DEVELOPMENT OF YIELD CRITERIA FOR DESCRIBING THE BEHAVIOR OF POROUS METALS WITH TENSION-COMPRESSION ASYMMETRY
By JOEL B. STEWART
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
1 c 2009 Joel B. Stewart
2 ACKNOWLEDGMENTS I would like to thank the members of my supervisory committee for their suggestions and support. I would like to especially thank Dr. Oana Cazacu for her support, her patience and her enthusiasm throughout this study. Her comments and direction have been invaluable to the completion of this research. I would also like to thank the Air Force Research Laboratory for providing the oppor- tunity to pursue a graduate education. Special recognition is extended to Dr. Lawrence Lijewski and Dr. Kirk Vanden for enthusiastically supporting my graduate research goals from the outset. I would also like to acknowledge the many helpful conversations and generous moral support received from Dr. Michael Nixon, Dr. Martin Schmidt and Dr. Brian Plunkett. Their steadfast support and honest critiques were always appreciated and a source of encouragement. I would like to extend special gratitude to Dr. Michael Nixon for reading through the dissertation and making a number of insightful comments and suggestions. Dr. Stefan Soare was of great assistance while he and I were studying for the qualifying exam; his helpful suggestions and engaging conversation were greatly appreciated. Finally, I would like to thank my friends and family who supported me throughout both this current research effort and earlier academic endeavors. Special appreciation is extended to my wife, Kelly, for supporting me throughout my academic journey. The process would have been much more painful without her unfailing support and encouragement.
3 TABLE OF CONTENTS page ACKNOWLEDGMENTS...... 3 LIST OF TABLES...... 7 LIST OF FIGURES...... 8 ABSTRACT...... 11
CHAPTER 1 INTRODUCTION...... 13 2 HOMOGENIZATION APPROACH...... 27 2.1 Kinematic Homogenization Approach of Hill and Mandel...... 28 2.2 Yield Criterion for the Matrix Material...... 30 2.3 Plastic Multiplier Rate Derivation...... 33 2.3.1 Plastic multiplier rate when J3 ≤ 0...... 35 2.3.2 Plastic multiplier rate when J3 ≥ 0...... 38 2.3.3 General plastic multiplier rate expression...... 41 3 PLASTIC POTENTIAL FOR HCP METALS WITH SPHERICAL VOIDS... 46 3.1 Limit Solutions...... 48 3.1.1 Zero porosity and equal yield strengths limiting cases...... 48 3.1.2 Exact solution for a hydrostatically-loaded hollow sphere...... 49 3.1.2.1 Strain-displacement relations...... 49 3.1.2.2 Strain compatibility...... 50 3.1.2.3 Equations of motion...... 51 3.1.2.4 Elastic constitutive relation: Hooke’s law...... 52 3.1.2.5 Ultimate pressure...... 52 3.2 Choice of Trial Velocity Field...... 57 3.3 Calculation of the Local Plastic Dissipation...... 62 3.4 Development of the Macroscopic Plastic Dissipation Expressions...... 66 4 NUMERICAL IMPLEMENTATION OF THE MATRIX YIELD CRITERION. 78 4.1 Return Mapping Procedure...... 78 4.2 First Derivatives...... 85 4.2.1 Isotropic CPB06 first derivatives: general loading...... 87 4.2.2 Isotropic CPB06 first derivatives: biaxial loading...... 88 4.3 Second Derivatives...... 89 4.3.1 Von mises second derivatives...... 89 4.3.2 Isotropic CPB06 second derivatives: general loading...... 92 4.3.3 Isotropic CPB06 second derivatives: biaxial loading...... 96
4 5 ASSESSMENT OF THE PROPOSED SPHERICAL VOID MODEL BY FI- NITE ELEMENT CALCULATIONS...... 97 5.1 Modeling Procedure...... 97 5.2 Finite Element Results...... 102 5.3 Concluding Remarks...... 104 6 PLASTIC POTENTIALS FOR HCP METALS WITH CYLINDRICAL VOIDS 122 6.1 Limit Solutions...... 123 6.1.1 Zero porosity and von mises material limiting cases...... 123 6.1.2 Analysis of a hydrostatically-loaded hollow cylinder...... 124 6.1.2.1 Strain-displacement relations...... 125 6.1.2.2 Strain compatibility...... 126 6.1.2.3 Equations of motion...... 127 6.1.2.4 Elastic constitutive relation: Hooke’s law...... 128 σ 6.1.2.5 Relation between J3 and Σm ...... 128 6.2 Choice of Trial Velocity Field...... 133 6.3 Parametric Representation of the Porous Aggregate Yield Locus for Ax- isymmetric Loading...... 137 6.3.1 β ≥ 1: the matrix yield strength in tension is greater than in com- pression...... 138 Σ 6.3.1.1 Σm > 0 and J3 < 0...... 138 6.3.2 Discussion...... 144 6.4 Proposed Closed-Form Expression for a Plane Strain Yield Criterion.... 145 6.4.1 Calculation of the local plastic dissipation...... 145 6.4.2 Development of the macroscopic plastic dissipation expressions... 148 7 ASSESSMENT OF THE PROPOSED CYLINDRICAL VOID MODEL BY FI- NITE ELEMENT CALCULATIONS...... 162 7.1 Modeling Procedure...... 162 7.2 Finite Element Results...... 166 7.3 Concluding Remarks...... 168 8 ANISOTROPIC PLASTIC POTENTIAL FOR HCP METALS CONTAINING SPHERICAL VOIDS...... 180 8.1 Kinematic Homogenization Approach of Hill and Mandel...... 181 8.2 Yield Criterion for the Matrix Material...... 183 8.3 Choice of Trial Velocity Field...... 187 8.4 Calculation of the Local Plastic Dissipation...... 188 8.5 Development of the Macroscopic Plastic Dissipation Expression...... 189 8.6 Assessment of the Proposed Anisotropic Criterion through Comparison with Finite Element Calculations...... 192 8.7 Concluding Remarks...... 197
