Utah State University DigitalCommons@USU All Graduate Theses and Dissertations Graduate Studies 2014 Uncertainty Analysis of Mechanical Properties from Miniature Tensile Testing of High Strength Steels Deepthi Rao Malpally Utah State University Follow this and additional works at: http://digitalcommons.usu.edu/etd Part of the Mechanical Engineering Commons Recommended Citation Malpally, Deepthi Rao, "Uncertainty Analysis of Mechanical Properties from Miniature Tensile Testing of High Strength Steels" (2014). All Graduate Theses and Dissertations. Paper 4029. This Thesis is brought to you for free and open access by the Graduate Studies at DigitalCommons@USU. It has been accepted for inclusion in All Graduate Theses and Dissertations by an authorized administrator of DigitalCommons@USU. For more information, please contact dylan.burns@usu.edu. UNCERTAINTY ANALYSIS OF MECHANICAL PROPERTIES FROM MINIATURE TENSILE TESTING OF HIGH STRENGTH STEELS by Deepthi Rao Malpally A thesis submitted in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE in Mechanical Engineering Approved: Dr. Leijun Li Major Professor Dr. Thomas H. Fronk Committee Member Dr. Steven L. Folkman Committee Member Dr. Nicholas A. Roberts Committee Member Dr. Mark R. McLellan Vice President for Research and Dean of the School of Graduate Studies UTAH STATE UNIVERSITY Logan, Utah 2014 ii Copyright c Deepthi Rao Malpally 2014 All Rights Reserved iii Abstract Uncertainty Analysis of Mechanical Properties from Miniature Tensile Testing of High Strength Steels by Deepthi Rao Malpally, Master of Science Utah State University, 2014 Major Professor: Dr. Leijun Li Department: Mechanical and Aerospace Engineering Boat samples extracted from scheduled maintenance shutdowns of piping and pressure vessels provide opportunities for testing for mechanical properties of the service exposed components. However, it is not clear whether testing of miniature specimens machined from boat samples which are about 2 in. long can be a viable replacement for the standard-sized mechanical testing. Three steels, stainless steel Type 304, sensitized Type 304, and SA516 Grade 70 carbon steel, are tested by standard-sized specimen and miniature specimen tensile tests. Mechanical properties as affected by the specimen geometry and tensile testing procedure for miniature specimen testing are compared to that of conventional specimens tested according to ASTM A370-10. The miniature tensile testing results are analyzed by using Monte Carlo Method (MCM) for uncertainty estimation in order to quantify the probability distribution of mechanical properties. For the steels under study, miniature specimens with a cross-sectional area of 3 mm2 and 12 mm gauge length are found to produce equivalent mechanical properties as tested from standard-sized specimens. (59 pages) iv Public Abstract Uncertainty Analysis of Mechanical Properties from Miniature Tensile Testing of High Strength Steels by Deepthi Rao Malpally, Master of Science Utah State University, 2014 Major Professor: Dr. Leijun Li Department: Mechanical and Aerospace Engineering This Miniature mechanical testing study is concerned with the use of miniature specimens to identify the mechanical properties of stainless steel Type 304, sensitized Type 304 and SA516 Grade 70 carbon steel as a viable replacement for the standard sized mechanical testing. The study aims at obtaining suitable specimen geometry and tensile testing procedure for miniature mechanical testing whose mechanical properties are comparable to that of conventional specimens of ASTM A370-10 of the same steel. All specimens are flat and the gauge length cross section will be varied to obtain suitable geometry. The miniature tensile testing results are further validated by using Monte Carlo Method (MCM) for uncertainty estimation in order to know the probability distribution of mechanical properties. Miniature specimens with a cross section of 3 mm2 and 12 mm gauge length are found to produce equivalent mechanical properties as tested from standard-sized specimens. If a reasonable agreement is received, it will provide us with a very useful tool to evaluate mechanical properties of degraded materials, which cannot be removed from service for standard testing, for repair and service life evaluation. v Dedicated to my dearest parents and brother... vi Acknowledgments I would like to express my highest regards and gratitude to my major professor, Dr. Leijun Li, for his continued guidance and advice despite several constraints. I will be forever indebted to him for believing in a novice like me right from the beginning of my career as a research assistant at USU. I would like to thank my committee members, Dr. Thomas H. Fronk, Dr. Steven L. Folkman and Dr. Nick Roberts, for their invaluable suggestions and also Dr. Barton Smith for his advice in uncertainty analysis. I would like to express my sincere gratitude to the graduate advisor, Christine Spall, for her constant encouragement and for helping in fulfilling all the requirements to accomplish graduate studies. I would like to thank all past members of the Materials Processing & Testing Laboratory, Andrew for helping me prepare the specimens, Jacob, Zhifen, Yin and Bishal for their friendship. Special thanks to Dayakar Naik for being my mentor academically and emotionally. Motivation and belief in oneself is of umpteen importance when one is far away from home. I would like to thank my dear parents, Ravi and Anjana, for believing in me and boosting my confidence. I hope to make you proud everyday. Thank you my late grandmother, Usha ajji, for expressing immense happiness in all my endeavours. Whatever little accomplishments I have today is all because of your blessings and encouragement. Thank you my little brother, Sanjeev, for filling my absence at home and all my well wishers. My experience as a graduate student at USU has been wonderful and fun filled only with the presence of many fellow graduate students who eventually became friends for life. Thank you Ravi, for helping me beilieve in myself, Neeraj, Bidisha, Saptarishi, Rajee, Ashish, Swati, Ruchir, Manju, Joe Shope, Kurt, Scott and many others. I would like to thank my friends back home, Prerita, Rahul and Danny, for helping me in many ways than I could describe. I am thankful to God for showing me the right direction in life. Deepthi Rao Malpally vii Contents Page Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Public Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Overview of the field of inquiry . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Parameters effecting the mechanical properties of miniature specimens . . 2.3 Advances to miniature testing techniques . . . . . . . . . . . . . . . . . . 3 3 6 7 3 Experimental Procedure . . . . . . . . . 3.1 Component Element Properties . . 3.2 Sensitization . . . . . . . . . . . . 3.3 Tensile Specimen Preparation . . . . . . . .... . . . . . . . . . .... . . . . . . . . . .... . . . . . . . . . . . . . 4 Experimental Results and Discussion . . . . . . . . . . . . . 4.1 Conventional (Macro-sized specimen) Tensile Testing . 4.2 Miniature Tensile Testing . . . . . . . . . . . . . . . . 4.3 Optimum Specimen Size . . . . . . . . . . . . . . . . . .... . . . . . . . . . . . . . .... . . . . . . . . . .... . . . . . . . . . ..... . . . . . . . . . . . . .... . . . . . . . . . .... . . . . . . . . . .... . . . . . . . . . . . . . 9 9 11 13 ..... . . . . . . . . . . . . . . . . 17 . . . 17 . . . 17 . . . 19 5 Uncertainty analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Tolerance intervals in Sample Populations . . . . . . . . . . . . . . . . . . 5.2 Propagation of Mechanical Properties by Monte Carlo Method (MCM) . . 5.2.1 General Approach for MCM . . . . . . . . . . . . . . . . . . . . . . 5.2.2 Propagation of Mechanical Properties . . . . . . . . . . . . . . . . 5.3 Convergence study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 . . . 29 . . . 32 . . . 32 . . . 34 . . . 42 6 Conclusion and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 viii List of Tables Table Page 3.1 Component element properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 4.1 Conventional (Macro-sized) results . . . . . . . . . . . . . . . . . . . . . . . . . 17 4.2 Results obtained from SEMTester 1000 EBSD . . . . . . . . . . . . . . . . . . . 18 4.3 Mechanical properties for Stainless Steel Type 304 miniature specimens . . . . 20 4.4 Mechanical properties for Sensitized Stainless Steel Type 304 miniature specimens 21 4.5 Mechanical properties for SA516 grade 70 carbon steel miniature specimens . . 22 5.1 Specimen area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 5.2 Load values for (YS) and (UTS) in specimens of gauge area (1x3) mm2 . . . . 36 5.3 Variables and their uncertainties - SS grade 304 . . . . . . . . . . . . . . . . . . 36 5.4 Final length of specimens of gauge cross section (1x3) after tensile testing - SS grade 304 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.5 Variables and their uncertainties - Sensitized SS grade 304 . . . . . . . . . . . . 39 5.6 Final length of specimens of gauge cross section (1x3) after tensile testing Sensitized SS grade 304 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 5.7 Variables and their uncertainties - carbon steel SA516 grade 70 . . . . . . . . . 