n . Ht“. »nnu. mafl$n.:. x v.9. Nauru-I. . intiupn! z. u‘ . are: themahfikub H.513 . E. nu 5% 1.x»... .3 h.bv...uw. 2 . - 4.1mm .3zc..«.q u .p. ...fiymn.unw.wmvmfi$_nd.fiund .. v DO..WQOa|r!|c ”Hwy“. I ’IE.‘I . ‘ . ;. 43“..-.ufifimwww. . . unurflnnwu..hnmv.H..o.1:vn. .. . fl.fiv:.vh.fihfl. . :0 v1 aft av- . n .. .‘ . o a... tum-s . - t.~.m.m.m=.fi..rw.m~.m»%l. Hahn. mun...‘ figmfikfiuxflfimuwwwfimm‘ufin an... . u... . .- f: . ho)“. ....H.~..h.l $0.1!” 3.5!..w v.1.- HIM.» OI I‘vbni.nu... . r. w‘g..h.....\t13.. . . .11 «Dun. . at... v. «5.-.: . ,i. . .Rflfiuflpun. .t. thud... . . . v‘ luv ovfl..c-v0-s I.D~ntl I-QIIOI .0. Ifiku‘ J. v4 :3. gummfluiflvrfi 3..- n... .._.!...yuu.u.l.h.-. 2! 1 3. 4...»: «'34.. u . . o. .l .u' .1flmu1n......v. vivid. . . :tflWMWq $943.?!) naming? q . . .a ..... ... 1.95.... I a. . .v. . ....J..nuon.n1-1hvlu~nyw. , v . luv. n)‘.-I . .v...JPVVP‘NDOON’JOIIH‘VJ-Vounow 'IAI ‘t‘lj "“..I .6... 11:7: if .. vi... 1.] x}...l\.! . . Lo. when“... - ..r. 1 “Vi . . .. W. t... ‘lI\-.v Wall pilot-“WV all“... . o . .u Kind. , 'llnu. [‘5‘ r \l IL.:a.L‘ . 3.“ -...umfluu.nbjuonc I 'I‘nllIll lirl‘l!'{’t'{q‘tfi v oao..n:xuvo.luv~lv . . I'Vt .. ‘ - vn‘tnb.!tl.l 1.43.1. ll .04. u . vs . $000.0. .yrnt v ORCJ‘VVM ‘- 14'); 65””. A... ll! n .n‘ :3‘.‘ its: .- nip . . n .n c it: v . lllllll 95": ulrbb |§ II no 1. Ctla: .. . I‘vtvlovuk‘hu . .ofi .5 1.11100. 7.0-1 V: a: Q 31.53 h tviusolnlJ' . . ‘Oul A .315... $3 51‘. . . Ca ! 0‘14. .11.! I .1“ uvna. unatttfi. ' I. 'i A . 1|, ......... I :21 w? it, ‘I VOSIIVOOio gilt! . , ‘ .. ;. ,~ :11 o .hohbkvatohvvoll. ‘ Intoltla tvtlnIOlE ll .lutb‘lv‘brhlu .0 o c 90 llv v. .. . . RWIHI‘HflrKI..cK.Ho. so. . fir. u.l;.........71p.h,fifinhm.a 1 . 1Q..ur....vinluou u ““3? in. .... he? {Hula . . . ., . :tktJL.i..WuTt .;.nlu...ronu....m.n.(o.1 .u .. .. .. Amman“. . . . . h v T - cm!- I: haw}. . 3r .- I '59: t -‘,;{I,- . N 1% £13. . 'HJ'O: , H: . ! Haw. nu! 3k.) ' 4 'O‘ '55? ‘ I v.1...flltl I. ‘0‘!»L'bvo. ‘natl‘.0fi+c0~vy ‘0 ‘r .c .Q‘ . 4“. pin u... bolcov1v 0.. . 3;}. ‘ ; .11"..;':3,g.,f'lé£. an, . 5!!!! w :xyz'Ll, ”7 ' 31‘5§!E%;1k!nif't§§;3 =. 1 ' I .; i ‘. .17- é?! ‘ .! ' "~x-‘ti .54 4,0.- 'l 9%. U' 5 X ' l,;‘ L’, “'4‘.” 1' ‘ ’ mmfl. . THESlS Z Illllllllllllllllllllllllllllllllllllllllllllllllllllllll 3 1293 0169434 This is to certify that the thesis entitled In-Situ Sem Study of Fatigue Crack Initiation in KaowoolTM/SaffilTM Discontinuous Fiber Reinforced Al-Si Alloy Matrix Composites presented by Prasad Alavilli has been accepted towards fulfillment of the requirements for ' . i ien e ' Master 3 degree 1n Mater als Sc c l %Z /@ Major professor DateW/JJJ Z/gyvj 0-7639 MS U is an Affirmative Action/Equai Opportunity Institution LIBRARY Mlchlgan State Unlverslty PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE 1m CDMpGS-p.“ IN-SITU SEM STUDY OF FATIGUE CRACK INITIATION IN KAOWOOLTM/SAFFILTM DISCONTINUOUS FIBER REINFORCED Al-Si ALLOY MATRIX COMPOSITES By Prasad Alavilli Thesis Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE IN ENGINEERING Department of Materials Science and Mechanics Michigan State University East Lansing, MI 1997 ABSTRACT IN-SITU SEM STUDY OF FATIGUE CRACK INITIATION IN KAOWOOLTM/SAFFILTM DISCONTINUOUS FIBER REINFORCED Al-Si ALLOY MATRIX COMPOSITES By Prasad Alavilli Specimens of Al-Si alloys reinforced with discontinuous KaowoolTM (50% A1203- 50% SiOz) and SaffilTM (97% A1203-3% Si02) fibers were fatigued under strain- controlled repeated bending loading conditions, using an in-situ SEM fatigue stage to identify the fatigue crack initiation site(s). From the series of pictures recorded from particular locations on the surface of the specimen, at various stages of the cyclic pro- cess, the crack initiation site(s) was identified. The results from the study of 15 vol.% KaowoolTM composites showed that crack initiation occured at processing defects called “shot” particles, very early in the cyclic process and at other microstructural features. Cracks from multiple sites were observed to grow and coalesce into a single major crack that led to specimen fracture after about 250,000 cycles. The premature cracking of these shot particles motivated further study of the behavior of these shot particles under tensile loading using an in-sz’tu SEM tensile testing apparatus. In the case of 15 vol.% SaffilTM composites, cracks initiated at fiber/ matrix interfaces of transversely oriented fibers as early as 5 cycles of loading. Fracture occured after only about 2500 cycles due to link up of several such cracks along the interfaces. The difference in behavior of the fiber/ matrix bond for the KaowoolTM and SaffilTM composites may be attributed to the differences in chemical reaction products at the interface due to different amounts of silica in the two fibers. ACKNOWLEDGEMENTS Dr. Martin Crimp, my graduate advisor at the department of Materials Science and Mechanics, Michigan State University, is not only a fine researcher himself but also a great inspiration and morale booster. With out his guidance and tremendous patience, this wouldn’t have been a success. I thank him very much for the same. I would like to thank my other advisors Dr. William J. Baxter and Dr. Anil Sachdev at General Motors Research Laboratories for their continual technical guidance and suggestions towards making this research focussed on real-life metallurgical problems. I have been truly challenged and motivated through out the course of this research. I would like to acknowledge help from a graduate student at MSU, Mr. Jung Kook Park, towards metallographic polishing and from Mr. Kurt Wong at General Motors, with EPMA chemical analysis of the composites. I would like to thank them for their time, effort and sincerity. Special thanks to my Parents, family & friends for their support and encouragement and also thanks to General Motors Research Laboratories and State of Michigan Research Excellence Fund for the financial support towards this research. Finally, I also appreciate the suggestions and comments from the graduate commit— tee members Dr. Ivona Jasuik and Dr. Subramanian at the Department of Materials Science and Mechanics, Michigan State University. iii TABLE OF CONTENTS iv LIST OF TABLES vi LIST OF FIGURES vii 1 Introduction 1 1.1 Background ................................ 1 1.2 Objective and Scope of Study ...................... 3 Review of Literature 6 2.1 Processing and Microstructure Development in Composites ...... 6 2.1.1 Matrix Material .......................... 7 2.1.2 Reinforcement Material ...................... 8 2.1.3 Various Processing Techniques/ Technique Used ........ 10 2.1.4 Fiber/ Matrix Interface ...................... 15 2.1.5 Heat Ti'eatment .......................... 17 2.1.6 Microstructure Development ................... 18 2.2 Deformation and Damage under Monotonic Tensile Stressing ..... 21 2.3 Cyclic Stress-Strain Behavior ...................... 25 2.3.1 Stress Control vs. Strain Control for Fatigue Testing ..... 32 2.3.2 Theories on Fatigue Crack Initiation .............. 36 2.3.3 Crack Initiation / Damage Examination Under Cyclic Loading 42 2.4 Design of Fatigue Testing Machines ................... 45 2.5 Design of Fatigue Specimen ....................... 49 2.5.1 Specimen Size Effect ....................... 49 Experimental Procedure 51 3.1 Design of Fatigue Stage, Tensile Stage and Test Specimens ...... 51 3.2 Finite Element Stress Analysis ...................... 54 3.3 Calibration of Strain in Fatigue Stage .................. 57 3.4 Metallography ............................... 57 3.4.1 Polishing .............................. 59 3.4.2 Optical Microscopy and Image Analysis ............. 59 3.5 Chemical Analysis ............................ 59 3.6 Description of the Experimental Setup Used .............. 61 3.7 Description of the Testing Procedure .................. 63 3.7.1 Procedure for In-Situ Fatigue Testing .............. 63 3.7.2 Procedure for In-Sz'tu Tensile Testing .............. 65 4 Experimental Results & Discussion 67 4.1 Optical Microscopy and Image Analysis ................. 67 4.2 Results from Electron Probe MicroAnalysis ............... 72 4.3 Observation of Fatigue Crack Initiation Site/ Damage ......... 80 4.3.1 Crack Initiation/ Damage in Unreinforced Material ....... 81 4.3.2 Crack Initiation/Damage in KaowoolTM Composites ..... 83 4.3.3 Crack Initiation/Damage in SaffilTM Composites ....... 96 4.4 Results from In-Sz'tu Tensile Testing of KaowoolTM Composites . . . 102 4.5 Results from Hacture Analysis ..................... 105 5 Conclusions and Recommendations for Future Work 111 A 115 A1 Composition and Mechanical Properties of Unreinforced Alloy and Composite Materials ........................... 115 A11 The composition of Al-Si matrix alloy (A339): ......... 115 A.1.2 Material properties: ........................ 115 B 116 B.1 Finite Element Analysis to Demonstrate the Stress / Strain Distribution on the Surface of the Specimen ..................... 116 B.1.l Out-line of the Finite Element Analysis Work: ........ 116 C 120 CI Observations from the Image Analysis of Composite Materials . . . . 120 C.1.1 Out-line of the Analysis: ..................... 120 C.1.2 Data/output from Image Analysis: ............... 121 BIBLIOGRAPHY 122 LIST OF TABLES 1.1 Key material properties for automotive engine components (Dinwoodie [10]) ..................................... 4 2.1 Typical physical properties of fibers used in automotive industry (B. Peters [2].) ................................. 10 4.1 Table comparing the observations from KaowoolTM and SaffilTM fa- tigue testing. ............................... 101 A.1 Mechanical properties of the matrix alloy and composites (Reference: General Motors Research Laboratories) ................. 115 vi 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 3.1 3.2 3.3 3.4 3.5 LIST OF FIGURES Schematic of squeeze casting process (B. Peters [2].) .......... Phase diagram of Al-Si alloys (Metals Handbook [6]). ......... Schematic illustrating the basics of a hysteresis loop (B. Sandor [40].) a) Fatigue strain amplitude vs. Number of cycles to failure curves for ductile, strong and tough materials, b) Stress-Strain response of a ductile, strong and tough material (B. Sandor [40].) .......... A schematic illustrating the necessity of strength and ductility segre- gation in a composite model (B. Sandor [40]). ............. Schematic comparing the stress-strain responses of a discontinuously reinforced composite and an unreinforced alloy a) at high stresses, b) at intermediate stresses (Allison & Bonnen [56]) ............. Cycle dependent material responses under a) stress-control, b) strain- control (B. Sandor [40]). ......................... Stress vs. Strain response indicating cyclic hardening under a) stress- control, b) strain-control (B. Sandor [40]). ............... Stress vs. Strain response indicating cyclic softening under a) stress- control, b) strain-control (B. Sandor [40]). ............... a) Formation of long slip lines from short slip lines by cross-slip, b) Local softening by cross-slip during cyclic loading (N.F. Mott [47]). a) Fatigue crack initiation model based on Stage I type cracking, b) Fatigue crack initiation model based on void nucleation (Masuda & Tanaka [50]). ............................... Various types of repeated bending fatigue testing machines (Manual [59]) ..................................... Fatigue stage for in-sz’tu examination ................... Geometry of fatigue specimen ....................... Geometry of tensile specimen. ...................... Deflection vs. Strain plot from fatigue stage calibration ......... Procedure for metallographic polishing .................. vii 14 20 26 28 30 31 33 34 35 39 48 53 55 56 58 60 3.6 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 Experimental setup for fatigue and tensile testing ............ 62 a) Optical micrograph showing the microstructures at the interface between an unreinforced alloy (A) and a SaffilTM reinforced compos- ite (B), b) Optical micrograph showing the microstructure of a 15% KaowoolTM reinforced composite. .................... 69 Histogram showing the orientation of the fibers on the surface of the composite specimen ............................ 71 SEM micrographs taken from the surface of a KaowoolTM specimen showing a) a shrinkage hole, b) a hollow shot particle, and c) a solid shot particle. ............................... 73 a) Backseattered electron image, and b) Elemental Al distribution map, from the same area on the surface of a KaowoolTM composite specimen. Features marked A, B & C contain FeNi, CuNi and MgFeNi, respectively. 75 a) Elemental Si distribution map, b) Elemental Mg distribution map, from the same area on the surface of a KaowoolTM composite specimen as in Figure 4.4a. ............................. 76 a) Elemental Cu distribution map, b) Elemental Fe distribution map from the same area shown in Figure 4.4a ................. 77 a) Elemental Ni distribution map, b) Elemental 0 distribution map, from the same area shown in Figure 4.4a ................. 78 a) Backscattered electron image, b) Elemental Si distribution map, from the same area on the surface of a SaffilTM composite specimen. . 79 a) Optical micrograph showing several fatigue cracks in a narrow zone on the surface of an unreinforced alloy specimen after 250,000 cycles of loading, b) Optical micrograph showing fatigue crack initiation along primary aluminum dendrites (B) and at intermetallics (A) in unre- inforced alloy specimen. Stress axis is indicated by a double headed arrow. ................................... 82 4.10 Optical micrographs showing an area on the surface of a KaowoolTM 4.11 composite, a) before cyclic loading, and b) after fatigue fracture. . . . 84 a) Secondary electron SEM image showing extensive crack growth af- ter 25,000 cycles of loading on the surface of a 15 vol.% KaowoolTM composite, b) Back scattered electron SEM micrograph from the same area at 5cycles of loading, and c) Secondary electron image from the same area at 0 cycles of loading. Stress axis is indicated by a double headed arrow. ............................... 85 viii 4.12 4.13 4.14 4.15 4.16 4.17 4.18 4.19 4.20 SEM micrographs showing crack initiation and propagation from a hollow shot particle in a 15 vol.% KaowoolTM composite, a) after 10,000 cycles, b) after 2500 cycles, and c) after 5 cycles of loading. Stress axis is indicated by a double headed arrow. ................. Higher magnification view of the area adjacent to hollow shot particle showed in Figures 4.12, a) Area on the left side of the shot, b)Area on the right side of the shot .......................... SEM micrographs showing crack initiation near a solid shot particle in KaowoolTM composite, a) Secondary electron image, b) Back scattered electron image. Stress axis is indicated by a double headed arrow. . . SEM micrographs showing crack initiation and propagation from a shrinkage hole in KaowoolTM composite, a) after 25,000 cycles, b) after 500 cycles, and c) after 5 cycles of loading. Stress axis is indicated by a double headed arrow ........................... SEMmicrographs showing fatigue crack initiation from fiber/intermetallic interface in lTM composite a) after 25,000 cycles, b) after 500 cycles, and c) after 5 cycles of loading. Stress axis is indicated by a double headed Kaowoo arrow. ................................... SEM micrographs showing fatigue crack initiation from several fiber/intermetallic interfaces in KaowoolTM composites a) after 25,000 cycles, b) after 500 cycles, and c) after 5 cycles of loading. Stress axis is indicated by a double headed arrow. ................. Schematic showing the bahavior of a fatigue crack on approaching a bimaterial interface. A crack from harder side to softer side would penetrate the interface, while a crack from softer side to harder side would be deflected away from the interface. Cracks are indicated by broken arrows ................................ Schematic showing the orientation of the fiber with the stress axis. A crack approaching the fiber at angles less than 40-50” would penetrate the fiber, while a crack approaching the fiber at an angle greater than 40-50" would seek the fiber/ matrix interface. .............. a)&b) SEM micrographs showing debonding at fiber/ matrix interfaces in SaffilTM composites after 5 cycles of loading. Stress axis is indicated by a double headed arrow. ........................ 87 88 89 91 92 93 95 97 98 4.21 4.22 4.23 4.24 4.25 4.26 B.1 a)&b) SEM micrographs showing debonding at fiber/ matrix interfaces in SaflilTM composites after 5 cycles of loading at a higher magnifica- tion.A thin layer of unknown interfacial product is expected to have caused this abnormal debonding. .................... SEM micrographs showing crack initiation at a hollow shot particle in a KaowoolTM composite under tensile loading conditions, at stress levels of a) 15 ksi, b) 29 ksi, and c) 35 ksi. Stress axis is indicated by a double headed arrow ........................... SEM micrographs showing crack initiation at a hollow shot particle in a KaowoolTM composite under tensile loading conditions, at stress levels of a) 15 ksi, b) 23 ksi, c) 29 ksi, and d) 34 ksi. Stress axis is indicated by a double headed arrow. .................. a) SEM micrograph showing shot particle on the specimen surface near fracture region of a KaowoolTM composite specimen, b) SEM image showing secondary cracks on the fracture surface originating from shot particles inside the KaowoolTM specimen ................. a)&b) SEM micrographs showing bare fibers exposed on the surface of a SaffilTM composite near fatigue fracture. ............... Comparison of the fracture surfaces of a) 15 vol.