Attribution Non-Commercial (BY-NC)

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

96 Aufrufe

Attribution Non-Commercial (BY-NC)

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

- Mar 103 Experimental Elastomer Analysis
- Rapid Prototyping
- A Handbook 0f Building Structure = Allan Hodgkinson
- Automobile Engg-UNIT-2
- Mae160_chap14
- 7207839 Ansys Structural Analysis Guide
- NISTNCSTAR1Draft
- Mae160_chap13
- Derivation of the Shell Element , Ahmed Element , Midlin Element in Finite Element Analysis- Hani Aziz Ameen
- Nonlinear Micro Finite Element Analysis of Human Trabecular Bone 2003
- Mathcad Users Guide
- Chapter 2
- 17192061-Phd-Thesis-1997-D
- Chapter 4
- Chapter 6
- Chapter 8
- Microsoft Word - Creep Behaviour of Reformer Tubes
- 6250099 Strength of Materials
- Nonlinear Solid Mechanics
- Dynamics of Smart Structure

Sie sind auf Seite 1von 22

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

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

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

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

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

be inverted to solve u

when f

for a rigid body translation u

= a

= 0

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

.

The unknown degrees of freedom u

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

way. Reorganizing components of f

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

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

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

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

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

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/

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

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

The derivatives of u

and w

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

= 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.

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

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

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

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

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

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.

= 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.

= 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

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

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]

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.

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

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

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

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

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

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

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

As any other column, A

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

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

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

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

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

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

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

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

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

119

undeformed conguration ( = t0) x0 = (

deformed conguration ( = t) x = (

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 =

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

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

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

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

=

_

_

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

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

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

C. The material compliance matrix S

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

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

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

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

= 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

d

{cos()

1

6

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

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

_

_

_

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

_

V0

_

L

0w

_T

t

C

_

L

0u

_

t

dV0 = fe0(w

) w

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

written as F = F

+ F and P = P

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

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

values are written as Fn = (I +L

u) F

between C

c and Cc, and L

u = ( u)c, with u = x

_

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

displacement components are stored in the column ()u

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

in the column x

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

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

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) =

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) =

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

= 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

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

_

_

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

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

J = (F11F22 F12F21)F33 = J1 +J

T

2

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)

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

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

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

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

w

= H B

d

= H B

(4)

F

= HF

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

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

%**********************************************************************

- Mar 103 Experimental Elastomer AnalysisHochgeladen vonDan Wolf
- Rapid PrototypingHochgeladen vonAnkit
- A Handbook 0f Building Structure = Allan HodgkinsonHochgeladen vonNiques Dawson
- Automobile Engg-UNIT-2Hochgeladen vonMuthuvel M
- Mae160_chap14Hochgeladen vonaraveeti
- 7207839 Ansys Structural Analysis GuideHochgeladen vonPandharkar
- NISTNCSTAR1DraftHochgeladen vonjrod
- Mae160_chap13Hochgeladen vonaraveeti
- Derivation of the Shell Element , Ahmed Element , Midlin Element in Finite Element Analysis- Hani Aziz AmeenHochgeladen vonHani Aziz Ameen
- Nonlinear Micro Finite Element Analysis of Human Trabecular Bone 2003Hochgeladen vonSIMULIACorp
- Mathcad Users GuideHochgeladen vonPandharkar
- Chapter 2Hochgeladen vonradoslav121
- 17192061-Phd-Thesis-1997-DHochgeladen vonawalmanuhuruk
- Chapter 4Hochgeladen vonradoslav121
- Chapter 6Hochgeladen vonradoslav121
- Chapter 8Hochgeladen vonradoslav121
- Microsoft Word - Creep Behaviour of Reformer TubesHochgeladen voneibrahim1
- 6250099 Strength of MaterialsHochgeladen vonReDo Distira
- Nonlinear Solid MechanicsHochgeladen vonAzam Rashdi
- Dynamics of Smart StructureHochgeladen vonTran Tuan Kiet
- 27324976-Mathcad-User-s-BookHochgeladen vonshabin007
- Lab creep testingHochgeladen vonjatyt
- Strength and Material 1 Project 2 Photo Elasticity by Ryan ParkHochgeladen vonRyan Horyang Park
- Notes for Finite Element Method Thermal AnalysisHochgeladen vonManikanth Valavoju
- 2008 Biot Mechanics of Incremental DeformationsHochgeladen vonarich82
- Introduct Fatigue PJHHochgeladen vonMartin Stone
- Ordinary and Partial Differential EquationsHochgeladen vonAmit Yadav
- Nair ThesisHochgeladen vonSuneel Kanhar
- Materials Methodology Stress FactorsHochgeladen von538540

- Determination of Peak X-Ray TubeHochgeladen vonJulio Fonso Reyes
- ATV312 Programming Manual en BBV46385 03Hochgeladen vonFrancisco Mendoza
- cspiii-160830160014Hochgeladen vonMahmoud Eissa
- S11003QUO01r0Hochgeladen vonFARHAN
- Zeta Potential of Air Bubbles Conditioned With Typical Froth Flotation ReagentsHochgeladen vonYesi Ceballos
- Characterizing the Bistable Flow, Of BWR, As a Bifurcation (Pitchfork Type) in the Navier-Stokes Equation SolutionHochgeladen vonIJAERS JOURNAL
- Jantar MantarHochgeladen vonspirit571
- JDZX22-10(3_6)C1G-A(N)Hochgeladen vonmuhammad nazir
- ScienceOlympiad5.0Hochgeladen vonhector15013
- ChemistryHochgeladen vonSanjay Verma
- 1.2.1 - Different Types of Gasifiers and Their Integration With Gas TurbinesHochgeladen vonRene Gonzalez
- Standards & StandardsHochgeladen vonNaresh
- Chapter 7 - Finite Element Method for Solid Mechanics (2016!10!10)Hochgeladen vonRick
- PHYSICS.pdfHochgeladen vonpavan
- Fundamental Theorem of Finite Abelian GroupsHochgeladen vonMao Ze Dong
- Manual telescopeHochgeladen vonNikola Ignjatovic
- resume gHochgeladen vonapi-336438736
- Chap 016Hochgeladen vonAbdullah Hamadi
- Identification and Quantification of the Free Sugars in Cocoa BeansHochgeladen vonAnne Suellen
- NEW BIOCOMPATIBLE POLYESTERS BASED ON BILE ACIDS.docHochgeladen vonChriss Bnhjyu
- Chemical Anchor Technical Handbook 072008Hochgeladen vonamadhubalan
- Final PPTHochgeladen vonvasu_koneti5124
- The Dymaxion World of Buckminster FullerHochgeladen vonNg Calin
- Ic TransformerHochgeladen vonax33m144
- Climate Change: Lines of EvidenceHochgeladen vonNational Academies of Science, Engineering, and Medicine
- Principles of the Theory of Vortex Gravitation - Sa OrlovHochgeladen vonmichaelcad
- Design Basis for Prestressed ConcreteHochgeladen vonPavethira Kumar Manokaran
- 05. Material Properties of Metals (Lectured)Hochgeladen vonJiann Hau
- Brochure - Bed GuardsHochgeladen vonAmit Goyal
- Olympus Om88 Om101 InstructionsHochgeladen vonfreddlutz765

## Viel mehr als nur Dokumente.

Entdecken, was Scribd alles zu bieten hat, inklusive Bücher und Hörbücher von großen Verlagen.

Jederzeit kündbar.