41 5.8 Final length of specimens of gauge cross section (1x3) after tensile testing carbon steel SA516 grade 70 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 ix List of Figures Figure Page 2.1 Shear punch test fixture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 Indigenously developed table top ball-indentation test system . . . . . . . . . . 4 3.1 Microstructure of SA516 steel etched with 4% Nital . . . . . . . . . . . . . . . 10 3.2 Microstructure of 304/304L steel etched with Vilella’s reagent . . . . . . . . . . 10 3.3 Microstructure of sensitized 304/304L steel etched with Vilella’s reagent . . . . 11 3.4 Microstructure of 304/304L steel . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.5 Microstructure of sensitized 304/304L steel . . . . . . . . . . . . . . . . . . . . 12 3.6 SEMTester 1000 EBSD instrument for miniature testing . . . . . . . . . . . . . 14 3.7 Specimen cross section at gage length is (1x0.2) mm . . . . . . . . . . . . . . . 15 3.8 Variation in specimen gage thickness from 3 mm to 0.5 mm . . . . . . . . . . . 15 3.9 (a) Tinius Olsen H50KS (b) Specimen test in progress . . . . . . . . . . . . . . 16 4.1 Chauvenet’s criterion for rejecting a reading obtained from the book Experimentation, Validation and Uncertainty Analysis for Engineers by Coleman and Steele [1] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 4.2 Tensile strength vs specimen gauge cross section . . . . . . . . . . . . . . . . . 24 4.3 Yield strength vs specimen gauge cross section . . . . . . . . . . . . . . . . . . 25 4.4 Engineering stress vs strain curves for specimen gauge cross section (1x3) . . . 26 5.1 Factors for two-sided tolerance interval [1] . . . . . . . . . . . . . . . . . . . . . 30 5.2 Schematic for MCM for uncertainty propagation when random standard uncertainties for individual variables are used [1] . . . . . . . . . . . . . . . . . . . . 33 Distribution of MCM results for yield strength of carbon steel SA516 grade 70. Expanded uncertainties for each variable being calculated at 95% confidence . . 43 Convergence study for MCM value of UF for 95% combined uncertainty for each variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 5.3 5.4 x Acronyms ASTM american standards of testing and materials SEM scanning electron microscope EBSD electron backscatter diffraction MEMS microelectromechanical systems MMST modified miniature specimen test FE finite element EDM electric discharge machining FIB focused ion beam LVDT linear variable differential transformer MCM monte carlo method DRE data reduction equation SS stainless steel YS yield strength UTS ultimate tensile strength 1 Chapter 1 Introduction Steels and alloys experience material degradation when encountered with accidents or exposed to elevated services, especially after long-term exposure. Depending on the types of materials, the degradation could include embrittlement, creep, sensitization, phase precipitation, other phase transformation/formation, corrosion, oxidation, etcetra. When materials experience degradation, it is important to evaluate material properties for repair, extended use, or life prediction. For the material evaluation, it is often to prefer to obtain the actual material mechanical properties. However, for many important components, such as vessels in refinery industry, it is not realistic to obtain sufficient material to perform standard mechanical testing due to the relative large sample size requirement, which typically needs window cut from vessels. Window cut from service vessels will involve significant challenges of its replacement from cost and code requirements, and thus, it is typically not feasible to obtain materials for standard mechanical testing. Nevertheless, it is more feasible to obtain boat samples, such as about 2 in. (50mm) long, from serviced vessels, and it would be highly valuable for material evaluation if mechanical testing could be performed on boat samples, which requires miniature mechanical testing method. Miniature mechanical testing has become an important aspect of industrial and research labs, especially for analyzing the mechanical properties. The requirement of miniature testing depends on many factors such as availability. A number of tensile testers have been employed for miniature testing with samples of different sizes and geometries. For example, MTI SEMTester can test miniature specimens. Universal Tensile Testing machines can also be employed for miniature testing depending on the minimum cross head travel as miniature specimens have very short gauge length. However, there is no study to verify the miniature mechanical testing results against the standard mechanical testing results. 2 The current study will verfiy the miniature mechanical testing for stainless steel Type 304, sensitized Type 304 and SA516 Grade 70 carbon steel as a viable replacement for standard sized mechancial testing. Miniature tensile specimen designs are developed in a combination of varying gauge cross-sectional area, and then tested on two different tensile testers, Tinius Olsen H50KS and MTI SEMTester 1000 EBSD. It is shown in this study that miniature specimens of a particular gauge cross section exhibit similar mechanical properties to that of its equivalent conventional size. Since only a minimum number of specimens are tested, there is a possibility of variation from the expected conventional results from miniature specimens. This uncertainty in the results could be due to a number of factors, either machine oriented, which classify as systematic uncertainties, or procedure oriented, which classify as random uncertainties, in a broad sense. Also, the variation in miniature measurements reflects local variations in the measurements that are not measured with full sized specimens. This is due to lesser presence of voids and defects in miniature specimens when compared to standard full sized specimens. An estimation of uncertainty analysis will help generalize the mechanical properties of the materials being reviewed, and hence can be compared to conventional testing on a general scale which has not been covered by any study so far. The objectives for this study are as follows: • To identify an optimum procedure for extraction of miniature tensile specimens of high strength steels • To obtain a suitable specimen geomtery relevant to that obtained from a boat sample, and the miniature tensile tester utilized for the purpose • To analyze the data obtained with miniature specimens and compare the results with the tensile tests of equivalent standard specimen results • To propagate the mechanical properties obtained after tensile testing of miniature specimens by conducting a Monte Carlo uncertainty estimation 3 Chapter 2 Literature Review 2.1 Overview of the field of inquiry The origin of miniaturization of specimens is in the nuclear industry because of expensive and limited irradiation space, specimen sizes had to be miniaturized for experimental purposes of irradiation programs in reactors. However, in some cases there is a need to keep the test specimen similar in size to structural components [2]. The miniature specimen test techniques are broadly classified as follows [3]: • Tests that are based on miniaturization of conventional specimen sizes such as miniature tensile, fatigue, impact and fracture toughness tests • Tests based on novel techniques using disk sized specimens such as disk bend tests, shear punch/small punch tests as shown in Figure 2.1. The small specimen test techniques for disk bend tests and small punch tests do not have conventional counter parts. These techniques therefore need to be validated before being effectively used • Tests based on ball indentation techniques as shown in Figure 2.2 Fig. 2.1: Shear punch test fixture 4 Fig. 2.2: Indigenously developed table top ball-indentation test system A few methodologies that evolved in the past on miniature testing and its requirement in various fields are briefly described as follows: 1. Service life of a component is greatly dependent on the extent of monitoring the material degradation of the component when in service. In order to monitor the service life, nondestructive testing is required to evaluate material properties while keeping the component in service. For this purpose miniature mechanical test named small punch test has been used [4]. Sample punch test is claimed to be the most effective test as the specimens can be extracted from the highly stressed zones of the components and also, the samples can be extracted as many times as required allowing continuous monitoring of the component while in service. 2. Micro-scale testing has also been developed to characterize the performance and reliability of Micro-Electro-Mechanical Systems (MEMS). Various testing techniques have been developed to measure the mechanical response of small specimens. Different methodologies aiming at determining the mechanical properties of small scale samples have evolved in the recent past. A survey describes that a technique involving instrumented 5 indentation, most specifically, nano-indentation, coupled with careful loading and unloading response of the material make it possible to measure both the hardness and elastic modulus of material [5]. However, strength and strain hardening effects cannot be measured with indentation techniques. For manufacture of MEMS, a number of machining processes such as photolithography, vapor deposition and electro plating are available in exclusive foundries of MEMS. For machining miniature specimens from bulk specimens, traditional machining processes are applicable such as wire EDM, laser machining, chemical mechanical polishing, and FIB milling may be used [5]. 