Sie sind auf Seite 1von 22

Computational Material Models

Lecture notes - course 4K620


dr.ir. P.J.G. Schreurs
Eindhoven University of Technology
Department of Mechanical Engineering
Materials Technology
April 19, 2011
Contents
1 Introduction 1
2 Truss structures 3
2.1 Homogeneous truss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Elongation and contraction . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.2 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Linear deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Linear strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2 Linear elastic behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.4 Solution procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Nonlinear deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Strains for large elongation . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.2 Mechanical power for an axially loaded truss . . . . . . . . . . . . . . 11
2.3.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.4 Solution procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Finite element method for linear truss . . . . . . . . . . . . . . . . . . . . . . 20
2.4.1 FE program tr2dL . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.5 Weighted residual formulation for nonlinear truss . . . . . . . . . . . . . . . . 28
2.6 Finite element method for nonlinear truss . . . . . . . . . . . . . . . . . . . . 31
2.6.1 FE program tr2d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3 One-dimensional material models 39
3.1 Material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.1 Time history plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.2 Tensile curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.1.3 Discrete material models . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2 Elastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.1 Elastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2.2 Hyper-elastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2.3 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.4 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 Elastoplastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.1 Elastoplastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3.2 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
I
II
3.3.3 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.4 Linear viscoelastic material behavior . . . . . . . . . . . . . . . . . . . . . . . 65
3.4.1 Viscoelastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.4.2 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.3 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.4.4 Viscoelastic : dierential formulation . . . . . . . . . . . . . . . . . . . 74
3.4.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.5 Creep behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.5.1 Creep models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.5.2 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.5.3 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.5.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.6 Viscoplastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.6.1 Viscoplastic (Perzyna) model . . . . . . . . . . . . . . . . . . . . . . . 86
3.6.2 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.6.3 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.6.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.7 Nonlinear viscoelastic material behavior . . . . . . . . . . . . . . . . . . . . . 93
3.7.1 Nonlinear viscoelastic model . . . . . . . . . . . . . . . . . . . . . . . 94
3.7.2 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.7.3 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.7.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4 Vectors, tensors, columns, matrices 105
4.1 Summary of vector and tensor operations . . . . . . . . . . . . . . . . . . . . 105
4.1.1 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.1.2 Second-order tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.1.3 Fourth-order tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.2 Column and matrix notation . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.2.1 Matrix/column notation for second-order tensor . . . . . . . . . . . . 107
4.2.2 Matrix notation of fourth-order tensor . . . . . . . . . . . . . . . . . . 109
4.2.3 Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5 Kinematics 117
5.1 Material coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.2 Position vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.3 Lagrange - Euler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.4 Time derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.5 Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.5.1 Deformation tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.5.2 Volume change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.5.3 Area change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.5.4 Elongation and elongational strain . . . . . . . . . . . . . . . . . . . . 123
5.5.5 Shear and shear strain . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.5.6 Right Cauchy-Green deformation tensor . . . . . . . . . . . . . . . . . 125
5.5.7 Right stretch tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
III
5.5.8 Rotation tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.5.9 Right polar decomposition . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.5.10 Strain tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.6 Deformation rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.6.1 Spin tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.6.2 Deformation rate tensor . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.6.3 Elongation rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.6.4 Volume change rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.6.5 Area change rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.7 Special deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.7.1 Inverse deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.7.2 Planar deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.7.3 Axi-symmetric deformation . . . . . . . . . . . . . . . . . . . . . . . . 133
5.7.4 Homogeneous deformation . . . . . . . . . . . . . . . . . . . . . . . . . 134
5.8 Linear deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.8.1 Elongational strain and shear strain . . . . . . . . . . . . . . . . . . . 136
5.8.2 Principal strains and directions . . . . . . . . . . . . . . . . . . . . . . 137
5.8.3 Linear strain : Cartesian components . . . . . . . . . . . . . . . . . . 137
5.8.4 Linear strain : cylindrical components . . . . . . . . . . . . . . . . . . 140
6 Stresses 145
6.1 Stress vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.2 Cauchy stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.3 Principal stresses and directions . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.3.1 Stress transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.3.2 Mohrs circles of stress . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.4 Resulting force on arbitrary material volume . . . . . . . . . . . . . . . . . . 151
6.5 Resulting moment on arbitrary material volume . . . . . . . . . . . . . . . . . 152
6.6 Special stress states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
7 Balance or conservation laws 157
7.1 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.2 Balance of momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.3 Balance of moment of momentum . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.4 Equilibrium equations in components . . . . . . . . . . . . . . . . . . . . . . . 161
7.4.1 Cartesian components . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.4.2 Cylindrical components . . . . . . . . . . . . . . . . . . . . . . . . . . 163
7.4.3 Special equilibrium states . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.5 Balance of energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
7.5.1 Mechanical energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
7.5.2 Thermal energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.5.3 Kinetic energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.5.4 Internal energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.5.5 Energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.5.6 Energy equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.5.7 Mechanical power for three-dimensional deformation . . . . . . . . . . 169
IV
8 Constitutive equations 171
8.1 Equations and unknowns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.2 General constitutive equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.2.1 Locality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.2.2 Frame indierence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.3 Invariant stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.4 Invariant stress tensors and their rates . . . . . . . . . . . . . . . . . . . . . . 176
9 Linear elastic material 177
9.1 Material symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9.1.1 Monoclinic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9.1.2 Orthotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9.1.3 Quadratic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
9.1.4 Transversal isotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
9.1.5 Cubic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.1.6 Isotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.2 Planar deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.2.1 Plane strain and plane stress . . . . . . . . . . . . . . . . . . . . . . . 183
9.3 Engineering parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.3.1 Isotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.3.2 Compliance and stiness matrix . . . . . . . . . . . . . . . . . . . . . 186
9.4 Isotropic material tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.4.1 Column/matrix notation of Hookes law . . . . . . . . . . . . . . . . . 187
9.5 Thermo-elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
9.5.1 Isotropic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
9.5.2 Plane strain/stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
10 Elastic limit criteria 191
10.1 Yield function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
10.2 Principal stress space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
10.3 Yield criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
10.3.1 Maximum stress/strain . . . . . . . . . . . . . . . . . . . . . . . . . . 194
10.3.2 Maximum principal stress (Rankine) . . . . . . . . . . . . . . . . . . . 195
10.3.3 Maximum principal strain (Saint Venant) . . . . . . . . . . . . . . . . 196
10.3.4 Tresca . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.3.5 Von Mises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
10.3.6 Beltrami-Haigh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
10.3.7 Mohr-Coulomb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
10.3.8 Drucker-Prager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
10.3.9 Other yield criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
11 Solution strategies 207
11.1 Governing equations for unknowns . . . . . . . . . . . . . . . . . . . . . . . . 207
11.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
11.2.1 Saint-Venants principle . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.2.2 Superposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11.3 Solution : displacement method . . . . . . . . . . . . . . . . . . . . . . . . . . 209
V
11.3.1 Navier equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
11.3.2 Axi-symmetric with ut = 0 . . . . . . . . . . . . . . . . . . . . . . . . 210
11.4 Solution : stress method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
11.4.1 Beltrami-Mitchell equation . . . . . . . . . . . . . . . . . . . . . . . . 211
11.4.2 Beltrami-Mitchell equation for thermal loading . . . . . . . . . . . . . 212
11.4.3 Airy stress function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
11.5 Weighted residual formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
11.5.1 Weighted residual formulation for linear deformation . . . . . . . . . . 215
11.5.2 Total Lagrange formulation . . . . . . . . . . . . . . . . . . . . . . . . 215
11.5.3 Updated Lagrange formulation . . . . . . . . . . . . . . . . . . . . . . 218
11.6 Finite element method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
11.6.1 FE program plaxL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
11.6.2 FE program plax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
12 Three-dimensional material models 227
12.1 Elastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
12.1.1 Isotropic elastic material . . . . . . . . . . . . . . . . . . . . . . . . . . 228
12.1.2 Hyper-elastic material . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
12.1.3 Incompressible elastic material . . . . . . . . . . . . . . . . . . . . . . 232
12.1.4 Incremental analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
12.1.5 Linear P-E model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
12.1.6 Linear -A model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
12.1.7 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
12.2 Elastoplastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 243
12.2.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
12.2.2 Constitutive relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
12.2.3 Constitutive model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
12.2.4 Incremental analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
12.2.5 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
12.2.6 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
12.3 Linear viscoelastic material behavior . . . . . . . . . . . . . . . . . . . . . . . 253
12.3.1 Constitutive relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
12.3.2 Incremental analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.3.3 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
12.3.4 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
12.3.5 Isotropic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.3.6 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
12.4 Viscoplastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
12.4.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
12.4.2 Constitutive relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
12.4.3 Constitutive model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
12.4.4 Incremental analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
12.4.5 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
12.4.6 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
12.4.7 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
12.4.8 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
12.4.9 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
VI
12.5 Nonlinear viscoelastic material behavior . . . . . . . . . . . . . . . . . . . . . 276
12.5.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
12.5.2 Constitutive relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
12.5.3 Constitutive model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
12.5.4 Incremental analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
12.5.5 Stress update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
12.5.6 Stiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
12.5.7 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
A FE program tr2dL a1
A.1 Example input le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a1
A.2 The program tr2dL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a4
B FE program tr2d a9
B.1 Example input le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a9
B.2 The program tr2d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a11
C Stiness and compliance matrices a15
C.1 Orthotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a15
C.1.1 Voigt notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a16
C.1.2 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a16
C.1.3 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a17
C.2 Transversal isotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a17
C.2.1 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a18
C.2.2 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a19
C.3 Isotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a19
C.3.1 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a19
C.3.2 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a20
C.3.3 Axi-symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a20
D Planar elements a23
D.1 Four-node quadrilateral element . . . . . . . . . . . . . . . . . . . . . . . . . . a23
D.1.1 Cartesian coordinate system . . . . . . . . . . . . . . . . . . . . . . . . a24
D.1.2 Cylindrical coordinate system . . . . . . . . . . . . . . . . . . . . . . . a26
D.1.3 Numerical integration . . . . . . . . . . . . . . . . . . . . . . . . . . . a27
D.2 Eight-node quadrilateral element . . . . . . . . . . . . . . . . . . . . . . . . . a28
D.2.1 Numerical integration . . . . . . . . . . . . . . . . . . . . . . . . . . . a29
E FE program plaxL a31
E.1 Example input le . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a31
E.2 The program plaxL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a33
F FE program plax a39
F.1 The program plax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a39
Chapter 1
Introduction
For a good description of physical phenomena, proper constitutive equations, describing the
material behaviour, are essential. Although many constitutive equations are used since long
and mostly successful, they need continuous adaptation and extension due to both the use
of new materials and the more extreme situations of their use. Experimental techniques to
determine material parameters (-functions) need to be more and more sophisticated.
In the course CMM, attention goes to implementation of the constitutive equations in
nite element software. When this is not done properly, the accuracy of the numerical solution
and the eciency of the solution process will be detrimental.
Subsequently attention is focussed onto procedures, which are needed for various material
behaviour 1) to calculate stresses and 2) to calculate the current material stiness. Attention
is given to material models for: hypo- and hyper-elastic behavior, elastoplastic behavior,
viscoelastic behavior and viscoplastic behavior.
First, the material behavior is characterised and analysed with one-dimensional discrete
mechanical models, made of springs, dashpots and friction elements. Calculation of stress
response for a prescribed strain excitation is done with Matlab les.
The background of the nite element method (weighted residuals, interpolation, numeri-
cal integration, assemblage, partitioning) is summarized for truss, 2D (plane strain and plane
stress) and axisymmetric elements. Implementation of the material models is done in both
a truss element and in plane strain(/stress) and axisymmetric elements. The nite element
software is available online as a set of Matlab command and function les. Input les for
demo problems are also available. In the appendix of these lecture notes, some les are listed
with material model procedures.
1
2
Chapter 2
Truss structures
A truss is a mechanical element whose dimension in one direction the truss axis is much
larger than the dimensions in each direction perpendicular to the axis. A truss structure is
an assembly of trusses, which are connected mutually and to the suroundings with hinges.
The truss can transfer only axial forces along its axis, so bending is not possible, and the axis
must be and remain straight.
In this chapter, we rst consider small elongation and rotation of a truss. The material
behaves linearly elastic and the resulting equilibrium equation is linear. The nite element
method is used to model truss structures and to solve the resulting set of equilibrium equa-
tions.
Large elongations and rotations lead to a set of nonlinear equations. Moreover, the ma-
terial behavior is likely to be nonlinear as well. Solution of the set of equations must be done
iteratively. Implementation of various material models in the nite element software is the
subject of the next chapter.
2.1 Homogeneous truss
We consider a truss to be oriented with its axis along the global x-axis. Its undeformed length
is l0. The undeformed cross-sectional area has a uniform value A0. It is assumed that the
material of the truss is isotropic and homogeneous.
A0
l0
x
z
y
Fig. 2.1 : Homogeneous truss
3
4
2.1.1 Elongation and contraction
In the deformed state the length of the truss is l and its cross-sectional area is A. The
elongation is described by the axial elongation factor . The change in cross-sectional area is
described by the contraction . It is assumed that the load, which provokes the deformation,
is such that the deformation is homogeneous. This means that and are the same in each
point of the truss. The volume change is described by the volume ratio J.
A0
A
x
x
l0
l
z
z
y
Fig. 2.2 : Deformation of a homogeneous truss
elongation factor =
l
l0
=
l0 +l
l0
= 1 +
l
l0
contraction =
_
A
A0
volume change J =
lA
l0A0
=
2
2.1.2 Stress
The deformation of the truss is caused by an external axial force N. In each cross-section
(x) of the truss, an internal axial force N(x) exists. With no volume load, the cross-sectional
load will be the same in each cross-section. If a volume load is applied, this is not the case,
but we will not consider such loading here.
The axial load is such that it causes only axial deformation and no bending. In the
absence of a volume load the deformation will be homogeneous.
x
x
z
N N
y
5
x
N N
N(x) N(x)
Fig. 2.3 : Axial loading of a homogeneous truss
N(x) = N
The cross-sectional force is the result of the axial cross-sectional stress. For a homogeneous
material with no volume loads, the stress is uniform over the cross-section. This leads to
the denition of true stress, being the axial force divided by the deformed (= real) cross-
sectional area. In many (engineering) applications the engineering or nominal stress is used,
dened as the ratio of the axial force and the undeformed cross-sectional area. True stress
and engineering stress, are related by the contraction .
In literature a truss is sometimes called a tie when it carries a tensile force and a strut
when it is loaded in compression.
x

N N
N N
N N
z
x
y
z
Fig. 2.4 : Cross-sectional stress in an axially loaded homogeneous truss
axial stress = (y, z)
cross-sectional force N(x) = N =
_
A
(y, z) dA
stress uniform in cross-section N =
_
A
dA = A
true stress =
N
A
engineering stress n =
N
A0
relation =
N
A
=
A0
A
N
A0
=
1
2
n
6
2.2 Linear deformation
When the elongation of the truss is very small, the contraction is even smaller so that the
deformed cross-sectional area can be taken to be equal to the initial cross-sectional area.
Consequently there is no dierence between the true stress and the engineering stress.
2.2.1 Linear strain
Elongation is generally described by the strain . For small elongation and rotation, the linear
strain is used. For the elongation, this strain is related to the elongation facor and for the
contraction to the contraction .
linear strain = l = 1
contractive linear strain d = 1
2.2.2 Linear elastic behavior
The linear elastic material behavior is characterized by two matarial constants: Youngs
modulus and Poissons ratio. Youngs modulus relates the axial stress to the axial strain.
Poissons ratio relates the contractive strain to the axial strain. For most materials Poissons
ratio is about 0.3. For small elongations this value is constant. For small deformation the
volume change factor J can be expressed in the linear strain. For incompressible material
J = 1 implying =
1
2. The table lists values of Youngs modulus and Poissons ratio for some
materials.
axial stress strain = E
contraction strain = 1 d = 1 = = ( 1)
volume change J = ( + 1)( + 1)
2
(1 2) + 1
2.2.3 Equilibrium
We consider a truss with length l0 and cross-sectional area A0 with its axis along the global
x-axis. One end (x = 0) is xed and the other (x = l0) can be displaced in x-direction only.
The elongation of the truss equals this displacement u. The displacement is caused by an
external axial force fe. In the deformed state the length of the truss is l = l0 + u and its
cross-sectional area is A. The material of the truss is homogeneous.
When the external axial force fe is prescribed, the elongation l = u of the truss can
be determined by solving the equilibrium equation in point P, which states that the internal
force must be equal to the external force. The internal force fi is a function of the elongation,
a relation which is determined by the material behavior. It represents the resistance of the
truss against elongation.
7
u
fi l, A
P
P fi
l, A
l0, A0
fe
fe
Fig. 2.5 : Equilibrium of external and internal axial force
external force fe
internal force fi = fi(u)
equilibrium of point P fi(u) = fe
When the deformation (= elongation) is very small, there is virtually no dierence between
the undeformed and the deformed geometry. Such deformation is referred to as being geo-
metrically linear. The true axial stress = N/A approximately equals the engineering stress
n = N/A0, where N is the axial force.
When, moreover, the material behavior is not inuenced by the deformation, as is the
case for linear elastic behavior, this is referred to as physical linearity the total deformation
is linear and the internal force fi can be linearly related to the displacement u.
The equilibrium equation can be solved directly, yielding the displacement u.
2.2.4 Solution procedure
Because the relation between the external force fe and the axial displacement u is linear,
the latter can be solved directly from the equilibrium equation fi = fe, yielding the exact
solution uexact = us. The stiness K of the truss depends on the Youngs modulus E and on
the initial geometry (A0 and l0).
fi = nA0 = EA0 =
EA0
l0
u = Ku
fi = fe Ku = fe u = us =
fe
K
=
l0
EA0
fe
f fi
u us
0
fe
Fig. 2.6 : Solution of linear equilibrium equation
8
Proportionality and superposition
Two important characteristics hold for linear problems :
the deformation is proportional to the load : when the external force fe is multiplied by
a factor, say , the elongation u is also multiplied by .
superposition holds : when we determine the elongation u1 and u2 for two separate
forces, fe1 and fe2 respectively, the elongation for the combined loading fe1 +fe2 is the
sum of the separate elongations : u1 +u2.
2.3 Nonlinear deformation
When deformation and/or rotation of the truss are large, various strains and stresses can be
dened and related by material laws. The material behavior can be expected to be no longer
linearly elastic.
2.3.1 Strains for large elongation
The deformation of the truss can be characterized uniquely by the two elongation factors
and . However, it is common and useful to introduce deformation variables which are a
function of the elongation factors : the strains. A wide variety of strain denitions is possible
and used.
All strain denitions must obey some requirements, one of which is that they have to
result in the same value for small elongations, being the value of the linear strain. When we
plot the various strains as a function of the elongation factor, it is immediately clear that the
strains, which are dened here, obey this requirement.
It is obvious that one and the same strain denition must be used throughout the same
specimen and analysis. This implies that the contraction strain is dened analogously to the
elongational strain. These strains are related by a material parameter, the Poissons ratio .
It is assumed, until stated otherwise, that this parameter is constant.
linear strain = l = 1
logarithmic strain = ln = ln()
Green-Lagrange strain = gl =
1
2(
2
1)
9
1
2(2 1)
1
ln()

1
0
Fig. 2.7 : Three strain denitions as function of the elongation factor
Linear strain
The linear strain denition results in unrealistic contraction, when the elongation is too large.
The cross-sectional area of the truss can become zero, which is of course not possible.
linear strain = l = 1 =
l
l0
contraction strain d = 1 = l = ( 1)
change of cross-sectional area
=
_
A
A0
= 1 ( 1) A = A0{1 ( 1)}
2
restriction of elongation
1 ( 1) > 0 1 <
1

<
1 +

Logarithmic strain
The logarithmic strain denition does not lead to unrealistic values for the contraction. There-
fore it is very suitable to describe large deformations. 1
logarithmic strain = ln = ln()
contraction strain d = ln() = ln = ln
change of cross-sectional area
1 ln x =
e
log(x) = y x = e
y
10
=
_
A
A0
= e
ln = e
ln()
=
_
e
ln()
_
=

A = A0
2
A deformation process may be executed in a number of steps, as is often done in forming
processes. The start of a new step can be taken to be the reference state to calculate current
strains. In that case the logarithmic strain is favorably used, because the subsequent strains
can be added to determine the total strain w.r.t. the initial state.
1 2
01 12
0
Fig. 2.8 : Two-step deformation process
l0l1 l(01) =
l1l0
l0
ln(01) = ln(
l1
l0
)
l1l2 l(12) =
l2l1
l1
ln(12) = ln(
l2
l1
)
l0l2 l(02) =
l2l0
l0
= l(01) +l(12)
ln(02) = ln(
l2
l0
) = ln(
l2
l1
l1
l0
) = ln(01) +ln(12)
Green-Lagrange strain
Using the Green-Lagrange strain leads again to restrictions on the elongation to prevent the
cross-sectional area to become zero.
Green-Lagrange strain = gl =
1
2(
2
1)
contraction strain d =
1
2( 1) = ln =
1
2 (
2
1)
change of cross-sectional area
1 (
2
1) > 0 <
_
1 +

11
2.3.2 Mechanical power for an axially loaded truss
The gure shows a tensile bar which is elongated due to the action of an axial force F.
Its undeformed cross-sectional area and length are A0 and l0, respectively. In the deformed
conguration the cross-sectional area and length are A and l.
At constant force F an innitesimal small increase in length is associated with a change
in mechanical energy per unit of time (power) : P = F l. The elongation rate l can be
expressed in various strain rates.
l0, A0
l, A F
e1
e2
e3
Fig. 2.9 : Axial elongation of homogeneous truss
linear strain l = 1 l = =
l
l0
logarithmic strain ln = ln() ln =
1
=
l
l
Green-Lagrange strain gl =
1
2 (
2
1) gl = =
l
l0
=
2
l
l
P = F = F0 l =
F
A0
A00 l =
F
A0
V0 l
P = F = F ln =
F
A
A ln =
F
A
V ln
P = F = F0 l =
F
A
A
0

l =
F
A
V
1
l
P = F = F2 gl =
F
A
A
2
gl =
F
A
V
2
gl
Various stress denitions automatically emerge when the mechanical power is considered in
the undeformed volume V0 = A0l0 or the current volume V = Al of the tensile bar. The
stresses are :
: Cauchy or true stress
n : engineering or nominal stress
p1 : 1st Piola-Kirchho stress = n
: Kirchho stress
p2 : 2nd Piola-Kirchho stress
12
P = = = V0n l
P = V ln = V0(J) ln = V0 ln
P = V (1) l = V0(J1) l = V0p1 l
P = V (2) gl = V0(J2) gl = V0p2 gl
specic mechanical power : P = V0
W0 = V W
W0 = n l = ln = p1 l = p2 gl
W = = ln = 1 l = 2 gl
2.3.3 Equilibrium
Deformations may be so large that the geometry changes considerably. This and/or non-
linear boundary conditions render the deformation problem nonlinear. Proportionality and
superposition do not hold in that case. The internal force fi is a nonlinear function of the
elongation u.
Nonlinear material behavior may also result in a nonlinear function fi(u). This nonlin-
earity is almost always observed when deformation is large.
Solving the elongation from the equilibrium equation is only possible with an iterative
solution procedure.
u
fi(u)
fe
uexact
Fig. 2.10 : Nonlinear internal load and constant external load
external force fe
internal force fi = A = fi(u)
equilibrium of point P fi(u) = fe
13
2.3.4 Solution procedure
It is assumed that an approximate solution u for the unknown exact solution uexact exists.
(Initially u = 0 is chosen.)
The residual load r is the dierence between f(u) and fe. For the exact solution this
residual is zero. What we want the iterative solution procedure to do, is generating better
approximations for the exact solution so that the residual becomes very small (ideally zero).
u u
f
i
fi(u)
r
fe
uexact
Fig. 2.11 : Approximation of exact solution
analytic solution fi(uexact) = fe fe fi(uexact) = 0
approximation u fe fi(u

) = r(u

) = 0
residual r

= r(u

)
The unknown exact solution is written as the sum of the approximation and an unknown
error u. The internal force fi(uexact) is then written as a Taylor series expansion around u
and linearized with respect to u. The rst derivative of fi with respect to u is called the
tangential stiness K. Subsequently u is solved from the linear iterative equation. The
solution is called the iterative displacement.
fi(uexact) = fe
uexact = u +u
_
fi(u

+u) = fe
fi(u

) +
dfi
du

u
u = fe f

i +K

u = fe
K

u = fe f

i = r

u =
1
K
r

14
u
u
u
K
f
i
fi(u)
fe
r
Fig. 2.12 : Tangential stiness and iterative solution
With the iterative displacement u a new approximate solution u can be determined by
simply adding it to the known approximation.
When u is a better approximation than u, the iteration process is converging. As the
exact solution is unknown, we cannot calculate the deviation of the approximation directly.
There are several methods to quantify the convergence.
uexact
u
u
K
f
i
fi(u)
fe
r
u u
Fig. 2.13 : New approximation of the exact solution
new approximation u

= u

+u
error uexact u

error smaller convergence


Convergence control
When the new approximation u is better than u, the residual r is smaller than r. If its
value is not small enough, a new approximate solution is determined in a new iteration step.
If its value is small enough, we are satised with the approximation u for the exact solution
and the iteration process is terminated. To make this decision the residual is compared to a
convergence criterion cr. It is also possible to compare the iterative displacement u with a
15
convergence criterion cu. If u < cu it is assumed that the exact solution is determined close
enough.
When the convergence criterion is satised, the displacement u will not satisfy the nodal
equilibrium exactly, because the convergence limit is small but not zero. When incremental
loading is applied, the dierence between fi and fe is added to the load in the next increment,
which is known as residual load correction.
u
u
fi(u)
r
u
fe
f
i
u
Fig. 2.14 : New residual for approximate solution
residual force |r

| cr stop iteration
iterative displacement |u| cu stop iteration
u
fi(u)
fe
Fig. 2.15 : Converging iteration process
Residual and tangential stiness
The residual and the tangential stiness can be calculated from the material model, which
describes the material behavior. It is assumed that this is a relation between the axial Cauchy
stress and the elongation factor or stretch ratio =
l
l0
: = (). It is also necessary to
now the relation between the cross-sectional area A and .
16
internal nodal force f

i = N(

) = A

tangential stiness K

=
fi
u

u
=
N()
u

u
=
dN
d

d
du
geometry = 1 +
l
l0
= 1 +
1
l0
u
d
du
=
1
l0
tangential stiness K

=
dN
d

u
=
dN
d

1
l0
=
dN
d

1
l0
=
1
l0
d
d
(A)

=
1
l0
d
d

+
1
l0

dA
d

Incremental loading
The external loading may be time-dependent. To determine the associated deformation, the
time is discretized : the load is prescribed at subsequent, discrete moments in time and
deformation is determined at these moments. A time interval between two discrete moments
is called a time increment and the time dependent loading is referred to as incremental loading.
This incremental loading is also applied for cases where the real time (seconds, hours) is not
relevant, but when we want to prescribe the load gradually. One can than think of the time
as a ctitious or virtual time.
fe
t 0
fi
fe
tn tn+1 t
fe
f
un un+1 u
Fig. 2.16 : Incremental loading
Non-converging solution process
The iteration process is not always converging. Some illustrative examples are shown in the
next gures.
17
fi(u)
fe
u
fe
fi(u)
fi(u)
fe
Fig. 2.17 : Non-converging solution processes
Modied Newton-Raphson procedure
Sometimes, it is possible to reach the exact solution by modifying the Newton-Raphson iter-
ation process. The tangential stiness is then not updated in every iteration step. Its initial
value is used throughout the iterative procedure.
The gure shows a so-called snap-through problem, where no convergence can be
reached due to a cycling full Newton-Raphson iteration process. With modied Newton-
Raphson, iteration proceeds to the equilibrium fi = fe.
18
u
u
fi(u)
fi(u)
fe
fe
Fig. 2.18 : Modied Newton-Raphson procedure
Path-following solution algorithm
The iteration procedure can be combined with a path-following algorithm to control the
load step.
Each increment starts with a load step g1 = fe = 1fef where fef is the nal load
and 1 is a known fraction. The initial iterative displacement u1 can be calculated and the
initial incremental and total displacement approximations are known. The internal load fi1
and tangent stiness K1 can be calculated.
u1
K0 K1
g1
fi1
Fig. 2.19 : Path-following solution procedure : rst iterative step
19
K0u1 = fe0 +1fef = g1 u1 = K
1
0 g1
u1 = u1 ; u1 = u0 +u1 fi1 , K1
In the second iteration step the incremental load is calculated again as a fraction of the nal
load, but now the fraction 2 is unknown. It can be calculated from the requirement that the
length of [u2 g2] has to be C, which is a known constant value for the current increment
and is referred to as the arc-length. For the two resulting solutions the product of u2 and
u1 is calculated. The solution resulting in the largest value is selected. With 2 known, the
new approximate solution is determined. Obviously the incremental load step g2 is changed
compared to the initial value g1.
The iteration process is continued until convergence is reached. The convergence norm
can be calculated from the residual load, the iterative displacement or a combination of the
two.
After convergence a new increment can be started with a modied initial fraction 1.
The procedure is stopped when convergence is reached after applying the nal load fef .
K0 K1
fi1
g1
g2
u2
Fig. 2.20 : Path-following solution procedure : reduction of external incremental load
K1u2 = g2 fi2 = fe0 +2fef fi2
u2 = K
1
1 (fe0 +2fef fi2)
_
u2 g2

_
u2
g2
_
= (u2)
2
+ (g2)
2
= C
2

_
u1 +K
1
1 fe0 +K
1
1 fef2 K
1
1 fi2
_2
+ (fe0 +2fef)
2
= C
2

2 u2 u2 , u2 fi3 , K2
20
2.4 Finite element method for linear truss
When a truss structure is loaded by external forces or prescribed displacements, its defor-
mation can sometimes be calculated analytically, especially when the structure is statically
determinate. When the structure is statically indeterminate, this is only possible for very
simple cases. Practical problems can be solved numerically, using the nite element method.
When the trusses in the structure show small elongation and rotation, and when more-
over their material behavior is linearly elastic, the whole problem is linear and the nite
element method can be explained rather straightforwardly.
In the following we restrict ourselves to two-dimensional structures.
Truss element
A truss element e with two nodal points is oriented with its axis in the 1-direction of a
two-dimensional coordinate system. Both nodes move in this direction being by denition
positive , leading to an elongation of the truss : its initial length le
0 becomes le.
This elongation is resisted by the material of the truss, leading to reaction forces in both
nodes : the internal nodal point forces, again dened to be positive in the positive 1-direction.
In absence of distributed axial load the axial force N in the truss is constant and a function
of the elongation.
u1 u2
le
0
le
1 2
1 2
1
2
1 2
fi1 fi2

N(0) N(le)
Fig. 2.21 : One-dimensional truss element
nodal point displacements u

e
=
_
u1
u2
_
=
_
u(0)
u(le)
_
internal nodal point forces
f

e
i
=
_
fi1
fi2
_
=
_
N(0)
N(le)
_
=
_
k(u2 u1)
k(u2 u1)
_
=
_
k k
k k
_ _
u1
u2
_
; k =
EA0
l0
In the initial situation an angle 0 may exist between the axis of a two-dimensional truss
element and the 1-direction of the coordinate system. The displacements and forces of/in
the nodal points have two components, and are dened positive in the positive coordinate
directions. Due to the deformation, the current angle of the axis is . For small deformations
and rotations we have 0.
The internal force components can be expressed in the axial force N and the cosine and
21
sine of the angle . For a linear element the nodal forces f

e
i
can be related to the elongation,
expressed in the nodal displacements in the direction of the element axis, denoted as uL
i ,
which are related to the displacement components of the nodal points u

e. This relation is
expressed by the element stiness matrix K
e
.
1
u12
u11
u22
u21
fi12
fi22
fi21
fi11

2
Fig. 2.22 : Two-dimensional truss element
nodal point displacements u

eT
=
_
u11 u12 u21 u22

internal forces with c = cos() ; s = sin()


f

e
i
(u

) =
_

_
fi11
fi12
fi21
fi22
_

_
=
_

_
cN(0)
sN(0)
cN(e)
sN(e)
_

_
= k(u
L
2 u
L
1 )
_

_
c
s
c
s
_

_
= k(u21c +u22s u11c u12s)
_

_
c
s
c
s
_

_
= k
_

_
c2 cs c2 cs
cs s2 cs s2
c2 cs c2 cs
cs s2 cs s2
_

_
_

_
u11
u12
u21
u22
_

_
= K
e
u

e
Assembling
For the truss structure, the internal nodal forces are stored in a column f

i
and are related
to the nodal displacement components in the column u

. This relation is expressed by the


structural or global stiness matrix K.
The contributions of the individual element stinesses to the structural stiness are
added in the assembling procedure.
element contribution K
e
u

e
= f

e
i
assembled equation K u

= f

i
22
Equilibrium of the truss structure requires that the internal nodal point forces f

i
are equal
to the external nodal point forces f

e
. This leads to a linear system of equations for the nodal
displacement components u

. It is, however, not yet possible to solve this set of equations,


because some essential boundary conditions have to be incorporated rst.
f

i
= f

e
Ku

= f

e
= f

Boundary conditions
The equilibrium equations can only be solved uniquely, when proper boundary conditions are
prescribed. These boundary conditions are suppressed displacements, prescribed displace-
ments and prescribed forces.
It is always needed to prevent rigid body motions, because otherwise no (unique) so-
lution can be determined. The algebraic system of equations Ku

= f

has to be solved to
determine the nodal displacements u

. However, the stiness matrix K is singular and cannot


be inverted to solve u

when f

is known. This singularity is obvious as we must conclude that


for a rigid body translation u

= a

= 0

the nodal forces are zero.


rigid translation u

= a

no forces needed K a

= 0

with a

= 0

K singular
To get a non-singular matrix we have to suppress the rigid body movement of the construction,
by prescribing enough nodal displacements. Besides boundary conditions to suppress rigid
body motion, some more nodal displacements may be prescribed, as well as some nodal
forces. When in a node a displacement component is prescribed, the associated external force
component is unknown and vice versa.
Prescribed nodal displacement components are often denoted as kinematic boundary
conditions and prescribed nodal forces as dynamic boundary conditions.
The prescribed degrees of freedom u

p are associated with unknown force components f

p
.
The unknown degrees of freedom u

u are associated with the known (prescribed or zero) force


components f

u
. The components of u

and f

are reorganized.
u

=
_
u

u
u

p
_
; f

=
_
f

u
f

p
_
Reorganizing components of u

implies that columns of K have to be reorganized in the same


way. Reorganizing components of f

implies that rows of K have to be reorganized in the


same way. The components associated with the various parts of u

and f

can be placed in
sub-matrices of the resulting matrix K. The reorganization of columns and matrix described
above is called partitioning.
As we can see, this partitioning leads to two sets of equations. Only the rst set is
relevant for the calculation of the unknown u

u. After determination of these unknowns, the


second set is used to calculate the unknown reaction forces f

p
.
23
linear equation system Ku

= f

partitioning
_
Kuu Kup
Kpu Kpp
_ _
u

u
u

p
_
=
_
f

u
f

p
_

Kuuu

u +Kupu

p = f

u
Kpuu

u +Kppu

p = f

p
_
solving u

u Kuuu

u = f

u
Kupu

p u

u = K
1
uu(f

u
Kupu

p)
calculating unknown f

p
f

p
= Kpuu

u +Kppu

p
Links
Links (or tyings) are relations between some of the components of u

. In these relations we
make a dierence between independent and dependent components. Dependent or linked
components can be calculated from the independent ones after these have been solved. The
linked components are removed from the equation system, as will be seen later. The inde-
pendent components are not, so they are called retained components. Components which are
not part of link relations are simply denoted as free. To identify the various components we
use the indices l (linked), r (retained) and f (free).
Associated with the linked degrees of freedom are nodal forces, which ensure the rela-
tionship. They are calculated by requiring that the total virtual energy, associated with the
links, is zero.
The column u

is reorganized such that free, retained and linked components are grouped
in columns u

f , u

r and u

l. The right-hand column f

is reorganized in the same way. The


matrix K is adapted accordingly by moving rows and columns.
_
_
Kff Kfr Kfl
Krf Krr Krl
Klf Klr Kll
_
_
_
_
u

f
u

r
u

l
_
_ =
_

_
f

f
f

r
+ f

r
f

l
+ f

l
_

_
Kff u

f +Kfru

r +Kflu

l = f

f
Krf u

f +Krru

r +Krlu

l = f

r
+ f

r
Klf u

f +Klru

r +Kllu

l = f

l
+ f

l
The relation between u

l and u

r is denoted with a matrix Llr. Imposing the link relations will


result in a change of the corresponding components of f

. In a mechanical system f

r
and f

l
may be seen as forces which are needed to realize the links between u

r and u

l. The resulting
work of these forces at a random change in u

r and u

l must be zero, which results in a relation


between f

r
and f

l
.
u

l = Llru

r
f

T
l
u

l + f

T
r
u

r = 0 {u

l, u

r}
f

T
l
Llr + f

T
r
= 0

T
L
T
lr
f

l
+ f

r
= 0

r
= L
T
lr
f

l
= Lrl
f

l
24
Implementation of the link relations results in two systems of algebraic equations from which
u

r and u

l can be solved.
Kff u

f + (Kfr +KflLlr)u

r = f

f
Krf u

f + (Krr +KrlLlr)u

r = f

r
Lrl
f

l
Klf u

f + (Klr +KllLlr)u

r = f

l
+ f

l
elimination of f

l
_
_
_

Kff u

f + (Kfr +KflLlr)u

r = f

f
(Krf +LrlKlf )u

f + (Krr +KrlLlr +LrlKlr +LrlKllLlr)u

r = f

r
+Lrlf

l
_

_
Kff Kfr +KflLrl
Krf +LrlKlf Krr +KrlLlr +LrlKlr +LrlKllLlr
_ _
u

f
u

r
_
=
_
f

f
f

r
+Lrlf

l
_

Ku

= f

Program structure
A nite element program starts with reading data from an input le and initialization of
variables and databases.
Subsquently, a loop over all elements is started to calculate Ke for each individual element
and place it at the proper location in the structural matrix K (assembly). After taking into
account the link relations and boundary conditions, the unknown nodal displacements are
calculated.
Subsequently, another loop over all elements is entered to calculate the element strains,
stresses and internal nodal forces f

e
i
. The latter are assembled in the column f

i
, which then
contains the reaction forces of the system.
Finally some calculated values are stored for post-processing.
read input data from input file
calculate additional variables from input data
initialize values and arrays
for all elements
calculate initial element stiffness matrix
assemble global stiffness matrix
end element loop
determine external load from input
take tyings into account
take boundary conditions into account
calculate nodal displacements
25
for all elements
calculate stresses from material behavior
calculate element internal nodal forces
assemble global internal load column
end element loop
store data for post-processing
2.4.1 FE program tr2dL
The Matlab program tr2dL is used to model and analyze two-dimensional truss structures. It
is described in detail in appendix A. The input data, which must be provided by the user, are
a.o. the coordinates of the nodes, the location of the trusses between nodes, element material
data, link relations and prescribed nodal displacements and forces.
In this section, a few examples of two-dimensional truss structures are shown, which are
modelled and analyzed with the program.
Simple two-dimensional truss structure
A simple truss structure is shown in the left gure below. The length of the horizontal truss
[1] is 100 mm and the length of truss [2] is 200/

3 mm. Cross-sectional areas are 10 and 20


mm2, respectively. Youngs modulus is 200 and 150 GPa and Poissons ratio is = 0.3. The
prescribed force F = 100 N leads to the deformation {u2x, u2y} = {0.0071, 0.0222} mm,
which is shown in the right part of the gure. The real deformation is very small, which is in
accordance with the theory, so it is enlarged 1000 times.
1
2
3
F
y
x 1
2
1
2
3
tr2dL2bardef
Fig. 2.23 : Deformation of a truss structure ( 1000).
Transformation of nodal coordinate system
It is possible to prescribe nodal displacements and/or forces in a local nodal coordinate system,
which is rotated over an angle w.r.t. the global system. An example is shown below, where
26
in node 2, the local coordinate axes are rotated over 45o w.r.t. the xy-axes. The length of
the horizontal truss is 100 mm and the length of the vertical truss is 50 mm. Cross-sectional
areas are 1 mm2. Youngs moduli and Poissons ratios are 100 GPa and 0.25. The external
load is F = 100 N.
x
y
xL
yL
1 2
3
F
1
2
3
tr2dL2btrdef
Fig. 2.24 : Deformation of a truss structure ( 250), with a transformes nodal
coordinate system.
Tyings
The gure shows a rigid beam hanging from three trusses, which have equal stiness k. The
load P will cause an elongation of the trusses, which can be calculated, using link relations.
First the governing equations will be presented and solved analytically. Subsequently
the solution of the nite element program will be presented.
The two equilibrium relations are not sucient to solve the problem. Deformation and
thus material behavior (= stiness k) has to be taken into account. Still the nal set of
equations cannot be solved.
+
k1
F2 F3 F1 l l
k2 k3
P
F1
F2
F3
1 2 3
truss stiness k1 = k2 = k3 = k
27
equilibrium F1 +F2 +F3 P = 0 ; F12l F2l = 0
deformation v1 =
F1
k
; v2 =
F2
k
; v3 =
F3
k
equilibrium equations in displacements
kv1 kv2 kv3 P = 0 ; 2lkv1 +lkv2 = 0
Due to the rigidity of the beam, the displacements v1, v2 and v3 are not independent. The
dependency represents a link relation. Displacement v2 is linked to the displacements v1 and
v3. Displacement v2 is eliminated from the equation system and v1 and v3 are retained.
link relation v2 =
1
2 (v1 +v3) v2 =
_
1
2
1
2

_
v1
v3
_
elimination of v2 equation for retained displacements

3
2kv1
3
2kv3 P = 0
5
2lkv1 +
1
2lkv3 = 0 v1 =
1
5v3
_
_
_

3
10kv3
3
2kv3 P = 0
6
5kv3 P = 0
v3 =
5
6
P
k
v1 =
1
6
P
k
link v2 =
1
3
P
k
The nite element solution is calculated. The undeformed and deformed structure is shown
in the gure below. Both for the analytic and the numerical calculation, we nd the next
values for the nodal displacements, when setting k = 100 N/mm and P = 10000 N.
v1 =
100
6
= 16.66mm ; v2 =
100
3
= 33.33mm ; v3 =
500
6
= 83.33mm
1
2
3
4
5
6
tr2dL3blnkdef
Fig. 2.25 : Deformation of a truss structure with applied links ( 10).
28
2.5 Weighted residual formulation for nonlinear truss
In the initial conguration a truss has length 0. In the current conguration the truss is
subjected to an axial load: concentrated forces N0 and N in begin and end point, and a
volume load q(s) per unit of length. It has length and is rotated with respect to the initial
conguration. The coordinate along the truss axis is s and the direction of the axis is indi-
cated by the unit vector n.
In each point of the truss the equilibrium equation has to be satised. The equilibrium
equation is derived under assumption of static loading conditions. It is a dierential equa-
tion, for which analytical solutions do only exist for rather simple boundary conditions. For
practical problems we have to be satised with an approximate solution.
The error represented by the approximation can be smeared out along the axis of the
truss, by integrating the product of this error and a so-called weighting function over the
length of the truss.
N0 s
N
s0
q(s)

0
Fig. 2.26 : Inhomogeneous truss
equilibrium equation
d N
ds
+q(s) =0
d(An)
ds
+q(s) =0 s [ 0, ]
approximate solution and A
d(A n)
ds
+q(s) = (s) =0 s [ 0, ]
weighted error (s) is smeared out over [0, ]
_ s=
s=0
w(s) (s) ds
29
Weighted residual formulation
The product of the left-hand side of the equilibrium equation and a weighting function w(s)
can be integrated over the element length, resulting in the weighted residual integral. The
principle of weighted residuals now states that :
the requirement that the equilibrium equation is satised in each point
of the truss, is equivalent to the requirement that the weighted residual
integral is zero for every possible weighting function.
The rst term in the integral is integrated by parts to reduce the continuity requirements of
the axial stress. This results in the so-called weak form of the weighted residual formulation.
The right hand part of the resulting integral equation represents the contribution of the
external loads.
_ s=
s=0
w
_
d(An)
ds
+q
_
ds = 0 w(s)
_ s=
s=0
d w
ds
(An) ds =
_ s=
s=0
w q ds +
_
w() N() w(0) N(0)
_
= fe( w) w(s)
State transformation
Because the current length of the truss is not known, the integration can not be carried out.
Also the derivatives with respect to the coordinate s can not be evaluated. These problems
can be circumvented by a transformation. In this case we transform everything to the initial
conguration on time t0 where the truss is undeformed. This procedure is generally referred
to as the Total Lagrange approach. When transformation is carried out to the last known
conguration, we would have followed the Updated Lagrange approach, which will not be
considered here.
d( )
ds
=
ds0
ds
d( )
ds0
=
1

d( )
ds0
; ds = ds0
_ s0=0
s0=0
d w
ds0
(An) ds0 = fe0( w) w(s0)
The current stress , cross-sectional area A and axis direction n have to be determined
such that the integral is satised for every weighting function. Following a Newton-Raphson
iteration procedure, the exact solutions are written as the sum of an approximation and a
deviation. Subsequently linearisation is carried out.
30
_ s0=0
s0=0
d w
ds0
(

+)(A

+A)(n

+n) ds0 = fe0( w) w(s0)


Linearisation with assumption A 0 leads to an iterative weighted residual integral.
_ s0=0
s0=0
d w
ds0
A

ds0 +
_ s0=0
s0=0
d w
ds0

nds0
= fe0( w)
_ s0=0
s0=0
d w
ds0

ds0 w(s0)
Material model iterative stress change
The material model relates the stress to the elongation . Using this relation the iterative
change can be expressed in the iterative displacement u.
= () =
d
d

=
d
d

d(s)
ds0
=
d
d

d(u)
ds0
Rotation iterative orientation change
Due to the rotation of the truss, the axis direction vector n is also a function of the iterative
displacement. The vector m is a unit vector perpendicular to n.
n =
dx
ds
=
ds0
ds
dx
ds0
=
1

dx
ds0
n =
_

1
2
dx
ds0
_
+
_
1

_
d(x)
ds0
=
_

n
_
+
_
1

_
d(x)
ds0
=
_

nn
_

d(u)
ds0
+
_
1

_
d(u)
ds0
=
_
(I nn)
1

d(u)
ds0
=
_
m m
1

d(u)
ds0
Iterative weighted residual integral
The expressions for and n are substituted into the iterative weighted residual integral.
_ s0=0
s0=0
d w
ds0

_
d
d

d(u)
ds0
_
A

ds0 +
_ s0=0
s0=0
d w
ds0

_
m

d(u)
ds0
_
ds0
= fe0( w)
_ s0=0
s0=0
d w
ds0

ds0 w(s0)
31
_ s0=0
s0=0
d w
ds0
n

_
d
d

_
n

d(u)
ds0
ds0 +
_ s0=0
s0=0
d w
ds0
m

A
1

_
m

d(u)
ds0
ds0
= fe0( w)
_ s0=0
s0=0
d w
ds0

ds0 w(s0)
2.6 Finite element method for nonlinear truss
The mechanical behavior of truss structures, which are build from nonlinear trusses, which
may show large elongations and (thus) large rotations, can be analyzed with the nite element
method. Individual truss elements are considered rst, which means that the structure is
discretized. Later the contributions of all trusses will be combined in an assembling procedure.
Element equation
We start with the weighted residual integral for one truss element, which length is e
0 in the
initial state and e in the current state. First, the global coordinate s0 is replaced by a local
element coordinate .
s
fe
1
fe
2
s0
q(s)
e
e
0
Fig. 2.27 : Inhomogeneous truss element in undeformed state
local coordinate : 1 1 ; ds0 =
l0
2
d ;
d( )
ds0
=
2
l0
d( )
d
_ =1
=1
d w
d
n

_
d
d

A
2
l0
_
n

d(u)
d
d +
_ =1
=1
d w
d
m

A
1

2
l0
_
m

d(u)
d
d
= f
e
e0( w)
_ =1
=1
d w
d

d
32
The vectors in the weighted residual integral are written in components with respect to a
vector basis.
_ =1
=1
dw

T
d
n

_
d
d

A
2
l0
_
n

T d(u

)
d
d +
_ =1
=1
dw

T
d
m

A
1

2
l0
_
m

T d(u

)
d
d
= f
e
e0(w

)
_ =1
=1
dw

T
d

d
Interpolation
Both the iterative displacement and the weighting function components are interpolated be-
tween their values in the element nodes. Here we use a linear interpolation between two nodal
values. The element nodes are located in the begin and end points of the element. Following
the Galerkin procedure, the interpolation functions for u

and w

are taken to be the same.


The derivatives of u

and w

can also be interpolated directly.


u

T
=
_
u1 u2

=
_
u111 +u212 u121 +u222

w

T
=
_
w111 +w212 w121 +w222

with
1
() = 1
2(1 ) ;
2
() = 1
2(1 +)
derivatives
d(u

)
d
=
_

_
d(u1)
d
d(u2)
d
_

_ =
_
d1
d 0
d2
d 0
0
d1
d 0
d2
d
_
_

_
u11
u12
u21
u22
_

_
= 1
2
_
1 0 1 0
0 1 0 1
_
u

e
dw

T
d
=
_
dw1
d
dw2
d
_
=
_
w11 w12 w21 w22

_
d1
d 0
0
d1
d
d2
d 0
0
d2
d
_

_
= w

eT 1
2
_

_
1 0
0 1
1 0
0 1
_

_
Substitution of the interpolated variables leads to an element integral equation, where the
internal nodal forces and the element tangential stiness matrix can be recognized.
w

eT
_ =1
=1
_

_
1 0
0 1
1 0
0 1
_

_
_
c
s
_
1
4
_
d
d

A
2
l0
_
_
c s

_
1 0 1 0
0 1 0 1
_
d u

e
+
w

eT
_ =1
=1
_

_
1 0
0 1
1 0
0 1
_

_
_
s
c
_
1
4
_

A
1

2
l0
_
_
s c

_
1 0 1 0
0 1 0 1
_
d u

e
33
= f
e
e0(w

e
) w

eT
_ =1
=1
_

_
1 0
0 1
1 0
0 1
_

_
1
2
_
c
s
_
(

) d
w

eT
_ =1
=1
_
1
2
d
d

A
1
l0
_
_

_
c
s
c
s
_

_
c s c s

d u

e
+
w

eT
_ =1
=1
_
1
2

A
1

1
l0
_
_

_
s
c
s
c
_

_
s c s c

d u

e
= f
e
e0(w

e
) w

eT
_ =1
=1
1
2 (

)
_

_
c
s
c
s
_

d
w

eT
_ =1
=1
_
1
2
d
d

A
1
l0
_
_

_
c2 cs c2 cs
cs s2 cs s2
c2 cs c2 cs
cs s2 cs s2
_

d u

e
+
w

eT
_ =1
=1
_
1
2

A
1

1
l0
_
_

_
s2 cs s2 cs
cs c2 cs c2
s2 cs s2 cs
cs c2 cs c2
_

d u

e
= f
e
e0(w

e
) w

eT
_ =1
=1
1
2 (

)
_

_
c
s
c
s
_

d
With the introduction of some proper matrices and columns, the element equation can be
written in short form.
w

eT
__ =1
=1
_
1
2
d
d

A
1
l0
_
d M

L
_
u

e
+w

eT
__ =1
=1
_
1
2

A
1

1
l0
_
d M

N
_
u

e
= f
e
e0(w

e
) w

eT
_ =1
=1
1
2 (

) V

d
w

eT
K
e
u

e
= w

eT
f

e
e0
w

eT
f

e
i
= w

eT
r

e
34
Integration
Integation over the element length is needed to determine the element stiness matrix K
e
and the internal force column f

e
i
.
For a homogeneous element, e.g. an element with uniform cross-sectional area and ma-
terial properties, this leads to the following expressions.
K
e
=
_
d
d

A
1
l0
_
_

_
c2 cs c2 cs
cs s2 cs s2
c2 cs c2 cs
cs s2 cs s2
_

+
_

A
1
l
_
_

_
s2 cs s2 cs
cs c2 cs c2
s2 cs s2 cs
cs c2 cs c2
_

e
i
=

_
c
s
c
s
_

Assembling
The contributions of the individual elements are added in the assembling procedure. The re-
sult is an integral equation for the total system, which, according to the principle of weighted
residuals, has to be satised for every column with nodal weighting function values. This
requirement leads to a system of algebraic equations from which the iterative nodal displace-
ment components must be solved.
element contribution w

eT
K
e
u

e
= w

eT
f

e
e0
w

eT
f

e
i
= w

eT
r

e
assembled equation w

T
K

= w

T
f

e0
w

T
f

i
= w

T
r

iterative equation system K

= r

Boundary conditions
Boundary conditions are only applied at the beginning of an incremental step. Links
relations between degrees of freedom can be incorporated as usual, but now of course for
the iterative displacements.
Program structure
A nite element program starts with reading data from an input le and initialization of
variables and databases.
The loading is prescribed as a function of the (ctitious) time in an incremental loop.
In each increment the system of nonlinear equilibrium equations is solved iteratively.
In each iteration loop the system of equations is build. In a loop over all elements, the
stresses are calculated and the material stiness is updated. The element internal nodal force
35
column and the element stiness matrix are assembled into the global column and matrix.
After taking tyings and boundary conditions into account, the unknown nodal displace-
ments and reaction forces are calculated.
When the convergence criterion is not reached, a new iteration step is performed. After
convergence output data are stored and the next incremental step is carried out.
read input data from input file
calculate additional variables from input data
initialize values and arrays
while load increments to be done
for all elements
calculate initial element stiffness matrix
assemble global stiffness matrix
end element loop
determine external incremental load from input
while non-converged iteration step
take tyings into account
take boundary conditions into account
calculate iterative nodal displacements
calculate total deformation
for all elements
calculate stresses from material behavior
calculate material stiffness from material behavior
calculate element internal nodal forces
calculate element stiffness matrix
assemble global stiffness matrix
assemble global internal load column
end element loop
calculate residual load column
calculate convergence norm
end iteration step
store data for post-processing
end load increment
36
2.6.1 FE program tr2d
The Matlab program tr2d is used to model and analyze two-dimensional truss structures,
where large deformations and nonlinear material behavior may occur. The program is de-
scribed in detail in appendix B.
In this section, examples of two-dimensional truss structures are shown. The material
behavior is always elastic and described by a linear relation between the Cauchy stress and
the linear strain. Other material models have also been implemented in the program, but
this is the subject of the next chapter.
Large deformation of a truss structure
A structure is made of ve trusses. The vertical truss is 0.5 m and the horizontal truss is 1
m in length. Cross-sectional areas are 100 mm2. The modulus is 2.5 GPa. Contraction is
not considered ( = 0). The vertical displacement of node 4 is prescribed to increase from
0 to -0.25 m. The reaction force, the horizontal displacement of node 4 and the vertical
displacement of node 2 is plotted against the ctitious time t.
1
2
3
4
tr2d1def
0 0.5 1 1.5
20000
15000
10000
5000
0
5000
t
f y (4) [m
m
]
0 0.5 1 1.5
0.025
0.02
0.015
0.01
0.005
0
0.005
t
u x (4) [m
m
]
0 0.5 1 1.5
1
0
1
2
3
4
5
x 104
t
u y (2) [m
m
]
Fig. 2.28 : Large deformation of a truss structure.
Buckling
Large rotations occur when buckling leads to a sudden increase in deformation. The theoret-
ical buckling load can be calculated analytically for a simple systems as shown here.
The numerical calculation starts with a very small imperfection being an initial vertical
displacement of the inner node(s) of 0.0001 m. This allows us to reach not only the rst and
smallest buckled state, the symmetric shape, but also the second mode, the anti-symmetric
shape. Also a larger imperfection is analized for both buckling modes.
The horizontal trusses have a high stiness of kt = (EA)/l = (100e9)(100e 6)/1 N/m,
37
while the springs have a very low stiness of k = 1 N/m. The displacement in node 4 is
prescribed to increase from 0 to -0.02 m.
F
F k k
l l l
Fig. 2.29 : Symmetric and anti-symmetric buckling.
symm : Fc =
kl
3
; anti-symm : Fc = kl
0.02 0.015 0.01 0.005 0
1
0.8
0.6
0.4
0.2
0
disp [m]
F
[N
]
tr2dbuckuf


symm.perf
symm.imperf
antisymm.perf
antisymm.imperf
tr2dbuckdef
Fig. 2.30 : Buckling forces versus displacement (left). Symmetric and anti-symmetric
buckling shapes (right).
38
Chapter 3
One-dimensional material models
In the following sections material behavior is described and used in a one-dimensional context.
The material behavior is modeled, using a discrete mechanical model of springs, dashpots and
friction sliders. The axial stress is related to the axial strain by one or more (dierential)
equation(s) from which the stress response must be calculated when the strain excitation is
prescribed. This stress update procedure is implemented in Matlab les.
The various material models are incorporated in a nite element program, which is used
to model and analyze the mechanical behavior of truss structures, subjected to prescribed
displacements and/or forces. In the iterative solution procedure, the material stiness plays
an essential role and must be derived from the material law.
3.1 Material behavior
Characterization of the mechanical behavior of an unknown material almost always begins
with performing a tensile experiment. A stepwise change in the axial stress may be prescribed
and the strain of the tensile bar can be measured and plotted as a function of time. From
these plots important conclusions can be drawn concerning the material behavior.
Another way of representing the measurement data of the tensile experiment is by plot-
ting the stress against the strain, resulting in the stress-strain curve. The relation between
stress and strain may be linear or nonlinear. Also, the relation may be history dependent,
due to changes in the material structure. Dierent behavior in tensile and compression may
be observed.
All these features must be captured by the material model, which describes the experi-
mentally observed behavior as accurate as possible in a range required by the application.
3.1.1 Time history plots
For elastic material behavior the strain follows the stress immediately and becomes zero after
stress release. For elastoplastic material behavior the strain also follows the stress immediately,
but there is permanent deformation after stress release. When the material is viscoelastic
the strain shows time delayed response on a stress step, which indicates a time dependent
behavior. When time dependent behavior is accompanied by permanent deformation, the
behavior is referred to as viscoplastic.
39
40
t t

t2

t1 t2 t1
t t

t1 t2 t1 t2
e
p
t t2 t

t1 t2 t1
t2 t t

t1 t2 t1
Fig. 3.1 : Strain response for a stress-step for a) elastic, b) elastoplastic, c)
viscoelastic and d) viscoplastic material behavior
3.1.2 Tensile curves
Tensile curve : elastic behavior
When elastic behavior is well described by a linear relation between a stress and a strain, the
elastic behavior is referred to as linear.

