Structural Health Monitoring for Aerospace Vehicles Structural Health Monitoring for Aerospace Vehicles Mark Hedley CSIRO ICT Centre Australia Research Partners: CSIRO Industrial Physics CSIRO Manufacturing & Infrastructure Technology NASA Langley Research Center Boeing Phantom Works Defence Science & Technology Organisation (DSTO) 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Commonwealth Scientific and Industrial Research Organisation (CSIRO) 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Where have we come from in the last 100 years? Where are we now? Where will we be in 50-100 years? Boeing 747 1970s Wright Bros., Kitty Hawk, 17 December 1903 29 October 2004 CENS Seminar Boeing 7E7 2008 Structural Health Monitoring for Aerospace Vehicles Future aerospace vehicles will Re-configurable (morphing) Structural Self Assessment Self Repair Intelligent adaptive response Bio/Nano/Thinking/Sensing Vehicle This requires multi-functional material and structures Active/sensory/structural Embedded intelligence Biomimetic functionality is being explored for ways to achieve this 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles What is structural health monitoring? The ability to monitor damage, assess structural health and diagnose damage conditions What is structural health management? Taking action is response to damage, form a prognosis, make a decision and take remedial action How do they differ from current practice? Currently based on periodic inspection First Step - Conditioned Based Maintenance 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Outline of talk: 1. Motivation 2. General requirements and principles of future SHM systems • Requirements for a SHM system • Proposed Architectures • Agents and Sensing 3. Example: the CSIRO multi-agent test-bed • • • • Objectives and simple damage scenario Architecture and hardware Multi-agent algorithms Current system status 4. Summary and conclusions 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Acknowledgements: A team of ~ 20 people at CSIRO (CIP and the ICT Centre) has contributed the ideas and done all the work I will describe here. The major contributors have been: Tony Farmer Andy Scott Graeme Edwards Mark Hedley Mark Johnson Chris Lewis Phil Valencia Nigel Hoschke Mikhail Prokopenko Peter Wang Vadim Gerasimov Geoff Poulton Geoff James 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Technology Advancement General requirements and principles of future SHM systems. Where do we want to go with SHM? A few bonded sensors connected to a processor or data logger 29 October 2004 • Integrated / embedded sensors • Mobile sensors • Large numbers of sensors (10ⁿ) • Autonomous diagnostics & prognostics • Intelligent decision-making • Remediation strategies / self-repair • Sensory / active materials • Robust, adaptive, reconfigurable. Time CENS Seminar Structural Health Monitoring for Aerospace Vehicles Functional Requirements for a SHM System: Identification of threats What is the operational nature and environment of the structure? What are the potential sources of damage? Where might they occur? Can the threat be detected and avoided before damage occurs? Monitoring Detection (or deduction) of damage Sensing, interpretation, learning Evaluation of damage Diagnosis of damage Decision-making Prognosis for structure Remediation decision Damage models, prior knowledge, learning Options for action: repair or remediation. Actions What responses are appropriate, achievable? Nothing, report only, modify operational conditions, repair, abandon ship!! 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Application Requirements for a SHM System: Robustness: Must be able to operate effectively in the presence of damage Scalability: System may contain a very large number of sensors, processors, structural elements, etc. Reliability: Must be more reliable than the vehicle structure Validation/certification is a major issue as systems become more complex This is a problem for all types of control and safety systems 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Specific Characteristics Needed for a SHM System: 1. Wide range of reaction/response times: Milliseconds – impact damage, pressure leaks, … Seconds – cracks, disbonds, composite degradation, … Hours (or longer) – fatigue, corrosion, creep, wear, … In many cases, the required reaction time depends on how early the damage (or threat) was detected. 