% KaowoolTM, and b) 15 vol.% SaffilTM composites from fatigue testing ............ Finite element meshing of the specimen model ............. 99 103 104 107 108 109 117 B2 Boundary conditions applied to simulate the conditions in fatigue stage 118 B3 Strain distribution on the surface of the specimen ........... 119 CHAPTER 1 Introduction 1 .1 Background The quest for lighter, stronger materials for automotive and aerospace applications has finally led us to a new world of materials called metal matrix composites (MMC’s), where the physical and mechanical pr0perties of the composite material are a blend of those provided by the reinforcement and the matrix. The reinforcement, usually with higher strength and modulus than the matrix, is often a brittle second phase either in the form of continuous fibers or as discontinuous reinforcement such as short fibers, whiskers, nodules or particles. The use of continuous fibers offers advantages in applications where highly direc- tional properties are desired. In contrast, composites with randomly oriented rein- forcements are more or less isotropic and are generally used to produce the highest strength discontinuous reinforced composites. MMC’s, comprised of discontinuous high strength ceramic fiber reinforcement in a matrix of aluminum, are being devel- oped for high performance automotive applications due to their light weight, lower thermal expansion coefficient, high strength, stiffness, wear resistance and more im- portantly, superior high temperature performance [1]. These materials are currently showing tremendous promise for automotive design applications. The increasing use of low cost ceramic fibers and a metal fabrication technique called squeeze casting bring a new hope for the possibility of making automotive components economically from MMC’s. Briefly looking back at the historical development of these ceramic fiber reinforced composite materials, the first application came in 1983 in a diesel piston by Toyota. It replaced the conventional Ni-Resist cast iron ring that has been used to reduce wear of the upper piston ring groove [2]. This technology soon expanded to replace conven- tional materials in other components such as connecting rods, piston pins and rocker arms in the engine. In the 90’s, Toyota continued to lead the way towards more ap- plications by replacing the cast iron hub of a crankshaft damper pulley with an MMC to reduce weight and engine vibration. Engineers at Honda have successfully devel- oped an aluminum engine block with an integrally cast aluminum matrix composite liner to replace the existing die-cast aluminum blocks with cast iron linings. This change has effectively solved the problems of poor abrasion resistance and warpage at elevated temperatures that was experienced with aluminum blocks [3]. Table 1.1 shows the key material properties for various engine components [4] that challenged the conventional materials and ultimately paved the way for MMC’s. Many of the above applications still need to be commercialized. Nevertheless, research is being currently performed at a rapid pace by the auto industry in US and elsewhere to expand the application areas of these composites and to reduce the fabrication cost and enhance their reliability assessment, which have been the major disadvantages hindering their growth. While designing a structure or component, tensile strength is the most widely used measure and is of primary importance in many applications. However, the tensile strength, although it provides vital information, is most often not a true representation of a material’s capability. The stress-strain response of a material cyclically loaded either under controlled stress or strain condition (fatigue) would be different from the tensile stress-strain response. When a material is cyclically loaded, cyclic hardening or softening takes place, shifting the stress-strain curve higher or lower than the monotonic stress-strain curve. Moreover, it is a very well-known fact that while the presence of extrinsic defects in a material may have little effect on it’s tensile properties, these defects often reduce the fatigue life considerably. The yield strength from the tensile test might appear safe to specify a critical design stress, but not adequate considering the fact that the material might be responding to the cyclic stress-strain end. 1.2 Objective and Scope of Study Metal matrix composites are currently limited from critical safety applications (ap- plications where the material failure could endanger the driver or occupants) in au- tomobiles because their cyclic stress-strain behaviors are not well understood. The fatigue life curve for composite materials, as obtained by mechanical testing, is a strong function of many factors, including the presence of small defects, inclusions and inhomogeneities, which act as stress concentrators. This fatigue life curve encom- passes three different stages, namely the crack initiation stage, the crack propagation stage and the final fracture. It is a simple minded observation that the fatigue life can be extended by controlled delay of the initiation process or by slowing down the propagation. Since a crack must nucleate before it can propagate, an ideal approach to improve fatigue performance would be based on understanding the factors which are responsible for crack initiation. However, the overall fatigue performance can be improved only if both the aspects of initiation and prOpagation of the crack are thor- oughly understood. This improvement in performance under cyclic loading should not however sacrifice high performance under monotonic loading conditions. Hence, .. .. 8“ 83 25> * * I. coo 33> 6:: .. .. .. .. as as: $330 * I. * owH Em :Smi .. ... .. 822 as $5880 mwom Em .. .. .. .. .. 83mm 385 was 5520 Semi 33:26:00 ooqgflmom commammxm 388:8 Bantam wmoabSm fimfizam 12525. use?» no .mooO fiwaobm mcmvnom ESQ. onwEwm .QEoH.an BozanoO do: «Boogie munoaomfioo 2:ch «$505335. 8m moEomoE 3:82: max H HA 2an in order to design a better composite material, the dominant mechanisms leading to failure under cyclic as well as monotonic loading should be understood in relation to the microstructure. Based on this understanding, the microstructure can be tailored during processing to improve the overall mechanical performance. The objective of this study is to investigate the micromechanisms of crack initia- tion in two composite systems based on Al-Si alloy, one reinforced with discontinuous SaffilTM fibers and the other with discontinuous KaowoolTM fibers, under strain con- trolled cyclic loading at room temperature conditions. Monotonic tensile tests are also performed to identify the stress levels at which crack initiation takes place at specific microstructural features on the surface of the specimen. These tests are conducted with the intention of supporting the on-going effort towards applying these materi- als in automobile engine blocks at General Motors Research Laboratories (GM). A fatigue stage, which has been specially designed and built for this purpose at GM, has been utilized for this study. The design and implementation of this stage has limited the scope of this study to understanding the micromechanisms of failure only, with no effort at gathering data from mechanical testing. These studies have led to identification of the critical crack initiation sites in the composites processed under varying conditions. Accordingly, assessment of the microstructures in terms of cyclic loading behavior has been made. CHAPTER 2 Review of Literature Literature available from work that has already been carried out by experimentalists around the world would be a good starting point for any research. The understanding of which, if not contradicting, would save time and effort due to avoidance of repeat work. The issues relevant to the understanding of this research work are presented below with importance stressed equally among Processing and Microstructure Devel- opment (Section2.1), Fatigue Crack Initiation and Damage Mechanisms (Section2.3) and Design of Test Equipment & Testing process (Section 2.4 & 2.5). 2.1 Processing and Microstructure Development in Composites The processing technique employed for manufacturing a composite influences the microstructure that is developed in the material, which in turn dictates the compos- ite’s mechanical prOperties. In order to achieve important mechanical properties in a particular material, it is necessary to understand the correlation between the mi- crostructure and processing. Depending on the required mechanical performance, a processing technique can be evaluated for its applicability. 2.1.1 Matrix Material The matrix material employed in the present study is an Al-Si alloy, commercially the most widely used castable aluminum alloy and a very useful automotive material. The principal use of these alloys is in the pistons and cylinder blocks of automobiles where elevated temperature properties are a criterion. Aluminum, apart from being a light weight metal, has several advantages such as low melting temperatures, negligible solubility for all gases except hydrogen, good surface finish and more importantly, high strength to weight ratios in certain alloys because of its ability to age or precipitation harden. The addition of silicon imparts high fluidity due to the formation of a A1- 12.7 wt.%Si eutectic, reduces the coefficient of thermal expansion and also lowers the surface tension of the molten alloy by 0.66N/m/wt.% of silicon added [5]. The Al-Si alloy used in this study is designated as A339 with several additions in- cluding copper, magnesium, iron, nickel and manganese similar to the composition of the more famous automotive alloy A336. Alloy A339 is cheaper but performs similar to A336 at room temperature. However, it does not have the superior high temper- ature performance of A336. The exact composition of A339 is given in Appendix A. The presence of different elements in small quantities in the Al-Si matrix alloy improves either the pre-processing or post-processing characteristics [6] as described below. Addition of copper substantially increases the strength and hardness in the as- cast and heat-treated conditions, but decreases the castability, ductility, corrosion resistance and hot tear-resistance of the alloy. Magnesium is the basis for strength and hardness development in the heat-treated Al-Si alloys. It combines with silicon and forms a hardening phase, Mggsi, which shows a solubility limit corresponding to approximately 0.70 wt.% Mg, beyond which either no further strengthening occurs or matrix softening takes place. Addition of magnesium also gives a bright surface finish to the material. Hot tear-resistance along with elevated temperature strength can be improved by addition of iron to the alloy. Increase in strength is due to the several insoluble phases that form in the alloy like FeA13, FeMnA16 and aAlFeSi. However, the formation of these phases drastically reduces the flowability and feeding characteristics of the casting. Nickel usually enhances the elevated temperature properties while reducing the coefficient of thermal expansion. Manganese, although considered an impurity in the casting process, has been observed to be beneficial in influencing the internal casting soundness and also the response of the alloy to chemical finishing and anodizing. 2.1.2 Reinforcement Material The reinforcing material in a composite is usually harder than the matrix material, is often assumed to wet and form a good bond with the matrix, and should reinforce at low cost. Fiber type, volume and degree of fiber orientation are dictated by the prop- erties or functional requirements of the MMC component. Some of these properties [2] include wear resistance, frictional characteristics, seizure resistance, yield strength, tensile strength, fatigue strength, stiffness, dimensional stability and thermal resis- tance. Various fibers have been tested for efficient reinforcement of aluminum alloys. Extensive literature is available on reinforcements carbon fibers, SiC and derivatives of alumina-silicate fibers. The main application to date of alumina-silicate fibers, in the amorphous form, is in the thermal insulation of high temperature furnaces. Two varieties FibermaxTM, an amorphous fiber with a typical composition of 49% alumina and 51% silica and FiberfraxTM, a polycrystalline fiber with 72% alumina and 28% silica, chemi- cally known as mullite were primarily developed as reinforcing materials. However, alumina-silicate fibers, particularly in the amorphous phase, are structurally unstable [7], which can either lead to poor wetting of the reinforcing fibers with the molten matrix alloy or if well wetted, to deterioration of the fibers due to chemical reaction at the interface. The principle behind crystallization is the formation of mullite (3A1203.28102) crystals embedded in amorphous alumina-silicate upon heating the amorphous fibers above 1253K (980°C). During this process, Si“ is replaced by Al3+ in each alumina- silicate unit cell. This can be represented by the chemical reaction: A1325116030 - 4874+ + 4Al3+ — 202— = Al355’11207g. (2.1) Alumina-silicate fibers in crystalline form are currently being developed as reinforce- ments in MMCs. The relatively low cost of these fibers and their availability in large quantities has attracted the attention of material design engineers even though their strength is not as high as SiC and other reinforcements. Table 2.1 compares the properties and cost of fibers tested [2] for automotive applications. The fibers that are used in present study are based on alumina-silicate, with vary- ing amounts of cristobalite (8102) and mullite (3A1203.28102), in crystalline form. Their trade names are KaowoolTM and SaffilTM . KaowoolTM is made up of 50 wt.% alumina (6 phase) and 50 wt.% silica and SaffilTM contains 97 wt.% alumina (6 phase) and 3 wt.% silica. The procedure adopted for fiber processing is described by Canumalla [8] and briefly explained here. The fibers are processed by first blowing the molten alumina-silicate through an orifice, against a stream of hot air. The im— properly fiberized or unfiberized alumina-silicate is retained as particulate inclusions, also called ’shot’, with varying sizes. These shot particles can be solid or hollow depending on whether air is entrapped within the fibrous material during the blow- ing process. Finally, the fibers are chopped to the required length and filtered to separate them from the shot. The presence of 8102 aids in the development of fine grained microstructure in the fiber. It also controls the transformation of 6 alumina 10 into a alumina, facilitating the removal of porosity and by acting as a crystal growth inhibitor once the transformation to a alumina occurs [9]. The transformation of 6 alumina to a alumina is accompanied by an increase in hardness, modulus and den- sity and a decrease in strength due to step increase in crystal size [10]. The properties of the fiber can be tailored by controlling the concentration of oz alumina in the fiber. Table 2.1 : Typical physical properties of fibers used in automotive industry (B. Pe- ters [2].) Property FiberfraxTM FibermaxT Carbon fibers SiC whiskers Tensile strength, Ksi 250 120 550 1200 Young’s Modulus, Msi 15 22 33 80 Density, lb/in3 0.10 0.11 0.06 0.11 Fiber Diameter, mils 0.10 0.12 0.28 0.01 Fiber Length, mils 30 30 continuous 4.9 Price, $/lb 1.00 16.00 21.00 300 2.1.3 Various Processing Techniques/ Technique Used The basic requirement for manufacturing a sound ceramic reinforced MMC is to de- velop a strong bond between the reinforcement and the matrix material. Improved wettability and hence bonding at the matrix/ reinforcement interface can be achieved in some systems by a chemical reaction that forms spinels or oxides isostructural with spinels (MgAle4) [11]. It has been reported that the wetting properties of ceramics by liquid metals are governed by a number of variables, including heat of forma- tion, stoichiometry, valence electron concentration in the ceramic phase, interfacial chemical reactions, temperature, and contact time [12]. This problem in developing a strong interface and also a uniform dispersion of reinforcement through out the matrix can be approached by either a powder processing route or one of the widely 11 applied foundry techniques. Advancements in powder metallurgy technology leading to production of fine, rapidly solidified metal powders is paving the way to a variety of new alloying pos- sibilities which can result in microstructural refinements and mechanical property improvements compared with the traditional cast products. However, the extent to which powder metallurgy will be applied to composite processing will depend on how well the costs can be brought down. Composites made from powder metallurgy tech- niques demonstrate properties equivalent to that of the cast material, while extruded powder metallurgy components show significantly enhanced mechanical properties, especially in the axial direction, and strong bonding between fiber and matrix [10]. However, advancements till this date can only use fibers of small length (approx. 