Fig. 3.2 : Tensile curves for elastic material behavior


Tensile curve : viscoelastic behavior
Viscoelastic behavior is time-dependent. The stress is a function of the strain rate. There
is a phase dierence between stress and strain, which results in a hysteresis loop when the
loading is cycling in time.
41

0
0
Fig. 3.3 : Tensile curve and hysteresis loop for viscoelastic material behavior
Tensile curve : elastoplastic behavior
When a material is loaded or deformed above a certain threshold, the resulting deformation
will be permanent or plastic. When time (strain rate) is of no importance, the behavior is re-
ferred to as elastoplastic. Stress-strain curves may indicate dierent characteristics, especially
when the loading is reversed from tensile to compressive.


Fig. 3.4 : Tensile curves for elastoplastic material behavior
Tensile curve : viscoplastic behavior
A combination of plasticity and time-dependent phenomena is called viscoplastic behavior.
This behavior is often observed for polymeric materials. For some polymers the stress reaches
a maximum and subsequently drops with increasing strain. This phenomenon is referred to
as intrinsic softening. In a tensile experiment it will provoke necking of the tensile bar.
42

Fig. 3.5 : Tensile curves for viscoplastic material behavior


Tensile curve : damage
Structural damage inuences the material properties. The onset and evolution of damage
can be described with a damage model. For materials like concrete and ceramics, the onset
and propagation of damage causes softening. Because damage is often associated with the
initiation and growth of voids, the stress-strain curve is dierent for tensile and compressive
loading.

Fig. 3.6 : Tensile curve for damaging material with dierent behavior in tension and
compression
3.1.3 Discrete material models
Material models relate stresses to deformation and possibly deformation rate. For three-
dimensional continua the material model is often represented by a (large) number of coupled
(dierential) equations. As a simplied introduction, we will present material models rst in
a one-dimensional setting. The material behavior is represented by the behavior of a one-
dimensional, discrete, mechanical system of springs, dashpots and friction sliders. For such a
system the relation between the axial stress and the axial strain can be derived.
When the model is employed in a truss, the stress will be used to calculate the internal
axial force.
43

Fig. 3.7 : Discrete elements : spring, dashpot and friction slider


44
3.2 Elastic material behavior
When a material behaves elastically, the current stress can be calculated directly from the
current strain, because there is no path and/or time dependency. When the stress is released,
the strain will become zero, so there is no permanent deformation at zero stress. All stored
strain energy is released and there is no dissipation. For the one-dimensional case the elastic
behavior is described by a relation between the stress and the elongation factor or the
strain .
unloading
loading
0

1
Fig. 3.8 : Non-linear elastic material behavior
Large strain elastic behavior
For large deformations, nonlinear elastic behavior can be observed in polymers, elastomeric
materials (rubbers) and, on a small scale, in atomic bonds, when a tensile/compression test
is carried out and the axial force F is plotted as a function of . In a material model we
want to describe such behavior with a mathematical relation between a stress and a strain.
Consideration of the stored elastic energy per unit of volume learns that each stress denition
is associated with a certain strain denition, so these should be combined in a material model.
However, when the observed material behavior is described accurately by another stress/strain
combination, it can be used as well.
For three-dimensional models more considerations have to be taken into account. Care
has to be taken that the material model does not generate stresses for large rigid body
rotations of the material.
45

0 1

0 1
Fig. 3.9 : Non-linear stress-strain relations for an elastomeric material and for an
atomic bond
Small strain elastic behavior
For small elongations, all strain denitions are the same, as are all stress denitions. The
relation between stress and strain is linear and the constant material parameter is the Youngs
modulus.
strain = gl = ln = l = 1
stress =
F
A
=
F
A0
= n
linear elastic behavior = E = E( 1)
modulus E = lim
1
d
d
= lim
0
d
d
3.2.1 Elastic models
The discrete one-dimensional model for elastic material behavior is a spring. The behavior is
modeled with a relation between the stress and the elongation factor or a strain . The
material stiness C is the derivative of w.r.t. the stretch ratio . The derivative w.r.t. the
strain results in the stiness C.
Consideration of the stored elastic energy per unit of material volume learns that, in
a material model, true stress should be combined with logarithmic strain ln, engineering
stress n with linear strain l or 2nd-Piola-Kirchho stress p2 with Green-Lagrange strain
gl. Experimentally observed tensile behavior can often be described with a linear relation
between a certain stress and its associated strain.
46

Fig. 3.10 : Spring


constitutive equation = ()
stiness C =
d
d
=
d
d
d
d
= C
d
d
elastic models (examples)
_
_
_
linear true-log. = C ln() = Cln
linear eng.-lin. n = C( 1) = Cl
3.2.2 Hyper-elastic models
Elastomeric materials (rubbers) show very large elastic deformations (elongation up to 5).
The material models for these materials are therefore referred to as hyper-elastic. They are
derived from an elastic energy function, which has to be determined experimentally. The
so-called Rivlin or Mooney type of these functions are expressed in the principal elonga-
tion factors i, i = 1, 2, 3. Experimental observations indicate that elastomeric materials are
incompressible, so that we have 123 = 1.
W =
n
i
m
j
Cij (I1 3)
i
(I2 3)
j
with C00 = 0
I1 =
2
1 +
2
2 +
2
3
I2 =
2
1
2
2 +
2
2
2
3 +
2
3
2
1 =
1
2
3
+
1
2
1
+
1
2
2
The incremental change of the elastically stored energy per unit of deformed volume, can be
expressed in the principal stresses and the principal logarithmic strains.
dW = 1dln1 +2dln2 +3dln3
Mooney models
For incompressible materials like elastomers (rubber) the stored elastic energy per unit of de-
formed volume is specied and tted onto experimental data. Several specic energy functions
are used.
47
Neo-Hookean W = C10 (I1 3)
Mooney-Rivlin W = C10 (I1 3) +C01 (I2 3)
Signiorini W = C10(I1 3) +C01(I2 3) +C20(I1 3)
2
2nd-order invariant model W = C10(I13)+C01(I23)+C11(I13)(I23)+C20(I13)
2
Yeoh W = C10(I1 3) +C20(I1 3)
2
+C30(I1 3)
3
Klosner-Segal W = C10(I1 3) +C01(I2 3) +C20(I1 3)
2
+C03(I2 3)
3
Third-order model of James, Green and Simpson
W = C10(I1 3) +C01(I2 3) +C11(I1 3)(I2 3) +
C20(I1 3)
2
+C02(I2 3)
2
+C21(I1 3)
2
(I2 3) +
C30(I1 3)
3
+C03(I2 3)
3
+C12(I1 3)(I2 3)
2
Ogden models
For slightly compressible materials the Ogden specic energy functions are used. Because
the volume change is not zero, these functions depend on the volume change factor J. The
second part of the energy function accounts for the volumetric deformation. Because the
volumetric behavior is characterized by a constant bulk modulus, the model is conned to
slightly compressible deformation.
For highly compressible materials like foams, specic energy functions also exist. The
rst part of the energy function also describes volume change.
slightly compressible W =
N
i=1
ai
bi
_
J

bi
3
_

bi
1 +
bi
2 +
bi
3
_
3
_
+ 4.5K
_
J
1
3 1
_2
highly compressible W =
N
i=1
ai
bi
_

bi
1 +
bi
2 +
bi
3 3
_
+
N
i=1
ai
ci
(1 J
ci )
One-dimensional models
For tensile (or compressive) loading of a homogeneous and isotropic truss, where the axial
direction is taken to be the 1-direction, we have : 1 = and 2 = 3 = 1/

. In this case
there is only an axial stress 1 = , so that we have
dW = dln =
dW
dln
=
dW
d
d
dln
=
dW
d

48
The Neo-Hookean model is the simplest model as it contains only one material parameter.
Axial stress and axial force F can be calculated easily. From statistical mechanics it is
known that for an ideal rubber material the stress is :
=
RT
M
_

_
with : density
R : gas constant = 8.314 JK1mol1
T : absolute temperature
M : average molecular weight
Most rubber materials cannot be characterized well with the Neo-Hookean model. The more
complex Mooney-Rivlin model yields better results. The stiness C is a function of the
elongation factor . The initial stiness E is often referred to as the modulus.
Neo-Hookean
W = C10
_

2
+
2

3
_
= C10
_
2
2
2
_
= 2C10
_

_
C =

= 2C10
_
2 +
1
2
_
; E = lim
1

= 6C10
F = A =
1

A0 = 2C10A0
_

1
2
_
Mooney-Rivlin
W = C10
_

2
+
2

3
_
+C01
_
1
2
+ 2 3
_
= 2C10
_

_
+ 2C01
_

_
1

C =

= 2C10
_
2 +
1
2
_
+ 2C01
_
1 +
2
3
_
; E = lim
1

= 6(C10 +C01)
F = A =
1

A0 = A0
1

_
2C10
_

_
+ 2C01
_

_
1

_
3.2.3 Stress update
The relation between and can be used directly to update the stress when the strain is
known.
= (t +t) = ((t +t)) = ()
49
3.2.4 Stiness
The material stiness is determined by taking the derivative of the stress with respect to the
elongation ratio or the strain.
C =
d
d
3.2.5 Examples
Strain excitation
With the strain prescribed, the stress is calculated for various elastic models. This is done
using the stress calculation procedures for the elastic models.. The stress is presented versus
the stretch ratio , which is varied between 0.5 and 3.5. For all elastic models the material
constant is : C = 100000 MPa.
The models with a linear relation between stress ( or P) and Green-Lagrange strain,
clearly lack a physically realistic description of the material behavior during compression.
0.5 1 1.5 2 2.5 3 3.5
0.5
0
0.5
1
1.5
2
2.5
x 10
5

[M
P
a]
delasex11xxls
0.5 1 1.5 2 2.5 3 3.5
1
0.5
0
0.5
1
1.5
x 10
5

[M
P
a]
delasex12xxls
Fig. 3.11 : Stress versus for l and ln models
0.5 1 1.5 2 2.5 3 3.5
1
0
1
2
3
4
5
6
x 10
5

[M
P
a]
delasex13xxls
0.5 1 1.5 2 2.5 3 3.5
5
0
5
10
15
20
x 10
5

[M
P
a]
delasex14xxls
Fig. 3.12 : Stress versus for gl and P gl models
50
For the elastomeric models the material constants are : C10 = 20000 MPa and C01 =
20000 MPa.
0.5 1 1.5 2 2.5 3 3.5
1
0
1
2
3
4
5
x 10
5

[M
P
a]
delasex21xxls
0.5 1 1.5 2 2.5 3 3.5
4
2
0
2
4
6
8
x 10
5

[M
P
a]
delasex22xxls
Fig. 3.13 : Stress versus for Neo-Hookean and Mooney-Rivlin models
The stress update procedure is implemented in the FE truss program. Also the material
stiness is updated in the iterative solution procedure.
A truss is loaded axially with a prescribed elongation/force. The initial length l0 of the
truss is 100 mm and the initial cross-sectional area A0 is 10 mm2. The axial force/elongation
is calculated. The cross-sectional area will change as a function of the elongation.
Fig. 3.14 : Tensile loading of truss element
For all elastic models the elastic constant is taken C = 100000 MPa and Poissons ratio is
0.3. Some material models will result in its value to become zero and even negative, which
clearly shows the limited use of the model.
1 1.2 1.4 1.6 1.8 2
0
1
2
3
4
5
x 10
5

F
[N
]
1Delasex11lf
1 1.2 1.4 1.6 1.8 2
5
6
7
8
9
10

A
[m
m
2 ]
1Delasex11la
Fig. 3.15 : Axial force and cross-sectional area versus the elongation for l model
51
0.5 1 1.5 2 2.5 3 3.5
1.5
1
0.5
0
0.5
1
x 10
6

F
[N
]
1Delasex12lf
0.5 1 1.5 2 2.5 3 3.5
4
6
8
10
12
14
16

A
[m
m
2 ]
1Delasex12la
Fig. 3.16 : Axial force and cross-sectional area versus the elongation for ln model
0.5 1 1.5 2 2.5 3 3.5
14
12
10
8
6
4
2
0
2
x 10
6

F
[N
]
1Delasex13lf
0.5 1 1.5 2 2.5 3 3.5
25
20
15
10
5
0
5
10
15

A
[m
m
2 ]
1Delasex13la
Fig. 3.17 : Axial force and cross-sectional area versus the elongation for gl model
0.5 1 1.5 2 2.5 3 3.5
5
0
5
10
15
20
x 10
6

F
[N
]
1Delasex14lf
0.5 1 1.5 2 2.5 3 3.5
25
20
15
10
5
0
5
10
15

A
[m
m
2 ]
1Delasex14la
Fig. 3.18 : Axial force and cross-sectional area versus the elongation for P gl model
For the Neo-Hookean and the Mooney-Rivlin model, the axial force is calculated for a pre-
scribed axial elongation. Material parameters are : C01 = 20000 MPa and C10 = 20000
MPa.
52
0.5 1 1.5 2 2.5 3 3.5
1.5
1
0.5
0
0.5
1
1.5
x 10
6

F
[N
]
1Delasex21lf
0.5 1 1.5 2 2.5 3 3.5
5
4
3
2
1
0
1
2
x 10
6

F
[N
]
1Delasex22lf
Fig. 3.19 : Axial force and cross-sectional area versus the elongation for Neo-Hookean
and Mooney-Rivlin model
53
3.3 Elastoplastic material behavior
Below a certain load (stress) value, the deformation of all materials will be elastic. When
the stress exceeds a limit value, plastic deformation occurs, which means that permanent
elongation is observed after release of the load. At increased loading above the limit value, the
stress generally increases with increasing elongation, a phenomenon referred to as hardening.
Reversed loading will rst result in elastic deformation, but after reaching a limit value of
the stress, plastic deformation will be observed again. Looking at the stress-strain curve after
a few loading reversals, it can be seen that elastoplastic material behavior is history dependent:
the stress is not uniquely related to the strain; its value depends on the deformation history.
The total stress-strain history must be taken into account to determine the current stress.


Fig. 3.20 : Stress-strain curves for elastoplastic material behavior
When a tensile bar, with undeformed length l0 and cross-sectional area A0, is subjected to
a tensile test, the axial force F and the length l can be measured. The axial strain can
be calculated from the elongation factor . To calculate the true stress = F/A, the actual
cross-sectional area of the tensile bar must be measured during the experiment. The nominal
stress n = F/A0 can be calculated straightforwardly.
The nominal stress n can be plotted against the linear strain l. Until point P (n
= P, the proportionality limit) is reached, the material behavior is assumed to be linear
elastic : n = E, where E is Youngs modulus. When the stress exceeds the initial yield
stress y0 > P , unloading will reveal permanent (= plastic) deformation. In practice y0 is
taken to be the stress where a plastic strain of 0.2 % remains. In the following however, we
will assume that y0 is exactly known and that y0 = P.
The axial force and therefore the nominal stress will reach a maximum value. At that
point necking of the tensile bar will be observed. The maximum nominal stress is the tensile
strength T . In forming processes strains can be much higher than in a tensile test, because
of the compression in certain directions.
54
Experiments have shown that during plastic deformation the volume of the metals and
metal alloys remains constant : plastic deformation is incompressible.
p e
A
A
B
y
T
F

n
P
y0
y0
0.2
Fig. 3.21 : Stress-strain curve during tensile test
P proportional limit
n = y0 yield
y0 initial yield stress
y0 strain at y0 : y0 = y0/E
0.2 0.2-strain : p = 0.2% = 0.002
T tensile strength
F fracture strength
F fracture strain ( 5% = 0.05 (metals))
For metal alloys a compression test instead of a tensile test will reveal that rst yield will occur
at n = y0. The initial material behavior is the same in tension and compression.
When the axial load is released at A with y0 < A < T , the unloading stress-strain
path is elastic and characterized by the initial Youngs modulus E. The permanent or plastic
elongation is represented by the plastic strain p. The dierence between the total strain in
point A and the plastic strain is the elastic strain e = A p =
A
E .
When after unloading, the bar is again loaded with a tensile force, the elastic line BA
will be followed where = E = Ee holds. The Youngs modulus E is assumed to be
not aected by the plastic deformation. For A( A) further elastoplastic deformation
takes place and the stress-strain curve will be followed as if unloading had not occurred. The
stress A is the current yield stress y, which is generally larger then the initial yield stress
y0. The increase, referred to as hardening, is related to the plastic strain by a hardening
law. The plastic strain is referred to as a history parameter.
55
To study the hardening phenomenon, the tensile bar is not reloaded in tension but in
compression. Two extreme observations may be done, illustrated in the gure.
In the rst case we can observe that the elastic trajectory increases in length due to plastic
deformation : AA > Y0Y
0. The elastic trajectory is symmetric about = 0 (BA = BA).
What we observe is isotropic hardening.
In the second case it is observed that the elastic trajectory remains of constant length :
AA = Y0Y
0. It is symmetric about the line OC (CA = CA). After unloading the yield stress
under compression is dierent than the yield stress under tension. This is called kinematic
hardening. The stress in point C, the center of the elastic trajectory, is the shift stress = q.
A
B
A
Y0
Y
0
0

q
A
A
0
Y
0
Y0
B
C

y0
Fig. 3.22 : Isotropic and kinematic hardening
isotropic hardening
- elastic area : larger & symmetric w.r.t. = 0
tensile : = y
compression : = y
_
kinematic hardening
- elastic area : constant & symmetric w.r.t. = q
- shift stress : q shift elastic area
- Bauschinger eect
tensile : = q +y0
compression : = q y0
_
combined isotropic/kinematic hardening
tensile : = q +y
compression : = q y
_
56
Isotropic hardening could be described by relating the yield stress y to the plastic strain
p. However, this would lead to the unrealistic conclusion that the yield stress could increase
while the plastic strain decreases. To prevent this problem, the eective plastic strain p
is taken as the history parameter. It is a measure of the total plastic strain, be its change
positive or negative, and as such cannot decrease.
p =

|p| =
=t
=0
|p|
t
t =
_ t
=0
| p| d =
_ t
=0
p d
3.3.1 Elastoplastic models
The elastoplastic deformation characteristics can be represented by a discrete mechanical
model. A friction element represents the yield limit and a hardening spring stiness H
(H > 0) provides the stiness reduction after reaching the yield limit. The elastoplastic
model describes rate-independent plasticity there is no dashpot in the discrete model , so
the time is ctitious and rate is just referring to momentary change.
In an elastoplastic material model a yield criterion is used to decide at which stress level
a purely elastic deformation will be followed by elastoplastic deformation. A yield criterion
f is used to determine whether a stress state is elastic or elastoplastic.
During elastoplastic deformation the total strain rate ( ) is additively decomposed in
an elastic ( e) and a plastic ( p) part. The plastic strain rate p is related to
f
by the
rate of the plastic multiplier , the so-called consistency parameter . During ongoing plastic
deformation the consistency equation f = 0 must be satised, because f must remain zero.
The hardening law relates the current yield stress y to the initial yield stress y0 and
the eective plastic strain p. The shift stress q, which may decrease, is related to the plastic
strain p.
In the one-dimensional situation considered here, the eective stress is simply the axial
stress : = .
H
y
e p

E
Fig. 3.23 : Discrete model for elastoplastic material behavior
57
f = ( q)
2

2
y with f < 0 | f = 0 f < 0 elastic
f = 0 f = 0 elastoplastic
y = y(y0, p) ; q = q(p)
= e + p
= Ee e =
1
E

p =
f

= 2 ( q) ; p = | p| = 2 | q|
p =
_ t
=0
p d
Constitutive equations
From the constitutive relations a set of constitutive equations can be derived. Using y =
dy
d p
p = H p, q =
dq
dp
p = K p and p = | q|, the rates p, and q can be related to the
strain rate .
= E e = E( p) = E{ 2 ( q)}
f = 0
_
_
_
Hardening laws
For one-dimensional stress states encountered in the axial loading of a truss, several hardening
laws are formulated, based on experimental observations. They can be generalized to three-
dimensional stress-strain states. The current yield stress is related to the eective plastic
strain and the initial yield stress. The hardening parameter is H =
dy
d p
.
Linear and power law hardening laws
Many hardening laws represent a linear or exponential relationship between stress and strain.
Von Mises (1913) y = y0 ideal plastic H = 0
Ludwik (1909) y = y0 +y0
_
p
y0
_n
(0 n 1)
H = n
y0
y0
_
p
y0
_n1
= nE
_
p
y0
_n1
mod. Ludwik y = y0
_
1 +m
n
p
_
H = y0mn
n1
p
58
Swift (1952) y = C(m+ p)
n
with C =
y0
mn
H = Cn(m+ p)
n1
Ramberg-Osgood (1943) p =
y
E
_
1 +
_
y
y0
_m1
_
(m 0; 3
7 )
Asymptotically perfect hardening laws
Some hardening laws are formulated in such a way as to result in no hardening (ideal plastic
behavior) for large strain values.
Prager (1938) y = y0 tanh
_
E p
y0
_
H =
y0
y0
_
sech
_
p
y0
__2
= E
_
sech
_
p
y0
__2
Betten I (1975) y = y0
_
tanh
_
E p
y0
_m_1/m
(m > 1)
H = E
_
p
y0
_m1 _
tanh
_
p
y0
_m_ 1
m1 _
sech
_
p
y0
_m_2
Voce (1949) y = C
_
1 ne
m p
_
with C =
y0
1 n
(m > 1)
H = Cnme
m p
Betten II (1975) y = y0 + (E p)
_
1 +
_
p
y0
_m_1/m
H = E
_
1 +
_
p
y0
_m_ 1
m
_
1
_
p
y0
_m_
1 +
_
p
y0
_m_1
_
3.3.2 Stress update
Elastoplastic material behavior is history dependent. This makes the stress-strain relation
non-linear. Numerical analysis of mechanical behavior must be done iteratively eg. with
a Newton-Raphson scheme. Following an incremental procedure the total loading time is
subdivided into a discrete number of increments, which we assume to be of equal length
t. All relevant variables {, , p, p, y, q} are assumed to be known at the beginning tn of
the current increment. Starting the new increment, it is assumed that the material stiness
equals the Youngs modulus. This provides us with a rst value for the strain at the end
of the current increment, = n+1 = (tn+1). Whether the assumption of elastic material
behavior was correct has to be evaluated during the calculation of the stress n+1.
Depending of = n further elastoplastic deformation or elastic unloading can
occur. Several possibilities are indicated in the gure.
59
(3)
(1)
n + 1
n + 1
n

n
(4)
n
n + 1

(2)
n + 1
n

Fig. 3.24 : Various incremental stress-strain changes


Elastic stress predictor
Because it is not known a priori whether (ongoing) elastoplastic deformation or elastic unload-
ing will have taken place in the current increment tntn+1, the stress calculation starts from
the assumption that the strain increment is completely elastic. The elastic stress predictor e
is calculated and subsequently the yield criterion is evaluated with the yield function f.
e = n +E( n)
1. f = (e qn)2 2
yn 0 elastic increment
2. f = (e qn)2 2
yn > 0 elastoplastic increment
Elastic increment
When the increment is fully elastic, the end-increment stress equals the calculated elastic
stress. As no plastic deformation has occurred during the increment, the eective plastic
strain and the yield stress remain unchanged.
(tn+1) = e ; p(tn+1) = p(tn) = pn
y(tn+1) = y(tn) = yn ; q(tn+1) = q(tn) = qn
60
Elastoplastic increment
If the elastic stress predictor indicates that the yield criterion is violated, the increment is
elastoplastic. The end-increment stress has to be determined by integration of the constitu-
tive equations, such that at the end of the increment the stress satises the yield criterion.
Integration of the stress can be carried out following an explicit or an implicit method.
Explicit stress integration
In the explicit procedure, the change of plastic strain and stress as a function of the change in
total strain, must be known. These relations can be derived from the constitutive equations.
= E{ 2 ( q)}
( q) ( q) q yH p = 0
_

( q){E( 2 ( q))} ( q)K p yH p = 0
( q){E( 2 ( q))} ( q)K2 ( q) yH2 | q| = 0
2 =
( q)E
( q)2E + ( q)2K +yH| q|

p = 2 ( q) =
( q)2E
( q)2E + ( q)2K +yH| q|
=
E
E +K +H
= Sp
= E( p) =
( q)2KE +yHE| q|
( q)2E + ( q)2K +yH| q|
=
E(K +H)
E +K +H
= S
q = K p =
( q)2EK
( q)2E + ( q)2K +yH| q|
=
EK
E +K +H

A scaling factor is calculated from the requirement that ( n) brings us to the yield
curve where = f . This strain to yield f is determined, and the current stress is calculated,
using the current stiness. Calculation of is generalized for tension and pressure. Notice
that sign() is the sign of .
Relevant state variables are calculated subsequently. The index n indicates the begin-
increment time tn. The values for S and Sp depend on the hardening law.
61

f
yn +qn
e

n
qn
Fig. 3.25 : Explicit stress integration by increment splitting
=
| sign( n)yn (n qn)|
|e n|

f
= n +( n)
= n +(e n) +S (
f
)
p = pn +Sp (
f
) ; p = pn +|p pn|
q = q(p) ; y = y(y0, p)
Implicit stress integration
The increments of stress, , and plastic multiplier, , are determined such that the end-
increment stress satises the yield criterion. Because and are not independent, an
iterative procedure has to be used. Only linear kinematic hardening is considered here.
constitutive equations
= E( 2 )
f = 0
_
_
_
backward Euler
= E 2E
f = 0
_
_
_
stress update (iteratively)

+ = E 2E(

+)(

+)
f

+f = 0
_
_
_
+ 2E

+ 2E

+E 2E

f = f

_
_
_
62
The variation of the yield function f can be expressed in variations of and , using
p
= ||.
(1 + 2E

) + (2E

) =

+E 2E

(2

) (2

y)H(2||

) = f

_
_
_
3.3.3 Stiness
The current material stiness can be determined as the derivative of the stress with respect
to the strain.
d
d
= C = S =
E(H +K)
E +H +K
with H =
dy
d p
and K =
dq
dp
3.3.4 Examples
Cyclic loading
The stress-strain behavior is calculated for a prescribed cyclic strain, for linear isotropic and
linear kinematic hardening. Parameters are :
Youngs modulus E 100000 MPa
Poissons ratio 0.3 -
initial yield stress y0 250 MPa
hardening coecient H 5000 MPa
hardening coecient K 5000 MPa
The isotropic hardening leads to an increasingly larger elastic trajectory. After many load
reversals, the behavior will become purely elastic. The kinematic hardening results in a steady
state hysteresis loop.
0.01 0.005 0 0.005 0.01
600
400
200
0
200
400
600

[M
P
a]
delplexihexes
0.01 0.005 0 0.005 0.01
300
200
100
0
100
200
300

[M
P
a]
delplexkhexes
Fig. 3.26 : Cyclic stress-strain curve for isotropic and kinematic hardening
63
A truss, with initial length l0 = 100 mm and cross-sectional area A0 = 10 mm2, is loaded
with a prescribed axial elongation/force. The axial force/elongation is calculated. The cross-
sectional area will change as a function of the elongation.
Fig. 3.27 : Tensile loading of truss element
A cyclic axial strain is prescribed and the resulting axial stress is calculated for linear isotropic
and linear kinematic hardening. Material parameters are listed in the table.
0.01 0.005 0 0.005 0.01
600
400
200
0
200
400
600

[M
P
a]
1Delplexih
0.01 0.005 0 0.005 0.01
300
200
100
0
100
200
300

[M
P
a]
1Delplexkh
Fig. 3.28 : Cyclic stress-strain behavior for linear isotropic and kinematic hardening
Example
A prismatic truss is clamped between two rigid walls, as is shown in the gure. The length L
of the truss is 2 1000 [mm], its cross-sectional area A is 100 mm2. The material is elastic
with Youngs modulus E = 200000 N/mm2 as long as the axial stress is below the initial yield
stress of 200 N/mm2. Above this value the material shows linear isotropic hardening with
hardening coecient H = 1000 M/mm2.
In the middle of the truss, in point Q, a point load F is applied, which rst increases
and then is decreased to zero. The displacement of point Q is calculated with tr2d. The
force F as function of the displacement is shown in the gure below.
F
Q
L L
64
0 5 10 15 20 25 30
0
1
2
3
4
5
x 10
4

F
[N
]
1Delplopgmm1
Fig. 3.29 : Force F versus displacement of point Q.
Example
A structure of three trusses is loaded by a vertical displacement. When the axial stress exceeds
a certain limit value, a trusses will deform plastically.
L
L
a
b
c

F
L
0 0.2 0.4 0.6 0.8
0
100
200
300
400
500
600
700

F
[N
]
1Delpl3baref
0 0.2 0.4 0.6 0.8
0
50
100
150
200
250
300

F
[N
]
1Delpl3bareff
Fig. 3.30 : Force F versus displacement .
65
3.4 Linear viscoelastic material behavior
Viscoelastic materials show time dependent behavior. When during a tensile test the stress/-
strain is prescribed stepwise, the strain/stress will not react immediately, but show a delayed
response, which is called creep/relaxation.
The gures show characteristic behavior for a so-called viscoelastic liquid (permanent
strain for t) and a viscoelastic solid (no strain for t).
Here, we only consider linear viscoelastic behavior and also assume that strains are small.
constant
0
0
p

t t
t t
Fig. 3.31 : Creep and relaxation for viscoelastic liquid and viscoelastic solid
Boltzmann integral
Linear viscoelasticity is characterized by two properties, which are explained by considering
a strain response to a stress excitation.
1. proportionality : At every time the strain response is proportional to the amplitude
of a constant stress step which is applied at t = t0 : i(t) = i D(t t0) for t t0
For a unit step at t = 0 the strain response is identied as the creep function D(t).
2. superposition : The strain response to two subsequently (at time t = t0 and t = t1)
applied constant amplitude (0 and 1) stress steps equals the sum of the separate
responses for t > t1 : (t) = 0D(t t0) +1D(t t1) for t t1
Every stress excitation can be seen as an innite sequence of innitesimal small stress steps.
The superposition property then leads to the Boltzmann integral, expressing the strain re-
sponse. This integral is also called Duhamel or memory integral.
For strain excitation and stress response the same observations can be made. In that
case the relaxation function E(t) describes the characteristic behavior.
66

t2 t1 t t0

t t0
0
Fig. 3.32 : Superposition of strain responses to subsequent stress excitations
strain response on stress excitation
(t) = 0D(t t0) +1D(t t1) +2D(t t2) +..
=
n
i=1
iD(t ti) limit n (t)
=
_ t
=t
0
D(t ) d() =
_ t
=t
0
D(t )
d()
d
d
(t) =
_ t
=t
0
D(t ) () d
stress response on strain excitation
(t) =
_ t
=t
0
E(t ) () d
Creep and relaxation
Step excitations are important for the (experimental) characterization of viscoelastic materi-
als. A unit step (amplitude = 1) can be described with the Heaviside function.
The derivative of the unit step function is the Dirac function or unit pulse. It has the
important property that integration of the product of a function f() and (, t) over an
interval which contains = t, the location of the Dirac pulse, results in the value f(t).
67
H
t t 0
1
t t 0

Fig. 3.33 : Unit step and unit pulse


Heaviside function
_
t < t : H(t, t) = 0
t > t : H(t, t) = 1
Dirac function (t, t

) =
d
dt
{H(t, t

)}
properties
_ t>t
=0
(, t

) d = 1 ;
_ t>t
=0
f()(, t

) d = f(t

)
The strain response to a stress step excitation at t = 0 having an amplitude 0 equals 0D(t).
The creep function D(t) can be measured and experiments have revealed the following prop-
erties :
D(t) 0 t 0 D(t) < 0 t 0
The measured strain response can be used to t a proposed model for D(t).
The stress response to a strain step excitation at t = 0 having an amplitude 0 equals
0E(t). The relaxation function E(t) can be measured and experiments have revealed the
following properties :
E(t) 0 t 0 E(t) > 0 t 0

_
t=0
E(t) dt 0 limt
E(t) = 0
The measured stress response can be used to t a proposed model for E(t).
stress step (t) = 0H(t, 0) (t) = 0(t, 0)
(t) =
_ t
=0
D(t ) () d =
_ t
=0
D(t )0(, 0) d = 0D(t)
strain step (t) = 0H(t, 0) (t) = 0(t, 0)
(t) =
_ t
=0
E(t ) () d =
_ t
=0
E(t )0(, 0) d = 0E(t)
68
3.4.1 Viscoelastic models
The response of a viscoelastic material is given by the Boltzmann integral and to calculate
it we need the creep and/or relaxation functions D(t) and E(t). Mathematical expressions
can be chosen for these functions taking into account some general requirements. The chosen
functions can than be tted onto data from creep and relaxation tests. Instead of choosing
rather arbitrary functions, they are generally derived from the behavior of one-dimensional
mechanical spring-dashpot systems. Simple systems like the Maxwell, Kelvin-Voigt and Stan-
dard Solid element, are not always useful, because the lack of parameters prohibits a good t
of experimental data. In practice Generalized Maxwell or Generalized Kelvin-Voigt models
are used.
Because creep and relaxation tests may need a long experimental time period and ac-
curacy is not high, harmonic excitation tests are carried out to determine the storage and
loss modulus E() and E() or the storage and loss compliance D() and D(). These
parameters can be converted to E(t) and D(t). These experiments are generally known as
D(ynamic) M(echanical) A(nalysis) or D(ynamic) M(echanical) T(hermal) A(nalysis), be-
cause time-temperature superposition is mostly used.
In the following we will study some mechanical models. Their behavior is described by
a dierential equation. Solving this for stress or strain excitations results in the viscoelastic
material functions.
Maxwell model
One of the simplest models to describe linear viscoelastic material behavior is the Maxwell
model. It consists of a spring (modulus E) and a dashpot (viscosity ) in series.
The stress and strain for the Maxwell element are related by a rst-order dierential
equation.

Fig. 3.34 : Maxwell model


= E + = E + =

E +

For a general stress and strain excitation the dierential equation of the Maxwell model can be
solved using appropriate initial conditions. These general solutions are Boltzmann integrals,
which can be used to calculate strain/stress responses to stress/strain excitations.
69
The creep and relaxation functions of the Maxwell model are readily recognized in the
integrals. Response to step excitations reveals that the Maxwell model describes viscoelastic
uid behavior, characterized by a time constant =

E [s].
(t) =
1

_ t
=
_
(t ) +

E
_
() d =
_ t
=
D(t ) () d
(t) =
_ t
=
_
Ee

E
(t)
_
() d =
_ t
=
E(t ) () d
Kelvin-Voigt model
The Kelvin-Voigt model is a simple model for the description of linear viscoelastic material
behavior. It consists of a spring (modulus E) parallel to a dashpot (viscosity ). The stress
and strain for the Kelvin-Voigt element are related by a rst-order dierential equation. For
strain excitation, this equation directly describes the stress response.

Fig. 3.35 : Kelvin model


= E + = E +
The general solution for the strain response to a stress excitation is given by a Boltzmann
integral, in which we recognize the creep function of the Kelvin-Voigt element. Strain response
to a step excitation of stress reveals that the Kelvin-Voigt model describes viscoelastic solid
behavior, characterized by the time constant =

E [s]. A stepwise strain excitation leads to
innite stress.
(t) =
1
E
_ t
=
_
1 e

E
(t)
_
() d =
_ t
=
D(t ) () d
(t) = E(t) + (t)
70
Standard Solid model
The Standard Solid model consists of a parallel arrangement of a Maxwell element (modulus
E, viscosity ) and a linear spring (modulus E). This model incorporates the Maxwell
model (E = 0) and the Kelvin-Voigt model (E = 0). The stress-strain relation is described
by a dierential equation, which can be solved resulting in Boltzmann integrals for strain and
stress.
E
E

v e
Fig. 3.36 : Standard Solid model
constitutive relations
= +ve
= v + e
v =
1

ve
ve = Ee e =
1
E
ve
=
1
E

constitutive equation
= +ve = E + v
= E +( e) = E +
ve
E
= E +

E
( E )
+

E
= E +
(E +E)
E

In the Boltzmann integrals for strain and stress, the creep and relaxation functions of the
Standard Solid element are readily recognized.
The time constant for creep is dened as c =

E
+

E
and the time constant for relax-
ation as r =

E
. They represent the intersection point of the tangent to the creep/relaxation
curve at t = 0 and the asymptote for strain (
0
E
) and stress (0 E), respectively.
71
(t) =
_ t
=
_
1
E

E
E(E +E)
e

EE
(E+E) (t)
_
() d =
_ t
=
D(t ) () d
(t) =
_ t
=
_
E +Ee

E
(t)
_
() d =
_ t
=
E(t ) () d
Generalized Maxwell model
Both the Maxwell and the Kelvin-Voigt models are too simple to describe the viscoelastic
behavior of real materials. Combining a number of Maxwell elements in a parallel congura-
tion, leads to the generalized Maxwell model, which mostly also has an extra parallel spring
for the correct description of long-term behavior of viscoelastic solid materials. Such a model
is generally used for experimental characterization of the behavior of linear viscoelastic ma-
terials in a Dynamic Mechanical (Thermal) Analysis (DM(T)A) test.
The creep function E(t) is easily determined and has a number of time constants to
characterize the viscoelastic material response. A model like the generalized Maxwell model
is therefore also referred to as multi mode.
E1 E2
1 2
E
Fig. 3.37 : Generalized Maxwell model
E(t) = E +
N
i=1
Eie

t
i ; i =
i
Ei
equilibrium modulus E = lim
t
E(t)
glass modulus Eg = lim
t0
E(t) = E +
N
i=1
Ei
Generalized Kelvin model
The generalized Kelvin model consists of a number of Kelvin-Voigt elements arranged in
series. An extra spring sometimes a dashpot is also provided.
72
Eg
E1 E2
2 1
Fig. 3.38 : Generalized Kelvin model
D(t) = Dg +
N
i=1
Di(1 e

t
i ) ; Di =
1
Ei
; i =
i
Ei
glass compliance Dg =
1
Eg
= lim
t0
D(t)
equilibrium compliance D = lim
t
D(t) = Dg +
N
i=1
Di
3.4.2 Stress update
The current stress is given by a Boltzmann integral over the strain history.
Using a Generalized Maxwell model to specify the relaxation function E(t), an expression for
(t) can be derived.
(t) =
t _
=0
E(t ) () d
E(t) = E +
N
i=1
Eie

t
i
_
_
_

(t) =
t _
=0
_
E +
N
i=1
Eie

t
i
_
() d = E(t) +
N
i=1
t _
=0
Eie

t
i () d
= E(t) +
N
i=1
i(t)
Time discretization
In the numerical analysis of the time dependent behavior, the total time interval [0, t] is
discretized :
[0, t] [t1 = 0, t2, t3, .., tn, tn+1 = t]
73
The timespan between two discrete moments in the time interval is a time increment. It is
assumed that these increments are of equal length :
t = tj+1 tj ; j = 1, ..., n
The hereditary integral is now split in an integral over [0, tn] and an integral over the last or
current increment [tn, tn+1 = t].
(t) = E(t) +
N
i=1
i(t) = E(t) +
N
i=1
t _
=0
Eie

t
i () d
= E(t) +
N
i=1
_
_e

t
i
tn _
=0
Eie

tn
i () d +Ei
t _
=tn
e

t
i () d
_
_
Linear incremental strain
For further evaluation of (t) it is assumed that the strain is a linear function of time in each
time increment. For the current increment we have :
() = (tn) + ( tn)

t
() =

t
The integral over the current increment can now be evaluated very easily.
t _
=tn
e
t
i () d =

t
t _
=tn
e
t
i d =

t
i
_
1 e
t
i
_
Stress
Calculating the current stress does not mean that the Boltzmann integral has to be evaluated
over the total deformation history. When results are stored properly we can easily update
the stress (t).
(t) = E(t) +
N
i=1
i(t)
= E(t) +
N
i=1
_
_e
t
i
tn _
=0
Eie

tn
i () d+ Eii
_
1 e
t
i
_

t
_
(t) = E(t) +
N
i=1
_
e
t
i i(tn) +Eipi
_
74
with pi =
i
t
_
1 e
t
i
_
3.4.3 Stiness
The current stiness of the material is the derivative of the stress with respect to the strain.
(t) = E(t) +
N
i=1
_
e
t
i i(tn) +Eipi
_

d
d
= C = E +
N
i=1
Eipi
3.4.4 Viscoelastic : dierential formulation
The dierential equation for a viscoelastic material model can be solved numerically. This is
illustrated for the Standard Solid model.
constitutive equation
+

E
= E +
(E +E)
E

+A = B +C
stress update, implicit backward Euler
t +A = tB +C
(t +A) = An +tB +C
=
1
t +A
[An +tB +C]
3.4.5 Examples
Strain step excitation
A strain step with an amplitude of 0.1 is applied and the stress response is calculated for the
Maxwell and the Standard-Solid models. Rather ctitious values for the material parameters
are chosen.
Maxwell E = 0 E1 = 1 1 = 0.01
Standard-Solid E = 1 E1 = 1 1 = 0.01
75
The initial stress can be veried, using the strain amplitude and the initial stiness. The nal
stress value can be veried, using the strain amplitude and the equilibrium modulus.
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0.02
0.04
0.06
0.08
0.1
t

[M
P
a]
dvielexmaiets
0 0.05 0.1 0.15 0.2
0
0.05
0.1
0.15
0.2
t

[M
P
a]
dvielexssiets
Fig. 3.39 : Stress response for Maxwell and Standard-Solid model
Response of various models
When stress or strain is prescribed as a function of time, the strain or stress can be calculated.
The examples show stress responses for a prescribed strain excitation, being a strain step
(0 = 0.01) followed by a constant strain rate ( = 0.1 [s1]).
The stress response is calculated, using a Maxwell, a Kelvin-Voigt, a Standard-Solid and
a 2-mode model. Parameter values are listed in the table.
E E1 1 E2 2
Maxwell 0 100 0.1 0 0 0
Kelvin-Voigt 100 1010 1012 0 0 0
Standard-Solid 100 100 0.1 0 0 0
2-mode 100 100 0.1 100 0.1 0
0 0.5 1 1.5 2 2.5
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
t [s]

1Dvielexet
Fig. 3.40 : Prescribed strain
76
0 0.5 1 1.5 2 2.5
0
0.2
0.4
0.6
0.8
1
t [s]
[M
P
a]
0 0.5 1 1.5 2 2.5
0
2
4
6
8
10
12
14
16
18
t [s]
[M
P
a]
Fig. 3.41 : Stress response for Maxwell and Kelvin-Voigt model
0 0.5 1 1.5 2 2.5
0
2
4
6
8
10
12
14
16
18
t [s]
[M
P
a]
0 0.5 1 1.5 2 2.5
0
5
10
15
20
t [s]

[M
P
a]
Fig. 3.42 : Stress response for Standard-Solid and 2-mode model
Multi-mode model response
An axial strain step with amplitude 0.01 is prescribed on an tensile bar with initial cross-
sectional area A0 = 10 mm2. The stress response is calculated for a 12-mode generalized
Maxwell model. The modal parameters are listed in the table.
E [MPa] [s] E [MPa] [s]
1 3.0e6 3.1e-8 2 1.4e6 3.0e-7
3 3.9e6 3.0e-6 4 5.4e6 2.9e-5
5 1.3e6 2.8e-4 6 2.3e5 2.7e-3
7 7.6e4 2.6e-2 8 3.7e4 2.5e-1
9 3.3e4 2.5e+0 10 1.7e4 2.4e+1
11 8.0e3 2.3e+2 12 1.2e4 2.2e+3
77
0 0.5 1 1.5
3000
3500
4000
4500
5000
5500
t [s]

[M
P
a]
1Dvielex12m
Fig. 3.43 : Tensile stress versus time
78
3.5 Creep behavior
The phenomenon called creep is the deformation under constant load. Viscoelastic material
shows this behavior. The term creep however is especially reserved for deformation at tem-
peratures, which are considered to be high with respect to the melting temperature of the
material, e.g. T > 0.2Tm. Such high temperatures are encountered in eg. (jet) engines and
heat exchangers. Some materials with a low melting point, like lead (Pb), show creep at room
temperature.
For the range T < 0.4Tm, the so-called stage I or primary creep is observed, where the
strain rate decreases as a function of time. At higher temperatures, 0.5Tm < T < 0.6Tm,
the strain rate is constant for stage II or secondary creep. At still higher temperatures,
0.6Tm < T < 0.8Tm, we have stage III or tertiary creep. This leads eventually to creep
fracture.
III II

t
I
Fig. 3.44 : Creep strain as a function of time at constant stress
Creep strain rate
Creep behavior is described by relating the creep strain rate c to stress, temperature and time.
The temperature dependency could be included by making material parameters a function
of temperature. It is more convenient, however, to implement the temperature dependency
explicitly in the creep model. Much used are so-called power law models.
general model c = Af()f(c)fT (T)ft(t)
power law model c = A
m

n
c T
p
_
qt
q1
_
Primary creep
Primary creep also referred to as stage I, transient creep or delayed elastic eect is
observed at T < 0.4Tm. Mechanisms, which are associated with this behavior are dislocation
coalescence and dislocation entanglement, leading to slip steps (jogs). Dislocations may pile-
up at grain boundaries and impurities. All this leads to macroscopic hardening. When
temperature is higher, 0.4Tm < T < 0.5Tm, the thermal activity of dislocations is higher and
79
a transition to secondary creep is seen.
Primary creep can be modeled as time dependent plasticity. A simple model for the
creep strain rate is :
c = C() exp[()t]
Secondary creep
Secondary creep also referred to as stage II, steady-state creep or viscous ow is observed
at 0.5Tm < T < 0.6Tm. The hardening, which is apparent in primary creep is balanced by
recovery, leading to thermal softening. The thermal energy leads to vacancy movement (self
diusion) and this causes dislocation movement (climb). The moving dislocations can anni-
hilate, align and/or pass obstacles. More drastic recovery may be caused by recrystallization,
which can occur when internal stresses exist.
The temperature dependency is generally included with an Arrhenius function exp(Qc/kT),
where Qc is the creep actuation energy and k is Boltzmanns constant. The available evidence
indicates that stage II creep is diusion controlled, and so in the models the activation energy
for creep, Qc, can often be replaced by the activation energy for self diusion Qsd.
Most models to for stage II creep are based on the ve-power-law creep law. For tem-
peratures below 0.5-0.6 Tm a transition toward primary creep is observed, which in reference
to the modeling is called power-law-breakdown. Sometimes a threshold stress is introduced
below which no creep can be measured.
c = A exp
_

Qc
kT
__

E
_n
(n 5)
Tertiary creep
Tertiary creep also referred to as stage III or accelerating creep is observed at 0.6Tm <
T < 0.8Tm and is associated with geometric instabilities and damage.
One mechanism is grain boundary sliding and subsequent void initiation and coalescence,
leading to inter granular cracks.
Another mechanism is diusional ow, which occurs mainly at higher temperatures and
lower stresses. Two possibilities are : 1) diusion through grains (Nabarro-Herring creep) with
slow vacancy jump frequencies along many paths, and 2) diusion along grain boundaries
(Coble creep) with high vacancy jump frequencies along a few paths.
Stage III creep is often modeled with continuum damage mechanics, where a damage
variable () is used to model internal damage, which inuences the creep strain rate. An
evolution equation is required to control the damage growth as a function of stress and/or
strain.
80
c
c0
=
(/0)n
(1 )m
;

0
=
(/0)
(1 )
(n )
with = m = n
c
c0
=
_
(/0)
(1 )
_n
;

o
=
(/0)
(1 )n
integrating failure time t( = 1) =
1
n + 1
1
0
_
0

_
3.5.1 Creep models
The discrete mechanical model for creep is a Maxwell element with a non-linear dashpot.
The viscous or creep strain rate may be a function of stress , total creep strain c, absolute
temperature T and time t.
c e
E

Fig. 3.45 : Creep model


constitutive relations
= e + c
= Ee e =
1
E

c = Af() fc(c) fT (T) ft(t) = f(, c, T, t)
constitutive equation
= E e = E E c = E Ef(, c, T, t)
Stress functions
Several authors have reported various functions f to implement the inuence of the stress.
Norton; Bailey (1929) c = K
n
Hooke-Norton c =

E
+K
n
81
Johnson et.al. (1963) c = D1
n1 +D2
n2
Dorn (1955) c = Bexp()
Soderberg (1936) c = B
_
exp
_

0
_
1
_
Prandtl (1928) c = Asinh
_

0
_
Garofalo (1965) c = A
_
sinh
_

0
__n
Lemaitre, Chaboche (1985) c =
_

0
_N0
exp
_

N0+1
_
Temperature functions
Several authors have reported various functions fT to implement the inuence of the temper-
ature. These creep models also take into account the dependency of stress and (sometimes)
time.
Kauzmann (1941) c = A exp
_

H
RT
_
Lifszic (1963) c =

T
exp
_

H
RT
_
Dorn, Tietz (1949/55) c = f
_
t exp
_

H
RT
__
f()
Penny, Marriott (1971) c =
_
t exp
_

H
RT
__n
f()
Boyle, Spence (1983) c = C exp
_

H
RT
_
t
m

n
Time functions
Several authors have reported various functions ft to implement the inuence of the time.
Andrade (1910) c = ln
_
1 +t
1
3
_
+kt
Andrade (small ) c = t
1
3 +kt t
1
3
Bailey (1935) c = Ft
n
Graham, Walles (1955) c =
M
j=1
ajt
mj
82
McVetty (1934) c = G(1 exp(qt)) +Ht
Findley et.al. (1944) c = 1 +2t
n
(n < 1)
Pugh (1975) c =
a1t
1 +b1t
+
a2t
1 +b2t
+ mt
3.5.2 Stress update
The constitutive equation, can be solved explicitly or implicitly. For the latter case, a Newton
iteration procedure must be implemented to calculate the stress.
constitutive equation = E Ef(, c, T, t)
explicit = n +E En t Ef(, cn, Tn, tn)
implicit (backward Euler)
= n +E En t Ef(, c, T, t) = e t Ef(, c, T, t)
g() = +t Ef(, c, T, t) = e
g

+
g

= e
_
1 +t E
f

_
= e g

3.5.3 Stiness
In a structural analysis the equilibrium state at a given load is found iteratively. In each
iterative step not only the stress has to be calculated from the deformation, but also the
stiness has to be known. The material stiness is the relation between a variation of the
stress and a variation of the strain.
constitutive equation = E Ef(, c, T, t)
implicit integration
n E +En +t Ef(, c, T, t) = 0
+t Ef(, c, T, t) E n +En = 0
variation
+t E
f

E = 0
_
1 +t E
f

_
= E

=
_
1 +t E
f

_1
E
83
3.5.4 Examples
Creep versus viscoelasticity
Linear viscoelastic behavior can be modeled with a multi-mode Maxwell model, represented
by a mechanical system, which has a number of parallel Maxwell elements and one parallel
spring. Springs and dashpots are linear.
Creep behavior is modeled with one Maxwell model with a nonlinear dashpot. The
viscosity is a nonlinear function of stress, creep strain, temperature and time.
The Norton model for secondary creep can be made equivalent to the linear Maxwell
model.
Maxwell model (E, )
= e +c ; E(t) = Ee
t/
; =

E
; c =

; e =

E
Norton model (A, m)
= e +c ; c = f(, c, T, t) c = A
m
; e =

E
equivalence
Maxwell E = 109 = 109 = 1
Norton E = 109 A =
1
= 109 m = 1
0 2 4 6 8 10
0
2
4
6
8
10
x 10
6
t

[M
P
a]
dvielexmciets
0 2 4 6 8 10
0
2
4
6
8
10
x 10
6
t

[M
P
a]
delviexnlimts
Fig. 3.46 : Stress response for Maxwell viscoelastic and Norton creep model
General creep model for SnAg-solder
Evans and Wilshire and later Maruyama and Oikawa proposed a general model, which de-
scribes the primary, secondary and tertiary creep of alloys. Parameters in the model must be
tted onto experimental data.
84
The creep strain at time t is described by two terms. The rst one describes the hard-
ening or primary creep stage and the second describes the weakening or tertiary creep stage.
Combined, they characterize also the transition region, the secondary creep.
c(t) = 0 +A()
_
1 e
(,T)t
_
+B(, T)
_
e
(,T)t
1
_
(, T) = c1 [sinh()]
n1 e

Q1
RT
A() = c2
n2 ; B(, T) = c3
n3e

Q2
RT
From the general model for the creep strain the creep strain rate c can be calculated and
subsequently the initial creep rate c,i, the time tm for the minimum creep rate c,m to occur
and the strain c,m at that time.
With the universal gas constant R = 8.314 and stress in MPa and Qin kJ/mol, parameter
values for SnAg-solder are tted on experimental data and listed in the table. The absolute
temperature is assumed to be T = 398 [K].
c = Ae
t
+Be
t
c,i = c(t = 0) = (A +B) ; tm =
1
2
ln
_
A
B
_
c,m = c(t = tm) = 2

AB ; c,m = c(t = tm) = 0 +A B


0 0
c1 1.73 105 n1 4.66
0.095 Q1 70
c2 2.06 103 n2 1.1
c3 9.65 104 n3 2.38
Q2 17.8
0 50 100 150 200
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
t [s]

Fig. 3.47 : Creep strain at constant stress = 20 MPa


85
Special creep model for SnAg-solder
Several creep models for SnAg-solder have been published in literature [?]. The 2-term model
of Wiese (2005) is one of them and its parameter values have been tted on experimental
data for Sn4Ag0.5Cu solder material. Parameter values are listed in the table. Temperature
(T) is in oK and equivalent stress () is in MPa. The absolute temperature is assumed to be
T = 398 [K].
c = A1
m1e
e1/T
+A2
m2e
e2/T
E = 59.533 66.667 T
A1 = 4.107 m1 = 3 e1 = 3223
A1 = 1.1012 m1 = 12 e1 = 7348
0 50 100 150
0
5
10
15
20
25
30
t [s]

[M
P
a]
delviexwiimts
Fig. 3.48 : Relaxation stress for constant strain = 0.001
86
3.6 Viscoplastic material behavior
In many forming processes the deformation rates are small enough to consider the material
behavior to be independent of strain rate and to use an elastoplastic material model. For
high strain rates this assumption leads to faulty results. In a tensile test the yield stress is
seen to increase with higher strain rates.

