SOLID MECHANICS

Fundamentals and engineering concepts

of elasticity-based design
in mechanical engineering

MARC GEERS
PIET SCHREURS

Eindhoven University of Technology

Department of Mechanical Engineering
MECHANICS of MATERIALS
 C ONTENTS

Contents i

Preface vii

1 Tensor calculus 1
1.1 Scalars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2.2 Vector operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.3 Vector basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.4 Vector components; matrix notation of a vector . . . . . . . . . . . . . . 6
1.3 Second order tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Some properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.3 Tensor components; matrix notation of a tensor . . . . . . . . . . . . . . 10
1.3.4 Scalar functions of a tensor . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.5 Eigenvalues and eigenvectors of a tensor . . . . . . . . . . . . . . . . . . 13
1.3.6 Some special tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Spatial derivatives of scalar and vector fields . . . . . . . . . . . . . . . . . . . . 18
1.4.1 Derivatives of functions of one or more variables . . . . . . . . . . . . . 18
1.4.2 Components of the position vector . . . . . . . . . . . . . . . . . . . . . 19
1.4.3 The gradient operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Tensor formula involving first and second-order tensors . . . . . . . . . . . . . 22
1.6 Higher-order and fourth-order tensors . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.2 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.3 Tensor rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.4 Matrix condensation of the double inner product with a second-order
tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2 Kinematics 31
2.1 General aspects of deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1.1 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1.2 Lagrangian versus Eulerian coordinates . . . . . . . . . . . . . . . . . . . 32
2.1.3 The displacement vector . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.1.4 The displacement gradient tensor and deformation gradient tensor . . 34
2.2 Strains and strain tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

i
C ONTENTS

2.2.1 Strain definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2.2 Deformation of infinitely small material line elements . . . . . . . . . . 38
2.3 Linearization of deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.3.1 Linearized strain definitions . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.2 The infinitesimal linear strain tensor . . . . . . . . . . . . . . . . . . . . 42
2.3.3 The infinitesimal rotation tensor . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Applicability of the theory of small displacements . . . . . . . . . . . . . . . . . 44
2.5 Physical interpretation of the strain components . . . . . . . . . . . . . . . . . . 46
2.5.1 Normal or extensional strains . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.5.2 Shear strains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.5.3 Principal strains and strain invariants . . . . . . . . . . . . . . . . . . . . 49
2.5.4 Volume and shape changes . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.6 Compatibility conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.7 Special deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.7.1 Rigid body motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.7.2 Homogeneous deformation . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.7.3 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.7.4 Axisymmetric deformation . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3 Force and stress 61

3.1 Forces and stresses in a deformable continuum . . . . . . . . . . . . . . . . . . 61
3.1.1 External forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.1.2 Internal forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.1.3 The stress vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.1.4 The stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.2 Components of the stress tensor with respect to a basis . . . . . . . . . . . . . . 64
3.2.1 Normal stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.2 Shear stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3 Principal stresses and stress invariants . . . . . . . . . . . . . . . . . . . . . . . . 64
3.4 Hydrostatic and deviatoric stresses . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.4.1 Hydrostatic pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.4.2 Deviatoric stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5 Stress state with respect to a plane . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.5.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.5.2 Octahedral stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.5.3 Maximum shear stress in a material point . . . . . . . . . . . . . . . . . 69
3.6 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.6.2 Mohr’s stress circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.6.3 Mohr’s stress circle for three-dimensional stress states . . . . . . . . . . 73
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4 Mechanical equilibrium 77
4.1 Balance of mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.2 Balance of momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.2.1 General derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.2.2 Translational equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

ii
 Contents

4.2.3 Equilibrium in a Cartesian and cylindrical basis . . . . . . . . . . . . . . 81

4.3 Balance of moment of momentum . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.1 General derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.2 Moment or rotational equilibrium . . . . . . . . . . . . . . . . . . . . . . 83
4.3.3 Rotational equilibrium in a Cartesian and cylindrical basis . . . . . . . 84
4.4 Special deformation states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.1 Planar deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.4.2 Axisymmetric deformation . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.5 Mechanical work and internal energy . . . . . . . . . . . . . . . . . . . . . . . . 86
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

5 Three-dimensional linear elasticity 89

5.1 General aspects of constitutive equations . . . . . . . . . . . . . . . . . . . . . . 89
5.2 The concept and nature of linear elasticity . . . . . . . . . . . . . . . . . . . . . . 89
5.2.1 The elastic response of materials . . . . . . . . . . . . . . . . . . . . . . . 90
5.2.2 The elastic stored energy function . . . . . . . . . . . . . . . . . . . . . . 90
5.3 General 3D linear elastic solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.3.1 Generalized Hooke’s law for anisotropic solids . . . . . . . . . . . . . . . 91
5.4 Material symmetry and anisotropy . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4.1 Orthotropic materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.4.2 Transverse isotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.5 Linear elastic isotropic materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.5.1 The isotropic elasticity tensor . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.5.2 Engineering elastic moduli . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.5.3 Stored energy density for isotropic linear elastic materials . . . . . . . . 103
5.5.4 Coincidence of principal axes of stress and strain . . . . . . . . . . . . . 103
5.5.5 Plane strain and plane stress . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.6 The superposition principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

6 Limit design criteria 109

6.1 Failure modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2 Standard tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.2.1 Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.2.2 Torsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.3 Elastic limit criteria for isotropic materials . . . . . . . . . . . . . . . . . . . . . . 112
6.3.1 Rankine criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.3.2 Tresca yield criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.3.3 Von Mises yield criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.3.4 Mohr-Coulomb limit criterion . . . . . . . . . . . . . . . . . . . . . . . . 120
6.3.5 Drucker-Prager limit criterion . . . . . . . . . . . . . . . . . . . . . . . . 123
6.3.6 General criteria for isotropic solids . . . . . . . . . . . . . . . . . . . . . 127
6.4 Limit criteria for anisotropic materials . . . . . . . . . . . . . . . . . . . . . . . . 127
6.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.6 Elastic limit behaviour: Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

7 Mechanical design: solutions 133

7.1 Vector/tensor equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

iii
C ONTENTS

7.2 Three-dimensional scalar equations . . . . . . . . . . . . . . . . . . . . . . . . . 134

7.2.1 Cartesian components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.2.2 Cylindrical components . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.2.3 Material law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.3 Planar deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.3.1 Cartesian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.3.2 Cylindrical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.3.3 Cylindrical : axisymmetric + u t = 0 . . . . . . . . . . . . . . . . . . . . . . 137
7.4 Stress inconsistency under plane stress . . . . . . . . . . . . . . . . . . . . . . . 138
7.5 Solution strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.5.1 Governing equations for unknowns . . . . . . . . . . . . . . . . . . . . . 138
7.5.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.6 Solution : displacement method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.6.1 Planar, Cartesian : Navier equations . . . . . . . . . . . . . . . . . . . . . 140
7.6.2 Planar, axisymmetric with u t = 0 . . . . . . . . . . . . . . . . . . . . . . . 140
7.6.3 Planar, axisymmetric with u t = 0, isotropic . . . . . . . . . . . . . . . . . 141
7.7 Weighted residual formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.7.1 Weighted residual formulation for 3D deformation . . . . . . . . . . . . 142
7.7.2 Weighted residual formulation : linear . . . . . . . . . . . . . . . . . . . 142
7.7.3 Finite element method for 3D deformation . . . . . . . . . . . . . . . . . 143

8 Mechanical design: analytical 147

8.1 Homogeneous plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.2 Axisymmetric cylinders and discs . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2.1 Prescribed edge displacement . . . . . . . . . . . . . . . . . . . . . . . . 149
8.2.2 Edge load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.3 Circular hole in infinite medium . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.3.1 Pressurized hole in an infinite medium . . . . . . . . . . . . . . . . . . . 152
8.3.2 Stress-free hole in bi-axially loaded infinite medium . . . . . . . . . . . 152
8.4 Rotating disc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
8.4.1 Solid disc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
8.4.2 Disc with central hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.4.3 Disc fixed on rigid axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.4.4 Rotating disc with variable thickness . . . . . . . . . . . . . . . . . . . . 157
8.5 Large thin plate with a central hole . . . . . . . . . . . . . . . . . . . . . . . . . . 158

9 Mechanical design: numerical 161

9.1 MSC.Marc/Mentat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
9.2 Cartesian, planar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
9.2.1 Tensile test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
9.2.2 Shear test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
9.2.3 Orthotropic plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.3 Axisymmetric, u t = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.4 Axisymmetric, planar, u t = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
9.4.1 Prescribed edge displacement . . . . . . . . . . . . . . . . . . . . . . . . 165
9.4.2 Edge load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
9.4.3 Centrifugal load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
9.4.4 Large thin plate with a central hole . . . . . . . . . . . . . . . . . . . . . 168

iv
 Contents

10 Large deformations & nonlinear behaviour 169

10.1 Nonlinear axial deformation of trusses . . . . . . . . . . . . . . . . . . . . . . . . 169
10.1.1 Strains for large elongation . . . . . . . . . . . . . . . . . . . . . . . . . . 169
10.1.2 Mechanical power and stress definitions . . . . . . . . . . . . . . . . . . 171
10.1.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
10.2 Nonlinear material models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
10.2.1 Material behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
10.2.2 Discrete material models . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
10.3 Large deformation elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
10.3.1 Elastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
10.3.2 Hyper-elastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
10.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
10.4.1 Example 1: limitations of a linear relation between the Cauchy stress
and the Green-Lagrange strain . . . . . . . . . . . . . . . . . . . . . . . . 186
10.4.2 Example 2: instability & localisation . . . . . . . . . . . . . . . . . . . . . 188
10.4.3 Example 3: inflating a spherical balloon . . . . . . . . . . . . . . . . . . 189

11 Plasticity 191
11.1 Elasto-plastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
11.1.1 Uniaxial test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
11.1.2 Hardening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
11.2 Elasto-plastic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
11.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
11.3.1 Example 1: Reversed plasticity in a tensile bar . . . . . . . . . . . . . . . 201
11.3.2 Example 2: parallel truss structure . . . . . . . . . . . . . . . . . . . . . . 203
11.3.3 Example 3: serial truss structure . . . . . . . . . . . . . . . . . . . . . . . 204

A Stiffness and compliance matrices 207

A.1 General orthotropic material law . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
A.1.1 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
A.1.2 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
A.1.3 Plane strain/stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
A.2 Linear elastic orthotropic material . . . . . . . . . . . . . . . . . . . . . . . . . . 211
A.2.1 Voigt notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
A.2.2 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
A.2.3 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
A.3 Linear elastic transversal isotropic material . . . . . . . . . . . . . . . . . . . . . 213
A.3.1 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
A.3.2 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
A.4 Linear elastic isotropic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
A.4.1 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
A.4.2 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

B Mechanical design problems 217

B.1 Governing equations and general solution . . . . . . . . . . . . . . . . . . . . . . 218
B.2 Disc, edge displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
B.3 Disc/cylinder, edge load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
B.4 Rotating solid disc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221

v
C ONTENTS

B.5 Rotating disc with central hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

B.6 Rotating disc fixed on rigid axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

vi
 P REFACE

The functionality, strength and reliability of a mechanical component, product or device

are controlled by the deformations or loads imposed through its service conditions. Defor-
mations are caused by the (external) loads, giving rise to stresses within the material. In most
engineering design problems, deformations and rotations remain small and reversible (elas-
tic), whereby no energy is dissipated in the material. This regime is adequately described
with the theory of linear elasticity, which is widely used as the basic tool for designing and
optimizing products and defining its service limits.
The modeling of complex products is inevitably a three-dimensional problem. The fact
that strains and stresses have to be characterized in a 3D-space, is certainly one of the goals
of this course. Within a 3D-space, vectors and tensors are indispensable. A whole chap-
ter has therefore been devoted to this subject. Next, the kinematical framework for gen-
eral (large) deformations is set up, and consequently linearized in the case of infinitesimal
displacements. The stress tensor is defined and the 3D constitutive laws for isotropic and
anisotropic linear elastic materials are given. Particular attention is given to the various
forms in which the three-dimensional Hooke’s law appears, as well as to the relations be-
tween the various elastic moduli. Combined with the governing equilibrium equations, de-
sign problems can be solved through the elaboration of appropriate design rules. Particular
cases, for which analytical solutions are available, are treated in the the course. Elemen-
tary aspects of numerical solutions techniques for more complex problems are presented
and simple problems are solved by the students in the exercise sessions. The course ends
with some one-dimensional basics on nonlinear deformation (large displacements and ro-
tations) and more complex material behaviour (plasticity). This last part serves as a bridge
to the mechanics of manufacturing processes, in which the production steps are designed,
shaped, controlled and optimized. This subject is treated in follow-up courses.
These lecture notes constitute the core of the course ’Solid Mechanics’. In spite of our
best efforts, some errors doubtless remain. We would highly appreciate to be informed on
any error found or any other comment. We sincerely hope that this monograph stimulates
the student’s interest in the field of solid mechanics and its application in the design of many
products.

Marc Geers
Piet Schreurs

January 2015

vii
 CHAPTER
1
V ECTOR AND TENSOR CALCULUS

In this chapter, attention is given to the definition and manipulation of vector and tensor
fields. This mathematical description is used in many physical disciplines, and mechanics
in particular. Vectors and tensors enable a compact notation and manipulation, without the
necessity to resort to individual components with respect to a particular basis. The latter is
needed only at the point that quantitative calculations need to be done, which is generally
postponed to the last step. Special attention is given to Cartesian tensor calculus, which pro-
vides the basis for the elaboration of coordinate-free equations. Within this specific frame-
work, no distinction is made between contravariant and covariant tensors while the metric
tensor is not introduced neither. The quantification into components measured with respect
to a specific basis, and the related matrix calculus is also elaborated.

1.1 Scalars
A scalar a is a quantity which has only a measure but no direction. It is the most trivial
tensor, the tensor of order zero. The value of a scalar does not depend on the choice of a
basis. Well-known examples are the temperature T or the density ρ in a material point.

1.2 Vectors
1.2.1 Definition
A vector is a physical quantity possessing both a magnitude (or length) and a direction. In
case of a vector ~
a:
~
a = ||~
a ||~
e (1.1)

the length is given by k~a k, and the direction is given by a unit vector ~
e , having a unit length.
A vector is a first-order tensor. The zero vector ~0 is a special vector having zero length. The
so-called index notation of a vector ~ a is given by

~
a =⇒ ai (1.2)
1. T ENSOR CALCULUS

r
a

Figure 1.1: Magnitude and direction of a vector ~

a.

A vector is thus indicated with a single index (i.c. i ), the value of which ranges between 1 and
the dimension of the Euclidian space it is used in (commonly this dimension equals 3). The
order of this tensor (i.e., 1) thus equals its number of indices. The index notation obtains a
specific meaning once a basis has been chosen, since the components of a vector are trivially
identified with this notation, see 1.2.3.

1.2.2 Vector operations

Multiplication of a vector ~
a = k~ e by a scalar α yields a vector ~
a k~ b with length αk~ a k in the
direction ~
e:
~
b = α~ a (1.3)
The sum of two vectors ~a and ~
b is a new vector ~c , equal to the diagonal of the parallelogram
r
a r
aa

r
-a

Figure 1.2: Multiplication of a vector: ~

b = α~
a.

a and ~
spanned by ~ b.
~ a +~
c =~ b (1.4)
scalar product, inner product or dot product 1 of two vectors is a scalar quantity, defined as:
r
r c
b

r
a

Figure 1.3: The sum of two vectors: ~ a +~

c =~ b.

a ·~
~ a || ||~
b = ||~ b|| cos(φ) (1.5)

a and ~
where φ is the smallest angle between ~ b. Note that the dot product of two nonzero
mutually perpendicular vectors equals zero. The product is commutative, i.e.:
1
inwendig produkt

2
 1.2. Vectors

r
b

r
j a

a ·~
Figure 1.4: The inner product of two vectors: ~ b.

a ·~
~ b =~
b·~
a (1.6)

and linear
a ·(~
~ b +~ a ·~
c) = ~ b +~
a ·~
c (1.7)
Using index notations, Einstein’s summation convention has to be introduced. This con-
vention states, that a summation is made over all repeated indices, where the range of the
summation equals the dimension n of the problem. For the inner product, this is written as
3
X
a ·~
~ b = ai bi = ai bi (1.8)
i =1

The summation symbol is generally not written, but here simply added to clarify the con-
vention.
The scalar product of a vector with itself yields the square of its length:

~
a ·~ a ||2 ≥ 0
a = ||~ (1.9)

The vector product or cross product of two vectors is a vector quantity:

a ×~
~ b =~
c (1.10)

The length of the vector ~

c is given by:

a || ||~
c k = ||~
k~ b|| sin(φ) (1.11)

where φ is the smallest angle between ~ a and ~

b. Its value equals the area of the parallelogram
spanned by the vectors ~ ~
a and b. The direction of ~ c is perpendicular to both ~a and ~b, such
that the vectors ~ ~
a , b and ~
c form a righthanded system. The linearity of the vector product is
expressed through
a × (~
~ b +~ a ×~
c) = ~ b +~ a ×~
c (1.12)
The vector product is not commutative:

~
b ×~ a ×~
a = −~ b (1.13)

which can be concluded from the definition. It can be shown that:

a × (~
~ c ) = (~
b ×~ c )~
a ·~ a ·~
b − (~ b)~
c (1.14)

3
1. T ENSOR CALCULUS

r r
c b

r
a

Figure 1.5: The cross product of two vectors: ~ a ×~

c =~ b.

The triple product2 of three vectors is a scalar quantity, defined as:

a ×~
~ b·~ a || ||~
c = ||~ b|| sin(φ)(~ c)
n ·~
a || ||~
= ||~ b|| sin(φ)||~
c || cos(ψ) (1.15)

where ~n is the unit vector in the direction ~a ×~ a and ~

b, φ is the smallest angle between ~ b and
ψ is the smallest angle between ~ c and ~
n . The absolute value of the triple product equals the

r
c
r
b
y
r
n j r
a
a ×~
Figure 1.6: The triple product of three vectors ~ c.
b·~

volume of the parallelepiped spanned by the three vectors. If ~ a ×~ c 6= 0, the vectors are not
b·~
lying within one plane, and are called mutually independent.
The tensor product3 or open product of two vectors ~ a~
b is a linear vector transformation
called a dyad. Application of the dyad ~a~ ~ yields a vector ~
b to a vector p a , multiplied by the
~
scalar (b·~
p ):
a~
~ p =~
b·~ a (~ p)
b·~ (1.16)

or in index notation
(ai b j )p j = ai (b j p j ) (1.17)

The linearity of the transformation is expressed by:

a~
~ p + β~
b·(α~ a (~
q ) = α~ a (~
p ) + β~
b·~ q)
b·~ (1.18)
2
tripel produkt
3
dyadisch produkt

4
 1.2. Vectors

rr
ab
r
p
r r r
(b × p) a

a~
Figure 1.7: The tensor product: ~ p.
b·~

The transpose of a dyad (also called conjugate dyad or adjoint dyad)4 is defined as:

a~
(~ b)T = ~
b~ a~
a 6= ~ b (1.19)

a~
If (~ a~
b)T = ~ a~
b, the dyad ~ b is called symmetric.