5 9 CONCLUSIONS...... 210
APPENDIX A PARAMETRIC REPRESENTATION DERIVATION OF THE AXISYMMET- RIC YIELD LOCUS...... 213 A.1 General Form of Equations...... 213 A.1.1 Plastic multiplier rate branches...... 213 A.1.2 Macroscopic plastic dissipation and derivatives...... 217 A.2 β ≥ 1: The Matrix Yield Strength in Tension is Greater than in Compres- sion...... 220 Σ A.2.1 J3 < 0...... 220 A.2.1.1 Σm > 0...... 220 A.2.1.2 Σm < 0...... 221 Σ A.2.2 J3 > 0...... 223 A.2.2.1 Σm > 0...... 223 A.2.2.2 Σm < 0...... 226 A.3 β ≤ 1: The Matrix Yield Strength in Tension is Less than in Compression. 229 Σ A.3.1 J3 < 0...... 230 A.3.1.1 Σm > 0...... 230 A.3.1.2 Σm < 0...... 233 Σ A.3.2 J3 > 0...... 236 A.3.2.1 Σm > 0...... 236 A.3.2.2 Σm < 0...... 237 B RELATIONSHIP BETWEEN HILL48 AND CPB06 COEFFICIENTS..... 240 B.1 Determine The Hill48 Coefficients Given The CPB06 Coefficients..... 245 B.2 Determine The CPB06 Coefficients Given The Hill48 Coefficients..... 246 REFERENCES...... 247 BIOGRAPHICAL SKETCH...... 251
6 LIST OF TABLES Table page 2-1 k-relations...... 44 2-2 z-parameters...... 44
5-1 k = 0, J3 > 0 and f0 = 0.01 spherical void computational test matrix...... 108
5-2 k = 0, J3 < 0 and f0 = 0.01 spherical void computational test matrix...... 108
5-3 k = −0.3098, J3 > 0 and f0 = 0.01 spherical void computational test matrix... 108
5-4 k = −0.3098, J3 < 0 and f0 = 0.01 spherical void computational test matrix... 109
5-5 k = 0.3098, J3 > 0 and f0 = 0.01 spherical void computational test matrix.... 109
5-6 k = 0.3098, J3 < 0 and f0 = 0.01 spherical void computational test matrix.... 109
5-7 k = 0, J3 > 0 and f0 = 0.04 spherical void computational test matrix...... 111
5-8 k = 0, J3 < 0 and f0 = 0.04 spherical void computational test matrix...... 111
5-9 k = −0.3098, J3 > 0 and f0 = 0.04 spherical void computational test matrix... 111
5-10 k = −0.3098, J3 < 0 and f0 = 0.04 spherical void computational test matrix... 112
5-11 k = 0.3098, J3 > 0 and f0 = 0.04 spherical void computational test matrix.... 112
5-12 k = 0.3098, J3 < 0 and f0 = 0.04 spherical void computational test matrix.... 112
5-13 k = 0, J3 > 0 and f0 = 0.14 spherical void computational test matrix...... 114
5-14 k = 0, J3 < 0 and f0 = 0.14 spherical void computational test matrix...... 114
5-15 k = −0.3098, J3 > 0 and f0 = 0.14 spherical void computational test matrix... 114
5-16 k = −0.3098, J3 < 0 and f0 = 0.14 spherical void computational test matrix... 115
5-17 k = 0.3098, J3 > 0 and f0 = 0.14 spherical void computational test matrix.... 115
5-18 k = 0.3098, J3 < 0 and f0 = 0.14 spherical void computational test matrix.... 115 7-1 k = 0 plane strain cylindrical void computational test matrix...... 172 7-2 k = −0.3098 plane strain cylindrical void computational test matrix...... 172 7-3 k = 0.3098 plane strain cylindrical void computational test matrix...... 173 8-1 Transversely isotropic CPB06 constants...... 198
7 LIST OF FIGURES Figure page 1-1 Particle cracking example...... 24 1-2 Particle decohesion example...... 25 1-3 Void in single crystal...... 26 2-1 CPB06 yield condition π-plane representation...... 43 2-2 CPB06 plastic multiplier rate π-plane representation...... 45 3-1 Ductile crack in aluminum plate...... 74 3-2 Representative volume element for a sphere containing a spherical void...... 75 3-3 Hydrostatically-loaded hollow sphere...... 75 3-4 Macroscopic ductile yield surfaces with tension-compression asymmetry...... 76 3-5 Evolution of macroscopic ductile yield surfaces with porosity...... 77 5-1 Axisymmetric unit cell for the spherical void...... 106
5-2 f0 = 0.01 axisymmetric finite element mesh for the unit cell...... 107
5-3 f0 = 0.04 axisymmetric finite element mesh for the unit cell...... 110
5-4 f0 = 0.14 axisymmetric finite element mesh for the unit cell...... 113 5-5 k = 0 with f = 0.01: FE versus analytical...... 116 5-6 k = 0 with f = 0.01, f = 0.04 and f = 0.14: FE versus analytical...... 116 5-7 k = −0.3098 with f = 0.01: FE versus analytical...... 117 5-8 k = 0.3098 with f = 0.01: FE versus analytical...... 117 5-9 k = −0.3098 with f = 0.04: FE versus analytical...... 118 5-10 k = 0.3098 with f = 0.04: FE versus analytical...... 118 5-11 k = −0.3098 with f = 0.14: FE versus analytical...... 119 5-12 k = 0.3098 with f = 0.14: FE versus analytical...... 119 5-13 k = 0 axisymmetric FE data versus analytical yield curves...... 120 5-14 k = −0.3098 axisymmetric FE data versus analytical yield curves...... 120 5-15 k = 0.3098 axisymmetric FE data versus analytical yield curves...... 121
8 6-1 Voids and shear band in titanium...... 157 6-2 Cylindrical RVE...... 158
6-3 σT /σC = 1.21 yield curve parametric representation for one quadrant...... 159
6-4 σT /σC = 1.21 yield curve parametric representation...... 159
6-5 σT /σC = 0.82 yield curve parametric representation...... 160 6-6 f = 0.01 cylindrical macroscopic yield curves...... 160 6-7 f = 0.04 cylindrical macroscopic yield curves...... 161 6-8 f = 0.14 cylindrical macroscopic yield curves...... 161 7-1 Plane strain geometry used in finite element calculations...... 169 7-2 Valid triaxiality angles for plane strain...... 170