Fig. 3.49 : Strain rate dependent plastic behavior


Polymers and certain metallurgical alloys show softening behavior immediately after reaching
the yield point. At larger strains the softening is followed by hardening. The complete
stress-strain behavior is strain rate dependent, but the initial yield stress is constant.


Fig. 3.50 : Softening and resumed hardening
3.6.1 Viscoplastic (Perzyna) model
The Perzyna model is a genuine viscoplastic model because it has a yield criterion, expressed
with a yield function f. The model is called an over-stress model because f > 0 may occur.
This is dierent compared to elastoplastic models, which always require f 0. The rate of
the viscoplastic multiplier , , cannot be calculated from a consistency equation, but is given
by a separate equation. We will only consider isotropic hardening.
The discrete mechanical model for viscoplastic material behavior consists of a spring
E in series with a parallel arrangement of a hardening spring H, a linear dashpot and a
friction slider, opening at = y.
87

E
y
e vp
H

Fig. 3.51 : Discrete model for viscoplastic material behavior


f = y with f < 0 elastic
f 0 viscoplastic
y = y(y0, vp) = ||
= e + vp
= Ee e =
1
E

vp =
f

=


; vp = | vp|
vp =
_ t
=0
vp d
= (f) = (f/y0)
N
Constitutive equations
From the constitutive relations a set of constitutive equations can be derived. The stress and
the viscoplastic multiplier must be determined by integration of these equations.
= E e = E( vp) = E{


}
= = (f/y0)
N
_
Hardening laws
Various hardening laws can be used in the Perzyna model. Polymer materials may show
softening, as is the case with Polycarbonate (PC). Parameters must be determined experi-
mentally in a compression test, because the material softening poses problems of localization
(necking) in a tensile test.
88
linear hardening y = y0 +H p
softening y = y0 +H vp +a
2
vp +b
3
vp +c
4
vp +d
7
vp
3.6.2 Stress update
In the viscoplastic Perzyna model the stress and viscoplastic multiplier have to be solved
from a set of dierential equations. This equations are nonlinear although the viscosity in the
model is constant.
Numerical analysis of mechanical behavior must be done iteratively, e.g. with a Newton-
Raphson scheme. Following an incremental procedure the total loading time is subdivided
into a discrete number of increments, which we assume to be of equal length t. All rele-
vant variables {, , vp, vp, y} are assumed to be known at the beginning tn of the current
increment. Starting the new increment it is assumed that the material stiness equals the
Youngs modulus. This provides us with a rst value for the strain at the end of the current
increment, = n+1 = (tn+1). Whether the assumption of elastic material behavior was
correct has to be evaluated during the calculation of the stress n+1.
Depending of = n further elastoviscoplastic deformation or elastic unloading can
occur.
Elastic stress predictor
Because it is not known a priori whether (ongoing) elastoviscoplastic deformation or elastic
unloading will occur in the current increment tntn+1, the stress calculation starts from the
assumption that the strain increment is completely elastic. The elastic stress predictor e is
calculated and subsequently the yield criterion is evaluated with the yield function f.
elastic stress e = n +E( n)
1. f = e yn 0 elastic increment
2. f = e yn > 0 elastoviscoplastic increment
Elastic increment
When the increment is fully elastic, the end-increment stress equals the calculated elastic
stress. As no viscoplastic deformation has occurred during the increment, the eective vis-
coplastic strain and the yield stress remain unchanged.
(tn+1) = e ; vp(tn+1) = vp(tn) = vpn
y(tn+1) = y(tn) = yn
89
Elastoviscoplastic increment
If the elastic stress predictor indicates that the yield criterion is violated, the increment
is elastoviscoplastic. The end-increment stress has to be determined by integration of the
constitutive equations. Integration of the stress is done following an implicit method.
Implicit stress integration
When the increment appears to be elastoviscoplastic, the end-increment stress must be up-
dated from the elastic trial stress. The viscoplastic multiplier and the stress are determined
such that the constitutive equations are satised. Because and are not independent,
an iterative procedure has to be used. Here, a simple stepping algorithm is shown. The
equations could also be solved fully coupled.
constitutive equations
= E
_

_


__
=
_
_
_
backward Euler
= E E
_


_
= t
_
_
_
stress update (iteratively)

+ = E E(

+)
_

+ = t (

+)
_
_
_
The variation of the function can be expressed in variations of and , using its denition
and
vp
=
_


_
.
+E
_


_
= E

+ = t

+t

+t


_
_
_

=
d
df
df
dy
y
d p
p
d
=
d
df
df
dy
_

=
d
df
df
d
=
d
df
_


_
+E
_


_
= E

+E E

_
t
d
df
_


__
+
_
1 +t
d
df
_


_
dy
d vp
_
=

+t

_
_
_
d
df
= N
_
f
y0
_N1
1
y0
;
dy
d p
= H
90
+E
_


_
= E

+E E

_
t N
_
f
y0
_N1
1
y0
_


_
_
+
_
1 +t N
_
f
y0
_N1
1
y0
_


_
H
_
=

+t

_
_
_
3.6.3 Stiness
The material stiness is calculated as C = C =
d
d . For this purpose the iterative stress
update algorithm is considered again.
= E E


= t
_
_
_
n = E( n) E( n)


n = t
_
_
_
= E E
1
= t
d
df
f

+t
d
df
f
y
dy
d vp

_
_
_
= E E
_


_
_
t
d
df
f

_
_
1 t
d
df
f
y
dy
d vp
_
= E E
_
t
d
df
_
_
1 +t
d
df
dy
d vp
_
_
_1 +
_
E t
d
df
_
_
1 +t
d
df
dy
d vp
_
_
_ = E
_
_
1 +t
d
df
dy
d vp
+E t
d
df
1 +t
d
df
dy
d vp
_
_ = E
d
d
= C =
E
_
1 +t
d
df
dy
d vp
_
1 +t
d
df
dy
d vp
+E t
d
df
91
3.6.4 Examples
Tensile test at various strain rates
The Perzyna model is loaded with a linearly increasing strain. The hardening model and
tabulated data for polycarbonate are used. The strain rate is varied.
y = v0 +H vp +a
2
vp +b
3
vp +c
4
vp +d
7
vp
E 1800 MPa 0.37 -
y0 37 MPa H -200 MPa
0.001 1/s N 3 -
a 500 MPa b 700 MPa
c 800 MPa d 30000 MPa
=
_
10
3
, 10
2
, 10
1
, 1
_
1 1.1 1.2 1.3 1.4
0
50
100
150
200
250
300
350
400

[M
P
a]
dperzexpcxxes


dGldt = 1
dGldt = 0.1
dGldt = 0.01
Fig. 3.52 : Stress versus stretch ratio for dierent strain rates
A truss is loaded axially with a prescribed elongation. In the initial state the length of the
truss is l0 = 100 mm and its cross-sectional area is A0 = 10 mm2. The axial force/elongation
is calculated for various material models. The cross-sectional area will change as a function
of the elongation.
Fig. 3.53 : Tensile loading of truss element
92
The Perzyna model for viscoplastic material behavior is used to calculate the axial stress
response for a prescribed axial elongation, which increases linearly in time. The elongation
rate is varied between 0.01 and 1. Parameter values for polycarbonate (PC) are used. Their
values are listed in the table.
E 1800 MPa 0.37 -
y0 37 MPa H -200 MPa
0.001 1/s N 3 -
a 500 MPa b 700 MPa
c 800 MPa d 30000 MPa
strain rate = {0.01, 0.1, 1}[s
1
]
1 1.1 1.2 1.3 1.4
0
50
100
150
200
250
300
350
400

[M
P
a]
1Dperzexls


dGldt = 1
dGldt = 0.1
dGldt = 0.01
1 1.1 1.2 1.3 1.4
0
500
1000
1500
2000
2500
3000

F
[N
]
1Dperzexlf


dGldt = 1
dGldt = 0.1
dGldt = 0.01
Fig. 3.54 : Stress-stretch and force-elongation for PC
93
3.7 Nonlinear viscoelastic material behavior
A polymeric material can be loaded in compression with a constant logarithmic strain rate.
The true stress - absolute value - reaches a maximum value (B) after which softening occurs
(BC) due to structural evolution. Subsequent hardening (CD) results in an increase of the
stress, with increasing strain - absolute value - due to orientation of molecules.
Until the maximum stress level (B) is reached, the deformation is fully reversible. Ini-
tially the material behavior is linear viscoelastic (OA) but from a certain strain, nonlinear
viscoelastic behavior (AB) is observed. After reaching the maximum stress (B), plastic ow
occurs and therefore this stress is called the yield stress y.
For a number of polymers, like polycarbonate (PC), polymethylmethacrylate (PMMA),
polystyrene (PS) and polyethenetereftalate (PET), the above typical stress-strain behavior is
observed.
Two time-dependent processes can be observed, one related to the deformation kinetics
(strain rate dependency) and another related to the aging kinetics.

ln O
A
B
C
D
Fig. 3.55 : Softening and resumed hardening
When the uniaxial compression test for PC is carried out at a higher strain rate, the increase
in stress is equal for each strain value.
This is shown in a graph, where the stress maximum yma, the stress minimum after
softening ymi and the dierence between those two values y, are plotted against the
logarithm of the true strain rate. The stress maximum is referred to as the upper yield stress,
the stress minimum after softening is called the lower yield stress and the dierence is the
yield drop, which is constant for PC.

ln
yma
y
101 105

ymi
Fig. 3.56 : Strain rate dependent stress-strain for PC
94
The strain rate dependency of PMMA is dierent from that of PC. The increase of the stress
with higher strain rates is not the same for each strain value. The upper yield stress increases
more than the lower yield stress. The yield drop is a function of the strain rate.

ln

ymi
y
105 101
yma
Fig. 3.57 : Strain rate dependent stress-strain for PMMA
The yield drop appears to be a function of the history of the material. When the specimen is
quenched after processing, there is no yield drop. Softening is observed after a certain time, a
phenomenon which is called aging. The time is characteristic for the polymer in question : 15
minutes for PS, 1 day for PMMA and about 3 weeks for PC. Aging and the resulting softening
characteristic, can be neutralized by mechanical deformation, indicated as rejuvenation.
ageing
ln

Fig. 3.58 : Aging


3.7.1 Nonlinear viscoelastic model
The complete model for nonlinear viscoelastic behavior is based on the models, which are
used to describe the mechanical behavior at increasing stress level.
Linear viscoelastic behavior
For small strains the material behavior of polymers is linear viscoelastic and can be described
by a Boltzmann integral with multi-mode Maxwell relaxation function. When more molecular
processes are relevant, the relaxation functions for the separate processes can be added.
95
E1 E2
1 2
E
Fig. 3.59 : Generalized Maxwell model for linear viscoelastic behavior
(t) =
t _
=
E(t ) () d ; E(x) = E +
N
i=1
Eie

x
i ; i =
i
Ei
ln

Fig. 3.60 : Stress-strain relation in linear viscoelastic range


Nonlinear viscoelastic behavior
For higher strains, but before yielding, the behavior is nonlinear viscoelastic and the relaxation
function becomes a function of the stress. Fortunately this inuence can generally be modeled
by using time-stress superposition and adaptation of time variables using a time-stress shift
factor.
(t) =
t _
=
E(

) () d
=
t _
=
d
a{()}
;

=
_
=
d
a{()}
E(x) = E +
N
i=1
Eie

x
i() = E +
N
i=1
Eie

x
ia() ; a =
/0
sinh (/0)
96
ln

Fig. 3.61 : Stress-strain relation in nonlinear viscoelastic range


Creep
When the yield stress (= maximum stress) is reached, stress-activated plastic ow occurs,
described by a semi-empirical relation for the viscous strain rate. The stress level depends
on the strain rate and the temperature. When the stress is below the initial yield stress, the
viscosity is very high and the material behavior is considered to be linear elastic with stiness
E.
This behavior can be modeled with a Maxwell model with a linear spring (stiness E)
and a nonlinear dashpot (viscosity ). The associated relaxation function can be written as :
E(t) = E exp
_

t
()
_
; () =
()
E
v e

() E
Fig. 3.62 : Nonlinear creep model
= e + v
= Ee e =
1
E

v = f(, v, T) =

( , T)
; = ||

ln
Fig. 3.63 : Stress-strain relation for creep range
97
Softening
After reaching the initial yield stress the stress decreases asymptotically toward a nal value.
This softening is taken into account by decreasing the viscosity with an internal (damage)
variable D. Initially D = D0 and nally D reaches a saturation value D. The value of
D is determined by an evolution equation, which relates D to v, with v = | v| for the
one-dimensional case.
v e

() E
Fig. 3.64 : Model for nonlinear creep
= e + v
= Ee e =
1
E

v =
1
( , T, D)
; = ||
D =
_
1
D
D
_
h v ; v = | v|

ln
Fig. 3.65 : Softening
Hardening
In a compression test it is observed that the softening is followed by hardening. This can be
modeled by decomposing the stress additively. In the discrete mechanical element a linear
spring (stiness H) is placed parallel to the Maxwell element with linear spring (stiness E)
and nonlinear dashpot (viscosity ).
The total strain is additively decomposed in an elastic strain e and a viscous strain v.
The total axial stress is the sum of the viscoelastic stress s and the hardening stress w. The
viscoelastic stress is related to the stiness E, but also to the viscosity .
98
s
w
(s) E
H

e v
Fig. 3.66 : Model for nonlinear viscoelastic behavior
= e + v
= s +w = Ee +H
v =
1
( s, T, D)
s ; s = |s|
D =
_
1
D
D
_
h v ; v = | v|

= s +w
w
s
ln
Fig. 3.67 : Stress-strain curve for nonlinear
viscoelastic behavior
Aging and hardening
A dierent model to describe aging and softening is based on additive decomposition of the
stress, where the total stress is the sum of the ow stress s, the hardening stress w and
the aging stress y, which is determined by an aging characteristic function S(t, v). This
function is taken to be the product of a time-dependent function Sa(t), which describes the
aging kinetics and a softening function R( v) which describes the softening kinetics. The
viscosity is now a function of this function S(t, v).

ln
y
w
s
y
Fig. 3.68 : Aging and hardening
99
= e + v
= s +y +w = Ee +y +H
v =
1
( s, T, S)
s ; s = |s|
S(t, v) = Sa(t)R( v)
R( v) =
__
1 +
_
r0e
v
_r1_
/ {1 +r
r1
0 }
r21
r1 ; 0 < R < 1
S(t) = Sa(teff ) = c0 +c1 ln
_
teff +ta
t0
_
teff (T, s) =
_ t
=0
d
T (T())( s())
ta = exp
_
S(0) c0
c1
_
y = y(t) y0 =
c
c1
{S(t) c0}
Nonlinear viscoelastic model
The nonlinear viscoelastic material behavior is described by some relations, which can be
combined. The resulting constitutive equations must be solved simultaneously.
s
w
(s) E
H

e v
Fig. 3.69 : Model for nonlinear viscoelastic behavior
= e + v
= s +w = Ee +H
v =
1
( s, T, D)
s ; s = |s|
D =
_
1
D
D
_
h v ; v = | v|
100
constitutive equations
e = v =
1
2( s,T,D) s =
E
2( s,T,D) e
= Ee +H
D =
_
1
D
D
_
h v
_
_
_
Volumetric eects
To include volumetric deformation eects in the one-dimensional model the stress s is ad-
ditively decomposed in a hydrostatic stress sh, associated with volume change J, and a
deviatoric stress sd, determined by isochoric elastic strain d
e. The hardening stress is related
to the total deviatoric strain d.
The compression modulus , the shear modulus G and the hardening modulus Gw are
used as elastic parameters :
=
E
3(1 2)
; G =
E
2(1 +)
; Gw =
H
2(1 +)
When using the logarithmic strain denition, the volume change J = 2 can be written as
J = e(12)e . With the linear strain denition we nd J = (1 2)e + 1.
constitutive relations
= e + v
= s
h
+s
d
+w = (J 1) +G
d
e +Gw
d

d
e =
2
3 (1 +)e ;
d
=
2
3 (1 +)
v =
1
2( s, T, D)
s
d
D =
_
1
D
D
_
h v
constitutive equations
e = v =
1
2( s,T,D) sd =
G
2( s,T,D) d
e
= (J 1) +Gd
e +Gwd
D =
_
1 D
D
_
h v
_
_
_
101
Eyring viscosity
For each material the proper relation for the viscosity has to be chosen. For polymers the
Eyring viscosity function is used. The Eyring viscosity is sometimes dened in a dierent
way with parameters A0 and .
= A0
s

3 sinh
_
s/(

30)
_ exp
_
H
RT
+
p
0
D
_
s = |s| ; p =
1
3s ; 0 =
RT
V
alternative formulation
_
A0e
H
RT
_
e
p
0 e
D
= A0e
p
e
D
A0 = A0 e

H
RT ; = 0
Bodner-Partom viscosity
For metals the Bodner-Partom viscosity function is used. Values for steel and aluminum are
listed in the table.
=

120
exp
_
1
2
_
Z
s
_2n
_
Z = Z1 + (Z0 Z1) exp [m p]
3.7.2 Stress update
The stress is related to the strain rate by a dierential equation, which has to be solved
together with the damage evolution equation. The elastic parameters , G and G
w depend
on the formulation.
For the 1D formulation :
= 0 ; G

= E ; G

w = H
For the 1D version with volumetric eects :
=
E
3(1 2)
; G =
E
2(1 +)
; Gw =
H
2(1 +)
G

= 2
3 (1 +)G ; G

w = 2
3(1 +)Gw
After updating e the stress is calculated directly.
constitutive equations
e = v =
1
2( s,T,D) sd =
G
2( s,T,D) e
= (J 1) +Ge +G
w
D =
_
1 D
D
_
h v
_
_
_
102
stress update (iteratively)
e =
_
n +en
_
/
_
1 +
Gt
2n
_
s
h
= (J 1) ; s
d
= G

e ; w = G

w
= s
h
+s
d
+w ; s = |s| ; p =
1
3s
v = e v =

v vn

D =
_
Dn +h v
D +h v
_
D
= ( s, T, D)
3.7.3 Stiness
The material stiness is calculated as C =
d
d

, where is determined from the
constitutive equations.
= (J 1) +G

e +G

w
= J +G

e +G

w
=
_

dJ
de
+G

_
e +G

w
=
_

dJ
de
+G

_
( v) +G

w
d =
_

dJ
de
+G

_
(d dv) +G

wd
d
d
= C =
_

dJ
de
+G

__
1
v

_
+G

w
3.7.4 Examples
Polymer materials are characterized by an Eyring viscosity. Parameters for various materials
are experimentally determined and listed in the table. The Eyring viscosity is sometimes
dened in a dierent way with parameters A0 and . The temperature determines the value
for A0. For the values in the table the temperature is chosen to be T = 285 K. The universal
gas constant is R = 8.314 J/(mol.K).
For steel and aluminum, the Bodner-Partom viscosity function is used. The material
parameters are listed in the table.
103
PET PC PS PP
E 2400 2305 3300 1092 MPa
0.35 0.37 0.37 0.4 -
H 15 29 13 3 MPa
h 13 270 100 0 -
D 11 19 14 - -
A0 3.8568E-27 9.7573E-27 4.2619E-34 2.0319E-29 s
H 2.617E+05 2.9E+05 2.6E+5 2.2E+5 J/mol
0.0625 0.06984 0.294 0.23 -
0 0.927 0.72 2.1 1.0 MPa
A0 3.5661E21 1.3877E27 1.9230E14 4.2740E11 MPa
0.0674 0.0970 0.140 0.23 -
steel Al
G 7.8E4 2.6E4 MPa
K 1.52E5 7.8E4 MPa
0 1.0E8 1.0E8 s2
n 1.82 3.4 -
m 20 13.8 -
Z0 810 81.4 MPa
Z1 930 170 MPa
Tensile test with increasing strain rates
The model is loaded with a linearly increasing strain. The tabulated data for polycarbonate
are used. The strain rate is varied.
=
_
10
3
, 10
2
, 10
1
, 1
_
0 0.5 1 1.5
0
20
40
60
80
100

[M
P
a]
dleonexpcxxes


0.001
0.01
0.1
1
Fig. 3.70 : Stress versus strain in polycarbonate for dierent strain rates
Tensile test for various polymers
The model is also loaded with a strain rate = 101 s1, using the tabulated parameter
values for polycarbonate, polypropylene, polystyrene and PET.
104
0 0.5 1 1.5
0
20
40
60
80
100

[M
P
a]
dleonexpsxxes


PC
PP
PET
PS
Fig. 3.71 : Stress versus strain for dierent polymers at strain rate 0.1
Chapter 4
Vectors, tensors, columns, matrices
In mechanics and other elds of physics, quantities are represented by vectors and tensors.
Essential manipulations with these quantities will be summerized in this section. For quan-
titative calculations and programming, components of vectors and tensors are needed, which
can be determined in a coordinate system with respect to a vector basis. The three compo-
nents of a vector can be stored in a column. The nine components of a second-order tensor
are generally stored in a three-by-three matrix.
A fourth-order tensor relates two second-order tensors. Matrix notation of such relations
is only possible, when the 9 components of the second-order tensor are stored in columns.
Doing so, the 81 components of a fourth-order tensor are stored in a 9 9 matrix. For
some mathematical manipulations it is also advantageous to store the 9 components of a
second-order tensor in a 9 9 matrix.
4.1 Summary of vector and tensor operations
In this section we give a summary of the most important manipulations and properties of
vectors and tensors and their operations.
4.1.1 Vectors
Essential denitions and manipulations with vectors are summarized below. Three indepen-
dent vectors in three-dimensional space constitute a vector base. In a Cartesian coordinate
system, these base vectors are independent of the Cartesian coordinates {x, y, z}. In a cylin-
drical and a spherical coordinate system, some of the base vectors are a function of one of
more coordinates.
vector: length and direction a = ||a||e ; ||e|| = 1
scalar multiplication a =b
summation a +b =c
scalar product1 a b = ||a||||b|| cos()
vector product c =a b =
_
||a||||b||
_
sin() n ; ||n|| = 1
105
106
triple product a b c =
_
||a||||b|| sin()
_
||c|| cos()
tensor product2 ab = dyad ; q =ab p = p (ab)
c
orthonormal vector base {e1, e2, e3} ; ei ej=i = 0 ; ei ei = 1
vector components column3 a = a

T
e

= e

T
a

1) = smallest angle between a and b


2) index c indicates conjugation
3) index T indicates transposition
4.1.2 Second-order tensors
Essential denitions and manipulations with second-order tensors are summarized below.
second-order tensor A =

i
iai
bi ; A p = q
tensor components matrix A = e

T
Ae

unity tensor unity matrix I a =a a I = e

T
I e

conjugate tensor A
c
=

i
i
biai ; A p = p A
c
scalar product A = B
summation A+B = C
inner product B A = C
double inner product A : B = A
c
: B
c
= scalar
1st invariant J1(A) = tr(A)
2nd invariant J2(A) = 1
2
_
tr
2
(A) tr(A A)
_
3rd invariant J3(A) = det(A) ; det(A) = 0 A = singular
inverse tensor A
1
A = I ; A = regular
symmetric tensor A
c
= A
skew-symmetric tensor A
c
= A
positive denite a A a > 0 a =0
orthogonal tensor (A a) (Ab) =a b a,b
adjugate tensor (A a) (Ab) = A
a
(a b) a,b
4.1.3 Fourth-order tensors
Essential denitions and manipulations with fourth-order tensors are summarized below.
fourth-order tensor
4
A =

i
iai
bici
di ;
4
A : B = C
unity tensor
4
I : A = A A
inner product
4
A B =
4
C
107
4.2 Column and matrix notation
Three-dimensional continuum mechanics is generally formulated initially without using a
coordinate system, using vectors and tensors. For solving real problems or programming,
we need to use components w.r.t. a vector basis. For a vector and a second-order tensor,
the components can be stored in a column and a matrix. In this section a more extended
column/matrix notation is introduced, which is especially useful, when things have to be
programmed.
4.2.1 Matrix/column notation for second-order tensor
The components of a tensor A can be stored in a matrix A. For later purposes it is very
convenient to store these components in a column. To distinguish this new column from the
normal column with components of a vector, we introduce a double under-wave. In this
new column A

the components of A are located on specic places.


As any other column, A

can be transposed. Transposition of the individual column


elements is also possible. When this is the case we write : A

t
.
3 3 matrix of a second-order tensor
A = eiAijej A =
_
_
A11 A12 A13
A21 A22 A23
A31 A32 A33
_
_
column notation
A

T
=
_
A11 A22 A33 A12 A21 A23 A32 A31 A13

T
t
=
_
A11 A22 A33 A21 A12 A32 A23 A13 A31

conjugate tensor
A
c
Aji A
T
=
_
_
A11 A21 A31
A12 A22 A32
A13 A23 A33
_
_ A

t
Column notation for A : B
With the column of components of a second-order tensor, it is now very straightforward to
write the double product of two tensors as the product of their columns.
C = A : B
= eiAijej : ekBklel = AijjkilBkl = AijBji
= A11B11 +A12B21 +A13B31 +A21B12 +A22B22 +A23B32 +
A31B13 +A32B23 +A33B33
=
_
A11 A22 A33 A21 A12 A32 A23 A13 A31

_
B11 B22 B33 B12 B21 B23 B32 B31 B13
T
= A

T
t
B

= A

T
B

t
108
idem
C = A : B
c
C = A

T
t
B

t
= A

T
B

C = A
c
: B C = A

T
B

= A

T
t
B

t
C = A
c
: B
c
C = A

T
t
B

= A

T
B

t
Matrix/column notation C = A B
The inner product of two second-order tensors A and B is a new second-order tensor C.
The components of this new tensor can be stored in a 33 matrix C, but of course also in a
column C

.
A matrix representation will result when the components of A and B can be isolated.
We will store the components of B in a column B

and the components of A in a matrix.


C = A B = eiAikek elBljej = eiAikklBljej = eiAikBkjej
C =
_

_
A11B11 +A12B21 +A13B31
A11B12 +A12B22 +A13B32
A11B13 +A12B23 +A13B33
A21B11 +A22B21 +A23B31
A21B12 +A22B22 +A23B32
A21B13 +A22B23 +A23B33
A31B11 +A32B21 +A33B31
A31B12 +A32B22 +A33B32
A31B13 +A32B23 +A33B33
_

_
C

=
_

_
C11
C22
C33
C12
C21
C23
C32
C31
C13
_

_
=
_

_
A11B11 +A12B21 +A13B31
A21B12 +A22B22 +A23B32
A31B13 +A32B23 +A33B33
A11B12 +A12B22 +A13B32
A21B11 +A22B21 +A23B31
A21B13 +A22B23 +A23B33
A31B12 +A32B22 +A33B32
A31B11 +A32B21 +A33B31
A11B13 +A12B23 +A13B33
_

_
The column C

can be written as the product of a matrix A and a column B

which contain
the components of the tensors A and B, respectively. To distinguish the new matrix from
the normal 33 matrix A, which contains also the components of A, we have introduced a
double underline.
The matrix A can of course be transposed, giving A
T
. We have to introduce, however,
three new manipulations concerning the matrix A. First it will be obvious that the individual
matrix components can be transposed : Aij Aji. When we do this the result is written as
: A
t
, just as was done with a column C

.
109
Two manipulations concern the interchange of columns or rows and are denoted as ( ) c
and ( ) r. It can be easily seen that not each row and/or column is interchanged, but only :
(4 5), (6 7) and (8 9).
C

=
_

_
A11 0 0 0 A12 0 0 A13 0
0 A22 0 A21 0 0 A23 0 0
0 0 A33 0 0 A32 0 0 A31
0 A12 0 A11 0 0 A13 0 0
A21 0 0 0 A22 0 0 A23 0
0 0 A23 0 0 A22 0 0 A21
0 A32 0 A31 0 0 A33 0 0
A31 0 0 0 A32 0 0 A33 0
0 0 A13 0 0 A12 0 0 A11
_

_
_

_
B11
B22
B33
B12
B21
B23
B32
B31
B13
_

_
= AB

idem
C = A B C

= AB

= A
c
B

t
; C

t
= A
r
B

= A
rc
B

t
C = A B
c
C

= AB

t
= A
c
B

C = A
c
B C

= A
t
B

= A
tc
B

t
C = A
c
B
c
C

= A
t
B

t
= A
tc
B

4.2.2 Matrix notation of fourth-order tensor


The components of a fourth-order tensor can be stored in a 9 9 matrix. This matrix has
to be dened and subsequently used in the proper way. We denote the matrix of 4A as A.
When the matrix representation of 4A is A, it is easily seen that right- and left-conjugation
results in matrices with swapped columns and rows, respectively.
4
A = eiejAijklekel
A =
_

_
A1111 A1122 A1133 A1112 A1121 A1123 A1132 A1131 A1113
A2211 A2222 A2233 A2212 A2221 A2223 A2232 A2231 A2213
A3311 A3322 A3333 A3312 A3321 A3323 A3332 A3331 A3313
A1211 A1222 A1233 A1212 A1221 A1223 A1232 A1231 A1213
A2111 A2122 A2133 A2112 A2121 A2123 A2132 A2131 A2113
A2311 A2322 A2333 A2312 A2321 A2323 A2332 A2331 A2313
A3211 A3222 A3233 A3212 A3221 A3223 A3232 A3231 A3213
A3111 A3122 A3133 A3112 A3121 A3123 A3132 A3131 A3113
A1311 A1322 A1333 A1312 A1321 A1323 A1332 A1331 A1313
_

_
4
A
c
A
T
;
4
A
rc
A
c
;
4
A
lc
A
r
110
Matrix/column notation C = 4A : B
The double product of a fourth-order tensor 4A and a second-order tensor B is a second-
order tensor, here denoted as C.
The components of C are stored in a column C

, those of B in a column B

. The
components of 4A are stored in a 9 9 matrix.
Using index-notation we can easily derive relations between the fore-mentioned columns.
C =
4
A : B
eiCijej = eiejAijmnemen : epBpqeq
= eiejAijmnnpmqBpq = eiejAijmnBnm
C

= A
c
B

= AB

t
idem
C = B :
4
A
eiCijej = epBpqeq : emenAmnijeiej
= BpqqmpnAmnijeiej = BnmAmnijeiej
C

T
= B

T
A
r
= B

T
t
A
Matrix notation 4C = 4A B
The inner product of a fourth-order tensor 4A and a second-order tensor B is a new fourth-
order tensor, here denoted as 4C. The components of all these tensors can be stored in
matrices. For a three-dimensional physical problem, these would be of size 9 9. Here we
only consider the 5 5 matrices, which would result in case of a two-dimensional problem.
4
C =
4
A B = eiejAijklekel epBpqeq
= eiejAijkleklpBpqeq = eiejAijklBlqekeq
= eiejAijkpBplekel
C =
_

_
A111pBp1 A112pBp2 A113pBp3 A111pBp2 A112pBp1
A221pBp1 A222pBp2 A223pBp3 A221pBp2 A222pBp1
A331pBp1 A332pBp2 A333pBp3 A331pBp2 A332pBp1
A121pBp1 A122pBp2 A123pBp3 A121pBp2 A122pBp1
A211pBp1 A212pBp2 A213pBp3 A211pBp2 A212pBp1
_

_
=
_

_
A1111 A1122 A1133 A1112 A1121
A2211 A2222 A2233 A2212 A2221
A3311 A3322 A3333 A3312 A3321
A1211 A1222 A1233 A1212 A1221
A2111 A2122 A2133 A2112 A2121
_

_
_

_
B11 0 0 B12 0
0 B22 0 0 B21
0 0 B33 0 0
B21 0 0 B22 0
0 B12 0 0 B11
_

_
= AB
cr
= A
c
B
c
C
r
= A
r
B
r
= A
cr
B
111
Matrix notation 4C = B 4A
The inner product of a second-order tensor and a fourth-order tensor can also be written as
the product of the appropriate matrices.
4
C = B
4
A = eiBijej epeqApqrseres
= eiBijjpeqApqrseres = eieqBijAjqrseres
= eiejBipApjklekel
C =
_

_
B1pAp111 B1pAp122 B1pAp133 B1pAp112 B1pAp121
B2pAp211 B2pAp222 B2pAp233 B2pAp212 B2pAp221
B3pAp311 B3pAp322 B3pAp333 B3pAp312 B3pAp321
B1pAp211 B1pAp222 B1pAp233 B1pAp212 B1pAp221
B2pAp111 B2pAp122 B2pAp133 B2pAp112 B2pAp121
_

_
=
_

_
B11 0 0 0 B12
0 B22 0 B21 0
0 0 B33 0 0
0 B12 0 B11 0
B21 0 0 0 B22
_

_
_

_
A1111 A1122 A1133 A1112 A1121
A2211 A2222 A2233 A2212 A2221
A3311 A3322 A3333 A3312 A3321
A1211 A1222 A1233 A1212 A1221
A2111 A2122 A2133 A2112 A2121
_

_
= BA = B
c
A
r
C
r
= B
r
A
c
= B
cr
A
cr
Matrix notation 4C = 4A : 4B
The double inner product of two fourth-order tensors, 4A and 4B, is again a fourth-order
tensor 4C. Its matrix, C, can be derived as the product of the matrices A and B.
4
C =
4
A :
4
B = eiejAijklekel : epeqBpqrseres
= eiejAijkllpkqBpqrseres = eiejAijqpBpqrseres
= eiejAijqpBpqklekel
C =
_

_
A11qpBpq11 A11qpBpq22 A11qpBpq33 A11qpBpq12 A11qpBpq21
A22qpBpq11 A22qpBpq22 A22qpBpq33 A22qpBpq12 A22qpBpq21
A33qpBpq11 A33qpBpq22 A33qpBpq33 A33qpBpq12 A33qpBpq21
A12qpBpq11 A12qpBpq22 A12qpBpq33 A12qpBpq12 A12qpBpq21
A21qpBpq11 A21qpBpq22 A21qpBpq33 A21qpBpq12 A21qpBpq21
_

_
=
_

_
A1111 A1122 A1133 A1112 A1121
A2211 A2222 A2233 A2212 A2221
A3311 A3322 A3333 A3312 A3321
A1211 A1222 A1233 A1212 A1221
A2111 A2122 A2133 A2112 A2121
_

_
_

_
B1111 B1122 B1133 B1112 B1121
B2211 B2222 B2233 B2212 B2221
B3311 B3322 B3333 B3312 B3321
B2111 B2122 B2133 B2112 B2121
B1211 B1222 B1233 B1212 B1221
_

_
= AB
r
= A
c
B
112
Matrix notation fourth-order unit tensor
The fourth-order unit tensor 4I can be written in matrix-notation. Following the denition
of the matrix representation of a fourth-order tensor, the matrix I may look a bit strange.
The matrix representation of A = 4I : A is however consistently written as A

= I
c
A

.
In some situations the symmetric fourth-order unit tensor 4I
s
is used.
4
I = eiejiljkekel
I =
_

_
1111 1212 1313 1211 1112
2121 2222 2323 2221 2122
3131 3232 3333 3231 3132
1121 1222 1323 1221 1122
2111 2212 2313 2211 2112

_

_
=
_

_
1 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0
0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 1 0 0
0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 1 0
_

_
4
I
s
= 1
2
_
4
I +
4
I
rc_
I
s
= 1
2
_
I +I
c
_
= 1
2
_

_
2 0 0 0 0
0 2 0 0 0
0 0 2 0 0
0 0 0 1 1
0 0 0 1 1

_

_
Matrix notation II
In some relations the dyadic product II of the second-order unit tensor with itself appears.
Its matrix representation can easily be written as the product of columns I

and its transposed.


II = eiijejekklel = eiejijklekel
II =
_

_
1 1 1 0 0 0 0 0 0
1 1 1 0 0 0 0 0 0
1 1 1 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
_

_
= I

T
113
Matrix notation 4B = 4I A
The inner product of the fourth-order unit tensor 4I and a second-order tensor A, can be
elaborated using their matrices.
4
B =
4
I A = eiejiljkekel epApqeq = A
4
I
B =
_

_
A1111 A1212 A1313 A1211 A1112 ..
A2121 A2222 A2323 A2221 A2122 ..
A3131 A3232 A3333 A3231 A3132 ..
A1121 A1222 A1323 A1221 A1122 ..
A2111 A2212 A2313 A2211 A2112 ..
.. .. .. .. .. ..
_

_
=
_

_
A11 0 0 A12 0 ..
0 A22 0 0 A21 ..
0 0 A33 0 0 ..
0 A12 0 0 A11 ..
A21 0 0 A22 0 ..
.. .. .. .. .. ..
_

_
= A
c
Summary and examples
Below the tensor/matrix transformation procedure is summarized and illustrated with a few
examples. The storage of matrix components in columns or blown-up matrices is easily done
with the Matlab .m-les m2cc.m and m2mm.m. (See appendix ??.)
x x

A A ; A

; A
4A A
4I I
x

=
_
_
x1
x2
x3
_
_ ; A =
_
_
A11 A12 A13
A21 A22 A23
A31 A32 A33
_
_
A

=
_

_
A11
A22
A33
A12
A21
..
_

_
; A =
_

_
A11 0 0 0 A12 ..
0 A22 0 A21 0 ..
0 0 A33 0 0 ..
0 A12 0 A11 0 ..
A21 0 0 0 A22 ..
.. .. .. .. .. ..
_

_
A =
_

_
A1111 A1122 A1133 A1112 A1121 ..
A2211 A2222 A2233 A2212 A2221 ..
A3311 A3322 A3333 A3312 A3321 ..
A1211 A1222 A1233 A1212 A1221 ..
A2111 A2122 A2133 A2112 A2121 ..
.. .. .. .. .. ..
_

_
; I =
_

_
1 0 0 0 0 ..
0 1 0 0 0 ..
0 0 1 0 0 ..
0 0 0 0 1 ..
0 0 0 1 0 ..
.. .. .. .. .. ..
_

_
In de Matlab programs the name of a column or matrix indicates its structure as is indicated
below. Some manipulations are introduced, which are easily done in Matlab.
114
A A : 3 3 matrix mA
A A

: column with components of A ccA = m2cc(mA,9)


A A : blown-up matrix mmA = m2mm(mA,9)
A
c
A
T
: transpose mAt = mA
A
c
A

t
: transpose all components ccAt = m2cc(mAt,9)
A
c
A
t
: transpose all components mmAt = m2mm(mA)
4A
lc
A
r
: interchange rows 4/5, 6/7, 8/9 mmAr = mmA([1 2 3 5 4 7 6 9 8],:)
4A
rc
A
c
: interchange columns 4/5, 6/7, 8/9 mmAc = mmA(:,[1 2 3 5 4 7 6 9 8])
c = A : B c = A

T
t
B

C = A B C

= AB

C = 4A : B C

= AB

t
C = B : 4A C

T
= B

T
t
A
4C = 4A B C = AB
cr
4C = 4A : 4B C = AB
r
4I I
II I

T
4.2.3 Gradients
Gradient operators are used to dierentiate w.r.t. coordinates and are as such associated with
the coordinate system. The base vectors unit tangent vectors to the coordinate axes in
the Cartesian system, {ex, ey, ez}, are independent of the coordinates {x, y, z}. Two base
vectors in the cylindrical coordinate system with coordinates {r, , z}, are a function of the
coordinate : {er(), et(), ez}. This dependency has ofcourse to be taken into account when
writing gradients of vectors and tensors in components w.r.t. the coordinate system, using
matrix/column notation.
The gradient of a vector is denoted as a conjugate tensor, whose components can be
stored in a matrix or in a column : La = ( a)
c
La L

a
.
Cartesian = ex

x
+ey

y
+ez

z
=
_
ex et ez
_
_

z
_

_
e

115
cylindrical = er

r
+et
1
r

+ez

z
=
_
er et ez
_
_

r
1
r

z
_

_
= e

Gradient of a vector in Cartesian coordinate system


a = L
c
a =
_
ex

x
+ey

y
+ez

z
_
(axex +ayey +azez)
= exax,xex +exay,xey +exaz,xez +eyax,yex +
eyay,yey +eyaz,yez +ezax,zex +ezay,zey +ezaz,zez
La =
_
_
ax,x ax,y ax,z
ay,x ay,y ay,z
az,x az,y az,z
_
_
L

T
a
=
_
ax,x ay,y az,z ax,y ay,x ay,z az,y az,x ax,z

Gradient of a vector in cylindrical coordinate system


a = L
c
a =
_
er

r
+et
1
r

+ez

z
_
(arer +atet +azez)
= erar,rer +erat,ret +eraz,rez +et
1
r
ar,ter +et
1
r
at,tet +et
1
r
az,tez +et
1
r
aret et
1
r
ater
ezar,zer +ezat,zet +ezaz,zez
La =
_
_
ar,r
1
r ar,t
1
r ar,z
at,r
1
r at,t +
1
r ar at,z
az,r
1
r az,t az,z
_
_
L

T
a
=
_
ar,r
1
r at,t +
1
r ar az,z
1
r ar,t
1
r at at,r at,z
1
r az,t az,r ar,z

Divergence of a tensor in cylindrical coordinate system


A = ei i(ejAjkek)
= ei (iej)Ajkek +ei ej(iAjk)ek +ei ejAjk(iek)
= ei (iej)Ajkek +ij(iAjk)ek +ijAjk(iek)
iej = i21j
1
r
et i22j
1
r
er
116
= ei (i21j
1
r
et i22j
1
r
er)Ajkek +ij(iAjk)ek +ijAjk(i21k
1
r
et i22k
1
r
er)
= ei (i21j
1
r
et i22j
1
r
er)Ajkek +ij(iAjk)ek + (i21k
1
r
et i22k
1
r
er)Ajkij
= et (1j
1
r
et 2j
1
r
er)Ajkek + (jAjk)ek + (j21k
1
r
et j22k
1
r
er)Ajk
= 1j
1
r
Ajkek + (jAjk)ek + (j21k
1
r
et j22k
1
r
er)Ajk
=
1
r
A1kek + (jAjk)ek +
1
r
(A21et A22er)
= (
1
r
A11
1
r
A22)e1 + (
1
r
A12 +
1
r
A21)e2 +
1
r
A13e3 + (jAjk)ek
= gkek +jAjkek
= g

T
e

+ (

T
A)e

= (

T
A)e

+g

T
e

with g

T
=
1
r
_
(A11 A22) (A12 +A21) A33

Chapter 5
Kinematics
The motion and deformation of a three-dimensional continuum is studied in continuum me-
chanics. A continuum is an ideal material body, where the neighborhood of a material point is
assumed to be dense and fully occupied with other material points. The real micro structure
of the material (molecules, crystals, particles, ...) is not considered. The deformation is also
continuous, which implies that the neighborhood of a material point always consists of the
same collection of material points.
Kinematics describes the transformation of a material body from its undeformed to its
deformed state without paying attention to the cause of deformation. In the mathematical
formulation of kinematics a Lagrangian or an Eulerian approach can be chosen. (It is also
possible to follow a so-called Arbitrary-Lagrange-Euler approach.)
The undeformed state is indicated as the state at time t0 and the deformed state as the
state at the current time t. When the deformation process is time- or rate-independent, the
time variable must be considered to be a ctitious time, only used to indicate subsequent
moments in the deformation process.
t
O
P
P
t0
Fig. 5.1 : Deformation of continuum
117
118
5.1 Material coordinates
Each point of the material can be identied by or labeled with material coordinates. In a
three-dimensional space three coordinates {1, 2, 3} are needed and sucient to identify a
point uniquely. The material coordinates of a material point do never change. They can be
stored in a column

T
=
_
1 2 3

.
P

t
P

t0
Fig. 5.2 : Material coordinates
5.2 Position vectors
A point of the material can also be identied with its position in space. Two position vectors
can be chosen for this purpose : the position vector in the undeformed state, x0, or the posi-
tion vector in the current, deformed state, x. Both position vectors can be considered to be
a function of the material coordinates

.
Each point is always identied with one position vector. One spatial position is always
occupied by one material point. For a continuum the position vector is a continuous dier-
entiable function.
Using a vector base {e1, e2, e3}, components of the position vectors can be determined
and stored in columns.
A
V0
A0 t
V
O
e3
e2
e1
x0
t0
x

Fig. 5.3 : Position vector


119
undeformed conguration ( = t0) x0 = (

, t0) = x01e1 +x02e2 +x03e3


deformed conguration ( = t) x = (

, t) = x1e1 +x2e2 +x3e3


5.3 Lagrange - Euler
When a Lagrangian formulation is used to describe state transformation, all variables are
determined in material points which are identied in the undeformed state with their initial
position vector x0. When an Eulerian formulation is used, all variables are determined in
material points which are identied in the deformed state with their current position vector
x. For a scalar quantity a, this can be formally written with a function A.
scalar quantity Lagrange : a = A(x0, t)
scalar quantity Euler : a = A(x, t)
Spatial variation is described with a dierent gradient operator for both procedures. The
dierence da of a scalar quantity a in two adjacent points P and Q can be calculated in both
the Lagrangian and the Eulerian framework. This leads to the denition of two gradient
operators, 0 and , respectively.
da = aQ aP = A(x +dx, t) A(x, t) = dx ( a)

t
da = aQ aP = A(x0 +dx0, t) A(x0, t) = dx0 ( 0a)

t
gradient operators
= e1

x1
+e2

x2
+e3

x3
0 = e1

x01
+e2

x02
+e3

x03
For a vectorial quantity a, the spatial dierence da in two adjacent points, can also be
calculated, using either 0 or . For the position vectors, the gradients result in the unity
tensor I.
(x) = I ; 0(x0) = I
5.4 Time derivatives
A time derivative of a variable expresses the change of its value in time. This change can
be measured in one and the same material point or in one and the same point in space. In
120
the rst case, the observer of the change follows the material, and, in the second case, he is
located in a xed spatial position.
This dierence of observer position leads to two dierent time derivatives, the material
time derivative and the spatial time derivative. Using a material time derivative is associated
with the Lagrangian formulation, while in the Eulerian formulation the spatial time derivative
is generally used. Below, we consider the time derivatives of a scalar vatiable a.
material time derivative
Da
Dt
= a = lim
t0
1
t
{A(x0, t +t) A(x0, t)}
velocity of a material point v = v(x0) = x
spatial time derivative
a
t
= lim
t0
1
t
{A(x, t +t) A(x, t)}
velocity eld v = v(x, t)
A relation between the material and the spatial time derivative can be derived. The material
velocity enters this relation and represents the velocity of the observer. The material time
derivative can be written as the sum of the spatial time derivative and the convective time
derivative.
Da
Dt
= lim
t0
1
t
{A(x0, t +t) A(x0, t)}
= lim
t0
1
t
{A(x +dx, t +t) A(x, t)}
= lim
t0
1
t
{A(x +dx, t +t) A(x, t +t) +A(x, t +t) A(x, t)}
= lim
t0
1
t
{dx ( a)

t+t
+A(x, t +t) A(x, t)}
= lim
t0
{
dx
t
( a)

t+t
} + lim
t0
1
t
{A(x, t +t) A(x, t)}
= v ( a) +
a
t
= (convective time derivative) + (spatial time derivative)
= (material time derivative)
5.5 Deformation
Upon deformation, a material point changes position from x0 to x. This is denoted with a
displacement vector u. In three-dimensional space this vector has three components : u1, u2
and u3.
The deformation of the material can be described by the displacement vector of all the
material points. This, however, is not a very suitable procedure. Instead, we consider the
deformation of an innitesimal material volume in each point, which can be described with a
deformation tensor.
121
x
A0
A
V
x0
V0
u
O
e1
e2
e3
Fig. 5.4 : Deformation of a continuum
displacement : u = x x0 = u1e1 +u2e2 +u3e3
5.5.1 Deformation tensor
To introduce the deformation tensor, we rst consider the deformation of an innitesimal
material line element, between two adjacent material points. The vector between these points
in the undeformed state is dx0. Deformation results in a transformation of this vector to dx,
which can be denoted with a tensor, the deformation tensor F. Using the gradient operator
with respect to the undeformed state, the deformation tensor can be written as a gradient,
which explains its much used name : deformation gradient tensor.
dx = F dx0
= X(x0 +dx0, t) X(x0, t) = dx0
_
0x
_
=
_
0x
_c
dx0 = F dx0
F=
_
0x
_c
=
__
0x0
_c
+
_
0u
_c_
= I +
_
0u
_c
In the undeformed conguration, an innitesimal material volume is uniquely dened by
three material line elements or material vectors dx01, dx02 and dx03. Using the deformation
tensor F, these vectors are transformed to the deformed state to become dx1, dx2 and dx3.
These vectors span the deformed volume element, containing the same material points as in
the initial volume element. It is thus obvious that F describes the transformation of the
material.
122
t t0
P
P
F
dx1
dx01
dx2
dx02
dx3
dx03
Fig. 5.5 : Deformation tensor
dx1 = F dx01 ; dx2 = F dx02 ; dx3 = F dx03
5.5.2 Volume change
The three vectors which span the material element, can be combined in a triple product. The
resulting scalar value is positive when the vectors are right-handed and represents the volume
of the material element. In the undeformed state this volume is dV0 and after deformation the
volume is dV . Using the deformation tensor F and the denition of the determinant (third
invariant) of a second-order tensor, the relation between dV and dV0 can be derived.
t t0
P
P
F
dx1
dx01
dx2
dx02
dx3
dx03
Fig. 5.6 : Volume change
undeformed conguration dV0 = dx01 dx02 dx03
current conguration dV = dx1 dx2 dx3
= (F dx01) (F dx02) (F dx03)
= det(F){dx01 dx02 dx03}
= det(F)dV0
123
volume change factor J = det(F) =
dV
dV0
5.5.3 Area change
The vector product of two vectors along two material line elements represents a vector, the
length of which equals the area of the parallelogram spanned by the vectors. Using the
deformation tensor F, the change of area during deformation can be calculated.
dAn = dx1 dx2 = (F dx01) (F dx02)
dAn (F dx03) = (F dx01) (F dx02) (F dx03)
= det(F)(dx01 dx02) dx03 dx03
dAn F = det(F)(dx01 dx02)
dAn = det(F)(dx01 dx02) F
1
= det(F)dA0 n0 F
1
= dA0 n0
_
F
1
det(F)
_
5.5.4 Elongation and elongational strain
The elongation of a material line element is completely described by the stretch ratio = l
l0
,
where l is its current length and l0 its undeformed length. When there is no deformation, we
have = 1. It is often convenient to describe the elongation with a so-called elongational
strain, which is zero when there is no deformation. A strain is dened as a function of ,
which has to satisfy certain requirements. Much used strain denitions are the linear, the
logarithmic, the Green-Lagrange and the Euler-Almansi strain. One of the requirements of a
strain denition is that it must linearize toward the linear strain.
linear l = 1
logarithmic ln = ln()
Green-Lagrange gl =
1
2(
2
1)
Euler-Almansi ea =
1
2
_
1
1
2
_
0
1
1
ln()
1
2(2 1)
f()
Fig. 5.7 : Strain denitions
124
The elongation factor of a material line element dx0 can be expressed in F and e0, the unity
direction vector of a material vector in the undeformed state. The Green-Lagrange strain can
be expressed in the so-called Green-Lagrange strain tensor E.
e0
F
dx
dx0
e
Fig. 5.8 : Elongation of material line element
(e0) =
||dx||
||dx0||
=

dx dx

dx0 dx0
=

dx0 F
c
F dx0

dx0 dx0
=
||dx0||
||dx0||
_
e0 F
c
F e0
=
_
e0 F
c
F e0
1
2
_

2
(e0) 1
_
= e0
_
1
2 (F
c
F I)

e0 = e0 E e0 E =
1
2 (F
c
F I)
5.5.5 Shear and shear strain
We consider two material vectors in the undeformed state, dx01 and dx02, which are perpen-
dicular. The shear deformation is dened as the cosine of , the angle between the two
material vectors in the deformed state. The shear deformation can be expressed in F and e01
and e02, the unit direction vectors of dx01 and dx02
F

dx1
dx2
dx01
dx02
e1
e2
e01
e02
Fig. 5.9 : Shear of two material line elements
(e01, e02) = sin(

2
) = cos() =
dx1 dx2
||dx1||||dx2||
125
=
e01 F
c
F e02
(e01) (e02)
=
e01 [ F
c
F I ] e02
(e01) (e02)
=
_
2
(e01) (e02)
_
e01 E e02
5.5.6 Right Cauchy-Green deformation tensor
The general transformation of a material line element from the undeformed to the deformed
state is uniquely described by the deformation (gradient) tensor F. The true deformation
consists of elongation of material line elements and mutual rotation of line elements, which is
also referred to as shear.
The true deformation, represented by the expressions for and , is described by the
product F
c
F, which is called the right Cauchy-Green deformation tensor C. This important
tensor has two properties, which are easily recognized : 1) it is symmetric and 2) it is positive
denite.
These properties imply that C has real-valued eigenvectors and eigenvalues, of which
the latter must be positive. The eigenvectors are mutually perpendicular or can be chosen to
be so. Taking them as a vector basis, the tensor C can be written in spectral form.
1. symmetric C
c
= C
2. positive denite
a C a =a F
c
F a = (F a) (F a)
F is regular F a =0 if a =0
a C a > 0 a =0
3. eigenvalues and eigenvectors real
eigenvalues positive
eigenvectors (choice)
spectral representation C = 1 m1 m1 +2 m2 m2 +3 m3 m3
Eigenvectors and eigenvalues
The physical meaning of the eigenvalues and eigenvectors of C becomes clear if we consider
again the expressions for stretch and shear, but now using the spectral representation of
C. For these expressions to have a physical relevant meaning, the eigenvectors of C must
characterize a material direction in the undeformed state. They are denoted as n0i, i = 1, 2, 3.
C = 1 m1 m1 +2 m2 m2 +3 m3 m3
= 1n01n01 +2n02n02 +3n03n03
126
Eigenvectors
Two eigenvectors of C are mutually perpendicular and represent the direction of two mate-
rial elements in the undeformed state. The shear deformation between these two material
directions is zero. i.e. the material line elements remain perpendicular during deformation.
They are called principal directions of deformation or principal strain directions.
(n01, n02) =
n01 C n02

n01 C n01

n02 C n02
= 0
Eigenvalues
The eigenvalues of C appear to be the squared stretch ratios of the material line elements
oriented in the direction of the eigenvectors of C. They are called the principal elongation
factors. The right Cauchy-Green deformation tensor is fully dened in the undeformed state.
It is therefore characterized as a Lagrangian tensor.
(n01) =
_
n01 C n01
=
_
n01 (1n01n01 +2n02n02 +3n03n03) n01 =

1
C =
2
1 n01n01 +
2
2 n02n02 +
2
3 n03n03
5.5.7 Right stretch tensor
Based on the right Cauchy-Green deformation tensor, a new tensor, the right stretch tensor
U, is simply dened as the square root of C. It is obvious that U, like C, is symmetric,
positive denite and regular.
U =

C = 1n01n01 +2n02n02 +3n03n03


1. symmetric : U
c
= U
2. positive denite : a U a > 0 a
3. regular : U
1
=
1
1
n01n01 +
1
2
n02n02 +
1
3
n03n03
4. det(C) = det(U U) = det(F
c
F) = det
2
(F)
det(U) = 123 = det(F) = J
The stretch tensor U can be used to transform perpendicular material line elements dx01, dx02
and dx03. The resulting material vectors dx
01, dx
02 and dx
03, will have changed in length and
will also be no longer perpendicular, when the original line elements do not coincide with the
principal deformation directions. It can be concluded that U describes the real deformation,
so elongation and shear.
127
dx03
dx02
dx01 dx
02
U
dx
03
dx
01
Fig. 5.10 : Transformation by U
dx

01 = U dx01 ; dx

02 = U dx02 ; dx

03 = U dx03
Total transformation
The total transformation from the undeformed to the deformed state, is not described by U
but by F. It seems that there must be another part of the total transformation, which is not
described by U. This missing link between U and F is a tensor R = F U
1
.
dx03
dx02
dx01
R
U
dx
03 dx
02
dx
01
dx3
dx2
dx1
F
Fig. 5.11 : Total transformation
128
dx
01 = U dx01 dx01 = U
1
dx
01
dx1 = F dx01
_

dx1 = F U
1
dx

01 = R dx

01 R = F U
1
5.5.8 Rotation tensor
The tensor R = F U
1
has some properties which renders it to have a physical meaning :
it is a rotation tensor and describes the rigid body rotation of the material volume element
during the transformation from the undeformed to the current, deformed state.
R = F U
1
1.
R
c
R = U
c
F
c
F U
1
= U
c
U U U
1
= U
c
U
c
U U
1
= I R is orthogonal
2.
det(R) = det(F U
1
)
= det(U) det(U
1
) = det(U U
1
)
= det(I) = 1 R is rotation tensor
5.5.9 Right polar decomposition
The total transformation described by F is decomposed into a true deformation, described
by U and a rigid body rotation, described by R. This decomposition is denoted as the right
polar decomposition of the deformation tensor. This decomposition is unique and both U and
R can be determined from F.
F = R U
5.5.10 Strain tensors
The stretch ratio of a material line element in the direction e0 could be determined using the
right Cauchy-Green deformation tensor C. For a strain denition = f() we would like to
have a strain tensor , such that the strain of a material line element in the direction e0 can
be calculated according to : (e0) = e0 e0.
stretch ratio (e0) =
_
e0 C e0
129
strain tensor
strain measure (e0) = e0 e0 = f((e0))
shear measure (e01, e02) = e01 e02
Linear strain tensor
The linear strain tensor E is dened as E = U I. The linear strain of a material line
element in the direction e0 cannot be calculated with this tensor. This is only possible for a
line element in a principal deformation direction n0i.
linear strain tensor E = U I
random strain e0 E e0 = e0 U e0 e0 I e0 = e0 U e0 1 = (e0) 1
principal strain n0i E n0i = n0i U n0i 1 = (n0i) 1 = i 1
Logarithmic strain tensor
The logarithmic strain tensor is dened as = ln(U). The logarithmic strain of a material
line element in the direction e0 cannot be calculated with this tensor. This is only possible
for a line element in a principal deformation direction n0i.
logarithmic strain tensor = ln(U)
random strain e0 e0 = e0 ln(U) e0 = ln((e0))
principal strain n0i n0i = n0i ln(U) n0i = ln((n0i)) = ln(i)
Green-Lagrange strain tensor
The Green-Lagrange strain tensor E is dened as E =
1
2 (C I). For a material line element
in the initial direction e0 the Green-Lagrange strain can be calculated using the Green-
Lagrange strain tensor.
Green-Lagrange strain tensor E =
1
2 (C I)
random strain e0 E e0 =
1
2 (e0 C e0 1) =
1
2
_

2
(e0) 1
_
Innitesimal linear strain tensor
The innitesimal strain tensor is the linearized fraction of the Green-Lagrange strain tensor
E. For innitesimal displacements, the rst partial derivatives of the displacement compo-
nents are so small that all involved squares and products are negligible with respect to the
linear terms. The non-linear terms in E can than be neglected.
For innitesimal displacements the change in position vector of a material point is not
relevant. This means that the dierence between gradient operators vanishes.
130
Green-Lagrange strain tensor
E =
1
2 (F
c
F I) =
1
2
_
( 0u) + ( 0u)
c
+ ( 0u) ( 0u)
c
_
linearisation innitesimal strain tensor
= 1
2
_
( 0u) + ( 0u)
c
_
= 1
2 (F +F
c
) I = 1
2
_
( u) + ( u)
c
_
5.6 Deformation rate
The rate of deformation of a material line element is the material time derivative we follow
the same line element in time of a material vector dx in the current state. This derivative
can be related to dx with a tensor L, the velocity gradient tensor. This tensor is decomposed
into a symmetric and a skewsymmetric part, the deformation rate tensor D and the spin
tensor , respectively.
d x = F dx0 = F F
1 dx = L dx = ( v)c dx
=
1
2{L+L
c
} dx +
1
2{LL
c
} dx
= D dx + dx
5.6.1 Spin tensor
The spin tensor describes only rotation rate of the material line element. This follows
directly from the fact that the spin tensor is skewsymmetric and has a unique associated
axial vector .
= 1
2
_
F F
1
( F F
1
)
c
_
= 1
2
__
v
_c

_
v
__
= skewsymmetric dx = dx = velocity dx = rotation rate
dx

dx
Fig. 5.12 : Rotation rate of material line element
131
5.6.2 Deformation rate tensor
The deformation rate tensor does not what its name suggests. For a random material vector
dx the product D dx is a vector which is not along dx. The deformation rate tensor describes
the rate of elongation but also partly the rate of rotation of dx. Only for material line elements
in the direction of one if its eigenvectors the tensor D describes purely elongation rate.
D =
1
2
_
F F
1
+ ( F F
1
)
c
_
=
__
v
_c
+
_
v
__
D = D
c
D = 111 +222 +333
1. : vector dx along 1 : dx = dx11
D dx = dx1D 1 = dx111 = 1dx
2. : random vector : dx = dx11 +dx22 +dx33
D dx = dx111 +dx222 +dx333
1dx
1
2
3
dx
dx
1
D dx
3
2
dx
e dx
Fig. 5.13 : Deformation rate of material line element
5.6.3 Elongation rate
The elongation rate of a material line element can be expressed in the time derivative of the
elongation factor .