3. Recent development in nanomaterials and metallic glasses are facing challenges with preparation, handling and testing of small volumes of materials and also lack clear understanding of test results. Hence miniaturization has become a general trend [2]. In the past, Modified Miniature Specimen Test (MMST) was employed to determine yield strength, tensile strength and uniform elongation of unirradiated and irradiated reactor pressure vessel steels [6]. 4. Quasi static tensile tests were performed for 1.2mm gauge TRIP800 steel sheets with miniature tensile sample to determine the effects of sample geometry and loading rate on tensile ductility [7]. The samples with smaller gauge length will yield higher level of ultimate ductility since the post uniform elongation is achieved mainly in very narrow neck region. This is because in the miniature sample geometry, there is the localized nature of the neck during the post-uniform elongation just before fracture [7]. Similar increase in ductility was observed in miniature samples of AA5182 samples under quasi static loading in comparison with full size samples. Also, higher loading rate yielded higher upper yield strength compared to the results for TRIP steels reported by other authors. This is different from the reported strain rate sensitivity results for mild steel where a typical reduction of ductility is observed at high strain rates. However these discussions are valid for quasi static conditions of the sample being tested. 6 5. In one of the miniature tensile behavioral studies, the samples were made of ultrafinegrained Cu having the same width of 1mm and varying gauge length while keeping thickness constant and vice versa were studied. The results showed that thinner samples are susceptible to shear failure, resulting in smaller reduction in area [8]. However, the gauge length has no evident influence on the failure mode or area reduction. This article further summarizes that shorter and thicker specimens tend to be more ductile. The thickness effect is mainly seen during necking/fracture modes. The gauge length effect originates from strain. 6. In another study it is shown that tensile testing of very thin sheet metal, usually specimens lesser than 10µm (in this case Cu 99.9%), show trend in decreasing mechanical properties [9]. However, many authors have mentioned the need for efficient modelling of the constitutive behavior and the use of simulation and FE analysis [10]. 2.2 Parameters effecting the mechanical properties of miniature specimens 1. Miniaturization causes scaling effect, which leads to a different material behavior in the microscale compared to the macroscale. In general, this effect is pronounced largely and relates to specimen size and geometry and other factors such as microstructural constraints (grain size through specimen thickness, microstructural anisotropy, microstructural inhomogeneity, etc.), surface effect, residual stress [2]. 2. Non uniform stress distributions during testing provide an additional length parameter that affects the size-effect. The tensile specimen size-effect should be considered when comparing mechanical properties, such as tensile ductility measured on non-standardized dog-bone specimens [8]. To minimize any undue stress, the sample must be aligned with the center line of the two test machine grips [11]. Furthermore, with an ever increasing popularity of miniaturized tensile specimens in research of materials, there is a need for a standardized protocol to be adopted. 7 3. Specimen dimensions and strain measurement methods largely influence the tensile stress-strain curves. Miniature specimens are too small for use with conventional extensometers so that the strains are usually calculated from the crosshead displacements. Uniform elongation and post necking elongation increase with decreasing gauge length and increasing specimen thickness [12]. With a decrease in thickness, the gauge part is effectively transformed from a bulk to sheet geometry and the stress state within the gauge changes from a more or less biaxial condition to uniaxial stress state condition under tension, thereby resulting in a change from diffuse necking to localized necking. A reduction in gauge length prolongs the stress-strain curves to higher fracture strains or higher apparent ductility primarily by prolonging the uniform elongation. However, in [13], the yield stress is known to be thickness independent for thickness larger than the critical thickness, usually 6 to 10 times the average grain size for ferrous materials. 4. The variation in micro-sample measurements reflects local variations in the material that are not measured with full sized specimens. This is due to lesser presence of voids and defects in miniature samples when compared to the standard full size samples [14]. All the above listed effects are not limited to strength dependence on area, many times it relates to various microstructural constrains [2, 8–10, 12–27]. 2.3 Advances to miniature testing techniques For tensile specimens of most components in service, the dimensions are larger than the internal microstructural features, and thus the importance of external size-effects on mechanical properties can be a minor issue and hence can be neglected. However, with the advent of MEMS, and other miniature testing requirement module samples with microstructural dimensions, external size effects have become a more prominent feature to look at. Thus many small scale devices have been designed to get reliable results from these miniature samples [5]. Nevertheless, there is a need for a handbook for collecting accurate and reliable values of elastic, plastic, fracture, and fatigue properties of materials at different gage lengths. 8 Optical techniques have emerged replacing the grip system to hold the specimen in position as the grip displacement do not give correct strain measurements most of the time [5]. Results from these miniature tests give an insight into size effects such as elastic, plastic brittle and ductile behavior at miniature level. Also, the influence of voids and boundary cracks and surface roughness may also have an effect on the mechanical properties that are being tested for. With the need of requirement of understanding the mechanical behavior at small scale level, the trend towards miniature testing will increasingly continue. 9 Chapter 3 Experimental Procedure 3.1 Component Element Properties The as-received samples of stainless steel Type 304 and SA516 Grade 70 carbon steel are initially tested for actual composition. The component element properties are as shown in Table 3.1. One of the sample slabs of stainless steel Type 304 is furnace heat treated to promote sensitization. Table 3.1: Component element properties A: SA516 Steel, 1/2” Plate C Mn P S Si Cr Ni Mo Cu 0.14-0.20 0.70-1.00 0.040 max 0.050 max 0.17 0.98 0.015 0.007 0.18 0.11 0.11 0.02 0.26 B: 304/304L. Steel, 1/2” Plate 0.03 max 2.00 max 0.045 max 0.03 max 1.00 max 18.0-20.0 8.0-20.0 0.02 1.62 0.029 0.02 0.40 17.8 8.5 0.02 0.26 C: 304/304L. Steel, 1/2” Plate Sensitized Same as plate B but furnace heat treated (CRESS electric furnace, model C162012/SD) at 650◦ C(1200◦ F) for 7 days (168 hours) Microstructures of the as-received samples are captured as shown in Figure 3.1, 3.2 and 3.3. For the preparation of material sample, small portion (about an inch) of the samples are cut with a band saw. These sample pieces are placed in a mold on to which resin and hardener mixture is poured and is allowed to solidify. The casting on the sample piece aided in grinding and polishing for the preparation of flat polished sample. After etching on the flat polished samples, traces of its microstructure are revealed and the image is captured using an optical microscope. 10 Fig. 3.1: Microstructure of SA516 steel etched with 4% Nital Fig. 3.2: Microstructure of 304/304L steel etched with Vilella’s reagent 11 Fig. 3.3: Microstructure of sensitized 304/304L steel etched with Vilella’s reagent 3.2 Sensitization Sensitization occurs when stainless steel is exposed to very high temperatures (425◦ C to 815◦ C). Sensitization effect causes precipitation of chromium carbides at grain boundaries as chromium carbides are insoluble at high temperatures. For carbide to precipitate, it must get chromium from the surrounding material causing chromium depleted zone around the grain boundaries. This chromium depleted zone will be less corrosion resistant, specifically to intergranular corrosion. Sensitization is important at welded joints. This is because the welded zones experience very high temperatures causing sensitization [28]. The difference in the grain boundaries is evident in the microstructures. The Grain boundaries in the sensitized microstructure are darker due to the precipitation of carbides. The microstructures of stainless steel as-received vs sensitized stainless steel taken at 20x magnification is shown in Figure 3.4 & 3.5. 