7-3 f0 = 0.01 plane strain finite element mesh for the unit cell...... 170
7-4 f0 = 0.04 plane strain finite element mesh for the unit cell...... 171
7-5 f0 = 0.14 plane strain finite element mesh for the unit cell...... 171 7-6 Plain strain yield point determination...... 173 7-7 k = 0, no fitting parameters: FE versus analytical...... 174 7-8 k = −0.3098, no fitting parameters: FE versus analytical...... 175 7-9 k = 0.3098, no fitting parameters: FE versus analytical...... 176 7-10 k = 0, with fitting parameters: FE versus analytical...... 177 7-11 k = −0.3098, no fitting parameters: FE versus analytical...... 178 7-12 k = 0.3098, no fitting parameters: FE versus analytical...... 179 8-1 Transversely isotropic RVE...... 198 8-2 Anisotropic effective stress versus effective strain curves...... 199
8-3 Plane stress yield loci for materials A, B and C with σT = σC ...... 199
8-4 Plane stress yield loci for materials A, B and C with σT < σC ...... 200
8-5 Plane stress yield loci for materials A, B and C with σT > σC ...... 200 8-6 k = 0 Material A deviatoric plot...... 201 8-7 k < 0 Material A deviatoric plot...... 201
9 8-8 k > 0 Material A deviatoric plot...... 202 8-9 k = 0 Material B deviatoric plot...... 202 8-10 k < 0 Material B deviatoric plot...... 203 8-11 k > 0 Material B deviatoric plot...... 203 8-12 k = 0 Material C deviatoric plot...... 204 8-13 k < 0 Material C deviatoric plot...... 204 8-14 k > 0 Material C deviatoric plot...... 205 8-15 Material A theoretical yield curves versus FE data...... 206 8-16 Material B theoretical yield curves versus FE data...... 207 8-17 Material C theoretical yield curves versus FE data...... 208 8-18 Anisotropic ductile yield surfaces with tension-compression asymmetry...... 209
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEVELOPMENT OF YIELD CRITERIA FOR DESCRIBING THE BEHAVIOR OF POROUS METALS WITH TENSION-COMPRESSION ASYMMETRY By Joel B. Stewart August 2009 Chair: Oana Cazacu Major: Mechanical Engineering A significant difference between the behaviors in tension versus compression is obtained at the polycrystal level if either twinning or non-Schmid effects are included in the description of the plastic deformation at the single crystal level. Examples of materials that exhibit tension-compression asymmetry include hexagonal close packed (HCP) polycrystals and intermetallics (e.g., molybdenum compounds). Despite recent progress in modeling the yield behavior of such materials, the description of damage by void growth remains a challenge. This dissertation is devoted to the development of macroscopic plastic potentials for porous metallic aggregates in which the void-free, or matrix, material displays tension- compression asymmetry. Using a homogenization approach, new analytical plastic potentials for a random distribution of voids are obtained. Both spherical and cylindrical void geometries are considered for void-matrix aggregates containing an isotropic matrix, while spherical voids are considered for the case of an anisotropic matrix material. The matrix plastic behavior in all cases is described by a yield criterion that captures strength differential effects and can account for the anisotropy that may be exhibited in the void- free material. For the case when the matrix material is isotropic, the developed analytical potentials for the void-matrix aggregate are sensitive to the second and third invariants of the stress deviator and display tension-compression asymmetry. Furthermore, if the matrix material
11 has the same yield strength in tension and compression, the developed criteria reduce to the classical Gurson criteria for either spherical or cylindrical voids. It has also been demonstrated that the developed isotropic criterion for porous aggregates containing spherical voids captures the exact solution of a hollow sphere loaded in hydrostatic tension or compression. Finite element cell calculations with the matrix material obeying an isotropic yield criterion and displaying tension-compression asymmetry were performed and the comparison between finite element calculations and theoretical predictions demonstrate the versatility of the proposed formulations. A new anisotropic potential for the porous aggregate was also developed for the case when the matrix material is anisotropic and displays tension-compression asymmetry. If the matrix is isotropic, the proposed analytical anisotropic criterion reduces to the isotropic criterion developed in this dissertation for a void-matrix aggregate containing spherical voids. Comparison between finite element calculations and theoretical pre- dictions show the predictive capabilities of the developed anisotropic formulation. The yield criteria developed in this dissertation are the only criteria available to capture the influence of damage by void growth in HCP metals and other materials that exhibit tension-compression asymmetry.