2
= e0 C e0
D
Dt
(
2
) =
D
Dt
(e0 C e0)
2 = e0
D
Dt
(C) e0 = e0
D
Dt
(F
c
F) e0
= e0 { F
c
F +F
c
F} e0
= e0 F
c
{F
c
F
c
+ F F
1
} F e0
= (F e0) {( F F
1
)
c
+ F F
1
} (F e0)
= (e) (2 D) (e)

= e D e
132
5.6.4 Volume change rate
The rate of change of the material volume, the material time derivative of the volume change
factor J, is the product of J itself and the trace of the deformation rate tensor D. To derive
this relation, we consider a material volume element in the undeformed and the deformed
state. In the undeformed state the sides of the element coincide with the pricipal deformation
directions {n01, n02, n03}.
n03
t t0
n1 n2
n3
n01
n02
Fig. 5.14 : Volume change rate of material cube
tr(D) = n1 D n1 +n2 D n2 +n3 D n3
=
1
1
+
2
2
+
3
3
=
D
Dt
{ln(1) + ln(2) + ln(3)} =
D
Dt
{ln(123)}
=
D
Dt
[ln{det(U)}] =
D
Dt
[ln{det(F)}] =
D
Dt
{ln(J)} =
J
J

J = Jtr(D) = J
_
v
_
5.6.5 Area change rate
The rate of change of a material area dA with unit normal vector n can also be expressed in
the velocity gradient tensor L.
D
Dt
(dAn) =
D
Dt
_
det(F)dA0n0 F
1
_
=
D
Dt
{det(F)} dA0n0 F
1
+ det(F)dA0n0 F
1
= J dA0n0 F
1
J dA0n0 F
1
L
= tr(L)JdA0n0 F
1
J dA0n0 F
1
L
= J tr(L)F
c
dA0n0 J L
c
F
c
dA0n0
133
= J (tr(L)I L
c
) F
c
dA0n0
= (tr(L)I L
c
) dAn
5.7 Special deformations
5.7.1 Inverse deformation
The determinant of the deformation tensor, being the quotient of two volumes, is always a
positive number. This implies that the deformation tensor is regular and that the inverse
F
1
exists. It represents the transformation of the deformed state to the undeformed state.
The gradient operators and 0 are related by the (inverse) deformation tensor.
det(F) = J > 0 F regular inverse deformation F
1
inverse deformation
dx0 = F
c
dx
I = F
c
F
c
x = F
c
( 0x) = F
c
0
5.7.2 Planar deformation
It often happens that (part of) a structure is loaded in one plane. Moreover the load is often
such that no bending out of that plane takes place. The resulting deformation is referred to
as being planar.
Here it is assumed that the plane of deformation is the (x1, x2)-plane. Note that in this
planar deformation there still can be displacement perpendicular to the plane of deformation,
which results in change of thickness.
The in-plane displacement components u1 and u2 are only a function of x1 and x2. The
out-of-plane displacement u3 may be a function of x3 as well.
u1 = u1(x1, x2) ; u2 = u2(x1, x2) ; u3 = u3(x1, x2, x3)
Cartesian ux = u1(x, y) ; uy = u2(x, y) ; uz = u3(x, y, z)
cylindrical ur = u1(r, ) ; ut = u2(r, ) ; uz = u3(r, , z)
5.7.3 Axi-symmetric deformation
Many man-made and natural structures have an axi-symmetric geometry, which means that
their shape and volume can be constructed by virtually rotating a cross section around the
axis of revolution. Points are indicated with cylindrical coordinates {r, , z}. When material
properties and loading are also independent of the coordinate , the deformation and resulting
stresses will be either. With the additional assumption that no rotation around the z-axis
takes place, all state variables can be studied in one half of the cross section through the
z-axis.
134
z
r
P
r
z
r

Fig. 5.15 : Axi-symmetric deformation

= 0 u = ur(r, z)er() +ut(r, z)et() +uz(r, z)ez


extra
ut = 0 u = ur(r, z)er() +uz(r, z)ez
5.7.4 Homogeneous deformation
The deformation tensor describes the deformation of an innitesimal material volume, initially
located at position x0. The deformation tensor is generally a function of the position x0.
When F is not a function of position x0, the deformation is called homogeneous. In that
case, each innitesimal material volume shows the same deformation. The current position
vector x can be related to the initial position vector x0 and an unknown rigid body translation
t.
Fig. 5.16 : Homogeneous deformation
135
0x = F
c
= constant tensor x = (x0 F
c
) +t = F x0 +t
5.8 Linear deformation
In linear elasticity theory deformations are very small. All kind of relations from general
continuum mechanics theory may be linearized, resulting for instance in the linear strain
tensor . The deformations are in fact so small that the geometry of the material body in
the deformed state approximately equals that of the undeformed state.
t
O
t0
t
x x0
q
V V0
Fig. 5.17 : Small deformation
F = I + ( 0u)
c
( 0u)
c
= F I O
E =
1
2
_
( 0u)
c
+ ( 0u) + ( 0u)
c
( 0u)
_

1
2
_
0u)
c
+ ( 0u)
_

1
2
_
u)
c
+ ( u)
_
=
Not only straining and shearing must be small to allow the use of linear strains, also the rigid
body rotation must be small. This is immediately clear, when we consider the rigid rotation
of a material line element around a xed point P. The x- and y-displacement of point Q, u
and v respectively, are expressed in the rotation angle and the length of the line element
dx0. The nonlinear Green-Lagrange strain is always zero. The linear strain, however, is only
zero for very small rotations.
136

P Q
x0
Q
dx0
Fig. 5.18 : Rigid rotation of a line element
u = [cos() 1] dx0 ; v = [sin()] dx0
u
x0
= cos() 1 ;
v
x0
= sin()
Green-Lagrange strain gl =
u
x0
+ 1
2
_
u
x0
_2
+ 1
2
_
v
x0
_2
= cos() 1 +
1
2 [cos() 1]
2
+
1
2 sin
2
() = 0
linear strain l =
u
x0
= cos() 1 = 0
small rotation l 0
With respect to an orthogonal basis, the linear strain tensor can be written in components,
resulting in the linear strain matrix. From its denition it is clear that the tensor and the
matrix are symmetric.
=
_
_
11 12 13
21 22 23
31 32 33
_
_ ;
21 = 12
32 = 23
31 = 13
_
_
_

T
=
5.8.1 Elongational strain and shear strain
For small deformations and rotations the elongational and shear strain can be linearized and
expressed in the linear strain tensor . The volume change ratio J can be expressed in linear
strain components and also linearized.
137
elongational strain
1
2
_
2(e01) 1
_
= e01 E e01

(e01) 1 = e01 e01


shear strain (e01, e02) = sin
_

2
_
=
_
2
(e01) (e02)
_
e01 E e02

2 = 2e01 e02
volume change J =
dV
dV0
=
ds1ds2ds3
ds01ds02ds03
= 123 = (1 + 1)(2 + 1)(2 + 1)

J = 1 +2 +3 + 1 = tr() + 1
5.8.2 Principal strains and directions
Because the linear strain tensor is symmetric, it has three real-valued eigenvalues {1, 2, 3}
and associated eigenvectors {n1, n2, n3}. The eigenvectors are normalized to have unit length
and they are mutually perpendicular, so they constitute an orthonormal vector base. The
strain matrix w.r.t. this vector base is diagonal.
The eigenvalues are referred to as the principal strains and the eigenvectors as the prin-
cipal strain directions.
spectral form = 1n1n1 +2n2n2 +3n3n3
principal strain matrix P =
_
_
1 0 0
0 2 0
0 0 3
_
_
5.8.3 Linear strain : Cartesian components
The linear strain components w.r.t. a Cartesian coordinate system are easily derived using
the expression for the gradient operator
= ex

x
+ey

y
+ez

z
and also
u = uxex +uyey +uzez
For derivatives a short notation is used : ( )i,j =
( )i
xj
.
linear strain tensor =
1
2
_
( u)
c
+ ( u)
_
= e

T
e

linear strain matrix


138
=
_
_
xx xy xz
yx yy yz
zx zy zz
_
_ =
1
2
_
_
2ux,x ux,y +uy,x ux,z +uz,x
uy,x +ux,y 2uy,y uy,z +uz,y
uz,x +ux,z uz,y +uy,z 2uz,z
_
_
Strain displacement
The strain-displacement relations for the elongation of line elements can be derived by con-
sidering the elongational deformation of an innitesimal cube of material e.g. in a tensile
test.
x
x
x
l0
l
z
y
P Q
P Q
P P Q Q x
y
z dx0
dy0
dz0
dx
dy
dz
y
z
Fig. 5.19 : Homogeneous elongation of a truss
xx = xx 1 =
dx
dx0
1 =
dx dx0
dx0
=
uQ uP
dx0
=
u(x0 +dx0) u(x0)
dx0
=
u
x0
=
u
x
yy =
v
y
; zz =
w
z
The strain-displacement relations for the shear of two line elements can be derived by con-
sidering the shear deformation of an innitesimal cube of material e.g. in a torsion test.
139
Q
P
dy0
dz0 v

P x
y
z z
y
x
dx0
u
Q
R R
Fig. 5.20 : Shear of a three-dimensional material cube
xy =

2 xy = + sin() + sin()
=
v
dx0
+
u
dy0
=
vQ vP
dx0
+
uR uP
dy0
=
v(x0 +dx0) v(x0)
dx0
+
u(y0 +dy0) u(y0)
dy0
=
v
x0
+
u
y0
=
v
x
+
u
y
yz =
w
y
+
v
z
; zx =
u
z
+
w
x
Compatibility conditions
The six independent strain components are related to only three displacement components.
Therefore the strain components cannot be independent. Six relations can be derived, which
are referred to as the compatibility conditions.
2xx
y2
+
2yy
x2
= 2
2xy
xy
2yy
z2
+
2zz
y2
= 2
2yz
yz
2zz
x2
+
2xx
z2
= 2
2zx
zx
2xx
yz
+
2yz
x2
=
2xz
xy
+
2xy
xz
2yy
zx
+
2zx
y2
=
2yx
yz
+
2yz
yx
2zz
xy
+
2xy
z2
=
2zy
zx
+
2zx
zy
Plane strain
When the boundary conditions and the material behavior are such that displacement of
material points are only in the (x, y)-plane, the deformation is referred to as plane strain
in the (x, y)-plane. Only three relevant strain components and one relevant compatibility
relation remain.
140
plane strain ux = ux(x, y) ; uy = uy(x, y) ; uz = 0
zz = 0 ; xz = yz = 0
compatibility xx,yy +yy,xx = 2xy,xy
5.8.4 Linear strain : cylindrical components
The linear strain components w.r.t. a cylindrical coordinate system are derived straightfor-
wardly using the expression for the gradient operator
= er

r
+et
1
r

+ez

z
and also
u = urer() +utet() +uzez
linear strain tensor =
1
2
_
( u)
c
+ ( u)
_
= e

T
e

linear strain matrix


=
_
_
rr rt rz
tr tt tz
zr zt zz
_
_ =
1
2
_
_
2ur,r
1
r (ur,t ut) +ut,r ur,z +uz,r
1
r (ur,t ut) +ut,r 2
1
r (ur +ut,t)
1
r uz,t +ut,z
uz,r +ur,z
1
r uz,t +ut,z 2uz,z
_
_
Strain displacement
Strain-displacement relations can be derived geometrically in the cylindrical coordinate sys-
tem, as we did in the Cartesian coordinate system.
We consider the deformation of an innitesimal part in the (r, )-plane and determine
the elongational and shear strain components. The dimensions of the material volume in
undeformed state are dr rd dz.
141
ur +ur,rdr

ur ut +ut,td
ur,td
ut
ut +ut,rdr
ut
r (r +dr)
Fig. 5.21 : Deformation of a cylindrical material volume
rr =
ur,rdr
dr
= ur,r
tt =
(r +ur)d rd
rd
+
(ut +ut,td) ut
rd
=
ur
r
+
1
r
ut,t
rt =

2
= + =
_
ut,r
ut
r
_
+
_
1
r
ur,t
_
Axi-symmetry
In many cases the geometry, boundary conditions and material behavior is such that no state
variable depends on the circumferential coordinate :

= 0. The strain-displacement
relations are simplied considerably.
In many axi-symmetric deformations the boundary conditions are such that there is no
displacement in the circumferential direction : ut = 0. In these cases there are only four
relevant strain components.
linear strain matrix
=
1
2
_
_
2ur,r
1
r (ut) +ut,r ur,z +uz,r

1
r (ut) +ut,r 2
1
r (ur) ut,z
uz,r +ur,z ut,z 2uz,z
_
_
linear strain matrix for ut = 0
=
1
2
_
_
2ur,r 0 ur,z +uz,r
0 2
1
r (ur) 0
uz,r +ur,z 0 2uz,z
_
_
142
Axi-symmetric plane strain
When boundary conditions and material behavior are such that displacement of material
points are only in the (r, )-plane, the deformation is referred to as plane strain in the (r, )-
plane.
plane strain deformation
ur = ur(r, )
ut = ut(r, )
uz = 0
_
_
_
zz = rz = tz = 0
linear strain matrix
=
1
2
_
_
2ur,r ut,r
1
r (ut) 0
ut,r
1
r (ut)
2
r (ur) 0
0 0 0
_
_
plane strain deformation with ut = 0
ur = ur(r)
uz = 0
_
=
1
2
_
_
2ur,r 0 0
0
2
r (ur) 0
0 0 0
_
_
Example : inhomogeneous deformation
A cubic block of material (length of all sides is 1) is deformed, as shown in the gure. The
basis {e1, e2, e3} is orthonormal.
h
e2 e2
e1
h0 h0
e3 e3 e1
l0 l
The position vector of an arbitrary material point in undeformed and deformed state, re-
spectively is :
x0 = x01e1 +x02e2 +x03e3 ; x = x1e1 +x2e2 +x03e3
There is no deformation in e3-direction. Deformation in the (e1e2)-plane is such that straight
lines remain straight during deformation.
143
The deformation tensor can be calculated from the relation between the coordinates
of the material point in undeformed and deformed state.
x1 =
l
l0
x01 ; x2 = x02 +
h h0
h0l0
x01x02 ; x3 = x03
F
c
=
_
0x
_
=
_
e01

x01
+e02

x02
+e03

x03
_
(x1e1 +x2e2 +x3e3)
=
_
e01

x01
+e02

x02
+e03

x03
_
__
l
l0
x01
_
e1 +
_
x02 +
h h0
h0l0
x01x02
_
e2 + (x03) e3
_
=
_
l
l0
_
e01e1 +
_
h h0
h0l0
x02
_
e01e2 +
_
1 +
h h0
h0l0
x01
_
e02e2 +e03e3
Example : strain gages
Strain gages are used to measure strains on the surface of a thin walled pressure vessel.
Three gages are glued on the surface, the second perpendicular to the rst one and the
third at an angle of 45o between those two. Measured strains have values g1, g2 and g3.
The linear strain tensor is written in components w.r.t. the Cartesian coordinate system
with its x-axis along the rst strain gage. The components xx, xy and yy have to be
determined from the measured values.
To do this, we use the expression which gives us the strain in a specic direction,
indicated by the unit vector n.
n = n n
Because we have three dierent directions, where the strain is known, we can write this
equation three times.
g1 = ng1 ng1 = n

T
g1 n

g1 =
_
1 0

_
xx xy
yx yy
_ _
1
0
_
= xx
g2 = ng2 ng2 = n

T
g2 n

g2 =
_
0 1

_
xx xy
yx yy
_ _
0
1
_
= yy
g3 = ng3 ng3 = n

T
g3 n

g3 =
1
2
_
1 1

_
xx xy
yx yy
_ _
1
1
_
=
1
2(xx + 2xy +yy)
The rst two equations immediately lead to values for xx and yy and the remaining
unknown, xy can be solved from the last equation.
xx = g1
yy = g2
xy = 2g3 xx yy
= 2g3 g1 g2
_
_
_
=
_
g1 2g3 g1 g2
2g3 g1 g2 g2
_
144
The three gages can be oriented at various angles with respect to each other and with
respect to the coordinate system. However, the three strain components can always be
solved from a set of three independent equations.
Chapter 6
Stresses
Kinematics describes the motion and deformation of a set of material points, considered here
to be a continuous body. The cause of this deformation is not considered in kinematics.
Motion and deformation may have various causes, which are collectively considered here
to be external forces and moments.
Deformation of the material not its motion alone results in internal stresses. It is
very important to calculate them accurately, because they may cause irreversible structural
changes and even unallowable damage of the material.
6.1 Stress vector
Consider a material body in the deformed state, with edge and volume forces, which are in
equilibrium. The body is divided in two parts, where the cutting plane passes through the
material point P. To restore equilibrium of the two parts, an edge load is introduced in their
cutting planes. In two associated points (= coinciding before the cut was made) in the cutting
plane of both parts, these loads are of opposite sign, but have equal absolute value.
The resulting force per unit of area, p, is referred to as the stress vector in point P on
the associated cutting plane through this point.
p
A
q
V
q p
n
k
p
P
A
Fig. 6.1 : Cross-sectional stresses and stress vector on a plane
145
146
resulting cross-sectional force in point P : k
denition stress vector in point P on plane : p = lim
A0
k
A
6.2 Cauchy stress tensor
The stress vector can be calculated, using the stress tensor , which represents the stress
state in point P. The plane is identied by its unity normal vector n. The stress vector is
calculated according to Cauchys theorem, which states that in each material point such a
stress tensor must uniquely exist. (! : there exists only one.)
Theorem of Cauchy : ! tensor such that : p = n
The stress vector p can be written as the sum of two other vectors. The rst is the normal
stress vector pn in the direction of the normal n. The second vector is in the plane and is
called the shear stress vector ps.
The length of the normal stress vector is the normal stress pn and the length of the shear
stress vector is the shear stress ps.
pn n

ps
P
p
Fig. 6.2 : Decomposition of stress vector in normal and shear stress
normal stress : pn = p n = ( n) n = n c n
tensile stress : positive ( <

2 )
compression stress : negative ( >

2 )
normal stress vector : pn = pnn
shear stress vector : ps = p pn
shear stress : ps = || ps|| =
_
|| p||2 p2
n
The Cauchy stress matrix stores the components of the Cauchy stress tensor w.r.t. an
orthonormal vector base {e1, e2, e3}. The components of the Cauchy stress matrix are com-
147
ponents of stress vectors on the planes with unit normal vectors in the coordinate directions.
With our denition, the rst index of a stress component indicates the direction of the
stress vector and the second index indicates the normal of the plane where it is loaded. As
an example, the stress vector on the plane with n = e1 is considered.
p3 p
e1
e2
e3
p1
p2
Fig. 6.3 : Components of stress vector on a plane
p = e1 p

= e

p = e

e1 = e

(e

T
e

) e1 = e

e1
_
_
p1
p2
p3
_
_ =
_
_
11 12 13
21 22 23
31 32 33
_
_
_
_
1
0
0
_
_ =
_
_
11
21
31
_
_
The components of the Cauchy stress matrix can be represented as normal and shear stresses
on the side planes of a stress cube.
12
e1
e3
e2
11
21
31
13
23
33
32
22
Fig. 6.4 : Stress cube
=
_
_
11 12 13
21 22 23
31 32 33
_
_
148
Cartesian stress cube
In the Cartesian coordinate system the stress cube sides are parallel to the Cartesian coordi-
nate axes. Stress components are indicated with the indices x, y and z.
yz
zz
ex
ey
ez
xz
zy
yy
xy
xx
yx
zx
Fig. 6.5 : Cartesian stress cube
=
_
_
xx xy xz
yx yy yz
zx zy zz
_
_
Cylindrical stress cube
In the cylindrical coordinate system the stress cube sides are parallel to the cylindrical
coordinate axes. Stress components are indicated with the indices r, t and z.
tz
r
z

rr
tt
zr
tr
rt
zt
zz
rz
Fig. 6.6 : Cylindrical stress cube
=
_
_
rr rt rz
tr tt tz
zr zt zz
_
_
149
6.3 Principal stresses and directions
It will be shown later that the stress tensor is symmetric. This means that it has three real-
valued eigenvalues {1, 2, 3} and associated eigenvectors {n1, n2, n3}. The eigenvectors are
normalized to have unit length and they are mutually perpendicular, so they constitute an
orthonormal vector base. The stress matrix w.r.t. this vector base is diagonal.
The eigenvalues are referred to as the principal stresses and the eigenvectors as the
principal stress directions. The stress cube with the normal principal stresses is referred to
as the principal stress cube.
Using the spectral representation of , it is easily shown that the stress tensor changes
as a result of a rigid body rotation Q.
3
z
y
O x
t0
t
P P
2
3
1
1
2
Fig. 6.7 : Principal stress cube with principal stresses
spectral form = 1n1n1 +2n2n2 +3n3n3
principal stress matrix P =
_
_
1 0 0
0 2 0
0 0 3
_
_
rigid body rotation
= 1n1n1 +2n2n2 +3n3n3

= 1n

1n

1 +2n

2n

2 +3n

3n

= Q Q
c
6.3.1 Stress transformation
The normal and shear stresses on each plane in between the principal stress planes can be
calculated from the principal stresses 1 and 2. Calculation can be based on transformation
of the stress matrix but also on equilibrium.
150
2
1

area 1
e2
e1
area cos
area sin
Fig. 6.8 : Normal and shear stress on a plane
equilibrium
sin()1 + cos() sin() = 0
cos()2 + sin() + cos() = 0
eq. 1 sin(); eq. 2 cos()
= sin
2
()1 + cos
2
()2
= cos() sin()1 + cos() sin()2
6.3.2 Mohrs circles of stress
From the relations for the normal and shear stress on a plane in between two principal stress
planes, a relation between these two stresses and the principal stresses can be derived. The
resulting relation is the equation of a circle in the (, )-plane, referred to as Mohrs circle for
stress. The radius of the circle is
1
2(1 2). The coordinates of its center are {
1
2 (1 +2), 0}.
Stresses on a plane, which is rotated over w.r.t. a principal stress plane, can be found
in the circle by rotation over 2.
Because there are three principal stresses and principal stress planes, there are also three
stress circles. Each stress state in the material is located on one of the circle or in the shaded
area.
= sin
2
()1 + cos
2
()2
= 1(1
2 1
2 cos(2)) +2(1
2 + 1
2 cos(2))
=
1
2(1 +2)
1
2(1 2) cos(2)
_

1
2(1 +2)
_2
=
_
1
2(1 2)
_2
cos
2
(2)
= cos() sin()1 + cos() sin()2 =
1
2 (2 1) sin(2)

2
=
_
1
2(2 1)
_2
sin
2
(2)
add equations
_

1
2(1 +2)
_2
+
2
=
_
1
2(1 2)
_2
151

1 2

m
2
1

3 2
Fig. 6.9 : Mohrs circles
6.4 Resulting force on arbitrary material volume
A material body with volume V and surface area A is loaded with a volume load q per unit of
mass and by a surface load p per unit of area. Taking a random part of the continuum with
volume V and edge A, the resulting force can be written as an integral over the volume, using
Gauss theorem. The load q is a volume load per unit of volume, where is the density of
the material.
A
V
q
V
p
p
A
Fig. 6.10 : Forces on a random section of a material body
resulting force on V K =
_
V
q dV +
_
A
p dA =
_
V
q dV +
_
A
n
c
dA
Gauss theorem
K =
_
V
_
q +
c
_
dV
152
6.5 Resulting moment on arbitrary material volume
The resulting moment about a xed point of the forces working in volume and edge points of
a random part of the continuum body can be calculated by integration.
x
p
A
V
A
q
p
V
O
Fig. 6.11 : Moments of forces on a random section of a material body
resulting moment about O
MO =
_
V
x q dV +
_
A
x p dA
6.6 Special stress states
Uni-axial stress
An unidirectional stress state is what we have in a tensile bar or truss. The axial load N
in a cross-section (area A in the deformed state) is the integral of the axial stress over A.
For homogeneous material the stress is uniform in the cross-section and is called the true or
Cauchy stress. When it is assumed to be uniform in the cross-section, it is the ratio of N
and A. The engineering stress is the ratio of N and the initial cross-sectional area A0, which
makes calculation easy, because A does not have to be known. For small deformations it is
obvious that A A0 and thus that n.
153
x
z
N N
y
P
y
x
z
P
xx xx
x
Fig. 6.12 : Stresses on a small material volume in a tensile bar
true or Cauchy stress =
N
A
= xx = xxexex
engineering stress n =
N
A0
Hydrostatic stress
A hydrostatic loading of the material body results in a hydrostatic stress state in each material
point P. This can again be indicated by stresses (either tensile or compressive) on a stress
cube. The three stress variables, with the same value, are normal to the faces of the stress
cube.
y
z
x
x
z
p
p
p
p
p
yy = p
p
y
xx = p
yy = p
zz = p
xx = p
zz = p
= p(exex +eyey +ezez)
Fig. 6.13 : Stresses on a material volume under hydrostatic loading
154
Shear stress
The axial torsion of a thin-walled tube (radius R, wall thickness t) is the result of an axial
torsional moment (torque) T. This load causes a shear stress in the cross-sectional wall.
Although this shear stress has the same value in each point of the cross-section, the stress
cube looks dierently in each point because of the circumferential direction of .
= zx
x
z
y

= yx
= zx
= yx T
Fig. 6.14 : Stresses on a small material volume in the wall of a tube under shear loading
shear stress in cross section =
T
2R2t
= (eiej +ejei) with i = j
Plane stress
When stresses on a plane perpendicular to the 3-direction are zero, the stress state is referred
to as plane stress w.r.t. the (1, 2)-plane. Only three stress components are relevant in this
case.
e1
e3
e2
11
22
12
21
Fig. 6.15 : Stress cube for plane stress in e1, e2-plane
155
33 = 13 = 23 = 0 e3 =0
relevant stresses : 11, 22, 12
u1 = u1(x1, x2) ; u2 = u2(x1, x2) ; u3 = u3(x1, x2, x3)
156
Chapter 7
Balance or conservation laws
In every physical process, so also during deformation of continuum bodies, some general ac-
cepted physical laws have to be obeyed : the conservation laws. During deformation the total
mass has to be preserved and also the total momentum and moment of momentum. Because
we do not consider dissipation and thermal eects, we will not discuss the conservation law
for total energy.
7.1 Conservation of mass
The mass of each nite, randomly chosen volume of material points in the continuum body
must remain the same during the deformation process. Because we consider here a nite
volume, this is the so-called global version of the mass conservation law.
From the requirement that this global law must hold for every randomly chosen volume,
the local version of the conservation law can be derived. This derivation uses an integral
transformation, where the integral over the volume V in the deformed state is transformed
to an integral over the volume V0 in the undeformed state.
A
V
V0
t0
A0
V V0
t
Fig. 7.1 : Random volume in undeformed and deformed state
Global version
_
V
dV =
_
V0
0 dV0 V
157
158
Integral transformation (with dV = J dV0)
_
V0
(J 0) dV0 = 0 V0
Local version : continuity equation
J = 0 x V (t)
The local version, which is also referred to as the continuity equation, can also be derived
directly by considering the mass dM of the innitesimal volume dV of material points.
The time derivative of the mass conservation law is also used frequently. Because we
focus attention on the same material particles, a so-called material time derivative is used,
which is indicated as ().
dM = dM0 dV = 0dV0
J = 0 x V (t)
J + J = 0
7.2 Balance of momentum
According to the balance of momentum law, a point mass m which has a velocity v, will change
its momentum i = mv under the action of a force K. Analogously, the total force working
on a randomly chosen volume of material points equals the change of the total momentum of
the material points inside the volume. In the balance law, again a material time derivative
is used, because we consider the same material points. The total force can be written as a
volume integral of volume forces and the divergence of the stress tensor.
q
A
V
V
t
A
p
V
Fig. 7.2 : Forces on random section of a material body
Global version
K =
Di
Dt
=
D
Dt
_
V
v dV V
159
=
D
Dt
_
V0
vJ dV0 =
_
V0
D
Dt
(vJ) dV0 =
_
V0
_
vJ + vJ +v J
_
dV0
mass balance : J + J = 0
=
_
V0
vJ dV0 =
_
V
v dV V
_
V
_
q +
c
_
dV =
_
V
v dV V
From the requirement that the global balance law must hold for every randomly chosen volume
of material points, the local version of the balance of momentum can be derived, which must
hold in every material point. In the derivation an integral transformation is used.
The local balance of momentum law is also called the equation of motion. For a stationary
process, where the material velocity v in a xed spatial point does not change, the equation
is simplied. For a static process, where there is no acceleration of masses, the equilibrium
equation results.
local version : equation of motion
c
+ q = v =
v
t
+v
_
v
_
x V (t)
stationary
_
v
t
= 0
_

c
+ q = v
_
v
_
static : equilibrium equation
c
+ q =0
7.3 Balance of moment of momentum
The balance of moment of momentum states that the total moment about a xed point of
all forces working on a randomly chosen volume of material points ( MO), equals the change
of the total moment of momentum of the material points inside the volume, taken w.r.t. the
same xed point ( LO).
160
x
A
V
V
t
A
p
V
q
O
Fig. 7.3 : Moment of forces on a random section of a material body
MO =
D LO
Dt
=
D
Dt
_
V
x v dV V
_
V
x q dV +
_
A
x p dA =
_
V
x v dV V
To derive a local version, the integral over the area A has to be transformed to an integral
over the enclosed volume V . In this derivation, the operator 3 is used, which is dened such
that
a b =
3
: ab
holds for all vectors a and b.
_
A
x p dA =
_
A
3
: (x p) dA =
_
A
3
: (x n) dA
=
_
A
3
: (x) ndA =
_
A
n {
3
: (x)}
c
dA
=
_
V
{
3
: (x)}
c
dV
=
_
V
{(x)
c
:
3

c
} dV =
_
V
{(
c
x) :
3

c
} dV
=
_
V
_
(
c
)x :
3

c
+ ( x) :
3

c
_
dV
=
_
V
_
(
c
)x :
3

c
+ :
3

c
_
dV =
_
V
_
3
: x(
c
) +
3
:
c
_
dV
161
=
_
V
_
3
:
c

+x (
c
) dV
Substitution in the global version and using the local balance of momentum, leads to the local
version of the balance of moment of momentum, which simply states that the Cauchy stress
tensor is symmetric.
_
V
x q dV +
_
V
3
:
c
dV +
_
V
x (
c
) dV =
_
V
x v dV V
_
V
x
_
q + (
c
) v
_
dV +
_
V
3
:
c
dV =0 V
_
V
3
:
c
dV =0 V
3
:
c
=0 x V
Because the components of 3 equal 1 if the permutation {i, j, k} is even, -1 if it is odd and 0
if indices are repeated, it can be derived that
_
_
32 23
13 31
21 12
_
_ =
_
_
0
0
0
_
_
So the local version states that the Cauchy stress tensor is symmetric.
c = x V (t)
7.4 Equilibrium equations in components
Balance of momentum and moment of momentum result in the equilibrium equations for
forces and moments.

c
+ q =0 ; =
c
7.4.1 Cartesian components
The local balance of momentum law can be written in components w.r.t. a Cartesian vector
basis. This results in three partial dierential equations, one for each coordinate direction.
xx,x +xy,y +xz,z +qx = 0
yx,x +yy,y +yz,z +qy = 0
zx,x +zy,y +zz,z +qz = 0
=
T
xy = yx ; yz = zy ; zx = xz
162
The equilibrium equations in the three coordinate directions can be derived by considering
the force equilibrium of the Cartesian stress cube.
yz +yz,zdz
xz +xz,zdz
zz +zz,zdz
yz
xz
zz
z
y
x
yy
yx
yy +yy,ydy
xy
zy
xx
zx
xy +xy,ydy
zy +zy,ydy
xx +xx,xdx
yx +yx,xdx
zx +zx,xdx
Fig. 7.4 : Equilibrium of forces working on the faces of a Cartesian stress cube
Equilibrium in x-direction :
(xx +xx,xdx)dydz + (xy +xy,ydy)dxdz + (xz +xz,zdz)dxdy
(xx)dydz (xy)dxdz (xz)dxdy +qxdxdydz = 0
The forces, working on the Cartesian stress cube, have a moment w.r.t. a certain point in
space. The sum of all the moments must be zero. We consider the moments of forces in
the xy-plane w.r.t. the z-axis through the center of the cube. Anti-clockwise moments are
positive.
xx(x) xx(x +dx)
yy(y +dy)
xy(y)
xy(y +dy)
yx(x +dx) yx(x)
yy(y)
Fig. 7.5 : Equilibrium of moment of forces working on the faces of a Cartesian stress cube
163
yxdydz 1
2 dx +yxdydz 1
2dx +yx,xdxdydz 1
2dx
xydxdz
1
2dy xydxdz
1
2dy xy,xdxdydz
1
2dy = 0
yx xy = 0 yx = xy
7.4.2 Cylindrical components
Writing tensor and vectors in components w.r.t. a cylindrical vector basis, results in compo-
nent versions of the equilibrium equations. It is noted that the cylindrical base vectors er and
et are a function of the coordinate , so they have to be dierentiated, when expanding the
divergence term.
rr,r +
1
r
rt,t +
1
r
(rr tt) +rz,z +qr = 0
tr,r +
1
r
tt,t +
1
r
(tr +rt) +tz,z +qt = 0
zr,r +
1
r
zt,t +
1
r
zr +zz,z +qz = 0
=
T
rt = tr ; tz = zt ; zr = rz
The equilibrium equations in the three coordinate directions can be derived by considering
the force equilibrium of the cylindrical stress cube. Here only the equilibrium in r-direction
is considered. The stress components are a function of the three cylindrical coordinates r,
and z, but only the relevant (changing) ones are indicated.
r
dr
d
rr(r)
rr(r +dr)
tt()
tt( +d)
rt()
rt( +d)
tr(r)
tr(r +dr) rz(z)
rz(z +dz)
qr
Fig. 7.6 : Equilibrium of forces working on the faces of a cylindrical stress cube
164
rr(r)rddz rz(z)rdrd rt()drdz tt()dr
1
2 ddz
+rr(r +dr)(r +dr)ddz +rz(z +dz)rdrd
+rt( +d)drdz tt( +d)dr1
2 ddz +qrrdrddz = 0
rr,rrdrddz +rrdrddz +rz,zrdrddz +rt,tdrddz
tt()drddz +qrdrddz = 0
rr,r +
1
r
rr +rz,z +
1
r
rt,t
1
r
tt +qr = 0
The forces, working on the cylindrical stress cube, have a moment w.r.t. a certain point in
space. The sum of all the moments mus be zero. We consider the moments of forces in the
(r, )-plane w.r.t. the z-axis through the center of the cube.
r
dr
d
rr(r)
rr(r +dr)
tt()
tt( +d)
rt()
rt( +d)
tr(r)
tr(r +dr) rz(z)
rz(z +dz)
qr
Fig. 7.7 : Equilibrium of moment of forces working on the faces of a cylindrical stress cube
tr(r)rddz
1
2dr +tr(r +dr)(r +dr)ddz
1
2dr
rt()drdz
1
2rd rt( +d)drdz
1
2rd = 0
trrdrddz rtrdrddz = 0 tr = rt
7.4.3 Special equilibrium states
The three-dimensional equilibrium equations can be simplied for special deformation or
stress states, such as plane strain, plane stress and axisymmetric cases.
Planar deformation
It is assumed here that the z-direction is the direction where either the strain or the stress
is zero. Only stresses and strains in the plane perpendicular to the z-direction remain to
165
be determined from equilibrium. The strain or stress in the z-direction can be calculated
afterwards, either directly from the material law or iteratively during the solution procedure.
Cartesian components
xx,x +xy,y +qx = 0
yx,x +yy,y +qy = 0
xy = yx
Cylindrical components
rr,r +
1
r
rt,t +
1
r
(rr tt) +qr = 0
tr,r +
1
r
tt,t +
1
r
(tr +rt) +qt = 0
rt = tr
Axisymmetric deformation
In many cases the geometry, boundary conditions and material behavior is such that no
state variable depends on the circumferential coordinate :

= 0. For such axisymmetric
deformations, the equilibrium equations can be simplied considerably.
In many axisymmetric deformations the boundary conditions are such that there is no
displacement in the circumferential direction : ut = 0. In these cases there are only four
relevant strain and stress components and only three equilibrium equations.
rr,r +
1
r
(rr tt) +rz,z +qr = 0
tr,r +
2
r
(tr) +tz,z +qt = 0 (if ut = 0)
zr,r +
1
r
zr +zz,z +qz = 0
rt = tr ; tz = zt (if ut = 0)
zr = rz
When boundary conditions and material behavior are such that displacement of material
points are only in the (r, )-plane, the deformation is referred to as plane strain in the (r, )-
plane.
When stresses on a plane perpendicular to the z-direction are zero, the stress state is
referred to as plane stress w.r.t. the (r, )-plane.
166
rr,r +
1
r
(rr tt) +qr = 0
tr,r +
2
r
(tr) +qt = 0 (if ut = 0)
rt = tr (if ut = 0)
7.5 Balance of energy
The rst law of thermodynamics states that the total amount of energy supplied to a material
body is converted to kinetic energy (Uk) and internal energy (Ui). The supplied energy is
considered to be 1) work done by external mechanical loads (Ue), and 2) thermal energy
supplied by internal sources or external uxes (Ut). The internal energy can be of very
dierent character, such as elastically stored energy and dissipated energy due to plastic
deformation, viscous eects, crack growth, etcetera.
D
Dt
(Ue +Ut) =
D
Dt
(Uk +Ui)
7.5.1 Mechanical energy
When a point load k is applied in a material point and the point moves with a velocity v,
the work of the load per unit of time is Ue = k v. For a random volume V with edge A
inside a material body the mechanical work of all loads per unit of time can be calculated.
Using Gauss theorem, this work can be written as an integral over the volume V . Also the
equation of motion is used to arrive at the nal result.
q
A
V
V
t
A
p
V
Fig. 7.8 : Mechanical load on a material volume
Ue =
_
V
q v dV +
_
A
p v dA =
_
V
{ q v + (
c
v)} dV
(
c
v) = (
c
) v + : ( v)
167
= v v q v + : D+ :
=
_
V
( v v + : D) dV
7.5.2 Thermal energy
Thermal energy can be produced by internal sources. The heat production per unit of mass
is r [J kg1].
Heat can ow in or out of a material body or in the body from one part to another. In
a material point P the heat ux vector is H [J]. The heat ux density vector in P through a
plane with area A is
h = lim
A0
H
A
[J m2]
The resulting heat ux in P through the plane is n h [J m2] , where n is the unit normal
vector on the plane.
For a random volume V having edge A with unit normal outward vector n, the increase
in thermal energy at time t is Ut.
r
A
V
V
t
A
V
h
r
Fig. 7.9 : Heat sources in and heat ux into a material volume
Ut =
_
V
r dV
_
A
n hdA =
_
V
(r h) dV
7.5.3 Kinetic energy
The kinetic energy of a point mass m with velocity v is
Uk =
1
2 m ||v||
2
=
1
2 m v v
For a random volume V of material points, having density and velocity v, the total kinetic
energy Uk can be calculated by intergration.
168
Uk(t) =
_
V
1
2 v v dV Uk =
_
V
v v dV
7.5.4 Internal energy
The internal energy per unit of mass is . The total internal energy of all material points in
a random volume V of a material body, Ui, can be calculated by integration.
Ui(t) =
_
V
dV Ui =
_
V
dV
7.5.5 Energy balance
The energy balance or rst law of thermodynamics for a random volume of material points in
a material body, can be written as an integral equation. It is the global form of the balance
law, because a nite volume is considered.
Ue + Ut = Uk + Ui
global version of energy balance
_
V
( v v + : D+r h) dV =
_
V
( v v + ) dV V
_
V
dV =
_
V
( : D+r h) dV
V
7.5.6 Energy equation
The local version of the energy balance, also called the energy equation, is easily derived by
taking into account the fact that the global version must be valid for each volume V .
The specic internal energy can be written as the product of the specic heat Cp
(assumed to be constant here) and the absolute temperature T.
The heat ux density h is often related to the temperature gradient T accoding to
Fouriers law, where the thermal conductivity k (assumed to be constant here) is a material
parameter.
local version of energy balance : energy equation
= : D+r h x V (t)
with = Cp
T (Cp : specic heat)
169
Cp
T = : D+r h x V (t)
with h = k T (k : thermal conductivity)
Cp
T k
2
T = : D+r x V (t)
7.5.7 Mechanical power for three-dimensional deformation
Elastic deformation of a three-dimensional continuum leads to storage of elastic energy, which
can be calculated per unit of undeformed (W0) or deformed (W) volume. Dierent expressions
for the strain rate can than be combined with dierent stress tensors, which are all a function
of the Cauchy stress tensor . The starting point is the change of stored energy per unit of
deformed volume.
W = : D = Cauchy stress tensor
W0 = [J] : D
= : D = Kirchho stress tensor
W0 = J : D = J :
1
2
_
F F
1
+ ( F F
1
)
c
_
=
= J :
_
F F
1
_
= J
_
F
1

_
: F = S : F = S : U
= S : E S = 1st-Piola-Kirchho stress tensor
W0 = J : D = J :
_
F
c
E F
1
_
= J
_
F
1
F
c
_
: E
= P : E P = 2nd-Piola-Kirchho stress tensor
These stress tensors can be interpreted as forces per unit of area in the undeformed or the
deformed state.
170
Chapter 8
Constitutive equations
Stresses must always satisfy the balance laws, which are considered to be laws of physics in
the non-quantum world, where we live our lives together with our materials and structures.
Balance laws must apply to each material, of which the deformation is studied. It is obvious,
however, that various materials will behave very dierently, when subjected to the same
external loads. This behavior must be incorporated in the continuum mechanics theory and
is therefore modelled mathematically. The resulting equations are referred to as constitutive
equations. They can not fully be derived from physical principles, although the theory of
thermodynamics tells us a lot of how they must look like. The real mathematical formulation
of the material laws is however based on experimental observations of the deformation of the
material.
In later sections, the behavior of a wide range of materials is modelled and used in a
three-dimensional context. In this chapter, the more general aspects of constitutive equations
are discussed.
8.1 Equations and unknowns
Although it is obvious that material laws must be incorporated to describe the behavior of
dierent materials, they are also needed from a purely mathematical point of view. This has
to do with the number of unknown variables and the number of equations, from which they
must be solved. Obviously, the number of equations has to be the same as the number of
unknowns.
The local balance laws for mass, momentum and moment of momentum have to be
satised in every material point of the continuum body at every time during the deformation
process.
The mass balance law is a scalar equation. The balance of momentum or equation of
motion is a partial dierential equation. It is a vector equation. The balance of moment of
momentum is a tensor equation.
mass J = 0
momentum
c
+ q = v
171
172
moment of momentum
c
=
The unknown variables, which appear in the balance laws, are the density of the material,
the position vector x of the material point and the stress tensor .
density
position vector x
Cauchy stress tensor
The continuity equation can be used to express the density in the deformation tensor F,
which is known, when the position x of the material point is known. So we can skip the mass
balance from our equation set and the density from the set of unknowns.
The moment of momentum equation can be used directly to state that there are only 6
unknown stress components instead of 9. So we loose three equations and three unknowns.
The number of unknowns is now 9 and the number of equations is 3, so 6 constitutive
equations are required. These equations are relations between the stress components and the
components of the position vector.
= N(x)
8.2 General constitutive equation
The most general constitutive equation states that the stress tensor in point x at (the cur-
rent) time t, is a function of the position of all material points at every previous time in
the deformation process. This implies that the complete deformation history of all points is
needed to calculate the current stress in each material point.
This constitutive equation is far to general to be useful. In the following it will be speci-
ed by incorporating assumptions about the material behavior. In practice these assumptions
must of course be based on experimental observations.
(x, t) = N{ x, | x V ; t}
O
x
x
x0
t0 t

x() x0
P
P
P
Fig. 8.1 : Deformation history of a continuum
173
8.2.1 Locality
A wide range of materials and deformation processes allow the assumption of locality. In that
case the stress in a point x is determined by the position of points in its direct neighborhood,
so points with position vector x+dx. This can be written in terms of the deformation tensor
F.
(x, t) = N{ x, | x V ; t}
x = x +dx = x +F(x) dx0
_
(x, t) = N(x, F(x, ), | t)
O
P
t
x
x
Fig. 8.2 : Deformation with local inuence
8.2.2 Frame indierence
The stress state in a material point will not change when the material body is translated
and/or rotated without (extra) deformation, i.e. when it moves as a rigid body. The vari-
ables in the constitutive equation may, however, change. The constitutive equation must be
formulated such that these changes do not aect the stress state in a material point.
A rigid body translation is described with a displacement vector, which is equal for all
material points. The stress is not allowed to change, so it is easily seen that the constitutive
function N cannot depend on the position x.
(x, t) = N(F(x, ), | t)
The symmetric Cauchy stress tensor can be written in spectral form. When the deformed body
is subjected to a rigid rotation, described by the rotation tensor Q, the principal stresses do
not change, but the principal directions do. This means that the Cauchy stress tensor changes
due to rigid rotation of the material.
The deformation tensor F will also change as a consequence of rigid rotation, which can
be easily seen from the polar decomposition.
174
n1
t
t
n
1
Fig. 8.3 : Rigid body rotation of a continuum
= 1n1n1 +2n2n2 +3n3n3
rigid rotation Q
_

= 1n

1n

1 +2n

2n

2 +3n

3n

3
n

1 = Q n1 ; n

2 = Q n2 ; n

3 = Q n3
_

= 1Q n1n1 Q
c
+2Q n2n2 Q
c
+3Q n3n3 Q
c
= Q [1n1n1 +2n2n2 +3n3n3] Q
c
= Q Q
c
F = R U
rigid rotation Q
_

F

= R

U = Q R U F

= Q F
The relation between and F

must be the same as that between and F, which results


in a requirement for the constitutive equation. (We skip the x-dependency of F.)

(x, t) = N(F

() | t)
Q(t) (x, t) Q
c
(t) = N(Q() F() | t)
(x, t) = Q(t)
c
N(Q() F() | t) Q(t) = N(F() | t)
N(F

() | t) = Q(t) N(F() | t) Q
c
(t) Q
8.3 Invariant stress tensor
For convenient constitutive modeling where stress (rate) is related to deformation (rate), we
need stress tensors which are invariant with rigid rotation. Also their time derivative must
answer this requirement.
A stress tensor S = A A
c
can be dened, where A is to be specied later, but always
has to obey A

= A Q
c
. It follows that the stress tensor S is invariant for rigid rotations.
175
S = A A
c
S

= A

A
c
= A

Q Q
c
A
c
dene A

= A Q
c
_

S

= A Q
c
Q Q
c
Q A
c
= A A
c
= S
Also its time derivative S is inveriant.
S = A A
c
+A A
c
+A A
c
S

= A

A
c
+A

A
c
+A

A
c
= ( A Q
c
+A Q
c
) Q Q
c
Q A
c
+
A Q
c
( Q Q
c
+Q Q
c
+Q Q
c
) Q A
c
+
A Q
c
Q Q
c
(Q A
c
+ Q A
c
)
= A A
c
+A Q
c
Q A
c
+A Q
c
Q A
c
+
A A
c
+A Q
c
Q A
c
+A A
c
+
A Q
c
Q A
c
= A A
c
+A A
c
+A A
c
+A Q
c
Q A
c
+
A Q
c
Q A
c
+A Q
c
Q A
c
+
A Q
c
Q A
c
= S
The time derivative of S can also be expressed in the Cauchy stress tensor and its rate. As
a short notation the Cauchy stress rate

is introduced, which is a function of , A and A.
This tensor has the same transformation upon rigid body rotation than the Cauchy stress
tensor .
S = A A
c
+A A
c
+A A
c
= A (A
1
A) A
c
+A A
c
+A (A
1
A)
c
A
c
= A
_
(A
1
A) + (A
1
A)
c
+
_
A
c
= A

A
c

= + (A
1
A) + (A
1
A)
c

+ (A
1
A

(A
1
A

)
c
A

= A Q
c
A
1
= A
1
= Q A
1
A

= A Q
c
+A Q
c
176
A
1
A

= Q A
1
A Q
c
+Q Q
c
=

+Q A
1
A Q
c

+Q Q
c

Q (A
1
A)
c
Q
c
+

Q Q
c
= Q Q
c
+Q Q
c
+Q Q
c
+
Q A
1
A Q
c
+Q Q
c
Q
c
Q
c
+
Q (A
1
A)
c
Q
c
+Q Q
c
Q Q
c
= Q [ + (A
1
A) + (A
1
A)
c
] Q
c
= Q

Q
c
8.4 Invariant stress tensors and their rates
The tensor A is now specied, which results in some alternative invariant stress tensors.
With each tensor a so-called objective rate of the Cauchy stress tensor is associated. choosing
A {F
1
, Q
1
, F
c
, R
c
} results in the Truesdell, Jaumann, Cotter-Rivlin and Dienes tensor
and rate.
general tensor S = O = A A
c
S = O = A

O A
c
general rate

O = + (A
1
A) + (A
1
A)
c
Truesdell tensor T = F
1
F
c
T = F
1


T F
c
Truesdell rate

T =

= L L
c
Jaumann tensor J = Q
1
Q
c
with Q = Q
J = Q
1


J Q
c
Jaumann rate

J =

=
c
Cotter-Rivlin tensor C = F
c
F
C = F
c


C F
Cotter-Rivlin rate

C =

= +L
c
+ L
Dienes tensor D = R
c
R with F = R U
D = R
c


D R
Dienes rate

D =

= ( R R
c
) ( R R
c
)
c
Chapter 9
Linear elastic material
For linear elastic material behavior the stress tensor is related to the linear strain tensor
by the constant fourth-order stiness tensor 4C :
=
4
C :
The relevant components of and w.r.t. an orthonormal vector basis {e1, e2, e3} are stored
in columns

and

. Note that we use double waves to indicate that the columns contain
components of a second-order tensor.