2. Range of decision-making processes and response types: “Panic” or reflex response (no reasoning) Considered, reasoned (intelligent) response 3. Broad spectrum of environments, risks 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles The ideal network architecture is a Sensor Network: 29 October 2004 Robust network topology Scalable Distributed Processing Wired or Wireless Limited Network Power CENS Seminar Structural Health Monitoring for Aerospace Vehicles The ideal distributed processing architecture is a Complex Multi-Agent System: 29 October 2004 autonomous agents that interact with each other and with their environment each agent is incapable of solving a problem alone can display self-organization, or complex emergent behaviour inherent redundancy, no single point of failure well suited for handling evolving, dynamic problems CENS Seminar Structural Health Monitoring for Aerospace Vehicles Complex Systems Features Complex vs Complicated Characterised by Emergent Behaviour No processor has global view Issue Design agent properties to obtain desired emergent behaviour using only local knowledge We design complexity out of engineering structures because we don’t yet know how to control it! 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Agents in a Multi-agent-based SHM system: Agents are autonomous, but are only aware of their local environment – an agent can’t see the “big picture”, can’t solve the “big problem”. Agent functions: Controls a suite of sensors and/or active elements Processes sensor data to infer local damage information Communicates with other agents (neighbours only) Contributes to emergent response Agents may be static (embedded) or mobile (robotic) 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Sensing in a Multi-agent-based SHM system: Sensor characteristics: Direct vs. indirect (damage inference/likelihood) Passive vs. active Local vs. remote Embedded vs. mobile Sensing strategies in a SHM system may be very different from “traditional” periodic inspection-based NDE Continuous monitoring vs. periodic inspection – Aim to detect damage at early stage Primary vs. secondary sensing Dense vs. representative sensing Use of network as a sensor The only certainty is that a multi-sensor strategy will be needed!! 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Outline of talk: 1. Motivation 2. General requirements and principles of future SHM systems • Requirements for a SHM system • Proposed Architecture • Agents and Sensing 3. Example: the CSIRO multi-agent test-bed • • • • Objectives and simple damage scenario Architecture and hardware Multi-agent algorithms Current system status 4. Summary and Conclusions 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Example: the CSIRO/NASA Multi-Agent Test-bed Objectives: • Purpose: to experiment with and demonstrate concepts for an intelligent vehicle health monitoring system within a relatively simple environment. The prime focus, initially, is on systems issues (conversion of data to information, diagnosis, prognosis, intelligent decision-making, … ), rather than on particular sensing issues. • Intended as a research tool and demonstrator for concepts and techniques – to explore the possibilities. • NOT intended to be a prototype of a practical system. Constraints of weight, cost, power consumption, EMI, … not considered. • High level of processing power for maximum flexibility. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Damage scenario: • Simple damage environment: impacts from fast particles, such as micrometeoroids or space debris. • 1st stage of development: aim to detect, locate and evaluate the effects of particle impacts anywhere within the aluminium skin of a “vehicle”. • Later stages will develop diagnosis, prognosis and remediation decision-making, incorporate other sensors & strategies, damage scenarios, etc. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Damage Simulation: • As the hardware will not be flown in space, other sources of impacts are required: • A light-gas gun was used to fire 1 mm diameter stainless steel ball bearings to speeds up to 1.5 km/s • The focused pulse from a Nd:YAG laser (wavelength 1.06 nm) with duration 8 ns and energy up to 0.5 J provides a good simulation of a normally incident non-penetrating impact 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Architecture and Hardware: Serial Communication Links 800 mm 400 mm UDP Communication Link Concept Demonstrator Hardware Containing Sensors and Physical Cells 29 October 2004 PC Cluster for Simulating Cells Workstation for Initialising and Monitoring Test-bed Vital statistics: • Hexagonal prism • 192 agents (cells) • 1 mm Al sheet skin • Cells ~100 x100 mm • Regular mesh • 4 x 2.5 mm dia. PVDF sensors/cell • Sensors on ~60 mm square • Passive sensing only • Laser pulse “impacts” CENS Seminar Structural Health Monitoring for Aerospace Vehicles Sensor Signals: 200 Laser Pulses Sensor signals (mV) Particles V ~ 200 m/s 100 25 0 0 -100 -25 -200 -20 -50 -30 -10 0 10 20 20 10 10 5 0 0 -10 -5 -20 -30 -20 -10 Time (µs) 29 October 2004 V ~ 1 km/s 50 0 10 -20 -20 -10 0 10 -10 0 10 20 Time (µs) CENS Seminar Structural Health Monitoring for Aerospace Vehicles Cell Architecture: Skin Sensors 29 October 2004 Electronics Electronics Preprocessing Preprocessing Analysis Analysis Communications Communications