10pm) and a major technical target for future would be to increase the aspect ratio without sacrificing alignment after extrusion processing. A detailed description of important foundry techniques for composite fabrication can be found in Metals Hand book [6]. However, for the sake of completeness, impor- tant aspects of these techniques will be presented in this section with emphasis on the squeeze casting (liquid forging) technique, which has been used for the manufacture of the composites under study. Mixing techniques that are generally used for introducing and dispersing a discon- tinuous phase homogeneously in a melt, use external force to transfer a nonwettable ceramic phase into a melt and create a homogeneous suspension of the ceramic in the melt. Some of these techniques are: 1. Addition of the discontinuous phase to a vigorously agitated fully or partially molten alloy [13]. 2. Injection of the discontinuous phase into the melt with an injection gun [14]. 3. Addition of powders to an electromagnetically stirred melt. l2 4. Centrifugal dispersion of the discontinuous phase in the melt. The melt-dispersoid slurry can then be cast either by conventional foundry tech- niques, such as gravity, pressure die, centrifugal casting, or by novel techniques such as squeeze casting, spray codeposition, melt spinning or laser melt-dispersoid injec- tion [6]. The distribution of the discontinuous ceramic phase in the cast structure will be different for each of these processes and will ultimately influence the mechanical properties of the composites. The important variables among all the techniques (sand casting, die casting, centrifugal casting, compocasting and pressure die casting) used to solidify the melt-dispersoid slurries are [6]: o cooling rate; 0 viscosity of the solidifying melt; o shape, size, and volume fraction of dispersoid; o reinforcement and melt specific gravities; 0 thermal properties of reinforcement and the matrix alloy; 0 chemistry and morphology of the crystallized phases and their interactions with the dispersoid, nucleation of primary phases on ceramics, entrapment or pushing of dispersoid by solidifying interfaces; flocculation (clustering) of the dispersoid. The higher solidification rates obtained by changing from simple sand casting to die casting will generally give rise to homogeneous distribution of the reinforcement in the cast matrix [15]. In compocasting, discontinuous fibers are incorporated into vigorously agitated partially solidified alloy slurry. The discontinuous phase is me- chanically entrapped between the proeutectic phase present in the alloy slurry, which 13 is held between its liquidus and solidus temperatures [11]. Apparently, a higher pr0portion of the proeutectic solid phase in the hypo-eutectic alloy slurries aids the separation and dispersion of the discontinuous reinforcement. This technique leads to significant fiber breakage due to the grinding effect of the stirrer and the partially solid alloy slurry if the process is not carefully controlled. With pressure die-casting techniques, large and intricate component shapes can be produced rapidly at relatively low pressures (approx. 15 Mpa). The application of pressure puts fewer restrictions on the choice of the matrix and reinforcement due to possible infiltration with out altering their chemistry. It also increases the heat transfer from the melt to the die by lowering the contact resistance between them, for faster solidification and refined grain size. It has been reported that high pressures, short infiltration paths, and columnar solidification toward the gate produce void free composite castings [6]. Squeeze casting is a recent invention which involves unidirectional pressure in- filtration of fiber preforms in order to produce void free, near-net shape castings of composites. Some of the processing variables governing the evolution of microstruc- tures in squeeze cast MMC’s are [16]: 0 fiber preheat temperature; interfiber spacing; infiltration temperature; infiltration speed; metal-superheat temperature. A schematic of the squeeze casting process is shown in Figure 2.1. In this process, preheated particles or fiber preforms are inserted into a metal die and infiltrated with molten metal under high pressure (70 to 200 MPa) followed by solidification under 14 SN— Boaom .5 $805 moans 381% no oSQEozom n fin ousmmh 15 pressure. Several papers have been published describing the squeeze casting process [17, 18]. The preforms are made by slurrying fibers of controlled length in water along with small amounts of inorganic binder and silica, and forming the required shape in a “paper making-type” process. The procedure can be used for controlling the lay- down which determines the fiber orientation in the composite. The preform density determines the volume fraction of the fibers. The silica binder provides the necessary handling strength and enables machining to tight tolerances. Among other properties, good compressive strength is necessary so that there is resistance to deformation during the high pressure casting process. Composites are made by placing the preforms into a die cavity and infiltrating with molten alloy by means of a ram assembly. The preforms, die, and ram are preheated before introduction of the melt at a known degree of super heat. The fiber preheat temperature should be controlled within limits. Too low a temperature will lead to porous castings while too high a temperature would lead to excessive metal/fiber reaction that degrades the casting properties. The ceramic preforms are usually poorly wetted by metallic alloys (contact angles less than 90°) and hence their infiltration requires considerable hydrostatic pressure to overcome the capillary pressure. Also, the application of high pressure increases the solidus temperature. Thus, solid forms at a higher temperature, which minimizes the contact time between the molten metal and fibers and reduces the time for interfacial reaction [19]. The final shape of the component produced by squeeze casting promotes a fine, equiaxed grain structure because of large undercooling and rapid heat extraction. 2.1.4 Fiber/ Matrix Interface The processing conditions employed for composite fabrication significantly alter the formation and composition of the reaction products at the fiber/matrix interface. The reaction products have been investigated by several researchers [7, 20, 21] and 16 it has been established that alumina is unstable in aluminum alloys containing Mg. Two possible chemical reactions that can occur at the interface are: and Al203 + 3M9 = 3MgO + 2A1 (2.3) Such interface reactions are undesirable for the following reasons: they make it difficult to control the matrix composition; they remove magnesium from the matrix and hence reduce the age hardening capability of the matrix; they increase the viscosity of the composite melt; they may degrade the fiber/ matrix interface strength. Investigation [7] aimed at understanding the interface in the case of amorphous alumina-silicate fiber reinforced aluminum alloys revealed the presence of A1203.MgO (spinel) and 2MgO.SiOz reaction products extending into the entire section of the fiber by diffusion mechanism. However, only 2MgO.Si02 has been detected in the reac- tion zone between the crystallized alumina-silicate short fibers and the matrix, with the reaction product limited to the surface of the fiber, as the more reactive 8102 rich region on the surface forms a strong bond with the magnesium in the matrix [7]. Mogilevsky et al.[21] reported that there was no evidence of reaction product at the interface between SaffilTM (96%A1203,4%Si02) or FiberfraxTM (45-55 wt.% mixture of alumina-silica) fibers and pure aluminum or aluminum with magnesium and silicon additions [22]. Reactions were limited to the binder, exterior to the fiber, which contains both magnesium and silicon. During T5 heat treatment (described in 17 Section 2.1.5), magnesium migrated to the fiber/ matrix interface due to the reaction between magnesium and silica binder, where Mg reduces the 8102 to form MgO and the reduced Si diffuses out of the binder to from the elemental Si phase in the ma- trix. The absence of a reaction product at the fiber/ matrix interface is typical for all processing techniques where short solid-liquid contact times are allowed, like in the squeeze casting technique [23]. 2.1.5 Heat Treatment The effect of processing depends on how it alters the local composition (segregation, inclusions etc.), alters the local microstructure (type, orientation, size) and alters or introduces self-stresses or residual stresses, which are known to have a serious effect on the fatigue resistance. A serious complication would arise if a processing technique develops a variation in microstructure through the specimen. Hence, it is impor- tant to deve10p Optimum processing conditions where the refined surface structure is maintained throughout the dimensions of the specimen. The effects of fabrica- tion, thermal and mechanical processing procedures can be physically represented by material effects and / or surface stress effects. Material effects include cold work (increased hardness), microstructure, composi- tion, hot work, and electrochemical alteration. When cold work (plastic flow) occurs locally in a gradient, it also influences the stress field by creating compressive or tensile residual stresses that are concentrated at the surface. These residual stresses can also be created by thermal gradients during heat treatment. Whatever the pro- cessing procedure, after the completion of processing, surface yielding may lead to a steep gradient of compressive residual stresses at the surface that are reacted by much lower elastic tensile stresses in the core. Because of this gradient, fatigue cracks may initiate at subsurface regions [24]. Residual stresses have been shown to have the equivalent effect of a mechanically applied mean stress, hence, it is possible to 18 simulate the residual stress effect by changing the applied mean stress. Also, these residual stresses tend to relax under the influence of alternating stress [25, 26]. Increase in fatigue strength from heat treatment of steel parts has been attributed to the increase in hardness or tensile strength and the creation of compressive residual stresses in the surface layer. Similarly, for Al alloys, the fatigue strength should increase if the heat treatment improves the matrix strength, once again, stressing the importance of the presence of Mg in the matrix, which increases the age hardening capability of these alloys. The two most commercially important heat treatments are T5 artificial aging heat treatment (heating the MMC for 11hr. at 200°C in air followed by air cooling) and a T6 heat treatment (involving a solutionizing treatment to about 500°C for few hours, a water or oil quench, and an artificial aging treatment). The former produces a small improvement in mechanical properties and also improves the dimensional stability of the matrix, while the latter gives peak strengthening of the aluminum alloy matrix [21]. Diesel engine pistons are often limited to T5 heat treatment because they contain a cast-in iron alloy insert in the ring groove area that may debond because of thermal stresses induced during the quenching cycle of T6 heat treatment. The high temperature solutionizing of the T6 treatment is intended to uniformly redistribute most of the alloying elements in the solution. These are then re—precipitated during artificial aging to strengthen the matrix. The main purpose of giving T5 treatment in the present study is to improve hardness for enhanced machinability and also to eliminate any permanent changes in dimensions from residual growth due to aging at operating temperatures. 2.1.6 Microstructure Development The phase diagram shown in Figure 2.2 illustrates the simple eutectic system for Al-Si alloy along with the cast microstructures of the pure components and alloys of 19 various compositions. The microstructures [6] typically contain primary aluminum along with the proeutectic a phase and the eutectic phase in the hypoeutectic alloys. Silicon contents ranging from 4% to the eutectic level of 12% permit the production of more intricate designs with greater variations in section thickness, and yield castings with higher surface and internal quality [6]. The required amounts of eutectic formers depends on the type of casting process used. The strength and ductility of these alloys, especially those with higher silicon, can be substantially improved by modifying the Al-Si eutectic with suitable elemental additions. The microstructural features that strongly effect the mechanical properties are: 0 grain size and shape; 0 dendrite-arm spacing; 0 size and distribution of second phase particles and inclusions. The finer the grain sizes, interdendritic spacings, and dispersion of inclusions, and second phase particles, the better will be the mechanical performance [6]. Large masses of oxides or intermetallic compounds produce excessive brittleness. While grain size and interdendritic spacing can be controlled by cooling and solidification rate control, the size and shape of microconstituents can be controlled by changing the composition. It is also possible to minimize the growth of microconstituents by using rapid cooling techniques. The control over size and distribution of primary and intermetallic phases is more complex and often use is made of modifiers and refiners, which act as heterogeneous grain nucleation sites, to influence eutectic and hypereutectic structures in these alloys. The most widely used grain refiners are master alloys of titanium, or titanium and boron, in aluminum. Addition of certain elements such as calcium, sodium, strontium, and antimony to hypoeutectic compositions suppresses the growth of silicon crystals within the eutec- tic, modifying the normally occuring eutectic structure into finer lamellar or fibrous 20 -. , 0". ._'.‘l H}... "" jit.’ ,“‘-I -63 A 8 8 \: . \fi ‘ \ 3 :1 m \ .. 3 ‘\\\»\”§’ a” \ ..,...+s~ 151M" ii ‘ § + i" 3 E 3 § 1. mutant»; 99.95% Al Figure 2.2 : Phase diagram of Al-Si alloys (Metals Handbook [6]). 21 eutectic network [6]. In the case of hypereutectic compositions, the elimination of large, coarse primary silicon crystals that are harmful during casting and machining, is the primary function of refinement. Addition of phosphorus affects the form and distribution of primary silicon phase. The addition of ceramic fibers or whiskers to alloys during the fabrication of com- posites not only strengthen the material through reinforcement, but also tends to modify or refine the microstructure [27]. The eutectic silicon in aluminum-silicon alloys is modified due to the presence of graphite particles and primary silicon is re- fined when solidification occurs in the presence of a high volume fraction of ceramic phase [15]. The microstructures of these MMCs suggest that primary aluminum, crystallizing from the melt in hypoeutectic alloys, tends to avoid discontinuous ce- ramic phases and nucleates in the interstices between particles or fibers unless special surface modification techniques are employed to promote heterogeneous nucleation on the fiber surface. Primary silicon and the eutectic in aluminum-silicon alloys tend to concentrate on particle or fiber surfaces [28]. 2.2 Deformation and Damage under Monotonic Tensile Stressing Experiments done on specimens under monotonic tension provide basic but valuable design data relating stress applied in tension to the elongation produced due to the applied stress. They also provide information relating the tensile stress to the amount of material deformation or damage accumulation. In-sz'tu observations of the surface damage under powerful optical microscopes or scanning electron microscopes during tensile testing, followed by fractography of the fractured specimens, reveal information that can be used for improving material design and performance. Investigations by 22 several researchers have assessed the evolution of damage under tensile loading in Al and Al-Si alloy composites with reinforcements such as SiC, alumina, alumina- silicate in the form of particulates or short fibers. The majority of these investigations performed under tensile loading did not study the damage accumulation prior to macroscopic crack formation, but instead concentrated on crack propagation, mainly due to the difficulty in locating the crack initiation site [29]. Initial investigations [30] on 6-alumina or SaflilTM short fiber reinforced alu- minum/ magnesium and aluminum/silicon alloys demonstrated good interfacial wet- ting and hence strong bonding between fiber and matrix alloy as a result of squeeze casting. This wetting was achieved with out the necessity of applying any coating to the fiber. No reaction layer was detected at the fiber/ matrix interface even after long periods of heat treatment at close to liquidus temperatures. The fiber rein- forcement enhanced the stiffness of the alloys and strength at elevated temperature. However, the room temperature strength was limited by the lack of ductility. Alloys that showed little or no improvement in strength at room temperature due to the addition of reinforcement usually exhibited increased strength at elevated temper- atures [30, 31]. The reason for this behavior has been attributed to the increased ductility of the matrix at higher temperatures. A matrix with high ductility has been shown to enhance the strength of the composite. It has been shown [31, 32] that short fibers oriented perpendicular to the stress direction play an important role in the generation of strength in the composites. The composite strength is maximum and increases with volume fraction when the bond strength is larger than the matrix strength. It is a minimum and weakly dependent or independent of volume fraction when the bond strength is weaker than the matrix. An attempt has been made by You at al. [33] to understand the failure mecha- nism in a 20% SiC particulate 2124 aluminum alloy composite made by a powder metallurgy technique. Tensile testing was conducted on the heat treated specimens 23 at a strain rate of 1x10‘2 / min. Examination of the fracture surfaces revealed ma- trix failure as the dominant mechanism leading to failure in these composites. The reinforcement contributed to the failure process primarily by imposing high levels of constraint on matrix deformation and by raising the stress in the matrix to a level significantly greater than that normally associated with matrix failure. Cracking of the reinforcement and decohering of the matrix/ reinforcement interface occured only as a result of matrix failure. Detailed analysis showed that cracked particles outnum- bered decohered particles by a ratio of 2:1. You et al.’s proposed failure mechanism anticipates failure to occur through the matrix, not specifically avoiding the reinforce— ment nor specifically linking them. They speculate that a reinforcement optimized for both strength and toughness along with a ductile matrix alloy would ensure the best mechanical properties of the composite. Manoharan et al.