T
= [11 22 33 12 21 23 32 31 13]

T
= [11 22 33 12 21 23 32 31 13]
The relation between these columns is given by the 9 9 matrix C, which stores the compo-
nents of 4C and is referred to as the material stiness matrix. Note again the use of double
underscore to indicate that the matrix contains components of a fourth-order tensor.
_

_
11
22
33
12
21
23
32
31
13
_

_
=
_

_
C1111 C1122 C1133 C1121 C1112 C1132 C1123 C1113 C1131
C2211 C2222 C2233 C2221 C2212 C2232 C2223 C2213 C2231
C3311 C3322 C3333 C3321 C3312 C3332 C3323 C3313 C3331
C1211 C1222 C1233 C1221 C1212 C1232 C1223 C1213 C1231
C2111 C2122 C2133 C2121 C2112 C2132 C2123 C2113 C2131
C2311 C2322 C2333 C2321 C2312 C2332 C2323 C2313 C2331
C3211 C3222 C3233 C3221 C3212 C3232 C3223 C3213 C3231
C3111 C3122 C3133 C3121 C3112 C3132 C3123 C3113 C3131
C1311 C1322 C1333 C1321 C1312 C1332 C1323 C1313 C1331
_

_
_

_
11
22
33
12
21
23
32
31
13
_

_
The stored energy per unit of volume is :
W =
1
2 :
4
C : =
_
1
2 :
4
C :
c
=
1
2 :
4
C
c
:
which implies that 4C is total-symmetric : 4C = 4C
c
or equivalently C = CT .
As the stress tensor is symmetric, = c, the tensor 4C must be left-symmetric :
4C = 4C
lc
or equivalently C = C
LT
. As also the strain tensor is symmetric, = c, the
constitutive relation can be written with a 6 6 stiness matrix.
177
178
_

_
11
22
33
12
23
31
_

_
=
_

_
C1111 C1122 C1133 [C1121 +C1112] [C1132 +C1123] [C1113 +C1131]
C2211 C2222 C2233 [C2221 +C2212] [C2232 +C2223] [C2213 +C2231]
C3311 C3322 C3333 [C3321 +C3312] [C3332 +C3323] [C3313 +C3331]
C1211 C1222 C1233 [C1221 +C1212] [C1232 +C1223] [C1213 +C1231]
C2311 C2322 C2333 [C2321 +C2312] [C2332 +C2323] [C2313 +C2331]
C3111 C3122 C3133 [C3121 +C3112] [C3132 +C3123] [C3113 +C3131]
_

_
_

_
11
22
33
12
23
31
_

_
The components of C must be determined experimentally, by prescribing strains and measur-
ing stresses and vice versa. It is clear that only the summation of the components in the 3rd,
4th and 5th column can be determined and for that reason, it is assumed that the stiness
tensor is right-symmetric : 4C = 4C
rc
or equivalently C = C
RT
.
_

_
11
22
33
12
23
31
_

_
=
_

_
C1111 C1122 C1133 2C1121 2C1132 2C1113
C2211 C2222 C2233 2C2221 2C2232 2C2213
C3311 C3322 C3333 2C3321 2C3332 2C3313
C1211 C1222 C1233 2C1221 2C1232 2C1213
C2311 C2322 C2333 2C2321 2C2332 2C2313
C3111 C3122 C3133 2C3121 2C3132 2C3113
_

_
_

_
11
22
33
12
23
31
_

_
To restore the symmetry of the stiness matrix, the factor 2 in the last three columns is
swapped to the column with the strain components. The shear components are replaced
by the shear strains : 2ij = ij. This leads to a symmetric stiness matrix C with 21
independent components.
_

_
11
22
33
12
23
31
_

_
=
_

_
C1111 C1122 C1133 C1121 C1132 C1113
C2211 C2222 C2233 C2221 C2232 C2213
C3311 C3322 C3333 C3321 C3332 C3313
C1211 C1222 C1233 C1221 C1232 C1213
C2311 C2322 C2333 C2321 C2332 C2313
C3111 C3122 C3133 C3121 C3132 C3113
_

_
_

_
11
22
33
12
23
31
_

_
9.1 Material symmetry
Almost all materials have some material symmetry, originating from the micro structure,
which implies that the number of independent material parameters is reduced. The following
names refer to increasing material symmetry and thus to decreasing number of material
parameters :
monoclinic orthotropic quadratic transversal isotropic cubic isotropic
9.1.1 Monoclinic
In each material point of a monoclinic material there is one symmetry plane, which we take
here to be the (e1, e2)-plane. Strain components w.r.t. two vector bases e

= [e1 e2 e3]T and


e

= [e1 e2 e3]T must result in the same stresses. It can be proved that all components of
the stiness matrix, with an odd total of the index 3, must be zero. This implies :
179
C2311 = C2322 = C2333 = C2321 = C3111 = C3122 = C3133 = C3121 = 0
A monoclinic material is characterized by 13 material parameters. In the gure the directions
with equal properties are indicated with an equal number of lines.
Monoclinic symmetry is found in e.g. gypsum (CaSO42H2O).
1
2
3
Fig. 9.1 : One symmetry plane for
monoclinic material symmetry
C =
_

_
C1111 C1122 C1133 C1112 0 0
C2211 C2222 C2233 C2212 0 0
C3311 C3322 C3333 C3312 0 0
C1211 C1222 C1233 C1212 0 0
0 0 0 0 C2323 C2331
0 0 0 0 C3123 C3131
_

_
9.1.2 Orthotropic
In a point of an orthotropic material there are three symmetry planes which are perpendicular.
We choose them here to coincide with the Cartesian coordinate planes. In addition to the
implications for monoclinic symmetry, we can add the requirements
C1112 = C2212 = C3312 = C3123 = 0
An orthotropic material is characterized by 9 material parameters. In the stiness matrix,
they are now indicated as A, B, C, Q, R, S, K, L and M.
Orthotropic symmetry is found in orthorhombic crystals (e.g. cementite, Fe3C) and in
composites with bers in three perpendicular directions.
180
1
2
3
Fig. 9.2 : Three symmetry planes for
orthotropic material symmetry
C =
_

_
A Q R 0 0 0
Q B S 0 0 0
R S C 0 0 0
0 0 0 K 0 0
0 0 0 0 L 0
0 0 0 0 0 M
_

_
9.1.3 Quadratic
If in an orthotropic material the properties in two of the three symmetry planes are the same,
the material is referred to as quadratic. Here we assume the behavior to be identical in the e1-
and the e2-directions, however there is no isotropy in the (e1e2)-plane. This implies : A = B,
S = R and M = L. Only 6 material parameters are needed to describe the mechanical
material behavior.
Quadratic symmetry is found in tetragonal crystals e.g. TiO2 and white tin Sn.
1
2
3
Fig. 9.3 : Quadratic material
C =
_

_
A Q R 0 0 0
Q A R 0 0 0
R R C 0 0 0
0 0 0 K 0 0
0 0 0 0 L 0
0 0 0 0 0 L
_

_
9.1.4 Transversal isotropic
When the material behavior in the 12-plane is isotropic, an additional relation between pa-
rameters can be deduced. To do this, we observe a initial rectangular material area in the
181
12-plane, which is deformed by a shear stress . Because the material is isotropic in the
12-plane, the deformed material area is symmetric w.r.t. the directions 1 and 2, which are
principal strain directions.
For small isochoric (= without volume change) deformation in the 12-plane we can de-
duce an expression for the shear 12 from the geometry.
p

1
2

l
a
a

2
1
2
p
Fig. 9.4 : Pure shearing of a material element
a

2
a

2 +l
=
a
a +p
l = p

2 1 =
l
a

2
=
p
a
V = 0 1 +2 +3 = 0 2 = 1 =
p
a
= 12 = 2 sin() = 2
p
a
The 1- and 2-directions are principal strain and principal stress directions. With respect to
the 12-axes the stresses can be expressed in the strains with the same stiness components
as are already used for the 12-axes, because of the isotropy of the 12-plane. It is also known
from Mohrs stress circle, that 1 = and 2 = . This results in a relation between two
stiness parameters.
1 = A1 +Q2 = = K
2 = Q1 +A2 = = K
_

(AQ)(1 2) = 2K
1 = 2 =
1
2
_

(A Q) = 2K K =
1
2(A Q)
Examples of transversal isotropy are found in hexagonal crystals (CHP, Zn, Mg, Ti) and
honeycomb composites. The material behavior of these materials can be described with 5
material parameters.
182
1
2
3
Fig. 9.5 : Transversal material
C =
_

_
A Q R 0 0 0
Q A R 0 0 0
R R C 0 0 0
0 0 0 K 0 0
0 0 0 0 L 0
0 0 0 0 0 L
_

_
with K =
1
2(A Q)
9.1.5 Cubic
In the three perpendicular material directions the material properties are the same. In the
planes there is no isotropic behavior. Only 3 material parameters remain.
Examples of cubic symmetry are found in BCC and FCC crystals (e.g. in Ag, Cu, Au,
Fe, NaCl).
1
2
3
Fig. 9.6 : Cubic material
C =
_

_
A Q Q 0 0 0
Q A Q 0 0 0
Q Q A 0 0 0
0 0 0 L 0 0
0 0 0 0 L 0
0 0 0 0 0 L
_

_
9.1.6 Isotropic
In all three directions the properties are the same and in each plane the properties are
isotropic. Only 2 material parameters remain.
Isotropic material behavior is found for materials having a micro structure, which is suf-
ciently randomly oriented and distributed on a very small scale. This applies to metals with
a randomly oriented polycrystalline structure, ceramics with a random granular structure and
composites with random ber/particle orientation.
183
1
2
3
Fig. 9.7 : Isotropic material
C =
_

_
A Q Q 0 0 0
Q A Q 0 0 0
Q Q A 0 0 0
0 0 0 L 0 0
0 0 0 0 L 0
0 0 0 0 0 L
_

_
with L =
1
2(A Q)
9.2 Planar deformation
In many cases the state of strain or stress is planar. Both for plane strain and for plane stress,
only strains and stresses in a plane are related by the material law. Here we assume that
this plane is the 12-plane. For plane strain we than have 33 = 23 = 31 = 0, and for plane
stress 33 = 23 = 31 = 0. The material law for these planar situations can be derived from
the three-dimensional stress-strain relations. In the following sections the result is shown
for orthotropic material. For cases with more material symmetry, the planar stress-strain
relations can be simplied accordingly.
The planar stress-strain laws can be derived either from the stiness matrix C or from
the compliance matrix S.
_

_
11
22
33
12
23
31
_

_
=
_

_
A Q R 0 0 0
Q B S 0 0 0
R S C 0 0 0
0 0 0 K 0 0
0 0 0 0 L 0
0 0 0 0 0 M
_

_
_

_
11
22
33
12
23
31
_

_
;
_

_
11
22
33
12
23
31
_

_
=
_

_
a q r 0 0 0
q b s 0 0 0
r s c 0 0 0
0 0 0 k 0 0
0 0 0 0 l 0
0 0 0 0 0 m
_

_
_

_
11
22
33
12
23
31
_

_
S = C
1
k =
1
K
; l =
1
L
; m =
1
M
9.2.1 Plane strain and plane stress
For a plane strain state with 33 = 23 = 31 = 0, the stress 33 can be expressed in the
planar strains 11 and 22. The material stiness matrix C

can be extracted directly from


C. The material compliance matrix S

has to be derived by inversion.


184
33 = R11 +S22 =
r
c
11
s
c
22
C

=
_
_
A Q 0
Q B 0
0 0 K
_
_ =
_
_
A Q 0
Q B 0
0 0 K
_
_
S

=
_
_
a q 0
q b 0
0 0 k
_
_ = C
1
=
1
Q2 BA
_

_
B Q 0
Q A 0
0
Q2 BA
K
_

_
=
1
c
_
_
ac r2 qc rs 0
qc rs bc s2 0
0 0 kc
_
_
For the plane stress state, with 33 = 23 = 31 = 0, the two-dimensional material law can
be easily derived from the three-dimensional compliance matrix S

. The strain 33 can be


directly expressed in 11 and 22. The material stiness matrix has to be derived by inversion.
33 = a11 +s22 =
R
C
11
S
C
22
S

=
_
_
a q 0
q b 0
0 0 k
_
_ =
_
_
a q 0
q b 0
0 0 k
_
_
C

=
_
_
A Q 0
Q B 0
0 0 K
_
_ = S
1

=
1
q2 ba
_

_
b q 0
q a 0
0
q2 ba
k
_

_
=
1
C
_
_
AC R2 QC RS 0
QC RS BC S2 0
0 0 KC
_
_
In general we can write the stiness and compliance matrix for planar deformation as a 3 3
matrix with components, which are specied for plane strain (p = ) or plane stress (p = ).
C
p
=
_
_
Ap Qp 0
Qp Bp 0
0 0 K
_
_ ; S
p
=
_
_
ap qp 0
qp bp 0
0 0 k
_
_
185
9.3 Engineering parameters
In engineering practice the linear elastic material behavior is characterized by Youngs moduli,
shear moduli and Poisson ratios. They have to be measured in tensile and shear experiments.
In this chapter these parameters are introduced for isotropic material by analyzing a tensile
test and a shear test.
For orthotropic and transversal isotropic material, the stiness and compliance matrices,
expressed in engineering parameters, can be found in appendix C.
9.3.1 Isotropic
For isotropic materials the material properties are the same in each direction. The mechan-
ical behavior is characterized by two independent material parameters. Youngs modulus E
characterizes the tensile stiness and Poissons ratio determines the contraction. The shear
modulus G describes the shear behavior and is not independent but related to E and .
To express the compliance matrix in the parameters E, and G, three simple tests are
considered : a tensile test along the 1-axis, a shear test in the 13-plane and an hydrostatic
volume change.
_

_
11
22
33
12
23
31
_

_
=
_

_
A Q Q 0 0 0
Q A Q 0 0 0
Q Q A 0 0 0
0 0 0 L 0 0
0 0 0 0 L 0
0 0 0 0 0 L
_

_
_

_
11
22
33
12
23
31
_

= C

tensile test

T
=
_
d d 0 0 0

T
=
_
0 0 0 0 0

= A + 2Qd
0 = Q + (A +Q)d d =
Q
A+Q
=
_
_
_
= A 2Q = (A2Q) = E
Q(1 ) = A
A2Q = E
_
A =
(1 )E
(1 +)(1 2)

Q =
E
(1 +)(1 2)
= ; L =
1
2 (AQ) =
E
2(1 +)
=
shear test

T
=
_
0 0 0 0 0

T
=
_
0 0 0 0 0

= L =
E
2(1 +)
= G
volume change J = 112233 11 +22 +33
=
1 2
E
(11 +22 +33) =
1
K
1
3tr()
186
Besides the Youngs modulus, the Poisson ratio and the shear modulus, other material pa-
rameters can be used to characterize isotropic linear elastic behavior. The next tables list the
relations between all those parameters.
E, , G K, G E, G E, K
E E
(2G+3)G
+G
9KG
3K+G E E

2(+G)
3K2G
2(3K+G)
E2G
2G
3KE
6K
G
E
2(1+) G G G
3KE
9KE
K
E
3(12)
3+2G
3 K
EG
3(3GE) K

E
(1+)(12)
3K2G
3
G(E2G)
3GE
3K(3KE)
9KE
E, G, , K K,
E E 2G(1 +)
(1+)(12)

9K(K)
3K 3K(1 2)

E+

(E+)2+82
4

3K
G
3+E+

(3E)2+8E
4 G
(12)
2
3(K)
2
3K(12)
2(1+)
K
E3+

(E3)212E
6
2G(1+)
3(12)
(1+)
3 K K
2G
12 3K
1+
9.3.2 Compliance and stiness matrix
Now that the parameters A, Q and L are known as a function of engineering parameters,
these expressions can be substituted in the stiness matrix C. The compliance matrix S can
then be derived by inversion. Also the reduced matrices for plane strain and plane stress can
be derived straightforwardly by substitution, with reference to section 9.2.
All relevant compliance and stiness matrices for isotropic, orthotropic and transversally
isotropic material, are listed in appendix C.
9.4 Isotropic material tensors
Isotropic linear elastic material behavior is characterized by only two independent material
constants, for which we can choose Youngs modulus E and Poissons ratio .
187
9.4.1 Column/matrix notation of Hookes law
In column/matrix notation the strain components are related to the stress components by a
6 6 compliance matrix. Inversion leads to the 6 6 stiness matrix, which relates strain
components to stress components. It should be noted that shear strains are denoted as ij
and not as ij, as was done before.
_

_
11
22
33
12
23
31
_

_
=
1
E
_

_
1 0 0 0
1 0 0 0
1 0 0 0
0 0 0 1 + 0 0
0 0 0 0 1 + 0
0 0 0 0 0 1 +
_

_
_

_
11
22
33
12
23
31
_

= S

_
11
22
33
12
23
31
_

_
=
E
(1 +)
_

_
(1 )
(1 2)

(1 2)

(1 2)
0 0 0

(1 2)
(1 )
(1 2)

(1 2)
0 0 0

(1 2)

(1 2)
(1 )
(1 2)
0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
_

_
_

_
11
22
33
12
23
31
_

= C

The stiness matrix is written as the sum of two matrices, which can then be written in
matrix form.
C =
_

_
E
(1 +)(1 2)
_

_
1 1 1 0 0 0
1 1 1 0 0 0
1 1 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
_

_
+
E
(1 +)
_

_
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
_

_
_

_
_

_
11
22
33
12
23
31
_

_
Tensorial notation
The rst matrix is the matrix representation of the fourth-order tensor II. The second matrix
is the representation of the symmetric fourth-order tensor 4I
s
. The resulting fourth-order
material stiness tensor 4C contains two material constants c0 and c1, which are related to
E and .
=
_
c0II +c1
4
I
s
: =
4
C :
with
4
I
s
=
1
2
_
4
I +
4
I
rc_
and c0 =
E
(1 +)(1 2)
; c1 =
E
1 +
188
Hydrostatic and deviatoric strain/stress
The strain and stress tensors can both be written as the sum of an hydrostatic - (.)h - and a
deviatoric - (.)d - part. Doing so, the stress-strain relation can be easily inverted.
=
4
C : =
_
c0II +c1
4
I
s
:
= c0tr()I +c1 = c0tr()I +c1
_

d
+ 1
3tr()I
_
= (c0 + 1
3c1)tr()I +c1
d
= (3c0 +c1)1
3 tr()I +c1
d
= (3c0 +c1)
h
+c1
d
=
h
+
d

=
h
+
d
=
1
3c0 +c1

h
+
1
c1

d
=
1
3c0 +c1
1
3tr()I +
1
c1
_

1
3tr()I
_
=
c0
(3c0 +c1)c1
tr()I +
1
c1
=
_

c0
(3c0 +c1)c1
II +
1
c1
4
I
s
_
:
=
4
S :
9.5 Thermo-elasticity
A temperature change T of an unrestrained material invokes deformation. The total strain
results from both mechanical and thermal eects and when deformations are small the total
strain can be written as the sum of mechanical strains m and thermal strains T . The ther-
mal strains are related to the temperature change T by the coecient of thermal expansion
tensor A. For thermally isotropic materials only the linear coecient of thermal expansion
is relevant.
The stresses in terms of strains are derived by inversion of the compliance matrix S.
= m +T =
4
S : +AT

m +

T = S

+AT
=
4
C : ( AT)

= C (

AT)
9.5.1 Isotropic material
For mechanical and thermal isotropy, the stress strain relation can be simplied, using the
usual elasticity parameters.
189
=
4
C : ( T I)
= c0tr( TI)I +c1( TI)
= c0tr()I +c1 (3c0 +c1)T I
= tr()I + 2 (3 + 2)T I
=
E
(1 +)(1 2)
tr()I +
E
1 +

E
1 2
T I
9.5.2 Plane strain/stress
When the material is in plane strain or plane stress state, only three strains and stresses are
relevant, here indicated with indices 1 and 2. The thermo-elastic stress-strain law can than be
expressed in the elasticity parameters and the coecient of thermal expansion. The relation
is presented for orthotropic material. It is assumed that the thermal expansion is isotropic.
_
_
11
22
12
_
_ =
_
_
ap qp 0
qp bp 0
0 0 k
_
_
_
_
11
22
12
_
_ T
_
_
1
1
0
_
_
11 = Ap(11 T) +Qp(22 T) = Ap11 +Qp22 (Ap +Qp)T
22 = Qp(11 T) +Bp(22 T) = Qp11 +Bp22 (Qp +Bp)T
12 = K12
_
_
11
22
12
_
_ =
_
_
Ap Qp 0
Qp Bp 0
0 0 K
_
_
_
_
11
22
12
_
_ T
_
_
Ap +Qp
Bp +Qp
0
_
_
190
Chapter 10
Elastic limit criteria
Loading of a material body causes deformation of the structure and, consequently, strains
and stresses in the material. When either strains or stresses (or both combined) become too
large, the material will be damaged, which means that irreversible micro structural changes
will result. The structural and/or functional requirements of the structure or product will be
hampered, which is referred to as failure.
Their are several failure modes, listed in the table, each of them associated with a failure
mechanism. In the following we will only consider plastic yielding. When the stress state
exceeds the yield limit, the material behavior will not be elastic any longer. Irreversible
microstructural changes (crystallographic slip in metals) will cause permanent (= plastic)
deformation.
failure mode mechanism
plastic yielding crystallographic slip (metals)
brittle fracture (sudden) breakage of bonds
progressive damage micro-cracks growth coalescence
fatigue damage/fracture under cyclic loading
dynamic failure vibration resonance
thermal failure creep / melting
elastic instabilities buckling plastic deformation
10.1 Yield function
In a one-dimensional stress state (tensile test), yielding will occur when the absolute value
of the stress reaches the initial yield stress y0. This can be tested with a yield criterion,
where a yield function f is used. When f < 0 the material behaves elastically and when
f = 0 yielding occurs. Values f > 0 cannot be reached.
191
192
f() =
2

2
y0 = 0 g() =
2
=
2
y0 = gt
gt = limit in tensile test
y0 = l
y0
y0
= n
Fig. 10.1 : Tensile curve with initial yield stress
In a three-dimensional stress space, the yield criterion represents a yield surface. For elastic
behavior (f < 0) the stress state is located within the yield surface and for f = 0, the
stress state is on the yield surface. Because f > 0 cannot be realized, stress states outside
the yield surface can not exist. For isotropic material behavior, the yield function can be
expressed in the principal stresses 1, 2 and 3. It can be visualized as a yield surface in the
three-dimensional principal stress space.
f() = 0 g() = gt : yield surface in 6D stress space
f(1, 2, 3) = 0 g(1, 2, 3) = gt : yield surface in 3D principal stress space
1
2
3
Fig. 10.2 : Yield surface in three-dimensional principal stress space
193
10.2 Principal stress space
The three-dimensional stress space is associated with a material point and has three axes, one
for each principal stress value in that point. In the origin of the three-dimensional principal
stress space, where 1 = 2 = 3 = 0, three orthonormal vectors {e1, e2, e3} constitute a
vector base. The stress state in the material point is characterized by the principal stresses
and thus by a point in stress space with coordinates 1, 2 and 3. This point can also be
identied with a vector , having components 1, 2 and 3 with respect to the vector base
{e1, e2, e3}.
The hydrostatic axis, where 1 = 2 = 3 can be identied with a unit vector ep.
Perpendicular to ep in the (e1, ep)-plane a unit vector eq can be dened. Subsequently the
unit vector er is dened perpendicular to the (ep, eq)-plane.
The vectors eq and er span the so-called -plane perpendicular to the hydrostatic axis.
Vectors ep, eq and er constitute a orthonormal vector base. A random unit vector in the
-plane can be expressed in eq and er.

ep
e1
e2
e3
ep
er
eq
et()

hydrostatic axis ep =
1
3

3(e1 +e2 +e3) with ||ep|| = 1


plane perpendicular to hydrostatic axis
e

q = e1 (ep e1)ep = e1
1
3(e1 +e2 +e3) =
1
3(2e1 e2 e3)
eq =
1
6

6(2e1 e2 e3)
er = ep eq =
1
3

3(e1 +e2 +e3)


1
6

6(2e1 e2 e3) =
1
2

2(e2 e3)
vector in -plane et() = cos()eq sin()er
A stress state can be represented by a vector in the principal stress space. This vector can be
written as the sum of a vector along the hydrostatic axis and a vector in the -plane. These
vectors are referred to as the hydrostatic and the deviatoric part of the stress vector.
= 1e1 +2e2 +3e3 =
h
+
d

h
= ( ep)ep =
h
ep = 1
3

3(1 +2 +3)ep =

3mep

h
=
1
3

3(1 +2 +3)
194

d
= ( ep)ep
= 1e1 +2e2 +3e3
1
3

3(1 +2 +3)
1
3

3(e1 +e2 +e3)


= 1e1 +2e2 +3e3 1
3(1e1 +2e1 +3e1 +1e2 +2e2 +3e2 +1e3 +2e3 +3e3)
=
1
3 {(21 2 3)e1 + (1 + 22 3)e2 + (1 2 + 23)e3}

d
= ||
d
|| =

d d
= 1
3
_
(21 2 3)2 + (1 + 22 3)2 + (1 2 + 23)2
=
_
2
3(2
1 +2
2 +2
3 12 23 31)
=

d : d
Because the stress vector in the principal stress space can also be written as the sum of three
vectors along the base vectors e1, e2 and e3, the principal stresses can be expressed in h and
d.
=
h
+
d
=
h
ep +
d
et()
=
h
ep +
d
{cos()eq sin()er}
=
h 1
3

3(e1 +e2 +e3) +


d
{cos()
1
6

6(2e1 e2 e3) sin()


1
2

2(e2 e3)}
= {
1
3

3
h
+
1
3

6
d
cos()}e1 +
{
1
3

3
h

1
6

6
d
cos()
1
2

2
d
sin()}e2 +
{
1
3

3
h

1
6

6
d
cos() +
1
2

2
d
sin()}e3
= 1e1 +2e2 +3e3
10.3 Yield criteria
In the following sections, various yield criteria are presented. Each of them starts from a
hypothesis, stating when the material will yield. Such a hypothesis is based on experimental
observation and is valid for a specic (class of) material(s).
The yield criteria can be visualized in several stress spaces:
the two-dimensional (1, 2)-space for plane stress states with 3 = 0,
the three-dimensional (1, 2, 3)-space,
the -plane and
the (, )-plane, where Mohrs circles are used.
10.3.1 Maximum stress/strain
The maximum stress/strain criterion states that
yielding occurs when one of the stress/strain components exceeds a limit value.
195
This criterion is used for orthotropic materials.
maximum stress (11 = sX) (22 = sY ) (12 = sS)
maximum strain (11 = eX) (22 = eY ) (12 = eS)
10.3.2 Maximum principal stress (Rankine)
The maximum principal stress (or Rankine) criterion states that
yielding occurs when the maximum principal stress reaches a limit value.
The absolute value is used to arrive at the same elasticity limit in tension and compression.
The Rankine criterion is used for brittle materials like cast iron. At failure these materials
show cleavage fracture.
max = max(|i| ; i = 1, 2, 3) = maxt = y0
The gure shows the yield surface in the principal stress space for a plane stress state with
3 = 0.
1
2
Fig. 10.3 : Rankine yield surface in two-dimensional principal stress space
In the three-dimensional stress space the yield surface is a cube with side-length 2y0.
1
2
3
Fig. 10.4 : Rankine yield surface in three-dimensional principal stress space
196
In the (, )-space the Rankine criterion is visualized by to limits, which can not be exceeded
by the absolute maximum of the principal stress.
y0 y0

0
Fig. 10.5 : Rankin yield limits in (, )-space
10.3.3 Maximum principal strain (Saint Venant)
The maximum principal strain (or Saint Venant) criterion states that
yielding occurs when the maximum principal strain reaches a limit value.
From a tensile experiment this limit value appears to be the ratio of uni-axial yield stress and
Youngs modulus.
For 1 > 2 > 3, the maximum principal strain can be calculated from Hookes law
and its limit value can be expressed in the initial yield value yo and Youngs modulus E.
1 =
1
E
1

E
2

E
3 =
y0
E
1 2 3 = y0
For other sequences of the principal stresses, relations are similar and can be used to construct
the yield curve/surface in 2D/3D principal stress space.
max = max(|i| ; i = 1, 2, 3) = maxt =
y0
E
197
2
1
1 2 = y0
2 1 = y0
y0 y0
Fig. 10.6 : Saint-Venants yield curve in two-dimensional principal stress space
10.3.4 Tresca
The Tresca criterion (+ Coulomb, Mohr, Guest (1864)) states that
yielding occurs when the maximum shear stress reaches a limit value.
In a tensile test the limit value for the shear stress appears to be half the uni-axial yield stress.
max = 1
2 (max min) = maxt = 1
2 y0 TR = max min = y0
Using Mohrs circles, it is easily seen how the maximum shear stress can be expressed in the
maximum and minimum principal stresses.
For the plane stress case (3 = 0) the yield curve in the 12-plane can be constructed
using Mohrs circles. When both principal stresses are positive numbers, the yielding occurs
when the largest reaches the one-dimensional yield stress y0. When 1 is positive (= tensile
stress), compression in the perpendicular direction, so a negative 2, implies that 1 must
decrease to remain at the yield limit. Using Mohrs circles, this can easily be observed.
198
2
1
2 = y0
1 = 0
2 = 0
1 = y0
1 = 0
2 = y0
2 1
2
1 = y0
Fig. 10.7 : Tresca yield curve in two-dimensional principal stress space
1 0 ; 2 0 max = 1|2 =
1
2y0
1 0 ; 2 < 0 max =
1
2 (1 2) =
1
2 y0
Adding an extra hydrostatic stress state implies a translation in the three-dimensional prin-
cipal stress space
{1, 2, 3} {1 +c, 2 +c, 2 +c}
i.e. a translation parallel to the hydrostatic axis where 1 = 2 = 3. This will never result in
yielding or more plastic deformation, so the yield surface is a cylinder with its axis coinciding
with (or parallel to) the hydrostatic axis.
In the -plane, the Tresca criterion is a regular 6-sided polygonal.
199
1 = 2 = 3
30o
2 1
3
1
2
3
1 = 2 = 3
Fig. 10.8 : Tresca yield surface in three-dimensional principal stress space and the
-plane
max
max

min

Fig. 10.9 : Mohrs circles and Tresca yield limits in (, )-space


10.3.5 Von Mises
According to the Von Mises elastic limit criterion (+ Hubert, Hencky (1918)),
yielding occurs when the specic shape deformation elastic energy reaches a critical
value.
The specic shape deformation energy is also referred to as distortional energy or deviatoric
energy or shear strain energy. It can be derived by splitting up the total specic elastic energy
W into a hydrostatic part Wh and a deviatoric part Wd.
The deviatoric Wd can be expressed in d and the hydrostatic Wh can be expressed in
the mean stress m =
1
3tr().
200
W = 1
2 : = 1
2 :
4
S : = 1
2
_

h
+
d
_
:
4
S :
_

h
+
d
_

h
= 1
3tr()I

d
=
1
3tr()I I :
d
= tr(
d
) = 0
4
S =

E
II +
1 +
E
4
I
s
=
1
2
_

h
+
d
_
:
_

E
tr()I +O +
1 +
3E
tr()I +
1 +
E

d
_
=
1
2
_

h
+
d
_
:
_
1 2
3E
tr()I +
1 +
E

d
_
=
1
2
_
1 2
3E
tr
2
() +
1 +
E

d
:
d
_
=
1
18K
tr
2
() +
1
4G

d
:
d
= W
h
+W
d
The deviatoric part is sometimes expressed in the second invariant J2 of the deviatoric stress
tensor. This shape deformation energy Wd can be expressed in the principal stresses. For the
tensile test the shape deformation energy Wd
t can be expressed in the yield stress y0. The
Von Mises yield criterion Wd = Wd
t can than be written as V M = y0, where V M is the
equivalent or eective Von Mises stress, a function of all principal stresses.
The equivalent Von Mises stress V M is sometimes replaced by the octahedral shear
stress oct =
1
3

2 V M.
W
d
=
1
4G

d
:
d
=
1
4G
2J2
=
1
4G
_

1
3 tr()I
_
:
_

1
3tr()I
_
=
1
4G
_
:
1
3tr
2
()I

=
1
4G
_

2
1 +
2
2 +
2
3
1
3 (1 +2 +3)
2
_
=
1
4G
1
3
_
(1 2)
2
+ (2 3)
2
+ (3 1)
2

W
d
t =
1
4G
2
3
2
t =
1
4G
2
3
2
y0
W
d
= W
d
t
1)
1
2
_
(1 2)
2
+ (2 3)
2
+ (3 1)
2
_
=
2
y0
V M =
_
1
2 {(1 2)2 + (2 3)2 + (3 1)2} = y0
2)
1
2
d
:
d
=
1
3
2
y0 V M =
_
3
2d : d =
_
3J2 = y0
The Von Mises yield criterion can be expressed in Cartesian stress components.
201
2
3
2
V M = tr(
d

d
) with
d
= 1
3tr()I
=
_
2
3 xx
1
3 yy
1
3 zz
_2
+
2
xy +
2
xz +
_
2
3 yy
1
3 zz
1
3xx
_2
+
2
yz +
2
yx +
_
2
3 zz
1
3xx
1
3yy
_2
+
2
zx +
2
zy
= 2
3
_

2
xx +
2
yy +
2
zz
_
2
3 (xxyy +yyzz +zzxx) + 2
_

2
xy +
2
yz +
2
zx
_
For plane stress (3 = 0), the yield curve is an ellipse in the 12-plane. The length of the
principal axes of the ellipse is

2y0 and
_
2
3y0.
2
1
Fig. 10.10 : Von Mises yield curve in two-dimensional principal stress space
The three-dimensional Von Mises yield criterion is the equation of a cylindrical surface in
three-dimensional principal stress space. Because hydrostatic stress does not inuence yield-
ing, the axis of the cylinder coincides with the hydrostatic axis 1 = 2 = 3.
In the -plane, the Von Mises criterion is a circle with radius
_
2
3y0.
1 = 2 = 3
30o
2 1
3
1
2
3 1 = 2 = 3
_
2
3y0
Fig. 10.11 : Von Mises yield surface in three-dimensional principal stress space and
the -plane
202
10.3.6 Beltrami-Haigh
According to the elastic limit criterion of Beltrami-Haigh,
yielding occurs when the total specic elastic energy W reaches a critical value.
W =
1
18K
tr
2
() +
1
4G

d
:
d
=
1
18K
(1 +2 +3)
2
+
1
4G
_

2
1 +
2
2 +
2
3
1
3 (1 +2 +3)
2
_
=
_
1
18K

1
12G
_
(1 +2 +3)
2
+
1
4G
_

2
1 +
2
2 +
2
3
_
Wt =
_
1
18K

1
12G
_

2
+
1
4G

2
=
1
2E

2
=
1
2E

2
y0
2E
_
1
18K

1
12G
_
(1 +2 +3)
2
+
2E
4G
_

2
1 +
2
2 +
2
3
_
=
2
y0
The yield criterion contains elastic material parameters and thus depends on the elastic
properties of the material. In three-dimensional principal stress space the yield surface is an
ellipsoid. The longer axis coincides with (or is parallel to) the hydrostatic axis 1 = 2 = 3.
2
1
3
1 = 2 = 3
2
1
Fig. 10.12 : Beltrami-Haigh yield curve and surface in principal stress space
10.3.7 Mohr-Coulomb
A prominent dierence in behavior under tensile and compression loading is seen in much
materials, e.g. concrete, sand, soil and ceramics. In a tensile test such a material may have
a yield stress ut and in compression a yield stress uc with uc > ut. The Mohr-Coulomb
yield criterion states that
yielding occurs when the shear stress reaches a limit value.
203
For a plane stress state with 3 = 0 the yield contour in the 1, 2-plane can be constructed
in the same way as has been done for the Tresca criterion.
ut
uc
uc
1
2
ut
1
ut

2
uc
= 1
Fig. 10.13 : Mohr-Coulomb yield curve in two-dimensional principal stress space
The yield surface in the three-dimensional principal stress space is a cone with axis along the
hydrostatic axis.
The intersection with the plane 3 = 0 gives the yield contour for plane stress.
1 = 2 = 3
30o
2 1
3
1
2
3
1 = 2 = 3
Fig. 10.14 : Mohr-Coulomb yield surface in three-dimensional principal stress space
and the -plane
10.3.8 Drucker-Prager
For materials with internal friction and maximum adhesion, yielding can be described by the
Drucker-Prager yield criterion. It relates to the Mohr-Coulomb criterion in the same way as
the Von Mises criterion relates to the Tresca criterion.
204

3 J1 +
_
3J2 = y0
For a plane stress state with 3 = 0 the Drucker-Prager yield contour in the 12-plane is a
shifted ellipse.
2
1
Fig. 10.15 : Drucker-Prager yield curve in two-dimensional principal stress space
In three-dimensional principal stress space the Drucker-Prager yield surface is a cone with
circular cross-section.
1 = 2 = 3
30o
2 1
3
1
2
3
1 = 2 = 3
Fig. 10.16 : Drucker-Prager yield surface in three-dimensional principal stress space
and the -plane
205
10.3.9 Other yield criteria
There are many more yield criteria, which are used for specic materials and loading condi-
tions. The criteria of Hill, Homan and Tsai-Wu are used for orthotropic materials. In these
criterion, there is a distinction between tensile and compressive stresses and their respective
limit values.
Parabolic Drucker-Prager
_
3J2 +

3 y0J1
_1
2
= y0
Buyokozturk
_
3J2 +

3 y0J1 0.2J
2
1
_1
2
= y0
Hill
2
11
X2

1122
XY
+
2
22
Y 2
+
2
12
S2
Homan
_
1
Xt

1
Xc
_
11 +
_
1
Yt

1
Yc
_
22 +
_
1
XtXc
_

2
11 +
_
1
YtYc
_

2
22 +
_
1
S2
_

2
12
_
1
XtXc
_
1122 = 0
Tsai-Wu
_
1
Xt

1
Xc
_
11 +
_
1
Yt

1
Yc
_
22 +
_
1
XtXc
_

2
11 +
_
1
YtYc
_

2
22 +
_
1
S2
_

2
12 + 2F12 1122 = 0
with F
2
12 >
1
XtXc
1
YtYc
206
Chapter 11
Solution strategies
11.1 Governing equations for unknowns
The deformation of a three-dimensional continuum in three-dimensional space is described
by the displacement vector u of each material point. Due to the deformation, stresses arise
and the stress state is characterized by the stress tensor . For static problems, this tensor
has to satisfy the equilibrium equations. Solving stresses from these equations is generally
not possible and additional equations are needed, which relate stresses to deformation. These
constitutive equations, which describe the material behavior, relate the stress tensor to the
strain tensor , which is a function of the displacement gradient tensor ( u). Components
of this strain tensor cannot be independent and are related by the compatibility equations.
unknown variables
displacements u = u(x)
deformation tensor F =
_
0x
_C
stresses
equations
compatibility
2
{tr()} ( )
c
= 0
equilibrium
c
+ q = u ; =
T
material law = (F)
11.2 Boundary conditions
Some of the governing equations are partial dierential equations, where dierentiation is
done w.r.t. the spatial coordinates. These dierential equations can only be solved when
proper boundary conditions are specied. In each boundary point of the material body, ei-
ther the displacement or the load must be prescribed. It is also possible to specify a relation
207
208
between displacement and load in such a point.
When the acceleration of the material points cannot be neglected, the equilibrium equa-
tion becomes the equation of motion, with u as its right-hand term. In that case a solution
can only be determined when proper initial conditions are prescribed, i.e. initial displacement,
velocity or acceleration. In this section we will assume u =0.

c
+ q =0 x V
u = up x Au
p = n = pp x Ap
11.2.1 Saint-Venants principle
The so-called Saint-Venant principle states that, if a load on a structure is replaced by a
statically equivalent load, the resulting strains and stresses in the structure will only be
altered near the regions where the load is applied. With this principle in mind, the real
boundary conditions can often be modeled in a simplied way. Concentrated forces can for
instance be replaced by distributed loads, and vice versa. Stresses and strains will only dier
signicantly in the neighborhood of the boundary, where the load is applied.
b
P
b
x

(x)
Fig. 11.1 : Saint-Venant principle
P =
_
A
(x) dA = A ; A = b t
11.2.2 Superposition
Under the assumption of small deformations and linear elastic material behavior, the govern-
ing equations, which must be solved to determine deformation and stresses (= solution S)
209
are linear. When boundary conditions (xations and loads (L)), which are needed for the
solution, are also linear, the total problem is linear and the principle of superposition holds.
The principle of superposition states that the solution S for a given combined load
L = L1 + L2 is the sum of the solution S1 for load L1 and the solution S2 for L2, so :
S = S1 +S2.
F1
F2
u1
u2
F2
F1
u1 +u2
Fig. 11.2 : Principle of superposition
11.3 Solution : displacement method
In the displacement method the constitutive relation for the stress tensor is substituted in the
force equilibrium equation.
Subsequently the strain tensor is replaced by its denition in terms of the displacement
gradient. This results in a dierential equation in the displacement u, which can be solved
when proper boundary conditions are specied.
In a Cartesian coordinate system the vector/tensor formulation can be replaced by index
notation. It is elaborated here for the case of linear elasticity theory.
210

c
+ q =0
=
4
C :
_


_
4
C :
_c
+ q =0
=
1
2
__
u
_c
+
_
u
__
_
_
_


_
4
C :
_
u
__c
+ q =0 u
Cartesian index notation
ij,j +qi = 0i
ij = Cijlklk
_

Cijkllk,j +qi = 0i
lk =
1
2 (ul,k +uk,l)
_

Cijklul,kj +qi = 0i ui ij ij
11.3.1 Navier equations
The displacement method is elaborated for planar deformation in a Cartesian coordinate sys-
tem. Linear deformation and linear elastic material behavior is assumed. Elimination and
substitution results in two partial dierential equations for the two displacement components.
For the sake of simplicity, we do not consider thermal loading here.
xx,x +xy,y +qx = ux ; yx,x +yy,y +qy = uy
xx = Apxx +Qpyy
yy = Qpxx +Bpyy
xy = 2Kxy
_
_
_
Apxx,x +Qpyy,x + 2Kxy,y +qx = ux
2Kxy,x +Qpxx,y +Bpyy,y +qy = uy
_
Apux,xx +Qpuy,yx +K(ux,yy +uy,xy) +qx = ux
K(ux,yx +uy,xx) +Qpux,xy +Bpuy,yy +qy = uy
_
Apux,xx +Kux,yy + (Qp +K)uy,yx +qx = ux
Kuy,xx +Bpuy,yy + (Qp +K)ux,xy +qy = uy
_
11.3.2 Axi-symmetric with ut = 0
Many engineering problems present a rotational symmetry w.r.t. an axis. They are axi-
symmetric. In many cases the tangential displacement is zero : ut = 0. This implies that
there are no shear strains and stresses.
displacements ur = ur(r) ; uz = uz(r, z)
strains rr = ur,r ; tt =
1
r ur ; zz = uz,z
stresses tz = 0 ; rz 0 ; tr = 0
211
eq. of motion rr,r +
1
r
(rr tt) +qr = ur
The radial and tangential stresses are related to the radial and tangential strains by the planar
material law. Material parameters are indicated as Ap, Bp and Qp and can later be specied
for a certain material and for plane strain or plane stress. With the strain-displacement
relations the equation of motion can be transformed into a dierential equation for the radial
displacement ur
rr = Aprr +Qptt (Ap +Qp)T
tt = Qprr +Bptt (Bp +Qp)T
_
eq. of motion
ur,rr +
1
r
ur,r
Bp
Ap
1
r2
ur =

Ap
( ur qr) +
Ap +Qp
Ap
(T),r +
Ap Bp
Ap
1
r
T
ur,rr +
1
r
ur,r
2 1
r2
ur = f(r) with =

Bp
Ap
11.4 Solution : stress method
In the stress method, the constitutive relation for the strain tensor is substituted in the
compatibility equation, resulting in a partial dierential equation for the stress tensor. This
equation and the equilibrium equations constitute a set of coupled equations from which the
stress tensor has to be solved.
For planar problems, this can be elaborated and results in the Beltrami-Mitchell equation
for the stress components. It is again assmed that deformations are small and the material
behavior is linearly elastic.
Solution of the stress equation(s) is done by introducing the so-called Airy stress function.
11.4.1 Beltrami-Mitchell equation
The compatibility equation for planar deformation can be expressed in stress components,
resulting in the Beltrami-Mitchell equation.
xx,yy +yy,xx = 2xy,xy
xx = apxx +qpyy
yy = qpxx +bpyy
xy =
1
2 kxy
_
_
_

_
_
_
kxy,xy =
apxx,yy +qpyy,yy+
qpxx,xx +bpyy,xx
equilibrium
xx,x +xy,y = qx xy,xy +xx,xx = qx,x
yx,x +yy,y = qy xy,xy +yy,yy = qy,y
_

2xy,xy +xx,xx +yy,yy = qx,x qy,y
212
Beltrami-Mitchell equation
(k + 2qp) (xx,xx +yy,xx) + 2apxx,yy + 2bpyy,yy = k(qx,x +qy,y) +k( ux,x + uy,y)
11.4.2 Beltrami-Mitchell equation for thermal loading
With thermal strains, the compatibility equation for planar deformation can again be ex-
pressed in stress components. Combination with the equilibrium equations results in the
Belrami-Mitchell equation for thermal loading.
xx,yy +yy,xx = 2xy,xy
xx = apxx +bpyy +T
yy = bpxx +apyy +T
xy = (ap bp)xy
_
_
_

_
_
_
2(ap bp)xy,xy =
apxx,yy +bpyy,yy+
bpxx,xx +apyy,xx
(T),xx +(T),yy
equilibrium
2xy,xy +xx,xx +yy,yy = qx,x qy,y
Beltrami-Mitchell equation
xx,xx +yy,xx +xx,yy +yy,yy =
ap bp
a
(qx,x +qy,y)
1
ap
{(T)xx + (T)yy}
11.4.3 Airy stress function
In the stress function method an Airy stress function is introduced and the stress tensor is
related to it in such a way that the tensor obeys the equilibrium equations

c
=0
Using Hookes law, the strain tensor can be expressed in the Airy function. Substitution of
this () relation in one of the compatibility equations results in a partial dierential equation
for the Airy function, which can be solved with the proper boundary conditions.
In a Cartesian coordinate system the vector/tensor formulation can be replaced by index
notation.
The material compliance tensor is :
4
S =

E
II +
1 +
E
4
I
s
Airy stress function : (x)
= ( ) +
_

_
I
=
4
S :
_
_
_

=
4
S :
_
( ) + (
2
)I
_

2
(tr())
_

_c
= 0
_
_
_

2
(
2
) =
4
= 0
213
Cartesian index notation
Airy stress function : (xi)
ij = ,ij +ij,kk
ij = Sijlk kl
_
_
_

ij = Sijlk ( ,kl +kl,mm)


ii,jj ij,ij = 0
_
_
_
,iijj = 0 ij ij
Planar, Cartesian
The stress function method is elaborated for planar deformation in a Cartesian coordinate
system.
xx = ,xx +xx(,xx +,yy) = ,yy
yy = ,yy +yy(,xx +,yy) = ,xx
xy = ,xy
xx = apxx +qpyy
yy = qpxx +bpyy
xy =
1
2kxy
_
_
_

xx = ap,yy +qp,xx
yy = qp,yy +bp,xx
xy =
1
2k,xy
xx,yy +yy,xx = 2xy,xy
_
_
_
bp,xxxx +ap,yyyy + (2qp +k),xxyy = 0
isotropic ap = bp =
1
E
; qp =

E
; k =
2(1 +)
E

bi-harmonic equation ,xxxx +,yyyy + 2,xxyy = 0
Planar, cylindrical
In a cylindrical coordinate system, the bi-harmonic equation can be derived by transformation.
gradient and Laplace operator
= er

r
+et
1
r

2
= =
2
r2
+
1
r

r
+
1
r2
2
2
+
2
z2
2D

2
=
2
r2
+
1
r

r
+
1
r2
2
2
bi-harmonic equation
_
2
r2
+
1
r

r
+
1
r2
2
2
__
2
r2
+
1
r

r
+
1
r2
2
2
_
= 0
stress components
214
rr =
1
r

r
+
1
r2
2
2
; tt =
2
r2
rt =
1
r2


1
r
2
r
=

r
_
1
r

_
11.5 Weighted residual formulation
For an approximation, the equilibrium equation is not satised exactly in each material point.
The error can be smeared out over the material volume, using a weighting function w(x).
equilibrium equation
c
+ q =0 x V
approximation residual
c
+ q = (x) =0 x V
weighted error is smeared out
_
V
w(x) (x) dV
When the weighted residual integral is satised for each allowable weighting function w, the
equilibrium equation is satised in each point of the material.
_
V
w
_

c
+ q
_
dV = 0 w(x)
c
+ q =0 x V
In the weighted residual integral, one term contains the divergence of the stress tensor. This
means that the integral can only be evaluated, when the derivatives of the stresses are con-
tinuous over the domain of integration. This requirement can be relaxed by applying partial
integration to the term with the stress divergence. The result is the so-called weak formula-
tion of the weighted residual integral.
Gauss theorem is used to transfer the volume integral with the term .( ) to a surface
integral. Also p = n = n c and = c is used.
_
V
w
_

c
+ q
_
dV = 0
(
c
w) = ( w)
c
:
c
+ w (
c
)
_
_
_

_
V
_
(
c
w) ( w)
c
:
c
+ w q
_
dV = 0 w
_
V
( w)
c
:
c
dV =
_
V
w q dV +
_
A
n
c
wdA w
_
V
( w)
c
: dV =
_
V
w q dV +
_
A
w p dA = fe( w) w
215
11.5.1 Weighted residual formulation for linear deformation
When deformation and rotations are small, the deformation is geometrically linear. The
deformed state is almost equal to the undeformed state.
_
V0
( 0 w)
c
: dV0 =
_
V0
w q dV0 +
_
A0
w p dA0 = fe( w) w
The material behavior is described by Hookes law, which can be substituted in the weighted
residual integral, according to the displacement solution method.
=
4
C : =
4
C :
1
2
_
( 0u) + ( 0u)
c
_
=
4
C : ( 0u)
The weighted residual integral is now completely expressed in the displacement u. Approxi-
mate solutions can be determined with the nite element method.
_
V0
( 0 w)
c
:
4
C : ( 0u) dV0 =
_
V0
w q dV0 +
_
A0
w p dA0 = fe0( w) w
This tensor relation can be written in matrix/column notation. We use the column notation
of the vector gradient as it was introduced in section 4.2.3 : L0w = ( 0 w)
c
L

0w
and
L0u = ( 0u)
c
L

0u
where L

is the column with the derivatives w.r.t. the coordinates.


_
V0
_
L

0w
_T
t
C
_
L

0u
_
t
dV0 = fe0(w

) w

11.5.2 Total Lagrange formulation


When deformations are large geometrically nonlinear , the current volume of the material
is unknown, which means that the weighted residual integral can not be evaluated. Transfor-
mation of this integral is always possible. Besides the integral also the gradient operator must
be transformed. The conguration, which is the target of the transformation is the reference
conguration.
The rst thing we can think of is a transformation to the undeformed conguration t0.
This transformation results in the Total Lagrange formulation. The second Piola-Kirchho
stress tensor is mostly used in this case to represent the stress state.
216
_
V
( w)
c
: dV = fe( w) w(x)
transformation to undeformed conguration t0
= F
c
0 ( w)
c
= ( 0 w)
c
F
1
dV = det(F)dV0 = JdV0
weighted residual integral
_
V0
( 0 w)
c
F
1
: J dV0 = fe0( w) w(x)
P = JF
1
F
c
_
_
_

_
V0
( 0 w)
c
: (P F
c
) dV0 = fe0( w) w(x)
Iterative solution process
In the Total Lagrange formulation the weighted residual integral is transformed from the
current conguration Cc to the initial undeformed conguration C0. Unknown variables in
the integral are the total deformation tensor F and the 2nd-Piola-Kirchho stress tensor P.
To describe the essential steps of the iteration procedure, it is assumed that an approx-
imate state C
c is determined with values for F

and P

. The unknown current values are


written as F = F

+ F and P = P

+ P, where (.) indicates the dierence between C


c
and Cc. The iterative change of the deformation tensor F can be expressed in the iterative
displacement x = u.
_
V0
( 0 w)
c
: (P F
c
) dV0 = fe0( w) w(x)
F = ( 0x)
c
= { 0(x

+x)}
c
= ( 0x

)
c
+ ( 0x)
c
= F

+F = F

+L0u
P = P

+P
_
_
_

_
V0
( 0 w)
c
: (P

+P) (F

+L0u)
c
dV0 = fe0( w) w(x)
It is assumed that the iterative displacement, its gradient and the stress variation are very
small and then the weighted residual integral is linearized with respect to u. In analogy with
L0u, L0w = ( 0 w)c is introduced.
217
_
V0
L0w : (P

+P) (F

+L0u)
c
dV0 = fe0( w) w(x)
_
V0
L0w : (P

F
c
+P

L
c
0u +P F
c
) dV0 = fe0( w) w(x)
_
V0
L0w : (P

L
c
0u +P F
c
) dV0 =
fe0( w)
_
V0
L0w : (P

F
c
) dV0 = r

w(x)
Material model iterative stress change
The right-hand side of the iterative equation represents the residual load. To calculate r
and the term with P

in the left-hand integral, the stress P

(t) must be determined from the


constitutive equation. From the material model also a relation between P and L0u must be
derived. The iterative change of the 2nd-Piola-Kirchho stress P, must be expressed in the
iterative displacement u and substituted in the iterative weighted residual integral.
P =
4
M : L0u
_
V0
L0w :
_
P

L
c
0u + (
4
M : L0u) F
c
_
dV0 =
fe0( w)
_
V0
L0w : (P

F
c
) dV0 = r

w(x)
Matrix/column notation
We will now express the vectors and tensors in their components w.r.t. a basis of a coordinate
system. A matrix-column notation is used, which is explained elsewhere. The asterisk ( )
indicating an approximate value is omitted.
_
V0
L0w : (P L
c
0u +P F
c
) dV0 = fe0( w)
_
V0
L0w : (P F
c
) dV0
_
V0
[L0w : (P L
c
0u) +L0w : (F P
c
)
c
] dV0 = fe0( w)
_
V0
L0w : (F P
c
)
c
dV0
matrix/column notation
_
V0
__
L

0w
_T
t
P
_
L

0u
_
t
+
_
L

0w
_T
t
F
cr
P

_
dV0 = fe0(w

)
_
V0
_
L

0w
_T
t
F
cr
P

dV0
= fe0(w

) fi0(w

)
218
The material stiness, derived from the material model, must also be written in matrix/column
notation.
P

= M
0c
_
L

0u
_
t

_
V0
_
L

0w
_T
t
_
P +F
cr
M
0c
_ _
L

0u
_
t
dV0 =fe0(w

) fi0(w

)
11.5.3 Updated Lagrange formulation
In the Updated Lagrange formulation the reference conguration is chosen to be the start of
the current increment at tn.
_
V
( w)
c
: dV = fe( w) w(x)
transformation to begin increment conguration tn
= F
c
n n ( w)
c
= ( n w)
c
F
1
n
dV = det(Fn)dVn
weighted residual integral
_
Vn
( n w)
c
F
1
n : det(Fn) dVn = fen( w) w(x)
_
Vn
( n w)
c
: (F
1
n ) det(Fn) dVn = fen( w) w(x)
Iterative solution process
To describe the essential steps of the iteration procedure, it is assumed that an approximate
state C
c is determined with values for F

n, and the other variables. The unknown current


values are written as Fn = (I +L

u) F

n and = +, where (.) indicates the dierence


between C
c and Cc, and L

u = ( u)c, with u = x

the iterative displacement.


_
Vn
( n w)
c
: (F
1
n ) det(Fn) dVn = fen( w) w(x)
Fn = ( nx)
c
= { n(x

+x)}
c
= ( nx

)
c
+ ( nx)
c
= F

n +Fn = F

n + (

x)
c
( nx

)
c
= F

n +L

u F

n
= (I +L

u) F

n
=

+
_
_
_

219
_
Vn
( n w)
c
:
_
(F

n)
1
(I +L

u)
1
(

+) det{(I +L

u) F

n}] dVn
= fen( w) w(x)
Assuming that the iterative displacement and its gradient are very small, the weighted residual
integral can be linearized with respect to u. In analogy with L

u, L

w = ( w)c is introduced.
(I +L

u)
1
I L

u
det{(I +L

u) F

n} = det(I +L

u) det(F

n)
tr(I +L

u) det(F

n) = (1 +I : L

u) det(F

n)
weighted residual integral
_
Vn
( n w)
c
:
_
(F

n)
1
(I L

u) (

+)(1 +I : L

u) det(F

n)

dVn
= fen( w) w(x)
further linearisation
_
V
[L

w :

I : L
c
u +L

w : L

w : (
c
L
c
u )
c
] dV

=
f

e ( w)
_
V
L

w :

dV

= r

w(x)
Material model
The right-hand side of the iterative equation represents the residual load. To calculate r and
two terms in the left-hand integral, the stress (t) must be determined from the constitutive
equation.
The iterative change of the stress , must be expressed in the iterative displacement u
and substituted in the iterative weighted residual integral.
=
4
M : L

u
_
V
_
L

w :

I : L
c
u +L

w :
4
M : L

u L

w : (
c
L
c
u )
c

dV

=
f

e ( w)
_
V
L

w :

dV

w(x)
Matrix/column notation
We will now express the vectors and tensors in their components w.r.t. a basis of a coordinate
system. A matrix-column notation is used, which is explained elsewhere. The asterisk ( )
indicating an approximate value is omitted.
220
_
V
[Lw : I : L
c
u +Lw : Lw : (
c
L
c
u)
c
] dV = fe( w)
_
V
Lw : dV
matrix/column notation
_
V
__
L

w
_T
t

T
_
L

u
_
t
+
_
L

w
_T
t

_
L

w
_T
t

tr
_
L

u
_
t
_
dV = fe(w

)
_
V
_
L

w
_T
t

dV
= fe(w

) fi(w

)
The material stiness, derived from the material model, can also be written in matrix/column
notation.