12 CHAPTER 1 INTRODUCTION The aim of this dissertation is to develop analytical plastic potentials for metals that exhibit tension-compression asymmetry (e.g., certain BCC materials such as molybdenum and HCP metals such as α-titanium) and that contain either cylindrical or spherical voids. Closed-form yield criteria for porous materials exhibiting tension-compression asymmetry (i.e., that have different yield strengths in tension versus compression) do not currently exist in the literature. The successful derivation of analytic yield criteria for these types of materials should give researchers an additional and more accurate tool for dealing with damage in these materials. Ductile failure in metals occurs due to the nucleation, growth and coalescence of voids (e.g., see McClintock, 1968; Rousselier, 1987). Additionally, some localization phenomena that commonly lead to failure in metallic structures (e.g., adiabatic shear band formation) are thought to be influenced by micro-voids in the matrix material (see, for example Batra and Lear, 2005; Tvergaard, 1980). The void distribution in a material can exist because of pre-existing voids (e.g., manufacturing defects) or because of nucleation at second- phase particles. For example, Figure 1-1 illustrates void nucleation due to cracking of the inclusion (as well as some decohesion at the inclusion-matrix interface). Figure 1-2 shows an example of void nucleation due to decohesion of the matrix material at the inclusions. Voids can also nucleate in single crystals that contain neither pre-existing voids nor inclusions (see, for example, Cuiti˜noand Ortiz, 1996; Lubarda et al., 2004). Cuiti˜noand Ortiz(1996) proposed that vacancy condensation can act as the nucleating mechanism in single crystals for low strain rates. Lubarda et al.(2004) found that vacancy condensation cannot account for void nucleation at the extreme conditions seen, for example, in laser- driven shock experiments (with stress pulse duration on the order of 10 ns) and proposed dislocation emission as an alternate nucleating mechanism at this higher strain rate
13 regime. Figure 1-3 shows an example from Lubarda et al.(2004) of a void that has formed in monocrystalline copper. There have been a great deal of experimental investigations to assess void evolution under loading and its influence on the load-carrying capacity. For example, Benzerga et al. (2004a) performed a number of experiments on medium carbon low alloy steel in order to investigate the role of plasticity and porous microstructure on this material’s failure. The steel being investigated exhibited anisotropic behavior (both due to anisotropic void evolution and plastic anisotropy) as well as mild tension-compression asymmetry. Both round bar specimens were used (to characterize the material’s anisotropy and investigate failure properties) and notched round bar specimens (to investigate the influence of stress triaxiality on failure). In these tests, the initial microstructure was characterized and interrupted tests were performed to track the evolution of porous properties including average values of void volume fraction, void aspect ratio and void spacing ratio. The experiments show void nucleation occurring at very low strain levels at MnS and oxide inclusions either by particle cracking followed by decohesion or debonding after cracking (depending on the loading direction). At a certain point, the micro voids were observed to coalesce due to micro-necking of the ligaments between neighboring voids. The specimens also exhibited anisotropic crack propagation (i.e., the propagation pattern depended on the loading direction). For all specimens, the void growth rate was found to be mainly confined to the center region of the specimen with the extensional void growth rate dominating at low triaxialities and the radial void growth rate becoming dominant with increasing stress triaxiality. The authors also compared their experimental data to theoretical results using finite element calculations (see Benzerga et al., 2004b, and the discussion of the model used therein later in this section). Because of the relationship between material porosity and ductile failure, the ability to accurately describe the evolution of voids in a ductile metal is crucial to being able to accurately predict the failure of the material. Unfortunately, computational constraints
14 make it prohibitively expensive to model each of the micro-voids in most engineering structures; therefore, the method of explicitly tracking the evolution of each micro-void is not practical at this time. An alternative to explicitly tracking the evolution of each void (i.e., tracking microscopic, or local, quantities) is to incorporate the effects of the micro-voids into the macroscopic, or average, properties (such as macroscopic stress, strain, yielding, etc.). Since the rate of dilatation of the porous solid is related to the void growth rate, plastic potentials for the porous solid must be developed in order to describe the void growth. The most widely used plastic potential for porous solids was proposed by Gurson(1977). For the case of a spherical void geometry, the unit cell or Representative Volume Element (RVE) considered was a spherical shell, while a cylindrical tube RVE was used for the cylindrical void analysis. To derive an analytic expression for the plastic potential (and, thus, for the yield criterion, assuming associated plasticity), Gurson performed a limit load analysis on the RVE. The work in this dissertation mainly follows the homogenization procedure outlined in Gurson(1977) where Gurson developed analytic yield criteria for ductile materials containing either spherical or cylindrical voids. In Gurson’s analysis, it was assumed that the virgin material (void-free) obeys the classical von Mises yield criterion. To obtain the plastic potential, minimization of the plastic energy was done for a specific velocity field compatible with uniform strain rate boundary conditions. Thus, the obtained criterion is an upper bound of the exact plastic potential (since the plastic energy was minimized for only one velocity field rather than for the complete set of kinematically admissible velocity fields). Gurson’s yield criterion is given for spherical voids as
2 S Σe 3Σm 2 ΦG = + 2f cosh − 1 − f = 0 (1–1) σY 2σY and for cylindrical voids as √ 2 ! C Σe 3Σγγ 2 ΦG = Ceqv + 2f cosh − 1 − f = 0 (1–2) σY 2σY
15 where (1 + 3f + 24f 6)2 for plane strain Ceqv = (1–3) 1 for axisymmetry.