[34] studied in-situ in an SEM, the crack propagation in a 15 vol.% alumina particle reinforced 6061 aluminum alloy based composite under tensile deformation. They observed that micro cracking occured in a region of intense defor- mation surrounding the tip of the macrocrack. Crack growth occured by linking up of the macrocrack with these microcracks ahead of it when the distance between them is comparable to interparticle spacing. Thus, distribution of reinforcement seems to play an important role in the damage process. In the case of 12% A1203 short fiber/Al-Si alloy composites studied under tensile deformation by Jiang et al. [35], the composition of the matrix was close to alloy A336 and the composites were prepared by squeeze casting technique. It was concluded that these composites failed at room temperature essentially due to fiber/matrix debonding. This was not the case at elevated temperatures. Fibers parallel to the loading direction strengthen the composites while fibers that are perpendicular to the loading direction debonded easily and degraded the composite strength. Also, the plastic deformation was limited to a very small area near the fracture region. 24 Similarly, another study [36] concluded that the fracture origin was usually a bundle of fibers oriented perpendicular to the loading direction in specimens made of 15% SaffilTM reinforced ASI2UNG Al-Si alloy (Al—12.05%Si—1.24%Cu-0.98%Mg—1.05%Ni) composite produced by squeeze casting. The tensile deformation and fracture of 12.9% and 26.6% SaffilTM short fiber rein- forced Al-1% Cu matrix composites were studied by Anlas et al. [37]. The composites were fabricated by squeeze casting technique and demonstrated a strong fiber/ matrix bonding. They observed that fracture of the composites did not occur until after extensive fiber fragmentation occured. Fracture of several closely spaced fibers was seen to line up suggesting that fracture of one fiber led to the fracture of neighboring fibers, leaving the matrix in between continuous. Similar observations were reported by Clegg et al.[38]. Tensile studies performed on Al-Si SAE321 alloy hybrid reinforced with various volume fraction ratios of alumina-aluminosilicate short fibers made by squeeze casting showed that the tensile strength is higher than the unreinforced alloy at 300°C and is lower than the unreinforced alloy at room temperature [32]. A 3:2 fiber ratio of alumina-aluminosilicate showed optimum properties and it has been concluded that by prOper control, superior mechanical properties can be achieved in composites by replacing costlier alumina fibers with cheaper aluminosilicate fibers. Further insight and useful information was provided by TEM observations of the damage initiation due to tensile straining in a SiC whisker reinforced pure Al matrix composite [39]. Dislocations were seen to generate and pile-up at grain boundaries and at Sij-Al interfaces during the straining process. Microcracks initiated due to fracture of whiskers, debonding of Sij-Al interfaces, cracking of the matrix and at uninfiltrated zones that formed during the squeeze casting process. However, it has been observed that only the microcracks caused by matrix cracking can propagate and form the major crack which will lead to fracture of the specimen. 25 2.3 Cyclic Stress-Strain Behavior The base line stress-strain curves obtained under monotonic loading conditions can be used for a quality control check and also to deduce useful design parameters such as strength, ductility and toughness of the material. However, when the material under question is evaluated for an application where periodic reversal of stresses or strains occur, the actual response of the material under these conditions of cyclic loading should be analyzed. On the basis of stress and strain, a hysteresis loop best describes the behavior at a single location in the material during cyclic loading. Figure 2.3 shows one such generalized hysteresis loop [40]. A stabilized loop is generated if the same loop repeats for several cycles of loading. A cyclic stress-strain curve can be generated if the tip of these stabilized loops is joined for several specimens tested over a wide range of strain. In general, stability of a loop occurs after certain initial number of cycles of loading and until then there will be cycle dependent hardening or softening and the hysteresis loop gradually changes. The loop generated due to this softening or hardening behavior is strongly dependent on the type of control (stress or strain) used during testing. A complete cycle is defined when starting from any point on the loop and tracing the loop in a clockwise direction, the ending point coincides with the starting point. The tips A and B in Figure 2.3 represent the stress and strain limits of the cycle. The width of the loop between A and B represents the total strain which comprises of both the elastic strain and the plastic strain components. The width of the loop at its midsections, between points C and D represent the plastic strain. It is now well understood that fatigue occurs because of cyclic plastic deformation [41, 42] which causes irreversible changes in the material’s dislocation substructure. The type of irreversible change in the material’s substructure changes as the fatigue process progresses. Based on these changes, it is possible to divide the fatigue process 26 2o: 828m .5 goo. £88qu a Mo 8643 2: 9.5.9535 camp—ozom ” «zn Saw—m awd + mwd u moan: £83 .22 ... wd mococ wwmtm Hbq 27 into several stages as shown below 1. Cyclic hardening and / or softening, where the entire volume of the material may be affected by the change in the substructure. 2. Microcrack nucleation in the subsurface layer due to stress concentration effects at extrusions and intrusions. 3. Propagation of small cracks, of the order of the grain size of material. 4. Crack propagation resulting in final fracture. The rate of growth of such cracks is controlled by the magnitude of cyclic plastic deformation concentrated in the plastic zone at the crack tip. Consideration of all these stages results in the generation of fatigue-life curve. The stress amplitude vs. number of cycles to failure (S-N curve), elastic strain vs. number of cycles, plastic strain vs. number of cycles and total strain vs. number of cycles to failure, even though they are not independent of each other, would reveal unique information and are all common approaches to fatigue life analysis. For example, fatigue ductility properties can be obtained from the plastic strain component vs. life curve and the fatigue strength properties from the elastic strain component vs. life curve. The elastic strain component vs. life curve can also be readily converted to stress vs. life curve. Even though the total strain vs. life curve may be the most desirable information, it is often very difficult to obtain or measure because with increasing life, even the total strain becomes very small. In general, the relative fatigue behaviors of ductile, strong and tough materials are shown schematically in Figure 2.4a&b. From these figures, it can be deduced that a ductile material is best at resisting plastic deformation with out failure, while a strong material behaves quite elastically under a load and the tough material is idea] because it can deform plastically while sustaining high stresses. In a com- 28 :3. 865m .B 3:39: amzoa 6.8 mach—m 62.8.6 a .8 3.81%: :Bueméwueam 3 £383.: swag 6.8 moose .2526 e8 82.5 233.. 8 8—98 mo 82:52 .m> 6:23an 586 mammaam A.» “Ya wan—mum 3 3V .. - N Poems 2325 6:8 \ weak 9.2% INI 00— way \ swkx/ \ / 0:.an coach \ 9.25 ssang 29 posite structural material however, high ductility is required only at small regions of stress concentration, even though a combination of strength and ductility would raise the fracture toughness of the material [43] . This can be illustrated by considering the situation of a stress raiser (a reinforcement) in a composite material as shown schematically in Figure 2.5. The ideal situation here is to have strong and ductile regions segregated from each other. The strong regions carry most of the load, either static or cyclic with out much deformation, while the stress concentration at the rein- forcement is blunted by the plastic flow in the ductile region so that crack initiation and possible catastrOphic failure of the whole piece is prevented. At the same time, the plastic strains in the critical region are not excessive because the surrounding strong material acts as a constraint to deformation. Considering a composite material as a combination of ductile, strong and tough materials, a schematic comparison between the fatigue behaviors of composite and unreinforced material can be made (Figure 2.6). Usually the composite material experiences a lower total strain than the unreinforced alloy, both under high and low stresses, because of its higher modulus and higher work hardening rate. However, in some cases, the composite might experience a greater total plastic strain at low stresses, especially when the proportional limit of the composite is lower than that of the unreinforced alloy or when the stress-strain curve of the composite lies below that of the matrix alloy. In this situation, the fatigue performance of the composite would be close to the matrix alloy. The results obtained from fatigue testing vary depending on the technique that is employed, such as axial loading, repeated bending, rotating bending etc. and this difference in behavior has been explained due to difference in the stress gradient and the cross sectional area of the material experiencing the imposed stresses [44]. 30 ASE 26:3 .mv 362: 382.88 a E someewonwom 3:526 65 fimcmSm .o c3538: 2: 95833: 93828 < U ad ohm—mum .D »»»»+» nmcmmc £9.83 \/ \\\\\\©\\\\\\ T venom: 3.526 ~«.««« i0 31 écn— 550m Q 585$ 8883 35625an as 3 .8893 .3: 3 ? tho—E 38858:: as 6.8 Samoa—=8 BBSEE 32358586 a mo mom—8&8 583.883 2: mavens—8 oBannom ”ed 95w; 3V 3 3.28 .33 u q - u q q q q q d III-II L i 13...... \ \tll N /‘ I \ L \ ti \ 33133.3 \\ 3.038\ 1 i _lrt\ \ i m x a Q m x r c m t \ . \ \ i \ IUQ \ x xxx \ I \\ Esagu > . > > >-..>..,, ,, -..- ....... <-<-..., . u . o 32...... 632.2, Em 536$ - < < <-- 55522 .mwflcm-wc- -m_o>-o.- b -- -- -- -- w - £27.. 630:? 23539: \ --\// .8685; .8286. 228 - - . -- -- > --.<. .<... ..< <._ -2.<.>K>§ . -- -- .<- <2... .0 :65 60:83 62.2.:8 \/ \/l > > > row- 327.. 8:82 3:228 < < < < <7 ><><><><> 825 H 9N 0.53% 3 3 :E: IIII ® 383 mzmmmano - ®©© =8: 6 @® $83 238. 35 .23; Saga .5 3:28-5:22 3 ._o5:8-m8:m A: 31:: $28.8 :29? 35365 35:8: 553m .m> mmezm " ad mama 3V 3 @ 8‘6 % \- @ 2 _ @ a _ a b 36 The stress vs. number of cycles to failure or the stress-life approach can be used for getting rough estimates of the fatigue life, and this approach works very well for constant amplitude loading and long fatigue life situations, where plastic strains would be minimum. However, the stress-life method is completely empirical in nature as the true stress-true strain response of the material is ignored in favor of fictitious fully elastic strains and it does not distinguish between initiation and propagation. When plastic strains are significant, temperatures are high, amplitudes vary with time, and especially when small components are involved where initiation life is the primary concern, a strain-life approach is more appropriate. However, strain-life approach involves a more complicated level of analysis. True stress-true strain response has to be accounted for by measuring the notch root strains using Neuber analysis, finite element analysis or strain gage measurements. A critical analysis comparing and contrasting the stress-life and strain-life methods can be found in literature [45]. 2.3.2 Theories on Fatigue Crack Initiation The basic process of fatigue in single crystals or pure metals must be modified for alloys and composites because of the complicated deformations induced by metallur- gical changes in alloys and composites. These changes may lead to enhanced fatigue properties in these materials but do not essentially disqualify the basic mechanism itself. Many detailed and apparently conflicting observations have been reported on mechanisms that lead to fatigue crack nucleation in deforming materials. Any new theory proposed to explain the origin of fatigue cracks should be able to explain these experimental observations. The objective of this section is to summarize some of these important observations in order to aid in the understanding of the evolution of plastic damage and fatigue crack initiation in materials. To start with, the observation of fatigue in materials at liquid helium temperatures has ruled out thermal activation as one of the requirements for fatigue crack initiation. 37 Even though thermally activated processes like diffusion of vacancies or chemical processes such as attack by oxygen may play a role at high temperatures, they are not essential for fatigue damage accumulation. Also, since fatigue has been observed in single crystals, grain boundaries may modify the fatigue initiation and damage processes in polycrystalline materials, but are not necessary for fatigue crack initiation to take place. The room temperature behavior of a material under cyclic straining at constant amplitude shows three main stages [46, 47]. o a region of cycle dependent hardening. Fine slip and coarse slip may be observed in this stage. 0 a region in which no further hardening takes place. However, some slip lines broaden to become persistent slip bands and finally develop into cracks. Obser- vation of extrusions and intrusions in slip bands have been reported. 0 a stage in which one or more cracks that formed in stage two propagate from grain to grain. Each of the several theories [48] proposed a different model to explain how the slip bands broaden in the above stated second stage and turn into cracks. An excellent review on the various theories proposed to explain nucleation of fatigue cracks has been provided by Kennedy [49], which forms the basis for the following discussion. A theory on the formation of slip bands that has been accepted by many inves- tigators in the field states that, in polycrystalline face centered cubic metals and single crystals, as soon as the region of easy glide has passed, the first stage of slip is the formation of numerous short slip lines whose density and intensity depends on the substructure from which one starts. Each of these slip lines is thought to 38 originate in a Frank-Reed source and be terminated by a Cottrell-Lomer lock. Slip bands are due to cross-slip, which gets rid of the piled up dislocations at the ends of the lines, and so allows the sources (S, S', S" in Figure 2.10a) to generate more and more dislocations. By the same process, the slip lines broaden into bands in the later stages of deformation. Mott et al. [47] also assume that broadening takes place due to cross-slip. A schematic illustration of their explanation on how softening takes place due to cyclic slip is shown in Figure 2.10b. The continual backwards and forward movement of a dislocation on a long slip line AB past a neighboring short line CD will eventually lead to cross-slip and the consequent annihilation of the dislocation. This form of softening will allow S and other nearby sources to generate more dislocations. If the two slip planes are close enough together, a crack may be expected to form. According to Fujita, if two edge dislocations of opposite signs are on planes about 10Aor less apart, they attract each other so strongly that a crack is opened up and other dislocations moving along the same planes will move into the crack and make it wider. A simple model based on a single operative slip system, proposed by Wood ex- plains the formation of extrusions and intrusions as a result of different amounts of net slip on different planes during the forward and reverse stressing under cyclic loading. Unidirectional stress would cause layers of metal to slide in the same di- rection. However, this model fails to explain why, under an alternate loading, the slip continues to monotonically deepen the valley and raise the peaks as observed in experiments. Another theory proposed by Cottrell and Hull states that Frank-Reed sources exist on two intersecting slip planes and a complete cycle of forward and re- versed loading results in an extrusion and intrusion. Such a model would predict the extrusion and intrusion to form in neighboring slip bands and would be inclined to each other and hence could not explain the formation of extrusion and intrusion in the same slip band, parallel to each other. According to Mott [47], a screw disloca- 39 .22.. 