12 Fig. 3.4: Microstructure of 304/304L steel Fig. 3.5: Microstructure of sensitized 304/304L steel 13 3.3 Tensile Specimen Preparation The miniature specimens in this study are sheet specimens unlike in most previous research where only nano-scale or thin films have been studied. An important factor to be considered in specimen design for tensile tests is the aspect ratio. The ASTM standard specimen size has an aspect ratio (gauge length/diameter or in this case thickness) of 4:1. In order to get comparable results to that of the standard specimens, the miniature specimens were designed to have the same and multiples of the standard aspect ratio to study the effect of varied cross-sectional area. The gauge length is considered constant in all specimens to have uniform elongation. The first batch of specimens is tested on the SEMtester 1000 EBSD Figure 3.6 (designed to be fit inside a scanning electron microscope (SEM)). The minimum sample size (l × w) is (43 × 10) mm. Maximum displacement travel is 10mm with maximum load capacity of 1000lb (4500N). Suitable specimen design is created in Solid Edge software after careful examination of the grip area and the crosshead capacity of the miniature tensile tester. All specimens had constant gauge length of 18.21 mm. The dimensions were (1 to 2) mm (0.5 to 1) mm. Each material condition was tested for three specimen dimensions and 2 repeats in each combination. For machining miniature specimens, traditional machining processes such as wire EDM, laser machining, and chemical mechanical polishing, and FIB milling maybe used [5, 29, 30]. In this study, wire EDM and water jet machining were used to cut out miniature specimens. Wire EDM process is advantageous mainly because it produces a stress and burr-free cutting, has efficient production capabilities and is cost effective with an excellent finishing. The cutting mechanism in wire EDM is by bombarding the work piece with intense short pulses of electricity and each pulse leaves a tiny pit on the work piece [30]. This causes surface roughness on the specimen. Studies have proven that, the surface roughness that is induced during specimen preparation causes reduction in fatigue strength and areas of stress concentration. For example, fatigue strength of the SiC/Al metal matrix composite suffers due to the presence of surface roughness which is an aftermath of wire EDM machining. Therefore, the specimens were sanded before testing. Most of the specimens in the first batch did not fracture upon reaching maximum strain travel resulting in incomplete stress 14 Fig. 3.6: SEMTester 1000 EBSD instrument for miniature testing strain curves. Hence repeatability was very poor. Therefore, a second batch of specimens was designed with reduced gauge length and cross sectional dimensions to see complete failure of specimens and improve on the repeatability of the results. Keeping the aspect ratio comparable to that of the standard, the gauge length and width were maintained constant at 8 mm and 1 mm respectively. The thickness varied from 0.1 mm to 0.2 mm Figure 3.7. The sanding of the specimens causes slight decrease in thickness. From each material condition and 2 specimen designs, 2 repeats were conducted. Another batch of specimens was designed to be tested on Tinius Olsen machine Figure 3.9 in order to have a comparison of mechanical properties from two different tensile testers. The gauge length and width were maintained constant at 12 mm and 1 mm, respectively. The thickness varied from 0.5, 1, and 2 to 3 mm Figure 3.8. 15 Fig. 3.7: Specimen cross section at gage length is (1x0.2) mm Fig. 3.8: Variation in specimen gage thickness from 3 mm to 0.5 mm 16 (a) (b) Fig. 3.9: (a) Tinius Olsen H50KS (b) Specimen test in progress From each material condition and 4 specimen designs, 4 repeats were conducted to ensure repeatability. All the tests were conducted at a strain rate of 0.001s−1 . The Tinius Olsen H50KS is a 50KN capacity model having maximum crosshead travel of 1100 mm. Load measurement accuracy is ±0.5% of indicated load from 2% to 100% capacity and position measurement accuracy of ±0.01% of reading or 0.001 mm, whichever is greater. 17 Chapter 4 Experimental Results and Discussion 4.1 Conventional (Macro-sized specimen) Tensile Testing The same materials under study are tested on a conventional scale and the following results (from engineering stress-strain curves) are obtained from AZZ WRI facility. The specimens are designed with respect to ASTM A370-10 standard with a displacement rate of 0.075 in/min (1.9 mm/min) until 0.4% offset and post yield loading rate at 0.7 in/min (17.78 mm/min) until failure. Two tests per material is experimented on a Tinius Olsen tester. Table 4.1: Conventional (Macro-sized) results Specimen Material Stainless Steel (SS) grade 304/304L Sensitized (SS) grade 304/304L Carbon Steel 4.2 Average Yield Stress (YS) (MPa) Average Tensile Stress (UTS) (MPa) Young’s Modulus (GPa) Elongation (%) 276.48 610.19 187.88 61 269.59 617.08 165.82 57 344.39 529.17 204.08 35 Miniature Tensile Testing The miniature specimens designed for micro tensile tester are tested, and the Yield Strength (YS) and Ultimate Tensile Strength (UTS) obtained are as shown in Table 4.2. All specimens are flat and have a gauge length of 8 mm. The SEMTester 1000 EBSD is usually used for testing textiles, foods, biomaterials to paper and polymers. Hence, the results obtained for such high strength metals are not that accurate as that obtained from Tinius Olsen. Prestress values are taken care of during data analysis. All specimens are tested at a strain rate of 0.001s−1 . The results obtained from Tinius Olsen are as shown in Tables 4.3 18 Table 4.2: Results obtained from SEMTester 1000 EBSD Cross-sectional Yield Strength (YS) MPa Gauge Area(mm2 ) (Average value) Stainless Steel (SS) grade 304/304L 1x0.1 129.00 1x0.2 92.19 Sensitized (SS) grade 304/304L 1x0.1 575.00 1x0.2 172.02 Carbon Steel 1x0.1 47.57 1x0.2 48.34 Ultimate Tensile Strength (UTS) MPa (Average value) 781.84 1174.57 794.21 806.55 377.32 445.50 through 4.5. All specimens are flat and have a constant gauge length of 12 mm. Prestress values are subtracted (zeroed) during data analysis. All specimens are tested at a strain rate of 0.001s−1 . Application of Chauvenet’s Criterion: The experiment is affected in numerous ways such as human error while mounting the specimen and taking the readings, vibrations from surrounding instruments etc. All of these qualify as random errors. Certain errors like load cell defect and LVDT offsets qualify as systematic errors as the same error value is affecting every test. Also, in sample to sample experiments, the variability inherent in the samples themselves causes variations in measured values in addition to the random errors in the measurement system. A combination of these errors influence the results obtained from the instrument. Chauvenet’s criterion is applied to exclude the values which lie outside a certain range as dictated by the ratio of maximum acceptable deviation to standard deviation ∆Xmax Sx (4.1) From the table in Figure 4.1, Chauvenet’s criterion for rejecting a reading [1], for a set of four specimens in each cross-sectional gauge area, the absolute value of the ratio of Maximum Acceptable Deviation to Standard Deviation is 1.54. Hence, any ratio greater than 1.54 is considered as an outlier. The Chauvenet’s criterion can be applied on any stress value because 19 if one value is considered an outlier, the entire test on that specimen is discarded on the whole. Fig. 4.1: Chauvenet’s criterion for rejecting a reading obtained from the book Experimentation, Validation and Uncertainty Analysis for Engineers by Coleman and Steele [1] From the calculations in Tables 4.3 through 4.5, none of the tests had to be discarded. Thus, the experiment procedure was correct and all the results are valid. 4.3 Optimum Specimen Size The optimum miniature specimen size is the one whose results correspond to that of conventional tensile test results. From the comparison, the miniature stainless steel (SS) and Sensitized SS grade 304 specimens tested on Tinius Olsen give comparable yield and engineering tensile strength results. However, one can observe that there is a slight increase in yield strength (YS) values for miniature SS and Sensitized SS specimens when compared to its equivalent conventional sized specimens. 20 Table 4.3: Mechanical properties for Stainless Steel Type 304 miniature specimens Sl. No. Yield Strength (MPa) Specimen 1 2 3 4 X̄ 1 Sx 2 Interval Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Cross-Section 275.45 306.70 303.95 283.82 292.48 15.26 292.48±15.26 Cross-Section 296.41 301.92 306.97 285.99 297.82 8.99 297.82±8.99 Cross-Section 191.72 229.31 212.73 217.82 212.90 15.73 212.90±15.73 Cross-Section 157.69 147.38 160.38 173.45 159.73 10.73 159.73±10.73 1 2 Average Value Standard Deviation Xi − X̄ /Sx (1x3)mm2 1.12 0.93 0.75 0.57 (1x2)mm2 0.16 0.46 1.02 1.32 (1x1)mm2 1.35 1.04 0.01 0.31 (1x0.5)mm2 0.19 1.15 0.06 1.28 Ultimate Tensile Strength (MPa) Xi − X̄ /Sx Elongation (%) 691.00 640.85 640.42 611.00 645.81 33.20 645.81±33.20 1.36 0.15 0.16 1.05 58 66 63 66 63.25 3.77 63.25±3.77 690.07 620.00 639.10 592.10 635.32 41.29 635.32±41.29 1.33 0.37 0.09 1.05 50 58 42 46 49 6.83 49±6.83 508.00 579.06 524.00 588.04 549.76 39.71 549.76±39.71 1.05 0.74 0.65 0.96 50 46 50 42 47 3.83 47±3.83 543.00 435.00 500.00 511.00 497.25 45.33 497.25±45.33 1.01 1.37 0.06 0.30 25 25 46 25 30.25 10.50 30.25±10.50 21 Table 4.4: Mechanical properties for Sensitized Stainless Steel Type 304 miniature specimens Sl. No. Yield Strength (MPa) Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Cross-Section 301.38 262.31 267.01 305.85 284.14 22.65 284.14±22.65 Cross-Section 263.26 236.75 174.29 266.47 235.19 42.73 235.19±42.73 Cross-Section 149.71 197.64 105.84 156.14 152.33 37.58 152.33±37.58 Cross-Section 79.74 86.84 140.15 103.81 102.64 26.97 102.64±26.97 Xi − X̄ /Sx (1x3)mm2 0.76 0.96 0.76 0.96 (1x2)mm2 0.66 0.37 1.43 0.73 (1x1)mm2 