In the previous expressions, Σe is the macroscopic von Mises effective stress, σY is the yield strength of the virgin material, f is the void volume fraction (sometimes referred to as the porosity), Σm is the macroscopic mean stress and Σγγ is the sum of the in-plane stresses (e.g., Σγγ = Σ11 + Σ22 if the 3-direction is the out-of-plane direction). Note that Gurson’s criteria depends not only on the second invariant of the stress deviator, but also on the pressure (or the mean stress) and on the level of porosity in the material; therefore, the yield function incorporates the influence of voids on the plastic deformation in a material. The main premise behind incorporating this void effect is that a volume of virgin material will behave differently under an applied load than will a volume of material that has a reduced load-bearing area due to the presence of voids and their subsequent evolution (growth and coalescence). Gurson’s spherical void criterion is considered most often in the literature but the cylindrical void criterion is applicable to certain problems as well (e.g., the plane stress analysis of sheet metal). The void volume fraction, f, (the ratio of the void volume to the total volume) evolves both from the nucleation at second-phase particles and the growth of existing voids (see, for example, Chu and Needleman, 1980; Lemaitre and Desmorat, 2005) such that
˙ ˙ ˙ f = fgrowth + fnucleation (1–4) where the rate of change due to growth is determined from the assumption of plastic incompressibility as ˙ P fgrowth = (1 − f) dkk (1–5)
16 P with dij being the plastic part of the rate of deformation tensor. The void volume frac- tion’s rate of change due to nucleation is generally given as
B f˙ = A σ˙ + σ σ˙ (1–6) nucleation σ Y 3 kk
for stress-controlled nucleation, or as
˙ fnucleation = AN ¯˙ (1–7)
for plastic strain-controlled nucleation, where ¯ is obtained from the following relation:
P σijdij = (1 − f) σY .¯˙ (1–8)
In the case of plastic strain-controlled nucleation, the following statistical expression has been frequently used: " 2# fN 1 ¯− N AN = √ exp − (1–9) sN 2π 2 sN
where fN is the volume fraction of void-nucleating particles, N is the mean nucleation
strain and sN is the standard deviation. A modified version of Gurson’s spherical criterion was used in Spitzig et al.(1988) to compare with experimental tests on compacted iron specimens. The modified criterion can be written as Σ 2 3mΣ Φ = e + 2f m cosh m − 1 − f 2m = 0 (1–10) σ¯ 2¯σ whereσ ¯ is the flow stress and m is a material coefficient representing strain hardening (a power law matrix model was used in the paper to compare with the experimental data). A widely used modification of Gurson’s spherical yield criterion was suggested in Tvergaard(1981) and Tvergaard(1982) based on comparisons with finite element calculations of shear band instabilities (where the instability is determined by a loss of ellipticity of the governing equations). With finite element calculations, the minimization of the plastic energy is done over a larger set of kinematically admissible velocity fields than in Gurson’s analysis (in which a single kinematically admissible velocity field is
17 assumed); therefore, adjustments to Gurson’s yield criteria can be proposed based on these calculations. In Tvergaard(1982), the finite element calculations involved a cylindrical cell containing a spherical void which was compared with the modified Gurson spherical yield criterion (i.e., compared with the modified form proposed in Tvergaard, 1981). Specifically, these calculations were meant to represent a periodic array of spherical voids which were arranged such that hexagonal representative volume elements (RVEs) could be fit together to form the structure (the cylindrical RVE used in the calculations was an approximation of the hexagonal RVE). Based on these axisymmetric finite element calculations, the following modified form of Gurson’s yield criterion was suggested:
2 Σe 3Σm 2 Φ = + 2fq1 cosh q2 − 1 − q3f = 0 (1–11) σY 2σY
where the qi are the fitting parameters introduced by Tvergaard (all equal to one in
Gurson’s original expression). Tvergaard recommended values of q1 = 1.5, q2 = 1 and
2 q3 = q1 based on the finite element results. The introduction of these fitting parameters
(q1, q2 and q3) can be thought of as a necessary adjustment of the yield surface to account for the influence of neighboring voids. Leblond and Perrin(1990) used a self-consistent analysis of a void within a porous plastic sphere (thus explicitly accounting for two different void geometries and their interaction) to recommend values for Tvergaard’s
2 fitting parameters as q1 = 4/e ≈ 1.47, q2 = 1 and q3 = q1. Tvergaard and Needleman(1984) further modified Gurson’s spherical yield criterion to account for the onset of void coalescence leading to final material fracture. The authors used this modified yield criterion in both numerical and finite element calculations to compare with experimental data of a copper rod fracturing under uniaxial tension (exhibiting cup-cone fracture). This final modification is generally referred to as the GTN criterion (after the three authors) in the literature and is what is typically found in the
18 finite element codes. The GTN criterion is given below:
2 Σe ∗ 3Σm ∗ 2 Φ = + 2f q1 cosh q2 − 1 − q3 (f ) = 0 (1–12) σY 2σY
where f ∗ is the effective void volume fraction which is a function of the actual void volume fraction, f, and represents the modification of the yield criterion to account for final material failure. The effective void volume fraction is given as f for f ≤ fC ∗ f (f) = ∗U (1–13) f − fC fC + (f − fC ) for f > fC fF − fC where fC is the critical void volume fraction of a material, fF is the void volume fraction
∗U ∗ at final failure and f = f (fF ) is the ultimate value of the effective void volume fraction (i.e., the effective void volume fraction at which the macroscopic stress carrying capacity
∗U 2 vanishes) such that f = 1/q1 if q3 = q1 is used as suggested by Tvergaard. Richelsen and Tvergaard(1994) performed three-dimensional unit cell finite element calculations with proportional loading (i.e., the applied macroscopic stresses on the boundaries of the unit cell were constant multiples of each other for the duration of the calculation) to compare with the axisymmetric model of Equation (1–12) (note that Gurson’s spherical model is an axisymmetric model since the geometry of the RVE and the assumed velocity field were both axisymmetric in the limit load analysis). For equal applied transverse stresses, the
GTN model using q1 = 1.5 was found to agree well with the unit cell calculations. Unit cell calculations with unequal applied transverse stresses also showed good agreement with the GTN model; specifically, the unit cell results for largely unequal transverse stresses
were found to lie between the theoretical results of q1 = 1.5 and q1 = 1.0 in the GTN model. In many cases, the micro-voids that form within a loaded material are ellipsoidal rather than spherical due to the influences of asymmetrical loading and/or an anisotropic
19 microstructure. Gologanu et al.(1993) used a homogenization procedure on an axisym- metric, ellipsoidal RVE containing a confocal, prolate ellipsoidal void to arrive at an analytic expression for prolate, ellipsoidal voids rather than for spherical voids. The general expression is given as