3.32 :2: 332 0:28 mES: :::-%o:: 3 m:_:o80m .80: 3 d:m-mmo:o 3 8:: 37. team :5: 8:: 37. mi: .3 55:85.: A: mafia mun—ME g E 40 tion repeats its path through cross-slip. He considered a column of metal containing a single screw dislocation intersecting a free surface. When this dislocation travels a complete circuit, the volume contained in the circuit is translated parallel to the dislocation, which causes the metal to extrude. This mechanism however does not explain why the dislocation does not oscillate back and forth along the same path un- der cyclic stressing rather than traversing a closed circuit. Thompson et al.proposed an edge-screw interaction model for the initiation of extrusions. This model assumes a change in spacing of a pair of parallel screw dislocations after being cut by an edge dislocation. This cut causes jogs in both the edge and screw dislocations and will inhibit the change of spacing for the pair of screw dislocations. It is also hard to explain the formation of intrusions using this model. These extrusions and intrusions however do not dominate the fatigue damage process in composite materials. The crack initiation in composites is dominated by imperfections or defects that form during processing which act as stress concentrators and crack initiation occurs by fracture or delamination at the reinforcement /matrix interface or by nucleation, growth and / or coalescence of voids at or near these stress concentrators. The fatigue crack initiation models proposed by Masuda & Tanaka [50] for discontinuously reinforced composites based on their Sij / A2024 composite fractography results agree with the extrusion mechanism put forth by Forsyth [42] for pure aluminum. Figures 2.lla&b show their proposed fatigue crack initiation models. In Figure 2.1la, a fatigue crack (Stage 1) formed by a slip deformation mechanism at the specimen surface and propagated to the whisker rich zones, when the whiskers are very long compared to the plastic zone size. This crack either forms a dimple at the edge of the whiskers and propagate between the whisker/ matrix interface or intersects and cuts through the whiskers. Since this Stage I crack forms due to Mode II or Mode III type of loading, it rarely opens up and the presence of fibers would resist the Mode I fatigue crack initiation. The model shown in Figure 2.11b proposes 41 E . .29... 5:58 on «2535 55:29:: 29» :o :93: ESE 25:55 :03: 053:.”— 3 $538.8 2:: _ owaum :o .593 3:2: 53:53: :38 2:33.: A: n 3...“ 0.53m 3 EEO/i cosumtu coautxm xooco \ _ Z _ 32m . \‘ 30950: \\\_ _ _ _ _ , , 42 crack initiation at the voids situated at or beneath the specimen surface. 2.3.3 Crack Initiation / Damage Examination Under Cyclic Loading Most of the fatigue studies performed on composites have concentrated on character- izing the damage or plastic deformation after a crack has initiated and have not at- tempted to understand the micromechanisms involved in the crack initiation process. This has been mainly due to the difficulty in detecting the initiation site. Because of the reduced area experiencing maximum stress in the case of fatigue loading as compared to the tensile loading conditions, the detection of fatigue crack initiation site is challenging, but not impossible. From the few studies [51, 52] that have dealt with understanding the initiation, it can be summarized that in discontinuous]y-reinforced aluminum alloys some of the crack initiation sites have been large inclusions, porosity, matrix cracks, fractured fibers, cracked brittle intermetallics, fiber clusters, debonded interfaces and interfaces oriented perpendicular to the loading direction. Failure of these materials occurs by the linkage of the high density of surface microcracks and not by the growth of a stable fatal crack. A number of studies [52, 53] have focused attention on understanding the statistical nature of the microcrack length distributions to better understand fatigue fracture in composites. Bioden replicas were taken from the surface of the specimens at various intervals in fatigue testing, crack length measurements were made [52] under scanning electron microscope and the observations represented in the form of Weibull distribution. Some of the initial investigations [4] carried out on fatigue behavior of 6-alumina or SaffilTM short fiber reinforced aluminum alloy matrix composites revealed enhanced fatigue performance in these composites. Moreover, there has been less scatter in 43 the data obtained during stress-life testing for these composites compared to the unreinforced materials indicating a better control over design of composite engineering components. In general, discontinuously reinforced composites exhibit superior stress- controlled high cycle fatigue properties that improve with higher volume fraction of the reinforcement [50, 54]. The fatigue resistance however is lower for the composites compared to the unreinforced alloy when tested under strain-controlled conditions [55, 56]. A study [56] of SiC particulate or whisker reinforced composites showed that fatigue life can either increase or decrease depending on the strength of the matrix and the reinforcement/matrix interface. A strong dependence of the fatigue life on the mean stress has also been reported, in contrast to the negligible effect observed in continuous fiber reinforced metal matrix composites. When tested under stress- controlled conditions, at three different stress ratios of -l, 0.1 and 0.7, the composite specimens tested at high stress ratios had much shorter lives than those tested at low stress ratios, typical of mean stress effects in metals. Hence, conventional mean stress relationships can be applied to analyze the composite behavior. On stress-lifetime basis, the composite material exhibited improved fatigue life over the unreinforced matrix at intermediate and low stresses or under high cycle fatigue conditions. On application of a constant low stress, the significantly low strains generated in the composite compared to the unreinforced alloy due to substantially higher Young’s modulus of the composite led to this behavior. However, comparison of fatigue life data on a strain-lifetime basis indicated little difference between the composite and unreinforced material in the high cycle fatigue regime. At shorter life times, where plastic strains begin to dominate, the fatigue resistance of the composite is inferior to that of the unreinforced alloy. An explanation for this reduced fatigue resistance has been attributed to the earlier observation of higher levels of maximum local strain in the matrix region of a composite compared to an unreinforced matrix strained to 44 equivalent bulk strain levels. The influence of microstructure on the fatigue fracture surfaces of a 20 vol % SaffilTM fiber reinforced AA6061 alloy composite has been studied by Levin et al.[57]. In this study, an attempt has been made to quantify the geometry of the fracture sur- faces as it has been found to be very useful for understanding fatigue crack growth. Testing was performed under plane strain with nearly zero stress ratio using a sinu- soidal waveform at a frequency of 50 Hz. An important observation is that the fibers, and not the grain size, control the crack path in these composites. Levin et al.also observed that increased work hardening of these composites, both for monotonic and cyclic straining, did not correspond to any detectable increase in matrix microhard- ness. They speculate that any increased hardness of the matrix in the composite must be of a local character. The composites further showed higher resistance to deformation under compression than under tension. Another important observation, detection of a thin layer of aluminum on fibers on the fracture surface, suggests that interfacial separations do not really represent fracture at the fiber/ matrix interface. They also indicated a strong deviation of the fatigue crack towards the fibers. The fatigue behavior of Al-7Si-O.6Mg and Al-58i-3Cu—1Mg alloys with 20% A1203 short fiber reinforcements produced by squeeze casting technique was investigated by Liu and Bathias [51]. These composites were tested under tension-tension fatigue at room temperature with a stress ratio of 0.1. It was observed that high matrix strength led to high fatigue strength of the composite and the incorporation of fibers decreased the ductility of the composite. The fatigue damage in Al—7Si-O.6Mg matrix composite was dominated by fibers since the strain to failure of the fibers is much smaller than that of the matrix. In contrast, matrix dominated damage was revealed in the case of Al-5Si-3Cu-1Mg matrix composite as the strain to failure of the matrix and the reinforcement is nearly the same. Also, specimen surface roughness ranging from 0.1 to 2 pm did not influence the fatigue strength results. Crack initiation took 45 place at multiple locations such as large inclusions, porosity, cracked fibers and in- terfaces oriented perpendicular to the loading direction. Longitudinal and transverse (orthogonal) growth and linkage has led to the crack growth and failure of the com- posite. Similar results were reported in [52] from tests conducted on 20% SiC whisker reinforced 2124—T6 aluminum matrix composites. The inconsistencies expected between the results obtained from in-sz’tu surface damage observation studies and bulk damage examination studies due to presence of plane stress conditions on the free surface as compared to the plane strain conditions in the bulk of the specimen, were critically examined by Mummery and Derby [58]. The deformation at the surface of the specimen relieves the strain at the reinforcements by allowing them to deform out of the plane of the surface. This mechanism is clearly not present in the bulk of the material. From this work, Mummery and Derby concluded that even though a major part of the in-situ observations such as the onset of damage at the reinforcing phase prior to macrocrack initiation and the basic micromechanisms of crack propagation remain valid, some other observations like the quantitative information on reinforcement cracking, extent of damage ahead of the crack tip and the finer details of the crack path are invalid. 2.4 Design of Fatigue Testing Machines The basic idea behind the concept of fatigue testing is gathering critical information such as fatigue life and/or location of failure which will contribute to the design, construction and maintenance of structures in such a way that they are free from failures. Nearly 80% of machine failures are due to fatigue, a fact that makes present day fatigue testing extremely important. In this present study, since the identification of fatigue crack initiation sites and the observation of crack propagation are the critical topics of interest, the methods 46 available for this purpose will be briefly discussed here. It should be understood that while the initiation of fatigue crack is influenced only by conditions in a small volume near the point of origin, the propagation is afl'ected by conditions through out the cross section of the test piece. Hence, the process of initiation has to be studied separately from propagation in order to avoid erroneous conclusions. Some of the important testing procedures that could be adapted for examining initiation and growth of fatigue cracks are: 1. Microsc0pic tests - optical or electron and more recently, acoustic microscopy. 2. Magnetic particle testing. 3. Penetrant testing -fluorescent or dye penetrant, bubble methods. 4. Mechanical property tests. 5. Ultrasonic testing. 6. Electrical tests. The selection of a particular method for crack detection is based on several factors which are critical to the kind of information that is required. Some of these important factors are: c Desired sensitivity; for example, the smallest crack that needs to be detected. Type of fatigue testing machine; for example, a machine that allows in—sz’tu examination and which allows easy removal of specimen. Type of fatigue specimen employed; ASTM standard specimen for this project. Nature of the applied stress or strain; uniform strain in this project. Mode of fatigue stress or strain imposed; repeated bending in this case. 47 Type of material; magnetic or non-magnetic. Nature of detection method; non-destructive or destructive. Time available for crack examination. The equipment available for detection purposes. Fatigue testing machines, in general can be classified as to: 1. The controlling factor: Stress vs. Strain 2. Type of Stressing: Axial loading, Repeated bending, Rotational bending, etc. The fatigue stage that has been designed for this project is of the repeated bending type under displacement or strain-control. Schematics of various types of repeated bending fatigue testing machines [59] are given in Figure 2.12. The mechanically driven machine as seen in Figure 2.12a, makes use of a crank mechanism and it clearly demonstrates the simple principle behind the construction of fatigue stage used in this study. The specimen is held in the fatigue stage as a cantilever and with the help of a crank or cam, the movable end is given a known displacement applying bending loads on the specimen. The specimen, if it has a uniform width, will be subjected to a constant displacement and hence the bending moment increases linearly over the length of the specimen. These type of machines are commercially available, but depending on the use to which they are applied, the stage can have a complicated design and construction. In-sz'tu examination of the surface of the specimen requires the fatigue stage to be small enough to fit inside the specimen chamber of a scanning electron microscope. 48 .23. .9532: 8:23.: w:...m3 2&5... w::o:o: .8383: .o 89: 32:3 H «fin 0.5»...— ..:...=:.. 5.35014. .: ..:.......¢ 335—]. L 2:... :.U .8250]: .....:..:..E:2..l... 5.2.7:“: .2555... 2;... 23:9"; Q. 1...... 3...]... $.ch 2.7:]... 9.9.33.1. .333...— :c..:.:..>l£ 39:32": 25.3...»— .25.... 8...”...3”... 3.. fibix k OLDILDJ V\\\ \stfNL . \\ \ -.-..- , IWAW £OU U 0 I- . ..... W J.CU I- ram] . u < I I. o.— .1! .r-l - II. Ill-\I L..—n.llllle.- . 802 m flail-l uomz fiW-l- : :23 .556 N... u/m. a .— 45:52 ”3:59: 8.3.15 0. 22:625. .236 L: v.52... .9. 9.38.3.6 9:...usz 2...... 45:0 3v 13:53. 2:5: 35:7. .3 x . Il‘ Illlllll ilk-Id " .‘It‘ In. 4‘ .n \\\\. ..:::..:.2 u:::_u._ 2.2... 5:33... ”3 .232. I :3 . n...\\\l\\\b\\ n .MLM .. WI...» 65:32 .3. 2;..- u:o:w:.>. 3 \\. \\\\\-A _ .E:U .5 23:0 : .3 12:33. 1.82m 83.355 .3 m L 49 2.5 Design of Fatigue Specimen Standard specimen designs and the test conditions are specified in ASTM standards manual so that a common basis for comparison of properties between materials exists. Because standard specimens tend to impose minimum size requirements, it may be difficult to represent the microstructural and composition gradients that exist at critical areas and along crack paths using standard specimen sizes and geometries. Circular specimens used more frequently for fatigue testing are seldom used when repeated bending type machines are used for testing. Plate and sheet materials tested under repeated bending fatigue loading vary considerably in dimensions. If the specimen has a constant width, then the maximum stress is located at the fillet of the vibrating cantilever specimen where failure normally occurs. The specimens designed for this study, as shown in Figure 3.2 of Chapter 3, have been provided with a tapered section to produce a constant maximum stress over their gage lengths. The load is applied at the apex of the triangle formed by extending the sides of the tapered section with an intention of imposing maximum fatigue stresses on surface. It is also very important to have a smooth transition from the grip section to the test section of the specimen and also to design the grip section properly in order to eliminate fracture at the grip ends. The influence of a rough surface on fatigue crack initiation has already been stressed elsewhere and even though it is more critical for cylindrical specimens used in rotating beam fatigue testing, flat or plate specimens should also be polished to the specifications. 2.5.1 Specimen Size Effect Care should be taken when inferences are made on the performance of a large com- ponent based on testing done on smaller standard specimen. In other words, size effects have to be taken into consideration. The fatigue strength of large members 50 is typically lower than that of small specimens. The change of size usually results in variation of two factors. First, it increases the volume or surface area of the speci- men and secondly, for specimens tested in bending or torsion, there will be usually a decrease in stress gradient, thus increasing the volume of material that is highly stressed. It has been observed in the case of plain carbon steel specimens tested in fatigue, that size effects existed only when a notch has been introduced into an oth- erwise smooth specimen surface [60]. Since presence of the notch produces a stress gradient, it can be concluded that the size effect is primarily due to existence of a stress gradient. Actual failures in large parts usually are related to stress concentra- tions and it would be impossible to duplicate the same stress concentration and stress gradient in a laboratory specimen. CHAPTER 3 Experimental Procedure This chapter begins with a brief description of the design of in-situ fatigue stage, in-situ tensile stage, fatigue and tensile specimens that were used for this study and provides important information on the procedures adopted for calibrating strain in the fatigue stage, metallographic polishing and chemical analysis of the KaowoolTM and SaffilTM composite specimens. Finally, the procedures adapted for in-situ tensile and fatigue testing are given. 