, Dot projection photogrammetry and videogrammetry of Gossamer Space
structures Journal of Spacecraft and Rockets 40, 858 (2003).
4. Black .J.T, Pappa. R. S”Videogrammetry using projected circular targets proof-of-concept
Test”, NASA/TM-2003-212148.
7
8th International Symposium on NDT in Aerospace, November 3-5, 2016
Advances in Phased Array Ultrasonic Testing Analysis software
Bharath Kodumuru1 and Jay Amos2
1
Textron Aviation, Bangalore, Karnataka 560059, India;
Phone: +91 80 7140 9671; email- bkodumuru@textron.com
2
Textron Aviation, Wichita, KS 67215, USA; email- JMAmos@txtav.com
Abstract
With the recent advances in technology phased array ultrasonic testing (PAUT) has been widely used to
overcome the challenges in inspecting complex composite parts. Though many PAUT equipment manufacturers
(OEMs) integrate image display and basic processing they are not fully sufficient for faster, reliable and
consistent decisions. This paper describes various analysis tools to minimize operator dependency in
quantification of discontinuities that were detected and also introduces a tool for merging individual inspection
files to enable global view of the component being inspected that greatly aids engineering community
responsible for disposition & local repairs of the component without the understanding of PAUT inspections. It
also introduces a common software platform to analyze files that are saved by different OEMs.
Keywords- composites, phased array ultrasonic testing, image analysis
1. Introduction
Aircraft have many structural components that require testing without damaging the part. One
of the advanced methods of nondestructive testing is PAUT. This method inspects a bonded
or composite part for discontinuities such as voids, porosity, delaminations, disbonds, resin
starved/rich areas and cracks that over time can cause a loss of strength [1]. Usually OEMs
provide integrated software with basic analysis tools. Few also provide exclusive analysis
versions of the software offline (remotely at analysis workstation) to document the anomalies
detected within the part to determine its integrity with extra cost. Almost all OEMs have their
proprietary file formats in saving the inspection data which cannot be read by analysis
softwares of different OEMs. OEMs are slow to make the transition to saving data in the
approved DICONDE format (ASTM E2339 and E2663). Thus, it is very important to have a
common software platform that could import different file types (saved by OEMs) without
compromising the state-of-the-art imaging techniques.
To inspect a large aircraft structure such as an entire wing it is usually scanned by marking
grid lines to guide coverage of the whole structure. It is important to display all the C-scan
images (of individual files obtained from above) together in a merged image that greatly aids
the engineering community responsible for disposition & local repairs of the component
without the understanding of UT nondestructive inspections.
This paper introduces our in-house developed common software that can read files from
different OEMs and describes the concept of dynamic gating and multi-modal views which
minimizes operator dependency in decision making quickly and consistently, and merges
individual files without memory challenges that are being faced by leading OEMs.
2. Analysis Software
There are several complex composite structures and geometries in an aircraft with variations
in thicknesses (in same part), highly curved parts and stiffened panels etc. Analyzing PAUT
data for discontinuities such as porosity, voids, delaminations, disbonds, and resin
rich/starved areas in complex structures is highly operator dependent. Thus, we have
developed PAUT Analysis software; a custom-coded, flexible but powerful interface
language called InspectionWare, supported by UTEX Scientific Instruments Inc.
We use 3 conventional gates to monitor interface signal (I), discontinuities near front wall
(A), back wall gate (BW) and an advanced elastic gate to monitor discontinuity signals that
occur between front wall and back wall signals (Volume V). The Volume gate is an interface
following gate and its length is adjusted with reference to the back wall signal amplitude
automatically. Hence detection of discontinuities even in complex structures is fully
automatic minimizing operator dependency. Figure 1 displays A-scan with title showing the
type of time of flight selected for each gate respectively.
Figure 1. A-scan Display /Threshold Crossing; ^Peak Amplitude
2.1
Test Specimen
The usefulness of volumetric gate can be demonstrated by scanning a test specimen with a
composite fiber reinforced plastic wedge standard with 8 different thickness variations and
void inserts at a depth of 0.060 inch from the surface (refer Figure 2) being inspected at each
step using a near wall 5 MHz linear array ultrasonic transducer. After initial gate settings for
offset distances of the Volume (V) gate with reference to interface and back wall signal
amplitudes from I and BW, the amplitude and time of flight C-scan images of Gates BW & V
can be displayed throughout the part along with the A-scan, D-scan and B-scan (refer Figure
3).
Figure 2. CFRP Wedge with Void Inserts
2.2
Data Representation
Disposition of indications in complex composite material inspections require simultaneous
analysis of multiple representations of the UT data, for example; 1) B & D-scan view to
determine depth distribution of the indications (from front and end views), 2) C-scan
Amplitude to evaluate the anomaly attenuation, and 3) C-scan Time-of-Flight (TOF) to assess
the depth of anomalies 4) Views of multiple C-scan gates to determine response of
interlaminar anomalies relative to thru-thickness (back wall) response. Though several
OEMs’ analysis softwares display this type of multiple representations of data, few of them
do not have their cursor position synced with each display which is very important for A, B &
D-scans which correspond to the cursor position on C-scan. Others do not have their zoom
synchronized with each display, which is very useful for evaluation and documentation.
Thus a dynamic volume gate could detect signals that are very close to front wall and back
wall signals in addition to void, porosity or interlaminar anomaly signals in between front and
back wall signals without which, its prone to error or inconsistency between inspectors. It has
reduced the dependence on operator skills for detection of discontinuities. It can also improve
the probability of detection of near-surface discontinuities which may have gone undetected
without its availability and minimize false calls due indications from rough surfaces.
Figure 3. Multiple Representations Displayed in One Screen with Cursor & Zoom
Synchronized
2.3
Merge Utility
Depending on the size of the part to be inspected, the inspector makes a scan plan (raster
scan) & generates multiple images of the sections of the part. A merge utility has been
developed which can rotate and place these individual images onto a canvas at specific
locations to ensure 100% inspection scan coverage & evaluation. Figure 4 represents merged
image of aircraft shear panel structure with bonded stiffeners. This merged image file
contains the raw data as well i.e., a region of interest can be drawn to populate its statistics.
Figure 4. Merged Image of aircraft shear panels with bonded stiffeners
Merging individual files to visualize the large part under inspection aids the engineering
community responsible for disposition & local repairs of the component without the
understanding of PAUT inspections. Not to mention, that the merged file needs to have raw
data to analyze data at a particular region of interest (ROI). Several leading OEM softwares
have this utility and have been facing challenges with merging large data files described in
Table 1[2].
Table1. Challenges with Leading OEM Software
Challenge
Expected size of a merged file
Virtual memory size of the workstation
Draw Region of Interest (ROI) after
merging
Palette adjustments after merging
Merging both full waveform (3D) as well
as C-scan images (2D)
Performance
could not merge if > 2GB
could not merge if < 800 MB
not possible
not possible
not possible
Additional advanced capabilities of our software are listed in Table 2.
Table 2
Textron Aviation PAUT Analysis Software Capabilities
Merge Files