2 Σe κA :Σ 2 Φ = + 2fq1 cosh − 1 − q3f = 0 (1–14) σY σY where 1 √ ln (e /e )−1 κ = √ + 3 − 2 1 2 (1–15) 3 ln f and the void anisotropy tensor, A is given as
A = α (e2)(~ex ⊗ ~ex + ~ey ⊗ ~ey) + [1 − 2α (e2)]~ez ⊗ ~ez. (1–16)
In the previous equations, e1 and e2 are the eccentricities of the ellipsoidal void and
RVE, respectively, and should not be confused with the Cartesian basis vectors ~ex, ~ey
and ~ez. The evolution of the eccentricities is governed by the evolution of the void shape
parameter, S = ln(a1/b1), where a1 and b1 are the major and minor semi-axes of the void, respectively. The function, α, is given as
2 1 1 − ei −1 α (ei) = 2 − 3 tanh (ei). (1–17) 2ei 2ei
Gologanu showed that the model reduced to Gurson’s spherical criterion in the case of
a spherical void (e1 = e2 = 0) and to Gurson’s cylindrical criterion in the case of a
cylindrical void (e1 = e2 = 1). The formulas given in this paragraph are for Gologanu’s prolate ellipsoidal void analysis; Gologanu developed similar expressions for oblate voids (see, for example, Gologanu et al., 2001). Garajeu et al.(2000) extended the work of Gologanu by investigating the influence of void distribution evolution in addition to the influence of void shape evolution. Liao et al.(1997) developed a Gurson-type criteria for transversely isotropic metal sheets under plane stress conditions using Hill’s 1948 yield criterion (see Hill, 1948, 1950)
20 for the matrix material. The proposed model is as follows: s ! Σ 2 1 + 2R 3Σ Φ = e + 2f cosh m − 1 − f 2 = 0 (1–18) σY 2 (1 + R) 2σY
where the anisotropy parameter, R, is defined as “the ratio of the transverse plastic strain rate to the through-thickness plastic strain rate under in-plane uniaxial loading conditions.” Note that the previous model accounts for porosity in transversely isotropic materials and reduces to Gurson’s cylindrical criterion for isotropy (R = 1). Benzerga and Besson(2001) also developed a Gurson-type criteria, intended for orthotropic porous metals, by assuming a matrix material that could be characterized using Hill’s 1948 criterion. The model is given as
0 0 3Σ : H :Σ 3Σm 2 Φ = 2 + 2fq1 cosh q2 − 1 − q3f = 0 (1–19) 2σY hσY where h is called the anisotropy factor (h = 2 for isotropy) and is a function of the macroscopic anisotropy tensor, H. Equation (1–19) differs from Equation (1–18) in that it assumes spherical rather than cylindrical voids and applies to the more general case of orthotropic materials versus transversely isotropic materials. Besson and Guillemer-Neel(2003) extended the GTN model of Equation (1–12) to include mixed isotropic and kinematic hardening within a thermodynamical framework. State variables related to isotropic and kinematic hardening are employed in the model and are evolved in a thermodynamically-consistent manner using a dissipation potential (see Lemaitre and Chaboche, 1990). The extended model can be written as
2 Σe ∗ 3Σm ∗ 2 Φ = + 2f q1 cosh q2 − 1 − q3 (f ) = 0 (1–20) σ∗ 2σ∗
where σ∗ is an effective scalar stress and is a function of the state variables related to isotropic and kinematic hardening. Besson and Guillemer-Neel obtained good agreement between the previous model and finite element calculations of cyclically-loaded unit cells.
21 Benzerga et al.(2004b) used a yield criterion which combines many of the properties of Equations (1–14) and (1–19) in order to compare with the experimental results given in Benzerga et al.(2004a). The yield criterion, prior to void coalescence, used in the paper is as follows: 0 0 3Σ : H :Σ κA :Σ 2 Φ = 2 + 2fqw cosh − 1 − qwf = 0 (1–21) 2σ∗ hσ∗ where qw = 1 + (q1 − 1)/ cosh(S) was introduced by Gologanu (see, for example, Gologanu et al., 2001). In this yield criterion, σ∗ replaces σY in the denominators to account for hardening as in Equation (1–20). The authors use a different criterion to account for yield after coalescence (this approach can be viewed as analogous to the f ∗ parameter in the GTN model; an accurate ductile yield criterion must consider void coalescence separately from initial void growth). This post-coalescence criterion is given as q 3 Σ0 : H :Σ0 2 1 |3Σm| 3 2 Φ = + − 1 − χ Ψf (χ, S) = 0 (1–22) σ∗ 2 σ∗ 2
where χ is the ligament size ratio (related to the void spacing) and Ψf is a function of both χ and the void shape parameter, S. The authors showed good agreement between the experimental results obtained in Benzerga et al.(2004a) and the yield criteria given in this paragraph. Researchers at Los Alamos National Laboratory (LANL) have extended Gurson’s model (see, for example, Addesio and Johnson, 1993; Bronkhorst et al., 2006; Maudlin et al., 1999) to solve computational problems involving highly dynamic material behavior (i.e., high strain rates and high temperatures). This modified model is referred to as TEPLA and uses a Mie-Gruneisen equation of state to specify the pressure along with a mechanical threshold strength (MTS) model to account for rate and temperature effects. The above works and many others have done an enormous amount of research aimed at extending Gurson’s analysis both to more complex materials as well as to more complex void shapes and distributions. It is worth noting that all of the models mentioned above are applicable to metallic materials with cubic structure; i.e., the models were developed
22 for face-centered-cubic metals (e.g., aluminum) and body-centered-cubic metals (e.g., steel). The aim of this research is to develop models to describe the yielding and failure of hexagonal-closed-packed metals such as α-titanium, zirconium, and uranium (i.e., materials that have different yield strengths in tension versus compression). The outline of this dissertation is as follows. The kinematic homogenization approach of Hill-Mandel (Hill, 1967; Mandel, 1972) that is used to develop the analytical plastic potentials of the void-matrix aggregate is described in Chapter2. Chapter3 details the development of a closed-form plastic potential for an isotropic porous aggregate containing spherical voids when the matrix displays tension-compression asymmetry. The return mapping algorithms used in implementing the matrix, or void-free, plastic potential into a finite element code is outlined in Chapter4 and the necessary first and second derivatives are given for the chosen matrix plastic potential of Cazacu et al.(2006), which can account for tension-compression asymmetry in the void-free material. In Chapter 5, comparisons are made between the developed isotropic porous plastic potential for spherical voids and finite element unit cell calculations. The finite element calculations are designed such that a spherical void is explicitly meshed inside an axisymmetric cylinder whose response is governed by the Cazacu et al.(2006) plastic potential. Macroscopic plastic potentials for void-matrix aggregates containing cylindrical voids when the matrix displays tension-compression asymmetry are developed in Chapter6 and a proposed plane strain plastic potential is compared to finite element unit cell calculations in Chapter7. Finally, the isotropic plastic potential developed in Chapter3 for a void-matrix aggregate containing spherical voids and a matrix exhibiting tension-compression asymmetry is extended in Chapter8 to include the effects of matrix anisotropy on the yielding of the void-matrix aggregate. The developed anisotropic macroscopic plastic potential is compared to transversely isotropic unit cell finite element calculations.