0.07 1.21 1.24 0.10 (1x0.5)mm2 0.85 0.59 1.39 0.04 Ultimate Tensile Strength (MPa) Xi − X̄ /Sx Elongation (%) 740.00 683.42 548.39 612.00 645.95 83.51 645.95±83.51 1.13 0.45 1.17 0.41 66 58 66 42 58 11.31 58±11.31 546.00 526.00 443.60 595.00 527.65 63.09 527.65±63.09 0.29 0.03 1.33 1.07 50 42 58 58 52 7.66 52±7.66 472.00 485.95 407.00 455.00 454.99 34.40 454.99±34.40 0.49 0.90 1.40 0.00 58 58 42 46 51 8.25 51±8.25 299.00 341.00 398.00 373.00 352.75 42.76 352.75±42.76 1.26 0.27 1.06 0.47 38 50 33 42 40.75 7.18 30.25±10.50 22 Table 4.5: Mechanical properties for SA516 grade 70 carbon steel miniature specimens Sl. No. Yield Strength (MPa) Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Specimen 1 2 3 4 X̄ Sx Interval Cross-Section 301.99 291.92 282.37 287.34 290.91 8.36 290.91±8.36 Cross-Section 318.37 233.47 291.50 219.83 265.79 46.84 265.79±46.84 Cross-Section 180.81 208.90 106.05 161.29 164.26 43.45 164.26±43.45 Cross-Section 70.76 153.42 145.70 175.63 136.38 45.55 136.38±45.55 Xi − X̄ /Sx (1x3)mm2 1.33 0.12 1.02 0.43 (1x2)mm2 1.12 0.69 0.55 0.98 (1x1)mm2 0.38 1.03 1.34 0.07 (1x0.5)mm2 1.44 0.37 0.20 0.86 Ultimate Tensile Strength (MPa) Xi − X̄ /Sx Elongation(%) 465.44 440.00 432.00 433.00 442.61 15.63 442.61±15.63 1.46 0.17 0.68 0.61 25 33 25 25 27 4 27±4 461.00 362.00 465.00 494.00 445.50 57.58 445.50±57.58 0.27 1.45 0.34 0.84 25 33 21 25 26 5.03 26±5.03 328.00 350.00 296.01 307.00 320.25 23.86 320.25±23.86 0.32 1.25 1.02 0.56 42 21 17 25 26.25 10.99 26.25±10.99 203.00 240.00 250.00 352.00 261.25 63.79 261.25±63.79 0.91 0.33 0.18 1.42 17 25 25 13 20 6 20±6 23 This can be explained due to lesser presence of anomalies in the microstructure when compared to the macro-sized specimens. Also, lesser number of voids is present in a miniature specimen. Further, sensitized stainless steel is heat treated for several hours (168) leading to softening of the metal. This is explained due to the diffusion of Cr, C and N towards the grain boundaries from the grain body, thus indicating a fall in hardness of stainless steel. This is usually seen after 1 hour of sensitization. When the duration is longer than 1 hour, the hardness marginally increases due to the precipitation at the grain boundaries and also formation of martensitic microstructure [31]. The carbon steel specimens show more ductility owing to lesser yield strength (YS) values at miniature level. Such a ductile nature is seen only in carbon steel specimens unlike the stainless steel (SS) specimens hence carbon specimens were more sensitive to miniaturization. Also, the tensile strength values are lower than the conventional results. A plausible reasoning is due to the brittle nature of carbon steel and at miniature level the volume is much lesser than the conventional owing to lesser strength and higher ductility. Very thin cross sections that were tested on microtensile tester gave considerably high results for SS specimens and slightly comparable results for carbon steel specimens as shown in Figure 4.2 and 4.3. Considering all the above reasoning, the recommended miniature size for all three materials is a gauge area of (1 × 3) with an aspect ratio 4:1 which is the same as that of the standard ASTM testing for a constant gauge length of 12 mm. The overall specimen size was approximately 50 mm (2 in.). The engineering stress versus strain curve for four specimens of gauge dimension (1 × 3) from each material is as shown in Figure 4.4. 24 800 700 Tensile Strength (MPa) 600 500 400 300 200 100 0 0 0.5 1 1.5 2 2.5 3 3.5 3 3.5 3 3.5 Cross-Sectional Area (mm2) (a) Stainless Steel grade 304 800 700 Tensile Strength (MPa) 600 500 400 300 200 100 0 0 0.5 1 1.5 2 2.5 Cross-Sectional Area (mm2) (b) Sensitized Stainless Steel grade 304 600 Tensile Strength (MPa) 500 400 300 200 100 0 0 0.5 1 1.5 2 2.5 Cross-Sectional Area, (mm2) (c) Carbon Steel(SA5 16) grade 70 Fig. 4.2: Tensile strength vs specimen gauge cross section 25 350 Yield Strength (MPa) 300 250 200 150 100 50 0 0 0.5 1 1.5 2 2.5 3 3.5 Cross-Sectional Area (mm2) (a) Stainless Steel grade 304 350 Yield Strength (MPa) 300 250 200 150 100 50 0 0 0.5 1 1.5 2 2.5 Cross-Sectional Area (mm2) 3 3.5 (b) Sensitized Stainless Steel grade 304 350 Yield Strength (MPa) 300 250 200 150 100 50 0 0 0.5 1 1.5 2 2.5 3 3.5 Cross Sectional Area (mm2) (c) Carbon Steel(SA5 16) grade 70 Fig. 4.3: Yield strength vs specimen gauge cross section 26 800 700 Stress (MPa) 600 500 400 300 200 Specimen Specimen Specimen Specimen 100 0 0 0.1 0.2 0.3 0.4 0.5 Strain (mm/mm) 0.6 0.7 1 2 3 4 0.8 0.9 (a) Stainless Steel grade 304 800 700 Stress (MPa) 600 500 400 300 200 Specimen Specimen Specimen Specimen 100 0 0 0.1 0.2 0.3 0.4 0.5 Strain (mm/mm) 0.6 0.7 0.8 1 2 3 4 0.9 (b) Sensitized Stainless Steel grade 304 500 450 400 Stress (MPa) 350 300 250 200 150 Specimen Specimen Specimen Specimen 100 50 0 0 0.05 0.1 0.15 0.2 Strain (mm/mm) 0.25 0.3 1 2 3 4 0.35 (c) Carbon Steel(SA5 16) grade 70 Fig. 4.4: Engineering stress vs strain curves for specimen gauge cross section (1x3) 27 When gauge length is maintained constant, with decrease in thickness, the gauge part is effectively transformed from a bulk to sheet geometry and the stress state within the gauge changes from a more or less biaxial to a uniaxial stress state condition, thereby resulting in a change from diffuse necking to localized necking. In a tension test on a ductile material, a diffuse necking - so called because its spatial extension is much larger than the sheet thickness - begins to develop in the sample when the strain hardening is no longer able to compensate for the weakening due to the reduction of the cross-section. After some elongation under decreasing load, a localized neck usually appears in the region of the diffuse neck. In the localized neck, severe thinning occurs leading to ultimate failure. The transition from diffuse to localized neck happens much faster in a sheet specimen (in this case miniature specimen) when compared to a bulk specimen (conventional specimen). This causes reduction in yield strength since elastic zone is shortened and hence, material exhibits more ductility as a sheet miniature specimen than a conventional specimen. Miniature specimens have smaller volumes compared to conventional specimens. Therefore, it might be because of smaller volume, the interatomic forces give way too soon under tension (as they have small number of grains) causing decreased strain hardening effect.Yield stress is more sensitive to increase in strain rate than the tensile strength. High rates of strain cause a yield point to appear in specimens of low carbon steel which otherwise do not appear when tested under ordinary rates of loading [32]. All specimens were tested with a strain rate of 0.001s−1 . However, miniature carbon steel specimens show yield point as seen in Figure 4.4(c). The Modulus of elasticity in all three material specimens was found to be as low as about 6895 MPa (1000 ksi) in all the three specimens. There are various reasoning for this behavior of miniature specimens. The literature shows that with the increase of gauge length, the Youngs Modulus values also increase with different strain rates [2]. A strong dependence of elastic modulus on gauge length suggests that some shear might occur causing a deviation from the linear stress-strain relationship [2]. It is reported in many studies that the smaller the gauge length, the lower the slope of the 28 linear stress/strain relationship is detected for samples with shorter gauge length (with respect to the volume of the conventional specimens). However, shorter linear portion of stress strain graph does not influence the overall elongation of the material until maturation. Also, steel crystal is highly anisotropic; grain orientation anisotropy can be a valuable explanation for low Modulus values [33]. Youngs Modulus is the materials resistance to elastic deformation. The greater the modulus, the stiffer the material or smaller the elastic strain that results from application of a given stress. On atomic scale, macroscopic elastic strain is manifested as small changes in the interatomic spacing and the stretching of interatomic bonds. As a consequence, the magnitude of the modulus is a measure of resistance to separation of adjacent atoms which is interatomic bonding forces [34]. 