1.2.3 Vector basis

In three-dimensional space, a set of three vectors ~ c1 , ~
c2 and ~
c3 is called a basis, denoted by
{~
c1 ,~
c2 ,~
c3 }, if the vectors are independent, i.e. if the triple product of the vectors is nonzero:

{~
c1 ,~
c2 ,~
c3 } ⇒ ~
c1 ×~ c3 6= 0
c2 ·~ (1.20)

The three vectors ~

c1 , ~
c2 and ~
c3 are called base vectors. If the basis consists of mutually per-

r
e3
r
c3
r
c2
r
e2
r
e1
r
c1
(a) {~
c 1 ,~
c 2 ,~
c3 } (b) {~
e 1 ,~
e 2 ,~
e3}

Figure 1.8: A vector basis (a) and an orthonormal vector basis (b)

pendicular vectors, the basis is called an orthogonal basis.

If the basis consists of mutually perpendicular unit vectors, the basis is called an or-
thonormal basis. In the latter case the base vectors are usually denoted by ~e 1, ~
e 2 and ~
e 3.
Thus it holds that: ½
0 if i 6= j
{~
e 1 ,~ e3} ⇒ ~
e 2 ,~ e j = δi j =
e i ·~ (1.21)
1 if i = j
4
geconjugeerde dyade

5
1. T ENSOR CALCULUS

where δi j is called the Kronecker delta. A basis {~

e 1 ,~ e 3 } is called righthanded5 if the order
e 2 ,~
of ~
e 1, ~
e 2 and ~
e 3 is such that:

~ e2 = ~
e 1 ×~ e3 (1.22)
~ e3 = ~
e 2 ×~ e1 (1.23)
~ e1 = ~
e 3 ×~ e2 (1.24)

A Cartesian basis is an orthonormal basis in which the base vectors are independent of the
location in three-dimensional space. It is usually denoted by {~
e 1 ,~
e 2 ,~
e 3 } or {~
e x ,~
e y ,~
e z }. A cylin-

x3
r
e3
r
e2
r
e3 r
r e1
e2

r x2
e1

x1
Figure 1.9: A Cartesian orthonormal basis {~
e 1 ,~
e 2 ,~
e 3 }.

drical basis is an orthonormal basis as well. It is usually denoted by {~

e r ,~
e t ,~
e z }. The bases
{~
e 1 ,~
e 2 ,~
e 3 } and {~
e r ,~
e t ,~
e z } are mutually related by:

~
e r (θ) = cos(θ)~
e 1 + sin(θ)~
e2 (1.25)
~
e t (θ) = − sin(θ)~
e 1 + cos(θ)~
e2 (1.26)
~
ez = ~
e3 (1.27)

It is important to note that the base vectors ~

e r and ~
e t depend on the circumferential coordi-
nate θ. The derivatives of the vectors with respect to θ are given by:
e r (θ)
d~
=~e t (θ) (1.28)
dθ
e t (θ)
d~
= −~ e r (θ) (1.29)
dθ

1.2.4 Vector components; matrix notation of a vector

Let {~
c1 ,~ c3 } be a basis in three-dimensional space. Then any vector ~
c2 ,~ a can be expressed as
a unique linear combination of the base vectors. In case of an orthonormal basis {~ e 1 ,~
e 2 ,~
e 3}
we may write:
~
a = a1~
e 1 + a2~
e 2 + a3~
e 3 = ai ~
ei , (1.30)
5
rechtshandige basis

6
 1.2. Vectors

z r
ez r
x3 et
r
er
r
e3
r
e2 x2

r
e1
q
x1
r
Figure 1.10: A cylindrical orthonormal basis {~
e r ,~
e t ,~
e z }.

where we again used the Einstein summation convention, i.e. we sum over the repeated index
i . The coefficients ai are called the components of the vector ~
a with respect to the basis. Note

a3
r
a
r
e3
a2
r
e2
r
e1
a1

Figure 1.11: Components ai of a vector ~

a with respect to a Cartesian basis {~
e 1 ,~
e 2 ,~
e 3 }.

that, since the basis is orthonormal, a component ai of a vector ~

a with respect to the base
vector ~e i can be found by:

ai = ~
a ·~
ei (1.31)

The component ai represents the length of the projection of the vector ~a on the line with
direction ~
e i . For convenience we can use a matrix notation in which we store base vectors

7
1. T ENSOR CALCULUS

and components in a column:


a1
~ = a2
a (1.32)
 
a3
 
~
e1
e = ~
~
~
e 2 (1.33)
~
e3

Thus we may write:

 
~
£ ¤ e1 T
a = a1 a2
~ a3 ~
e 2 = a
~ ~ e
~ (1.34)
~
e3

or, equivalently:
 
£ ¤ a1
a = ~
~ e1 ~
e2 ~ eT a
e 3 a 2  = ~
~ ~ (1.35)
a3

The relations ai = ~ e i can now be written as:

a ·~

~ =~
a a ·~
e
~ (1.36)

1.3 Second order tensors

1.3.1 Definition
A second order tensor is a linear transformation of a vector into another vector:

p =~
A·~ q (1.37)

or in index notation
A i j p j = qi (1.38)

Note that we limit ourselves here to so-called Cartesian tensors, which allow orthogonal lin-
ear coordinate transformations only. In a three-dimensional space, a second-order tensor is
fully defined by the image of three independent vectors, see figure 1.12. Indeed, any set of
three independent vectors constitutes a vector basis by means of which any other vector can
be expressed. The image of an arbitrary vector can be found trivially through the images of
three such vectors. This property is a direct consequence of the linearity of the transforma-
tion, which is expressed by:

~ + β~
A·(αm ~ + βA·~
n ) = αA·m n (1.39)

For an arbitrary vector ~

a = a1~
e 1 + a2~
e 2 + a3~
e 3 the image vector thus equals

a = a1 (A·~
A·~ e 1 ) + a2 (A·~
e 2 ) + a3 (A·~
e 3) (1.40)

8
 1.3. Second order tensors

A r r
d = A ×a
r
a r
b r r
f = A ×c
r r
e = A ×b
r
c

Figure 1.12: A second-order tensor A in a three-dimensional space.

1.3.2 Some properties

Any second order tensor may be represented by a finite, nonunique, sum of dyads:

a1~
A = α1~ a2~
b 1 + α2~ a3~
b 2 + α3~ b3 + . . . (1.41)

Two special tensors are the zero tensor O and the unit tensor6 I . If applied to vector ~
p , these
tensors yield the following:
p = ~0
O·~ (1.42)
I ·~ ~
p =p (1.43)
The index notations for O and I are respectively

0 ⇔ 0i j (1.44)
I ⇔ δi j (1.45)

where the Kronecker delta δi j is used for the unit tensor.

The transpose or conjugate A T of a tensor A is defined by the following relation:

A T ·~ ~·A
p =p (1.46)

In index notation it holds that

A ⇔ Ai j =⇒ AT ⇔ A j i (1.47)

a~
For the special case A = ~ b, we find:

a~
(~ b)T ·~ a~
~·(~
p =p b)
= (~ a )~
p ·~ b
=~
b(~ a)
p ·~
=~
b(~ p)
a ·~
= (~
b~a )·~
p (1.48)
6
tweede-orde eenheidstensor

9
1. T ENSOR CALCULUS

showing that:
a~
(~ b)T = ~
b~a (1.49)
T
If A is written as a sum of dyads, the conjugate A is obtained by reversing the two vectors in
each dyad. Tensors can be multiplied by a scalar, summed with another tensor, or multiplied
by another tensor:
(αA)·~ v ),
v = A·(α~ (1.50)
(A + B )·~
v = A·~ v,
v + B ·~ (1.51)
(A·B )·~ v)
v = A·(B ·~ (1.52)
Note that (in general):
A·B 6= B ·A (1.53)
The inner product of a tensor with itself is called the square of a tensor, i.e.,
A 2 = A·A (1.54)
Likewise, higher-order exponents are defined through repeated inner products, e.g.,
A 3 = A·A·A (1.55)

1.3.3 Tensor components; matrix notation of a tensor

Consider a three-dimensional space with a basis {~ e 1 ,~
e 2 ,~
e 3 }. Then, according to (1.30), all
ai and ~
vectors ~ b i in relation (1.41) can be written with respect to this basis. Substitution of
these expressions in (1.41) results into:
A = A 11~
e 1~
e 1 + A 12~
e 1~
e 2 + A 13~
e 1~
e 3+
A 21~
e 2~
e 1 + A 22~
e 2~
e 2 + A 23~
e 2~
e 3+
A 31~
e 3~
e 1 + A 32~
e 3~
e 2 + A 33~
e 3~
e3 (1.56)
where the coefficients A i j are functions of the coefficients αi and the components of the
vectors ~a and ~
b. Note that the coefficients now coincide with the index notation. Thus the
tensor A can be expressed as a finite sum of 9 dyads of the base vectors. The coefficients A i j
are called the components of the tensor A with respect to the basis {~ e 1 ,~
e 2 ,~
e 3 }. In general, a
component A i j of a tensor A can be obtained by:
Ai j = ~
e i ·A·~
ej (1.57)
as can be verified from substitution of (1.56) in this equation. For convenience, we can store
the components A i j in a matrix A:
 
A 11 A 12 A 13
A =  A 21 A 22 A 23 
A 31 A 32 A 33
 
(~ e 1 ) (~
e 1 ·A·~ e 2 ) (~
e 1 ·A·~ e 3)
e 1 ·A·~
= (~ e 1 ) (~
e 2 ·A·~ e 2 ) (~
e 2 ·A·~ e 3 )
e 2 ·A·~
(~ e 1 ) (~
e 3 ·A·~ e 2 ) (~
e 3 ·A·~ e 3)
e 3 ·A·~
 
~
e1 £ ¤
 ~
= e 2  ·A· ~ e1 ~e2 ~ e3
~
e3
=~ eT
e ·A·~
~ ~ (1.58)

10
 1.3. Second order tensors

Using the matrix A, relation (1.56) can be rewritten as:

  
£ ¤ A 11 A 12 A 13 ~ e1
A = ~e1 ~
e2 ~
e 3  A 21 A 22 A 23 ~
  e 2
A 31 A 32 A 33 ~ e3
eT A ~
=~
~ e
~ (1.59)

Together, equations (1.58) and (1.59) describe the transformation from tensor to matrix and
back.
In (1.43), the unit tensor I was defined. After writing I in the general form (1.56) and
applying definition (1.43), it can be derived that the unit tensor can be written as:

I =~
e 1~ e 2~
e 1 +~ e 3~
e 2 +~ e3 (1.60)

Thus, for the unit matrix I belonging to the unit tensor I it holds:
 
1 0 0
I =~ e T = 0 1 0
e ·~
~ ~ (1.61)
0 0 1

1.3.4 Scalar functions of a tensor

Euclidean norm

The Euclidean norm7 m of a tensor, a measure of its magnitude, is defined by:

ek ∀ ~
m = kAk = max kA·~ e with k~
e k = 1. (1.62)
~
e

The norm is an invariant quantity8 , i.e. it is independent of the choice of the basis. With
respect to the norm of a tensor, the following properties hold:

kAk ≥ 0 (1.63)
kαAk = |α| kAk (1.64)
kA·B k ≤ kAk kB k (1.65)
kA + B k ≤ kAk + kB k (1.66)

The first invariant

The first invariant of a tensor A, also called the trace9 of A, is defined as:

J 1 (A) = tr(A)
1
= ~
c1 ×~
[~
c ·A·(~
c3 1
c2 ·~
c3 )+
c2 ×~
~
c2 ·A·(~ c1 ) +~
c3 ×~ c3 ·A·(~ c2 )]
c1 ×~ (1.67)

7
Euclidische norm
8
invariant
9
spoor

11
1. T ENSOR CALCULUS

where {~
c1 ,~
c2 ,~
c3 } represents a vector basis. Since J 1 (A) is an invariant quantity, it can also be
calculated using an orthonormal basis {~ e 1 ,~
e 2 ,~
e 3 }:

J 1 (A) = ~
e 1 ·A·~
e 1 +~
e 2 ·A·~
e 2 +~
e 3 ·A·~
e3 (1.68)
= A 11 + A 22 + A 33 (1.69)
= trA (1.70)

The following properties hold for the first invariant:

J 1 (A) = J 1 (A T ) (1.71)
J 1 (I ) = 3 (1.72)
J 1 (αA) = αJ 1 (A) (1.73)
J 1 (A + B ) = J 1 (A) + J 1 (B ) (1.74)

The second invariant

The second invariant of a tensor A is defined as:

J 2 (A) = 12 {tr2 (A)−tr(A 2 )}

1
= ~
c1 ×~
c2 ·~
c3
[ ~ c2 ) × (A·~
c1 ·(A·~ c3 )+
~ c3 ) × (A·~
c2 ·(A·~ c1 )+
~ c1 ) × (A·~
c3 ·(A·~ c2 )] (1.75)

The following properties hold for the second invariant:

J 2 (A) = J 2 (A T ) (1.76)
J 2 (I ) = 3 (1.77)
2
J 2 (αA) = α J 2 (A) (1.78)

The third invariant

The third invariant of a tensor A, also called the determinant of A, is defined as:
1
J 3 (A) = det(A) = ~
c1 ×~
c2 ·~
c3
(A·~
c1 )·(A·~
c2 ) × (A·~
c3 ) (1.79)

For the third invariant, the following properties hold:

J 3 (A) = J 3 (A T ) (1.80)
J 3 (I ) = 1 (1.81)
3
J 3 (αA) = α J 3 (A) (1.82)
J 3 (A·B ) = J 3 (A)J 3 (B ) (1.83)

The determinant of a tensor equals the determinant of its matrix with respect to an orthonor-
mal basis.
det(A) = det(A) (1.84)
If det(A) is nonzero, i.e. the triple product of the transformed base vectors is nonzero, the
tensor is called a regular tensor10 . If det(A) is zero, the tensor A is called a singular tensor11 .
10
reguliere tensor
11
singulier

12
 1.3. Second order tensors

In the latter case all vectors in three-dimensional space are transformed into vectors lying in
a plane (or on a line). Vectors perpendicular to that plane (or that line), are transformed into
the zero vector.

Scalar product

The scalar product, double dot product or double inner product or convolution12 of two ten-
sors A and B is the contraction of these two tensors, denoted by A : B , is a scalar, defined
as:
A : B = tr(A·B ) (1.85)
In index notation, this product is the full contraction of indices

A : B = Ai j B j i (1.86)

Some properties are:

A : I = tr(A) (1.87)
A:B = B : A (1.88)
= AT : B T (1.89)
= B T : AT (1.90)

Some properties of the invariants of a tensor

The three principal invariants of a tensor are related through the Caley-Hamilton theorem:

A 3 − J 1 (A)A 2 + J 2 (A)A − J 3 (A)I = 0 (1.91)

The conjugated tensor A T satisfies the same equation.

For any non-singular tensor it holds that:

J 2 (A)
J 1 (A −1 ) = (1.92)
J 3 (A)
J 1 (A)
J 2 (A −1 ) = (1.93)
J 3 (A)
1
J 3 (A −1 ) = (1.94)
J 3 (A)

Any function of the principal invariants of a tensor A is an invariant as well.

1.3.5 Eigenvalues and eigenvectors of a tensor

The eigenvalues13 λ and the eigenvectors14 ~ n of a tensor A are defined as those scalars and
vectors that satisfy the following equation:

n = λ~
A·~ n with n 6=~0
~ (1.95)
12
dubbel inwendig produkt
13
eigenwaarden
14
eigenvectoren

13
1. T ENSOR CALCULUS

To compute the eigenvalues and eigenvectors of a tensor, we proceed as follows. First equa-
tion (1.95) is rewritten as:

n = λ~
A·~ n ⇒
A·~ n = ~0
n − λ~ ⇒
A·~ n = ~0
n − λI ·~ ⇒
n = ~0
(A − λI )·~ (1.96)

n 6=~0, the tensor A − λI must be singular, meaning that:

Since ~

det(A − λI ) = 0 ⇒ det(A − λI ) = 0 (1.97)

Elaboration of the latter equation leads to the characteristic equation:

λ3 − J 1 (A)λ2 + J 2 (A)λ − J 3 (A) = 0 (1.98)

This equation has three roots λ1 , λ2 and λ3 . The eigenvector ~ n i corresponding with the
eigenvalue λi is found from the following set of linear equations:

n i = ~0
(A − λi I )·~ ⇒ (A − λi I )n 0
~i = ~ (1.99)

Since this is not an independent set of equations, it yields only the ratio n i 1 : n i 2 : n i 3 of the
components of the eigenvector. A unique solution is obtained by the additional requirement
that the eigenvector has unit length:

ni k = 1
k~ ⇒ n i21 + n i22 + n i23 = 1 (1.100)

The principal invariants of a tensor are then equal to:

J 1 (A) = λ1 + λ2 + λ3 (1.101)
J 2 (A) = λ1 λ2 + λ1 λ3 + λ2 λ3 (1.102)
J 3 (A) = λ1 λ2 λ3 (1.103)

1.3.6 Some special tensors

Inverse tensor

The inverse tensor A −1 of a tensor A is defined as the tensor that satisfies the following equa-
tion:
A −1 ·A = I (1.104)
A −1 only exists if A is regular. The components of the inverse tensor are given by the inverse
matrix A −1
−1 1 h i +j
iT
[A ]i j = (−1) Mi j (1.105)
det(A)
where Mi j is the minor of the component A i j , i.e., the determinant of the matrix that is
obtained by deleting row i and column j .
The inverse of the product of two tensors is given by:

(A·B )−1 = B −1 ·A −1 (1.106)

14
 1.3. Second order tensors

The proof is obtained by comparing the following expressions for ~

a:

a =~
(A·B )·~ b ⇒
a = (A·B )−1 ·~
~ b (1.107)

and

(A·B )·~
a = A·(B ·~a) = ~
b ⇒
a = A −1 ·~
B ·~ b ⇒
a = B −1 ·A −1 ·~
~ b (1.108)

Deviatoric part of a tensor

The deviatoric part15 A d of a tensor A is defined as:

A d = A − 13 tr(A)I (1.109)

where 31 tr(A)I is called the hydrostatic or spherical16 part of A. For the deviatoric part, the
following properties hold:

(A + B )d = A d + B d (1.110)
tr(A d ) = 0 (1.111)

Symmetric tensor

A tensor A is called symmetric if:

AT = A (1.112)
For the components A i j of a symmetric tensor A with respect to an orthonormal basis it
holds that:
Ai j = A j i (1.113)
i.e. the matrix A corresponding to A is symmetric too. The eigenvalues λi and eigenvectors
~
n i of a symmetric tensor have the following important properties:

• the eigenvalues and eigenvectors are real

• if two eigenvalues are different, the corresponding eigenvectors are mutually perpen-
dicular

• if two eigenvalues are equal, the corresponding eigenvectors can be chosen mutually
perpendicular

Thus, the set of eigenvectors can be used as an orthonormal basis. If a tensor A is considered
with respect to the basis of its eigenvectors, we find the spectral form17 of the tensor:

A = A·I (1.114)
n 1~
= A·(~ n 2~
n 1 +~ n 3~
n 2 +~ n3) (1.115)
= λ1~
n 1~
n 1 + λ2~
n 2~
n 2 + λ3~
n 3~
n3 (1.116)
15
deviatorisch deel
16
hydrostatisch of sferisch deel
17
spectrale vorm

15
1. T ENSOR CALCULUS

r
n3
r
n2

r
n1

Figure 1.13: The orthonormal basis of eigenvectors {~

n 1 ,~
n 2 ,~
n 3 } of a symmetric tensor.