3.1 Design of Fatigue Stage, Tensile Stage and Test Specimens The important criterion for fatigue stage design have been explained in Section 2.4. Figure 3.1 shows the fatigue stage used in this investigation, which has been specially designed at General Motors Research Laboratories. As intended, the stage can be fixed inside the specimen chamber of the SEM with ease and in-sz’tu investigation of the surface damage can be carried out. This fatigue stage uses an adjustable crank device to provide a constant displacement at the movable end which in turn bends the specimen back and forth. The specimen is held in position as a cantilever beam between the grips. With this kind of bending, a constantly increasing bending 51 52 moment is expected. However, in this research, due to the special design of the fatigue specimen (Figure 3.2), where the width of the specimen is made to decrease linearly, a larger part of the specimen area would experience a constant stress. A detailed explanation on the important features of fatigue stage design and specimen design has been given in sections 2.4 and 2.5. The chief advantage of this repeated bending type of fatigue stage is that the specimen surface preparation becomes less critical compared to other methods such as rotating bending type. The other advantages with this design are: 0 The static mean strain and alternating strain can be adjusted. The speed and hence, the frequency can be varied by adjusting the voltage applied to the motor. The stage is capable of running at rates of 0 to 1000 cycles per minute. The stage which is operated in-sz’tu, can be conveniently interrupted at periodic intervals for detailed examination of the surface. Since the fatigue stage remains in the SEM during testing, an area of interest on the specimen surface can be monitored continuously. It is possible to examine the cracks or other features on the specimen surface at maximum, minimum and intermediate strains in a single loading cycle. The tensile substage supplied as an accessory to the CamScan 44FE SEM has been used to strain the dog-bone type tensile specimens of the KaowoolTM composite, de- signed to specifications shown in Figure 3.3. This substage mounts securely inside the specimen chamber and is driven by a flexible shaft connected to the substage motor drive assembly which is installed on the door of the SEM. The specimen is mounted in the stage between the wedge clamps which are designed with a smooth taper on the back side and gripping serrations on the sample side to firmly hold the specimen dosgmfimxo 3.5:." 8L owfim onwmuam H H.” 05mg 54 in place. The load and the strain readings are displayed once electric connections are made through the feedthrough port on the door of the SEM. Internal calibration is provided through the help of a shunt resistor. Also, the rate of straining or the speed of the crosshead in the fixture can be adjusted. For the investigations aimed at determining the stress levels at which specific microstructural features (hollow and closed shot particles etc.) nucleate a crack, an elongation rate of 0.05 inches/min. has been used. 3.2 Finite Element Stress Analysis Finite element analysis (F EA) was performed using I-DEAS software to visualize the stress or strain distribution on the surface of the specimen for the tapered-gage specimen design and fatigue testing configuration described in the previous Section 3.1. The three dimensional finite element model of the specimen was accurately drawn and meshed using solid brick elements. The following boundary conditions were specified: 0 a three dimensional constraint along one of the edges to simulate the conditions at the fixed edge; 0 a uniform pressure on the top surface of the other edge to simulate the conditions of cantilever type static bending loading. From this analysis, the nature of stress or strain distribution produced on the spec- imen surface from deformation due to the bending loading can be visualized. The results from this analysis are reported in Appendix B. 55 doc—Ewan 253.: no baoEoco ” N6. 95w:— 56 X // // / / .5562? 28:8 .0 ate—=80 H m5 mun—ME 57 3.3 Calibration of Strain in Fatigue Stage The strain that is generated in the specimen is a function of amount of deflection the specimen is experiencing due to bending. Based on this fact, a simple technique has been devised which allows good estimation of the actual strains applied to the test specimens. The technique involves use of a normal test specimen, but with a strain gage applied to its surface, and a mechanical dial gage with a sensitivity of 0.001 inches. The surface of this specimen was polished metallographically before the application of strain gage, under the same conditions as specified for the actual test specimens. This specimen with the strain gage was then mounted in the fatigue stage and strain gage lead wires were connected to the strain indicator. With this configuration, strain readings were taken and the corresponding deflection at the movable end, as measured on the dial indicator, was noted. This was repeated for various configurations of specimen position in the stage, and a deflection vs. strain plot was made. This linear plot has been used as a reference to estimate strain for any amount of deflection produced in the stage. An actual plot obtained for a particular specimen setup in the stage is shown as an example in Figure 3.4. 3.4 Metallography On the basis of the fatigue stage design and the specimen design used, it is expected that the crack nucleation would take place on the surface of the specimen. As such, surface finish plays an important role in the fatigue crack nucleation. Cracks are known to start almost always ,at scratch marks or machining marks when ever they are present. Hence, care must be taken to avoid these obvious crack initiation sites by careful metallographic surface preparation which is briefly explained below. 58 633528 mwfim mamas.“ Eat 83 52am .m> ESSED H 06m 3202 599$ :26ng odm odm v.» 2:3,.” Qmm YIYTT—‘l' 'T‘l YIYTV‘rrY‘rTTj’IT] IIITYIYIYITI VITI' _ llllLlllllLlLlLLl lldLl rlllllllnlllllrl odonm odomm odomm odoom Odo Fm odcmm odomm odovm odomm odomm 0.8% 0.68m o comm odoov ugensomgw 59 3.4.1 Polishing The surface of the specimens were polished using diamond suspension of various grades starting from 30pm down to 1pm and finally with 0.06 pm colloidal silicon solution. The orientation of the specimens were changed by ninety degrees every time a different grade of diamond suspension / polishing cloth was used. The specimens were washed throughly under a stream of distilled water while moving from one polishing stage to the next. Finally, the edges of the specimens were beveled to eliminate crack nucleation at the edge of the specimens. Figure 3.5 shows the various stages involved in the polishing of the composite specimen. 3.4.2 Optical Microscopy and Image Analysis Optical microscopy was performed on the specimens mainly to check the surface finish. Any scratch marks in the transverse direction of the specimen are considered polishing defects and such specimens were either discarded or re-polished. Optical microscopy was also used to see if there were any large variations in whisker distributions and to identify any defect sites on the surface, like porosity. In general, care was taken to see that the surface of the specimens was free from pre—existing cracks and other surface defects. Finally, image analysis was performed to calculate the distribution, orientation and relative sizes of the whiskers and intermetallics on the surface of the specimen. 3.5 Chemical Analysis Chemical analysis was performed on the composites using the Cameca electron probe nlicroanalyzer (EPMA) at General Motors Research Laboratories. The X-ray exci- tation volume in EPMA is on the order of few micrometers, depending on the beam 30pm diamond suspension Ultrapad polishing cloth, Buehler 1 15pm diamond suspension Ultrapad polishing cloth, Buehler l (Sum diamond suspension Polimet polishing cloth, Buehler l lum diamond suspension Texmet polishing cloth, Buehler 1 0.06pm Colloidal silica suspension Microcloth, Buehler Figure 3.5 : Procedure for metallographic polishing. 61 energy and average sample composition. This obviously limits the ability to analyze submicron details and segregation phenomena. Although EPMA has lower sensitivity to detect light elements and has lower sensitivity to spatial resolution ratio compared to secondary ion mass spectroscopy (SIMS), it has been a reliable and a widely used technique for characterization of composites. The polished sections of the composites may have topographic surface relief at the matrix/fiber interface. This topography may cause problems in absorption corrections for X-rays in EPMA. Hence, the surface for all the specimens were polished and thoroughly cleaned according to the proce- dure described in Subsection 3.4.1 for consistency. Composition distribution maps for Al, Si, Mg, Cu, Fe, Ni and O were taken from the same location on each specimen for both KaowoolTM and SaflilTM reinforced composites. The maps obtained for O clearly reveal all the whiskers in a particular area, due to abundance of oxygen in the binder and whisker. 3.6 Description of the Experimental Setup Used The experimental setup used for in-sz’tu SEM studies was essentially the same for both tensile and fatigue testing. A schematic of the setup is shown in Figure 3.6. The setup can be basically divided into four parts. 0 The fatigue stage or tensile stage. 0 Means for controlling the mechanical deformation process: Mechanical aspects and electrical circuitry. 0 Means for observing the fatigue effects: Microscopy. 0 Means for storage and retrieval of the results. The fatigue or tensile stage was mounted inside the specimen chamber of a Cam- Scan 44FE field emission gun scanning electron microscope. Electrical connections .958“ 0:83 23 0:35 5.— 958 .3555qu 6d 0.53% W J 8283. 83> .5838 a?» Sun."— _U U I F L 62 2mm 5:83 Eur.— vv 509—30 \ 63 were made through the 25 pin feedthrough port on the door of the microscope. A fatigue stage controller consisting of a power supply and an integrated circuit for electronically counting the number of cycles of loading was used for controlling the fatigue in-situ. The frequency at which the specimen is fatigued was controlled by adjusting the voltage on the controller. A voltage of 5v was consistently applied for all the experiments to maintain the same frequency (540 cycles/min). The means for observing the fatigue process were the CRT of the scanning electron microscope and a video monitor. The results from the experiments were recorded on a video tape using a video cassette recorder and retrieved through a thermal video printer, either directly from the CRT of the SEM or from the play back of the recorded tape. 3.7 Description of the Testing Procedure 3.7.1 Procedure for In-Sz'tu Fatigue Testing Each specimen was taken through a series of identical steps for the sake of consistency in the evaluation of the results. Any change in the procedure would have the potential to bring in effects of unknown variables which might lead to experimental error. The following procedure was developed through experimentation and has been strictly adhered for all the specimens. First, the thickness of a specimen was measured with Vernier calipers at three different locations; at the center of gage length and at the two grip ends of the specimen. The specimen was then polished according to the procedure described in Subsection 3.4.1. The edges of the specimen along the gage length were beveled using a metallographic paper. The thickness was again measured using Vernier calipers, this time only at the end grip sections and not in the gage length. From the thickness values at the grip ends before and after polishing, it was possible to calculate the 64 thickness in the gage section of the specimen used for testing. Care was taken not to scratch the surface of the specimen or introduce cracks by bending the specimen. Optical microscopy and image analysis were performed on the specimen to check surface finish as well as the quality of the specimen. The specimen was then placed inside the scanning electron microscope and the entire gage length was scanned to detect any pre-existing cracks. Once the specimen passed through this procedure, it was ready to be mounted in the fatigue stage. While the specimen was being mounted in the stage, two accurately machined metal blocks were placed under the movable end of the fatigue stage to avoid appli- cation of any high stresses or possible bending. Once a specimen was mounted in the stage, it was considered to be at a state of zero cyclic loading. The surface of the specimen was rescanned after the fatigue stage was fixed inside the specimen chamber of the scanning electron microscope. The surface in the gage length area of the spec- imen was recorded on the VCR for later analysis. Having made sure that no cracks were introduced during the installation process, the fatigue stage was taken out of the chamber and the deflection measured at the movable end of the fixture using the dial gage indicator. This value was read five times for consistency. The amount of strain generated in the specimen was estimated from the measured deflection at the movable end and the plot between deflection vs. strain obtained during calibration as outlined earlier in section 3.3. Having applied five cycles of loading during deflection measurement, the fatigue stage with the specimen was once again placed in the SEM. The gage length area of the specimen was scanned once again. This condition represents five cycles of loading. Points of interest were recorded on the VCR and also placed in the SEM stage memory. The fatigue stage was operated using the controller with voltage set at 5v. The tests were stopped at 500 cycles of loading and the surface of the specimens were scanned and recorded on the VCR. This process was repeated after interrupting the tests at 65 2500 cycles. The reason for scanning the entire surface at intervals of cyclic loading was to identify the micro crack nucleation at multiple sites with increasing number of cycles. It has been observed that order (very few new microcrack initiations) was restored in the specimen after 2500 cycles. All the defect locations that developed in addition to possible crack nucleation site locations were memorized in the stage location storage system of the SEM. The tests were interrupted after every 5000 cycles of loading and all the memory locations were re—visited and recorded on the VCR. This process was continued until extensive crack growth or failure occured. Following failure or extensive cracking, it was possible to recall a series of pictures revealing the crack growth scenario at different numbers of cycles. From this information the crack initiation sites were identified. Failure analysis was also performed on the fractured specimens to identify the type of failure and to ascertain the crack nucleation site. It is also possible to infer if presence of sub surface features had any influence on the crack nucleation. 3.7 .2 Procedure for In-Sz'tu Tensile Testing The procedure followed for in-situ tensile testing of KaowoolTM reinforced composites using the Ernest Fullam tensile substage will be described here. The design features of the tensile stage and specimen were described earlier in Section 3.1. The procedure adopted is straight forward. The uniaxial dog-bone tensile speci- mens were polished on both sides using the procedure described previously (Section 3.4.1) and then mounted in the in-situ tensile stage. The stage with the specimen is fixed in the specimen chamber of the SEM. This allows one of the surfaces of the specimen to be monitored under SEM. The controls and output displays (force, elongation) for the test were located outside the microscope. The surface of the spec- imen is scanned under SEM before application of load and all the shot present were recorded in the video recorder. The locations of these shot particles was memorized 66 in the stage location storage system of the SEM. Specimens were tested under tensile loading at a strain rate of 0.05 inches/ min. The tests were interrupted every 30 sec- onds to revisit all the locations in SEM stage memory and record the damage at these locations. The stress as indicated by the output device was noted. This procedure was continued until fracture occured. The stress level at which a crack initiated from each shot particle was identified from the series of pictures that were taken earlier from the location containing that shot particle. Results obtained from in-situ fatigue and tensile testing on these composite ma- terials will be presented and discussed in Chapter 4. CHAPTER 4 Experimental Results & Discussion The important observations from this investigation on fatigue damage accumulation in Al-Si alloy matrix composite materials will be presented and analyzed in this Sec- tion. The discussion that follows will put together observations from this work and information from the literature survey and attempt to develop a better understanding of the mechanisms involved in the fatigue crack initiation in composite materials. To start with, the microstructural aspects of the unreinforced as well as composite spec- imens will be discussed, followed by discussion of the results from the in-sz’tu fatigue testing and fracture analysis of the failed specimens. Results from composition analy- sis performed using EPMA to identify the chemistry of the intermetallic constituents present in the microstructure, and tensile testing carried out to identify the stress level at which crack initiation takes place at some of the important microstructural features will also be discussed. 