29 Chapter 5 Uncertainty analysis 5.1 Tolerance intervals in Sample Populations In addition to the confidence interval for a sample population, there is a useful statistical interval called the tolerance interval. The tolerance interval gives information about the parent population that the specimens came from [1]. The Gaussian interval that contains 95% of the parent population is simply the X̄ ± 1.96σ, i.e the average value of the specimens being tested ± 1.96 times the random standard deviation. However, when we are dealing with specimens and the statistical quantities X̄ and Sx , the determination of the range that contains a portion of the parent population is not as straightforward. What one can estimate is the range that has a certain probability of containing a specified percentage of the parent population. Considering a set of four specimens of (1 × 3) cross-sectional area, and to find an estimate of 99% of the parent population with 95% confidence is X ± CT 95(99) (4) Sx (5.1) where CT is a factor which is given as a function of the number of readings in Table A.3 as in Figure 5.1 [1], for four specimens CT 95(99) factor is 8.299. (Here the number of specimens is four but note that if the number of specimens tends to infinity, the concept of a confidence level for the tolerance interval does not apply because the tolerance interval approaches the Gaussian interval.) For YS readings of SS grade 304 specimens with gauge length 12 mm, using the mean X 292.48 MPa and Sx 15.26 MPa from Table 4.3 in Eq.5.1 292.48 ± [8.299] [15.26] M P a (5.2) 30 Fig. 5.1: Factors for two-sided tolerance interval [1] prob (165.80 ≤ µ ≤ 419.17) M P a (5.3) For UTS readings, using the mean X 645.82 MPa and Sx 33.21 MPa in Eq.5.1 645.82 ± [8.299] [33.21] M P a (5.4) prob (370.25 ≤ µ ≤ 921.39) M P a (5.5) For YS readings of Sensitized SS grade 304 specimens with gauge length 12 mm, using the mean X 284.14 MPa and Sx 22.65 MPa from Table 4.4 in Eq.5.1 31 284.14 ± [8.299] [22.65] M P a (5.6) prob (96.17 ≤ µ ≤ 472.10) M P a (5.7) For UTS readings, using the mean X 645.95 MPa and Sx 83.50 MPa in Eq.5.1 645.95 ± [8.299] [83.50] M P a (5.8) prob (−47.04 ≤ µ ≤ 1338.93) M P a (5.9) For YS readings of carbon steel (SA5 16) grade 70 specimens with gauge length 12 mm, using the mean X 290.91 MPa and Sx 8.36 MPa from Table 4.5 in Eq.5.1 290.91 ± [8.299] [8.36] M P a (5.10) prob (221.56 ≤ µ ≤ 360.26) M P a (5.11) For UTS readings, using the mean X 442.61 MPa and Sx 15.63 MPa in Eq.5.1 442.61 ± [8.299] [15.63] M P a (5.12) prob (312.89 ≤ µ ≤ 572.33) M P a (5.13) Eq.5.3,5.7 and 5.11 gives the tolerance interval for 99% of the parent population with 95% confidence for YS. From Table 4.1, the conventional YS of SS specimens are 276.48 MPa, Sensitized SS specimens are 269.59 MPa and carbon steel specimens are 344.39 MPa. Eq.5.5,5.9 and 5.13 gives the tolerance interval for UTS. The conventional UTS of SS specimens are 610.19 MPa, Sensitized SS specimens are 617.08 MPa and carbon steel specimens are 529.17 MPa. Hence our miniature specimen tensile testing experiments give approximate values as that of the conventional testing as per ASTM A370-10 standard. 32 5.2 Propagation of Mechanical Properties by Monte Carlo Method (MCM) Monte Carlo simulation provides a distribution of errors for a result that is a function of multiple variables. In most cases, the result of our experiment or simulation will depend on several variables through a data reduction equation (DRE) or a simulation solution. This method is not limited to simple expressions but can also be used for highly complicated equations or for numerical solutions of advanced simulation equations. In this section, a general approach for the MCM for uncertainty propagation is discussed and this technique is applied to the data reduction equations of the current study. The mechanical properties of the three materials are propagated to parent population. It is seen that our experimental results, obtained by a set of four specimens in each gauge cross section, fall in the same interval as that of the propagated values of the parent population. The section concludes with a discussion for the convergence of the combined standard uncertainty (or relative uncertainty with respect to each iteration represnting a specimen) in this MCM analysis. 5.2.1 General Approach for MCM Figure 5.2 presents a flowchart that shows the steps involved in performing an uncertainty analysis by the MCM. The figure shows the sampling techniques of two variables, but the methodology is general for DREs or simulations that are functions of multiple variables. First, the assumed true value for each variable (load, area, stress etc.) in the result is the input. These would be the Xbest values that we have for each variable. Then, the estimates of the random standard uncertainty, s and the elemental systematic standard uncertainties bk for each variable are input. An appropriate probability distribution function is assumed for each error source. The random errors are usually assumed to come from a Gaussian distribution. The systematic errors are chosen based on the analyst’s judgment as it is upto the analyst to use the best information. 33 Fig. 5.2: Schematic for MCM for uncertainty propagation when random standard uncertainties for individual variables are used [1] For the flowchart in Figure 5.2, the random standard uncertainties for X and Y are assumed to come from Gaussian distributions and that each variable has three elemental systematic standard uncertainties, one Gaussian, one triangular, and one rectangular. For each variable, random values for the random errors and each elemental systematic error are found using an appropriate random number generator (Gaussian, triangular, rectngular, etc.). The individual error values are then summed and added to the true values of the variables to obtain ”measured” values. Using these measured values, the result is calculated. This process corresponds to running the simulation once. 34 5.2.2 Propagation of Mechanical Properties The Data Reduction Equations (DRE) for the study are the equations for YS, UTS and elongation (el) given by, σy (M P a) = load area (5.14) σ (M P a) = load area (5.15) el = lf − l0 l0 (5.16) where σy is YS, σ is UTS, lf is final gauge length after fracture and l0 is the original gauge length. The unit of load is in KN and that of area is mm2 . All lengths are in mm. However, elongation or strain is dimensionless. Before testing, the thickness and width of the specimens are carefully measured at three different locations in the gauge length. The average of the readings is considered as the thickness and the width values which are subsequently used for calculation of area. Random standard uncertainty for (95%) confidence interval (Eq.5.17) for area is calculated using Table 5.1. U ncertainty (95%) = 1.96 × Sx (5.17) 35 The uncertainty analysis is conducted only on optimum specimen size as it yields almost the same mechanical properties as that of a conventional specimen. Table 5.1: Specimen area Specimen No. Area (mm2 ) SS grade 304 - (1x3) specimens 1 2.90 2 2.89 3 2.88 4 2.88 Average 2.89 Standard Deviation 0.00957 Sensitized SS grade 304 - (1x3) specimens 1 4.26 2 2.84 3 2.74 4 2.83 Average 3.17 Standard Deviation 0.07297 Carbon Steel (SA5 16) grade 70 - (1x3) specimens 1 2.97 2 2.95 3 2.88 4 2.90 Average 2.93 Standard Deviation 0.04203 For SS 304 = 1.96 × 0.00957 ∴Uncertainty for area = 0.0188mm2 or 1.8% For Sensitized SS 304 = 1.96 × 0.07297. ∴Uncertainty for area = 0.14303mm2 or 14.30% For Carbon Steel 70 = 1.96 × 0.04203 ∴Uncertainty for area = 0.0824mm2 or 8.24% Below is the table for force applied by the load cell in each test for yielding to take place. Yielding point is crucial in every material testing as the specimen enters the plastic zone. Tensile strength does not depend on when the yielding takes place for a given material. The 36 load cell has a value of +/ − 0.5% of the applied load value. This specification is in the Tinius Olsen manual. Table 5.2: Load values for (YS) and (UTS) in specimens of gauge area (1x3) mm2 Specimen No. (YS) Load (KN) (UTS) Load (KN) SS grade 304 - (1x3) specimens 1 0.80 2.01 2 0.89 1.85 3 0.87 1.84 4 0.82 1.86 Average 0.85 1.89 Sensitized SS grade 304 - (1x3) specimens 1 1.28 3.15 2 0.74 1.94 3 0.73 1.50 4 0.86 1.73 Average 0.90 2.08 Carbon Steel (SA5 16) grade 70 - (1x3) specimens 1 0.90 1.38 2 0.86 1.30 3 0.81 1.24 4 0.83 1.25 Average 0.85 1.29 Stainless steel grade 304/304L Table 5.3: Variables and their uncertainties - SS grade 304 Variables Area (A)mm2 YS Load (F )KN UTS Load (Fu )KN Nominal Value 2.89 0.85 1.89 Uncertainty (95%) 1.8% 0.5% 0.5% Calculations for finding the uncertainty of YS using Eq.5.14 Uσy 2 ∂Uσy 2 ∂Uσy 2 2 = (UF ) + (UA )2 ∂F ∂A 2 −F 2 1 2 = (UF ) + (UA )2 A A2 (5.18) 37 The MCM nominal value (true value) for load and area (variables) are obtained from running the MCM algorithm for 10000 (M times) iterations in MATLAB. However, since the nominal value for area is very less, no significant changes were seen after running 10000 iterations, although the load values changed every single time. 10000 iterations is not just a random number chosen. It is later seen in this section that this number generates a converged value for relative uncertainty of the results. If we assume that all distributions are Gaussian, then the standard uncertainty inputs for the MCM will be previous percent uncertainties divided by 2. Plugging the nominal values obtained for the variables from the program, Eq.5.18 becomes Uσy 2 = 1 2.89 2 0.8452 × 103 × 0.005 1.96 2 + −0.85 × 103 2.892 2 2.9 × 1.8 × 10−2 1.96 2 = 0.5565 ± 7.3464 = 7.90 (5.19) In the above equation, (UF ) is 0.8452 × 103 (N ) × 0.005/1.96 where 0.8452 is the MCM nominal value for load to cause yielding and multiplied by one standard uncdertainty, i.e the standard uncertainty inputs for the MCM will be previous percent uncertainties divided by 2 or 1.96 for considering only the random uncertainties at 95% confidence. Similarly, (UA ) is (2.9 × 1.8 × 10−2 )/1.96 where 2.9 mm2 is the MCM nominal value for area. ∴ Uσy = 2.81 MPa Therefore, the uncertainty in the YS is the average value obtained from our experiment, 292.48 ± 2.81 MPa. Calculations for finding uncertainty in UTS using Eq.5.15 ∂Uσ 2 ∂Uσ 2 (UFu )2 + (UA )2 ∂Fu ∂A 2 −Fu 2 1 2 = (UFu ) + (UA )2 A A2 (Uσ )2 = (5.20) 38 The nominal value of the load changes as the load applied at breaking is different from the load applied at yielding. 