The matrix A of A with respect to this basis is a diagonal matrix:

 
λ1 0 0
A =  0 λ2 0  (1.117)
0 0 λ3
Using the spectral form, some functions of a tensor can easily be calculated. E.g. for the
inverse, the square root, and the natural logarithm of a tensor A it holds that:
A −1 = λ−1
1 ~ n 1 + λ−1
n 1~ 2 ~ n 2 + λ−1
n 2~ 3 ~n 3~
n3 (1.118)
1/2
A = λ1/2
1 ~ n 1 + λ1/2
n 1~ 2 ~ n 2 + λ1/2
n 2~ 3 ~ n 3~
n3 (1.119)
n 1~
ln A = ln(λ1 )~ n 2~
n 1 + ln(λ2 )~ n 3~
n 2 + ln(λ3 )~ n3 (1.120)

Skewsymmetric tensor

A tensor A is called skewsymmetric18 or antisymmetric if:

A T = −A (1.121)
For the components A i j of a skewsymmetric tensor A with respect to an orthonormal basis
it holds that:
A i j = −A j i ⇒ A i i = 0 (1.122)
If A is skewsymmetric and B is symmetric, then:
A : B = 0 especially A : I = trA = 0 (1.123)
The proof follows from:
A : B = A T : B T = −A : B (1.124)
19
If A is skewsymmetric, then a unique axial vector ~
ω exists, such that:
A·~ ω×~
q =~ q (1.125)
The proof of the existence of ~
ω is as follows:
A·~ ~
q =p ⇒
~ q ·A T ·~
q =~
q ·A·~ q
= −~
q ·A·~
q ⇒
~ q =0
q ·A·~ ⇒
~ p =0
q ·~ ⇒
~
q ⊥ ~
p ⇒ ∃~
ω such that p ω×~
~=~ q (1.126)
18
scheefsymmetrisch
19
axiaalvector

16
 1.3. Second order tensors

On the other hand, if:

A·~ ω×~
q =~ q ∀~
q (1.127)
comparison of the following expressions:
  
£ ¤ A 11 A 12 A 13 q1
q = ~
A·~ e1 ~
e2 ~
e 3 A 21 A 22 A 23
   q2 
A 31 A 32 A 33 q3
 
£ ¤ A 11 q 1 + A 12 q 2 + A 13 q 3
= ~ e1 ~
e2 ~
e 3  A 21 q 1 + A 22 q 2 + A 23 q 3  (1.128)
A 31 q 1 + A 32 q 2 + A 33 q 3

and

ω×~
~ q = (ω1~
e 1 + ω2~
e 2 + ω3~
e 3 ) × (q 1~
e 1 + q 2~
e 2 + q 3~
e 3)
= ω1 q 2 (~
e 3 ) + ω1 q 3 (−~
e 2 ) + ω2 q 1 (−~
e 3 ) + ω2 q 3 (~
e 1 )+
ω3 q 1 (~
e 2 ) + ω3 q 2 (−~
e 1)
 
£ ¤ ω2 q 3 − ω3 q 2
= ~
e1 ~ e2 ~ e 3 ω3 q 1 − ω1 q 3  (1.129)
ω1 q 2 − ω2 q 1

shows that:  
0 − ω3 ω2
A =  ω3 0 − ω1  (1.130)
− ω2 ω1 0
and thus that the tensor A is skewsymmetric.

Positive-definite tensor

A tensor A is called positive-definite20 if:

~ a 6=~0
a > 0 ∀~
a ·A·~ (1.131)

A positive-definite tensor A cannot be skewsymmetric. This follows from:

~ a ·A T ·~
a =~
a ·A·~ a
a 6=~0
> 0 ∀~ ⇒ (1.132)
T
~
a ·(A − A )·~
a =0 ∧ a 6=~0
~ (1.133)

If A would be skewsymmetric, the latter equation would yield zero, since then A T = −A.

Orthogonal tensor

A tensor A is called orthogonal21 if:

a )·(A·~
(A·~ a ·~
b) = ~ b a ,~
∀~ b (1.134)
20
positief definiet
21
orthogonale tensor

17
1. T ENSOR CALCULUS

For the special case ~

b =~
a we get:
(A·~ a) = ~
a )·(A·~ a ·~
a ⇒ a k = k~
kA·~ ak (1.135)
Thus, an orthogonal transformation preserves lengths of vectors and angles between vectors.
The inverse of an orthogonal tensor is identical to its conjugate:
A T = A −1 (1.136)
as follows from:
a )·(A·~
(A·~ a ·A T ·A·~
b) = ~ b
=~a ·~
b ⇒
A·A T = I ⇒
T −1
A = A (1.137)
Moreover, this equation shows that:
det(A·A T ) = [det(A)]2 = det(I ) = 1 ⇒
det(A) = ±1 (1.138)
Thus, an orthogonal tensor is a regular tensor. If det(A) = 1, A can be shown to describe a
rotation. If det(A) = −1, A can be shown to describe a reflection.

1.4 Spatial derivatives of scalar and vector fields

1.4.1 Derivatives of functions of one or more variables

f ( x + dx )

df
df
dx
dx
f ( x)
dx

x x + dx

Figure 1.14: A function f of one variable x and its derivative.

Consider a scalar function f (x) of one variable x. Using a Taylor series, the change d f of
the value of f due to a change d x in the parameter x can be expressed as:
d f = f (x + d x) − f (x)
¯ ¯
df ¯ d2 f ¯
= f (x) + d x ¯ d x + 12 ¯ d x2 + · · · −
d x2 x
f (x)
¯ x
df ¯
≈ dx ¯ d x (1.139)
x

18
 1.4. Spatial derivatives of scalar and vector fields

Extension to a function of two variables f (x, y) yields:

d f = f (x + d x, y + d y) − f (x, y)
¯ ¯
∂f ¯ ∂f ¯
= f (x, y) + ∂x ¯ d x + ∂y ¯ d y + · · · − f (x, y)
x,y x,y
¯ ¯
∂f ¯ ∂f ¯
≈ ∂x ¯ d x + ∂y ¯ d y (1.140)
x,y x,y

And for a function a(x1 , x2 , x3 ) of three variables, the variation d a of a, due to (infinitesimal)
variations d x1 , d x2 and d x3 can be written as:

∂a ∂a ∂a
d a = d x1 ∂x + d x2 ∂x + d x3 ∂x (1.141)
1 2 3

The index notation for differentiation with respect to one of the variables (i.c. xi ) is com-
monly denoted with a comma between or in front of the indices. The equation (1.141) for d a
with this convention reads
d a = d xi a,i (1.142)

1.4.2 Components of the position vector

In three-dimensional space, we may interpret the variables x1 , x2 and x3 in the function
a(x1 , x2 , x3 ) as the components of the position vector22 ~ x with respect to a Cartesian basis
{~
e 1 ,~ e 3 }. The position vector ~
e 2 ,~ x of a point with respect to a fixed point in space, the origin
O, can be expressed as:
~x = x1~e 1 + x2~
e 2 + x3~
e3 (1.143)
For the position vector ~ x of a neighbouring point we can write:
x + d~

x3
r
e3
r
dx
r
r e2
r x
e3 r
e1
r
e2 x2
r
e1
x1

Figure 1.15: Components xi of the position vector ~

x with respect to a Cartesian basis
{~
e 1 ,~
e 2 ,~
e 3 }.

~ x = (x1 + d x1 )~
x + d~ e 1 + (x2 + d x2 )~
e 2 + (x3 + d x3 )~
e3 (1.144)
22
positievector

19
1. T ENSOR CALCULUS

and thus the vector d~

x can be expressed as:

x = d x1~
d~ e 1 + d x2~
e 2 + d x3~
e3 (1.145)

From the latter relation it is obvious that the changes d x1 , d x2 and d x3 can be related to the
vector d~
x according to:

d x1 = d~
x ·~
e1 (1.146)
d x2 = d~
x ·~
e2 (1.147)
d x3 = d~
x ·~
e3 (1.148)

We could also express the function a(~ x ) in terms of the components r , θ, z of the position
vector ~
x with respect to a cylindrical basis {~
e r ,~
e t ,~
e z }. Then the position vector equals:

~ e r (θ) + z~
x = r~ ez (1.149)

Within the cylindrical basis, the position ~ x of a neighbouring point equals:

x + d~

x3 r
ez
z r
et
r
dx
r
r er
r
e3 x
r
e2 x2

r
e1
q r
x1

Figure 1.16: Components r , θ, z of the position vector ~

x with respect to a cylindrical basis
{~
e r ,~
e t ,~
e z }.

~ e r (θ + d θ) + (z + d z)~
x = (r + d r )~
x + d~ ez
d~
er
= (r + d r ){~
e r (θ) + dθ d θ} + (z + d z)~
ez
e r (θ) + z~
= r~ ez +
e t (θ)d θ + d r~
r~ e t (θ)d r d θ + d z~
e r (θ) +~ ez (1.150)

from which it can be concluded that, in a linear approximation, the vector d~

x equals:

d~ e r (θ) + r d θ~
x ≈ d r~ e t (θ) + d z~
ez (1.151)

20
 1.4. Spatial derivatives of scalar and vector fields

The changes d r , d θ and d z are related to the vector d~

x according to:

d r = d~
x ·~
er (1.152)
1
dθ = r d~
x ·~
et (1.153)
d z = d~
x ·~
ez (1.154)

1.4.3 The gradient operator

Considering a as a function of the position vector ~
x , we can investigate the variation d a due
to a variation d~
x:
∂a ∂a ∂a
d a = d x1 ∂x 1
+ d x2 ∂x 2
+ d x3 ∂x 3
∂a ∂a ∂a
= (d~ e 1 ) ∂x
x ·~ + (d~ e 2 ) ∂x
x ·~ + (d~ e 3 ) ∂x
x ·~
1 2 3
h i
∂a ∂a ∂a
= d~x· ~ e 1 ∂x1 +~ e 2 ∂x2 +~ e 3 ∂x3
x ·(~
= d~ ∇a) (1.155)

The symbol ~
∇ is called the gradient operator23 :
h i
~
∇= ~ e 2 ∂x∂2 +~
e 1 ∂x∂ 1 +~ e 3 ∂x∂3 (1.156)
eT ∇
=~
~ ~ (1.157)

and the expression ~∇a is called the gradient of a. The gradient of a vector function ~
a is
obtained similarly:
∂~
a ∂~
a ∂~
a
d~
a = d x1 ∂x 1
+ d x2 ∂x 2
+ d x3 ∂x 3
∂~
a ∂~
a ∂~
a
= (d~ e 1 ) ∂x
x ·~ 1
+ (d~ e 2 ) ∂x
x ·~ 2
+ (d~ e 3 ) ∂x
x ·~ 3
h i
∂~
a ∂~
a ∂~
a
= d~x· ~ e 1 ∂x +~ e 2 ∂x +~ e 3 ∂x
1 2 3

x ·(~
= d~ ∇~a) (1.158)

or in index notation
d a j = d xi a j ,i (1.159)
For the special case ~
a =~
x we find:
~
∇~x = I (1.160)
The gradient operator ~ ∇ can be applied to a vector ~ a in several manners, the result of which
24
is called the gradient of ~
a (grad ~
a ), the divergence of ~ a ), or the rotation25 or curl of ~
a (div ~ a
(rot ~
a ):
∂~
a ∂~
a ∂~
a
a =~
grad ~ ∇~a =~
e 1 ∂x 1
e 2 ∂x
+~ 2
e 3 ∂x
+~ 3
, (1.161)
∂~
a ∂~
a ∂~
a
a =~
div ~ a =~
∇·~ e 1 · ∂x 1
e 2 · ∂x
+~ 2
e 3 · ∂x
+~ 3
, (1.162)
∂~
a ∂~
a ∂~
a
a =~
rot ~ ∇×~
a =~
e 1 × ∂x e 2 × ∂x
+~ e 3 × ∂x
+~ (1.163)
1 2 3

23
gradiënt
24
divergentie
25
rotatie

21
1. T ENSOR CALCULUS

Note that div ~a yields a scalar, rot ~

a yields a vector, and grad ~
a yields a second order tensor.
Applied to tensors in general, the gradient increases the order by 1, the divergence dimin-
ishes the order by 1, while the order remains unaffected by the curl.
The expression for the gradient operator in cylindrical coordinates can be found follow-
ing a similar derivation as in the case of a Cartesian coordinate system:

d a = d r ∂a ∂a ∂a
∂r + d θ ∂θ + d z ∂z
= (d~ e r ) ∂a
x ·~ 1
∂r + ( r d~ e t ) ∂a
x ·~ ∂θ + (d~ e z ) ∂a
x ·~ ∂z
h i
= d~x· ~ e r ∂a
∂r
e t ∂a
+ r1~ ∂θ
+~ e z ∂a
∂z

x ·(~
= d~ ∇a) (1.164)

Thus the expression for the gradient operator in a cylindrical coordinate system equals:

∂ 1 ∂ ∂
~
∇=~
e r ∂r et
+~ e z ∂z
+~ (1.165)
r ∂θ

1.5 Tensor formula involving first and second-order tensors

a ,~
The following relations on differentiable scalar functions φ, vectors ~ b, tensors A, B ,C , in-
ner/outer products and their derivatives can be proven:

A : (B ·C ) = (A·B ) : C (1.166)
~ a ) = (~
∇(φ~ a + φ~
∇φ)~ ∇~a (1.167)
~
∇·(φ~ a ) = (~
∇φ)·~ a + φ~ a
∇·~ (1.168)
~
∇(~a ·~
b) = (~∇~a )·~ ∇~
b + (~ b)·~
a (1.169)
~ a~
∇·(~ b) = (~ a )~
∇·~ ∇~
a ·~
b +~ b (1.170)
~
∇·(φA) = (~∇φ)·A + φ~ ∇·A (1.171)
~ T
a ) = (~
∇·(A·~ a + A : (~
∇·A)·~ ∇~a) (1.172)
~ a = (~
∇·~ ∇~a ) : I = tr(~
∇~a) (1.173)

1.6 Higher-order and fourth-order tensors

1.6.1 Definition
Some problems invoke the transformation of tensors. Such transformations are carried out
by means of higher-order tensors. A general definition of a tensor could be: an invariant
object with an arbitrary number of indices. The number of indices is the order (or rank)
of the tensor. The transformation which is most relevant in the mechanics of materials is
the transformation of a second-order tensor into another second-order tensor by means of
a fourth-order tensor. An arbitrary tensor A (which is a linear transformation of vectors)
is mapped onto a different tensor B (which is another linear transformation of vectors) by
4
means of the double inner product with a fourth-order tensor C , see figure 1.17.
4
B = C:A (1.174)

22
 1.6. Higher-order and fourth-order tensors

4
r
c
r
a

r
b
C r
g
r
h

r
i

r r
e = A ×b A B r r
l = B ×i

r
r r k = B ×h
d = A×a
r r
r r j = B×g
f = A ×c
4
Figure 1.17: A fourth-order tensor C that maps a (linear) second-order tensor A onto a
second-order tensor B .

Using index tensor notations, the tensor A l k transforms to B i j according to

B i j = C i j kl A l k (1.175)

The tensor C i j kl which was used to perform this transformation has 4 indices.

1.6.2 Properties
Since a fourth-order tensor is built on quadrades rather than dyades, some new calculus
rules apply. Each fourth-order tensor can be written as a finite sum of quadrades. With
4
respect to a two-dimensional Cartesian basis, a tensor C would be given by si xt een terms,
i.e.,

4
C = C 1111~
e 1~
e 1~
e 1~
e 1 +C 1112~
e 1~
e 1~
e 1~
e 2 +C 1121~
e 1~
e 1~
e 2~
e 1 +C 1122~
e 1~
e 1~
e 2~
e2
C 1211~
e 1~
e 2~
e 1~
e 1 +C 1212~
e 1~
e 2~
e 1~
e 2 +C 1221~
e 1~
e 2~
e 2~
e 1 +C 1222~
e 1~
e 2~
e 2~
e2
(1.176)
C 2111~
e 2~
e 1~
e 1~
e 1 +C 2112~
e 2~
e 1~
e 1~
e 2 +C 2121~
e 2~
e 1~
e 2~
e 1 +C 2122~
e 2~
e 1~
e 2~
e2
C 2211~
e 2~
e 2~
e 1~
e 1 +C 2212~
e 2~
e 2~
e 1~
e 2 +C 2221~
e 2~
e 2~
e 2~
e 1 +C 2222~
e 2~
e 2~
e 2~
e2

4
In a three-dimensional space a tensor C has 3 × 3 × 3 × 3 = 81 components. The correct full
matrix representation of a higher-order tensor can only be achieved with multi-dimensional
matrices (which exist in modern computing languages). A fourth-order 3D tensor is then
represented by a [3 × 3 × 3 × 3] matrix.
4 T
Different types of transposed tensors are defined: the total transpose C , the left trans-
4 LT 4 RT 4
pose C or the right transpose C . If the tensor C is denoted C i j kl in index notation,

23
1. T ENSOR CALCULUS

each of these transposes is given by

4 T
C ⇔ Cl k j i (1.177)
4 LT
C ⇔ C j i kl (1.178)
4 RT
C ⇔ Ci j l k (1.179)

Different types of minor symmetries can be defined for a fourth-order tensor. A left symmet-
ric fourth-order tensor, is a tensor for which
4 4 LT
C = C =⇒ C i j kl = C j i kl (1.180)

In a 3D space, it has 81−1×3×3×3 = 54 distinct components. A right symmetric fourth-order

tensor is given by
4 4 RT
C = C =⇒ C i j kl = C i j l k (1.181)
In a 3D space, it has 81 − 3 × 3 × 3 × 1 = 54 distinct components. A tensor which is left and
right symmetric, has only 54 − 1 × 3 × 3 × 2 = 36 distinct components. A middle symmetric
fourth-order tensor complies with
4 4 T
C = C =⇒ C i j kl = C l k j i (1.182)

In a 3D space, it has 81−2×3×3×2 = 45 distinct components. A total symmetric fourth-order

tensor equals is left, right and central symmetric, i.e.,
4 4 LT 4 RT 4 T
C = C = C = C (1.183)

In a 3D space, it has 36 − 12 6 × 5 = 21 distinct components.