4.1 Optical Microscopy and Image Analysis The unreinforced and composite specimens were produced by squeeze casting tech- nique and later heat-treated and machined to the dimensions of the specimen as explained in Section 3.1. The surfaces of the specimens were polished as outlined in 67 68 Subsection 3.4.1 and observed under an optical microscope. Figure 4.1a is an optical micrograph showing the interface between an unrein- forced alloy (microstructure indicated by A) and a SaffilTM reinforced composite (mi- crostructure indicated by B). The micrograph in Figure 4.1b shows the microstructure of a KaowoolTM reinforced composite at higher magnification. The microstructure of the SaffilTM and KaowoolTM composites resemble each other due to the presence of identical chemical constituents and application of identical processing technique. Figure 4.1a mainly differentiates between the unreinforced and composite microstruc- tures. The microstructure of the unreinforced matrix alloy shows primary aluminum dendrites extending into a “needle-like” aluminum-silicon eutectic network and some intermetallic precipitates. This microstructure is clearly modified with the incorpora- tion of fibers. The composite microstructure is complicated by the presence of fibers, primary aluminum, primary silicon, a proeutectic, silicon eutectic along with several intermetallics. A random planar fiber orientation is expected due to the method of preform prepa- ration used, as discussed in Subsection 2.1.2. Observation of the microstructure at higher magnification (Figure 4.1b) reveals coarsening of the eutectic structure and preferential nucleation of primary silicon and eutectic at fiber boundaries. The grain refinement in composites (relative to the unreinforced materials) is clearly evident from the two micrographs. These observations are consistent with previously pub- lished work [27] as reported in Section 2.1. The net effect of these changes would lead to an increase in the interfiber matrix hardness and a decrease in the overall ductility [6]. Thermal effects during the squeeze casting process would lead to an increased grain size towards the top of the preform as reported by Clegg et al.[38]. Hence, the grain size might vary depending on which section of the preform the specimen is machined from, but it would always be more refined as compared to the unreinforced alloy. This refinement leads to enhanced mechanical performance of the composite as 69 dimes—E8 888E?“ Eiookoevm $2 a mo 2392883”: 2: wEBoam aaahwouflfi 33.50 3 Am: 2:89:00 88858 Siam a was A020 0088808: 00 m0 000200 000 80 0000 30800 0 8 8080 0838 18008 98.5000 8008888 $0500 A0 68 0.8me g E 83 4.3.2 Crack Initiation/Damage in KaowoolTM Composites For the KaowoolTM reinforced composites, initial investigations revealed that crack initiations occured at multiple locations on the surface of the specimens and frac- ture of the specimens occured due to growth and coalescence of these cracks. Figures 4.10a&b are optical micrographs showing the same area on the surface of a KaowoolTM composite before and after fatigue fracture. Figure 4.10b shows the tortuous fatigue fracture surface indicating crack growth from multiple initiation sites joining into a single crack leading to failure. In general, for KaowoolTM composites cracks initiated within the first 5 cycles of loading at surface defects such as hollow shot particles and shrinkage holes, and propagated into the matrix at rates influenced by the pres- ence of stress concentrations such as fibers and intermetallics adjacent to them. In the absence of such defects, cracks were seen to initiate at fiber/ matrix interface in the presence of intermetallics adjacent to and on the surface of the fibers and fiber/intermetallic interface. Fatigue cracks also initiated to a lesser extent due to the fracture of intermetallics, fracture of fibers, and also cracking of the matrix, in that order. Figure 4.11 shows a series of scanning electron micrographs from a particular location on the surface of a KaowoolTM composite taken at different number of cycles of loading. Figure 4.11a shows extensive cracking on the surface after 25,000 cycles of loading. Inspection of the micrographs taken earlier from the same location after 5 cycles (Figure 4.11b) and zero cycles (Figure 4.11c) of loading, identified the hole in the solid shot particle as the crack initiation site. This crack, once initiated, instead of fracturing the intermetallic (CuNi or FeNi) next to it (area A in Figure 4.11b), preferred to propagate along the fiber/intermetallic interface. Even on the other side (area marked B), the crack deviated along the fiber/intermetallic interface before propagating into the matrix. This crack however fractured the fiber (area marked 84 0.800000 00808 00000 3 000 $8008 020.8 00800 An 08000080 2s_00300& .0 Mo 00880 080 00 0000 00 w83000 mnaflwBEE E0500 ” ofiv muflma E 3 51 OH. . :1 gm 85 .3080 000008 03000 0 3 000.0808 0_ 008 00005 08008 00 0293 o 00 0000 08.00 000 800.0 0308 005020 30000000m A0 000 08002 00 0080.80 00 00.00 0800 000 808 00003088 2am 00800.0 00003000 02$ 3 08000800 V008030.0M 83> 2 0 m0 000M000 00... 00 w8008 .00 02030 ocofim 0000.0 530% x0000 0300038 m83000 0m88 Ema 005020 0.000000% A0 H :0 0.5me 86 B) and passed right through it. It needs to be pointed out that the micrograph in Figure 4.11c is taken from the recording on the video tape and hence reflects the poor quality of the image. Figures 4.12 a,b&c are scanning electron micrographs showing crack initiation and propagation from a hollow shot particle. The walls of the hollow shot particle cracked during the first 5 cycles of loading and the crack stopped after propagating a short distance into the matrix. However, the stress concentration associated with this crack has either fractured or debonded the fibers that were adjacent to the shot particle, resulting in continued crack propagation. A higher magnification image of the areas on either side of the shot particle, shown in Figures 4.13 a&b, clearly demonstrate the above observation. In general, cracks that initiated at a shot particle did not directly penetrate into the matrix to propagate, but created a stress concentration in the vicinity of the crack that either fractured or delaminated the fibers and/ or intermetallics that were present. The mode or nature of crack initiation observed at the shot particles during fatigue testing varied depending on their size, shape and location. Cracks in shot particles initiated by the cracking of the walls and/ or cracking of the base (bottom of the shell). Shot particles with thin walls are the easiest to crack while the solid shot particles hardly cracked. Cracks occasionally initiated in the close proximity of a solid shot particle and propagated into the matrix, while the shot particle itself remained intact. Figure 4.14 shows one such observation. The presence of mismatch in stiffness between the fiber and the matrix in a com- posite is known to lead to an increased stress concentration at the fibers, which leads to damage at / near the fibers. Even though the strength of the fibers or shot particles is far greater than the high stresses they are subjected to, a region of local plastic deformation exists near these inclusions which subjects the reinforcement to even higher strains. In this context, the strains that are applied rather than the stresses 87 .30000 000000 05000 0 3 0000808 00 00x0 008% $8008 00 000000 m 00000 A0 000 000000 comm 00000 3 00.003 83: 00:0 A0 680000000 3.0000300VH R200» 2 0 8 280000 0080 30:08 0 0000.0 080030000 000 080888 80000 M83000 0000030800 2mm 0 3.0 000mg 88 .0000 05 .00 080 8w: 0;» 00 000.003 .0000 00: .00 080 0.000 05 00 000< A0 .30 0000.30“ 8 003080 280000 0080 30:00 00 00000900 0000 080 00 38> 0800050300 0003: u 2.0 000mg 30 3 89 .30000 020000 03000 0 3 0000208 2 0000 0000i .0w008 0000020 000000000 200m 3 .0m008 0000020 0.000002% 00 6200000000 0000030032 8 220000 0020 0:00 0 0000 080888 20000 w83000 02000mo0800 2am “VH6 000mg 90 control the damage process near these shot particles. Figures 4.15a,b&c are a series of scanning electron micrographs showing crack initiation and propagation from a processing defect, a shrinkage hole. Cracks in KaowoolTM composites also initiated from fiber/intermetallic interfaces. Figure 4.16a is representative of several crack initiations at fiber/intermetallic inter- faces observed on the surface of the specimen. It shows a crack that propagated extensively into the matrix. The initiation site for this crack can be identified with reference to micrographs b&c in Figure 4.16, which show the same location after 500 cycles and 5 cycles. The presence of two closely spaced fibers near the crack initi- ation can also be seen on the micrograph. The small crack that initiated between the fibers (Figure 4.16b), did not however propagate. Crack initiation at an inter- metallic can also be seen in Figure 4.16a. This initiation occured towards the later stages of cyclic loading. Figure 4.17a shows several such cracks initiating from the KaowoolTM fiber/intermetallic interfaces perpendicularly oriented to the loading di- rection. Micrographs taken from the same location after 500 cycles and 5 cycles of loading are shown in Figure 4.17b&c for comparison. While the fibers themselves act as inclusions and create local stress concentrations in the material, the presence of less ductile intermetallics adjacent to them would lead to tensile residual stresses around fibers, intermetallics and the matrix and hence, increase the chances of crack nucleation. The fatigue crack initiation from shot particles has been speculated upon in lit- erature [10] and observed by Canumalla [8]. The modulus of the shot particle has been measured [8] using scanning acoustic microscopy and it has been confirmed that shot particles possess a Young’s modulus of 131.5 GPa, which is very low compared to 225 GPa exhibited by the fiber. The premature cracking of the shot particle has been attributed to this lower Young’s modulus. Further discussion on the early frac- ture of shot particles will follow the observations from in-sz’tu tensile testing of the 91 .3008 020000 20000 0 3 000000000 00 0000 008% @8002 00 020.8 0. 0000 A0 000 .0298 com 0000 3 02000 coo mm 0000 A0 080000000 2.020039% 8 200 03280000 0 00000 0800300000 000 0000888 020000 m83000 0000030200 2mm. . 3.0 000mg 92 $000de 00000: 03:00 0 3 0000000000 00 008 00000m 0000002 00 000000 m 00000 00 0000 .0398 com 00000 3 000000 80.3 00000 00 0000000000 2.0303030 E 0000000000 0500000000000<00£ 000000 0000000005 00000 0002000 9003000 0.0000mo0000002mm ” 3.0 005mg 93 .30000 0000000 03050 0 3 0000000000 00 0000 00000m $000002 00 0000.00 00 00000 00 0000 £000.00 com 00000 3 £00000 ooodm 00000 0 00000000 -0008 250003030 00 0000000000 0300000000000<00nm 00000000 8000 000000000000 00000 0003000 M00330 003003000000 Sam 0 bad 005M000 94 KaowoolTM composites in Section 4.4, where an effort has been made to determine the stress level that causes crack initiation at shot particles. Information on how a fatigue crack grows when it approaches a bimaterial inter- face has been investigated by Suresh et al. [62]. They concluded that a fatigue crack advancing in a normal direction towards the interface from the soft side will be de- flected away from the interface, while a crack approaching the interface from the hard side would penetrate the interface undeflected. Applying this observation to the shot particle/intermetallic bimaterial interface and shot particle/ matrix bimaterial inter- face as encountered in Figures 4.11 and 4.12, and assuming that a crack initiating and propagating from a spherical shot particle approaches the interface normally, a very important conclusion can be drawn. In Figure 4.11, the crack that initiated at the shot particle approached the shot / intermetallic interface normally, but instead of penetrating the intermetallic, deviated along the shot/intermetallic interface. This suggests that the intermetallic is harder than the shot particle. For the case repre- sented in Figure 4.12, the crack penetrates into the matrix where it is blunted. Here, the shot particle being harder than the matrix, represents the situation of a crack ap- proaching the interface from the harder side to the softer side. This does not explain why the advancing crack (in the area marked B in Figure 4.11) penetrated the harder fiber from the softer side (matrix). It is speculated that the crack that formed at the shot particle, once extended into the matrix, created a stress concentration in the vicinity of the fiber. This led the fiber to fracture first and the crack penetrated into the matrix and joined with the crack that originally started from the shot particle. These crack propagation scenarios can be easily visualized from the schematic shown in Figure 4.18. It is also important to understand how a crack behaves on approaching the in- terface at an arbitrary angle. This could possibly explain why a crack propagating in the matrix penetrates straight into a fiber on some occasions while seeking the 95 030:0 50.05 .3 00.0305 000 8.005 0000.25 0.: 80.0 0030 000000000 00 0.003 00.0 00000; 0. 00: 00008 80.0 0.0000 0 0:03 60000005 0.: 0.0000000 0.003 00: 00.000 00 007. 00000: E000 0.0000 < 00000005 00000055 0 $3008.50 :0 0.0000 0.03000 0 00 00300.0; 05 9:30.00 0:080:0m ” 3.9 0.5»...— l. I: . £00 0690‘ \ I "'I|‘ xEUE 96 fiber/ matrix interface some other times. Analyzing the crack-fiber interactions that were encountered during the fatigue studies in this research work, an optimum angle of 40-50“ between the load axis and the fiber, has been obtained. The definition of this angle can be easily visualized from the schematic diagram shown in Figure 4.193.. All the cracks that approach the fiber at angles greater than this value would follow the fiber/ matrix interface, while cracks that approach at angles less than 40-50" would penetrate through the fiber. Fibers oriented at right angles to the loading direction, represented by Figure 4.1% either had cracks along the fiber/matrix interfaces or have split in half along their length. 4.3.3 Crack Initiation/Damage in SaffilTM Composites In the case of SaffilTM reinforced composites, crack initiation occured very early in the fatigue process, within the first 5 cycles that were applied during deflec- tion measurement and mainly due to debonding at fiber/matrix interface. A few cracks along fiber/matrix interfaces were sometimes even seen prior to any cyclic loading being applied to the specimen. The strains induced during polishing seem to cause this debonding in SaffilTM composites, even though no such damage occured in KaowoolTM composites under similar conditions of polishing. Figures 4.20a&b show debonding at the fiber/ matrix interface for several trans- versely oriented fibers in a SaffilTM reinforced composite. Cracks were seen only along the transversely oriented fibers with no damage for the longitudinally oriented fibers. Observation of such interfacial cracks at higher magnification, shown in Fig.4.21 a&b, leads to the suspicion that these interfacial cracks may not be true fiber/ matrix sep- arations. The fibers could be covered with a thin layer of some material which leads to debonding. Such thin layers cannot be easily resolved from micrographs. Also EPMA chemical analysis is not sensitive enough either to pick up signal from this thin zone. Crack growth in SaflilTM composites occurs due to link up of several such 97 000000.... 50.0.5000... 00.0 0.000 0.003 00900. :0... 00000.» 0.000 :0 .0 0000 0... 050.0080? 0.0000 0 0.23 .0000 0... 000000.50 0.003 00900 :0... 0.00. 00.000 00 0000 0... 00.20.00.000 0.0000 < .0000 0000.0 0... 0.0.3 000.. 0... 00 00.00.0000 0... 0532.0. 00.00E0_.0m 3 00 .0 H 010 9:00.0— 3. .0. is SIXD SSSJ 98 .3005 000000 0.0000 0 3 00000.00. 0. 008 00005 00.000. .00 00.008 0 0000 00000000000 3.0.0.03 0. 000000000. 00000.0}0000 00 00.000000 00.3000 2.00%00080 2mm 3%? 0nd 053% 30 E 99 00.000000 000000000 0.00 000000 0060 00 000000000 0. 0000000 3.0000000. 03000.0: .00 00000. 0.00 00.00.080.038 0000.0 0 00 00.000. .00 00.008 0 00000 000.000.0000 000.5% 0. 000000000. x.00000\000.. 0.0 00.000000 00.3000 000000000000 saw 300?. 0 H040 005mg .0. . 