ROI & SNR tables

B-scan overlay

Both Horizontal and Vertical B-scans

can be displayed simultaneously
C-scan Display (Fit-to-screen or 1:1)

Automatic Gate Length adjustment to
display the interlaminar porosity/voids

even in the presence of material
thickness variations
Navigate through Cursor

Save/Import Gate Settings

Display Amplitude & ToF C-scans

simultaneously
Crop A/B-scan on time

Chose Filters on C-scans

Lock/Password protection

Display Amplitude + ToF C-scans

simultaneously w/A ,B & D-scans
Comment to each ROI

Cropping Scan Axis from Start and End

of a C-scan
Cropping Time Axis on B/D-scan

Display C/A-scans as Cartographs

(resizable displays)
Zoom Synchronize on all displays

2.4 Interoperability
Interoperability of NDI records is very important because aircraft manufacturers fabricate
parts either in-house or procure from a qualified supplier who often inspect with different
equipment. Often manufacturers of critical hardware need to review scan data of parts that
are being procured from multiple suppliers, and are unable to do so because files are stored in
proprietary formats. End users have many suppliers and it is not cost-effective to buy analysis
software from all of them. Thus common software that can analyze files from those OEMs
without compromising analysis capabilities is required. The software described has excellent
interoperability that can import and analyze files saved by several leading OEM softwares.
3. Recommendations
Due to mutliple suppliers and equipment, it is important for end users to have a common
software to analyze files saved by different equipment without retraining inspectors on
different software applications. Better still, it is vital that OEM’s work towards implementing
a common storage format such as DICONDE to eliminate these data conversion &
obsolesence issues. The flexibility of multiple data representations and merging big datasets
for large parts are valuable tools to make accept/reject decisions of parts being inspected.
References
1. J H Heida and D J Platenkamp, ‘Evaluation of Non-Destructive Inspection Methods for
Composite Aerospace Structures’, International Workshop of NDT Experts, Prague, 10-12
Oct 2011.
2. Olympus NDT white papers, manuals and tutorials.