4
The fourth-order null tensor 0 is defined by
4
0:A = 0 (1.184)
4
The fourth-order unit tensor I maps a second-order tensor A on itself
4
I:A = A (1.185)

This tensor is given in index notation by

4
I =⇒ I i j kl = δi l δ j k (1.186)

or with respect to a 3D Cartesian basis

4
e 1~
I =~ e 1~
e 1~ e 1~
e 1 +~ e 2~
e 2~ e 1~
e 1 +~ e 3~
e 3~
e1
~
e 2~
e 1~
e 1~ e 2~
e 2 +~ e 2~
e 2~ e 2~
e 2 +~ e 3~
e 3~
e2 (1.187)
~
e 3~
e 1~
e 1~ e 3~
e 3 +~ e 2~
e 2~ e 3~
e 3 +~ e 3~
e 3~
e3
4 s
The symmetric fourth-order unit tensor I maps a second-order tensor A on the sym-
metric part of A, denoted A s :
4 s
I : A = A s = 21 (A + A T ) (1.188)

24
 1.6. Higher-order and fourth-order tensors

This tensor is given in index notation by

4 s ¡ ¢
I ⇔ I isj kl = 1
2 δi l δ j k + δi k δ j l (1.189)

This tensor is a total symmetric tensor, i.e., left symmetric, right symmetric and central sym-
metric. Note that this tensor is also given by
4 s 4 4 RT
I = 12 ( I + I ) (1.190)