3 100 closely spaced debonded interfacial cracks and the specimens fail catastrophically, ex— hibiting short cyclic life. Even though fibers that are oriented parallel to the loading axis might provide positive reinforcement to the matrix by supporting the load, the debonding of transversely oriented fibers would weaken the material and hasten the fracture process. Since larger volume fractions of fibers would lead to closer spacing between the fibers, it is speculated that fracture should take place even faster for higher volume fractions of fiber reinforcement. The observations related to the contrasting failure behaviors exhibited by these composites under identical testing conditions are summarized in Table 4.1. The debonding at the fiber / matrix interfaces in the SaffilTM composites leads to a situation where the probability of finding a crack in the proximity of another is higher compared to the situation in KaowoolTM composites. This effectively controls the fatigue life of these composites as reflected in the Table. The different modes of crack initiation in these two composite systems can be explained with reference to the basic difference between the composition of two fibers. While KaowoolTM has 50% silica, SaffilTM is essentially alumina with 3% silica. The importance of the interaction between the silica in the fiber and the magnesium in the matrix has been stressed earlier in Section 2.1.4. The formation of a spine] MgO.A1203 in the interface leads to a loss of the magnesium in the matrix there by a decrease in matrix strength due to lower Mggsi precipitate formation. In SaffilTM, the lower contents of silica lead to the formation of a smaller interface layer compared to the KaowoolTM. In this case, Mg essentially remains in the matrix and provides increased hardness to the matrix. The differences in the strength and composition of the interface reaction product and relative strengths of fiber and the matrix are expected to lead to the differences in the observed fatigue crack initiations. The observations from the surface crack initiation studies for both KaowoolTM and SaffilTM composites are further corroborated by fracture mode analysis observations which will be discussed in Section 4.5. 101 $0.3 800 80.8 80.80 0000.00 00 00.0.00 000000000. x.00000\0000 000.0 800000.000 000 00.0000 .00 00000.1. 00.00000 000. 003000 0000 00.0000000n. 00.00.0.0. 0.0000 00.0.00 0 00.0.00 000-0 00 00.0.00 .00....000000000. 000000000. 000000000. x.00000\000.,m 05000000000. \ 000. n. 00.0.0000 000m 30 800005 2.800-800 0050 02000 010 ”0.00.. 00300 38000on 0080 0&0 005 ”0.00.. 00300 00000 0.0000 00.02.0000 2 0.0.00 00.02.0000 2 0.00303. 00000.00. 000.00>0000O 00.0000 000.00.. 2 5.0000 000 2 0.00303. 000.0 000.00>00000 000 00.009000 0.00... 0 H0 .0.an 102 4.4 Results from In-Sz'tu Tensile Testing of KaowoolTM Composites The observation of fatigue crack initiation at shot particles on the surface of a KaowoolTM reinforced composite specimens motivated the in-sz’tu tensile study of these composites. In-sz'tu tensile testing was conducted to elicit information relating the stress levels to the fatigue crack initiation at shot particles. Dog-bone type ten- sile specimen designed to ASTM specifications, shown in Figure 3.3, were used for testing in an Ernest H Fulham tensile substage mounted inside a scanning electron microscope. A brief description of the tensile substage and specimen design has been given in Section 3.1 and the procedure adopted for testing was discussed in Section 3.7.2. A constant elongation rate of 0.05 inches / min. was applied in tension and the microstructural features of interest on the surface of the specimen were monitored periodically under the scanning electron microscope. Figures 4.22 and 4.23 are scanning electron micrographs showing crack initiation at two different shot particles on the surface of the specimen under tensile loading. Figures 4.22a,b&c are taken from the same location during interruption of the tensile test at different levels of stress. Crack initiation occured from the hole, present at the base of the shot, perpendicular to the loading direction, at a stress level of 23- 29 ksi and propagated into the matrix. Close observation of the micrographs reveal that the wall of the shot particle has not nucleated a crack prior to 29 ksi even in the presence of a small hole on the inside of the wall. Figure 4.23 shows the crack initiation scenario at another shot particle. Micrographs 4.23a,b,c &d show the same location on the surface of the specimen at different levels of stress. The crack initiated early at the shot particle in the presence of an intermetallic next to it (Figure 4.23a). This crack probably deviated and cracked the thin wall on the right side of the shot particle, while the base of the shot opened up another crack (Figure 4.23b). At 29 103 .30000 00.000. 03:00 0 .3 00000.00. 0. 0.000 00000m ..00. mm .0 .000 ..00. mm 3 000. m. .0 .00 0.0>0. 000000 00 000.0608 M00000. 0..0000 00.000 00.009000 2 0.00303. 0 0. 0.0.0000 000.0 32.0.. 0 00 8.00.0.0. 0.0000 m0.300.0 00000000000 2.1mm “NNé 0009b 104 .30000 000000. 0.0000 0 .3 00000.00. 0. 0.000 000000 000. 0.0 A0 000 ..00. mm .0 000. mm 3 ..00. m. .0 .00 0.0>0. 000000 00 .000.0..0000 00.000. 2.0000 00.000 00.000080 2 0.00303. 0 0. 0.0.0000 000.0 30:00. 0 00 8.00.0.0. 0.0000 00.3000 00000000300 2am ”mad 0000a 105 ksi, the thick wall on the left side of the shot cracked. Finally, Figure 4.23d shows the pr0pagation of the crack at the thin wall into the matrix. The premature cracking of the shot particle in these composites during fatigue testing was explained earlier as due to lower Young’s modulus of the shot compared to the fiber. It can also be explained with reference to study performed by He at al. [63] investigating the effect of a reinforcement called microballoons, similar to hollow shot particles, on the tensile properties of the MMC. The principal use of the microballoon- reinforced metal matrix composites is in the marine industry where advantage is taken of their low density and high damping capacity. They predicted from finite element analysis that under tensile straining, the maximum principal stress occurs on the inside wall of the microballoon and hence crack initiation occurs there for both thin (wall thickness to radius ratio less than or equal to 0.2) and thick walled (thickness to radius ratio equal to 0.5) balloons (shot particles). The magnitude of principal stresses in the thin walled shot particles is similar to that of thick walled shot particles and hence, tensile failure is expected to initiate at a similar strain but at different stress levels. Even though these observations are generally agreeable, most of the cracks from shot particles in this present study were seen to initiate from the base of the shot either from the inside (between matrix and shot) or on the shot particle itself and at stresses as low as 23 ksi which is less than 27 ksi, the 0.2% yield strength lTM reported for 15% Kaowoo composite (Appendix A) and nearly 50% of the tensile strength of 44 ksi reported for this composite material (Appendix A). 4.5 Results from Fracture Analysis Fatigue failures usually leave patterns or features on the fracture surfaces from which useful crack initiation and propagation information can be obtained. In this research work, fatigue cyclic loading was continued until either extensive crack growth occured 106 or until the specimen fractured suddenly. Fracture analysis was performed on the fracture surfaces of the failed specimens to confirm the initiation site and to study the finer details of fracture such as brittle, ductile modes of failure, secondary cracking, presence of sub-surface features that aid the crack initiation process, extent of fiber pull-out etc. The scanning electron micrograph shown in Figure 4.24a supports the earlier ob- servation of crack initiation at shot particles in KaowoolTM composites from surface crack initiation studies. It reveals a shot particle on the surface of a specimen near fracture region that cracked and advanced into a major crack that ultimately caused the material failure. Also, the fracture surface in Figure 4.24b reveals a secondary crack that initiated from a cracked shot particle inside the specimen. The presence of secondary crack at the shot particle on the fracture surface further substantiates the earlier observation of crack initiation at shot particles. Cracking of the shot particles may not form a major crack, but will definitely aid in the fracture process. Figures 4.25 a&b show bare fibers that were exposed on the surface of the specimen near fracture region in a SaffilTM reinforced composite that strengthen the earlier observation of fiber/ matrix debonding from surface fatigue crack initiation studies. Figures 4.26 a&b are scanning electron micrographs comparing the fracture sur- faces for KaowoolTM and SaffilTM reinforced composites. The presence of an enor- mous number of fibers on the fracture surface for SaflilTM reinforced composite in- dicate a poor chemical bond between the fiber and the matrix. Fracture analysis in this case confirms the earlier observation from surface crack initiation studies, that failure occurs in SaffilTM composites due to delamination at fiber/ matrix interface. It can be clearly seen that there are more fibers that are transversely oriented to the loading direction than longitudinally oriented fibers. Thus, it can be concluded that cracks initiate due to debonding at the fiber/ matrix interface and the advancing crack front seeks the debonding interfaces at other transversely oriented fibers. The 107 008.0000 2 0.00300v. 00.0 0.0.00. 00.0.0000 000.0 0000. 050050.00 000.000 0000000. 00.0 00 00.0000 000000000 00.3000 0008. 2mm .0. 008.0000 00.000800 2 0.00300v. 0 .0 00.000 0000000. 0000 000.000 008.0000 00.0 00 0.0.0000 000.0 00.3000 00000000000 2mm .0 0 00.0 0000.0. .0 E Tl.‘ . .rr... -\ ., 3 . 108 0000000. 000000. 0000 0000000000 2 0....00 0 .0 000.000 000 00 00000000 0000.. 0000 00.3000 00000000000 2.00 300.0 0 00.0 0000.0. .0 E 933‘ , , ..... 0.3.030: facet}... 2:003: 5.2.0.3330. ’02:: 0.: 2.0.0000: \ 2200.030 comm .320 “c.0000: .000 . 2300.000 080 0.:- E..::.: 0.00 109 2300 89a 0800388 2i£0m 52 B 3 05 . SV .9503 2.039503” 88> 2 0 0o 080050 00:80: 05 00 ”80209050 0 and gamma 3 110 fiber/ matrix bonding is comparatively stronger for the KaowoolTM reinforced com- posites. Relatively few transversely oriented fibers and fewer longitudinally oriented fibers are seen on the fracture surface, an observation which confirms that fibers or lTM fiber/ matrix interfaces in Kaowoo composites do not play as significant a role in the fatigue crack initiation and fracture process as they do in SaffilTM composites. CHAPTER 5 Conclusions and Recommendations for Future Work The in-sz’tu fatigue studies performed on KaowoolTM and SaffilTM reinforced Al-Si alloy matrix composites using the specially designed SEM fatigue stage have allowed identification of fatigue crack initiation site(s) in these materials. The use of this fatigue stage allowed testing to be conducted under strain control with a range of 3200pe-3800ue and One as the lower strain limit, under room temperature conditions. In the KaowoolTM composites, multiple crack initiation sites were identified. Cracks from these sites were observed to grow extensively and coalesce into a single major crack leading to failure in these materials. Some of these sites were identified to be material processing defects such as “shot particles” and porosity. Cracks also initiated from microstructural features such as fiber/intermetallic interfaces and with a lower probability from intermetallics, fibers and matrix in that order. Cracks al- most always initiated from shot particles and porosity as early as five cycles of loading and continued to grow extensively into the matrix and joined with cracks from other locations until fracture occured after about 250,000 cycles of loading. The prema- ture crack initiation at shot particles motivated investigations in to the stress level at which shot particles initiate cracks. Tensile testing has been performed using dog- 111 112 bone type tensile specimens in-sz‘tu inside an SEM. From these observations, it was concluded that cracks initiate from shot particles in 15 vol.% KaowoolTM composites at stress levels as low as 23 ksi under tensile loading, which is lower than the the 0.2% yield strength of 27 ksi reported for this composite. In the case of SaffilTM composites, debonding occured at fiber/ matrix interfaces early in the fatigue cycling process. Cracks were seen along the fiber/ matrix interface of transversely oriented fibers. The specimens fractured due to link up of these transversely oriented cracks along the interfaces. Depending on the availability of favorably oriented fibers, crack propagation and failure of the specimen can occur as early as 2500 cycles of loading. Close observation of the fiber/matrix interfaces suggested the presence of a thin interfacial layer, which however, could not be detected by EPMA. The identification of this layer is beyond the scope of this work. Even though the composition of the fibers and the matrix was qualitatively similar, the relative contents of silica available at the fiber differ considerably and might lead to formation of different chemical constituents at the interface for the two systems. Silica in the fiber was shown to react more with magnesium in the matrix than alumina to develop a strong bond between the fiber and the matrix. The formation of 2MgO.Si02 due to this reaction acts as a blocking layer to prevent the diffusion of magnesium into the fiber. This variation in reaction products at the interface was seen to significantly influence the interface strength and hence the crack initiation mode in the composites used in this study. The following recommendations are made for future work on these materials to further understand the fatigue crack initiation mechanisms in these composites and develop them for useful automotive applications. Even though literature shows that reaction at the interface is limited to the binder and the magnesium in the matrix during T5 heat treatment in the composite systems under study, the presence of a strong bond in KaowoolTM and a weak bond in the 113 SaffilTM composites contradicts this observation. The interface between the fiber and matrix is seen to play an important role in the crack initiation in SaffilTM composites. The possible influence of silica content on the interfacial reaction products needs to be thoroughly understood to effectively tailor the interface in these composite systems. The importance of the fiber orientation apart from uniformity in their distribution has been stressed in Section 4.1. The orientation of the fibers at angles close to 5- 10" of the loading direction would allow uniform distribution of load between the fiber and the matrix. However, when the interface between the fiber and the matrix is weak, more fibers at angles 5-10° would lead to easy crack growth and fracture due to the link up of cracks along the fiber/matrix interfaces of the transversely oriented fibers. A process that allows control of the fiber orientation in the composite would significantly improve the mechanical performance of all or any discontinuously reinforced composites. In the KaowoolTM composites, the significance of crack initiation at shot particles was clearly characterized. From in-situ tensile testing experiments, the stress level at which these cracks occur at the shot particles was observed to be 23 ksi, which is lower than the nominal 0.2% yield strength of the composite . However, there is a significant variation in the behavior of these shot particles depending on their size, shape, wall thickness and depth. It would be useful to understand the influence of these parameters on the crack initiation at the shot particles. From such analysis, an effort can be made to eliminate only those shot particles that are detrimental to the performance of these composites. Finally, with the elimination of processing defects such as shot particles and poros- ity, cracks were seen to initiate from fiber/intermetallic interfaces. The presence of intermetallics on the surfaces of the fibers is a manifestation of the processing con- ditions used during composite fabrication. By using a faster fabrication technique, the preferential nucleation and segregation of the intermetallics on the fibers can be 114 avoided. Another solution would be use of a post processing treatment on the com- ponent such as laser surface processing where faster heat transfer occurs, there by preventing the intermetallic segregation on the fibers. APPENDICES APPENDIX A A.1 Composition and Mechanical Properties of Unreinforced Alloy and Composite Materials A.1.1 The composition of Al-Si matrix alloy (A339): Al-12.0%Si-1.0%Ni-1.0%Mg—2.25%Cu-1.2%Fe-O.3%Zn A.1.2 Material properties: Table A.1: Mechanical properties of the matrix alloy and composites (Reference: General Motors Research Laboratories) Material Yield strength Tensile strength Young’s modulus Matrix 140 MPa (20 ksi) 228 MPa (33 ksi) 73 GPa KaowoolTM fiber 120 GPa SaffilTM fiber 300 GPa 15% KaowoolTM composite 185 MPa (27 ksi) 80 GPa 15% SaffilTM composite 200 MPa (29 ksi) 300 MPa (44 ksi) 115 APPENDIX B B.1 Finite Element Analysis to Demonstrate the Stress/ Strain Distribution on the Surface of the Specimen B.1.1 Out-line of the Finite Element Analysis Work: The specimen was modeled using I-DEAS finite element analysis software and then the model was meshed with rectangular solid brick elements. A finer mesh was generated in the gage length region of the specimen as shown in Figure 8.1 by creating more nodes, to obtain stress variations with greater accuracy. The boundary conditions imposed were a three dimensional constraint at all nodes on one of the grip ends (towards the greater width of the taper) of the specimen and a uniform load of 5 lbf on the face of the other grip end (towards the smaller width of the taper). These conditions as shown in Figure 8.2 effectively simulate the static loading conditions on the specimen in the fatigue stage. The true material properties of the composite specimens were not readily available, so the default material properties of steel were used for analysis. The result from this analysis is presented as a strain distribution in Figure 33. Crack initiation can be expected to take place in the region at the center of the gage length where maximum strain occurs. 116 \\\\\\ “(fl \\\\\\\\\\\\\ \WMM W \\ 118 .0930 2.3.0.0 5 00030000 05 20.250 0. 00:00.0 0000:0000 b000=0m H Nd 0.5a:— 119 0088000 05 00 0000.50 05 00 0003506 Seam and 0.53m vanoo. Nonnoo. mwmfioo. vhmuoo. ownnoo. nvunoo. nvcfloo.o ux