Data Acquisition Sub-module Network Application Sub-Module CENS Seminar Structural Health Monitoring for Aerospace Vehicles Data Acquisition Sub-module: • 5 Sensor Channels (bidirectional) with analog filtering and amplification of input • 14-bit ADC up to 16 MSPS • 150 MIPS DSP • 256 kbyte FLASH 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Network Application Sub-module: • 400 MIPS DSP • 400k gate FPGA • 2 Mbytes FLASH • 8 Mbytes SDRAM • 5 communication ports • 1.2 W typical power consumption • 64-bit unique identifier 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Cell Hardware: + 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles System Hardware: 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Impact Localisation: • Lowest order extensional plate wave detected (about 5.3 km/s) • Localisation within cell based on triangulation using time of arrival difference for the four sensors • Table lookup for fast calculation • Error few mm within rectangle, grows quickly outside 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Communication Protocols: • Agent to Agent (only to neighbours) • Between agent and workstation (asymmetric) – not required for operational network (network query) • Agent to local neighbourhood • Between static and mobile agents • Don’t require communication between arbitrary agents! 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Communication Protocols: Reprogram Nodes Node and Node Network Control Status Data Transfer Monitor Link Control CD Flood Protocol Protocol Master Route Protocol & Master Flood Protocol Application Layer Agent Software Transport Layer Simple Agent Protocol Data Link Layer Data Link Layer UDP (Simulated Nodes) 29 October 2004 Network Layer Serial Port Device Driver CENS Seminar Physical Layer Structural Health Monitoring for Aerospace Vehicles Visualisation – Building the Network Network discovery by the visualizer of the 188 physical cells. There are four cells missing as a small window was left in the demonstrator to allow observation of the inside. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Visualisation - Current Flow Current flow in the demonstrator, showing flow into and out of cells, around absent cells, and the net current across cell edges 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Visualisation – 3D and Boundaries 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Multi-Agent Algorithms: Aim to develop local agent algorithms that will result in desirable emergent behaviour of the collective system of agents. Simple algorithms developed so far: – Impact boundaries – Damage networks – Clusters based on damage severity – Self-replication of damage region Entropy-based metrics developed to measure stability of emergent behaviour. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Impact Boundaries: simple, involves single cells only. White cells: direct impact damage Red cells: indirect damage Blue cells: form boundary between damaged and undamaged regions. White lines: form continuous, connected damage boundary. May be used to route messages around a damaged region. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Impact Networks: involves groups of cells. White cells: direct impact damage. Algorithm simulates ants foraging for food. Forms shortest path between “food sources” (and avoiding obstacles). Green cells: pheromone levels. May be useful for damage evaluation, e.g. guidance of a mobile sensor or repair agent. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Current System Status: • We have built a platform for undertaking research in SHM • Dense wired sensor network for impact detection • 192 Nodes and 4 Sensors per Node • We have undertaken initial research in the use of multi-agent systems for distributed processing for structural health diagnostics Underlying research issue is to develop techniques for multiagent-based knowledge management, learning and emergent decision-making. 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Current Developments: • Mobile robot for secondary inspection and repair • Wireless communication using acoustics through skin and RF (802.15.4) • Damage evaluation using active acoustic sensors 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Future Developments: • Secondary Inspection • Actions • Reflex response • Intelligent Response • Detection of other damage modes • Other materials (e.g. composites, metal foams) 29 October 2004 CENS Seminar Structural Health Monitoring for Aerospace Vehicles Conclusion: We can expect to see large qualitative as well as quantitative changes in SHM in coming years The extent and directions will depend on advances in: ― Multi-functional materials ― Embedded processing ― Intelligent systems This will be an enabling technology not just for more efficient maintenance, but for radically different aerospace vehicles 29 October 2004 CENS Seminar