or with respect to a Cartesian basis by

4 s
e 1~
I =~ e 1~ e 1 + 21~
e 1~ e 1~
e 2~ e 1 + 21~
e 2~ e 1~
e 2~ e 2 + 21~
e 1~ e 1~
e 3~ e 1 + 21~
e 3~ e 1~
e 3~
e 1~
e3
1 1
~~~~
2 e 2e 1e 1e 2 +
~ ~ ~ ~ e 2~
2 e 2 e 1 e 2 e 1 +~ e 2~e 2~e 2 + 21~e 2~
e 3~e 3~e 2 + 21~
e 2~
e 3~
e 2~
e3 (1.191)
1 1 1 1
~~~~
2 e 3e 1e 1e 3 +
~~~~ ~~~~ ~ ~ ~ ~ e 3~
2 e 3 e 1 e 3 e 1 + 2 e 3 e 2 e 2 e 3 + 2 e 3 e 2 e 3 e 2 +~ e 3~
e 3~
e3

Fourth-order tensors can be partially modified through inner products with second-order
tensors or double inner products with other fourth-order tensors. The inner product be-
4
tween a fourth-order tensor A and a second-order tensor B yields a fourth-order tensor
given by
4 4
C = A·B (1.192)
C i j kl = A i j km B ml (1.193)
4
The double inner product between a fourth-order tensor A and another fourth-order tensor
4 4
B again leads to a fourth-order tensor C according to
4 4 4
C = A: B (1.194)
C i j kl = A i j mn B nmkl (1.195)

Note that, in general, the double inner product used in combination with fourth-order ten-
sors is no longer commutative
4 4
A : B 6= B : A (1.196)
4 4 4 4
A : B 6= B : A (1.197)

1.6.3 Tensor rules

4 4
A : (B ·C ) = ( A·B ) : C (1.198)
4 s
A·B + B T ·A T = I : (A·B ) (1.199)
4 s
= ( I ·A) : B (1.200)

1.6.4 Matrix condensation of the double inner product with a

second-order tensor
In many numerical implementations, tensor products are condensed in a two-dimensional
matrix format once a Cartesian basis has been chosen. Making use of special properties

25
1. T ENSOR CALCULUS

of the tensors involved then permits to omit a considerable amount of computations for a
double inner product. Components of a second-order tensor B are normally represented as
a matrix
 
B 11 B 12 B 13
B = B 21 B 22 B 32  (1.201)
B 31 B 32 B 33

For the computation of a double inner product with a fourth-order tensor, it is often more
convenient to store the 9 components in a column B ~ Taking into account the symmetry of a
second-order tensor B permits to store the 6 distinct components in a column rather than
in a matrix, e.g.,
 
B 11
B 12 
 
B 
 13 
B 
 21 
 
B
~ = B 22  (1.202)
 
B 23 
 
B 31 
 
B 32 
B 33

On the basis of this notation, it is possible to obtain the components of second-order tensor
C as a similarly structured column, which is the double inner product between a fourth-
4
order tensor A and a second-order tensor B , as a matrix-column product, i.e., the compo-
nents of
4
C = A:B (1.203)

are obtained through

C
~ = AB~ (1.204)
 C 11   A 1111 A 1121 A 1131 A 1112 A 1122 A 1132 A 1113 A 1123

A 1133 B 11 
C 12 A 1211 A 1221 A 1231 A 1212 A 1222 A 1232 A 1213 A 1223 A 1233
 C 13   A 1311 A 1321 A 1331 A 1312 A 1322 A 1332 A 1313 A 1323
 B12
A 1333   B 13 
 C 21   A 2111 A 2121 A 2131 A 2112 A 2122 A 2132 A 2113 A 2123 A 2133   
C    B21 
 22  =   A 2211 A 2221 A 2231 A 2212 A 2222 A 2232 A 2213 A 2223 A 2233   B 22 
 (1.205)
 C 23   A 2311 A 2321 A 2331 A 2312 A 2322 A 2332 A 2313 A 2323 A 2333   B 23 
 C 31   A 3111 A 3121 A 3131 A 3112 A 3122 A 3132 A 3113 A 3123 A 3133   B 31 
C 32 A 3211 A 3221 A 3231 A 3212 A 3222 A 3232 A 3213 A 3223 A 3233 B 32
C 33 A 3311 A 3321 A 3331 A 3312 A 3322 A 3332 A 3313 A 3323 A 3333 B 33

4
Note that the matrix A has indeed 81 components (all the components of A), although this
matrix is only valid in conjunction with a double inner product with a second-order tensor.
In the particular case of mechanics, symmetric second-order tensors play an important
role. Taking into account the symmetry of a second-order tensor B permits to store the 6
distinct components in a smaller column with a somewhat different structure
 
B 11
B 
 22 
 
B 33 
B
~ =   (1.206)
B 12 
 
B 13 
B 23

26
 1.6. Higher-order and fourth-order tensors

If the resulting tensor C is not symmetric, it is represented by a column with 9 components.

The matrix A is restructured and condensed to a [9 × 6] matrix by adding up the appropri-
ate columns in equation (1.205), i.e., if B
~ has the format (1.206) and C ~ maintains the for-
4
mat (1.202) then C = A : B is represented in component form by

~ = AB
C ~ (1.207)
 C 11   A 1111 A 1122 A 1133 (A 1112 +A 1121 ) (A 1113 +A 1131 )

(A 1123 +A 1132 )
C 12 A 1211 A 1222 A 1233 (A 1212 +A 1221 ) (A 1213 +A 1231 ) (A 1223 +A 1232 ) 
 C 13   A 1311 A 1322 A 1333 (A 1312 +A 1321 ) (A 1313 +A 1331 )
 B11 
(A 1323 +A 1332 ) 
 C 21   B
   A 2111 A 2122 A 2133 (A 2112 +A 2121 ) (A 2113 +A 2131 ) (A 2123 +A 2132 )   22 
B
 C 22  =  A 2211 A 2222 A 2233 (A 2212 +A 2221 ) (A 2213 +A 2231 ) (A 2223 +A 2232 )   33  (1.208)

 C 23   A 2311 A 2322 A 2333 (A 2312 +A 2321 ) (A 2313 +A 2331 )
 B12
(A 2323 +A 2332 ) 
 C 31   A 3111 B 13
A 3122 A 3133 (A 3112 +A 3121 ) (A 3113 +A 3131 ) (A 3123 +A 3132 ) 
B 23
C 32 A 3211 A 3222 A 3233 (A 3212 +A 3221 ) (A 3213 +A 3231 ) (A 3223 +A 3232 )
C 33 A 3311 A 3322 A 3333 (A 3312 +A 3321 ) (A 3313 +A 3331 ) (A 3323 +A 3332 )

If the fourth-order tensor presents a symmetry, the equations can be simplified considerably.
A right symmetric tensor has only 54 distinct components (3 columns equal 3 other columns
in the matrix A). A left symmetric tensor has also only 54 distinct components (3 lines equal
3 other lines in the matrix A). A tensor which is left and right symmetric has only 36 inde-
pendent components (only 6 lines and columns remain in the matrix A). The case where
4 4
the tensor A is left symmetric results in a tensor C which is symmetric. Applying C = A : B
again to a symmetric B can then be done with columns B ~ and C
~ of the form (1.206). The
result in this particular case is then condensed as

~ = AB
C ~ (1.209)
 C 11   A 1111 A 1122 A 1133 A 1112 +A 1121 A 1113 +A 1131 A 1123 +A 1132

B 11 
C 22 A 2211 A 2222 A 2233 A 2212 +A 2221 A 2213 +A 2231 A 2223 +A 2232
 C 33   A 3311 A 3322 A 3333 A 3312 +A 3321 A 3313 +A 3331
 B22
A 3323 +A 3332   B 33 
 C 12  =  A 1211

A 1222 A 1233 A 1212 +A 1221 A 1213 +A 1231 A 1223 +A 1232   B 12  (1.210)
C 13 A 1311 A 1322 A 1333 A 1312 +A 1321 A 1313 +A 1331 A 1323 +A 1332 B 13
C 23 A 2311 A 2322 A 2333 A 2312 +A 2321 A 2313 +A 2331 A 2323 +A 2332 B 23

Other simplifications are easy to make and left to the reader.

27
1. T ENSOR CALCULUS

1.7 Summary

Vector operations

Multiplication ~
b = α~
a

Sum ~ a +~
c =~ b

Inner product a ·~
~ a || ||~
b = ||~ b|| cos(φ)
~ ~ ~
a ·b = b·~a

Cross product a ×~
~ b =~ a || ||~
c = ||~ b|| sin(φ)
~
b ×~
a = −~a ×~b

Triple product a ×~
~ b·~ a || ||~
c = ||~ b|| sin(φ)||~
c || cos(ψ)

Tensor product (dyad) a~

~ b
a~
~ a (~
p =~
b·~ p)
b·~
(~~ T ~
a b) = b~a

Gradient operators

Cartesian ~ e 1 ∂x∂1 +~
∇ =~ e 2 ∂x∂2 +~
e 3 ∂x∂3

∂~
a ∂~
a ∂~
a
grad ~
a a =~
grad ~ ∇~ a =~
e 1 ∂x 1
e 2 ∂x
+~ 2
e 3 ∂x
+~ 3
~
∇~a = a j ,i ~
e i~
ej

∂~
a ∂~
a ∂~
a
div ~
a a =~
div ~ a =~
∇·~ e 1 · ∂x 1
e 2 · ∂x
+~ 2
e 3 · ∂x
+~ 3

∂~
a ∂~
a ∂~
a
rot ~
a a =~
rot ~ ∇×~
a =~
e 1 × ∂x 1
e 2 × ∂x
+~ 2
e 3 × ∂x
+~ 3

∂ ∂ ∂
Cylindrical ~
∇ =~
e r ∂r e t r1 ∂θ
+~ e z ∂z
+~

28
 1.7. Summary

Second-order Tensors

Second order tensor p =~

A·~ q
A·(αm~ + β~ n ) = αA·m ~ + βA·~
n
(αA)·~v = A·(α~ v)
(A + B )·~
v = A·~ v + B ·~
v
(A·B )·~
v = A·(B ·~ v)

Conjugate of tensor A T ·~
p=~
p ·A

Unit tensor e 1~
I =~ e 2~
e 1 +~ e 3~
e 2 +~ e3

First invariant J 1 (A) = tr(A)

Second invariant J 2 (A) = 21 {tr2 (A) − tr(A 2 )}
Third invariant J 3 (A) = det(A)

Double dot product A : B = tr(A·B )

A : I = tr(A)
A : B = AT : B T

Inverse tensor A −1 ·A = I
(A·B )−1 = B −1 ·A −1

Deviatoric part A d = A − 13 tr(A)I

Symmetric tensor AT = A

Skew-symmetric tensor A T = −A

Positive definite tensor ~ a 6=~0

a >0 ∀~
a ·A·~

Orthogonal tensor a )·(A·~

(A·~ a ·~
b) = ~ b a ,~
∀~ b

29
1. T ENSOR CALCULUS

Fourth-order Tensors

4
Transformation B= C:A
Components [3 × 3 × 3 × 3] matrix

4 T
Total transpose C = [C i j kl ]T = C l k j i
4 LT
Left transpose C = [C i j kl ]LT = C j i kl
4 RT
Right transpose C = [C i j kl ]RT = C i j l k

4 4 LT
Left symmetric C= C
4 4 RT
Right symmetric C= C
4 4 T
Middle symmetric C= C
4 4 LT 4 RT 4 T
Total symmetric C= C = C = C

4
Fourth-order unit tensor I :A=A
4 s
Symmetric fourth-order unit I : A = A s = 12 (A + A T )

4 4
Products C = A·B
4 4 4
C= A: B
4
Matrix condensation B = C : A =⇒ B
~ =C A
~

30
 CHAPTER
2
K INEMATICS AND SMALL DEFORMATIONS

This chapter treats fundamental aspects of deformation in solid mechanics. Starting from
a general approach, in which the basic large deformation tensors are introduced, the limit
case for small deformations is treated as a special case. It is shown that the kinematic re-
lations used in linear elasticity are obtained through linearization of the general nonlinear
kinematic equations.

2.1 General aspects of deformation

Upon deformation of a material, a fundamental distinction is made between the material
(or Lagrangian) framework and the spatial (Eulerian) framework. The governing kinematic
relations in continuum mechanics are based on the one-to-one correspondence between
these two frameworks.

2.1.1 Continuity
The continuity of the displacements is one of the basic assumptions in classical continuum
mechanics. This assumption implies that the material is described in an average fashion,
without considering the physical discrete microstructure of the material (e.g., molecular or
crystalline structure). The neighbourhood of a material point is considered to be dense and
continuous. The kinematic and mechanical quantities that ensue from such a theory are
therefore considered as average values over the real microstructure. If a material deforms,
changes in distance between ’dense’ material points will occur. The transformation from
the undeformed state to the deformed state is a one-to-one mapping, where each material
point in the deformed state is associated with one and only one material point from the
undeformed state. The continuity of the deformation implies that neighbourhoods of a ma-
terial point P 0 in the initial region R 0 remain neighbourhoods of the transformed material
point P in the transformed region R, see figure 2.1. Fracture and damage can therefore not
be treated trivially in this framework. Solution strategies for these cases are sketched in the
courses of fracture and damage mechanics.
2. K INEMATICS

V0 V
R0 R
P
P0

t=0 t

Figure 2.1: Deformation of a continuum.

2.1.2 Lagrangian versus Eulerian coordinates

Consider a closed undeformed region R 0 in a material, which after deformation occupies a
deformed region R. The undeformed state or initial state is also the state at time t = 0, while
the deformed or actual state corresponds with time t . Note that the time t is in fact a specific
choice for a state parameter1, which is only a means to uniquely define the current state
of the continuum. For the sake of convenience, we use an independent Euclidean vector
basis to characterize the material points before and after deformation. Each point P 0 in the
undeformed subset R 0 is characterized by the position vector ~ X in the undeformed state,
with coordinates (X 1 , X 2 , X 3 ). These coordinates are the material or Lagrangian coordinates.
The material coordinates are

• unique2 , i.e. each point has different material coordinates

• constant, i.e. they never change since they do not depend on the deformation.

Hence, they can be considered as labels for the material points.

After deformation, this point P 0 will occupy a new position P , characterized by the posi-
tion vector ~
x , with coordinates (x1 , x2 , x3 ). These coordinates are called the spatial or Eule-
rian coordinates, since they determine a fixed point in space where different material points
may pass by in time. The material flows through space (i.e., a fixed set (x1 , x2 , x3 )) upon de-
formation.
The deformation from each point P 0 of the continuum to its corresponding point P , is
described by the so-called nonlinear deformation map φ. If the deformation is traced in
time, the locus of the point P defines a curve

x (t ) = φ( ~
t −→ ~ X ,t) for ~
X fixed (2.1)

For each vector ~ X , there exists one and only vector ~x for each state t . The transformation
from the vectors X ~ to the vectors ~x , or from the coordinates (X 1 , X 2 , X 3 ) to the coordinates
(x1 , x2 , x3 ), i.e., the nonlinear deformation map φ(t ), is also depicted in figure 2.2. Note
1
toestandsparameter
2
éénduidig

32
 2.1. General aspects of deformation

V0
V
P0 r
u P

t=0
r r t
x0 = X r
x

Figure 2.2: The nonlinear deformation map

that the nonlinear deformation map is not a vector but a point-to-point mapping (P 0 →
P (t ), ∀P 0 ). Mathematically speaking, this one-to-one correspondence may be expressed as

x = φ( ~
~ X ,t) (2.2)

or with the inverse deformation map φ−1

X = φ−1 (~
~ x, t ) (2.3)

The nonlinear deformation map φ has the following properties

• unique, i.e. the one-to-one correspondence between particles with position vector ~
x
~
at time t and material points X in the undeformed state.

• continuous, neighbourhoods remain neighbourhoods, no cracks occur. Mathemati-

cally this means that the nonlinear deformation map is differentiable.

This implies that both the coordinates (X 1 , X 2 , X 3 ) or (x1 , x2 , x3 ) could be used as inde-
pendent variables to describe the deformation. If the material coordinates (X 1 , X 2 , X 3 ) are
taken as independent variables, attention is fixed on a definite material particle that evolves
in space. This is the Lagrangian description. If the coordinates (x1 , x2 , x3 ) are taken as in-
dependent variables, one fixes in space a geometrical point (x1 , x2 , x3 ) and investigates what
particles pass through that point. This is the Eulerian description. A quantity α may there-
fore be either written in its material form α(X 1, X 2 , X 3 , t ) or its spatial form α(x1, x2 , x3 , t ). The
spatial form is typically used in fluid mechanics, while the material form is more common in
solid mechanics. The selection of either material or spatial coordinates for the development
of a continuum mechanics framework may have important practical consequences for the
large deformation theories. This topic is a subject of the follow-up course on large defor-
mations. In the present case, we will limit ourselves to problems of the small-deformation
theory. Since the material and the spatial coordinates are then so close together, the influ-
ence of the actual difference between them will become negligible upon linearization. Note
x0 , exactly equals the material vector ~
that the position vector at t = 0, i.e., ~ X . The subscript

33
2. K INEMATICS

0 is therefore often used as well to indicate the differences between the Lagrangian and the
Eulerian setting. Material coordinates are denoted X i or x I in index notation, while spatial
coordinates are then given by xi . In the case of x I , the uppercase index indicates that the
position vector is evaluated in the initial state, i.e., in the Lagrangian configuration.

2.1.3 The displacement vector

Upon deformation a material point changes position from ~ x0 to ~
x . This change in position
is characterized by the displacement vector (figure 2.2), which is given by

~
u =~
x −~
x0 (2.4)

The components of the displacement vector~

u with respect to an Euclidean basis are (u 1, u 2 , u 3 ),
also denoted by u i in index notation.

2.1.4 The displacement gradient tensor and deformation gradient tensor

Let’s consider two material points that are infinitely close, i.e., one material point is currently
in the position ~x , while its neighbour is in position ~
x +d~x . The principle of continuity tells us
that these two points were also neighbours in the initial undeformed state. Their positions
in that initial state are respectively given by ~ x0 and ~ x0 + d~x0 . The vector d~x0 that initially
interconnected these two points is thus transformed or deformed into a new vector d~ x , see
figure 2.3

F
P0

r r
dx0 = dX
r
t=0 P dx
t

Figure 2.3: The linearized deformation map - the deformation tensor F

The linear transformation of the vector d~

x0 into the vector d~
x is described by a second-
order tensor, called the deformation gradient tensor F . From the nonlinear deformation
map φ it appears that

x = (~
d~ x ) − (~
x + d~ x)
= φ(~ x0 ) − φ(~
x0 + d~ x0 )
∂φ
= ∂~
x0
x0
·d~
= F ·d~
x0 (2.5)

From the definition of the gradient operator, equation (1.158), it can be noticed that

d~ x0 ·~
x = d~ ∇0~
x (2.6)
= (~ x )T ·d~
∇0~ x0 (2.7)

34
 2.1. General aspects of deformation

The deformation gradient tensor F is thus given by

∂φ
F = (2.8)
∂~x0
= (~
∇0~x )T (2.9)

The second-order tensor F is thus the spatial derivative of the nonlinear deformation map in
the considered material point, which leads to a locally linear deformation map. The tensor is
the transposed Lagrangian gradient of the current position vector. The tensor thus expresses
the deformation with respect to the initial undeformed state. The gradient symbol ~∇0 clearly
indicates this with its subscript 0. Since F is a second-order tensor, it is fully identified by
means of the image (i.c. the deformation) of three (i.c. infinitesimal) vectors, which span an
infinitesimal volume as shown in figure 2.4.

r
dx2
r
dX 2 r
r dx3
dX 3
P0 r P r
dx1
dX 1

Figure 2.4: The deformation gradient tensor F

The matrix notation F of this tensor F with respect to an Euclidean basis is given by
 ∂x ∂x1 ∂x1

1
∂X 1 ∂X 2 ∂X 3
 ∂x2 ∂x2 ∂x2 

F =
 ∂X 1 ∂X 2 ∂X 3  (2.10)
∂x3 ∂x3 ∂x3
∂X 1 ∂X 2 ∂X 3

or in index format
Fi J = xi ,J (2.11)
Note that the second index is denoted with uppercase characters, which means that it refers
to the Lagrangian undeformed state, while the first index is in lowercase, corresponding to
the Eulerian deformed state. This convention will be respected in the large deformation part
of the kinematical framework, where the difference between these two states is essential.
As indicated in the chapter on tensor analysis, the comma between the indices in equa-
tion (2.11) indicates differentiation with respect to the second index variable.
The Lagrangian gradient of the displacement also leads to a second order tensor, called
the displacement gradient tensor (~ u )T . Like in equation (2.6) (i.e., the definition of the
∇0 ~
Lagrangian gradient of a vector), this tensor maps the initial interconnection vector d~ x0 on
the relative displacement vector between the two considered adjacent material points.

d~ x0 ·~
u = d~ ∇0 ~
u (2.12)
= (~ u )T ·d~
∇0 ~ x0 (2.13)

35
2. K INEMATICS

The relative displacement vector d~

u measures the displacement that occurred between these
two adjacent points with respect to their original relative positions
d~
u = d~
x − d~
x0 (2.14)
Using an Euclidean basis, the components of (~ u )T are commonly stored in a 3 × 3 matrix
∇0 ~
 ∂u ∂u ∂u

1 1 1

 ∂X 1 ∂X 2 ∂X 3

 ∂u2 ∂u2 ∂u2 
(2.15)
 ∂X 1 ∂X 2 ∂X 3 
∂u3 ∂u3 ∂u3
∂X 1 ∂X 2 ∂X 3

Evidently, the tensors F and (~ u )T are closely related

∇0 ~
F = (~ x )T
∇0~
= (~ u ])T
x0 + ~
∇0 [~
=~
∇0~x T + (~ u )T
∇0 ~
| {z 0}
I
= I + (~ u )T
∇0 ~ (2.16)

Since F and (~ u )T are both second-order tensors, they describe linear transformations.
∇0 ~
This means that the deformation along three independent vectors in (3D) material point
suffices to fully characterize these tensors. An initially cubic volume element will always be
deformed by F into a parallelepiped.
Consider a volume element that is spanned by three non-coplanar infinitesimal material
vectors d ~X 1, d ~
X 2 and d ~X 3 . After deformation these vectors are transformed into new vec-
tors d~
x1 , d~x2 and d~ x3 . The deformation tensor F gives the relation between the two sets of
vectors
d~ ~1
x1 = F ·d X
x2 = F ·d ~
d~ X2 (2.17)
x3 = F ·d ~
d~ X3
If both sets are known, then F can be determined.

2.2 Strains and strain tensors

2.2.1 Strain definitions
Strains are used to quantify the deformation relative to the undeformed state. It is often too
simply interpreted as a relation between the change in length of a line element and its initial
length. In fact, any definition of a strain is arbitrary and is largely a matter of convenience. In
this subsection we highlight most scalar strain definitions that apply to a one-dimensional
line element. These scalar strain measures are functions of the stretch ratio λ, which is the
ratio between the deformed l and undeformed length L = l 0 of a line element.
l l
λ= = (2.18)
L l0
A strain measure typically equals zero in the undeformed state, in contrast to the stretch ratio
which then equals 1. For a line element with an infinitesimal length, the stretch typically
reads d l /d l 0 .

36
 2.2. Strains and strain tensors

The engineering strain

A definition which is frequently used in engineering practice and which will play a particular
role in the linearization of other strain measures, is the engineering strain e. The engineering
strain e is a linear function of the stretch ratio λ, and simply equals

e = λ−1 (2.19)

This linear strain measure is bounded between −1 (since d l is always larger than zero) and
plus infinity.

The logarithmic strain

This definition is based on the instantaneous length of a line element, rather than the initial
length. Every infinitesimal increase of the strain, i.e., a strain increment d εl n , is related to
the infinitesimal increase in length d l of the line element with respect to its current length l

dl
d εl n = (2.20)
l

The current value of the logarithmic strain is thus given by

Zl µ ¶
dl l
εl n = = ln = ln λ (2.21)
l0 l l0

which justifies its name. This strain measure is often also called the natural or true strain. It
is bounded between −∞ and +∞.