= M
_
L

u
_
t

_
V
_
L

w
_T
t
_

tr
+M
_ _
L

u
_
t
dV = fe(w

) fi(w

)
_
V
_
L

w
_T
t
_
+M
_
L

u
_
t
dV =fe(w

) fi(w

)
11.6 Finite element method
Discretisation
The integral over the body volume V is written as a sum of integrals over smaller volumes,
which collectively constitute the whole volume. Such a small volume V e is called an element.
Subdividing the volume implies that also the surface with area A is subdivided in element
surfaces (faces) with area Ae. Of course for an updated Langragian formulation, we have to
consider the initial volume and surface area.
Fig. 11.3 : Finite element discretisation

e
_
V e
_
L

w
_T
t
_
W
_
L

u
_
t
dV
e
=

e
f
e
e (w

e
f
e
i (w

) w

221
Isoparametric elements
Each point of a three-dimensional element can be identied with three local coordinates
{1, 2, 3}. In two dimensions we need two and in one dimension only one local coordinate.
The real geometry of the element can be considered to be the result of a deformation
from the original cubic, square or line element with (side) length 2. The deformation can
be described with a deformation matrix, which is called the Jacobian matrix J. The de-
terminant of this matrix relates two innitesimal volumes, areas or lengths of both element
representations.
3
1
2
2
1
1
Fig. 11.4 : Isoparametric elements
isoparametric (local) coordinates (1, 2, 3) ; 1 i 1 i = 1, 2, 3
Jacobian matrix J =
_

T
_T
; dV
e
= det(J) d1d2d3
Interpolation
The value of the unknown quantity here the displacement vector u or the iterative displace-
ment vector u in an arbitrary point of the element, can be interpolated between the values
of that quantity in certain xed points of the element : the element nodes. Interpolation
functions are a function of the isoparametric coordinates.
The components of the vector ()u are stored in a column ()u

. The nodal (iterative)


displacement components are stored in the column ()u

e. The position x of a point within the


element is interpolated between the nodal point positions, the components of which are stored
in the column x

e. Generally, the interpolations for position and displacement are chosen to


be the same.
Besides ()u and x, the weighting function w also needs to be interpolated between
nodal values. When this interpolation is the same as that for the displacement, the so-called
Galerkin procedure is followed, which is generally the case for simple elements, considered
here.
We consider the vector function a to be interpolated, where nep is the number of element
nodes. Components ai of a w.r.t. a global vector base, can then also be interpolated.
222
a =
1
a
1
+
2
a
2
+ +
nep
a
nep
=

T
a

ai =
1
a
1
i +
2
a
2
i + +
nep
a
nep
i =
nep
=1

i =

T
a

e
i a

= a

e
The gradient of the vector function a also has to be elaborated. The gradient is referred
to as the second-order tensor L
c
, which can be written in components w.r.t. a vector basis.
The components are stored in a column L

. This column can be written as the product of


the so-called B-matrix, which contains the derivatives of the interpolation functions, and the
column with nodal components of a.
L
c
= a L

t
= Ba

e
Weighted residual integral
Interpolations for both the (iterative) displacement and the weighting function and their
respective derivatives are substituted in the weighted residual integrals of each element.
w

eT
_
_
_
V e
B
T
W BdV
e
_
_u

e
= w

eT
f

e
e
w

eT
f

e
i
Integration
Calculating the element contributions implies the evaluation of an integral over the element
volume V e and the element surface Ae. This integration is done numerically, using a xed set
of nip Gauss-points, which have s specic location in the element. The value of the integrand
is calculated in each Gauss-point and multiplied with a Gauss-point-specic weighting factor
cip and added.
_
V e
g(x1, x2, x3) dV
e
=
1 _
1=1
1 _
2=1
1 _
3=1
f(1, 2, 3) d1d2d3 =
nip
ip=1
c
ip
f(
ip
1 ,
ip
2 ,
ip
3 )
Resulting element quantities are the element stiness matrix K
e
, the element external force
column f

e
e
and the element internal force column f

e
i
.
w

eT
_

_
1=1 _
1=1
2=1 _
2=1
3=1 _
3=1
B
T
_
W

B det(J) d1d2d3
_

_ u

e
= w

eT
f

e
e
w

eT
f

e
i
w

eT
K
e
u

e
= w

eT
f

e
e
w

eT
f

e
i
223
Assembling
The weighted residual contribution of all elements have to be collected into the total weighted
residual integral. This means that all elements are connected or assembled. This assembling is
an administrative procedure. All the element matrices and columns are placed at appropriate
locations into the structural or global stiness matrix K and the load column f

e
.
Because the resulting equation has to be satised for all w

, the nodal displacements u

have to satisfy a set of equations.

e
w

eT
K
e
u

e
=

e
w

eT
f

e
e
w

eT
f

e
i

w

T
K u

= w

T
f

e
w

T
f

i
= w

T
r

Ku

= f

e
f

i
= r

Boundary conditions
The initial governing equations were dierential equations, which obviously need boundary
conditions to arrive at a unique solution. The boundary conditions are prescribed displace-
ments or forces in certain material points. After nite element discretisation, displacements
and forces can be applied in nodal points.
The set of nodal equations Ku

= f

e
cannot be solved yet, because the structural stiness
matrix K is singular and cannot be inverted. First some essential boundary conditions must
be applied, which prevent the rigid body motion of the material and renders the equations
solvable.
11.6.1 FE program plaxL
The Matlab program plaxL is used to model and analyze linear, planar and axi-symmetric
problems. So deformations are small and the material behavior is linear elastic. This means
that the weighted residual formulation from section 11.5.1 is used as a starting point. The
program is described in detail in appendix E. An example is shown below.
Thick-walled pressurized cylinder: plane stress
A thick-walled open cylinder is subjected to an internal pressure. Parameter values are listed
in the table below. The gures show the element model, both undeformed and deformed.,
and the plot of the radial displacement over the radius.
First a quarter of the cylinder is modeled and analyzed in plane stress.
a = inner radius b = outer radius h = height
E = Youngs modulus = Poissons ratio
pi = inner pressure pe = external pressure
a =0.25 b =0.5 h=0.5 E =250 =0.33 pi =100 pe =0
224
plaxLcylpsdef
0.25 0.3 0.35 0.4 0.45 0.5
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
x 10
4
r [m]
u r [m
]
plaxLcylpsrur
Fig. 11.5 : Deformation (1000) of the cylinder and radial displacement.
The thick-walled open cylinder is now analyzed as an axi-symmetric model.
0.25 0.3 0.35 0.4 0.45 0.5
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
x 10
4
r [m]
u r [m
]
plaxLcylaxrur
0 0.1 0.2 0.3 0.4 0.5
1
0.5
0
0.5
1
1.5
2
x 10
8 plaxLcylaxsig
r [m]

[P
a]


rr
tt
zz
Fig. 11.6 : Displacement and stresses in a pressurized cylinder.
11.6.2 FE program plax
The Matlab program plax is used to model and analyze planar and axisymmetrc problems.
Large deformations and rotations are allowed. Various nonlinear and time-dependent material
models are implemented, of which examples will be shown in later chapters.
Rigid rotation
When a material body is subjected to a rigid rotation, no stresses must be generated in the
material. The material model must be such that this requirement is satised.
In this example we rotate one element over 720o by prescribing all nodal point displace-
ments. A non-suitable material law, a linear relation between the Cauchy stress tensor
and the innitesimal strain tensor , will result in high reaction forces in the nodes (gure 1
225
below). A correctly formulated material law, such as a linear relation between the Cauvhy
stress tensor and the tensor
1
2(F F
c
I), will result in zero reaction forces (gure 2 below).
0 5 10 15
1.5
1
0.5
0
0.5
1
1.5

crd
1 node 3
plaxrigrot1ax
0 5 10 15
0
0.5
1
1.5
2
2.5
x 10
4

force node 3
plaxrigrot1af
Fig. 11.7 : Rigid rotation for non-objective elastic material model.
0 5 10 15
1.5
1
0.5
0
0.5
1
1.5

crd
1 node 3
plaxrigrot3ax
0 5 10 15
0
1
2
3
4
5
6
x 10
12

fo
rce
n
o
d
e
3
plaxrigrot3af
Fig. 11.8 : Rigid rotation for objective elastic material model.
226
Chapter 12
Three-dimensional material models
In this chapter we consider three-dimensional material models for various material behavior.
Implementation in nite element software is the main focus, which means that the calculation
of the stress and the stiness during the iterative solution procedure is paramount.
In the following sections we consider models for elastic, elastomeric, elastoplastic, linear
viscoelastic, viscoplastic and nonlinear viscoelastic materials behavior. Implementation in
FEM modules is explained and simple examples are calculated.
227
228
12.1 Elastic material behavior
For elastic materials the 2nd-Piola-Kirchho stress tensor P is related to the Green-Lagrange
strain tensor E. The Cauchy stress tensor can be written as a function of the right Cauchy-
Green strain tensor B. To assure the stress to be zero when there is no deformation, it is
more suitable to relate the Cauchy stress to the Finger tensor A.
P = G(E) with E = 1
2(C I) = 1
2(F
c
F I)
= J
1
F P F
c
= J
1
F G(E) F
c
with J = det(F)
= K(A) with A =
1
2(B I) =
1
2 (F F
c
I)
12.1.1 Isotropic elastic material
For an isotropic material a principal strain deformation of a material cube can only result in
normal stresses on its faces. Using its denition it can be shown that the principal directions
of the 2nd-Piola-Kirchho stress tensor coincide with the principal strain directions. It is
easily seen that the principal directions of the Cauchy stress tensor coincide with those of the
Finger tensor.
F 3
n1
n02
n03
n01
n2
n3
1
2
Fig. 12.1 : Deformation in principal directions
U = 1n01n01 +2n02n02 +3n03n03
R = n1n01 +n2n02 +n3n03
F = 1n1n01 +2n2n02 +3n3n03
= 1n1n1 +2n2n2 +3n3n3
P = JF
1
F
c
= J
_
1
2
1 n01n01 +2
2
2 n02n02 +3
2
3 n03n03
_
= s1n01n01 +s2n01n01 +s3n01n01
229
P E model
For a general isotropic material the principal directions of the 2nd-Piola-Kirchho stress
tensor P and the Green-Lagrange strain tensor E, coincide. As a result, P can be written
as a polynomial function of E.
Applying Cayley-Hamiltons theorem, a second-order polynomial relation remains. The
coecients i in this relation are not constant. For the isotropic material they are a function
of the invariants of E and have to be determined experimentally.
2nd-Piola-Kirchho stress tensor and Green-Lagrange strain tensor
P = s1n01n01 +s2n02n02 +s3n03n03
E = 1n01n01 +2n02n02 +3n03n03
isotropic elastic material model
P =

sin0in0i = G(E) =

G(i)n0in0i = a0I +a1E +a2E


2
+a3E
3
+...
Cayley-Hamiltons theorem E
3
= J1(E)E
2
J2(E)E +J3(E)I
constitutive equation
P = 0I +1E +2E
2
with i = i {J1(E), J2(E), J3(E)}
A model
For an isotropic material the principal directions of and A coincide, which implies that
can be written as a polynomial function of A. Applying Cayley-Hamiltons theorem results
in a second-order polynomial with coecients depending on the invariants of A.
Cauchy stress tensor and Finger tensor
= 1n1n1 +2n2n2 +3n3n3
A = A1n1n1 +A2n2n2 +A3n3n3
isotropic elastic material model
=

inini = K(A) =

K(Ai)nini = b0I +b1A+b2A


2
+b3A
3
+...
Cayley-Hamiltons theorem A
3
= J1(A)A
2
J2(A)A+J3(A)I
constitutive equation
= 0I +1A+2A
2
with i = i {J1(A), J2(A), J3(A)}
230
Isotropic linear elastic material
The above isotropic elastic material models are nonlinear. The polynomial functions have a
quadratic tensor term and, moreover, the coecients are functions of the three invariants of
the strain tensor. The rst invariant is a linear, the second a quadratic and the third a cubic
function of the tensor.
Simplication towards purely linear models is possible and allowed if it suits the exper-
imental observations.
P E model
When experiments show that the relation between P and E is linear, conclusions can be
drawn concerning the coecients i. The coecient of the quadratic term, 2, must be
zero. The coecient of the linear term, 1, must be a constant. The coecient of the unit
tensor may be a linear, isotropic function of the tensor E, which means it can be written as a
constant times the trace of E. No constant tensor is contained in the linear models, because
stress has to be zero at zero strain.
Substituting an (experimentally motivated) linear relation between P and E in the
denition relation of , results in a nonlinear relation between and B and vice versa.
P = 0I +1E +2E
2
with i = i {J1(E), J2(E), J3(E)}
linear model P = c0tr(E)I +c1E = (c0 +
1
3c1)tr(E)I +c1E
d
A model
The same holds for the relation between and A.
= 0I +1A+2A
2
with i = i {J1(A), J2(A), J3(A)}
linear model = c0tr(A)I +c1A = (c0 +
1
3c1)tr(A)I +c1A
d
12.1.2 Hyper-elastic material
When an explicit stored energy function is available for an elastic material, it is called hyper-
elastic. The stress tensor can then be calculated as the derivative of the energy function with
respect to the associated strain tensor.
When the stress-strain relation is not derived from a stored energy function, the elastic
model is called hypo-elastic. For large strains such a model predicts the elastic behavior not
correctly. In a closed cycle deformation loop residual stresses and elastic energy will remain.
A hyper-elastic material model describes large elastic strains correctly.
The elastic energy must of course always become zero when there is no deformation. The
function can be formulated with various strain tensors. The stress tensor can be derived by
dierentiation of the stored energy function with respect to the strain tensor.
The second Piola-Kirchho stress tensor P is derived from an energy function (E),
depending on the Green-Lagrange strain tensor. Instead of , a function W(C) is usually
231
specied. Although the stress tensor can now still be derived by dierentiation in this case
W w.r.t. C an additional requirement must be formulated or incorporated, namely that
stress must be zero (P = O) when there is no deformation (C = I).
elastic energy function = (E) W = W(C)
stress tensors P =
d(E)
dE
=
dW(C)
dC
:
dC
dE
= 2
dW(C)
dC
= G(E)
=
1
J
F P F
c
=
2
J
F
dW(C)
dC
F
c
Isotropic hyper-elastic material
For isotropic material the elastic energy function can be written as a function of the invariants
of E or C.
Isotropic hyper-elastic model : P E
For isotropic material the energy function (E) is only depending on the invariants of the
strain tensor. Again we see that P can be written as a second-order polynomial of E. The
coecients i are now functions of the derivatives of w.r.t. the invariants of E.
= (E) = {J1(E), J2(E), J3(E)} P =

J1
dJ1
dE
+

J2
dJ2
dE
+

J3
dJ3
dE
derivatives of invariants
dJ1
dE
= I ;
dJ2
dE
= J1I E ;
dJ3
dE
= J2I J1E +E
2

stress tensor
P =
_

J1
+

J2
J1 +

J3
J2
_
I +
_


J2


J3
J1
_
E +

J3
E
2
= 0I +1E +2E
2
Isotropic hyper-elastic model : P C
For isotropic material the energy function (E) is only depending on the invariants of the
strain tensor. Again we see that P can be written as a second-order polynomial of E. The
coecients i are now functions of the derivatives of w.r.t. the invariants of E.
W = W(C) = W{J1(C), J2(C), J3(C)} P = 2
_
W
J1
dJ1
dC
+
W
J2
dJ2
dC
+
W
J3
dJ3
dC
_
derivatives of invariants
dJ1
dC
= I ;
dJ2
dC
= J1I C ;
dJ3
dC
= J2I J1C +C
2

232
stress tensor
P = 2
_
W
J1
+
W
J2
J1 +
W
J3
J2
_
I + 2
_

W
J2

W
J3
J1
_
C + 2
W
J3
C
2
Isotropic incompressible hyper-elastic material
For a hyper-elastic material model the stress-strain relation is derived from an energy func-
tion W(C). For an isotropic material W is a function of the invariants of C. Due to the
incompressibility, the energy function cannot depend on the third invariant, which has always
the value 1.
From a given function W(C), the 2nd-Piola-Kirchho stress tensor can be determined
by dierentiation. Subsequently the Cauchy stress tensor can be calculated from P.
J = det(F) = 1 det(C) = J3(C) = 1 W(C) = W{J1(C), J2(C)}
P = 2
_
W
J1
dJ1
dC
+
W
J2
dJ2
dC
_
= 2
__
W
J1
+
W
J2
J1
_
I
W
J2
C
_
=
1
J
F P F
c
=
2
J
__
W
J1
+
W
J2
J1
_
B
W
J2
B
2
_
12.1.3 Incompressible elastic material
Elastic material behavior can be described by a relation between the Cauchy stress tensor
and the left Cauchy-Green strain tensor B = F F
c
. When the material is incompressible
and isotropic, the deformation will not be aected by the addition of a hydrostatic stress pI.
When the deformation is known, the stress cannot be determined, because the hydro-
static stress remains arbitrary. Only the so-called extra stress tensor depends solely on B
and can be calculated.
To determine the unknown hydrostatic stress pI the incompressibility condition must be
used.
p
p
p
Fig. 12.2 : Hydrostatic stress state
233
= pI +F P F
c
= pI + 2
__
W
J1
+
W
J2
J1
_
B
W
J2
B
2
_
= pI +1B +2B
2
= pI +H(B)
= pI +
Energy function
Energy functions W(C) are generally written as polynomials of (J1 3) and (J2 3) such
that W = 0 when there is no deformation (C = I J1 = J2 = 3).
The invariants of C can be expressed in the principal stretch ratios 1, 2 and 3.
The polynomial energy function W(C) can then also be written as a polynomial function of
these stretch ratios. This way of denoting these functions is often referred to as the Rivlin
formulation.
W(C) =
m
i=0
n
j=0
Cij{J1(C) 3}
i
{J2(C) 3}
j
with C00 = 0
J1 = tr(C) =
2
1 +
2
2 +
2
3
J2 =
1
2
_
tr
2
(C) tr(C
2
)
_
=
1
2
__

2
1 +
2
2 +
2
3
_2

4
1 +
4
2 +
4
3
__
=
2
1
2
2 +
2
2
2
3 +
2
3
2
1
Neo-Hookean model
The Neo-Hookean energy function has only one material parameter : C10. The model de-
scribes the mechanical behavior of natural rubbers rather well.
W = C10(J1 3) 1 = 2C10 ; 2 = 0
= pI + 2C10B = pI + 2C10(B I) = pI + 4C10A
Mooney-Rivlin model
The mechanical behavior of industrial rubbers cannot be captured well with the one-parameter
Neo-Hookean model. Instead the Mooney-Rivlin model is often used, which has two param-
eters.
234
W = C10(J1 3) +C01(J2 3) 1 = 2{C10 +C01tr(B)} ; 2 = 2C01
= pI + 2{C10 +C01tr(B)}B 2C01B
2
= pI + 2C10B + 2C01{tr(B)B B
2
}
= p

I + 2C10(B I) + 2C01
_
tr(B)B B
2
2I
_
A =
1
2(B I) B = 2A+I
A
2
=
1
4 B
2

1
2B +
1
4I B
2
= 4A
2
+ 2B I
= p

I + 4C10A+ 2C01
_
4tr(A)A+ 2tr(A)I 4A
2
+ 2A
_
= pI + 4C10A+ 4C01
_
(1 + 2tr(A))A2A
2
_
12.1.4 Incremental analysis
In nonlinear analysis, the load is applied in a number of steps, the load increments.
tn
tn+1
F(tn+1)
F(tn)
Fn(tn+1) t0
Fig. 12.3 : Incremental deformation
The end-increment state, i.e. deformation and stresses, must be determined such that equilib-
rium equations, material relations and boundary conditions are satised. Due to the nonlinear
character of deformation and material behavior, the equations must be solved iteratively. In
each iteration step both the stress and the material stiness must be updated.
12.1.5 Linear P-E model
Stress update
Elastic material behavior may be described by a linear relation between the second Piola-
Kirchho stress tensor P and the Green-Lagrange strain tensor E. This relation can be
derived from an elastic energy function and that is why this model is called hyper-elastic. For
isotropic materials we can derive
P = Ktr(E)I + 2G
_
E 1
3tr(E)I
_
= c0 tr(E)I +c1 E
where K is the bulk modulus and G is the shear modulus and c0 = K
2
3G and c1 = 2G.
For a given deformation the stress in the material can be calculated directly for an elastic
material. The Cauchy stress tensor can be calculated from P.
235
P = c0tr(E)I +c1E with E = 1
2(C I)
=
1
2c0C : II +
1
2c1C
1
2(3c0 +c1)I with C = F
c
F
matrix/column notation
P

=
1
2 c0C

T
I

t
I

+
1
2c1C

1
2 (3c0 +c1)I

with C

= F
t
F

Cauchy stress tensor


= J
1
F P F
c
= J
1
F (P F
c
) = J
1
F (F P
c
)
c
matrix/column notation

= J
1
F
_
F
r
P

t
_
= J
1
F F
r
P

t
Stiness
In the Newton-Raphson iterative solution procedure, the variation of the stress tensor must
be expressed in the iterative displacement of material points.
Starting from the P E elastic model, the relation between P and F is calculated.
The variation of the deformation tensor F can be expressed in the gradient of the iterative
displacement vector u = x :
F = L F = (F
c
L
c
)
c
with L
c
= u
The tensorial expression is transferred to matrix/column notation.
variation of 2nd-PK stress tensor
P =
1
2c0C : II +
1
2c1C
C = F
c
F C = F
c
F +F
c
F
= 1
2c0 (F
c
F +F
c
F) : II + 1
2c1 (F
c
F +F
c
F)
= c0(F
c
F) : II +
1
2 c1 (F
c
F +F
c
F)
= c0I(F
c
: F) +
1
2c1 {(F
c
F)
c
+ (F
c
F)}
matrix/column notation
P

= c0I

T
t
F

t
+
1
2c1
_
(F
t
F

)r + (F
t
F

)
_
= c0I

T
tc
F

+
1
2c1
_
F
tr
F

+F
t
F

_
=
_
c0I

T
tc
+
1
2c1(F
tr
+F
t
)
_
F

= M
0
F

= M
0
_
L

0
_
t
with F

= F
tr
L

t
236
P

=
_
c0I

T
tc
F
tr
+
1
2 c1
_
F
tr
F
tr
+F
t
F
tr
__
L

t
= M
1
L

t
Combining the variations P and leads to a relation between and L.
= J
1
F P F
c

= J
1
[JF P F
c
+F P F
c
+F P F
c
+F P F
c
]
J = Jtr(D) J = Jtr(L) = JL : I ; F = L F
= J
1
[(L : I)F P F
c
+ (L F) P F
c
+
F P F
c
+F P (F
c
L
c
)]
= (L : I) +L + L
c
+J
1
F P F
c
= (I : L) + (
c
L
c
)
c
+ L
c
+J
1
F (F P
c
)
c
matrix/column notation

T
L

t
+
tr
L

t
+L

t
+J
1
F F
r
P

t
=

T
L

t
+
tr
L

t
+L

t
+J
1
F F
rc
P

=
_

T
+
tr
+
_
L

t
+J
1
F F
rc
M
1
L

=
_

T
+
tr
+ +J
1
F F
rc
M
1
_
L

t
= M L

t
Uni-axial
For a tensile test only the axial component of P is non-zero and can be expressed in the
axial and cross-sectional stretch ratios and . Because stresses perpendicular to the axial
direction are zero, can be eliminated and the axial stress can be expressed in the axial
stretch .
The parameters c0 and c1 can be expressed in the more commonly used Youngs modulus
E and Poissons ratio .
P = E 1
2(
2
1)

2
1 = (
2
1)
_
c0 =
E
(1 +)(1 2)
; c1 =
E
1 +
P = 1
2c0(
2
+ 2
2
3) + 1
2c1(
2
1)
0 =
1
2c0(
2
+ 2
2
3) +
1
2c1(
2
1) =
1
2c0
2
+
1
2 (2c0 +c1)
2

1
2(3c0 +c1)

2
=
c02 + 3c0 +c1
2c0 +c1
=
c0
2c0 +c1

2
+
3c0 +c1
2c0 +c1
=
2
+
237
substitution in rst equation stress and axial force
P =
1
2 c0
2
+c0
_
c02 + 3c0 +c1
2c0 +c1
_

3
2 c0 +
1
2 c1(
2
1) =
2
+
K = A =
_

2
P
_

2
A0 = PA0 = (
2
+)A0
12.1.6 Linear -A model
Stress update
Elastic material behavior may be described by a linear relation between the Cauchy stress
tensor and the Finger tensor A =
1
2(B I) with B = F F
c
. The above relation cannot
be derived from an elastic energy function and is thus referred to as hypo-elastic.
Cauchy stress tensor
= c0tr(A)I +c1A with A =
1
2(B I)
=
1
2 c0B : II +
1
2c1B
1
2(3c0 +c1)I with B = F F
c
matrix/column notation

=
1
2c0B

T
I

t
I

+
1
2c1B

1
2 (3c0 +c1)I

with B

= FF

t
Stiness
The variation of the Cauchy stress tensor can be related to F, and consequently to the
gradient of the iterative displacement vector u.
variation of Cauchy stress tensor
=
1
2c0B : II +
1
2c1B
=
1
2c0 {(F F
c
)
c
+F F
c
} : II +
1
2c1 {(F F
c
)
c
+F F
c
}
= c0(F F
c
) : II +
1
2c1 {(F F
c
)
c
+F F
c
}
= c0IF : F
c
+
1
2c1 {(F F
c
)
c
+F F
c
}
with F = L F = (F
c
L
c
)
c
and L
c
= u
matrix/column notation

= c0I

T
F

+
1
2c1
_
F
r
F

t
+FF

t
_
= c0I

T
F

+
1
2c1
_
F
rc
F

+F
c
F

_
=
_
c0I

T
+
1
2c1
_
F
rc
+F
c
__
F

with F

=
_
F
t
L

t
_
r
= F
tr
L

t
238

=
_
c0I

T
F
tr
+
1
2c1
_
F
rc
F
tr
+F
c
F
tr
__
L

t
= M L

t
Plane strain, plane stress and axisymmetric
To describe planar deformation in the Cartesian coordinate system, components are now
renamed as follows :
1x ; 2y ; 3z
For plane strain it is assumed that no deformation occurs in the z-direction perpendicular
to the xy-plane. In this plane strain situation Fzz = 1, which implies that Lzz = 0. The
stress component zz can be expressed in the in-plane strain components Axx and Ayy and
calculated a posteriori.
In a plane stress situation the material is unrestrained in the z-direction and therefore
zz = 0. With this equation the zz-strain component can be expressed in the in-plane
components and thus eliminated from the constitutive equations.
In the axisymmetric conguration the components are renamed as follows :
1r ; 2z ; 3t()
plane strain Fzz = 1 Azz = 0 ; zz = c0(Axx +Ayy)
plane stress
zz = c0(Axx +Ayy +Azz) +c1Azz = 0
Azz =
c0
c0 +c1
(Axx +Ayy)
_
=
c0c1
c0 +c1
tr(A)I +c1A
F
2
zz = 2Azz + 1 =
c0
c0 +c1
(F
2
xx +F
2
xy +F
2
yx +F
2
yy) +
2c0
c0 +c1
+ 1
Uni-axial
For a tensile test only the axial component of is non-zero and can be expressed in the
axial and cross-sectional stretch ratios and . Because stresses perpendicular to the axial
direction are zero, can be eliminated and the axial stress can be expressed in the axial
stretch .
The parameters c0 and c1 can be expressed in the more commonly used Youngs modulus
E and Poissons ratio .
= E
1
2(
2
1)

2
1 = (
2
1)
_
c0 =
E
(1 +)(1 2)
; c1 =
E
1 +
239
stress components
=
1
2c0(
2
+ 2
2
3) +
1
2c1(
2
1)
0 = 1
2c0(
2
+ 2
2
3) + 1
2c1(
2
1) = 1
2c0
2
+ 1
2 (2c0 +c1)
2
1
2(3c0 +c1)
contraction follows from second equation

2
=
c02 + 3c0 +c1
2c0 +c1
=
c0
2c0 +c1

2
+
3c0 +c1
2c0 +c1
=
2
+
substitution in rst equation stress and axial force
=
_
1
2 c0
c2
0
2c0 +c1
+
1
2 c1
_

2
+
_
3
2c2
0 +
1
2 c0c1
c0 +
1
2 c1

3
2 c0
1
2 c1
_
=
2
+
K = A =
2
A0 = (
2
+)(
2
+)A0
12.1.7 Examples
A square plate is subjected to a tensile and a shear deformation. The two linear elastic
models, described before, are used to model the elastic behavior. Both plane stress and plane
strain states are considered.
Tensile test
A square plate or cylindrical bar is loaded uniaxially using dierent elastic material models.
Dimensions are listed in the table. For plane stress and axisymmetry, the loading is equivalent
to a tensile test.
Cartesian
initial width w0 100 mm
initial height h0 100 mm
initial thickness d0 0.1 mm
cylindrical
initial radius r0
_
(10/) mm
initial height h0 100 mm
The axial elongation is prescribed and the resulting axial force is calculated for various elastic
material models. Material parameter values are C = 100000 MPa and = 0.3.
240
0.5 1 1.5 2 2.5 3
8
6
4
2
0
2
4
6
x 10
5

F
[N
]
plax11ctpsglfa
0.5 1 1.5 2 2.5 3
0
2
4
6
8
10
12
14

A
[m
m
2 ]
plax11ctpsglaa
Fig. 12.4 : Tensile force and cross-sectional area versus elongation; plane stress;
model
0.5 1 1.5 2
5
0
5
x 10
5

F
[N
]
plax31ctpsglfa
0.5 1 1.5 2
0
2
4
6
8
10
12
14

A
[m
m
2 ]
plax31ctpsglaa
Fig. 12.5 : Tensile force and cross-sectional area versus elongation; plane stress; A
model
0.5 1 1.5 2 2.5
1
0
1
2
3
4
5
x 10
6

F
[N
]
plax41ctpsglfa
0.5 1 1.5 2 2.5
0
2
4
6
8
10
12
14

A
[m
m
2 ]
plax41ctpsglaa
Fig. 12.6 : Tensile force and cross-sectional area versus elongation; plane stress; P E
model.
The latter model is also used in a plane strain tensile test. Both Updated Lagrange and Total
Lagrange formulation are used. The results are the same, which should be the case.
241
0.5 1 1.5 2
0.5
0
0.5
1
1.5
2
2.5
3
3.5
x 10
6

F
[N
]
plax41ctpeglfa
0.5 1 1.5 2
2
0
2
4
6
8
10
12

A
[m
m
2 ]
plax41ctpeglaa
Fig. 12.7 : Tensile force and cross-sectional area versus elongation; plane strain;
P E model; Updated Lagrange formulation
0.5 1 1.5 2
0.5
0
0.5
1
1.5
2
2.5
3
3.5
x 10
6

F
[N
]
plax42ctpeglfa
0.5 1 1.5 2
2
0
2
4
6
8
10
12

A
[m
m
2 ]
plax42ctpeglaa
Fig. 12.8 : Tensile force and cross-sectional area versus elongation; plane stress; P E
model; Total Lagrange formulation
Shear test
The simple shear test is analyzed with one element, where the horizontal displacement/force
in the upper nodes is prescribed. Dimensions are listed in the table.
initial width w0 100 mm
initial height h0 100 mm
initial thickness d0 0.1 mm
Subsequently the material model A and P E are used in the analysis.
242
0.5 0 0.5 1 1.5 2
2
1
0
1
2
3
x 10
6

F
x [N
]
plax31shggfx
0.5 0 0.5 1 1.5 2
0
5
10
15
x 10
5

F
y [N
]
plax31shggfy
Fig. 12.9 : Shear and normal force versus shear strain; plane stress; A model
0.5 0 0.5 1 1.5
0.5
0
0.5
1
1.5
2
2.5
x 10
7

F
x [N
]
plax41shggfx
0.5 0 0.5 1 1.5
0
2
4
6
8
10
12
14
x 10
6

F
y [N
]
plax41shggfy
Fig. 12.10 : Shear and normal force versus shear strain; plane stress; P E model
243
12.2 Elastoplastic material behavior
The one-dimensional mechanical representation of an elastoplastic model consists of a spring
in series with a parallel arrangement of a spring and friction slider. The series-spring repre-
sents the purely elastic part of the deformation, when stress is below the yield stress. The
elastoplastic response becomes manifest when the stress exceeds the yield stress y.
After yielding the total strain is the sum of the elastic strain e and the plastic strain
p. The stress is not inuenced by the strain rate.
H
y
e p

E
Fig. 12.11 : Discrete model for elastoplastic material behavior
12.2.1 Kinematics
Transformation from the undeformed conguration at time t0 (position vector x0) to the
current conguration at time t (position vector x) is described by the deformation tensor
F = ( 0x)c, where 0 is the gradient operator with respect to the undeformed state.
The right and left Cauchy-Green strain tensors, C and B, are functions of F as is the
Green-Lagrange strain tensor E. The deformation rate is described by the velocity gradient
tensor L = ( v)c, where is the gradient operator with respect to the current state and v
is the velocity of the material volume.
The total deformation F is multiplicatively decomposed into an elastic and a plastic
contribution. For the velocity gradient tensor an additive decomposition into the symmetric
deformation rate tensor D and the skew-symmetric spin tensor is used. Both D and
can be split into an elastic and a plastic part.
To make the decomposition unique it is commonly assumed that the plastic rotation rate
during the current increment is zero, i.e. p = 0. Superimposed material rotations are thus
fully represented in Fe.
244
t0
t
F
Fp
Fe
Fig. 12.12 : Multiplicative decomposition of total deformation
F = ( 0x)
c
= Fe Fp
C = F
c
F ; B = F F
c
; E =
1
2(C I)
L = F F
1
= ( v)
c
= Le +Lp = (De +e) + (Dp +p) = (De +e) +Dp
12.2.2 Constitutive relations
Elastic deformation
The stress is related to the elastic strain with an elastic material model. In elastoplastic
deformation problems, it can often be assumed that elastic strains are small, which allows
the use of a hypo-elastic generalized Hookes law, relating the Cauchy stress tensor to the
logarithmic strain tensor .
The material is assumed to be isotropic in which case the elastic material behavior is
characterized by two material constants : the bulk modulus K and the shear modulus G. The
fourth-order unity tensor is dened as :
4
I = eiejiljkekel and 4I
rc
is its right conjugate.
linear elastic material model =
4
C : e
isotropic elastic material tensor
4
C = c0II +
1
2 c1(
4
I +
4
I
rc
) = KII + 2G
_
4
I
1
3II
_
Yield criterion and hardening
A yield function F is used to evaluate the stress state and to check whether the deformation
is purely elastic (F < 0) or elastoplastic (F = 0). The current stress state, represented by
the equivalent stress , is compared to a yield stress y. Its initial value is y0. This yield
stress changes with plastic deformation and is therefore related to the eective plastic strain
p. The relation between y and p is described by the hardening law. To decide whether
elastic or elastoplastic deformation occurs, the Kuhn-Tucker relations are used.
245
yield criterion F =
2

2
y( p)
eective plastic strain p =
t _
=0
p d ; p =
_
2
3Dp : Dp
hardening law y = y(y0, p) with
y
p
= H( p)
Kuhn-Tucker relations {(F < 0) (F = 0 F < 0)} elastic deformation
{(F = 0) ( F = 0)} elastoplastic deformation
Elastoplastic deformation
During elastoplastic deformation (F = 0) the plastic deformation rate Dp is related to the
stress by the ow rule. For a so-called normality or associative ow rule the direction of Dp
is perpendicular to the yield surface in stress space. The length of Dp is characterized by the
plastic multiplier .
= 0

Dp =
F

F = 0
F = 0
Fig. 12.13 : Associative ow rule
The value of the plastic multiplier can be determined from the requirement that the
stress state must always reside on the yield surface during elastoplastic deformation, so :
F = 0. This relation is referred to as the consistency condition.
associated ow rule Dp =
F

= a
consistency relation F = 2 2y y = 2 2yH p = 0
Invariant tensors
The current stress state must be determined from the elastoplastic constitutive model, which
is necessarily a rate formulation, i.e. a relation between a time derivative of the stress and the
deformation rate. To avoid problems with large rigid rotations, the constitutive equations are
246
formulated in invariant variables. A general invariant stress tensor A is introduced rst. It
can be proved that both A and A are invariant, when rigid body rotation (rotation tensor
Q) transforms A into A

according to : A

= A Q
c
.
A = A A
c
=

A with A

= A Q
c
Q
A = A
_
(A1 A) + (A1 A)
c
+
_
A
c
= A

A A
c
=

A
The elastic material law can now be reformulated, such that it obeys the objectivity require-
ment.

A =
4
C : De
12.2.3 Constitutive model
The material model can be summarized as a set of constitutive relations. In accordance with
A, invariant tensors DA and aA are dened. Also a new fourth-order material tensor 4CA
is introduced according to the requirement :
A
c

4
CA : DA A =
4
C : D A
The set of dierential equations must be integrated over the deformation history to determine
the current stress (t) when the current deformation F(t) is known. It is also used to derive
a relation between the variation of stress and deformation, which is an essential part of the
element stiness matrix.
{(F < 0) (F = 0 F < 0)} D = De p = 0

A =
4
C : D
A = (A A
c
)
_
4
CA : DA
_
(A A
c
)
{(F = 0) ( F = 0)} D = De +Dp

A =
4
C : (D a)
2 2yH p = 0
_
_
_

A = (A A
c
)
_
4
CA :
_
DA aA
__
(A A
c
)
2 2yH p = 0
_
_
_
y = y(y0, p) ; p =
_
2
3a : a
247
Von Mises plasticity
For the Von Mises yield criterion, the yield surface is a circular cylinder in principal stress
space. The normal to the yield surface can be expressed in the deviatoric stress d.
=
_
3
2d : d F =
3
2
d
:
d

2
y( p)
a =
F

=
F
d
:
d

=
_
3
d
:
4
I
_
:

_

1
3tr()I
_
=
_
3
d
:
4
I
_
:
_
4
I
1
3II

= 3
d
:
_
4
I
1
3II
_
= 3
d
F = 3
d
: 2yH p = 0
A = (A A
c
)
_
4
CA :
_
DA aA
__
(A A
c
)
3
d
A :
4
CA : DA
_
3
d
A :
4
CA : aA + 2yH
_
2
3 aA : aA
_
= 0
_
_
_
=
_
3
2d : d ; a = 3
d
; y = y(y0, p) ; p =

6d : d
12.2.4 Incremental analysis
The gure shows the relevant congurations in a large strain plastic deformation process.
Although the time t is used to identify various congurations, it is noted that the material
behavior is considered to be time independent. The variable t is thus a pseudo-time.
Starting from the undeformed conguration at t0 the external load is applied and the
deformation leads to the current conguration t. During a numerical analysis of this defor-
mation process the state of the material is determined at a nite number of discrete mo-
ments ti, i = 0, 1, .., n + 1. The period between two subsequent moments is an increment :
ti = ti+1 ti.
It is assumed that the analysis has brought us to t = tn, the beginning of the last incre-
ment and that all relevant variables are known and satisfying all governing equations (balance
laws, boundary conditions, constitutive relations). The state at the current time t = tn+1,
the end of the current increment has to be determined.
The incremental deformation is described by the deformation tensor Fn. The incre-
mental principle elongation factors and directions, ni and nni (i = 1, 2, 3), respectively, with
respect to the begin increment state, can be determined from Cn = F
c
n Fn. The incremental
stretch tensor Un and logarithmic strain tensor n can be expressed in ni and nni.
248
tn
tn+1
F(tn+1)
F(tn)
Fn(tn+1) t0
Fig. 12.14 : Incremental deformation
F() = Fn() F(tn) Fn() = ( nx)
c
= F() F
1
(tn)
D = 1
2
_
Fn F
1
n +F
c
n F
c
n
_
= 1
2Rn
_
Un U
1
n +U
1
n Un
_
R
c
n
=
1
2
_
Fn F
1
n F
c
n F
c
n
_
= Rn R
c
n +
1
2Rn
_
Un U
1
n U
1
n Un
_
R
c
n
Un =
3
i=1
ninninni ; n =
3
i=1
ln(ni)nninni
Rotation neutralized increment
Determination of the end-increment stress state implies integration of the constitutive equa-
tions over the increment.
Because the rigid rotation during the increment is not uniquely known, rotation neutral-
ized quantities are used. This implies the specication of the invariant tensors by choosing
A = R
c
n resulting in Dienes tensors and Dienes rates.
The complete elastoplastic model can now be formulated in rotation neutralized quanti-
ties D, DD and aD.
stress D = R
c
n Rn D = R
c
n

D Rn
deformation rate DD = R
c
n D Rn = 1
2
_
Un U
1
n +U
1
n Un
_
constitutive model
D =
4
CD :
_
DD aD
_
3
d
D :
4
CD : DD
_
3
d
D :
4
CD : aD + 2yH
_
2
3 aD : aD
_
= 0
_
_
_
It is assumed that there is no rigid body rotation during the increment. All rigid body rotation
will be taken into account at the end-increment time tn+1. The integrated stress tensor is the
so-called rotation neutralized stress tensor D.
When it is also assumed that the incremental principal strain directions are constant
during the increment, the tensors Un and U
1
n are commuting.
249
incremental rotation neutralized
tn < tn+1 : Rn = I ; DD = D ; aD = a ;
4
CD =
4
C
= tn+1 : Rn(tn+1) = F(tn+1) U
1
(tn+1)
constant incremental principal strain directions nni() = nni(tn)
Un() =
3
i=1
ni()nni(tn)nni(tn)
D = Un U
1
n =
3
i=1
_
ni()
ni()
_
nni(tn)nni(tn) = n
With this assumption, the constitutive equations for the rotation neutralized Dienes stress
D can now be used for integration.
D =
4
C :
_
n a
_
3
d
D :
4
C : n
_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_
= 0
_
_
_
12.2.5 Stress update
During the increment t = tn+1 tn the stress evolution equations are integrated using an
implicit Euler integration scheme.
evolution equations (at time tn < tn+1)
D = 4C :
_
n a
_
3
d
D :
4
C : n
_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_
= 0
_
_
_
implicit Euler integration scheme
f() = f(tn+1) =
1
t
{f(tn+1) f(tn)}
The derivative of the incremental logarithmic strain tensor is the end-increment value divided
by the time increment, because n(tn) = O.
incremental logarithmic strain rate
n =
1
t
{n n(tn)} =
1
t
n
250
incremental equations
D = D(tn) +
4
C : (n a)
3
d
D :
4
C : n
_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_
= 0
_
_
_
Elastic stress predictor
The stress integration procedure is always started with the calculation of an elastic stress pre-
dictor. It is assumed that the increment is fully elastic and that the begin-increment elasticity
tensor can be used to calculate the rotation neutralized Cauchy stress tensor. Subsequently
the elastic Cauchy stress tensor is calculated and used to evaluate the yield criterion with two
possible outcomes :
1. the increment is indeed fully elastic,
2. the yield criterion is violated which implies that during the increment further elasto-
plastic deformation has taken place.
elastic trial stress De = (tn) +
4
C(tn) : n e = Rn De R
c
n
yield criterion F =
3
2
d
e :
d
e
2
y(y0, p(tn))
1. F 0 elastic increment
2. F > 0 elastoplastic increment
matrix/column notation
C = KI

T
+ 2G
_
I
1
3 I

T
_
; n

De
=

(tn) +C
c

n
De e = Rn De R
T
n
F =
3
2
_

Dtr
_T _

Dtr
_

2
y( p)
Elastic increment
When the increment is fully elastic the end-increment Cauchy stress equals the calculated
elastic Cauchy stress. As no plastic deformation has occurred during the increment, the
eective plastic strain and the yield stress have not changed.
(tn+1) = e ; = 0
p(tn+1) = p(tn) ; y(tn+1) = y(tn)
251
Elastoplastic increment
If the elastic stress predictor violates the yield criterion, the increment is elastoplastic. The
stress and the elastic multiplier must be determined such that at the end of the increment,
the stress resides on the yield surface.
D = 4C :
_
n a
_
3
d
D :
4
C : n
_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_
= 0
_
_
_
The set of coupled nonlinear equations is solved iteratively following a Newton-Raphson
procedure. The derivative of a is :
a
D
=
a
d
D
:
d
D
D
=
a
d
D
:
_
4
I
1
3 II
_
= 3
4
I
From the coupled set of iterative equations D and leading to new values of D and .
The iteration process is stopped when the residuals s1 and s2 are small enough.
For a plane stress situation, the deformation tensor must be adapted during the stress
update procedure. This implies that the elastic trial stress will change as well. Excluding
plane stress situations, the elastic trial stress is constant in the stress update procedure, so
De = O.
4
R : D +t = s1
u : D +v = s2
_
_
_
with
4
R =
4
I + 3
4
C :
4
I
t =
4
C : a
u =
_
3
4
C II :
4
C
_
: n
_
_
3
4
C II :
4
C
_
: a + 4yH
_
2
3a : a
_
1
2 a :
4
I
_
v = 3
4
C : a :
d
D + 2yH
_
2
3a : a
s1 = D D(tn)
4
C : n +
4
C : a
s2 = 3
d
D :
4
C : n
_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_
12.2.6 Stiness
To evaluate the iterative Updated Lagrange weighted residual equation not only the Cauchy
stress , but also the relation between the stress variation and Lu = ( u)c has to be
known, i.e. = 4M : Lu.
252
The consistent stiness tensor 4M, eventually leads to the consistent stiness matrix.
It must be derived from the coupled nonlinear equations for and . Iterative changes
(variations) of and can be derived.
To simplify notation we omit again the upper index i, which indicates the iteration step
number.
update equations
D D(tn)
4
C : n +
4
C : a = 0
3
d
D :
4
C : n
_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_
= 0
_
_
_
variation
D = D(tn) +
4
C : n
4
C : a
4
C : a = 0
3
d
D :
4
C : n + 3
d
D :
4
C : n

_
3
d
D :
4
C : a + 2yH
_
2
3a : a
_

_
3
d
D :
4
C : a + 3
d
D :
4
C : a+
2yH
_
2
3 a : a + 2yH1
2 [ 2
3a : a]
1/2 4
3a : a
_
= 0
_
_
_
253
12.3 Linear viscoelastic material behavior
The modeling of linear viscoelastic material behavior is based on the principles of superpo-
sition and proportionality. Current stress and strain are given by a Boltzmann integral over
the strain or stress history. Fourth-order relaxation ( 4C) and creep ( 4S) tensors relate stress
to strain and vice versa.
Experiments show that long past history has less impact on the current stress than re-
cent history. This fading memory property motivates the use of Prony series for 4C and 4S.
In the one-dimensional case they represent the behavior of discrete spring-dashpot models.
E1 E2
1 2
E
Eg
E1 E2
2 1
Fig. 12.15 : Generalized Maxwell and Kelvin model
(t) =
t _
=0
4
C(t ) : () d ; (t) =
t _
=0
4
S(t ) : () d
4
C(t) =
4
C +
N
i=1
4
Cie

t
i ;
4
S(t) =
4
S +
N
i=1
4
Si
_
1 e

t
i
_
12.3.1 Constitutive relations
We now focus attention on the calculation of the current stress (t), because this is of im-
portance in a numerical procedure like the nite element method. The hereditary integral is
evaluated after substitution of the Prony series for 4C(t).
Using the Prony series expression for 4C(t) and assuming the initial strain to be zero
(( = 0) = O), an expression for (t) can be derived.
(t) =
t _
=0
4
C(t ) : () d
4
C(t) =
4
C +
N
i=1
4
Ci e

t
i
_
_
_

254
(t) =
t _
=0
_
4
C +
N
i=1
4
Cie

t
i
_
: () d =
4
C : (t) +
N
i=1
4
Ci :
t _
=0
e

t
i () d
=
4
C : (t) +
N
i=1
i(t)
12.3.2 Incremental analysis
It is immediately clear that calculation of the stress involves the evaluation of a (large) number
of integrals over the complete time history. For this reason the deformation time period is
subdivided into a discrete number of time increments.
tn
tn+1
F(tn+1)
F(tn)
Fn(tn+1) t0
Fig. 12.16 : Incremental deformation
In the numerical analysis of the time dependent behavior, the total time interval [0, t] is
discretized :
[0, t] [t1 = 0, t2, t3, .., tn, tn+1 = t]
The timespan between two discrete moments in the time interval is a time increment. It is
assumed that these increments are of equal length :
t = ti+1 ti ; i = 1, ..., n
The hereditary integral is now split in an integral over [0, tn] and an integral over the last or
current increment [tn, tn+1 = t]. Here we consider only the i-th term of the series : i(t).
i(t) =
4
Ci :
t _
=0
e

t
i () d =
4
Ci :
_
_
tn _
=0
e

t
i () d +
t _
tn
e

t
i () d
_
_
=
4
Ci :
_
_e

t
i
tn _
=0
e

tn
i () d +
t _
=tn
e

t
i () d
_
_
255
= e
t
i 4
Ci :
tn _
=0
e

tn
i () d +
4
Ci :
t _
=tn
e
t
i () d
= e

t
i i(tn) +i(t)
Linear incremental strain
The stress i(tn) is known from the previous increment. Calculation of i(t) can be done
analytically when it is assumed that the strain is a linear function of time in each time
increment. For the current increment we have :
() = (tn) + ( tn)

t
() =

t
i(t) =
4
Ci :
t _
=tn
e

t
i

t
d =
4
Ci :
t _
=tn
e

t
i d

t
=
4
Ci : i
_
1 e
t
i
_

t
12.3.3 Stress update
Calculating the current stress does not mean that the Boltzmann integral has to be evaluated
over the total deformation history. When results are stored properly we can easily update
the stress (t).
(t) =
4
C : (t) +
N
i=1
i(t)
=
4
C : (t) +
N
i=1
_
e
t
i i(tn) +
4
Ci : i
_
1 e
t
i
_

t
_
12.3.4 Stiness
The variation of (t) results in the consistent material stiness tensor.
=
_
4
C +
N
i=1
4
Ci
i
t
_
1 e

t
i
__
:
=
4
M :
256
12.3.5 Isotropic material
For an isotropic material the mechanical behavior is the same in each material direction and is
characterized by two material parameters, the Lame coecients and . The elastic stiness
tensor 4C can then be written as :
4
C = II + 2
4
I
s
where the fourth-order tensors II and 4I
s
have the following index equivalents :
II ijkl
2
4
I
s
=
4
I +
4
I
rc
iljk +ikjl
Using the above expression for 4C the hydrostatic and deviatoric parts of the stress tensor
can be decoupled and expressed in the hydrostatic and deviatoric strain tensor, respectively.
Instead of the Lame coecients other elastic material parameters are often used : Youngs
modulus E, Poissons ratio , shear modulus G and bulk modulus K. These parameters are
related as only two independent material parameters exist.
=
4
C :
=
_
II + 2
4
I
s
: =
_
II +
_
4
I +
4
I
rc_
: = I tr() + 2
= (3 + 2)
1
3 tr()I + 2
d
= (3 + 2)
h
+ 2
d
= 3K
h
+ 2G
d
=
h
+
d
K =
1
3 (3 + 2) =
E
3(1 2)
; = G =
E
2(1 +)
; =
E
(1 +)(1 2)
For a viscoelastic isotropic material the stress tensor is also split into an hydrostatic and a
deviatoric part. In analogy with the elastic model, time dependent bulk and shear moduli
are used, which are expressed in a Prony series.
(t) =
h
(t) +
d
(t)
= 3
t _
=0
K(t )
d
d
_

h
()
_
d + 2
t _
=0
G(t )
d
d
_

d
()
_
d
K(t) = K +
n
i=1
Ki e

t
i
1
3(1 2)
_
E +
n
i=1
Ei e

t
i
_
G(t) = G +
n
i=1
Gi e

t
i
1
2(1 +)
_
E +
n
i=1
Ei e

t
i
_
257
Stress update
Discretising the total time interval [0, t] in equal time increments t = ti+1 ti; i = 1..n
allows an ecient calculation of the stress where an integral only has to be evaluated over the
current increment, which moreover can be done rather straightforwardly when it is assumed
that the incremental strain rate is constant (= linear incremental strain).
(t) =
4
C : (t) +
N
i=1
i(t)
=
4
C : (t) +
N
i=1
_
e

t
i i(tn) +
4
Ci : i
_
1 e

t
i
_

t
_
= 3K
h
+ 2G
d
+
N
i=1
_
e

t
i i(tn) +
i
t
_
1 e

t
i
__
3Ki
h
+ 2Gi
d
__
Stiness
The relation between a small change in stress and a small change in strain is straightforwardly
derived from the incremental stress relation.
= 3K
h
+ 2G
d
+
N
i=1
i
t
_
1 e
t
i
__
3Ki
h
+ 2Gi
d
_
Matrix/column notation
The relation between the incremental stress and strain tensor can be written in indices with
relation to a vector basis. Components can then be stored in columns and matrices. For
a two-dimensional deformation the following columns for stress and strain components are
dened :

T
=
_
11 22 33 12 21

T
=
_
11 22 33 12 21

Hydrostatic and deviatoric stress/strain components can be related to total stress/strain com-
ponents with the following matrices :
A
h
=
_

_
1
3
1
3
1
3 0 0
1
3
1
3
1
3 0 0
1
3
1
3
1
3 0 0
0 0 0 0 0
0 0 0 0 0
_

_
; A
d
=
_

_
2
3
1
3
1
3 0 0

1
3
2
3
1
3 0 0

1
3
1
3
2
3 0 0
0 0 0 1 0
0 0 0 0 1
_

_
resulting in :

h
= A
h

d
= A
d

The stress column can then be rewritten.


258

(t) =
_
3KA
h
+ 2GA
d
_

+
N
i=1
_
e

t
i

i
(tn)+
i
t
_
1 e

t
i
__
3Ki A
h
+ 2Gi A
d
__

stress variation

(t) =
__
3KA
h
+ 2GA
d
_

+
N
i=1
i
t
_
1 e
t
i
__
3Ki A
h
+ 2Gi A
d
_
_

Initial stiness formulation


Some implementations of the linear viscoelastic model (e.g. MARC) are formulated in such a
way that the initial moduli K0 and G0 are required. The initial moduli are dened as
K0 = K +
N
i=1
Ki ; G0 = G +
N
i=1
Gi
The relation for stress increment and stress variation can be derived easily.
(t) = 3K0
h
+ 2G0
d

N
i=1
_
1
_
1 e
t
i
_
i
t
_ _
3Ki
h
+ 2Gi
d
_

N
i=1
_
1 e

t
i
__

h
i (tn) +
d
i (tn)
_
stress variation
= 3K0
h
+ 2G0
d

N
i=1
_
1
_
1 e
t
i
_
i
t
_ _
3Ki
h
+ 2Gi
d
_
12.3.6 Examples
Tensile test
An axial strain step with amplitude 0.01 is prescribed on an axisymmetric tensile bar with
initial cross-sectional area A0 = 10 mm2. The stress response is calculated for a 12-mode
generalized Maxwell model. The modal parameters are listed in the table.
259
E [MPa] [s] E [MPa] [s]
1 3.0e6 3.1e-8 2 1.4e6 3.0e-7
3 3.9e6 3.0e-6 4 5.4e6 2.9e-5
5 1.3e6 2.8e-4 6 2.3e5 2.7e-3
7 7.6e4 2.6e-2 8 3.7e4 2.5e-1
9 3.3e4 2.5e+0 10 1.7e4 2.4e+1
11 8.0e3 2.3e+2 12 1.2e4 2.2e+3
0 0.02 0.04 0.06 0.08 0.1 0.12
3
4
5
6
7
8
9
10
11
12x 104
t [s]
22 [N
/m
m
2 ]
Fig. 12.17 : Tensile stress versus time for axisymmetric element
260
12.4 Viscoplastic material behavior
The one-dimensional mechanical representation of the elastoviscoplastic Perzyna model con-
sists of a spring in series with a friction slider, a hardening spring and a linear viscous dash-
pot. The series-spring represents the elastic part of the material response. The viscoplastic
response, represented by the hardening spring and the viscous dashpot, becomes manifest as
soon as the friction slider opens when the stress exceeds a characteristic value. This limit
value is the current yield stress y.
After yielding, the total strain is the sum of the elastic strain e and the viscoplastic
strain vp.