23 Figure 1-1. Void nucleation at MnS inclusions in steel. (a) and (b) Longitudinal loading. (c) and (d) Transverse loading. (e) Decohesion at the poles of a tiny spherical MnS particle close to a grain boundary and 100 lm ahead of the crack tip (2% Nital solution etched). Field emission SEM imaging using back-scattered electrons in (a)(c) and (e) and secondary electrons with an in-lens detector in (d) to reveal the crack. [Reprinted with permission. Benzerga, Besson, and Pineau(2004a). Anisotropic ductile fracture. Part I: Experiments. Acta Materialia, 52 , 4623-4638.]
24 Figure 1-2. Reconstructed images of the same internal section at five different relevant deformation steps of a metal composite consisting of an aluminum matrix with embedded spherical ceramic particles (a) initial state; (b) = 0.065; (c) = 0.15; (d) = 0.35-0.48; (e) = 0.51-0.81. Detail (A) pre-existing holes induced by the extrusion process which start to grow during tension. Detail (B) particle/matrix decohesion during the tensile loading. Detail (C) coalescence. The tensile direction is vertical in the figure. The various true strain values on (d) and (e) convey the presence of necking in the sample. [Reprinted with permission. Babout, Maire, Buffi`ere,and Foug`eres(2001). Characterization by X-ray computed tomography of decohesion, porosity growth and coalescence in model metal matrix composites. Acta Materialia, 49, 2055-2063.]
25 Figure 1-3. TEM micrograph of laser-shocked monocrystalline copper: dark field image of an isolated void near the rear surface of the specimen and associated work-hardened layer (white rim). [Reprinted with permission. Lubarda, Schneider, Kalantar, Remington, and Meyers(2004). Void growth by dislocation emission. Acta Materialia, 52, 1397-1408.]
26 CHAPTER 2 HOMOGENIZATION APPROACH All engineering metals and alloys contain inclusions and second-phase particles at which micro-voids nucleate either by decohesion of the particle-matrix interface or by particle breaking. Subsequently, voids grow due to plastic deformation of the surrounding material until a localized internal necking of the intervoid matrix occurs that leads to the formation of some macroscopic crack. Investigations of the expansion of voids of cylindrical and spherical geometries in rigid ideal plastic matrices by McClintock(1968) and Rice and Tracey(1969) have established the effects of stress state (stress triaxiality) on the void growth rate. Gurson(1977) proposed approximate yield criteria and flow rules for ductile materials containing spherical or cylindrical cavities using an upper-bound approach. To better account for the interaction between voids, several modifications of Gurson’s criterion have been proposed based on results of two-dimensional finite-element studies (e.g., Koplik and Needleman, 1988; Tvergaard, 1981; Tvergaard and Needleman, 1984) or from rigorous estimates of the exact macroscopic potentials (for a review of those alternative approaches pioneered by Talbot and Willis 1985, see for example, Garajeu and Suquet 1997; Leblond et al. 1994). Recently, analytical criteria that account for the combined effects of void shape and matrix anisotropy on the macroscopic response of ductile porous solids were proposed (see for example Benzerga and Besson, 2001; Monchiet et al., 2008). In all the models mentioned it is assumed that the matrix has the same yield in tension and compression. However, in the absence of voids, some cubic materials such as high strength steels or molybdenum exhibit tension-compression asymmetry. This strength-differential (S-D) effect is a consequence of crystal slip that does not obey the well-known Schmid law (see for example Vitek et al., 2004). Also, twinning at the single crystal level may result in a strong tension-compression asymmetry at the aggregate level in some cubic metals (see Hosford and Allen, 1973). Hexagonal close packed (HCP)
27 metals exhibit tension-compression asymmetry as a result of twinning activation at the single crystal level. HCP metals can deform either by slip or twinning, with twinning becoming increasingly prominent with increasing strain rate. If a metal deforms by slip alone (a reversible shear mechanism), irreversible flow depends only on the magnitude of the resolved shear stress such that the strengths in tension and compression are equal; however, if twinning also exists as a deformation mechanism then a difference between the strengths in tension and compression will exist. This chapter is organized as follows. Section 2.1 introduces the homogenization approach due to Hill and Mandel (Hill, 1967; Mandel, 1972) that is used in developing the macroscopic plastic potentials for the void-matrix aggregates considered in this dissertation. Section 2.2 presents the isotropic version of the Cazacu et al.(2006) plastic potential that is used to describe the matrix material in all analyses except for those discussed in Chapter8 (which focuses on anisotropic plastic potentials). Section 2.3 presents the derivation of the plastic multiplier rate associated to the isotropic version of the Cazacu et al.(2006) plastic potential. The derivation of this plastic multiplier rate is the main challenge in determining the local plastic dissipation in the matrix. 2.1 Kinematic Homogenization Approach of Hill and Mandel
The current work will consider both a spherical and a cylindrical representative vol- ume element (RVE) containing a void of similar geometry (i.e., spherical and cylindrical, respectively). These RVEs were chosen both because the void geometries considered are typically seen in experiments and because the symmetry of the void and outer surface greatly simplifies the homogenization analysis. If the matrix material in the RVE is as- sumed to be rigid plastic, then any volume change is due exclusively to void evolution such that the rate of void growth is easily obtained from the rate of dilatation (see Equation (1–5)). Consider a representative volume element V , composed of a homogeneous rigid- plastic matrix and a traction-free void. The matrix material is described by a convex yield