The Green-Lagrange strain

The Green-Lagrange strain measure is one of the basic strain definitions used in continuum
mechanics. The Green-Lagrange scalar strain is a nonlinear function of the stretch λ, given
by
εg l = 12 (λ2 − 1) (2.22)

The Green-Lagrange strain is bounded between − 21 and +∞.

The Euler-Almansi strain

The Euler-Almansi strain, typically used within a Eulerian framework, is defined by

1
εea = 12 (1 − ) (2.23)
λ2

This strain is bounded between −∞ and + 12 .

The different strain definitions are represented in figure 2.5. The properties and differ-
ences between the cited strain definitions are thus well illustrated. It appears clearly that in
the vicinity of the undeformed state, all definitions are similar. For larger or smaller stretches,
important differences arise.

37
2. K INEMATICS

5
2
Green−Lagrange: 1/2(λ −1)
4
Linear: λ − 1
3
Strain definitions
2
Logarithmic: ln(λ)

Euler−Almansi: 1/2(1−1/λ2)
0

−1

−2
0 1 2 3 4 5
Stretch λ

Figure 2.5: Scalar strain definitions to characterize deformation.

2.2.2 Deformation of infinitely small material line elements

The one-dimensional case

Tensors and vectors simplify into scalars for the one-dimensional case, which makes the
equations particularly straightforward. Consider a one-dimensional line element with initial
length d L which is stretched to a new length d l . with respect to a one-dimensional Euclidean
basis, the deformation gradient simply equals
∂x
F =F =
∂X
∂u
= 1+ (2.24)
∂X
The squared new length d l 2 of the 1D line element equals

dl2 = dx dx
= FdX FdX
= F 2d L2
∂u 2 2
= (1 + ) dL (2.25)
∂X
The change in length of this line element can be measured with
µ· ¸ ¶
2 2 ∂x 2
1 1
2 (d l − d L ) = 2 − 1 d L2
∂X
 · ¸ 
∂u 1 ∂u 2 
 d L2

= + 
∂X 2 ∂X
= εg l d L 2 (2.26)

38
 2.2. Strains and strain tensors

where use has been made of the earlier defined Green-Lagrange strain εg l .

Generalization to three dimensions, the Green-Lagrange strain tensor

Let’s consider two neighbouring points P 0 and Q 0 in the undeformed region dV0 of a mate-
rial. Point P 0 has an initial position vector ~X =~ x0 , while its infinitely close neighbour Q 0 is
~ ~
characterized by an initial position X + d X = ~ x0 + d~x0 . The infinitesimal region dV0 is de-
formed into the region dV and the two points are now located in P and Q, see figure 2.6.
Their new positions are given by ~ x and ~
x + d~x . The vector d~ x0 = d ~
X is the vector that inter-

dV0 dV
r Q0
dX P r
P0 dx
Q
r r r r
x0 + dx0 = X + dX t
t=0
r
x
r r
r r
x0 = X
x + dx

0
Figure 2.6: Deformation in an infinitesimal volume

connects P 0 with Q 0 , while d~

x interconnects P with Q. The initial Euclidean length of this
material line element is d L = d l 0 , given by

d L2 = d X
~ ·d ~
X
= d Xi d Xi
= d X 12 + d X 22 + d X 32 (2.27)

Similarly, points P and Q are interconnected by the vector d~

x , for which the deformed length
d l is given by

d l 2 = d~
x ·d~
x
= d xi d xi
= d x12 + d x22 + d x32 (2.28)

x and d ~
Making use of the relation between d~ X , given by equation (2.5), the deformed length
d l can be written as

d l 2 = d~
x ·d~
x
~ )·(F ·d ~
= (F ·d X X)
~ ·(F T ·F )·d ~
= dX X (2.29)
~ ·C ·d ~
= dX X (2.30)

39
2. K INEMATICS

A new tensor C = F T ·F has thereby been introduced, which is called the right Cauchy-Green
deformation tensor. Specific properties of this tensor will be highlighted in the course of
large deformation kinematics. Here we will limit ourselves to its role towards the lineariza-
tion of deformations. Equation (2.30) permits to express the difference between d l 2 and d L 2
in terms of the initial material vector d ~
X (i.e., the material coordinates X i are the indepen-
dent variables)
1 2 2
2 (d l − d L ) = 12 (d ~
X ·(C )·d ~
X −d~ X ·d ~
X) (2.31)
1
~ ·( [C − I ])·d ~
= dX 2 X (2.32)

The tensor which appears between the material vectors d ~ X in the right-hand side is a non-
linear strain tensor, denoted the Green-Lagrange strain tensor E

E = 21 [C − I ] (2.33)
= 1 ~
2
u + (~
[∇0 ~ u )T
∇0 ~ +~ u ·(~
∇0 ~ u )T ]
∇0 ~ (2.34)

With respect to a Cartesian vector basis equation (2.32) can be rewritten as

2 2
1
2 (d l − d L ) = E 11 d x12 + E 22 d x22 + E 33 d x32 +
(2.35)
2E 12 d x1 d x2 + 2E 13 d x1 d x3 + 2E 23 d x2 d x3

where the individual Green-Lagrange strain components E I J are given by

µh i h i h i ¶
∂u1 1 ∂u1 2 ∂u2 2 ∂u3 2
E 11 = ∂X 1 + 2 ∂X 1 + ∂X 1 + ∂X 1
µh i h i h i ¶
∂u2 1 ∂u1 2 ∂u2 2 ∂u3 2
E 22 = ∂X + 2 ∂X + ∂X + ∂X
2 2 2 2
µh i2 h i2 h i ¶
∂u3 1 ∂u1 ∂u2 ∂u3 2
E 33 = ∂X + 2 ∂X + ∂X + ∂X (2.36)
3 3 3 3
³ ´
∂u1 ∂u2 ∂u1 ∂u1 ∂u2 ∂u2 ∂u3 ∂u3
E 12 = 21 ∂X + ∂X + ∂X + +
2 1 1 ∂X 2 ∂X 1 ∂X 2 ∂X 1 ∂X 2
³ ´
∂u1 ∂u3 ∂u1 ∂u1 ∂u2 ∂u2 ∂u3 ∂u3
E 13 = 21 ∂X + ∂X + ∂X + +
3 1 1 ∂X 3 ∂X 1 ∂X 3 ∂X 1 ∂X 3
³ ´
∂u2 ∂u ∂u1 ∂u1 ∂u2 ∂u2 ∂u3 ∂u3
E 23 = 21 ∂X 3
+ 3
∂X 2 + ∂X 2 ∂X 3 + ∂X 2 ∂X 3 + ∂X 2 ∂X 3

or in its compact index notation

1
¡ ¢
EIJ = 2
u I ,J + u J ,I + u K ,I u K ,J (2.37)

The difference between the squares of the length of material line elements is thus written
as

d l 2 − d L2 = 2 d ~
X ·E ·d ~
X (2.38)
= 2 Ei j d Xi d X j (2.39)

2.3 Linearization of deformations

The theory of linear elasticity is entirely built on the concept of small deformations. The
only strain definition given so far, which is a linear function of the stretch ratio, was the

40
 2.3. Linearization of deformations

engineering strain e. In the case of small deformations this definition equals the scalar linear
strain definition ε.
dl − dL
ε= = λ−1 (2.40)
dL
Furthermore, all other strain definitions equal ε as well in the limit for small deformations.
This will be shown for the scalar strain definitions first and generalized for the tensors after-
wards, which will lead to the linear strain tensor.

2.3.1 Linearized strain definitions

The linearization of the earlier presented scalar strain definitions will lead systematically to
the linear strain. This property could be expected by reconsidering figure 2.5, where only the
deformation in the vicinity of λ = 1 is represented. This result is shown in figure 2.7. The

0.06
Linear: λ − 1
2
Green−Lagrange: 1/2(λ −1)
Euler−Almansi: 1/2(1−1/λ2)
0.04
Logarithmic: ln(λ)

0.02
Strain definitions

−0.02

−0.04

−0.06
0.95 1 1.05
Stretch λ

Figure 2.7: Tangent character of the linear strain.

linear strain is clearly tangent to all other strain definitions in λ = 1. The mathematical proof
of this property is given in the following paragraphs.

Linearization of the logarithmic strain

The logarithmic strain εl n can be expanded as follows

εl n = ln(λ) = ln(1 + ε) (2.41)

For small deformations, it always holds that d l −d L < d L and the logarithm can be expanded
in the following series
1 1
ln (1 + ε) = ε − 12 ε2 + ε3 − ε4 + · · · (2.42)
3 4
In the limit case that deformations are infinitesimal, only the linear term in equation (2.42)
is to be preserved and one finds

εl n ≈ ε if |ε| ≪ 1 (2.43)

41
2. K INEMATICS

Linearization of the Green-Lagrange strain

The Green-Lagrange strain εg l can be expanded in d l as well

εg l = 12 (λ2 − 1)
¡ ¢
= 12 2ε + ε2 (2.44)

Linearization of this equations for infinitesimal deformations permits to neglect the higher-
order term in d l , which leads to the following approximation

εg l ≈ ε if |ε| ≪ 1 (2.45)

2.3.2 The infinitesimal linear strain tensor

Definition

The infinitesimal strain tensor ε is formally defined as the linearized fraction of the Green-
Lagrange strain tensor E . For infinitesimal displacements, the first partial derivatives of u i
are so small that all involved squares and products are negligible with respect to the linear
terms. In that case, it is allowed to neglect the nonlinear terms in the definition of E

1
¡ T
¢
E = 2 F ·F − I (2.46)
£ ¤
= 1 ~
~ ~ ~T
2 ∇0 u + (∇0 u ) +~ u ·(~
∇0 ~ u )T
∇0 ~ (2.47)

from which the linear strain tensor is extracted ε

£ ¤
ε= 1
2
~ u + (~
∇0 ~ u )T
∇0 ~ (2.48)

Note that this tensor equals the symmetric part of the displacement gradient tensor. Another
frequently used formula for the same definition is
£ ¤
ε= 1
2
F +FT −I (2.49)

Clearly, the distinction between the Lagrangian and Eulerian strain tensor now vanishes,
since it is now immaterial whether the derivatives of the displacements are determined at
the deformed or undeformed position of the considered point. Equation (2.38) thus leads to
one single expression where the difference between d X i and d xi is no longer relevant

d l 2 − d L2 = 2 d ~
X ·ε·d ~
X (2.50)
= 2 d~
x ·ε·d~
x
= 2 εi j d X i d X j = 2 εi j d xi d x j (2.51)

The fact that the change of the position vector is no longer relevant for infinitesimal dis-
placements is a fundamental observation, which permits to omit the double notation ~ X ↔~ x
~ ~ ~ ~
or even ∇ X~ = ∇0 ↔ ∇ = ∇~x . In this case, the notation ~ ~
x and simply ∇ are often used in liter-
ature if the kinematics apply to small displacements, since then ~ ∇ ≈~ ∇0 for most cases. The
index notation is now written as ³ ´
εi j = 21 u i , j + u j ,i (2.52)

42
 2.3. Linearization of deformations

Strain-displacement relations in Cartesian coordinates

∂ ∂
The gradient operator with respect to a Cartesian basis {~
e x ,~ e z } is given by ~
e y ,~ ∇ =~
e x ∂x e y ∂y
+~ +
∂
~
e z ∂z , while the components are the displacement vector with respect to this basis are u, v, w.
It is easy to show that the definition of the linear strain tensor (2.48) leads to the same equa-
tions that are obtained through direct linearization of the equations (2.36)

∂u
εxx =
∂x
∂v
εy y =
∂y
∂w
εzz =
∂zµ ¶ (2.53)
1 ∂u ∂v
εx y = 2 ∂y
+
∂x
µ ¶
1 ∂u ∂w
εxz = 2 ∂z
+
∂x
µ ¶
1 ∂v ∂w
ε yz = 2 ∂z
+
∂y

Strain-displacement relations in cylindrical coordinates

With respect to a cylindrical basis {~

e r (θ),~
e t (θ),~
e z } the gradient operator is given by

∂ 1 ∂ ∂
~
∇=~
er et
+~ ez
+~ (2.54)
∂r r ∂θ ∂z

The displacement vector with respect to a cylindrical basis reads ~ e r (θ) + v~

u = u~ e t (θ) + w~
ez
which by application of (2.48) yields

∂u
εr r =
∂r
u 1 ∂v
εt t = +
r r ∂θ
∂w
εzz =
∂z
µ ¶ (2.55)
1 1 ∂u ∂v v
εr t = 2 r ∂θ
+ −
∂r r
µ ¶
1 ∂u ∂w
εr z = 2 ∂z + ∂r
µ ¶
1 ∂v 1 ∂w
εt z = 2 ∂z + r ∂θ

2.3.3 The infinitesimal rotation tensor

Any tensor can be split into a symmetric tensor and a skew-symmetric tensor. For the dis-
placement gradient tensor, is was shown that the symmetric part has a specific meaning in
linear elasticity. The same holds for the skew-symmetric part. This part, denoted ω, equals
the mean rotation (displacement) tensor in a material point, under the condition that the

43
2. K INEMATICS

displacement field is infinitely small.

£ ¤T
~ u = 12 ([~
∇0 ~ u ]T + ~
∇0 ~ ∇0 u) + 21 ([~ u ]T − ~
∇0 ~ ∇0 u) (2.56)
= ε+ω (2.57)

The mean rotation in a material point is the mean with respect to all infinitesimal line ele-
ments that emanate from that point. Note that the name rotation tensor is not well-chosen,
since this tensor only yields the displacement vectors due to the rotation. The tensor ω dif-
fers in that sense from the more general (large displacement) rotation tensor Q, which is an
orthogonal tensor that yields a rotated material vector. In fact, in the case of infinitesimal
displacements, both tensors are related with

Q = I +ω (2.58)

Consider a material point P and a neighbouring point that is separated from P through an
x . Since ω is skew-symmetric, an axial vector ~
infinitely small material line element d~ ω exists
such that
ω·d~ x =~ω × d~ x (2.59)
The result is thus a vector that is always perpendicular to the that interconnects two neigh-
bouring material points. The resulting displacement vector from (2.59) is thus characteristic
for an infinitely small rotation around the considered material point P . Note that angular
displacements have to be infinitely small for this interpretation of ω.

2.4 Applicability of the theory of small displacements

On the basis of this additive split of the displacement gradient tensor, a better insight can be
provided in the limitations that apply to linear elasticity. The relation between the Green-
Lagrange strain tensor E and the infinitesimal strain tensor ε can be written with the use of
the infinitesimal rotation tensor ω.
¡ ¢
E = 21 ~ u + [~
∇0 ~ u ]T + ~
∇0 ~ u ·[~
∇0 ~ u ]T
∇0 ~ (2.60)
= ε + 12 [~ u ]T ·~
∇0 ~ ∇0 ~
u (2.61)

and using

ε + ω = [~ u ]T
∇0 ~ (2.62)
T
ε + ω = [~
∇0 ~
u] (2.63)

it follows that £ ¤
E = ε + 12 (ε + ωT )·(ε + ω) (2.64)
The latter relation clearly illustrates that it is not sufficient to have small strain |εi j | ≪ 1, since
then equation (2.64) simplifies to

E = ε + 12 ε·ωT + 21 ω·ε + 12 ωT ·ω (2.65)

which is not yet the linear strain tensor (only 21 ε·ε is negligible). The latter term only vanishes
if rotations are small. In order to use the infinitesimal strain tensor as a first-order approxi-
mation of the Green-Lagrange strain tensor, both strains and rotations must be small! This is

44
 2.4. Applicability of the theory of small displacements

particularly important for flexible bodies, such as thin plates and slender rods. The bending
of large thin sheets for instance, causes only small strains in the material. Yet, the displace-
ments and more specifically the rotations are very large. The linearization is thus not ap-
plicable. This is most easily illustrated with a simple example. Consider a two-dimensional
material line element with length L that interconnects A with B, see figure 2.8. The line ele-

B’
( x, y)

L
a B
L ( x0 , y0 )
A

Figure 2.8: Rigid rotation of a material line element

ment is simply rotated over an angle α around A towards its new position A −B ′ . A Cartesian
basis with its origin in A is next taken. A material point (x0 , y 0 ) of the initial line element will
be in a new position (x, y) given by
· ¸ · ¸· ¸
x cos α sin α x0
= (2.66)
y − sin α cos α y 0
The displacement components are thus given by
· ¸ · ¸ · ¸
u x − x0 (cos α − 1)x0 + (sin α)y 0
= = (2.67)
v y − y0 (− sin α)x0 + (cos α − 1)y 0
The matrix form of the two-dimensional linear strain tensor ε is then given by
 ³ ´
∂u 1 ∂u ∂v
·
εxx εx y
¸
∂x0 2 ∂y 0 + ∂x0
= ³ ´ 
ε yx ε y y 1 ∂v
+ ∂u ∂v
2 ∂x0 ∂y 0 ∂y 0
· 1 ¸
cos α − 1 2 (sin α − sin α)
= 1
2
(sin α − sin α) cos α − 1
· ¸
cos α − 1 0
= (2.68)
0 cos α − 1
Clearly the dilatational strain components (i.e., the components εxx and ε y y ) are not equal to
zero if the angle of rotation is not very small. Deviations become very large for large rotation
angles and the numerical values of the strain components completely lose their physical
meaning. A generalization of this rigid rotation example will be given in section 2.7.1. For
the components of the Green-Lagrange strain tensor it holds that
µh i h i ¶
2 2
∂u 1 ∂u ∂v
E xx = ∂x + 2 ∂x + ∂x
0 0 0

1 2 2
= cos α − 1 + 2 (cos α − 2 cosα + 1 + sin α)
=0 (2.69)

45
2. K INEMATICS

µh i2 h i2 ¶
∂v 1 ∂u ∂v
Eyy = ∂y 0 2 ∂y 0
+ ∂y 0

= cos α − 1 + 12 (sin2 α + cos2 α − 2 cos α + 1)

=0 (2.70)

³ ´
1 ∂u ∂v ∂u ∂u ∂v ∂v
E x y = E yx = 2 ∂y 0 + ∂x0 + ∂x0 ∂y 0 + ∂x0 ∂y 0
1
= 2 (sin α − sin α + sin α cos α − sin α − sin α cos α + sin α)
=0 (2.71)

which clearly shows that it is insensitive to rigid body rotations, see also section 2.7.1.
Small displacement theory better characterizes the limitations, since then both strains
and rotations have to be small. The small displacement theory should be used with caution,
since it will yield completely erroneous results if its underlying assumption is not satisfied.
Fortunately, for massive and thick products (not thin nor flexible) the linearization is well
permissible, for as long as the analysis is confined to the normal use of the products during
their lifetime. The manufacturing of the same products on the contrary is mostly accompa-
nied by large displacements. These nonlinear processes will therefore be treated separately
in the course on the mechanics of advanced manufacturing processes.

2.5 Physical interpretation of the strain components

2.5.1 Normal or extensional strains
Consider a Cartesian vector basis {~ e 1 ,~
e 2 ,~
e 3 }, with respect to which the components of the
strain tensor are analysed. The components ε11 , ε22 , ε33 of the infinitesimal strain tensor ε
are called the normal or extensional strains. They are determined along their corresponding
directions N~ 1, N
~ 2, N
~ 3 with e.g., for ε11
£ ¤
~ 1 ·ε·N
ε11 = N ~ 1 ·( 1 F + F T − I )·N
~1 = N ~1
2
¡ £ ¤ ¢
= 21 N~ 1 · F + F T ·N ~1 − 1
1 £ ¤
= d~ X 1 · F + F T ·d X ~1 − 1
2d L 1 d L 1
1 £ ¤
= d~X 1 ·d~x1 + d~x1 ·d ~
X1 − 1
2d L 1 d L 1
d L1 d l1
= N~ 1 ·~
n1 − 1 (2.72)
d L1 d L1

~ 1 to ~
For infinitely small displacements the change in angle from N n 1 is small and hence

~ 1 ·~
N n 1 = cos(θ1 ) ≈ 1 (2.73)

The extensional strain ε11 is then equal to

d l1
ε11 = −1 (2.74)
d L1
= λ1 − 1 (2.75)

46
 2.5. Physical interpretation of the strain components

This result is nothing else than the scalar linear strain along that direction. The diagonal
components ε11 , ε22 , ε33 of the linear strain tensor thus represent the extension along the
directions of the Cartesian base vectors, i.e., the change in length per unit length along that
direction, see figure 2.9. The normal extensional strain components ε11 , ε22 , ε33 equal the

¶v
v+ dy
¶y

dy u

v
¶u
u+ dx
dx ¶x
x

Figure 2.9: Physical interpretation of extensional or normal strains

linear strains in the directions of the coordinate axes, which are the relative elongation of the
fibers parallel to these axes. Evidently, this property of the linear strain tensor does no longer
hold if deformations and/or rotations are large.

2.5.2 Shear strains

The remaining components ε12 = ε21 , ε13 = ε31 , ε23 = ε32 of the strain tensor ε are called the
tensorial shear strains, since they are related to the change of angles between material line
elements. Let us consider two infinitely small material line elements that emanate from a
material point P, oriented along the Cartesian coordinate axes which are perpendicular, e.g.,
along ~
e 1 and ~e 2 . In the undeformed state, the associated material vectors are denoted d ~ X1
and d ~
X 2 . After deformation, the material line elements have deformed towards their new
positions d~ x1 and d~ x2 . The initial lengths of these material line elements are d L 1 and d L 2
respectively. After deformation the lengths equal d l 1 and d l 2 . The scalar product between
the deformed vectors then gives

π
d~ x2 = d l 1 d l 2 cos(
x1 ·d~ − θ) (2.76)
2

where θ is the total change in angle from the undeformed state to the current state between
the two line elements.
Applying the strain tensor to the undeformed perpendicular directions N ~ 1 and N~ 2 (N
~ 1 ·N
~2 =
~ ~
0) of the material vectors d X 1 and d X 2 and making use of equation (2.49) for the strain tensor

47
2. K INEMATICS

leads to
£ ¤
~ 1 ·ε·N
ε12 = N ~ 1 ·( 1 F + F T − I )·N
~2 = N ~2
2
£ ¤
~ 1 · F + F T ·N
= 12 N ~2
1 £ ¤
= d~X 1 · F + F T ·d ~ X2
2d L 1 d L 2
1 £ ¤
= d~X 1 ·d~
x2 + d~ ~2
x1 ·d X
2d L 1 d L 2
d L1 d l2 d l1 d L2
= ~ 1 ·~
N n2 + ~
n 1 ·N~2 (2.77)
2d L 1 d L 2 2d L 1 d L 2
For infinitesimal displacements, the differences between d L 1 ↔ d l 1 and d L 2 ↔ d l 2 only lead
to second-order contributions in equation (2.77), which allows to neglect them. Hence, the
component ε12 is given by
1~
ε12 = n 2 + 21 ~
2 N 1 ·~
~2
n 1 ·N
h
1 π π i
= cos( − θ2 ) + cos( − θ1 )
2 2 2
1
= 2 [sin(θ2 ) + sin(θ1 )]
1 θ
≈ 2 [θ2 + θ1 ]2
= (2.78)

Clearly, it appears that the strains ε12 , ε13 , ε23 , which are tensorial components, equal half

¶u
dy
¶y

dy ¶v
u dx
¶x
v

dx
x

Figure 2.10: Physical interpretation of shear strains

the change of angle between two material line elements oriented along the corresponding
coordinate axes in the undeformed state (~ e 1 -~
e 2 ,~
e 1 -~
e 3 ,~
e 2 -~
e 3 ). This is well depicted in fig-
ure 2.10, which shows the contributions to the shear strain with respect to a Cartesian basis.
In engineering practice, these tensorial strain components are frequently doubled, e.g., 2ε12
is used instead of ε12 . These doubled components are then called the shear strains γ12
γ12 = 2ε12
γ13 = 2ε31 (2.79)
γ23 = 2ε32

48
 2.5. Physical interpretation of the strain components

The shear strain γ12 equals the total change of angle after deformation between two line
elements. However, it should be emphasized that the extensional strains ε11 , ε22 , ε33 and the
shear strains γ12 , γ13 , γ23 do not form a tensor, which is mathematically inconvenient.

2.5.3 Principal strains and strain invariants

Since ε is a symmetric second-order tensor, its eigenvectors and eigenvalues are real, and
have a physical meaning. The eigenvalue problem of ε is easily solved from

n = ǫ~
ε·~ n (2.80)

where ǫ are the eigenvalues and ~

n the eigenvectors. The eigenvalues are then obtained from
the cubic characteristic equation

det(ε − ǫI ) = 0 (2.81)
3 2
ǫ − J 1 (ε)ǫ + J 2 (ε)ǫ − J 3 (ε) = 0 (2.82)

The three strain invariants J 1 (ε), J 2 (ε) and J 3 (ε) have thereby been introduced. The roots of
equation (2.82) are the principal strains ǫ1 , ǫ2 and ǫ3 . The eigenvectors, or principal strain
~ i are computed from
directions N

~ i = ~0
(ε − ǫi I )·N ∀i ∈ 1, 2, 3 (2.83)

in combination with the required unit length of the vectors ~ n i , i.e., |~

n i | = 1. If the three
principal strains are distinct, the three principal directions will be mutually orthogonal. If
two principal strain are equal, equation (2.83) has an infinite number of solutions, out of
which any pair of orthogonal vectors can be chosen as the principal directions. If all principal
strains are equal, each set of three mutually orthogonal vectors is a set of principal directions.