E
y
e vp
H

Fig. 12.18 : Discrete model for viscoplastic material behavior


12.4.1 Kinematics
Transformation from the undeformed conguration at time t0 (position vector x0) to the
current conguration at time t (position vector x) is described by the deformation tensor
F = ( 0x)c, where 0 is the gradient operator with respect to the undeformed state.
The right and left Cauchy-Green strain tensors, C and B, are functions of F as is the
Green-Lagrange strain tensor E. Material velocity is taken into account by the deformation
and rotation rate tensors D and , the symmetric and skew-symmetric parts of the velocity
gradient tensor L = ( v)c, where is the gradient operator with respect to the current state
and v is the velocity of the material volume.
In Perzynas model the total deformation F is decomposed multiplicatively into an elastic
and a viscoplastic contribution. Regarding the kinematics, this implies the introduction of
elastic and viscoplastic (rate) tensors.
To make the decomposition unique it is commonly assumed that the viscoplastic rotation
rate is zero, i.e. p = O. Superimposed material rotations are thus fully represented in Fe.
261
t0
t
F
Fvp
Fe
Fig. 12.19 : Multiplicative decomposition of total deformation
F = ( 0x)
c
= Fe Fvp
C = F
c
F ; B = F F
c
; E = 1
2 (C I)
L = F F
1
= ( v)
c
= Le +Lvp = (De +e) + (Dvp +vp) = (De +e) +Dvp
12.4.2 Constitutive relations
Elastic deformation
The stress is related to the elastic strain. Because we want to describe large elastic strains,
the elastic behavior must be described with a hyper-elastic model. In that case it is assumed
that an elastic strain energy function exists, which can be used to calculate the stress. The
2nd-Piola-Kirchho stress tensor P is related to the Green-Lagrange strain tensor E. The
current stress state is characterized by the Kirchho stress = F P F
c
.
An elastic energy function is chosen, which characterizes isotropic, compressible material
behavior. The fourth-order material tensor is completely determined by the volume ratio
J = det(F) and by the constant Lame coecients and , which are related to Youngs
modulus and Poissons ratio.
2nd-Piola-Kirchho stress tensor P =
W(Ee)
Ee
= 2
W
Ce
= F
1
F
c
elastic strain energy
W(1, 2, 3) =
1
2
_

2
1 +
2
2 +
2
3 3 2 ln(J)
_
+
1
2 {ln(J)}
2
with =
E
(1 +)(1 2)
; =
E
2(1 +)
262
Yield criterion and hardening
A yield function F is used to evaluate the stress state and to check whether the deformation
is purely elastic (F < 0) or viscoplastic (F 0). The current stress state, represented by the
equivalent or eective Kirchho stress , is compared to the current yield stress y, which
increases from its initial value y0 due to plastic deformation and is therefore related to the
eective viscoplastic strain vp. The relation between y and vp is described by the hardening
rule. To decide whether elastic or viscoplastic deformation occurs, the Kuhn-Tucker relations
are used.
Yield criterion F = y( vp)
eective viscoplastic strain vp =
t _
=0
vp d ; vp =
_
2
3 Dvp : Dvp
hardening law y = y(y0, vp) with
y
p
= H( p)
Kuhn-Tucker relations F < 0 elastic deformation
F 0 viscoplastic deformation
Viscoplastic deformation
During viscoplastic deformation the direction of the viscoplastic strain rate is dened by the
commonly used normality or associative ow rule : the viscoplastic strain rate is directed
normal to the yield surface in stress space. The length of Dvp is characterized by the rate of
the viscoplastic multiplier .
In contrast to elastoplastic models, stress states outside the yield surface can exist, which
explains why these viscoplastic models are often called over-stress models. The rate of the
viscoplastic multiplier is related to an over-stress function (F) by a uidity parameter and
a rate-sensitivity parameter N, which has to satisfy N 1 to make (F) convex.
The time-derivative of C can be related to D and its viscoplastic part to Dvp.
F = 0
Dvp =
F

= 0
Fig. 12.20 : Associative ow rule
263
Dvp =
F

= a
= (F) ; (F) =
_
F
y0
_N
_
_
_
C = 2 F
c
D F Cvp = 2 F
c
Dvp F
12.4.3 Constitutive model
The material model can be summarized as a set of constitutive equations. It must be used to
determine the stress when (an approximation of) the deformation is known. It is also used to
derive a relation between the variation of stress and deformation, which is an essential part
of the element stiness matrix.
F < 0 C = Ce
P = 2
2W
C
2 : C ; Cvp = O ; vp = 0
F 0 C = Ce + Cvp
P = 2
2W
C
2 :
_
C 2 F
c
a F
_
= (F) =
_
F
y0
_N
_
_
_
y = y(y0, vp) ; vp =
_
2
3 a : a
Von Mises plasticity
For the Von Mises yield criterion, the yield surface is a circular cylinder in principal stress
space. When hardening is assumed to be isotropic, the radius of the cylinder will change upon
viscoplastic deformation. The normal to the yield surface can be expressed in the deviatoric
stress d. It then follows that vp = .
=
_
3
2 d : d F =
_
3
2d : d y( vp)
a =
F

=
F
d
:
d

=
_
3
2
_
3
2
d
:
d
_1/2

d
:
4
I
_
:
d

=
_
3
2
_
3
2
d
:
d
_1/2

d
:
4
I
_
:
_

_

1
3tr()I
_
_
=
_
3
2
_
3
2
d
:
d
_1/2

d
:
4
I
_
:
_
4
I
1
3 II

264
=
3
2
_
3
2
d
:
d
_1/2

d
:
_
4
I
1
3 II
_
=
3
2
_
3
2
d
:
d
_1/2

d
=
3
2
1


d
Dvp = a = 3
2
1


d
vp =
_
2
3
2 3
2
3
2
1
2
d : d =
12.4.4 Incremental analysis
The gure shows the relevant congurations in a large strain viscoplastic deformation process.
Starting from the undeformed conguration at time t0 the external load is employed and
the deformation leads to the current conguration at time t. During a numerical analysis of
this deformation process the state of the material is determined at a nite number of discrete
moments ti, i = 0, 1, .., n+1. The period between two subsequent moments is an increment :
ti = ti+1 ti. The increments are assumed to be of equal length.
It is assumed that the analysis has brought us to t = tn, the beginning of the current
increment and that all relevant variables are known and satisfying all governing equations
(balance laws, boundary conditions, constitutive relations). The state at the current time
t = tn+1, the end of the current increment has to be determined.
The transformation during the current increment is described by the deformation tensor
Fn(), where indicates a moment in time during the last (= current) increment : tn
tn+1.
tn
tn+1
F(tn+1)
F(tn)
Fn(tn+1) t0
Fig. 12.21 : Incremental deformation
F() = Fn() F(tn) Fn() = F() F
1
(tn)
Fn = ( nx)
c
= Rn Un ; Jn = det(Fn) ; = F
c
n n
D =
1
2
_
( v)
c
+ ( v)
_
=
1
2
_
Fn F
1
n +F
c
n F
c
n
_
12.4.5 Stress update
During the increment t = tn+1 tn the stress evolution equations are integrated using an
implicit Euler integration scheme.
265
P = 2
2W
C
2 : ( C 2 F
c
a F)
= (F) =
_
F
y0
_N
_
_
_
implicit Euler integration scheme
f() = f(tn+1) =
1
t
{f(tn+1) f(tn)}
incremental equations
P = P(tn) + 2
2W
C
2 : {C C(tn) 2 F
c
a F}
= t(F)
_
_
_
The current or end-increment time t = tn+1 is not indicated further. Constitutive equations
are reformulated in the Kirchho stress tensor , using = F P F
c
, the incremental
deformation tensor Fn and the Almansi strain tensor en. A fourth-order elastic material
tensor 4H is introduced and can be calculated for the Neo-Hookean elastic energy function
W.
Fn = F F
1
(tn) C C(tn) = F
c
(I F
c
n F
1
n ) F = 2 F
c
en F
= F P F
c
= F P(tn) F
c
+F P P
c
= F F
1
(tn) (tn) F
c
(tn) F
c
+
2 F
_
2W
C
2 : {2F
c
en F 2 F
c
a F}
_
F
c
incremental equations
= Fn (tn) F
c
n +
4
H : (en a)
= t (F)
_
_
_
with
4
H = 4 F
_
F
2W
C
2 F
c
_lc,rc
F
c
= 2{ ln(J)}
4
I
rc
+II
266
Elastic stress predictor
The rst step in evaluating the end-increment stress is the calculation of the elastic stress
predictor. As a rst assumption the current increment is taken to be purely elastic, so = 0.
The elastic trial stress is used to evaluate the yield condition and to see if the assumption of
elastic deformation holds. There are two possibilities :
1. the increment is indeed fully elastic,
2. the yield criterion is violated which implies that during the increment further elastovis-
coplastic deformation has taken place.
elastic trial stress tr = Fn (tn) F
c
n +
4
H : en
yield criterion F =
_
3
2 (tr)
d
: (tr)
d
y(y0, vp(tn))
1. F < 0 elastic increment
2. F 0 elastoviscoplastic increment
matrix/column notation

tr
= A

+H
c
en

F =
_
3
2
_

tr
_T _

tr
_
t
()
with
_
_
_
H = 2 { ln(J)} I +I

T
en =
1
2
_
I F
T
n F
1
n
_
en
A = Fn (tn)F
T
n A

Elastic increment
When it is concluded that the current increment is purely elastic, the end-increment or current
stress equals the calculated elastic trial stress. Viscoplastic strain does not need updating and
is thus also known.
(tn+1) = tr ; = 0
vp(tn+1) = vp(tn) ; y(tn+1) = y(tn)
Elastoviscoplastic increment
When the yield criterion as evaluated with the elastic trial stress is violated, it must be con-
cluded that there is viscoplastic deformation during the current increment. The current stress
and viscoplastic multiplier must then be solved from the set of coupled nonlinear equations.
267
= Fn (tn) F
c
n + 4H : en 4H : a
= tr 4H : a
= t (F)
_
_
_
=
_
3
2 d : d ; vp =
_
2
3 Dvp : Dvp
a =
3
2
d

; F = y ; y = y(y0, vp)
The coupled set of equations is solved iteratively following a Newton-Raphson procedure. In
the stress update procedure it may be necessary to take into account the change in the elastic
trial stress and deformation. This is the case in a plane stress situation. Both J and tr
can then be expressed in and . New variables (J1, J2, M1, 4M2) are introduced, which
can be specied explicitly later.
From the coupled set of iterative equations and can be solved, whereupon new
(better) values of and are determined. The iteration process is stopped when the
residuals s1 and s2 are small enough.
When the iteration process has converged, the current values of and are known.
Then the Cauchy stress and the viscoplastic deformation rate Dvp can be determined. The
latter is used to calculate the eective viscoplastic strain vp. Subsequently the yield stress
is updated according to the hardening rule.
tr +
4
H : a +
4
H : a +
4
H : a = s1
t
_

F
_
a : t
_

F
__
F
vp
_
= s2
_
_
_
with
_
_
_
tr = M1 +
4
M2 :

4
H =
_
4H
J
_
J =
4
c J
a =
_
a

_
: =
4
b :
J = J1 +J2 :
coupled set of iterative equations
4
R : +t = s1
u : +v = s2
_
_
_
268
Derivatives
The variations of various variables are determined by dierentiation.
The hardening law relates the current yield stress to the equivalent viscoplastic strain.
To describe the intrinsic softening followed by hardening, the relation between y and vp is
taken to be a polynomial of 7th-order. Coecients are tted onto experimental data.
y = y0 +h vp +a
2
vp +b
3
vp +c
4
vp +d
7
vp
4H
J
= 2
1
J
4
I =
4
c c = 2
1
J
I
_
F
vp
_
=

vp
(y( vp)) = h 2a vp 3b
2
vp 4c
3
vp 7d
6
vp
_

F
_
=

F
__
F
y0
_N
_
=
N
y0
_
F(

, vp)
y0
_N1
a

=
a
d
:
d

a
d
=

d
_
3
2
1

d
_
=
3
2
_

2
d
_

d
+
3
2
1 4
I
d

_
1
3tr()I
_
=
4
I 1
3 II

d
=

d
__
3
2
d
:
d
_1/2
_
=
3
2
1

d
= a
=
_

1
aa +
3
2
1 4
I
_
:
_
4
I
1
3II
_
=
1
aa +
3
2
1 4
I
1
2
1 4
I : II =
4
b
b =
1
a

T
+
3
2
1
I
1
2
1
I

T
=
1
_
a

T
+
3
2 I
1
2 I

T
_
12.4.6 Stiness
To evaluate the iterative Updated Lagrange weighted residual equation, not only the Cauchy
stress , but also the relation between the stress variation and Lu = ( u)c has to be
known, i.e. = 4M : Lu.
The consistent stiness tensor 4M, eventually leads to the consistent stiness matrix.
It must be derived from the coupled nonlinear equations for and . Iterative changes
(variations) of and can be derived. To simplify notation we omit again the upper index
i, which indicates the iteration step number.
= Fn (tn) F
c
n + 4H : en 4H : a
= t (F)
_
_
_

269
=Fn (tn) F
c
n +Fn (tn) F
c
n +
4
H : (en a)+
4
H : en
4
H : a
4
H :
_
a

_
:
=
__
t
_

F
__
/
_
1 t
_

F
__
F
vp
___
a :
=c1a :
_
_
_

_
4
I +
4
H :
_
a

_
+c1
4
H : aa
_
: =
Fn (tn) F
c
n +Fn (tn) F
c
n+ =
4
H : (en a) +
4
H : en
To arrive at a relation between and Fn some new tensors are introduced which can be
specied later, when a coordinate system is chosen.
Fn (tn) F
c
n +Fn (tn) F
c
n =
4
T : Fn
J = det(Fn) = det(Fn +Fn) = J(1 +F
1
n : Fn) J = J F
1
n : Fn

4
H =
_
4H
J
_
J =
_
4H
J
_
_
JF
1
n : Fn
_
en =
1
2F
c
n F
1
n
1
2F
c
n F
1
n =
4
A1 : F
1
n
F
1
n = F
1
n Fn F
1
n =
4
A2 : Fn
_
_
_

en = (
4
A1 :
4
A2) : Fn =
4
P : Fn
Using the denition = J a relation between and Fn can be derived, which can be
transformed to =
4
M : Lu .
Kirchho stress variation
_
4
I +
4
H :
_
a

_
+c1
4
H : aa
_
: =
_
4
T +
_
4H
J
_
: (en a)JF
1
n +
4
H :
4
P
_
: Fn
4
V : =
4
E : Fn =
4
V
1
:
4
E : Fn
Cauchy stress variation
= J =
1
J

=
1
J
( J) =
1
J
_
4
V
1
:
4
E JF
1
n
_
: Fn =
4
C : Fn
=
4
C :
_
F
c
(tn) F
c
_c
=
4
C :
_
F
c
(tn) F
c
L
c
u
_c
=
4
M : Lu
270
4
T : Fn = Fn (tn) F
c
n +Fn (tn) F
c
n
= Fn A
c
+A F
c
n with A = Fn (tn)
4
T
rc
: F
c
n = (A F
c
n)
c
+A F
c
n
TFn

t
=
_
AFn

t
_
t
+AFn

t
= A
r
Fn

t
+AFn

t
=
_
A
r
+A
_
Fn

T = A
r
+A with A = Fn

T
(tn)
4
A1 : F
1
n = 1
2 F
c
n F
1
n + 1
2F
c
n F
1
n = 1
2
_
F
c
n F
1
n
_c
+ 1
2F
c
n F
1
n
A1Fn

1
t
=
1
2 Fn
1
tr
Fn

1
+
1
2Fn
1
t
Fn

1
A1
c
F

1
n
=
1
2
_
Fn
1
tr
+Fn
1
t
_
Fn

A1 =
1
2
_
Fn
1
trc
+Fn
1
tc
_
4
A2 : Fn = F
1
n Fn F
1
n = F
1
n
_
F
c
n F
c
n
_c
= F
1
n A
c
with A = F
c
n F
c
n
A2
c
Fn

1
= Fn
1
A

t
with A

= Fn
1
t
Fn

t
A

t
= Fn
1
tr
Fn

t
A2 Fn

t
= Fn
1
Fn
1
tr
Fn

t
A2 = Fn
1
Fn
1
tr
4
P =
4
A1 :
4
A2 P = A1
c
A2
Matrix/column notation
The matrix/column notation for the consistent stiness matrix is derived.
=
4
C : Fn

= C Fn

t
Fn =
_
F
c
(tn) F
c
_c
Fn

=
_
F1
t
(tn)F

t
_
t
Fn

t
= F
1
t
(tn)F

t
F
c
= F
c
L
c
u F

t
= F
t
Lu

t
nally

=
_
C F
1
t
(tn)F
t
_
Lu

t
= MLu

t
M = C F
1
t
(tn)F
t
C =
1
J
_
V
1
E
r
J

Fn

T
_
V = I +H
c
b +c1H
c
a

T
E = T 2I
_
en

__
Fn

1
_T
+H
c
P
271
12.4.7 Plane strain
For plane strain some terms in the stress update equations vanish. During viscoplastic defor-
mation the volume will not change, so J = 0. Also, the elastic trial stress will remain as it
is, i.e. tr = 0.
J = J1 +J2 : = 0
tr = M1 +
4
M2 : = O
Stress update
4
R : +t = s1
u : +v = s2
_
_
_
4
R =
4
I +
4
H :
4
b ; t =
4
H : a
u = t
_

F
_
a ; v = 1 t
_

F
__
F
vp
_
s1 = tr +
4
H : a ; s2 = t (F)
Matrix/column notation
It is assumed that there is no deformation in the x3-direction (u3 = 0), which results in the
plane strain deformation in the (x1x2)-plane. The plane strain case can be derived rather
straightforward from the three-dimensional formulation.
_
_
R
c
t

T
t
v
_
_
_
_

_
_ =
_
_
s

1
s2
_
_
R = I +H b
t
; t

= H a

t
u

= t
_

F
_
a

; v = 1 t
_

F
__
F
vp
_
s

1
=

tr
+Ha

t
; s2 = t (F)
Stiness
The plane strain stiness in tensorial notation is analogous to the three-dimensional relation.
=
4
C : Fn =
1
J
_
4
V
1
:
4
E JF
1
n
_
: Fn
4
V =
_
4
I +
4
H :
4
b +c1
4
H : aa
_
4
E =
_
4
T +
4
c : (en a)JF
1
n +
4
H :
4
P
_
Fn (tn) F
c
n +Fn (tn) F
c
n =
4
T : Fn
en =
4
P : Fn
272
Matrix/column notation
Matrix/column notation of the consistent stiness matrix for plain strain deformation.

= C
_
F

n
_
t
=
_
1
J
_
V
1
E
r

JF

T
n
__ _
F

n
_
t
V = I +H
c
b +c1H
c
a

T
E = T + 2I
_
e

_
JF

T
n
+H
c
P
12.4.8 Plane stress
For plane stress we have to take into account the variation of the trial stress and the defor-
mation.
Stress update
4
R : +t = s1
u : +v = s2
_
_
_
4
R =
4
I
4
M2 +
4
C : aJ2 +
4
H :
4
b ; t = M1 +
4
C : aJ1 +
4
H : a
u = t
_

F
_
a ; v = 1 t
_

F
__
F
vp
_
s1 = trial +
4
H : a ; s2 = t (F)
Matrix/column notation
Introduction of a suitable (problem dependent !) coordinate system leads to the transforma-
tion of vectors and tensors into their components, which are stored in columns and matrices.
_
_
R
c
t

T
t
v
_
_
_
_

_
_ =
_
_
s

1
s2
_
_
R = I M
2
+C a

r
J

T
2
+H b
r
; t

= M
1
+Ca

t
J1 +Ha

t
u

= t
_

F
_
a

; v = 1 t
_

F
__
F
vp
_
s

1
=

tr
+Ha

t
; s2 = t(F)
273
Stiness
The plane stress stiness in tensorial notation is analogous to the three-dimensional relation.
=
4
C : Fn =
1
J
_
4
V
1
:
4
E JF
1
n
_
: Fn
with
4
V =
_
4
I +
4
H :
_
a

_
+c1
4
H : aa
_
4
E =
_
4
T +
_
4H
J
_
: (e a)JF
1
n +
4
H :
4
P
_
and
Fn (tn) F
c
n +Fn (tn) F
c
n =
4
T : Fn
e =
4
P : Fn
Matrix-column notation
With the assumption that 13 = 23 = 33 = 0, the three-dimensional formulation reduces to
that for two-dimensional plane stress deformation in the (x1x2)-plane. Columns with relevant
components of stress and deformation rate are :

=
_
11 22 12 21
T
D

=
_
D11 D22 D12 D21
T
During the plane stress return mapping we have
F11 = F22 = F12 = F21 = 0 and

trial
= 0

As deformation in x3-direction is allowed, J can be expressed in F33 :


J = (F11F22 F12F21)F33 = J1 +J

T
2

which results in the set of iterative equations for

and .
12.4.9 Examples
Tensile test
A square plate or cylindrical bar is loaded uniaxially. Dimensions are listed in the table.
274
initial width w0 100 mm
initial height h0 100 mm
initial thickness d0 0.1 mm
initial radius r0
_
(10/) mm
initial height h0 100 mm
Viscoplastic model in tensile test
The Perzyna model parameter values for polycarbonate (PC) are used and listed in the table.
The axial elongation is prescribed as a linear function of time with a constant elongation
rate. The tensile bar is axisymmetric with initial cross-sectional area A0 = 10 mm2. The
axial stress and force are shown in the gure as a function of the elongation.
E 1800 MPa 0.37 -
y0 37 MPa H -200 MPa
0.001 1/s N 3 -
a 500 MPa b 700 MPa
c 800 MPa d 30000 MPa
elongation rate
l
h0
= {0.01, 0.1, 1} s
1
1 1.1 1.2 1.3 1.4 1.5
0
50
100
150
200
250
300
350
400

[M
P
a]
deps = 1
deps = 0.1
deps = 0.01
1 1.1 1.2 1.3 1.4 1.5
0
500
1000
1500
2000
2500
3000

F [N
]
deps = 1
deps = 0.1
deps = 0.01
Fig. 12.22 : Axial stress and force versus elongation for PC
275
Shear test
The simple shear test is analyzed with one element, where the horizontal displacement in the
upper nodes is prescribed. Because there are no unknown degrees of freedom, the stiness
matrix is not used. Only strains, stresses and reaction forces are calculated.
initial width w0 100 mm
initial height h0 100 mm
initial thickness d0 0.1 mm
Viscoplastic model in shear test
The shear force is calculated for polycarbonate (PC). The prescribed strain rate is constant.
strain rate =
u
h0
= 0.01 s1
1 2
3 4
0 0.2 0.4 0.6 0.8 1
0
200
400
600
800
1000
1200
1400
1600
1800

F x [N
]
Fig. 12.23 : Shear force versus shear strain for plane strain
276
12.5 Nonlinear viscoelastic material behavior
The one-dimensional mechanical representation of the nonlinear viscoelastic (Leonov) model
consists of a hardening spring in parallel with a Maxwell model, of which the viscosity is a
nonlinear function of the stress.
For some materials the viscosity is decreased using a damage parameter, to describe
intrinsic softening. Hardening at higher strains is described by the parallel spring.
In the compressible Leonov model with hardening, the Cauchy stress is additively
decomposed in an eective or driving stress s and a hardening stress w. This decomposition
reects the contribution of secondary interactions between polymer chains and that of the
entangled polymer network.
s
w
(s) E
H

e v
Fig. 12.24 : Model for nonlinear viscoelastic behavior
12.5.1 Kinematics
The deformation tensor F is multiplicatively decomposed into an elastic (Fe) and a plastic
(Fp) contribution : F = Fe Fp. This decomposition follows from the postulate of a stress-
free plastic intermediate conguration Cp. As the decomposition is not unique with respect
to rotational contributions, an extra assumption will later be needed regarding the rotations.
It is assumed that during plastic deformation the volume change is zero, i.e. Jp =
det(Fp) = 1 and thus J = det(F) = det(Fe). The elastic volume deformation is decoupled
from the isochoric distortional deformation by the denition of the tensor Fe according to
Fe = J1/3Fe.
The left Cauchy-Green strain tensor B = F F
c
is used as a strain measure. Its volume
invariant elastic part is given by Be = Fe F
c
e. The velocity gradient tensor L = ( v)c =
F F
1
can be written as the sum of the symmetric deformation rate tensor D and the
skew-symmetric spin tensor : L = D+. Using the decomposition of F, we can split L
in an elastic and a plastic part. This leads to associated tensors De, Dp, e and p.
To make the decomposition of F unique, p is chosen equal to the null tensor. It has
been shown by e.g. Boyce, that this specic choice regarding rotational contributions has no
signicant inuence on the overall stress-strain behavior.
277
F
Co
Fp
Cp
Cc
Fe
J1/3I
C
Fe
multiplicative decomposition of F
F = ( 0x)
c
= Fe Fp = J
1/3
I Fe Fp
C = F
c
F ; B = F F
c
Be = Fe F
c
e
L = F F
1
= ( v)
c
= Le +Lp = (De +e) + (Dp +p) = (De +e) +Dp
12.5.2 Constitutive relations
Stress decomposition
The deviatoric part of the eective stress, sd, is related to B
d
e through a generalized Hookean
relation. The hydrostatic part sh = pI (p = hydrostatic material pressure) is related to the
volumetric deformation.
Hardening is modeled according to Gaussian chain statistics as this model is applicable
to a large number of thermoplastic polymers, both amorphous and semi-crystalline, up to very
high extension ratios. A three-dimensional generalization of the Gaussian approach results in
a Neo-Hookean relation. Material parameters are : the shear modulus G, the bulk modulus
and the hardening modulus H. For the elastic part to be hyper-elastic, the shear modulus
G should be replaced by
G
J .
additive decomposition of = s +w = s
d
+s
h
+w
stress-strain relations s = G B
d
e +(J 1)I ; w = H B
d
Elastic deformation
As the model describes time- and history-dependent behavior, the elastic strain must be
updated by integration of appropriate evolution equations for Be. The expression for
Be can
be derived starting from Be = Fe F
c
e and using the decomposition F = Fe Fp and the
assumption p = O.
278
Be = Fe F
c
e
Be =
Fe F
c
e + Fe
F
c
e
F = Fe Fp Fe = F F
1
p
Fe =
F F
1
p + F F
1
p
Be =
_
F F
1
p + F F
1
p
_
F
c
e + Fe
_
F
c
p
F
c
+ F
c
p F
c
_
=
_
F F
1
p F
1
e + F F
1
p F
1
e
_
Be +
Be
_
F
c
e F
c
p
F
c
+ F
c
e F
c
p F
c
_
=
_
L+ Fe Fp F
1
p
F
1
e
_
Be + Be
_
L
c
+ F
c
e F
c
p F
c
p F
c
e
_
Fp F
1
p = I Fp F
1
p = Fp F
1
p
= ( L Dp) Be + Be ( L
c
Dp)
Viscoplastic deformation
The viscoplastic deformation rate Dp is related to the deviatoric stress sd, through the vis-
cosity, which is a nonlinear function of the equivalent stress s, the hydrostatic pressure p, the
absolute temperature T and the damage parameter D.
For polymers the Eyring viscosity function is successfully used and for metal alloys the
viscosity function of Bodner-Partom ([?],[?],[?]).
For the equivalent deviatoric stress s is the Von Mises denition is used.
Dp =
1
2
s
d
= ( s, p, T, D)
s =
_
3
2sd : sd
p = (J 1)I
_
_
_
Plastic strain rate
The current value of Be(t) can be determined by integration of
Be. However, the integrand
Be is not objective, so that rigid body rotations will inuence the results, which is of course
not allowed. The problem of non-objectivity of
Be can be circumvented by using an evolution
equation for the Cauchy-Green plastic strain tensor Cp, which is invariant.
Starting from F an expression for Cp can be derived, containing
Be. With the earlier
derived expression for
Be, Cp can be expressed in Be and Dp. The relation between Dp
and B
d
e allows Cp to be related to Be and Cp. This equation states that the direction of
the plastic strain rate is dened by the directional tensor A, while the plastic strain rate
magnitude is governed by the characteristic plastic deformation rate .
The plastic strain rate is invariant for rigid body rotations. It is shown in literature that
this formulation with a plastic predictor, can be used to apply an implicit, robustly stable
and ecient time integration procedure.
279
F = Fe Fp Cp = F
c
p Fp = F
c
B
1
e F
Cp = F
c
B
1
e
_
Be L
c
+ Be
B
1
e Be + L Be
_
B
1
e F
_
_
_
(1)
Be = ( LDp) Be + Be ( L
c
Dp)
Be
B
1
e = L Be L
c
B
1
e +Dp + Be Dp B
1
e
_
_
_
(2)
plastic strain rate (1) + (2)
Cp = F
c
B
1
e
_
Dp Be + Be Dp
_
B
1
e F
with Dp =
1
2
s
d
=
G
2
B
d
e
=
G

_
C
1
3 tr( Be)Cp
_
=
_
C
1

Cp
_
= A
12.5.3 Constitutive model
The material model can be summarized as a set of constitutive equations. The dierential
equations must be integrated to determine the current elastic strain and stress. Also the
variation of he stress must be derived from the constitutive model, representing the current
stiness.
J = det(F) F = J
1/3
F B = F F
c
w = H B
d
p = (J 1) s
h
= pI
Cp =
G

_
C
1
3 tr( Be) Cp
_
Be = F C
1
p F
c
_
_
_
s
d
= G B
d
e s =
_
3
2sd : sd
= s
d
+s
h
+w
Eyring viscosity
For polymer materials the plastic deformation rate tensor Dp is related to the deviatoric
stress sd by an Eyring viscosity . This is a function of the equivalent Von Mises stress
s, the hydrostatic stress p and the absolute temperature T. In the model, presented here,
the viscosity is depending on an intrinsic softening quantity D, determined by an evolution
equation, which has to be solved with the other constitutive relations.
Material parameters are :
280
A0 time constant
H activation energy
R universal gas constant
p hydrostatic pressure
parameter describing pressure dependence
V shear activation volume
D saturation value of D
=
A s

3 sinh
_
s

30
_
s =
_
3
2 sd : sd ; p =
1
3 tr()
A = A0 exp
_
H
RT
+
p
0
D
_
; 0 =
RT
V
evolution equation for softening parameter
D = h
_
1
D
D
_
s

6
; D [0, D]
Bodner-Partom viscosity
To describe viscoplastic behavior of metals, the plastic deformation rate tensor Dp is related
to the deviatoric stress sd by a Bodner-Partom viscosity . This is a function of the equivalent
Von Mises stress and Z, the resistance to plastic ow. 0 is a constant which reects the
smoothness of the transition from the elastic to the viscoplastic response and n characterizes
the rate sensitivity of the viscoplastic response. The plastic ow resistance Z depends on the
equivalent plastic strain p. Its lower and upper bounds are Z0 and Z1.
The Bodner-Partom model corresponds to isotropic hardening.
=
s

120
exp
_
1
2
_
Z

_2n
_
s =
_
3
2sd : sd ; p =
_
2
3Dp : Dp p
Z = Z1 + (Z0 Z1)e
m p
12.5.4 Incremental analysis
The plastic strain Cp at the current time t must be determined by integration of the dierential
equation for Cp(). In an incremental procedure the total deformation period is subdivided
into a number of sequential time increments : t = ti+1 ti ; i = 0 n. A solution for
the governing equations is determined for the discrete end-increment times, starting from the
known state with known values of all variables at the begin-increment time. This implies
281
that the dierential equation for Cp has to be solved for the last increment tn tn+1 assuming
that Cp(tn) is known. For simplicity we skip the indication of the current end-increment time
tn+1.
We now focus attention on the last increment [tn, tn+1]. It is assumed that at time tn
the conguration Cn is completely known and all equations are satised. The begin-increment
state Cn at = tn is taken as the reference conguration for deformation variables, which is
known as the Updated Lagrange procedure.
F
F(tn)
Co
Fp
Cn
Fn
Fnp
Fn
Cp
Cc
Fe
J1/3I
C
Fe
Fig. 12.25 : Incremental deformation
F() = Fn() F(tn) Fn() = F() F
1
(tn)
F() = Fn() F(tn)
Fn =
_
nx
_c
= Rn Un
Incremental plastic strain
Using the multiplicative decomposition, an expression for Cpn() can be derived. It contains
the tensor
B
1
en which is the rotation neutralized version of B
1
e :
B
1
en = R
c
n B
1
e Rn
where Rn is the incremental rotation tensor.
Cp() = F
c
p() Fp() = F
c
() B
1
e () F()
with F() = Fn() F(tn)
= F
c
(tn)
_
F
c
n() B
1
e () Fn()
_
F(tn)
= F
c
(tn) Cpn() F(tn)
282
incremental rotation neutralized plastic strain
Cpn() = F
c
n() B
1
e () Fn()
= Un()
_
R
c
n() B
1
e () Rn()
_
Un()
= Un()
B
1
en () Un()
Constitutive equations
With the incremental procedure the constitutive model is formulated in the incremental vari-
ables.
J = det(F) F = J
1/3
F B = F F
c
w = H B
d
p = (J 1) s
h
= pI
Cpn =
G

_
Cn 1
3 tr
_
Ben
_
Cpn
_
Ben = Un C
1
pn U
c
n Be = Rn
Ben R
c
n
D = h
_
1
D
D
_
s

6
= ( s, p, T, D)
_
_
_
s
d
= G B
d
e s =
_
3
2sd : sd
= s
d
+s
h
+w
12.5.5 Stress update
The incremental plastic strain rate Cpn() can be integrated over the last increment tn tn+1
to determine Cpn(tn+1). An implicit backward Euler integration scheme is used. With
Un(tn) = I we have Cpn(tn) =
B
1
en (tn) = B
1
en (tn).
The scalar is the so-called elasticity scalar, a state variable indicating the proportion of
incremental elastic/plastic strains with respect to the incremental total strains ( = 1, fully
elastic increment, and = 0, fully plastic increment). This parameter depends on and thus
on s and Cp. The isochoric elastic strain Be can be calculated from Cpn and Fn.
incremental plastic strain rate
Cpn() = ()
_
Cn()
1
n()
Cpn()
_
;
1
n
=
1
3tr
_
Ben
_
283
implicit backward Euler
f() = f(tn+1) =
1
t
{f(tn+1) f(tn)}
plastic strain
1
t
[Cpn Cpn(tn)] =
_
Cn
1
n
Cpn
_

Cpn =
n t
n +t
Cn +
n
n +t
Cpn(tn)
elasticity parameter : =
n
n +t
Cpn = n(1 ) Cn +Cpn(tn)
total isochoric elastic strain
Be = Rn
Ben R
c
n = Fn C
1
pn F
c
n
Sub-incremental plastic strain update
The dierential equation for the incremental plastic strain can be integrated more accurately
by subdividing the current increment t = tn+1 tn in a number (ns) of sub-increments t =
t/ns. The known iterative approximation for the end-increment deformation ( Fn Cn)
is also subdivided and subsequently values for C
j
pn are determined with a backward Euler
integration scheme.
The incremental rotation is not taken into account during this procedure but incor-
porated afterward at the end-increment time. It is also assumed that the principle strain
directions do not change during the integration procedure.
The sub-incremental integration scheme results in a more accurate determination of Cpn
and thus . It allows for larger incremental time steps.
Be aware that the nal j = ns+1 is not the elasticity parameter introduced earlier,
indicating the elastic part of the increment. This must be calculated without using sub-
increments or just according to
=
n
n +t
=
1
1 +t
where we assumed n = 1.
incremental plastic strain rate
Cpn() = ()
_
Cn()
1
n()
Cpn()
_
;
1
n
=
1
3tr
_
Ben
_
sub-incremental deformation : j = 1 ns + 1
j 1 = 0 : = tn ; j = ns + 1 : = tn+1
t = t/ns ; Cn =
_
Cn
_1/ns
; C
j
n =
_
Cn
_j
284
plastic strain
1
t
_
C
j
pn C
j1
pn

=
j
_
C
j
n
1

j
n
C
j
pn
_

C
j
pn =

j
n t j

j
n +t j
C
j
n +

j
n

j
n +t j
C
j1
pn
parameter :
j
=

j
n

j
n +t j
C
j
pn =
j
n(1
j
) C
j
n +
j
C
j1
pn
incremental plastic strain Cpn = Cpn(tn+1) = C
ns+1
pn
total isochoric elastic strain Be = Rn
Ben R
c
n = Fn C
1
pn F
c
n
Scalar variable update
The current plastic strain depends on two scalar variables : the elasticity parameter and the
softening parameter D. These are a function of the stress , which implies that the integration
has to be carried out iteratively. A Newton-Raphson iterative procedure is employed and the
resulting equation system involves partial derivatives of and D, which can be calculated
rather straightforwardly.
f

= 1 +t +t

= 1 +t t
G
2

= 1 +t t
G
2
_

_
1


1

30
__

f
D
= t

D
= t
G
2

D
= t
G
2
() = t
G
2

g

= t
D

= t
D

= t
_
D

30
_

g
D
= 1 t
D
D
= 1 t
_
D
h

6D
_
After convergence of the iterative process the (sub)incremental plastic strain and stress is
known, but beware that these are only approximations for the real end-increment values. The
update procedure is part of the iterative procedure which has to be repeated until convergence
is reached.
= 1/(1 +t) f(, D) = (1 +t) = 1
1
t {D D(tn)} = D g(, D) = D t D = D(tn)
285
Newton-Raphson iterative solution procedure
_

_
f

f
D
g

g
D
_

_
_

D
_

_
=
_

_
1 f
D(tn) g
_

_
=
_

_
r

r
D
_

_
partial derivatives
f

= 1 +t +t

= 1 +t t
G
2
_

_
1


1

30
__

f
D
= t

D
= t
G
2
= t
g

= t
D

= t
_
D

30
_

g
D
= 1 t
D
D
= 1 t
_
D
h

6D
_
Matrix/column notation
The tensors and vectors in the presented mathematics can be written in components w.r.t.
a vector basis. The components are stored in columns and matrices and the tensor formu-
lations are transferred into matrix/column formulations which can be implemented rather
straightforwardly in a computer code.
J = det(F) F = J
1/3
F B = F F
T
w = H B
d
p = (J 1) s
h
= pI
= 1/(1 +t)
1
t {D D(tn)} = D
_
_
_
, D
Cpn = (1 ) Cn +Cpn(tn)
Ben = Un C
1
pn
U
T
n
s
d
= G
Ben s =
_
3
2tr(sd sd)
= ( s, p, T, D)
_
_
_

Ben Be = Rn
Ben R
T
n s
d
= G B
d
e
= s
d
+s
h
+w
286
12.5.6 Stiness
The stress is related to the elastic isochoric strain Be, the volume change J and the total
isochoric strain B. Each of the three quantities will be considered separately and relations
between their variations and F will be derived.
The consistent material stiness tensor relates the iterative change of the Cauchy stress
tensor to the iterative displacement u. In the derivation of this relation it is assumed
that approximate end-increment values of all relevant variables are known. 4Sd, 4Sh and
4H are properly dened fourth-order tensors.
stress = s
d
+s
h
+w = G B
d
e +I(J 1) +H B
d
strains and deformation
Be = F C
1
p F
c
; Cp = (1 ) C +Cp(tn) ; F = J
1/3
F
stress variation
= s
d
+s
h
+w
= G B
d
e +IJ +H B
d
=
_
4
Sd +
4
Sh +
4
H
_
: F
=
4
S : F =
4
S
rc
: F
c
with F
c
= 0u = F
c
u = F
c
L
c
u
=
4
S
rc
: (F
c
L
c
u)
=
4
M : L
c
u
Elastic strain variation
The elastic strain Be must be calculated from the total deformation F and the plastic strain
Cp. Its variation is related to F and Cp which will be considered separately.
Be = F C
1
p F
c
Be = F C
1
p F
c
F C
1
p Cp C
1
p F
c
+ F C
1
p F
c
=
_
F C
c
p F
c
_c
F C
1
p
_
F C
c
p C
c
p
_c
+ F C
1
p F
c
=
_
M
(1)
F
c
_c
M
(2)

_
M
(1)
C
c
p
_c
+M
(2)
F
c
B

e
=
_
M
(1)
cr
+M
(2)
c
_
F

M
(2)
c
M
(1)
c
C

p
with M
(1)
= F C
T
p ; M
(2)
= F C
1
p
= A
(1)
F

+A
(2)
C

p
B
d
e = Be
1
3 tr( Be)I =
_
4
I
1
3II
_
: Be
B
d
e =
_
4
I 1
3II
_
: Be
B

d
e
=
_
I
1
3I

T
t
_
B

e
287
Plastic strain variation
The variation of the plastic strain Cp is related to F (via C) and . These variations will
be considered separately.
Cp = (1 ) C +Cp(tn)
Cp = (1 ) C +
_
Cp(tn) C
_
= (1 ) C +
_
Cp(tn) C
_

= (1 )
_
F
c
F + F
c
F
_
+
_
Cp(tn) C
_

= (1 )
__
F
c
F
_c
+ F
c
F
_
+
_
Cp(tn) C
_

p
= (1 )
_
F
tr
F

+ F
t
F

_
+
_
C

p
(tn) C

=
_
(1 )
_
F
tr
+ F
t
__
F

+
_
C

p
(tn) C

= C
(1)
F

+C

(2)

Deformation tensor variation


The variation of the isochoric deformation tensor F can be expressed in the variation of the
total deformation tensor F. The volume ratio J is assumed to be constant in this variation.
F = J
1/3
F
F = 1
6J
1/3
FI :
_
F F
1
+F
c
F
c
_
+J
1/3
F
=
1
3J
1/3
F
_
F
c
: F
c
_
+J
1/3
F
F

= 1
3J
1/3
F

_
F

1
_T
t
F

+J
1/3
F

=
_
1
3J
1/3
F

_
F

1
_T
t
+J
1/3
I
_
F

= F F

Elasticity scalar variation


The variation of the elasticity parameter can be expressed in Be and F, starting from
=
1
1 +t
=

+Gt
=
t
Gt +

The variation can be written as :
=


+

p
p +

D
D

+

D
D
p
p
=
3G2
2
B
d
e : B
d
e =
3G2
2
B
d
e : Be
p = Jtr(F) = JI : F
=
3G2
2
_


+

D
D

_
B
d
e : Be J
_

p
+

D
D
p
_
I : F
= h1
B
d
e : Be +h2I : F
288
A number of partial derivatives must be calculated to determine h1 and h2.
=
t
Gt +
= l1
B
d
e : Be +l2I : F
= l1
_
B

d
e
_T
t
B

e
+l2I

T
t
F

scalar parameters
l1 =
t h1
t G+
; l2 =
l1h2
h1
h1 =
3G2
2
_


+

D
D

_
; h2 = J
_

p
+

D
D
p
_
partial derivatives


=
_
1


1

30
_
;

p
=

0
;

D
=
D

=
t
D

1 t
D
D
;
D
p
=
t
D
p
1 t
D
D
D

=
D

30
;
D
p
=
D
0
;
D
D
= D
h

6 D
with D = h
_
1
D
D
_

6
Deviatoric stress variation
The variation of the deviatoric stress tensor is related to B
d
e and subsequently to F :
s
d
= G B
d
e =
4
Sd : F
The components are stored in the column s

d.
B

e
= A
(1)
F

+A
(2)
C

p
=
_
A
(1)
+A
(2)
C
(1)
_
F

+A
(2)
C

(2)
= B
(1)
F

+B

(2)

= B
(1)
FF

+l1B

(2)
_
B

d
e
_T
t
B

e
+l2B

(2)
I

T
t
F

e
=
_
I l1B

(2)
_
B

d
e
_T
t
_1 _
B
(1)
F +l2B

(2)
I

T
t
_
F

d
e
=
_
I
1
3 I

T
t
_
B

e
=
_
I
1
3I

T
t
__
I l1B

(2)
_
B

d
e
_T
t
_1 _
B
(1)
F +l2B

(2)
I

T
t
_
F

= B
(3)
F

deviatoric stress components


s

d
= G B

d
e
= GB
(3)
F

= S
d
F

289
Hydrostatic stress variation
The variation of the hydrostatic stress sh is related to the variation of the volume factor J.
The latter can be related to the variation of F, resulting in a relation between sh and F :
s
h
= IJ =
4
Sh : F
The components are stored in the column s

h.
J = J tr(D) = J
1
2 tr
_
F F
1
+
_
F F
1
_c_

J =
1
2J tr
_
F F
1
+F
c
F
c
_
=
1
2J
_
F
c
: F
c
_
+
1
2J
_
F
c
: F
c
_
= J F
c
: F
c
= J F
1
: F
= J
_
F

1
_T
t
F

h
= I

J = J I

_
F

1
_T
t
F

= S
h
F

Hardening stress variation


The hardening stress w is related to the deviatoric total volume invariant strain B
d
. The
variation w can be related to F.
w = H B
d
=
4
H : F
The components are stored in the column w

.
B = F F
c
B = F F
c
+ F F
c
B
d
= B
1
3 tr( B)I =
_
4
I
1
3II
_
: B
B
d
=
_
4
I
1
3II
_
:
__
F F
c
_c
+ F F
c
_
B

d
=
_
I
c

1
3 I

T
t
__
F
cr
F

+ F
c
F

_
=
_
I
c

1
3I

T
t
__
F
cr
+ F
c
_
FF

=
_
F
cr
+ F
c

2
3I

T
t
F
c
_
FF

= B
(4)
F

hardening stress components


w

= H B

d
= H B
(4)
F

= HF

Consistent material stiness tensor


The variation of the Cauchy stress is related to the variation of the deformation tensor
F. In the iterative weighted residual equation must be related to the gradient of the
iterative displacement Lu = ( u)c = ( x)c. The resulting fourth-order tensor 4M is the
consistent material stiness tensor.
The components of (in column

) and Lu (in column L

u
) are related by the consistent
stiness matrix M.
290
= s
d
+s
h
+w
=
_
4
Sd +
4
Sh +
4
H
_
: F =
4
S : F =
4
S
rc
: F
c
with F
c
= 0u = F
c
u = F
c
L
c
u
=
4
S
rc
: (F
c
L
c
u) =
4
M : L
c
u

= s

d
+s

h
+w

=
_
S
d
+S
h
+H
_
F

= SF

= S
c
F

t
with F

t
= F
t
_
L

u
_
t
= S
c
F
t
_
L

u
_
t
= M
_
L

u
_
t
12.5.7 Examples
Tensile test
A square plate or cylindrical bar is loaded uniaxially. Dimensions are listed in the table.
initial width w0 100 mm
initial height h0 100 mm
initial thickness d0 0.1 mm
initial radius r0
_
(10/) mm
initial height h0 100 mm
Viscoelastic model in tensile test
The axial elongation is prescribed with a constant elongation rate. The axial stress and
force are calculated for polycarbonate (PC). Parameter values are listed in the table. The
deformation is assumed to be plane strain.
291
E 2305 MPa 0.37 -
H 29 MPa h 270 -
D 19 - A0 9.7573E-27 s
H 2.9E5 J/mol 0.06984 -
0 0.72 MPa
elongation rate
l
h0
= 0.01 s
1
1 1.1 1.2 1.3 1.4 1.5
0
10
20
30
40
50
60
70
80
90

[M
P
a]
1 1.1 1.2 1.3 1.4 1.5
0
100
200
300
400
500
600
700
800

F [N
]
Fig. 12.26 : Axial stress and force versus elongation
Shear test
The simple shear test is analyzed with one element, where the horizontal displacement in the
upper nodes is prescribed. Because there are no unknown degrees of freedom, the stiness
matrix is not used. Only strains, stresses and reaction forces are calculated.
initial width w0 100 mm
initial height h0 100 mm
initial thickness d0 0.1 mm
Viscoelastic model in shear test
The shear force is calculated for polycarbonate (PC). The prescribed strain rate is constant.
292
strain rate =
u
h0
= 0.01 s1
1 2
3 4
0 0.2 0.4 0.6 0.8 1
0
100
200
300
400
500
600