28 function ϕ(σ) in the stress space and an associated flow rule
∂ϕ d = λ˙ , (2–1) ∂σ where σ is the Cauchy stress tensor, d = (1/2)(∇v +∇vT ) denotes the rate of deformation tensor with v being the velocity field, and λ˙ ≥ 0 stands for the plastic multiplier rate. The yield surface is defined as ϕ(σ) = 0. Let C denote the convex domain delimited by the yield surface such that C = {σ|ϕ(σ) ≤ 0} . (2–2)
The plastic dissipation potential of the matrix is defined as
w(d) = sup (σ : d) (2–3) σ∈C where “:” denotes the tensor double contraction. Uniform rate of deformation boundary conditions are assumed on the boundary of the RVE, ∂V , such that
v = Dx for any x ∈ ∂V (2–4) with D, the macroscopic rate of deformation tensor, being constant. For the boundary conditions of Equation (2–4), the Hill-Mandel (Hill, 1967; Mandel, 1972) lemma applies; hence,
hσ : diV = Σ : D, (2–5) where h i denotes the average value over the representative volume V , and Σ = hσiV . Furthermore, there exists a macroscopic plastic dissipation potential W (D) such that
∂W (D) Σ = (2–6) ∂D with
W (D) = inf hw(d)iV , (2–7) d∈K(D)
29 where K(D) is the set of incompressible velocity fields satisfying Equation (2–4) (for more details see Gologanu et al., 1997; Leblond, 2003). The matrix material being considered obeys the isotropic version of the pressure-insensitive yield criterion that captures strength differential effects of Cazacu et al.(2006). 2.2 Yield Criterion for the Matrix Material
Twinning and martensitic shear are directional deformation mechanisms and, if they occur, yielding will depend on the sign of the stress (see Hosford, 1993). Early polycrystalline simulations results by Hosford and Allen(1973) who analyzed deformation by twinning in random FCC polycrystals, predicted a yield stress in uniaxial tension 22% lower than that in uniaxial compression. Based on these simulations and more recent results concerning the effects of non-Schmid type yield criteria at the single-crystal level on the polycrystalline response (e.g., see Vitek et al., 2004), it can be concluded that yield loci with a strong asymmetry between tension and compression should be expected in any isotropic pressure-insensitive material that deforms either by twinning or directional slip. To account for strength differential effects in pressure insensitive materials, Plunkett (2005) and Cazacu et al.(2006) proposed the following isotropic form for yielding:
2 2 2 F = (|s1| − ks1) + (|s2| − ks2) + (|s3| − ks3) (2–8)
where si are the principal values of the Cauchy stress deviator. In Equation (2–8), k takes into account the tension-compression asymmetry and is given by
1 − h (σ , σ ) k = T C (2–9) 1 + h (σT , σC )
where v u 2 u 2 − σT u σC h (σ , σ ) = u (2–10) T C t 2 2 σT − 1 σC in the isotropic case. The terms σT and σC in the previous expressions are the yield strengths in tension and compression, respectively. The current work will sometimes refer
30 to Equation (2–8) along with the definitions in Equations (2–9) and (2–10) as the “CPB06 isotropic criterion” for notational convenience.
Let (e1, e2, e3) be a Cartesian coordinate system associated with the principal directions of the stress tensor. Due to the tension-compression asymmetry, the projection of the yield surface in Equation (2–8) in the deviatoric plane (the plane with normal n =
√1 e + √1 e + √1 e ) has threefold symmetry. Let f be the projections of the eigenvectors 3 1 3 2 3 3 i ei, i = 1...3 on the deviatoric plane. As an example, Figure 2-1 illustrates deviatoric
π-plane representations of the yield curves given by Equation (2–8) for σT /σC = 0.82 and σT /σC = 1.21, along with the von Mises yield locus for comparison. Note a very drastic departure of the yield locus from the Von Mises circle. This departure is due to a strong influence of the third invariant of the stress deviator on yielding, which results from tension-compression asymmetry. In the π-plane representation, the radial coordinate is related to the second invariant of the stress deviator while the angular coordinate is related to the third invariant of the stress deviator. Therefore, the von Mises circle is independent of the third invariant, while the CPB06 criterion for non-zero k depends on both the radial and angular coordinates (i.e., it is not a circle) such that it is dependent on both the second and third invariants of the stress deviator. The yield function (a.k.a. the stress potential), ϕ, is given as follows:
ϕ = φ − Y where φ is a scalar effective stress associated to Equation (2–8) and Y represents the material’s hardening condition. If Y is taken to be the yield strength in tension, σT , then the previous equation yields
ϕ (sα, k, σT ) = σe (sα, k) − σT (2–11)
where σe is now the effective stress formulated such that it reduces to the yield strength in tension for uniaxial loading. The explicit expression for this effective stress may be found
31 by writing Equation (2–8) for the uniaxial tension case and forcing σe equal to σT . The resulting expression is given as √ σe = m F v u 3 (2–12) uX 2 = mt (|si| − ksi) i=1 where s 9 1 m := . (2–13) 2 3k2 − 2k + 3 p Notice that for k = 0, m = 3/2 and σe simply becomes the von Mises effective stress. Also, note that for the uniaxial compression case,
r3k2 + 2k + 3 σ = σ e C 3k2 − 2k + 3 σT = σC σC
= σT such that the yield surface of Equation (2–11) is indeed satisfied. Since the Cazacu et al.(2006) criterion is homogeneous of degree one, Equation (2–1) yields (assuming rigid plastic behavior in the matrix)