An axis that is oriented along a principal direction is called a principal axis. A plane normal
to a principal direction is called a principal plane. If a Cartesian basis is taken with three axes
that coincide with the principal axes, the matrix form of ε takes a particular format
 
ǫ1 0 0
ε =  0 ǫ2 0  (2.84)
0 0 ǫ3

The shear strain components are clearly equal to zero along these directions. This implies
that material line elements oriented along these directions are only stretched or compressed.
Since the shear strain between principal axes vanishes, the principal axes are mutually or-
thogonal before and after deformation. Furthermore, the extensional strains attain an ex-
treme value along the principal axes. The largest and smallest extensional strain for all pos-
sible orientations of material line elements can always be found among the principal strains.
The three principal strain invariants J 1 (ε), J 2 (ε) and J 3 (ε) are computed from the princi-
pal strains through

J 1 (ε) = ǫ1 + ǫ2 + ǫ3 (2.85)
J 2 (ε) = ǫ1 ǫ2 + ǫ1 ǫ3 + ǫ2 ǫ3 (2.86)
J 3 (ε) = ǫ1 ǫ2 ǫ3 (2.87)

Note that any function of these three principal strain invariants is a strain invariant as well.

49
2. K INEMATICS

2.5.4 Volume and shape changes

Volumetric strain

The change of the volume of an infinitely small volume element is first analyzed on the basis
of the deformation tensor F . Linearization will then lead to a simplified expression, which
is applicable in linear elasticity. Consider an undeformed material parallelepiped that is
spanned by three vectors d ~ X 1, d ~
X 2 and d ~
X 3 . The undeformed volume of this parallelepiped
equals dV0 , and is given by
dV0 = (d ~X1 × d X~ 2 )·d ~
X3 (2.88)
After deformation, the deformed parallelepiped is spanned by the deformed vectors d~
x1 ,
x2 and d~
d~ x3 , which leads to the deformed volume dV :

dV = (d~ x2 )·d~
x1 × d~ x3
= (F ·d ~
X 1 ) × (F ·d ~ ~3)
X 2 )·(F ·d X
= det(F ) (d ~ X1 × d ~
X 2 ·d ~X 3)
= det(F ) dV0 (2.89)

The change in volume (for large deformations) is thus characterized and quantified with the
determinant of the deformation tensor, i.e., det(F ). This quantity is often denoted by J , and
is known as the volume ratio3 of the continuum:
dV
J = det(F ) = (2.90)
dV0
The volume of an infinitely small material volume element can physically never be negative
nor zero, nor reach infinity, so it holds that:

J = det(F ) > 0 and J <∞ (2.91)

In the case of infinitesimal displacements, the volume ratio J is easily linearized

J = det(F ) = det(I + ~
∇~uT )
≈ 1 + tr(~
∇~u T ) = 1 + tr(~
∇~u)
≈ 1 + tr(ε) (2.92)

This equation is easily retrieved with respect to a Cartesian basis (x, y, z) with infinitesimal
displacement components (u, v, w), for which
 ∂w 
1 + ∂u
∂x
∂v
∂x ∂x
T  ∂u ∂v ∂w

det(I + ~
∇u ) =  ∂y
 1 + ∂y ∂y 

∂u ∂v
∂z 1 + ∂w
∂z
∂z
∂u ∂v ∂w ∂u ∂v ∂u ∂w ∂v ∂w
= 1+ + + + + + +...
∂x ∂y ∂z ∂x ∂y ∂x ∂z ∂y ∂z
| {z }
higher-order terms
∂u ∂v ∂w
≈ 1+ + +
∂x ∂y ∂z
= 1 + tr(ε) (2.93)
3
volumeveranderingsfaktor

50
 2.6. Compatibility conditions

The first invariant or the trace of the infinitesimal linear strain tensor thus measures the
relative volume variation
dV − dV0
tr(ε) = (2.94)
dV0
The trace of ε is called the volumetric strain or cubical dilatation and often denoted e

e = tr(ε) = J 1 (ε) (2.95)

For an incompressible material (e.g., fluid) it holds that e = 0.

Shape changes

Many materials in deformation processes show a mechanical behaviour that is specially de-
pendent of the mean strain, which is simply the volumetric strain divided by 3
ε11 + ε22 + ε33 e 1
εm = = = 3 J 1 (ε) (2.96)
3 3
The mean strain tensor or volumetric strain tensor is denoted εv and equals

εv = εm I = 31 tr(ε) I (2.97)

The part of the strain tensor that is obtained after subtraction of εv is called the deviatoric
strain tensor εd
εd = ε − ε v (2.98)
Since by definition the trace of this deviatoric strain tensor is zero, there is no volume change
associated to it (infinitesimal displacements !). Its components are only related to the changes
in shape of a material volume element. The principal invariants of the deviatoric part of the
linear strain tensor are denoted J 1 (εd ), J 2 (εd ) and J 3 (εd ). Clearly the first invariant is zero

J 1 (εd ) = 0 (2.99)

It is easy to verify that the following relations exist between the invariants J 2 , J 3 of the total
strain tensor ε, the volumetric strain εm and the invariants of the strain deviator εd

J 2 (εd ) = 3ε2m − J 2 (ε) (2.100)

J 3 (εd ) = J 3 (ε) + εm J 2 (ε) − ε3m (2.101)

The second invariant J 2 of the strain deviator tensor εd is also given by

 
2
J 2 (εd ) = 1
(εd )]2 − tr(εd )
2 [t|r {z }
=0
= − 12 εd : εd (2.102)

2.6 Compatibility conditions

The deformation in a material point, characterized through the linear strain tensor ε, de-
pends on 6 components (taking into account the symmetry of ε). Yet, the definition of the
strain tensor clearly shows that all of these components are completely determined from the

51
2. K INEMATICS

3 components of the displacement vector field. This implies that the 6 strain components
are not independent and that 3 relations must exist between the strain components. These
3 relations are called the compatibility conditions. They are obtained through elimination
of the displacements in the strain-displacements equations to produce equations with only
strain components as unknowns. The equations are next elaborated for a Cartesian vector
basis. The first equation is found by taking the second derivatives of the components ε11 ,
ε22 , ε12 with respect to the Cartesian coordinates x1 , x2 :

∂2 ε11 ∂3 u 1
= (2.103)
∂x22 ∂x1 ∂x22
∂2 ε22 ∂3 u 2
= (2.104)
∂x12 ∂x12 ∂x2
∂2 ε12 ∂3 u 1 ∂3 u 2
2 = + (2.105)
∂x1 x2 ∂x1 ∂x22 ∂x12 ∂x2

Summing up the first two of these equations yields the third one, i.e.,

∂2 ε11 ∂2 ε22 ∂2 ε12

+ =2 (2.106)
∂x22 ∂x12 ∂x1 ∂x2

Repeating the same operation on the other components leads to two more equations which
have an identical format to equation (2.106) but with permuted indices

∂2 ε11 ∂2 ε33 ∂2 ε13

+ =2 (2.107)
∂x32 ∂x12 ∂x1 ∂x3
∂2 ε22 ∂2 ε33 ∂2 ε23
+ =2 (2.108)
∂x32 ∂x22 ∂x2 ∂x3

In a similar way further differentiation of the strain-displacement equations (2.53) also gives

∂2 ε11 ∂2 ε23 ∂2 ε13 ∂2 ε12

+ = + (2.109)
∂x2 ∂x3 ∂x12 ∂x1 ∂x2 ∂x1 ∂x3
∂2 ε22 ∂2 ε13 ∂2 ε23 ∂2 ε12
+ = + (2.110)
∂x1 ∂x3 ∂x22 ∂x1 ∂x2 ∂x2 ∂x3
∂2 ε33 ∂2 ε12 ∂2 ε13 ∂2 ε23
+ = + (2.111)
∂x1 ∂x2 ∂x32 ∂x2 ∂x3 ∂x1 ∂x3

which gives a total number of six equations, known as the St.Venant compatibility equations.
They are commonly represented in a compact index format

εkl ,i j + εi j ,kl = εi k,j l + ε j l ,i k (2.112)

This compact format stands for 81 equations, but only six of them are distinct while all the
other equations are redundant or equalities. The fact that 6 equations are found instead of
3 to restore the compatibility between strains and displacements, indicates that those 6 St.
Venant equations are not truly independent in the sense that further differential combina-
tions of these equations produce three connections between them (the Bianchi formulas).
However, since one cannot choose any three equations of the 6 St.Venant equations, one

52
 2.7. Special deformations

usually includes all 6 equations. It has been proven that these equations are necessary and
sufficient to ensure the existence of the displacement component functions that are related
to the linear strain tensor through (2.53).
The compatibility equations are sometimes derived in an another way, on the basis of
the continuity of matter. The strain field must be such that a closed loop before deformation
remains closed after deformation. The mathematical expression of this condition leads to
the same equations, but it explains why they are sometimes called the continuity equations
(they ensure the continuity of the displacement field for a given strain field).
Note that compatibility is automatically satisfied if strains are computed from the dis-
placement field, which is almost always the case in engineering practice.

2.7 Special deformations

2.7.1 Rigid body motion
A rigid body motion of a continuum occurs whenever external loads only lead to a transla-
tion and a rotation of the continuum (figure 2.11), without any deformation. Since there is
no deformation inside the continuum, relative displacement vectors of material points with
respect to other material points are equal to zero. A rigid body transformation of the ref-

r
Ni
+ = r
ni

t=0 t

Figure 2.11: Rigid body motion of a continuum.

erence to the current configuration may be considered as a pure rotation Q, followed by a

pure translation ~t of the continuum. The rotation tensor Q and the translation vector ~t are
constant in space, i.e., identical for all material points. If the centre of rotation is located at
~ , the current position ~
a spatial point P with initial position Y x of an arbitrary material point
in the continuum with initial position ~ X is given by:

~ ~ +~t +Q·( ~
x =Y ~)
X −Y (2.113)

The deformation tensor F for this rigid body motion is then given by

F = (~ x )T
∇0~
=Q (2.114)

Hence, the infinitesimal strain tensor is represented by

ε = 12 (Q +Q T ) − I (2.115)

A rotation tensor is an orthogonal tensor for which Q T = Q −1 . The infinitesimal strain tensor
is thus not equal to the null tensor, unless rotations are infinitely small (e.g., see section 2.4).

53
2. K INEMATICS

In that case only, Q approximates the unit tensor I . Clearly, this theory cannot be used for
large rigid body rotations. The Green-Lagrange strain tensor on the other hand, correctly
represents a zero strain state upon a rigid body rotation

E = 12 (C − I )
= 12 (F T ·F − I )
= 12 (Q T ·Q − I )
= 12 (Q −1 ·Q − I ) since Q is orthogonal
=0 (2.116)

2.7.2 Homogeneous deformation

If an arbitrary material vector d ~
X is deformed to the same d~
x in all material points, the de-
formation is independent of the considered material point and the continuum is deformed
homogeneously. The deformation tensor is then a constant tensor which is independent of
~ in the continuum. Since
the initial position X

d~ ~
x = F ·dX (2.117)

and since F is independent of ~

X , integration of (2.117) results in an equation that describes
the new position of an arbitrary material point in the continuum in terms of its reference
position
~ ~ +~t
x = F ·X (2.118)
where ~t represents the vectorial integration constant, i.e., a rigid translation. Since F is con-
stant, ε will be constant as well. All strain components are thus identical for all material
points.

2.7.3 Plane strain

The three-dimensional deformation state reduces to a plane deformation state if the strain
components εzz , εxz and ε yz are equal to zero. The existence of such a plane strain state
is frequently assumed for the analysis of flat specimens or structures, where the thickness
(which is taken in the z-direction) is so high that deformations in that direction become
negligible,
εzz = εxz = ε yz = 0 (2.119)
while the strain components in the plane are assumed to be independent of the z-direction.
The strain state is then fully described in the plane of the specimen by a plane strain tensor,
which has only 3 distinct components. This state of deformation is called a state of plane
strain with respect to the (x, y) plane.

2.7.4 Axisymmetric deformation

Many engineering components and structures have an axisymmetric geometry, which means
that their shape and volume can be constructed by virtually rotating a longitudinal cross sec-
tion around the axis of revolution. Points are indicated with cylindrical coordinates {r, θ, z}.
When material properties and loading are also independent of the coordinate θ, the defor-
mation and resulting stresses will be also independent of θ.

54
 2.7. Special deformations

z z

r r
r
θ

Figure 2.12: Axisymmetric deformation

∂
()=0 → ~
u = u r (r, z)~ e r (θ) + u t (r, z)~
e t (θ) + u z (r, z)~
ez
∂θ
 
2u r,r − 1r (u t ) + u t,r u r,z + u z,r
1
ε =  − 1r (u t ) + u t,r 2 r1 (u r ) u t,z 
2
u z,r + u r,z u t,z 2u z,z

If no rotation around the z-axis takes place (u t = 0), all state variables can be studied in
one half of a longitudinal cross section (containing the z-axis).

∂
( ) = 0 and u t = 0 → ~
u = u r (r, z)~
e r (θ) + u z (r, z)~
ez
∂θ
 
2u r,r 0 u r,z + u z,r
1
ε= 0 2 1r (u r ) 0 
2
u z,r + u r,z 0 2u z,z

Axisymmetric plane strain

When boundary conditions and material behavior are such that displacement of material
points are only in the r θ-plane, the deformation is referred to as plane strain in the r θ-plane.
plane strain deformation

u r = u r (r, θ) 
u t = u t (r, θ) → εzz = γr z = γt z = 0

uz = 0
linear strain matrix
 
2u r,r u t,r − r1 (u t ) 0
1
ε =  u t,r − r1 (u t ) 2
r (u r ) 0 
2
0 0 0
plane strain deformation with u t = 0

55
2. K INEMATICS

 
¾ 2u r,r 0 0
u r = u r (r ) 1 2
→ ε=  0 r
(u r ) 0 
uz = 0 2
0 0 0

56
 2.8. Summary

2.8 Summary

Kinematics

Nonlinear deformation map ~ ~ ,t)

x = φ( X
~ x )T
F = (∇0~

Deformation tensor d~x = F ·d~ x0

~
F = (∇0~ x )T
Fi J = xi ,J

Displacement gradient tensor υ = (~∇0 ~ u )T

T
d~u = υ ·d~ x0
F = I +υ
υi J = u i ,J

Strains and strain tensors

Engineering strain e = λ−1

Logarithmic strain εl n = ln λ
Green-Lagrange strain εg l = 21 (λ2 − 1)
Euler-Almansi strain εea = 21 (1 − λ12 )

Right Cauchy-Green tensor C = F T ·F

Green-Lagrange strain tensor E = 21 [C − I ]
d l 2 − d L2 = 2 d ~
X ·E ·d ~
X
T
Left Cauchy-Green tensor B = F ·F£ ¤
Euler-Almansi strain tensor E A = 21 I − B −1

57
2. K INEMATICS

Linearization of deformations

Linear strain ε = λ−1

Engineering strain ε=e
Logarithmic strain εl n ≈ ε if |ε| ≪ 1
Green-Lagrange strain εg l ≈ ε if |ε| ≪ 1
Euler-Almansi strain εea ≈ ε if |ε| ≪ 1
£ ¤
Linear strain tensor ε = 21 £υ + υT ¤
ε = 21 F + F T − I

Infinitesimal rotation tensor ω = 21 (~

∇0 u T − ~
∇0 u)

Properties of the linear strain tensor

Physical interpretation ε11 = λ1 − 1

ε12 =≈ 12 [θ2 + θ1 ]
Engineering shear strain γ12 = 2ε12
γ13 = 2ε31
γ23 = 2ε32

Principal strain ε1 , ε2 and ε3

Principal invariants J 1 (ε) = ε1 + ε2 + ε3
J 2 (ε) = ε1 ε2 + ε1 ε3 + ε2 ε3
J 3 (ε) = ε1 ε2 ε3

dV
Volumetric deformation J = det(F ) = dV 0
≈ 1 + tr(ε)
Cubical dilatation e = tr(ε) = dVdV
−dV0
0

Mean strain εm = 31 trε

Volumetric strain tensor εv = εm I
Deviatoric strain tensor εd = ε − ε v

58
 2.8. Summary

Compatibility

St.Venant compatibility εkl ,i j + εi j ,kl = εi k,j l + ε j l ,i k

Special deformations

Rigid body motion ~

x =Y~ +~t +Q·( ~ ~)
X −Y
1 T
ε = 2 (Q +Q ) − I ≈ 0, if Q small

Homogeneous deformation F constant, ε constant

Plane strain εzz = εxz = ε yz = 0

Axisymmetric deformation εzz = γr z = γt z = 0

59
 CHAPTER
3
F ORCE AND STRESS

This chapter is devoted to the fundamental concept of stress, which is here sketched within
the infinitesimal deformation context. The formal definition of the stress tensor is given,
together with the interpretation of its components in a Cartesian framework. Since it is a
second-order tensor, tensorial properties given for the strain tensor also apply to the stress
tensor.

3.1 Forces and stresses in a deformable continuum

Each device, no matter how small (e.g., micromechanical devices in a DVD or HDD) or large
(e.g., vehicles, bridges) is manufactured from a material in which all material points interact
with each other to sustain the external loads. In manufacturing processes or some particular
applications, the shape or volume of the component or device changes during this process.
For conventional use of a device or a mechanical component (e.g., engine parts) the initial
geometry may not change during use. In these cases, the combination of external loads and
the material response, mostly leads to cases where the theory of infinitesimal deformations
remains applicable. It is within this context, that we will here sketch the notion of the stress
tensor.

3.1.1 External forces

Consider a three-dimensional continuum, enclosed in a volume V with an external surface
Ω. The continuum is loaded by external loads which may have a different nature. Surface
~ are applied on material points located on the external boundary Ω, see figure 3.1.
forces p
These may be distributed loads (per unit of surface N /m 2 ), line loads or concentrated loads.
Body forces ~q act inside the continuum on individual material points (figure 3.1). Well-
known examples are gravitation forces (the weight of the material itself), or body forces in
the presence of transient electro-magnetic fields. Body forces ~ q (~
x ) are usually measured
with respect to unit of mass, i.e., N /kg . Per unit of volume, they can be written as
· ¸ · ¸· ¸
N kg N
ρ~q (~
x) 3
= 3
(3.1)
m m kg

61
3. F ORCE AND STRESS

r
p W
V

r
rq

Figure 3.1: External forces acting on a continuum

In real engineering materials, a true material point does not exist, since the material is
semi-continuous on a very small scale, e.g., atoms in a crystal lattice (metals) or molecular
chains (polymers) or grains (ceramics). Material points in the concept of the infinitesimal
deformation theory, are therefore best identified with an average response of the underlying
material microstructure. Once such material points are loaded, they will start to interact.
These interactions are visualized and measured through local deformations, while the resis-
tance against these deformations is expressed in terms of internal forces.

3.1.2 Internal forces

Each subvolume dV of the total volume V interacts with its immediate neighbourhood, once
external forces are applied. This implies that the material locally deforms, which is well de-
scribed with the already given kinematical framework. Internal forces arise from the atomic
or molecular interactions (e.g., deformation of the crystal lattice) upon deformation. Each
arbitrary material volume in a continuum deforms such that all forces applied to it equili-
brate one another. Internal forces are always in equilibrium with the internal forces applied
on the neighbouring material volumes (principle of action and reaction). If the considered
material volume is located on the boundary, the internal forces on the boundary segment
simply equal the external forces on that segment.

3.1.3 The stress vector

Based on the existence of these internal forces, a so-called stress vector can be defined. Let’s
consider a small volume ∆V inside the material, where we focus our attention on a part of
its boundary ∆S. Through this surface ∆S the considered volume exerts a force ∆ ~ f on the
neighbouring material, which is equal and opposite to the force acted by that neighbouring
material on that surface. If we now take the limit for ∆V and hence ∆S to zero, we find the
infinitesimal force d ~f acting on an infinitesimal material plane d S for a material point, see
figure 3.2. Since d S is plane in this limit, it can be identified by its outward normal ~
n . The
~
stress vector in this material point, acting on a plane with orientation n is the defined by

d~
f
~=
p (3.2)
dS

62
 3.1. Forces and stresses in a deformable continuum

r
p r
n
V dS
dV
P

Figure 3.2: The stress vector on a plane in a material point

The stress vector (also called the traction) is the internal force vector per unit of surface that
a material point undergoes on a surface with a given direction ~ n . The true or Cauchy stress
vector, as defined by(3.2), is defined with respect of the true (i.e., in the actual deformed
state) surface in the material. Hence, d S and ~ n are quantities measured with respect to the
deformed state. Since only infinitesimal displacements are considered in this framework,
the following simplifications may be made

d~
f d~f
~=
p ≈ ~
n ≈~
n0 (3.3)
dS d S0

Again, no distinction is made between the undeformed and deformed geometry on which
the internal forces act. This assumption is not valid in the presence of large deformations,
which will lead to different types of stress vectors, depending on the considered geometry.
The fact that the name ’Cauchy’ or ’true’ stress vector is used in the infinitesimal context
may be misleading, since one simply does not distinguish the ’true’ stress vector from any
other definition. No distinction between d S ↔ d S 0 or ~n ↔~n 0 will therefore be made in the
remainder of this chapter.
Note that each orientation corresponds to a specific stress vector.

3.1.4 The stress tensor

The stress vector is related to each possible orientation of a facet through a linear transfor-
mation, since the considered material volume on which it is defined is infinitely small. Such
a linear transformation is evidently described by a tensorial quantity, the stress tensor σ

~ = σ·~
p n (3.4)

The stress tensor σ gives the stress vector p ~ for each possible orientation ~ n of the infinitesi-
mal facet d S. It unifies all stress vectors for all possible orientations.
The fact that the transformation from the normal ~ n to the stress vector ~
p is described by a
second-order tensor, means that there is an exact similarity between concepts, components,
invariants that have been derived from σ compared to those derived from the strain tensor
ε. Another important property of the stress tensor is its symmetry, which will be proven
later in this chapter. Both ε and σ are thus symmetric second-order tensors. The following
paragraphs are therefore a natural extension of the previous chapter to the stress tensor.

63
3. F ORCE AND STRESS

3.2 Components of the stress tensor with respect to a basis

3.2.1 Normal stresses
The component σi i of the stress tensor σ along an arbitrary direction ~