F x [N
]
Fig. 12.27 : Shear force versus shear strain for plane strain
Bibliography
[1] Bathe, K.-J. Finite Element Procedures. Prentice Hall, New Jersey, 1996.
[2] Boyle, J.T.; Spence, J. Stress analysis for creep. Butterwort, 1983, pp 283.
[3] Criseld, M. Non-linear Finite Element Analysis of Solids and Structures, Vol. 1: Es-
sentials. John Wiley and Sons Ltd., West Sussex, England...
[4] Criseld, M. Non-linear Finite Element Analysis of Solids and Structures, Vol. 2: Ad-
vanced Topics. John Wiley and Sons Ltd., West Sussex, England...
[5] Gordon, J.E. The new science of strong materials. Penguin Books, 1976.
[6] Gordon, J.E. Structures. Penguin Books, 1978.
[7] Hughes, T. Numerical implementation of constitutive models: rate-independent devia-
toric plasticity. In: Theoretical Foundation for Large-scale Computations for Nonlinear
Material Behaviour, Ed: Nemat-Nasser, R. Asaro, G. Hegemier Martinus Nijho Pub-
lishers, Dordrecht, The Netherlands.., pp 29-57.
[8] Hunter, S.C. Mechanics of continuous media, 2nd edition. Ellis Horwood Limited, 1983.
[9] Simo, J.C.; Hughes, T. Computational Inelasticity. Interdisciplinary Applied Mathemat-
ics. Springer-Verlag, New York, 1998.
[10] Skrzypek, J. Plasticity and Creep - Theory, Examples and Problems. CRC Press, Inc.,
Florida, USA, 1993.
[11] Timoshenko, Stephen P. History of strength of materials: with a brief account of the
history of elasticity and theory of structures. London : McGraww-Hill, 1953, pp 452.
[12] Tschoegl, N.W. The phenomenological theory of linear viscoelastic behaviour; An intro-
duction. Springer-Verlag, 1989.
[13] Zienkiewicz, O.; Taylor, R. The Finite Element Method, Vol. 1, Basic Formulation and
Linear Problems. McGraw-Hill, London, UK, 1989.
[14] Zienkiewicz, O.; Taylor, R. The Finite Element Method, Vol. 2, Solid and Fluid Mechan-
ics, Dynamics and Non-linearity. McGraw-Hill, London, UK, 1989.
APPENDICES
Appendix A
FE program tr2dL
The Matlab program tr2dL allows the modelling and analysis two-dimensional truss struc-
tures, where trusses are homogeneous and linear elastic. Deformation and rotations must be
small, i.e. the behavior is geometrically linear.
Model geometry, topology (connectivity), geometrical and material parameters, bound-
ary conditions (prescribed displacements and point loads) and link relations (dependecies
between degrees of freedom) must be available as input data.
When the analysis is nished, output date are available in the data base and various
other data arrays.
In the following section an example input is presented, with explanatory comments.
Finally the program source is listed and explained in more detail with included comment.
A.1 Example input le
As an example, the two-bar truss structure, shown in the gure below, will be modelled,
loaded and analyzed.
1
2
3
F
y
x 1
2
Both trusses have dierent geometrical and material properties, which are given in the table
below.
a1
a2
truss 1 2
cross-sectional area A 10 20 [mm2]
Youngs modulus E 200 150 [GPa]
Poissons ratio 0.3 0.3 [-]
Now let us see which Matlab commands do the job. Before starting, it might be wise to close
all gures and clear the Matlab work space.
close all; clear all;
First we give the coordinates of the nodes in the array crd0. Now we have to decide on the
units and in this example we choose to model everything in mm.
crd0 = [ 0 0; 100 0; 0 100/sqrt(3) ];
The connectivity of the elements is dened in the array lok. This array has a row for each
element. The rst column contains the element type, which is 9 for a truss. The second
row is the element group number. Typically elements with the same properties are placed
in one and the same group. Because our two elements indeed have dierent properties, they
are placed in two dierent groups. The third and fourth column contain the rst and second
node of the element.
lok = [ 9 1 1 2 ; 9 2 2 3 ];
The geometrical and material properties are provided in the array elda (element data). For
each element group we have a row in elda. The rst column contains a zero (0), which is not
important for our use. The second column contains the material identication number. For
linear elastic material, which we will use here, this number is 11 (eleven). The third column
contains the cross-sectional area (in mm2). The fourth and fth column are not used for
our problems and always contain a zero (0). The sixth and seventh column contain Youngs
modulus and Poissons ratio. So for our example we have :
elda = [
0 11 10 0 0 200000 0.3
0 11 20 0 0 150000 0.3
];
Boundary conditions are prescribed nodal displacements and/or prescribed nodal forces. Pre-
scribed nodal displacements are always needed to prevent rigid body motions.
Prescribed displacements are provided in the array pp. For each prescribed displace-
ment component we have one row. The rst column contains the node, the second column
contains the direction (either 1 (= x = horizontal) or 2 (= y = vertical). The third column
contains the value.
For our example we have :
pp = [ 1 1 0; 1 2 0; 3 1 0; 3 2 0 ];
The prescribed forces are given in the array pf. Again each prescribed force component is
placed on a row, with the node in the rst, the direction in the second and the value in the
third column.
For our example :
a3
pf = [ 2 2 -100 ];
That is about all. The input is complete and the program can be executed to analyze the
behavior.
tr2dL;
When the analysis is completed successfully, we want to see some results. First the nodal
data, i.e. displacements and reaction forces. They are available in the arrays Mp and
M. Rows contain nodal data : displacement and forces in the rst (1 = x = horizontal)
and second (2 = y = vertical) directions. Just type the next commands in the Matlab shell.
Mp
Mfi
Element data, like stress and strain, are available in the data base. eldaC. For element e
we nd the date in row e of eldaC. Relevant date can be found at the following locations :
eldaC(e,1) = sine of angle between axis and 1-direction
eldaC(e,2) = cosine of angle between axis and 1-direction
eldaC(e,3) = length
eldaC(e,4) = cross-sectional area
eldaC(e,6) = linear axial strain
eldaC(e,7) = axial stress
eldaC(e,11) = axial stretch ratio
eldaC(e,12) = radial stretch ratio
eldaC(e,18) = axial force
To see some results for our example just type the following in the Matlab shell :
eldaC(1,6)
eldaC(1,7)
eldaC(1,18)
eldaC(2,6)
eldaC(2,7)
eldaC(2,18)
These values can ofcourse be stored and printed in various ways.
The above input commands can also be put in one single input le. The results are
shown as plot of he deformed structure, where the deformation is enlarged.
a4
1
2
3
tr2dL2bardef
A.2 The program tr2dL
The program tr2dL is listed below and is seeded with comments to explain variables and
actions.
%**********************************************************************
% tr2dL : 2-dimensional linear truss element
%======================================================================
%----------------------------------------------------------------------
% Calculate some parameters from the input data.
% nndof : number of nodal degrees of freedom
% nnod : number of nodes
% ndof : number of system degrees of freedom
% ne : number of elements
% nenod : number of element nodes
% nedof : number of element degrees of freedom
% negr : number of element groups
% lokvg : location of degrees of freedom of elements in structure
%----------------------------------------------------------------------
nndof = 2;
nnod = size(crd0,1);
ndof = nnod * nndof;
ne = size(lok,1);
nenod = size(lok,2)-2;
nedof = nndof * nenod;
negr = size(elda,1);
lokvg(1:ne,:) = ...
[ nndof*(lok(1:ne,3)-1)+1 nndof*(lok(1:ne,3)-1)+2 ...
nndof*(lok(1:ne,4)-1)+1 nndof*(lok(1:ne,4)-1)+2 ];
%----------------------------------------------------------------------
% Calculate transformation matrix Trm for local dofs, if needed.
%----------------------------------------------------------------------
Trm = eye(ndof);
if exist(tr),
ntr = size(tr,1);
for itr=1:ntr
a5
trp = round(tr(itr,1)); tra = tr(itr,2);
trc = cos((pi/180)*tra); trs = sin((pi/180)*tra);
k1 = nndof*(trp-1)+1; k2 = nndof*(trp-1)+2;
trm = [trc -trs ; trs trc];
Trm([k1 k2],[k1 k2]) = trm;
end;
else, ntr = 0; tr = []; end;
%----------------------------------------------------------------------
% Initialization of databases
% elpa : element parameters
% elda0 : initiel values
% eldaC : current values
% ety : element type ; egr : element group ;
% mnr : material number ; mcl : material class ; mty : material type
% l0 : initial element length
% s0 : sine of axis angle
% c0 : cosine of axis angle
%----------------------------------------------------------------------
for e=1:ne
ety = lok(e,1); egr = lok(e,2);
mnr = elda(egr,2); A0 = elda(egr,3);
E = elda(egr,6); Gn = elda(egr,7);
mcl = floor(mnr/10); mty = rem(mnr,10);
k1 = lok(e,3); k2 = lok(e,4);
x10 = crd0(k1,1); y10 = crd0(k1,2); x20 = crd0(k2,1); y20 = crd0(k2,2);
l0 = sqrt((x20-x10)*(x20-x10)+(y20-y10)*(y20-y10));
s0 = (y20-y10)/l0;
c0 = (x20-x10)/l0;
elpa(e,:) = [ety egr nenod nndof nedof];
elda0(e,:) = [ s0 c0 l0 A0 0 0 Gn E mcl mty ];
eldaC(e,:) = [ s0 c0 l0 A0 0 0 Gn E 0 0 ];
end; % element loop e
%----------------------------------------------------------------------
% Boundary conditions are reorganized.
% Additional arrays for later partitioning are made.
% npdof : number of prescribed degrees of freedom
% npfor : number of prescribed nodal forces
% nudof : number of unknown degrees of freedom
%
% Information for partitioning the system of equations associated
% with prescribed boundary conditions is made available in the arrays
% ppc, ppv, pfc and pfv.
%----------------------------------------------------------------------
if ~exist(pp), pp = []; ppc = []; ppv = []; end;
if ~exist(pf), pf = []; pfc = []; pfv = []; end;
npdof = size(pp,1);
npfor = size(pf,1);
nudof = ndof - npdof;
if npdof>0
ppc = [nndof*(round(pp(:,1))-1)+round(pp(:,2))];
a6
ppv = pp(:,nndof+1);
end;
if npfor>0,
pfc = [nndof*(round(pf(:,1))-1)+round(pf(:,2))];
pfv = pf(:,nndof+1);
end;
% Information for partitioning the system of equations associated
% with linked degrees of freedom is made available in the arrays
% plc and prc.
if ~exist(pl), pl = []; plc = []; end;
if ~exist(pr), pr = []; prc = []; end;
if ~exist(lim), lim = []; end;
npl = size(pl,1);
npr = size(pr,1);
if ~exist(lif), lif = zeros(1,npl); end;
if npl>0
plc = [nndof*(round(pl(:,1))-1)+round(pl(:,2))];
prc = [nndof*(round(pr(:,1))-1)+round(pr(:,2))];
end;
% Some extra arrays are made for later use.
pa = 1:ndof; pu = 1:ndof; prs = 1:ndof;
pu([ppc plc]) = [];
prs([ppc pfc plc]) = [];
% pe0 : column with prescribed initial displacements
% fe0 : array with prescribed initial forces
pe0 = zeros(ndof,1); pe0(ppc(1:npdof)) = ppv(1:npdof);
fe0 = zeros(ndof,1); fe0(pfc(1:npfor)) = pfv(1:npfor);
%----------------------------------------------------------------------
% Initialization to zero
% pe : column with nodal displacements
% p : column with nodal displacements
% fe : column with external (applied) nodal forces
% fi : column with internal (resulting) nodal forces
% #T : column with transformed components
%----------------------------------------------------------------------
pe = zeros(ndof,1); p = zeros(ndof,1); pT = zeros(ndof,1);
fe = zeros(ndof,1); fi = zeros(ndof,1);
feT = zeros(ndof,1); fiT = zeros(ndof,1);
%----------------------------------------------------------------------
% Loop over all elements to generate element stiffness matrix em
% Assemble em into structural stiffness matrix sm
% ec0 : initial coordinates of element nodes
% ec : current coordinates of element nodes
%----------------------------------------------------------------------
sm = zeros(ndof);
a7
for e=1:ne
ety = elpa(e,1); egr = elpa(e,2);
nenod = elpa(e,3); nedof = elpa(e,5);
ec0 = crd0(lok(e,3:2+nenod),:); ec = ec0;
em = zeros(nedof);
%----------------------------------------------------------------------
% Element stiffness matrix
s = eldaC(e,1) ; c = eldaC(e,2);
ML = [ c*c c*s -c*c -c*s ; c*s s*s -c*s -s*s
-c*c -c*s c*c c*s ; -c*s -s*s c*s s*s ];
%----------------------------------------------------------------------
l0 = elda0(e,3); A0 = elda0(e,4); E0 = elda0(e,8);
em = (A0/l0 * E0) * ML ;
sm(lokvg(e,1:nedof),lokvg(e,1:nedof)) = ...
sm(lokvg(e,1:nedof),lokvg(e,1:nedof)) + em;
end; % element loop e
%----------------------------------------------------------------------
% Transformation for local nodal coordinate systems
%----------------------------------------------------------------------
sm = Trm * sm * Trm;
%----------------------------------------------------------------------
% Boundary conditions and links
%----------------------------------------------------------------------
pe = pe0; fe = fe0; rs = fe;
if npl>0, rs = rs - sm(:,plc)*lif; end;
%----------------------------------------------------------------------
% Partitioning is done in the function fbibpartit.m
[sm,rs] = fbibpartit(1,sm,rs,ndof,pa,ppc,plc,prc,pe,lim);
%----------------------------------------------------------------------
% Solving the system of equations and take prescribed displacements
% and links into account.
% Update nodal point coordinates crd.
sol = inv(sm)*rs; % sol = sm\rs;
pe(pu) = sol;
if npl>0, pe(plc) = lim*pe(prc) + lif; end;
p = pe; pT = Trm * p;
crd = crd0 + reshape(pT,nndof,nnod);
%----------------------------------------------------------------------
% Calculate stresses and strains and the internal forces ef.
% Internal forces ef are assembled into fi, the structural
% internal forces, representing the reaction forces.
%----------------------------------------------------------------------
a8
fi = zeros(ndof,1);
for e=1:ne
ety = elpa(e,1); egr = elpa(e,2);
ec0 = crd0(lok(e,2+1:2+nenod),:);
ec = crd(lok(e,2+1:2+nenod),:);
ef = zeros(nedof,1);
%----------------------------------------------------------------------
% Element internal forces
s = eldaC(e,1) ; c = eldaC(e,2);
V = [ -c -s c s ];
%----------------------------------------------------------------------
l0 = elda0(e,3); A0 = elda0(e,4);
E0 = elda0(e,8); Gn0 = elda0(e,7);
x1 = ec(1,1); y1 = ec(1,2); x2 = ec(2,1); y2 = ec(2,2);
l = sqrt((x2-x1)*(x2-x1)+(y2-y1)*(y2-y1));
s = (y2-y1)/l; c = (x2-x1)/l;
Gl = l/l0; Ge = Gl-1;
Ged = -Gn0*Ge; Gm = Ged+1; A = Gm*Gm*A0; Gs = E0 * Ge; N = A * Gs;
eldaC(e,1:7) = [s c l A 0 Ge Gs];
eldaC(e,11:18) = [Gl Gm Ge 0 0 Gs 0 N];
ef = N * V;
fi(lokvg(e,1:nedof)) = fi(lokvg(e,1:nedof)) + ef;
end;
rs = fe - fi;
fi = Trm * fi; fiT = fi; fiT = Trm * fi; rsT = feT - fiT;
%----------------------------------------------------------------------
% Reshaping columns into matrices
Mp = reshape(p,nndof,nnod);
Mfi = reshape(fi,nndof,nnod); Mfe = reshape(fe,nndof,nnod);
Mrs = reshape(rs,nndof,nnod);
if ntr>=1
MpT = reshape(pT,nndof,nnod);
MfiT = reshape(fiT,nndof,nnod); MfeT = reshape(feT,nndof,nnod);
MrsT = reshape(rsT,nndof,nnod);
end;
%----------------------------------------------------------------------
%**********************************************************************
Appendix B
FE program tr2d
The Matlab program tr2d allows to model and analyze two-dimensional truss structures,
where trusses are homogeneous and can behave nonlinear. Deformation and rotations can be
large, i.e. the behavior is geometrically nonlinear.
Model geometry, topology (connectivity), geometrical and material parameters, bound-
ary conditions (prescribed displacements and point loads) and link relations (dependecies
between degrees of freedom) must be available as input data. Also the history of the pre-
scribed boundary conditions must be specied.
When the analysis is nished, output date are stored in the data base and various other
data arrays.
In the following section an example input is presented, with explanatory comments.
Finally the program itself is explained in more detail.
B.1 Example input le
As an example, the two-bar truss structure, shown in the gure below will be modelled, loaded
and analyzed.
1
2
3
F
y
x 1
2
Both trusses have dierent geometrical and material properties, The material of truss 1 be-
haves elastcally, according to a linear relation between the axial stress and the linear strain
a9
a10
= l = 1, specied by the modulus E and the Poissons ratio . Truss 2 behaves linearly
elastic upto the yield stress y0 after which isotropic linear hardening occurs with hardening
parameter H = E/20. The vertical force in node 4 is increased from 0 to -50000 N and
decreases back to zero again.
The Matlab workspace is cleared and gures are closed. The le loadincr.m, which de-
scribes the time history of the boundary conditions, is deleted. The le savedata.m, which
describes, which calculated values have to be saved for postprocessing, is also deleted.
clear all; close all;
delete(loadincr.m); delete(savedata.m);
Coordinates of the nodal points and the connectivity is given in the matrices crd0 and
lok. In the latter array, the rst column contains the element type and the second column
the element group.
crd0 = [ 0 0; 100 0; 0 100/sqrt(3) ];
lok = [ 9 1 1 2 ; 9 2 2 3 ];
element data are given in the array elda, which has one row for each property group. The
second column contains the material number, which has two digits. The rst digit is the
material class (mcl) and the second digit is the material type (mty). The next classes
and types are currently implemented :
mcl = 1 : elastic material
mcl = 2 : elastomeric material
mcl = 3 : elastoplastic material
mcl = 4 : linear viscoelastic material
mcl = 5 : viscoplastic material (Perzyna)
mcl = 6 : nonlinear viscoelastic material (Leonov)
mcl = 7 : elastoviscous material (creep)
For the element type you should look in the Matlab source le.
elda = [ 0 11 10 0 0 200000 0.3 0 0 0 ;
0 31 20 0 0 200000 0.3 250 10000 0 ];
We also have to indicate which hardening law (hm) is used and which stress update proce-
dure (pr, explicit or implicit).
hm = li; pr = ex;
The boundary conditions are prescribed, rst the incremental displacements, then the incre-
mental forces.
pp = [ 1 1 0; 1 2 0; 3 1 0; 3 2 0; ];
pf = [ 2 2 -100 ];
The load history is prescribed with a call to the function le mloin.m. In its source le, it
is explained how it must be used.
a11
[St,Sft,nic,GDt,tend] = mloin(0,0,400,1,pol,[0 0 200 50 400 0]);
Analysis data for postprocessing must be saved. This is done by making a le savedata.m,
which is called by the program tr2d at the end of each increment.
sada=fopen(savedata.m,w);
fprintf(sada,Sf2x(ic)=Mfi(2,1);Sf2y(ic)=Mfi(2,2); \n);
fprintf(sada,Su2x(ic)=MTp(2,1);Su2y(ic)=MTp(2,2); \n);
fprintf(sada,SGe1(ic)=eldaB(1,6);SGe2(ic)=eldaB(2,6); \n);
fprintf(sada,SGs1(ic)=eldaB(1,7);SGs2(ic)=eldaB(2,7); \n);
fclose(sada);
Now the input is complete and the program can be called for the analysis.
tr2d;
The result is shown in the gures below.
0.01 0 0.01 0.02 0.03 0.04
1000
800
600
400
200
0
200
400
600
tr2d2barse

[M
P
a]


elem 1
elem 2
0 100 200 300 400
8
7
6
5
4
3
2
1
0
tr2d2bartu
ic
disp [m
m
]


u2x
u2y
Element stress-strain relations and displacements of node 4
B.2 The program tr2d
The program tr2d is listed below. It is clear that the program calls a collection of com-
mand and function les, which can and should be inspected for through understanding of the
procedure. The program structure is however clearly shown in the listing.
%**********************************************************************
% tr2d : 2-dimensional truss element
%======================================================================
tr2dchkinp; % tr2dchkinp.m
fbiblcase; % fbiblcase.m
[Trm] = fbibtransbc(tr,ndof,nndof); % fbibtransbc.m
[elda0,eldaB,eldaC,mcl4] = tr2dinidat(ne,elgr,elda,neip,ts,lok,crd0,mm);
% tr2dinidat.m
tr2dinizer; % tr2dinizer.m
save([matf num2str(0)]);
crdB = crd0; crd = crd0;
a12
%======================================================================
% Incremental calculation
%======================================================================
ic = 1; ti = 0; it = 0; slow = 1;
while ic<=nic
%----------------------------------------------------------------------
fbibcutback; % fbibcutback.m
ti = ti + ts; it = 0; loadincr; % loadincr.m
pe = peC./slow; fe = feC;
rs = fe - fi;
Dp = zeros(ndof,1); Ip = zeros(ndof,1); IpT = zeros(ndof,1);
%======================================================================
% System matrix is assembled from element matrices
% System matrix is transformed for local nodal coord.sys.
%======================================================================
if (ic==1 | nl==1)
%----------------------------------------------------------------------
sm=zeros(ndof);
for e=1:ne
ec = crd(lok(e,3:nenod+2),:);
[ML,MN,V,eldaC] = tr2dgeom(e,ec,eldaC); % tr2dgeom.m
tr2dmat; % tr2dmat.m
em = CL * ML + CN * MN ; % element stiffness matrix
ef = (CI-CV) * V; % element internal load column
sm(lokvg(e,:),lokvg(e,:)) = sm(lokvg(e,:),lokvg(e,:)) + em;
end;
sm = Trm*sm *Trm; sm0 = sm;
%----------------------------------------------------------------------
end;
%======================================================================
% Iterative calculation
%======================================================================
nrm = 1000; it = 1;
while (nrm>ccr) & (it<=mit)
%----------------------------------------------------------------------
%======================================================================
% Links and boundary conditions are taken into account
% Unknown nodal point values are solved
% Prescribed nodal values are inserted in the solution vector
%======================================================================
%sm = sm0; % only used to test modified Newton-Raphson
if npl>0, rs = rs - sm(:,plc)*lif; end;
[sm,rs] = fbibpartit(it,sm,rs,ndof,pa,ppc,plc,prc,pe,lim);% fbibpartit.m
sol = inv(sm)*rs;
p = zeros(ndof,1); p(pu) = sol;
if it==1, p(ppc) = pe(ppc); end;
if npl>0, p(plc) = lim*p(prc) + lif; end;
Dp = p; Ip = Ip + Dp; Tp = Tp + Dp;
a13
%======================================================================
% Transformation dofs from local to global nodal coordinate systems
%======================================================================
DpT = Trm * Dp; IpT = IpT + DpT; TpT = TpT + DpT;
crd = crd0 + reshape(TpT,nndof,nnod);
%======================================================================
% Calculate stresses and strains.
% Make system matrix and internal force vector for next step.
%======================================================================
sm=zeros(ndof); fi=zeros(ndof,1); mcl4=0; mcl42=0; mcl9=0;
for e=1:ne
ec = crd(lok(e,3:nenod+2),:); % element nodal coordinates
[ML,MN,V,eldaC] = tr2dgeom(e,ec,eldaC); % tr2dgeom.m
tr2dmat; % tr2dmat.m
em = CL * ML + CN * MN ; % element stiffness matrix
ef = (CI-CV) * V; % element internal load column
sm(lokvg(e,:),lokvg(e,:)) = sm(lokvg(e,:),lokvg(e,:)) + em;
fi(lokvg(e,:)) = fi(lokvg(e,:)) + ef;
end;
sm=Trm*sm*Trm; fi=Trm*fi;
%======================================================================
% Calculate residual force and convergence norm
%======================================================================
rs = fe - fi;
nrm = fbibcnvnrm(cnm,pu,ppc,prs,Dp,Ip,rs,fi); % fbibcnvnrm.m
it = it + 1; % increment the iteration step counter
fbibwr2scr; % fbibwr2scr.m
%----------------------------------------------------------------------
end; %it
%======================================================================
% Transformation nodal forces from local to global nodal coord.sys.
%======================================================================
fiT = fi; feT = fe; rsT = rs;
fiT = Trm * fi; feT = Trm * fe; rsT = Trm * rs;
fbibcol2mat1; % fbibcol2mat1.m
crdB = crd; feB = fe; eldaB = eldaC; HGsB = HGsC;
savefile = [matf num2str(ic)]; savedata; % savedata.m
ic = ic + 1; % increment the increment counter
save([matf 00],ic); % save date to matf
%----------------------------------------------------------------------
end; %ic
%**********************************************************************
a14
Appendix C
Stiness and compliance matrices
In this appendix, the stiness and compliance matrices for orthotropic, transversal isotropic
and isotropic material are given.
C.1 Orthotropic
For an orthotropic material 9 material parameters are needed to characterize its mechanical
behavior. Their names and formal denitions are :
Youngs moduli : Ei =
ii
ii
Poissons ratios : ij =
jj
ii
shear moduli : Gij =
ij
ij
The introduction of these parameters is easily accomplished in the compliance matrix S. The
stiness matrix C can then be derived by inversion of S.
Due to the symmetry of the compliance matrix S, the material parameters must obey
the three Maxwell relations.
S =
_

_
E
1
1 21E
1
2 31E
1
3 0 0 0
12E
1
1 E
1
2 32E
1
3 0 0 0
13E
1
1 23E
1
2 E
1
3 0 0 0
0 0 0 G
1
12 0 0
0 0 0 0 G
1
23 0
0 0 0 0 0 G
1
31
_

_
with
12
E1
=
21
E2
;
23
E2
=
32
E3
;
31
E3
=
13
E1
a15
a16
C =
1

_
13223
E2E3
3123+21
E2E3
2132+31
E2E3
0 0 0
1332+12
E1E3
13113
E1E3
1231+32
E1E3
0 0 0
1223+13
E1E2
2113+23
E1E2
11221
E1E2
0 0 0
0 0 0 G12 0 0
0 0 0 0 G23 0
0 0 0 0 0 G31
_

_
with =
1 1221 2332 3113 122331 213213
E1E2E3
C.1.1 Voigt notation
In composite mechanics the so-called Voigt notation is often used, where stress and strain
components are simply numbered 1 to 6. Corresponding components of the compliance (and
stiness) matrix are numbered accordingly. However, there is more to it than that. The
sequence of the shear components is changed. We will not use this changed sequence in the
following.
stresses and strains

T
= [11 22 33 12 23 31] = [1 2 3 6 4 5]

T
= [11 22 33 12 23 31] = [1 2 3 6 4 5]
_

_
1
2
3
4
5
6
_

_
=
_

_
S11 S12 S13 0 0 0
S21 S22 S23 0 0 0
S31 S32 S33 0 0 0
0 0 0 S44 0 0
0 0 0 0 S55 0
0 0 0 0 0 S66
_

_
_

_
1
2
3
4
5
6
_

_
material parameters
S11 =
1
E1
S22 =
1
E2
S33 =
1
E3
S12 =
21
E2
S13 =
31
E3
S23 =
32
E3
S44 = 1
G23
S55 = 1
G31
S66 = 1
G12
C.1.2 Plane strain
For some geometries and loading conditions the strain in one direction is zero. Such defor-
mation is referred to as plane strain. Here we take 33 = 13 = 23 = 0. The stress 33 is not
zero but can be eliminated from the stress-strain relation and expressed in 11 and 22.
For plane strain the stiness matrix can be extracted directly from the three-dimensional
stiness matrix. The inverse of this 3x3 matrix is the plane strain compliance matrix.
a17
C =
1

_
_
13223
E2E3
3123+21
E2E3
0
1332+12
E1E3
13113
E1E3
0
0 0 G12
_
_
with =
1 1221 2332 3113 122331 213213
E1E2E3
S =
_
_
13113
E1

3123+21
E2
0

1332+12
E1
13223
E2
0
0 0
1
G12
_
_
33 =
1

_
1232 +13
E1E2
11 +
2113 +23
E1E2
22
_
= 13
E3
E1
11 +23
E3
E2
22
C.1.3 Plane stress
When deformation in one direction is not restricted, the stress in that direction will be zero.
This is called a plane stress situation. Here we assume 33 = 13 = 31 = 0. The strain 33 is
not zero but can be eliminated from the stress-strain relation and expressed in the in-plane
strains.
Such a plane stress state is often found in the deformation of thin plates, which are
loaded in their plane.
For plane stress the compliance matrix can be extracted directly from the three-dimensional
compliance matrix. The inverse of this 3x3 matrix is the plane strain stiness matrix.
S =
_
_
E
1
1 21E
1
2 0
12E
1
1 E
1
2 0
0 0 G
1
12
_
_
C =
1
1 2112
_
_
E1 21E1 0
12E2 E2 0
0 0 (1 2112)G12
_
_
33 = 13E
1
1 11 23E
1
2 22 =
1
1 1221
{(1223 +13)11 + (2113 +23)22}
C.2 Transversal isotropic
Considering an transversally isotropic material with the (e1e2)-plane isotropic, the Youngs
modulus Ep and the Poissons ratio p in this plane can be measured. The associated shear
modulus is related by Gp =
Ep
2(1 +p)
. In the perpendicular direction we have the Youngs
modulus E3, the shear moduli G3p = Gp3 and two Poisson ratios, which are related by
symmetry : p3E3 = 3pEp.
a18
S =
_

_
E1
p pE1
p 3pE
1
3 0 0 0
pE1
p E1
p 3pE
1
3 0 0 0
p3E1
p p3E1
p E
1
3 0 0 0
0 0 0 G1
p 0 0
0 0 0 0 G
1
p3 0
0 0 0 0 0 G
1
3p
_

_
with
p3
Ep
=
3p
E3
C =
1

_
13pp3
EpE3
3pp3+p
EpE3
p3p+3p
EpE3
0 0 0
p33p+p
EpE3
13pp3
EpE3
p3p+3p
EpE3
0 0 0
pp3+p3
EpEp
pp3+p3
EpEp
1pp
EpEp
0 0 0
0 0 0 Gp 0 0
0 0 0 0 Gp3 0
0 0 0 0 0 G3p
_

_
with =
1 pp p33p 3pp3 pp33p p3pp3
EpEpE3
C.2.1 Plane strain
For the plane strain case with 33 = 13 = 23 = 0. The stress 33 is not zero but can be
eliminated from the stress-strain relation and expressed in 11 and 22. The plane strain
stiness matrix can be extracted directly from the three-dimensional stiness matrix. The
inverse of this 3x3 matrix is the plane strain compliance matrix.
C =
1

_
13pp3
EpE3
3pp3+p
EpE3
0
p33p+p
EpE3
13pp3
EpE3
0
0 0 Gp
_

_
with =
1 pp p33p 3pp3 pp33p p3pp3
EpEpE3
S =
_

_
13pp3
Ep

3pp3+p
Ep
0

p33p+p
Ep
13pp3
Ep
0
0 0
1
Gp
_

_
33 =
1

3p(p + 1)
E2
p
(11 +22)
a19
C.2.2 Plane stress
For the plane stress state with 33 = 13 = 31 = 0, the strain 33 is not zero but can
be eliminated from the stress-strain relation and expressed in the in-plane strains. For plane
stress the compliance matrix can be extracted directly from the three-dimensional compliance
matrix. The inverse of this 3x3 matrix is the plane strain stiness matrix.
S =
_
_
E1
p pE1
p 0
pE1
p E1
p 0
0 0 G1
p
_
_
C =
1
1 pp
_
_
Ep pEp 0
pEp Ep 0
0 0 (1 pp)Gp
_
_
33 =
p3
Ep
(11 +22)
C.3 Isotropic
The linear elastic material behavior can be described with the material stiness matrix C or
the material compliance matrix S. These matrices can be written in terms of the engineering
elasticity parameters E and .
C =
E
(1 +)(1 2)
_

_
1 0 0 0
1 0 0 0
1 0 0 0
0 0 0
1
2 (1 2) 0 0
0 0 0 0
1
2(1 2) 0
0 0 0 0 0
1
2(1 2)
_

_
S =
1
E
_

_
1 0 0 0
1 0 0 0
1 0 0 0
0 0 0 2(1 +) 0 0
0 0 0 0 2(1 +) 0
0 0 0 0 0 2(1 +)
_

_
C.3.1 Plane strain
For some geometries and loading conditions the strain in z-direction is zero : zz = 0. Such
deformation is referred to as plane strain. With xz = yz = 0 we have for the stresses
xz = yz = 0. The stress zz is not zero but can be eliminated from the stress-strain relation
a20
and expressed in xx and yy.
From the stress-strain relation for plane strain it is immediately clear that problems will
occur for = 0.5, which is the value for incompressible material behavior.
_
_
xx
yy
xy
_
_ =
_
_
1 0
1 0
0 0
1
2(1 2)
_
_
_
_
xx
yy
xy
_
_
zz = (xx +yy) = (xx +yy)
with : =
E
(1 +)(1 2)
_
_
xx
yy
xy
_
_ =
1 +
E
_
_
1 0
1 0
0 0 2
_
_
_
_
xx
yy
xy
_
_
C.3.2 Plane stress
When deformation in z-direction is not restricted, the stress zz will be zero. This is called a
plane stress situation. With additional xz = zx = 0, we have xz = zx = 0. The strain zz
is not zero but can be eliminated from the stress-strain relation and expressed in the in-plane
strains.
Such a plane stress state is often found in the deformation of thin plates, which are
loaded in their plane.
_
_
xx
yy
xy
_
_ =
1
E
_
_
1 0
1 0
0 0 2(1 +)
_
_
_
_
xx
yy
xy
_
_
zz =
h
h0
=

E
(xx +yy) =

1
(xx +yy)
_
_
xx
yy
xy
_
_ =
E
1 2
_
_
1 0
1 0
0 0
1
2(1 )
_
_
_
_
xx
yy
xy
_
_
C.3.3 Axi-symmetry
In each point of a cross section the displacement has two components : u

T = [uruz]. The
stress and strain components are :

T
=
_
rr zz tt rz

T
=
_
rr zz tt rz

The stress-strain relation according to Hookes law can be derived from the general three-
dimensional case.
With the well-known strain-displacement relations, the stress components can be related
to the derivatives of the displacement components.
a21
_

_
rr
zz
tt
rz
_

_ =
E
(1 +)(1 2)
_

_
1 0
1 0
1 0
0 0 0
1
2(1 2)
_

_
_

_
rr
zz
tt
rz
_

_
a22
Appendix D
Planar elements
The 4-noded and 8-noded elements are described in the next sections of this appendix.
D.1 Four-node quadrilateral element
In two-dimensional nite element analysis the four-node element is used very much. In the
undeformed and deformed conguration the element sides are straight lines. As its name
indicates, it has four nodal points, which are located in its corners. The numbering of the
nodes is anti-clockwise by convention.
The shape functions, which are used to interpolate global coordinates and displacement
components and weighting function components between their respective nodal values, must
be linear along an element side. In the two-dimensional plane these functions are functions
of the isoparametric coordinates 1 and 2. These functions are not completely linear : they
have a term 12. This implies that their derivatives are not completely constant.
1 2
3
4
1
2
Four-node quadrilateral element

1
=
1
4(1 1)(1 2) ;
2
=
1
4(1 +1)(1 2)

3
= 1
4(1 +1)(1 +2) ;
4
= 1
4(1 1)(1 +2)
a23
a24
1 0 1
1
0
1
0
0.5
1
1 0 1
1
0
1
0
0.5
1
1 0 1
1
0
1
0
0.5
1
1 0 1
1
0
1
0
0.5
1
Linear interpolation functions in 4-node element
D.1.1 Cartesian coordinate system
In a Cartesian coordinate system the displacement of every point of a quadrilateral element
has two components, ux and uy. Both components are interpolated between the nodal dis-
placement components, using the shape functions. The element shape and the weighting
function is interpolated in the same way as the displacement.
_
ux
uy
_
=
_
1 0 2 0 3 0 4 0
0 1 0 2 0 3 0 4
_
_

_
u1
x
u1
y
u2
x
u2
y
u3
x
u3
y
u4
x
u4
y
_

_
u

= u

e
xi =
1
x
1
i +
2
x
2
i +
3
x
3
i +
4
x
4
i x

= x

e
x0i =
1
x
1
0i +
2
x
2
0i +
3
x
3
0i +
4
x
4
0i x

0 = x

e
0
wi =
1
w
1
i +
2
w
2
i +
3
w
3
i +
4
w
4
i w

= w

e
The columns L

u
and L

w
represent the components of the gradient of the displacement and the
weighting function, respectively. After interpolation of displacement and weighting function,
the so-called B-matrix appears.
a25
_

_
ux,x
uy,y
uy,x
ux,y
_

_
=
_

_
1
,x 0 2
,x 0 3
,x 0 4
,x 0
0 1
,y 0 2
,y 0 3
,y 0 4
,y
0 1
,x 0 2
,x 0 3
,x 0 4
,x
1
,y 0 2
,y 0 3
,y 0 4
,y 0
_

_
_

_
u1
x
u1
y
u2
x
u2
y
u3
x
u3
y
u4
x
u4
y
_

_
L

u
_
t
= Bu

e
The B-matrix contains derivatives of the shape functions {; = 1, 2, 3, 4} with respect to
the Cartesian coordinates x and y. The Jacobian matrix J contains the derivatives of the
Cartesian coordinates x and y with respect to the isoparametric coordinates 1 and 2.
_

_
1
,x 1
,y
2
,x 2
,y
3
,x 3
,y
4
,x 4
,y
_

_
=
_

_
1
,1 1
,2
2
,1 2
,2
3
,1 3
,2
4
,1 4
,2
_

_
_
_
1,x 1,y
2,x 2,y
_
_ =
_

_
1
,1 1
,2
2
,1 2
,2
3
,1 3
,2
4
,1 4
,2
_

_
J
T
J =
_
_
x,1 y,1
x,2 y,2
_
_ =
_
_
1
,1 2
,1 3
,1 4
,1
1
,2 2
,2 3
,2 4
,2
_
_
_

_
x1
e y1
e
x2
e y2
e
x3
e y3
e
x4
e y4
e
_

_
The deformation matrix can be calculated in each element integration point. Besides nodal
point coordinates in the current state, the coordinates in the reference state must be available.
F =
_

_
x
x0
x
y0
0
y
x0
y
y0
0
0 0 Fzz
_

_
_
x
x0
y
x0
x
y0
y
y0
_
=
_
1
,x0 2
,x0 3
,x0 4
,x0
1
,y0 2
,y0 3
,y0 4
,y0
_
_

_
x1
e y1
e
x2
e y2
e
x3
e y3
e
x4
e y4
e
_

_
=
_
1,x0 2,x0
1,y0 2,y0
_ _
1
,1 2
,1 3
,1 4
,1
1
,2 2
,2 3
,2 4
,2
_
_

_
x1
e y1
e
x2
e y2
e
x3
e y3
e
x4
e y4
e
_

_ = J
1
0 J
a26
D.1.2 Cylindrical coordinate system
In a cylindrical coordinate system the displacement of every point of a quadrilateral element
has two components, ur and uz. Both components are interpolated between the nodal dis-
placement components, using the shape functions. The element shape and the weighting
function is interpolated in the same way as the displacement.
_
ur
uz
_
=
_
1 0 2 0 3 0 4 0
0 1 0 2 0 3 0 4
_
_

_
u1
r
u1
z
u2
r
u2
z
u3
r
u3
z
u4
r
u4
z
_

_
u

= u

e
r =
1
r
1
+
2
r
2
+
3
r
3
+
4
r
4
=

T
r

z =
1
z
1
+
2
z
2
+
3
z
3
+
4
z
4
=

T
z

r0 =

T
r

0 ; z0 =

T
z

0
wi =
1
w
1
i +
2
w
2
i +
3
w
3
i +
4
w
4
i w

= w

e
The columns L

u
and L

w
represent the components of the gradient of the displacement and the
weighting function, respectively. After interpolation of displacement and weighting function,
the so-called B-matrix appears. For ease of programming, we swap the derivatives ut,t and
uz,z in the column L

u
and analoguously wt,t and wz,z in the column L

w
.
_

_
ur,r
uz,z
1
r ur
uz,r
ur,z
_

_
=
_

_
1
,r 0 2
,r 0 3
,r 0 4
,r 0
0 1
,z 0 2
,z 0 3
,z 0 4
,z
1
r 1 0 1
r 2 0 1
r 3 0 1
r 4 0
0 1
,r 0 2
,r 0 3
,r 0 4
,r
1
,z 0 2
,z 0 3
,z 0 4
,z 0
_

_
_

_
u1
r
u1
z
u2
r
u2
z
u3
r
u3
z
u4
r
u4
z
_

_
L

u
_
t
= Bu

e
The B-matrix contains derivatives of the shape functions {; = 1, 2, 3, 4} with respect to
the cylindrical coordinates r and z. The Jacobian matrix J contains the derivatives of the
cylindrical coordinates r and z with respect to the isoparametric coordinates 1 and 2.
_

_
1
,r 1
,z
2
,r 2
,z
3
,r 3
,z
4
,r 4
,z
_

_
=
_

_
1
,1 1
,2
2
,1 2
,2
3
,1 3
,2
4
,1 4
,2
_

_
_
_
1,r 1,z
2,r 2,z
_
_ =
_

_
1
,1 1
,2
2
,1 2
,2
3
,1 3
,2
4
,1 4
,2
_

_
J
T
a27
J =
_
_
r,1 z,1
r,2 z,2
_
_ =
_
_
1
,1 2
,1 3
,1 4
,1
1
,2 2
,2 3
,2 4
,2
_
_
_

_
r1
e z1
e
r2
e z2
e
r3
e z3
e
r4
e z4
e
_

_
The deformation matrix can be calculated in each element integration point. Besides nodal
point coordinates in the current state, the coordinates in the reference state must be available.
F =
_

_
r
r0
r
z0
0
z
r0
z
z0
0
0 0
r
r0
_

_
_
_
r
r0
z
r0
r
z0
z
z0
_
_ =
_
_
1
,r0 2
,r0 3
,r0 4
,r0
1
,z0 2
,z0 3
,z0 4
,z0
_
_
_

_
r1
e z1
e
r2
e z2
e
r3
e z3
e
r4
e z4
e
_

_
=
_
_
1,r0 2,r0
1,z0 2,z0
_
_
_
_
1
,1 2
,1 3
,1 4
,1
1
,2 2
,2 3
,2 4
,2
_
_
_

_
r1
e z1
e
r2
e z2
e
r3
e z3
e
r4
e z4
e
_

_
= J
1
0 J
D.1.3 Numerical integration
To generate the element stiness matrix and the residual force column, integration over the
element volume (2D : area) must be carried out. With quadrilateral elements this integration
cannot be done analytically, so numerical integration is necessary.
The numerical integration which we employ here is the Gauss quadrature integration.
The integrand is evaluated in a number of discrete points, the integration points or Gauss
points. The integration point values are multiplied by a weighting factor, , after which they
are added. The location of the integration points (= their isoparametric coordinates) and
the value of the weighting factor are determined in such a way that a polynomial of a certain
degree is integrated exactly.
The four-node quadrilateral integration point locations and weighting factor values are
shown in the table. Their choice is such that a polynomial of third order in each direction is
integrated exactly.
a28
4
3
2 1
1
4 3
2 1
2
ip 1 2
1
1
3

3
1
3

3 1
2 1
3

3 1
3

3 1
3
1
3

3
1
3

3 1
4
1
3

3
1
3

3 1
Integration points in a 4-node element
D.2 Eight-node quadrilateral element
The eight-node element has four sides (quadrilateral), which are straight lines in the un-
deformed conguration. The nodes 1 to 4 are located in the corners (corner nodes), the
nodes 5 to 8 are located in the middle of the sides (midpoint nodes). The numbering is
anti-clockwise.
Global coordinates, displacement components and weighting function components are
interpolated with shape functions which are quadratic along an element side. This implies
that in the deformed conguration these sides may be parabolic. The shape functions are
functions of the isoparametric coordinates 1 and 2.
4
3
2 1
1 8
7
6
5
2
Eight-node quadrilateral element

1
=
1
4 (1 1)(2 1)(1 2 1)

2
=
1
4 (1 + 1)(2 1)(1 +2 + 1)

3
=
1
4 (1 + 1)(2 + 1)(1 +2 1)

4
= 1
4 (1 1)(2 + 1)(1 2 + 1)

5
=
1
2 (
2
1 1)(2 1)

6
=
1
2 (1 1)(
2
2 1)

7
=
1
2 (
2
1 1)(2 1)
a29

8
=
1
2 (1 1)(
2
2 1)
The gures show the shape functions associated with the nodal points. The rst four plots
show the shape functions of the corner nodes and second series of four plots shows those of
the mid-side nodes.
1 0 1
1
0
1
0.5
0
0.5
1
1 0 1
1
0
1
0.5
0
0.5
1
1 0 1
1
0
1
0.5
0
0.5
1
1 0 1
1
0
1
0.5
0
0.5
1
1 0 1
1
0
1
0
0.5
1
1 0 1
1
0
1
0
0.5
1
1 0 1
1
0
1
0
0.5
1
1 0 1
1
0
1
0
0.5
1
Quadratic interpolation functions in 8-node element
D.2.1 Numerical integration
The table contains the location of the 9 integration (Gauss) points and their weighting func-
tions for the eight-node quadrilateral element. Their choice is such that a polynomial of fth
order in each direction is integrated exactly.
a30
2
5
6
7
1
1 2
3
4
8
6
1 2 3
5
7
8
9
4
a = 0.77459 ; p = 0.55556 ; q = 0.88889
ip 1 2
1 a a p p
2 0 a p q
3 a a p p
4 a 0 p q
5 0 0 q q
6 a 0 p q
7 a a p p
8 0 a p q
9 a a p p
Integration points in 8-node element
Appendix E
FE program plaxL
The program plaxL can be used to model and analyze linear elastic deformation of planar and
axisymmetric structures. The planar problems can be either plane strain or plane stress. For
axisymmetric problems, the rotation axis is the global y-axis and the radial axis is the global
x-axis. The half cross-section must be modelled for x 0. Both linear 4-node quadrilateral
and quadratic 8-node quadrilateral elements can be used.
E.1 Example input le
As an example, a tensile test is modelled and analyzed, both in plane stress and axisymmet-
ricly. Because the deformation is homogeneous, only one element is used.
Figures are closed and the Matlab work space is cleared.
close all; clear all;
The coordinates of the nodes are given in the matrix crd0. Units are chosen to be mm.
crd0 = [ 0 0; 100 0; 100 100; 0 100 ];
The location array lok contains information about the element type (rst column), element
group (second column) and connectivity (last four columns). The element type can be : 3 for
plane stress, 11 for plane strain and 10 for axisymmetry. As we have only one element, there
is only one element group. The element node numbers are 1, 2, 3 and 4.
lok = [ 3 1 1 2 3 4 ];
For each group we have to supply geometric and material data in the array elda. In the
subsequent columns we provide :
- 1 : random
- 2 : material number : 11 is isotropic linear material
- 3 : thickness
- 4 : unused
- 5 : unused
- 6 : Youngs modulus
- 7 : Poissons ratio
a31
a32
elda = [ 0 11 1 0 0 200000 0.3 ];
Prescribed displacements are provided in the array pp. For each prescribed displacement
component we have one row. The rst column contains the node, the second column contains
the direction (either 1 (= x = horizontal) or 2 (= y = vertical). The third column contains
the value.
When we want to prescribe the elongation of 1 mm, we have :
pp = [ 1 1 0; 1 2 0; 2 2 0; 4 1 0 ];
pp = [ pp; 3 2 1; 4 2 1 ];
We could also have prescribed a tensile force of 200000 N.
pp = [ 1 1 0; 1 2 0; 2 2 0; 4 1 0 ];
pf = [ 3 2 100000; 4 2 100000 ];
The input is complete and plaxL can be executed to analyze the behavior.
plaxL;
After the analysis the nodal displcacements are available in the array Mp. Element data
can be found in the database eidaC.
When an axisymmetric model is used, the axis is along the global y-axis, so vertical.
The coordinate and location array are then :
crd0 = [ 0 0; sqrt(100/pi) 0; sqrt(100/pi) 100; 0 100 ];
lok = [ 10 1 1 2 3 4 ];
When a tensile force is applied it will be dierent in the node on the central axis then in the
node on the outer edge.
The gure below shows the deformation in both cases.
1 2
3 4
1 2
3 4
1 2
3 4
a33
E.2 The program plaxL
The program plaxL is listed below and is seeded with comments to explain variables and
actions.
%**********************************************************************
% plaxL : 2-dimensional planar/axisym.; linear
%======================================================================
nndof = 2;
nnod = size(crd0,1);
ndof = nnod * nndof;
ne = size(lok,1);
negr = size(elda,1);
Trm = eye(ndof);
if exist(tr),
ntr = size(tr,1);
for itr=1:ntr
trp = round(tr(itr,1)); tra = tr(itr,2);
trc = cos((pi/180)*tra); trs = sin((pi/180)*tra);
k1 = nndof*(trp-1)+1; k2 = nndof*(trp-1)+2;
trm = [trc -trs ; trs trc];
Trm([k1 k2],[k1 k2]) = trm;
end;
else, ntr = 0; tr = []; end;
for e=1:ne
ety = lok(e,1); egr = lok(e,2);
if ety==3, nenod=4; neip=4; vrs=2;
elseif ety==11, nenod=4; neip=4; vrs=1;
elseif ety==10, nenod=4; neip=4; vrs=3; end;
nedof = nndof * nenod;
k=1;
for n=1:nenod
for v=1:nndof
lokvg(e,k) = nndof*(lok(e,2+n)-1)+v;
k=k+1;
end;
end;
elpa(e,1:7) = [ety egr nenod nndof nedof neip vrs];
mnr = elda(egr,2); thk = elda(egr,3);
gf2 = elda(egr,4); gf3 = elda(egr,5);
mcl = floor(mnr/10); mty = rem(mnr,10);
if mnr==11
E = elda(egr,6); Gn = elda(egr,7);
end;
for ip=1:neip
gip = neip*(e-1) + ip;
eida0(gip,:) = [ ety mnr thk gf2 gf3 E Gn ];
eidaC(gip,:) = [ ety mnr thk gf2 gf3 E Gn ];
end;
a34
end;
%----------------------------------------------------------------------
if ~exist(pp), pp = []; ppc = []; ppv = []; end;
if ~exist(pf), pf = []; pfc = []; pfv = []; end;
npdof = size(pp,1);
npfor = size(pf,1);
nudof = ndof - npdof;
if npdof>0
ppc = [nndof*(round(pp(:,1))-1)+round(pp(:,2))];
ppv = pp(:,nndof+1);
end;
if npfor>0,
pfc = [nndof*(round(pf(:,1))-1)+round(pf(:,2))];
pfv = pf(:,nndof+1);
end;
% Information for partitioning the system of equations associated
% with linked degrees of freedom is made available in the arrays
% plc and prc.
if ~exist(pl), pl = []; plc = []; end;
if ~exist(pr), pr = []; prc = []; end;
if ~exist(lim), lim = []; end;
npl = size(pl,1);
npr = size(pr,1);
if ~exist(lif), lif = zeros(1,npl); end;
if npl>0
plc = [nndof*(round(pl(:,1))-1)+round(pl(:,2))];
prc = [nndof*(round(pr(:,1))-1)+round(pr(:,2))];
end;
% Some extra arrays are made for later use.
pa = 1:ndof; pu = 1:ndof; prs = 1:ndof;
pu([ppc plc]) = [];
prs([ppc pfc plc]) = [];
% pe0 : column with prescribed initial displacements
% fe0 : array with prescribed initial forces
pe0 = zeros(ndof,1); pe0(ppc(1:npdof)) = ppv(1:npdof);
fe0 = zeros(ndof,1); fe0(pfc(1:npfor)) = pfv(1:npfor);
%----------------------------------------------------------------------
% Initialization to zero
% pe : column with nodal displacements
% p : column with nodal displacements
% fe : column with external (applied) nodal forces
% fi : column with internal (resulting) nodal forces
% #T : column with transformed components
a35
%----------------------------------------------------------------------
pe = zeros(ndof,1); p = zeros(ndof,1); pT = zeros(ndof,1);
fe = zeros(ndof,1); fi = zeros(ndof,1);
feT = zeros(ndof,1); fiT = zeros(ndof,1);
%----------------------------------------------------------------------
% Loop over all elements to generate element stiffness matrix em
% Assemble em into structural stiffness matrix sm
% ec0 : initial coordinates of element nodes
% ec : current coordinates of element nodes
%----------------------------------------------------------------------
sm=zeros(ndof);
for e=1:ne
ety = elpa(e,1); egr = elpa(e,2);
nenod = elpa(e,3); nedof = elpa(e,5); neip = elpa(e,6);
ec0 = crd0(lok(e,3:2+nenod),:); ec = ec0;
em = zeros(nedof);
[ksi,psi,psidksi,ipwf] = fbibfe2dq48(e,elpa(e,:)); % fbibfe2dq48.m
vrs = elpa(e,7);
for ip=1:neip
gip = neip*(e-1) + ip;
%----------------------------------------------------------------------
if vrs==3
r0 = psi(ip,:)*ec0(:,1); r = psi(ip,:)*ec(:,1);
thk0 = r0*2*pi; thk = r*2*pi;
mF33 = r/r0; ax = 1;
else
r0 = 1; r = 1;
thk0 = eida0(gip,3); thk = thk0;
mF33 = 1; ax = 0;
end;
dpsi(:,1) = psidksi(:,2*ip-1); dpsi(:,2) = psidksi(:,2*ip);
jc0 = dpsi * ec0; jci0 = inv(jc0);
jc = dpsi * ec; jci = inv(jc);
dt = det(jc); dt0 = det(jc0);
dfie0 = zeros(5,2*nenod);
dpsixy0 = dpsi * jci0 ;
dfie0(1,2*(1:nenod)-1) = dpsixy0(1:nenod,1);
dfie0(2,2*(1:nenod)) = dpsixy0(1:nenod,2);
dfie0(3,2*(1:nenod)-1) = ax.*psi(ip,1:nenod)/r0;
dfie0(4,2*(1:nenod)) = dpsixy0(1:nenod,1);
dfie0(5,2*(1:nenod)-1) = dpsixy0(1:nenod,2);
mF = eye(3); mF(1:2,1:2) = jci0*jc; mF(3,3) = mF33; mF = mF;
%----------------------------------------------------------------------
du = zeros(5,1);
[mmM,ccGs,ccGe,vm] = plaxelas1(eida0(gip,6:7),vrs,du);
% plaxelas1.m
em = em + dfie0 * mmM * dfie0 * thk0 * dt0 * ipwf(ip);
end;
sm(lokvg(e,1:nedof),lokvg(e,1:nedof)) = ...
a36
sm(lokvg(e,1:nedof),lokvg(e,1:nedof)) + em;
end; % element loop e
%----------------------------------------------------------------------
% Transformation for local nodal coordinate systems
%----------------------------------------------------------------------
sm = Trm*sm *Trm;
%----------------------------------------------------------------------
% Boundary conditions and links
%----------------------------------------------------------------------
pe = pe0; fe = fe0; rs = fe;
if npl>0, rs = rs - sm(:,plc)*lif; end;
%----------------------------------------------------------------------
% Partitioning is done in the function fbibpartit.m
[sm,rs] = fbibpartit(1,sm,rs,ndof,pa,ppc,plc,prc,pe,lim);
%----------------------------------------------------------------------
% Solving the system of equations and take prescribed displacements
% and links into account.
% Update nodal point coordinates crd.
sol = inv(sm)*rs; % sol = sm\rs;
pe(pu) = sol;
if npl>0, pe(plc) = lim*pe(prc) + lif; end;
p = pe; pT = Trm * p;
crd = crd0 + reshape(pT,nndof,nnod);
%----------------------------------------------------------------------
% Calculate stresses and strains and the internal forces ef.
% Internal forces ef are assembled into fi, the structural
% internal forces, representing the reaction forces.
%----------------------------------------------------------------------
fi = zeros(ndof,1);
for e=1:ne
ety = elpa(e,1); egr = elpa(e,2);
nenod = elpa(e,3); nedof = elpa(e,5); neip = elpa(e,6);
ec0 = crd0(lok(e,2+1:2+nenod),:);
ec = crd(lok(e,2+1:2+nenod),:);
Tpe = p(lokvg(e,1:nedof));
em = zeros(nedof); ef = zeros(nedof,1);
[ksi,psi,psidksi,ipwf] = fbibfe2dq48(e,elpa(e,:)); % fbibfe2dq48.m
vrs = elpa(e,7);
for ip=1:neip
gip = neip*(e-1) + ip;
%----------------------------------------------------------------------
if vrs==3
r0 = psi(ip,:)*ec0(:,1); r = psi(ip,:)*ec(:,1);
thk0 = r0*2*pi; thk = r*2*pi;
a37
mF33 = r/r0; ax = 1;
else
r0 = 1; r = 1;
thk0 = eida0(gip,3); thk = thk0;
mF33 = 1; ax = 0;
end;
dpsi(:,1) = psidksi(:,2*ip-1); dpsi(:,2) = psidksi(:,2*ip);
jc0 = dpsi * ec0; jci0 = inv(jc0);
jc = dpsi * ec; jci = inv(jc);
dt = det(jc); dt0 = det(jc0);
dfie = zeros(5,2*nenod);
dpsixy = dpsi * jci ;
dfie(1,2*(1:nenod)-1) = dpsixy(1:nenod,1);
dfie(2,2*(1:nenod)) = dpsixy(1:nenod,2);
dfie(3,2*(1:nenod)-1) = ax.*psi(ip,1:nenod)/r;
dfie(4,2*(1:nenod)) = dpsixy(1:nenod,1);
dfie(5,2*(1:nenod)-1) = dpsixy(1:nenod,2);
dfie0 = zeros(5,2*nenod);
dpsixy0 = dpsi * jci0 ;
dfie0(1,2*(1:nenod)-1) = dpsixy0(1:nenod,1);
dfie0(2,2*(1:nenod)) = dpsixy0(1:nenod,2);
dfie0(3,2*(1:nenod)-1) = ax.*psi(ip,1:nenod)/r0;
dfie0(4,2*(1:nenod)) = dpsixy0(1:nenod,1);
dfie0(5,2*(1:nenod)-1) = dpsixy0(1:nenod,2);
mF = eye(3); mF(1:2,1:2) = jci0*jc; mF(3,3) = mF33; mF = mF;
%----------------------------------------------------------------------
du = dfie0*Tpe;
[mmM,ccGs,ccGe,vm] = plaxelas1(eidaC(gip,6:7),vrs,du);
ccGs = ccGs; ccGe = ccGe;
thk = thk0 + thk0*ccGe(3);
eidaC(gip,3) = thk;
eidaC(gip,17:20) = ccGe(1:4);
eidaC(gip,21:24) = ccGs(1:4);
eidaC(gip,25) = dt;
eidaC(gip,90) = r0;
eidaC(gip,91) = r;
ef = ef + dfie * ccGs * thk * dt * ipwf(ip);
end;
fi(lokvg(e,1:nedof)) = fi(lokvg(e,1:nedof)) + ef;
end;
rs = fe - fi;
fi=Trm*fi; fiT = fi; fiT = Trm * fi; rsT = feT - fiT;
%----------------------------------------------------------------------
% Reshaping columns into matrices
Mp = reshape(p,nndof,nnod);
Mfi = reshape(fi,nndof,nnod); Mfe = reshape(fe,nndof,nnod);
Mrs = reshape(rs,nndof,nnod);
if ntr>=1
MpT = reshape(TpT,nndof,nnod);
a38
MfiT = reshape(fiT,nndof,nnod); MfeT = reshape(feT,nndof,nnod);
MrsT = reshape(rsT,nndof,nnod);
end;
%----------------------------------------------------------------------
%**********************************************************************
Appendix F
FE program plax
The Matlab program plax allows to model and analyze two-dimensional planar and axisym-
metric structures, where the deformation can be large and the material nonlinear..
Model geometry, topology (connectivity), geometrical and material parameters, bound-
ary conditions (prescribed displacements and point loads) and link relations (dependecies
between degrees of freedom) must be available as input data. Also the history of the pre-
scribed boundary conditions must be specied.
When the analysis is nished, output date are stored in the data base and various other
data arrays.
In the following section an example input is presented, with explanatory comments.
Finally the program itself is explained in more detail.
F.1 The program plax
The program plax is listed below. It is clear that the program calls a collection of com-
mand and function les, which can and should be inspected for through understanding of the
procedure. The program structure is however clearly shown in the listing.
%**********************************************************************
% plax : 2-dimensional planar/axisym.; nonlinear
%======================================================================
plaxchkinp; % plaxchkinp.m
fbiblcase; % fbiblcase.m
plaxinizer; % plaxinizer.m
[eida0,eidaB,eidaC,eismB,eismC,elip,neip] = ... % plaxinidat.m
plaxinidat(ne,elgr,elda,neip,GDt,mm);
if res==0, save([matf num2str(0)]); end;
crdB = crd0; crd = crd0;
%======================================================================
% Calculate shape functions and their derivatives % plaxq4.m plaxq8.m
%======================================================================
if nenod==4, [ksi,psi,psidksi,ipwf,lokvg] = plaxq4(lok,ne,nndof,neip);
elseif nenod==8, [ksi,psi,psidksi,ipwf,lokvg] = plaxq8(lok,ne,nndof,neip);
end;
%======================================================================
% Incremental calculation
%======================================================================
a39
a40
ic = 1; ti = 0; it = 0; slow = 1;
if res>0
ic = res; load([matf num2str(ic-1)]);
crd = crdB; eidaC = eidaB; it = 1;
end;
while ic<=nic
%----------------------------------------------------------------------
fbibcutback; % fbibcutback.m
ti = ti + GDt; it = 0;
loadincr;
pe = peC./slow; fe = feC; % fe = feB + feC./slow;
rs = fe - fi;
Dp = zeros(ndof,1); Ip = zeros(ndof,1);
GDt = GDt0/slow;
%======================================================================
% Element matrices are calculated if necessary.
%======================================================================
if (ic==1 | nl==1 | ic==res)
%----------------------------------------------------------------------
sm = zeros(ndof); % structural stiffness matrix
fi = zeros(ndof,1); % structural internal force column
elmalivi = 0; % counter of lin.viscoel. elements
gip = 0; % global integration point number
for e=1:ne
ety = elda(elgr(e),1); mat = elda(elgr(e),2);
em = zeros(nedof); ef = zeros(nedof,1);
ec0 = crd0(lok(e,3:nenod+2),:); ecB = crdB(lok(e,3:nenod+2),:);
ec = crd(lok(e,3:nenod+2),:); Tpe = Tp(lokvg(e,:));
vole0 = 0; voleC = 0;
if mat==8, elmalivi = elmalivi + 1; end;
plaxelem; % -> ef, em % plaxelem.m
% fi(lokvg(e,:)) = fi(lokvg(e,:)) + ef;
sm(lokvg(e,:),lokvg(e,:)) = sm(lokvg(e,:),lokvg(e,:)) + em;
end; %e
%----------------------------------------------------------------------
end;
%======================================================================
% Iterative calculation
%======================================================================
nrm = 1000; it = 1;
while (nrm>ccr) & (it<=mit)
%----------------------------------------------------------------------
%======================================================================
% Links and boundary conditions are taken into account
% Unknown nodal point values are solved
% Prescribed nodal values are inserted in the solution vector
%======================================================================
if npl>0, rs = rs - sm(:,plc)*lif; end;
[smp,rsp] = fbibpartit(it,sm,rs,ndof,pa,ppc,plc,prc,pe,lim);% fbibpartit.m
a41
sol = smp\rsp; dsol = smp\(smp*sol-rsp); soll = sol-dsol;
p = zeros(ndof,1); p(pu) = soll;
if it==1, p(ppc) = pe(ppc); end;
if npl>0, p(plc) = lim*p(prc) + lif; end;
Dp = p; Ip = Ip + Dp; Tp = Tp + Dp;
crd = crd0 + reshape(Tp,nndof,nnod);
%======================================================================
% Calculate stresses and strains.
% Make system matrix and internal force vector for next step.
%======================================================================
sm=zeros(ndof); fi=zeros(ndof,1); elmalivi=0; gip=0;
for e=1:ne
ety = elda(elgr(e),1); mat = elda(elgr(e),2);
em = zeros(nedof); ef = zeros(nedof,1);
ec0 = crd0(lok(e,3:nenod+2),:); ecB = crdB(lok(e,3:nenod+2),:);
ec = crd(lok(e,3:nenod+2),:); Tpe = Tp(lokvg(e,:));
vole0 = 0; voleC = 0;
if mat==8, elmalivi = elmalivi + 1; end;
plaxelem; % -> ef, em % plaxelem.m
fi(lokvg(e,:)) = fi(lokvg(e,:)) + ef;
sm(lokvg(e,:),lokvg(e,:)) = sm(lokvg(e,:),lokvg(e,:)) + em;
end; %e
%======================================================================
% Calculate residual force and convergence norm
%======================================================================
rs = fe - fi;
nrm = fbibcnvnrm(cnm,pu,ppc,prs,Dp,Ip,rs,fi); % fbibcnvnrm.m
it = it + 1;
%plaxipc; % plaxipc.m
fbibwr2scr; % fbibwr2scr.m
%----------------------------------------------------------------------
end; %it
%======================================================================
% Update and store values
%======================================================================
fbibcol2mat1; % fbibcol2mat1.m
crdB = crd; feB = fe; eidaB = eidaC; eismB = eismC;
savefile = [matf num2str(ic)]; savedata; % savedata.m
ic = ic + 1;
save([matf 00],ic);
%----------------------------------------------------------------------
end; %ic
%**********************************************************************