2 (Uσ ) = 1 2.89 2 1.8942 × 103 × 0.005 1.96 2 + −1.89 × 103 2.892 2 2.9 × 1.8 × 10−2 1.96 2 (5.21) = 2.7949 ± 36.3211 = 39.12 ∴ (Uσ ) = 6.25 MPa Therefore, the uncertainty in the UTS is the average value obatined from our experiment, 645.82 ± 6.25 MPa. For finding the uncertainty in elongation, the final length of each specimen needs to be calculated after testing. The final specimen length was measured from the neck of the gauge length of the fractured specimen. Table 5.4: Final length of specimens of gauge cross section (1x3) after tensile testing - SS grade 304 Specimen No. 1 2 3 4 Average Standard Deviation Final Length lf mm 19.00 20.00 19.50 20.00 19.63 0.4787 From Eq.5.17, uncertainty for lf is 0.9383 mm. Therefore, the final length is 19.63 ± 0.9383 mm. Using Eq.5.16, the mean of the final length as 19.63 and the original gauge length as 12 mm, elongation E is 0.636 or 64%. Calculations for finding uncertainty in E using Eq.5.16 are 2 ∂Uel 2 ∂Uel 2 Ulf + (Ul0 )2 ∂lf ∂l0 2 2 −lf 2 1 = Ulf + (Ul0 )2 2 l0 l0 (Uel )2 = (5.22) 39 The original length for all specimens is 12 mm nevertheless the least count of the measuring caliper is taken as the uncertainty. In this case the uncertainty value is 0.001. Upon substitution of nominal values for final and original length randomly generated by MATLAB for MCM analysis, (Uel )2 = 1 12 2 20.468 × 0.9383 1.96 2 + −19.63 122 2 11.99 × 0.001 1.96 2 (5.23) = 0.6667 ∴ (Uel ) = 0.8165 Therefore, the total uncertainty in the elongation is within the range of 0.636 ± 0.8165 or 64% ± 81.6% of the final length. This elongation can be really high if the material is ductile. A ductile material has low yielding point and hence undergoes yielding for a long period before failure. Another reason for high elongation is inclusions in the alloy material. Note that when 10000 iterations represents 10000 experimental trials that have random outcomes [35]. When applied to uncertainty estimation, random numbers are used to ramdomly sample parameters’ uncertainty space instead of point calculation carried out by a small test of experiments. Hence it is a method where propagation of uncertainty is achieved. Therefore, the possibility of elongation being greater than the value itself is almost insignificant. The range is a number interval within which elongation may lie without reaching to the extremes. Sensitized SS grade 304/304L Table 5.5: Variables and their uncertainties - Sensitized SS grade 304 Variables Area (A)mm2 YS Load (F )KN UTS Load (Fu )KN Nominal Value 3.17 0.90 2.08 Uncertainty (95%) 14.30% 0.5% 0.5% 40 Uncertainty calculation in YS is obtained by using Eq.5.14 and 5.18. The nominal value for load and area are obtained from running the MCM algorithm for 10000 iterations in MATLAB. Uσy 2 = 1 3.17 2 0.9105 × 103 × 0.005 1.96 2 + −0.90 × 103 3.172 2 2 3.16 × 14.30 × 10−2 1.96 (5.24) ∴ Uσy = 21.38 MPa Therefore, the uncertainty in the YS is the average value obtained in our experiment, 284.14 ± 21.38 MPa. Calculations for finding uncertainty in UTS using Eq.5.15 and 5.20 2 (Uσ ) = 1 3.17 2 2.0756 × 103 × 0.005 1.96 2 + −2.08 × 103 3.172 2 3.16 × 14.30 × 10−2 1.96 2 (5.25) ∴ Uσ = 47.77 MPa Therefore, the uncertainty in the UTS is the average value obtained in our experiment, 645.95 ± 47.77 MPa. Uncertainty in elongation is calculated based on the following table. Table 5.6: Final length of specimens of gauge cross section (1x3) after tensile testing - Sensitized SS grade 304 Specimen No. 1 2 3 4 Mean Standard Deviation Final Length lf mm 20.00 19.00 20.00 17.00 19.00 1.414 From Eq.5.17, uncertainty for lf is 2.7714 mm. Therefore, the final length is 19.00 ± 2.7714 mm. Using Eq.5.16, the mean of the final length as 19.00 and the original gauge length 41 as 12 mm, elongation E is 0.583 or 58%. Calculations for finding uncertainty in E using Eq.5.16 and Eq.5.22 2 (Uel ) = 1 12 2 18.11 × 2.7714 1.96 2 + −19 122 2 11.99 × 0.001 1.96 2 (5.26) ∴ Uel = 2.1339 Therefore, the uncertainty in the elongation is 0.5383 ± 2.1339. Carbon steel SA516 grade 70 Table 5.7: Variables and their uncertainties - carbon steel SA516 grade 70 Variables Area (A)mm2 YS Load (F )KN UTS Load (Fu )KN Nominal Value 2.93 0.85 1.29 Uncertainty (95%) 8.24% 0.5% 0.5% Uncertainty calculation in YS is obtained by using Eq.5.14 and 5.18. The nominal value for load and area are obtained from running the MCM algorithm for 10000 iterations in MATLAB. Uσy 2 = 1 2.93 2 0.8459 × 103 × 0.005 1.96 2 + −0.85 × 103 2.932 2 2.90 × 8.24 × 10−2 1.96 2 (5.27) ∴ Uσy = 12.09 MPa Therefore, the uncertainty in the YS is the average value obtained in our experiment, 290.91 ± 12.09 MPa. Calculations for finding uncertainty in UTS using Eq.5.15 and 5.20 2 (Uσ ) = 1 2.93 2 1.2923 × 103 × 0.005 1.96 2 + −1.29 × 103 2.932 2 2.90 × 8.24 × 10−2 1.96 2 (5.28) ∴ Uσ = 18.56 MPa 42 Therefore, the uncertainty in the UTS is the average value obtained in our experiment, 442.61 ± 18.56 MPa. Uncertainty in elongation is calculated based on the following table. 43 Table 5.8: Final length of specimens of gauge cross section (1x3) after tensile testing - carbon steel SA516 grade 70 Specimen No. 1 2 3 4 Mean Standard Deviation Final Length lf mm 15.00 16.00 15.00 15.00 15.25 0.5 From Eq.5.17, uncertainty for lf is 0.98 mm. Therefore, the final length is 15.25 ± 0.98 mm. Using Eq.5.16, the mean of the final length as 15.25 and the original gauge length as 12 mm, elongation E is 0.271 or 27%. Calculations for finding uncertainty in E using Eq.5.16 and 5.22 2 (Uel ) = 1 12 2 15.17 × 0.98 1.96 2 + −15.25 122 2 11.99 × 0.001 1.96 2 (5.29) ∴ Uel = 0.6321 Therefore, the uncertainty in the elongation is 0.271 ± 0.6321. 5.3 Convergence study The sampling process is repeated M times to obtain a distribution for the possible result values. The primary goal of the MCM propagation technique is to estimate a converged value for the standard deviation SM CM , of this distribution. 2s of this distribution is the resultant uncertainty at 95% confidence (assuming the distribution is Gaussian). An appropriate value for M is determined by periodically calculating SM CM during the MCM process and stopping the process when a converged value of SM CM is obtained. The SM CM is the combined standard uncertainty of the result (uncertainty in each reading/iteration). The number of iterations, in this case is 10000 (M ). We do not need to have a perfectly converged value of SM CM to have reasonable estimate of uncertainty. Once the SM CM values are converged to within 1-5%, then the value of SM CM 44 is a good approximation of the combined standard uncertainty of the result. Figure 5.3 gives the distribution for uncertainties calulated in Eq.5.27 for YS in carbon steel specimens. The resultant uncertainty is calulated using Eq.5.30 2UF F (5.30) M CM where, UF is the uncertainty calculated using the true value generated for a variable (load) in the Eq.5.27 and F is the true value generated by MCM process. This value of expanded uncertainty is stored with every iteration. In Figure 5.3, load and area uncertainties are included. The distribution is Gaussian, hence our assumption is valid. 350 300 No. of Specimens 250 200 150 100 50 0 2.5 2.6 2.7 2.8 2.9 3 3.1 Uncertainty, % 3.2 3.3 3.4 3.5 Fig. 5.3: Distribution of MCM results for yield strength of carbon steel SA516 grade 70. Expanded uncertainties for each variable being calculated at 95% confidence The averge uncertainty is 3.0344% and the standard deviation is 0.1200% for the entire range of specimens. The convergence plot is as shown in Figure 5.4. The plot describes the convergence of relative uncertainty of the specimens. The value of standard deviation, s of the resultant 45 uncertainty as in Eq.5.30 was calculated after every iteration, starting from the first iteration. The plot of these s values which represent combined standard uncertainty, SM CM is plotted in Figure 5.4. The value of relative uncertainty or the combined standard uncertainty was a fully converged value from a large number of iterations as seen. 0.08 0.07 Relative uncertainty, % 0.06 0.05 0.04 0.03 0.02 0.01 0 0 1000 2000 3000 4000 5000 6000 7000 8000 Monte Carlo Iterations, M (No. of specimens) 9000 10000 Fig. 5.4: Convergence study for MCM value of UF for 95% combined uncertainty for each variable We see that after 1000 iterations, the value has converged to within 3% and by about 2000 iterations to within less than 1% of the fully converged value. About 6.3% is the combined standard uncertainty for 10000 specimens. Since these are characteristic plots, similar trend is seen with mechanical properties of stainless steel and sensitized steel as well. Any further addition to 10000 would still give us a converged value and hence our selection of 10000 iterations is reasonable. 46 Chapter 6 Conclusion and Future Work 6.1 Conclusions This study clearly shows that the tensile testing results of miniature specimens are de- pendent not only on the property of the material itself and testing conditions (strain rate, tensile machine employed, etc.), but also on specimen size and geometry. Miniature specimens can be of any cross-section but to choose the best cross-section for a particular application is challenging. Four different specimen gauge cross-sections having the same or multiple of the ASTM standard aspect ratio is designed with a constant gauge length of 12mm. With the current facilities, the specimens were tested upon two different machines, Tinuis Olsen and SEMTester 1000 EBSD. From the results obtained, larger consistency in results was obtained from Tinius Olsen machine. The specimens with gauge cross- section (1 × 3) shows repeatability in results which is considerably agreeable to that of the ASME- SA 240 standard and the macro sized testing performed by AZZ WSI, hence considered an optimum specimen cross-section. The variation in results is present in all specimen lots but more pronounced in the smaller cross-section specimens. Variations are attributed to a number of factors such as stress concentration effects while machining the specimen, especially near the fillet area and voids present within. To minimize any undue stress, the specimen must be aligned with the center line of the two test machine grips [11]. Therefore, Chauvenets criterion is applied to remove the results that are most effected from the above mentioned factors so that more reliable mean and standard deviation is obtained. However, in the optimum specimen size considered, there was no big variation that was observed after the application of Chauvenets criterion. Tolerance interval for the parent population is estimated within 95% confidence and the method assumes to have a Gaussian parent population. To exclude the assumption made previously, MCM analysis for 47 uncertainty propagation is employed as the inputs to MCM analysis need not be Gaussian. The mean of the UTS for SS and Sensitized SS grade 304/304L specimens may lie within 370.25 to 921.39 MPa and 1338.93 MPa, respectively. The mean for the experimented lot in this study for SS specimens is 645.82 ± 6.25 MPa and for Sensitized SS specimens is 645.95 ± 47.77 MPa, which lies well within the range. The range of UTS within which Carbon Steel specimens may lie is 312.89 to 572.33 MPa and the mean for the experimented lot from this study is 442.61 ± 18.56 MPa which is well within the expected range. A conclusion can be drawn from this study that uncertainty analysis must be conducted in sample to sample experiments to know whether the range of the results obtained are agreeable with that of the standard results. In most studies, we see that a certain experiment is simply rejected for giving inconsistent results failing to understand the random and systematic errors that are prevalent in experimentation. Propagation of uncertainty is required in every research field to know whether the experimental results observe the same trend as that of the parent population. 6.2 Future Work The mechanical properties obatined with the set of specimens, categorized as optimum specimen design in this thesis gives a good approximation to that of the conventional specimens. The uncertainty analysis further ensures our study and validates the properties obtained on a large scale. However, the miniature specimens in this study was focussed on high strength steels. Many types of metals can be studied at miniature level, specimens obtained from boat samples to understand its behavior. Many a time, structures are subjected to high temperatures and is difficult to conduct a conventional test of an equivalent material at that temperature. Miniature mechanical properties should be determined at varied temperature levels. Further, SEM should be employed along with the SEMTester, miniature testing instrument to find out the physics at miniature level as many properties depend on a large number of variables. 48 References [1] Coleman, H. W. and Steele, W. G., Experimentation, Validation, and Uncertainty Analysis for Engineers, John Wiley & Sons, United States, 2009. [2] Sergueeva, A., Zhou, J., Meacham, B., and Branagan, D., “Gage Length and Sample Size Effect on Measured Properties during Tensile Testing,” Materials Science and Engineering: A, Vol. 526, No. 1, 2009, pp. 79–83. [3] Karthik, S. V., “Development of Miniature Specimen Test Techniques,” Post-Irradiation Examination Division, 2013, pp. 194–195. [4] Madia, M., Foletti, S., Torsello, G., and Cammi, A., “On the Applicability of the Small Punch Test to the Characterization of the 1CrMoV Aged Steel: Mechanical Testing and Numerical Analysis,” Engineering Failure Analysis, Vol. 34, 2013, pp. 189–203. [5] Hemker, K. and Sharpe Jr, W., “Microscale Characterization of Mechanical Properties,” Annu. Rev. Mater. Res., Vol. 37, 2007, pp. 93–126. [6] Lucas, G., “Review of Small Specimen Test Techniques for Irradiation Testing,” Metallurgical Transactions A, Vol. 21, No. 4, 1990, pp. 1105–1119. [7] Sun, X., Soulami, A., Choi, K. S., Guzman, O., and Chen, W., “Effects of Sample Geometry and Loading Rate on Tensile Ductility of TRIP800 Steel,” Materials Science and Engineering: A, Vol. 541, 2012, pp. 1–7. [8] Zhao, Y., Guo, Y., Wei, Q., Dangelewicz, A., Xu, C., Zhu, Y., Langdon, T., Zhou, Y., and Lavernia, E., “Influence of Specimen Dimensions on the Tensile Behavior of Ultrafinegrained Cu,” Scripta Materialia, Vol. 59, No. 6, 2008, pp. 627–630. [9] Hoffmann, H. and Hong, S., “Tensile Test of Very Thin Sheet Metal and Determination of Flow Stress Considering the Scaling Effect,” CIRP Annals-Manufacturing Technology, Vol. 55, No. 1, 2006, pp. 263–266. [10] Michel, J.-F. and Picart, P., “Size Effects on the Constitutive Behaviour for Brass in Sheet Metal Forming,” Journal of Materials Processing Technology, Vol. 141, No. 3, 2003, pp. 439–446. [11] LI, C.-T. and Langley, N. R., “Improvement in Fiber Testing of High-Modulus SingleFilament Materials,” Journal of the American Ceramic Society, Vol. 68, No. 8, 1985, pp. C–202. [12] Zhao, Y., Guo, Y., Wei, Q., Topping, T., Dangelewicz, A., Zhu, Y., Langdon, T., and Lavernia, E., “Influence of Specimen Dimensions and Strain Measurement Methods on Tensile Stress–Strain Curves,” Materials Science and Engineering: A, Vol. 525, No. 1, 2009, pp. 68–77. 49 [13] Decamp, K., Bauvineau, L., Besson, J., and Pineau, A., “Size and Geometry Effects on Ductile Rupture of Notched Bars in a C-Mn Steel: Experiments and Modelling,” International Journal of Fracture, Vol. 88, No. 1, 1997, pp. 1–18. [14] LaVan, D. and Sharpe Jr, W., “Tensile Testing of Microsamples,” Experimental Mechanics, Vol. 39, No. 3, 1999, pp. 210–216. [15] Jackson, J. and Freed, M., “The Effect of Specimen Geometry on the Tensile Strength of Graphite,” Carbon, Vol. 3, No. 3, 1965, pp. 257–259. [16] Osipov, V. and Lyafer, E., “Effect of Sheet-Specimen Geometry on Plasticity and the Deformation Curve,” Strength of Materials, Vol. 3, No. 8, 1971, pp. 974–980. [17] Goh, T. and Shang, H., “Effects of Shape and Size of Tensile Specimens on the StressStrain Relationship of Sheet-Metal,” Journal of Mechanical Working Technology, Vol. 7, No. 1, 1982, pp. 23–37. [18] Matic, P., Kirby III, G., Jolles, M., and Father, P., “Ductile Alloy Constitutive Response by Correlation of Iterative Finite Element Simulation with Laboratory Video Images,” Engineering fracture mechanics, Vol. 40, No. 2, 1991, pp. 395–419. [19] Pan, N., Chen, H., Thompson, J., Inglesby, M., Khatua, S., Zhang, X., and Zeronian, S., “The Size Effects on the Mechanical Behaviour of Fibres,” Journal of materials science, Vol. 32, No. 10, 1997, pp. 2677–2685. [20] Silva, F. d. A., Chawla, N., et al., “Tensile Behavior of High Performance Natural (Sisal) Fibers,” Composites Science and Technology, Vol. 68, No. 15, 2008, pp. 3438–3443. [21] Armstrong, R., “On Size Effects in Polycrystal Plasticity,” Journal of the Mechanics and Physics of Solids, Vol. 9, No. 3, 1961, pp. 196–199. [22] Wedberg, D., Modelling of High Strain Rate Plasticity and Metal Cutting, PhD thesis, Lule University of Technology, 2013. [23] Poling, W., Grain Size Effects in Micro-tensile Testing of Austenitic Stainless Steel , Colorado School of Mines, Golden, Colorado, 2012. [24] Alves Fidelis, M. E., Pereira, T. V. C., Gomes, O. d. F. M., de Andrade Silva, F., and Toledo Filho, R. D., “The Effect of Fiber Morphology on the Tensile Strength of Natural Fibers,” Journal of Materials Research and Technology, Vol. 2, No. 2, 2013, pp. 149–157. [25] Simons, G., Weippert, C., Dual, J., and Villain, J., “Size Effects in Tensile Testing of Thin Cold Rolled and Annealed Cu Foils,” Materials Science and Engineering: A, Vol. 416, No. 1, 2006, pp. 290–299. [26] Kohyama, A., Hamada, K., and Matsui, H., “Specimen Size Effects on Tensile Properties of Neutron-Irradiated Steels,” Journal of Nuclear Materials, Vol. 179, 1991, pp. 417–420. [27] KALP1681, Mechanical Behavior, Testing, and Manufacturing Properties of Materials, University of Notre Dame, Notre Dame, Indiana, 2009. 50 [28] Park, J.-W. and Lee, C.-K., “Mechanical Properties and Sensitization on Clad Steel Welding Design,” International Journal of Precision Engineering and Manufacturing, Vol. 14, No. 11, 2013, pp. 1939–1945. [29] Ramulu, M., Paul, G., and Patel, J., “EDM Surface Effects on the Fatigue Strength of a 15 vol% SiC/Al Metal Matrix Composite Material,” Composite Structures, Vol. 54, No. 1, 2001, pp. 79–86. [30] Daud, M. A. M., Omar, M. Z., and Sajuri, J. S. Z., “Effect of Wire-EDM Cutting on Fatigue Strength of AZ61 Magnesium Alloy,” Jurnal Mekanikal , Vol. 30, 2010, pp. 68–76. [31] Ghosh, S., Kain, V., Ray, A., Roy, H., Sivaprasad, S., Tarafder, S., and Ray, K., “Deterioration in Fracture Toughness of 304LN Austenitic Stainless Steel due to Sensitization,” Metallurgical and Materials Transactions A, Vol. 40, No. 12, 2009, pp. 2938–2949. [32] Dieter, G. E. and Bacon, D., Mechanical Metallurgy, Vol. 3, McGraw-Hill, New York, 1986. [33] Gandhi, U., “Investigation of Anisotropy in Elastic Modulus of Steel,” Toyota Technical Center , 2010, pp. 1–17. [34] Hertzberg, R. W., Deformation and Fracture Mechanics of Engineering Materials, Vol. 89, Wiley, United States, 1996. [35] Papadopoulos, C. and Yeung, H., “Application of Probabilistic Uncertainty Methods (Monte-Carlo Simulation) in Flow Measurement Uncertainty Estimation.” 2001.