n i , is called the normal
stress along that direction. The normal stress is determined with

σi i = ~
n i ·σ·~
ni (3.5)
=~
n i ·~
pi (3.6)

where p~ is the stress vector that acts on the infinitesimal surface perpendicular to N ~i .
Using a Cartesian vector basis {~ e 1 ,~
e 2 ,~
e 3 }, the normal stresses along the coordinate axes
are denoted σ11 , σ22 , σ33

σ11 = ~
e 1 ·σ·~
e1 (3.7)
σ22 = ~
e 2 ·σ·~
e2 (3.8)
σ33 = ~
e 3 ·σ·~
e3 (3.9)

The normal stress components σi i are thus located on the diagonal of the associated stress
matrix σ with respect to that basis. The normal stresses are (work-)conjugated to the normal
strains.

3.2.2 Shear stresses

The off-diagonal components σi j (i 6= j ) of the stress tensor σ are called the tensorial shear
stresses. The shear stress σi j between two perpendicular directions ~ n i and ~
n j (~ n j = 0) is
n i ·~
given by

σi j = ~
n i ·σ·~
nj
=~
n i ·~
pj (3.10)

The shear stress thus represents the component with respect to ~ ~j

n i of the stress vector p
acting on a surface perpendicular to ~
n j . Note that the vector ~
n i also lies in that surface.
Shear stresses are commonly denoted by the symbol τ. One may thus write

τ i j = σi j (3.11)

3.3 Principal stresses and stress invariants

Like ε, the stress tensor σ is a symmetric second-order tensor, for which the eigenvectors
and eigenvalues are real. The eigenvalue problem of σ is solved identically to section 2.5.3,
i.e., by
σ·M~ i = σi M
~i (3.12)
~ i the eigenvectors. The eigenvalues are again
where σi (i = 1, 2, 3) are the eigenvalues and M
obtained from the cubic characteristic equation for the stress tensor

det(σ − σI ) = 0 (3.13)
σ3 − J 1 (σ)σ2 + J 2 (σ)σ − J 3 (σ) = 0 (3.14)

64
 3.4. Hydrostatic and deviatoric stresses

The three stress invariants J 1 (σ), J 2 (σ) and J 3 (σ) of the stress tensor are those given in the
chapter on vectors and tensors. The roots of equation (3.14) are the eigenvalues, called the
principal stresses σ1 , σ2 and σ3 . The eigenvectors, or principal stress directions M ~ i are again
computed from
(σ − σi I )·M~ i = ~0 ∀i ∈ 1, 2, 3 (3.15)

where the vectors M ~ i are normalized, i.e., ||M ~ i || = 1. If the three principal stresses are dis-
tinct, the three principal stress directions will be mutually orthogonal. If two principal stresses
are equal, equation (2.83) has one exact solution and a set of an infinite number of solutions,
out of which any pair of orthogonal vectors can be chosen as the two remaining principal
stress directions. If all principal stresses are equal, each set of three mutually orthogonal
vectors is a set of principal stress directions. Such a state is called hydrostatic, since the
stress is equal in all directions as it is the case in static fluids. An axis that is oriented along a
principal stress direction is called a principal stress axis. A plane normal to a principal stress
direction is called a principal plane of σ. If a Cartesian basis is taken with three axes that
coincide with the principal stress axes, the matrix form of σ takes the same particular format
as for the strain tensor, i.e.,
 
σ1 0 0
σ =  0 σ2 0  (3.16)
0 0 σ3

The shear stress components are also equal to zero along these directions. Like for the anal-
ysis of the strain tensor, normal stresses attain an extreme value along the principal stress
axes. The largest and smallest normal stress for all possible orientations M ~ can always be
found among the principal stresses.
The three principal stress invariants J 1(σ), J 2 (σ) and J 3 (σ) can be computed directly from
the stress tensor, or alternatively from the principal stresses by

J 1 (σ) = σ1 + σ2 + σ3 (3.17)
J 2 (σ) = σ1 σ2 + σ1 σ3 + σ2 σ3 (3.18)
J 3 (σ) = σ1 σ2 σ3 (3.19)

Any scalar function of the three principal stress invariants is a stress invariant as well.
Note that in general the principal stress directions differ from the principal strain direc-
tions. Their mutual relationship depends on the constitutive behaviour of the considered
material.

3.4 Hydrostatic and deviatoric stresses

3.4.1 Hydrostatic pressure
For the analysis of strains, it was shown that volume changes lead to a volumetric strain that
equaled the trace of the infinitesimal strain tensor tr(ε). Such volume changes are mostly
mechanically associated to the existence of a scalar pressure that acts in all directions on a
material point. On the basis of this observation, one can split the stress tensor σ in two parts,
a hydrostatic part σh and a deviatoric part σd . The hydrostatic part is computed from the

65
3. F ORCE AND STRESS

pressure p that equals minus the mean stress in the material point, i.e.,
p = −σm = − 31 tr(σ) (3.20)
σ1 + σ2 + σ3
=− (3.21)
3
The hydrostatic stress tensor applies this pressure to all directions through
σh = −p I (3.22)
For many materials, there exists a clear constitutive relationship that relates the hydrostatic
stress tensor σh to its kinematical counterpart, the volumetric strain tensor εv .

y
p
σxx = p
σy y = p e x~
σ = p(~ e y~
e x +~ e z~
e y +~ ez )
p
σzz = p
p x

y
p
σxx = −p
σ y y = −p e x~
σ = −p(~ e y~
e x +~ e z~
e y +~ ez )
p σzz = −p
p x

Figure 3.3: Hydrostatic stress state

3.4.2 Deviatoric stresses

The remaining part of the stress tensor, after subtraction of its hydrostatic part, is called the
deviatoric stress tensor σd .
σd = σ − σh (3.23)
This part of the stress tensor has a zero mean stress (i.e., zero pressure). Again, for many ma-
terials there exists a direct relation between this deviatoric stress tensor and its kinematical
equivalent, the deviatoric strain tensor. It has been shown that shape changes of a material
volume element are associated to the latter tensor.
Like in section 2.5.4, the deviatoric part has three principal invariants as well J 1 (σd ),
J 2 (σd ) and J 3 (σd ). The first invariant is by definition zero
J 1 (σd ) = 0 (3.24)
while the following relations again apply for the two remaining invariants
J 2 (σd ) = 3 p 2 − J 2 (σ) (3.25)
= − 21 σd : σd (3.26)
J 3 (σd ) = J 3 (σ) + p J 2 (σ) − p 3 (3.27)

66
 3.5. Stress state with respect to a plane

A fully deviatoric stress state is typically encountered in the axial torsion of a thin-walled
tube (radius R, wall thickness t ), which is the result of an applied axial torsional moment
(torque) T . This load causes a shear stress τ in the cross-sectional wall. The shear stress
has the same value in each point of the cross-section, but changes its orientation with the
angular position of the point. Indeed, the shear stress τ follows the circumferential direction.

τ = σzx
y τ τ

τ
τ
x
T τ = σ yx T
z τ=
2πR 2 t
τ = −σzx
τ τ
τ
τ
τ = −σ yx

Figure 3.4: Deviatoric stress state in the wall of a tube under torsion

σd = σ = τ(~
e i~ e j~
e j +~ e i ) with i 6= j

3.5 Stress state with respect to a plane

3.5.1 Definition
The definition of the stress tensor is already based on the existence of resultant stress vector
~
p that acts on an infinitesimal plane in the material. The normal stress σnn on a plane is
clearly the normal stress along the normal ~ n on that plane. The shear stress τns on the plane
is the largest shear stress between the normal ~ n and all directions that lie in the plane. In
fact τns is the maximum shear stress in that particular plane with respect to its normal. The
in-plane direction according to which this shear stress is maximal is denoted with the unit
vector ~s, which is obtained through the orthogonal projection of the stress vector p ~ on the
~ is simply decomposed into two vectors, one oriented along
plane. In fact the stress vector p
the normal σnn ~ n and one in the plane τns~ s. The scalar values of the normal and shear stress
are given by

σnn = ~
n ·σ·~
n (3.28)
τns = σns = ~
s·σ·~
n (3.29)

It is convenient to express these quantities in terms of the principal stresses σ1 , σ2 , σ3 and

~ 1, M
the associated principal stress directions {M ~ 2, M
~ 3 }. Taking a Cartesian basis with axes

67
3. F ORCE AND STRESS

that coincide with the principal directions, permits to express the normal on an arbitrary
oblique plane as
~
n = m1 M ~ 1 + m2 M
~ 2 + m3 M
~3 (3.30)
with the normalisation condition
m 12 + m 22 + m 32 = 1 (3.31)
The normal stress σnn is expanded from (3.28) as
~ 1 + m2 M
σnn = (m 1 M ~ 2 + m3 M
~ 3 )·σ·(m 1 M
~ 1 + m2 M
~ 2 + m3 M
~ 3)
= m 12 σ1 + m 22 σ2 + m 32 σ3 (3.32)
Likewise, the shear stress along vector ~
s is given by
n − σnn ~
τns = ||σ·~ n || (3.33)
Since the vectors ~
n and ~
s are perpendicular, this relation simplifies to
τns = [ (σ·~
n )·(σ·~n ) − 2σ2nn + σ2nn ]0.5 = [ (σ·~ n ) − σ2nn ]0.5
n )·(σ·~
¡ ¢
= [ m 12 σ21 + m 22 σ22 + m 32 σ23 −
¡ 4 2 ¢
m 1 σ1 + m 24 σ22 + m 34 σ23 + 2m 12 m 22 σ1 σ2 + 2m 12 m 32 σ1 σ3 + 2m 22m 32 σ2 σ3 ]0.5
which is easily simplified by substituting m 14 by m 12 (1 − m 22 − m 32 ) and m 24 and m 34 by similar
expressions derived from (3.31)
q
τns = m 12 m 22 (σ1 − σ2 )2 + m 12 m 32 (σ1 − σ3 )2 + m 22 m 32 (σ2 − σ3 )2 (3.34)
If all principal stresses are equal, the stress state is fully hydrostatic and all directions may
be considered as principal. The latter equation shows that indeed all shear stresses are then
equal to zero, independent of the orientation of the considered plane.

3.5.2 Octahedral stresses

Octahedral planes are planes with normals that bisect the principal stress directions. If the
~ 1, M
principal stress directions are given by the set {M ~ 2, M
~ 3 }, the normals of the octahedral
planes are defined by
1 ¡ ¢
n oct = p ±M
~ ~1±M ~2±M ~3 (3.35)
3
Taking into account all possible signs in this relation yields 8 planes which may be repre-
sented in each octant by a triangle. All eight triangles form an octahedron, which is visual-
ized in figure 3.5. The shear stresses on these planes are the octahedral shear stresses. p
For
oct
the octahedral planes the direction cosines m 1 , m 2 , m 3 of the normal ~
n are simply ±1/ 3.
The octahedral normal stress σoct is computed from (3.32), which gives
1
σoct =
(σ1 + σ2 + σ3 )
3
1
= tr(σ) = −p (3.36)
3
i.e., minus the hydrostatic pressure in the considered point. The octahedral shear stress
computed with (3.34) is given by
q
1
τoct = 3 (σ1 − σ2 )2 + (σ1 − σ3 )2 + (σ2 − σ3 )2 (3.37)

68
 3.5. Stress state with respect to a plane

r
M3

r
noct
r
noct

r
M1

r
noct
r
M2

Figure 3.5: The octahedral planes oriented with respect to the principal stress axes in a ma-
terial point

3.5.3 Maximum shear stress in a material point

Since the analysis on a plane provides us the largest shear stress between the normal on that
plane and all possible orientations of line elements in the plane, it is convenient to trace
the maximum shear stress in a material point for all possible orientations of that plane. De-
pendent on the direction of the plane, the shear stress attains an extreme value if the partial
derivatives with respect to the independent direction cosines vanish. First m 1 and m 2 are
taken as the independent variables, while m 3 is dependent on them through (3.31). Substi-
tuting m 32 by 1 − m 12 − m 22 and taking next the partial derivatives yields

∂τns m1 £ 2 ¤
= m 2 (σ1 − σ2 )2 + (1 − 2m 12 − m 22 )(σ1 − σ3 )2 − m 22 (σ2 − σ3 )2
∂m 1 τns
=0 (3.38)

and

∂τns m2 £ 2 ¤
= m 1 (σ1 − σ2 )2 − m 12 (σ1 − σ3 )2 + (1 − m 12 − 2m 22 )(σ2 − σ3 )2
∂m 2 τns
=0 (3.39)

These equations are satisfied for the following orientations

p p
m1 = 0 m 2 = ±1/ 2 ⇒ m 3 = ±1/ 2
p p
m 1 = ±1/ 2 m 2 = 0 ⇒ m 3 = ±1/ 2

Note that m 1 = 0 and m 2 = 0 (and thus m 3 = 1) gives a shear stress equal to zero, which are
the principal directions themselves. In an analogous way, {m 1 , m 3 } and {m 2 , m 3 } could have

69
3. F ORCE AND STRESS

been taken as independent variables, which adds the following orientations

p p
m1 = 0 m 3 = ±1/ 2 ⇒ m 2 = ±1/ 2
p p
m 1 = ±1/ 2 m 3 = 0 ⇒ m 2 = ±1/ 2
p p
m2 = 0 m 3 = ±1/ 2 ⇒ m 1 = ±1/ 2
p p
m 2 = ±1/ 2 m 3 = 0 ⇒ m 1 = ±1/ 2
The planes of maximum shear stress are therefore characterised by the following normals
n τ1max = cos 45o M
~ ~ 1 + sin 45o M
~2
n τ2max = cos 45o M
~ ~ 1 − sin 45o M
~2
n τ3max = cos 45o M
~ ~ 1 + sin 45o M
~3
n τ4max = cos 45o M
~ ~ 1 − sin 45o M
~3
n τ5max = cos 45o M
~ ~ 2 + sin 45o M
~3
n τ6max = cos 45o M
~ ~ 2 − sin 45o M
~3

Two of these planes are depicted with respect to the principal directions in figure 3.6. The

r
M3

r
M2
45o 45o

r
M1

Figure 3.6: Planes with maximum shear stress in a material point

planes with maximum shear stresses always pas through one of the principal axes, while the
normal bisects the two remaining principal stress axes. The six values of the maximum shear
stresses are then given by
σ1 − σ2
τns max,1 = ± (3.40)
2
σ1 − σ3
τns max,2 = ± (3.41)
2
σ2 − σ3
τns max,3 = ± (3.42)
2

70
 3.6. Plane stress

3.6 Plane stress

3.6.1 Definition
A three-dimensional stress state approximately reduces to a plane stress state if the stress
components σzz , σxz and σ yz are equal to zero. The assumed existence of such a plane stress
state is typically justified for thin flat specimens or structures, where the thickness (in the
z-direction perpendicular to the plane of loading) is sufficiently small in order to neglect the
stresses in that direction, see figure 3.7.

σzz = σxz = σ yz = 0 (3.43)

The stress components in the plane are assumed to be independent of the thickness di-

y
y
s yy
t xy

s xx s xx z
x

t xy
s yy

Figure 3.7: A state of plane stress

rection (i.c. z-direction). The stress state is then fully characterized by its components in
the plane of the specimen, i.e., a plane stress tensor with only 3 distinct components. This
state of deformation is called a state of plane stress with respect to the (x, y) plane. Note
that the difference with the plane strain assumption resides in the thickness of the speci-
men, which must be small compared to the in-plane dimensions for plane stress and large
for plane strain.

3.6.2 Mohr’s stress circle

Since Mohr’s circle is a geometrical representation of the property of a symmetric second-
order tensor in plane loading states, it can be easily sketched for a plane stress state.
The equations which permit to elaborate the governing equations of Mohr’s stress circle
can be easily derived. Two Cartesian in plane coordinate systems are used, where the com-
ponents of the stress tensor can be defined with respect to both coordinate systems. The
original coordinate system is defined by the base vectors {~ e X ,~
e Y }, which are rotated to the
new base vectors {~ e y } over an angle θ, see figure 3.8. The following relations then apply
e x ,~

71
3. F ORCE AND STRESS

Y
Y x
s YY
y r
n
t XY
t xy s xx
s XX
q q
X q
X

Figure 3.8: Stress components on rotated planes in a material point

1
σxx = 2 (σ X X + σY Y ) + 21 (σ X X − σY Y ) cos 2θ + τ X Y sin 2θ (3.44)
1 1
σy y = 2 (σ X X + σY Y ) − 2 (σ X X − σY Y ) cos 2θ − τ X Y sin 2θ (3.45)
1
τx y = 2 (σY Y
− σ X X ) sin 2θ + τ X Y cos 2θ (3.46)

From these equations, the following transformations rules can be determined for the
stress components σxx , σ y y , σx y as a function of the principal stresses σ1 and σ2 (X and Y
are taken along the principal axes of stress)

1
σxx = 2 (σ1
+ σ2 ) + 21 (σ1 − σ2 )cos 2θ (3.47)
1 1
σy y = 2 (σ1 + σ2 ) − 2 (σ1 − σ2 )cos 2θ (3.48)
τx y = − 21 (σ1 − σ2 ) sin 2θ (3.49)

The angle θ is again measured counterclockwise with respect to the axis of the largest prin-
cipal stress σ1 . Stress components can be quantified for different orientations of the axes x
and y by simply changing θ. Mohr’s stress circle is now constructed with the normal stress
along x on the horizontal axis and the shear stress between x and y along the vertical axis,
see figure 3.9. The locus of points for σxx and τx y is then given by

(σxx − a)2 + τ2x y = R 2 (3.50)

with

1
a = 2 (σ1
+ σ2 ) (3.51)
1
R = 2 (σ1 − σ2 ) (3.52)

A state of stress measured with respect to axes x, y rotated over an angle θ is found on the
circle on an angle 2θ measured counterclockwise from the horizontal axis.

72
 3.6. Plane stress

(s xx , t xy )

s1 -s 2
2

s2 2q s xx
s1

s1 -s 2
2

s1 +s2
t xy 2

Figure 3.9: Mohr’s circle for plane stress

3.6.3 Mohr’s stress circle for three-dimensional stress states

Let us consider a three-dimensional stress state, characterized by the stress tensor σ with
its three principal stresses σ1 , σ2 , σ3 . Otto Mohr showed that the normal stress σnn and the
shear stress τns acting on a plane with a 3D orientation, can be plotted on a plane with σnn
and τns as coordinates as shown in figure 3.10. The locus of the points that represents all
possible 3D stress states on the different planes necessarily falls in a closed domain, which
is represented by the shaded area in the figure. This area is bounded by the three plane cir-
cles. This result confirms the results that were found earlier for the maximum shear stresses.
Indeed, if σ1 Ê σ2 Ê σ3 then (σ1 − σ3 )/2 is the largest shear stress, and the plane on which it
acts bisects the principal directions M ~ 1 and M
~ 3.

73
3. F ORCE AND STRESS

s3 s2 s1 s nn

tn s
Figure 3.10: Mohr’s circle for three-dimensional stress states

74
 3.7. Summary

3.7 Summary

External forces

Body forces ρ~
q (~
x)

Surface forces ~
p

The stress tensor

Stress vector ~
p = σ·~
n

Normal stresses σi i = N~ i ·~
pi
Shear stresses τ i j = σi j = N~ i ·~
pj

Principal stresses σ·M~ i = σi M

~i
det(σ − σI ) = 0
~ i =~0
(σ − σi I )·M ∀i ∈ 1, 2, 3

Hydrostatic pressure p = −σm = − 13 tr(σ)

Hydrostatic stress tensor σh = −p I
Deviatoric stress tensor σd = σ − σh

Stress state with respect to a plane

Normal stress σnn = ~

n ·σ·~
n
Shear stress τns = σns = ~s·σ·~
n
q
τns = m 1 m 2 (σ1 − σ2 )2 + m 12 m 32 (σ1 − σ3 )2 + m 22 m 32 (σ2 − σ3 )2
2 2

Octahedral stresses σoct = −p

p
τoct = 13 (σ1 − σ2 )2 + (σ1 − σ3 )2 + (σ2 − σ3 )2

Maximum shear stresses τns max,1 = ± σ1 −σ

2
2

σ −σ
τns max,2 = ± 1 2 3
τns max,3 = ± σ2 −σ
2
3

75
3. F ORCE AND STRESS

Plane stress

Plane stress σzz = σxz = σ yz = 0

Mohr’s stress circle (σxx − a)2 + τ2x y = R 2

R = 21 (σ1 − σ2 )
a = 12 (σ1 + σ2 )

76
 CHAPTER
4
B ALANCE AND EQUILIBRIUM LAWS

In previous chapters we have studied forces acting on a continuum, the resulting deforma-
tion of the continuum, and the stresses that are caused by the deformation. The relation
between forces, deformation and stress is governed by two types of equations. The first type,
the balance laws, are mathematical formulations of universal physical principles: balance
of mass, momentum, and moment of momentum. This chapter is devoted to these balance
equations for mechanical problems in a small deformation context.

4.1 Balance of mass

We consider a body with a volume V0 , a surface area A 0 and a density ρ 0 in the reference
state t0 . In the deformed state t the body occupies a volume V while having a surface area
A and a density ρ. To derive the law of balance of mass, we concentrate on an arbitrary,

t0 t

A0
V̄0 V̄ A
V0

Figure 4.1: A body in the initial state t0 and the deformed state t , with arbitrary subregion V̄ .

fixed amount of material points in a subregion V̄0 , i.e. we use a Lagrangian approach. In the
initial state these material points occupied a volume V̄0 while having a surface area Ā 0 and a
density ρ 0 . According to the law of balance of mass, the mass of these material points must

77
4. M ECHANICAL EQUILIBRIUM

have remained constant during the deformation:

Z Z
ρ dV = ρ 0 dV0 (4.1)
V̄ V̄0

This equation is called the global formulation of the law of balance of mass. From the theory
of kinematics, we know that the volumes before and after deformation are related as d V̄ =
J d V̄0 , where J represents the determinant of the deformation tensor. Thus we can transform
the left-hand integral into an integral over the undeformed volume:
Z Z
ρ J dV0 = ρ 0 dV0 (4.2)
V̄0 V̄0

which can also be written as: Z

(ρ J − ρ 0 ) dV0 = 0 (4.3)
V̄0

Since this integral must hold for any arbitrary volume V̄0 , it follows that the integrand must
be zero:
ρ J = ρ0 (4.4)

This equation is called the local formulation of the law of balance of mass.

4.2 Balance of momentum

4.2.1 General derivation
Let’s consider a fixed amount of matter having a volume V , a surface A and a density ρ. The
matter is acted upon by body forces ~ p . The total momentum ~
q and surface forces ~ i of the

~
p

~
q Ā
V̄ A V̄

Figure 4.2: Forces ~

q and ~
p acting on a subvolume V̄ of a body.

matter equals:
Z
~
i= ρ~
v dV (4.5)
V̄

78
 4.2. Balance of momentum

The total resulting force ~

f , acting on the matter, is the sum of all body forces and surface
forces:
Z Z
~
f = ρ~ q dV + p ~d A
V̄ Ā
Z Z
= ρ~
q dV + ndA
σ·~
V̄ Ā
Z
= (ρ~ ∇·σT ) dV
q +~ (4.6)
V̄

where we have applied Gauss’s theorem to convert the surface integral into a volume integral.
According to the law of balance of momentum, this resulting force is equal to the rate of
change of the momentum of the matter in V̄ :
Z Z
T d
(ρ~ ~
q + ∇·σ ) dV = ρ~v dV (4.7)
dt
V̄ V̄

To elaborate the differentiation with respect to time, we convert the right-hand volume inte-
gral to the undeformed situation:
Z Z
d d
ρ~
v dV = v d V¯0
ρ 0~
dt dt
V̄ V
Z 0
= ρ 0~ v˙ d V¯0
V0
Z
= v˙ d V̄
ρ~ (4.8)
V

where we used the local form of the law of balance of mass (ρdV = ρ 0 dV0 ). The global for-
mulation of the law of balance of momentum can thus be written as:
Z Z
~ T
q + ∇·σ ) dV = ρ~
(ρ~ v˙ dV (4.9)
V̄ V̄

Now again, because the volume V̄ may be chosen arbitrarily, the integrand must be zero,
yielding the local formulation of the law of balance of momentum:

∇·σT + ρ~
~ v˙
q = ρ~ (4.10)

These equations are also called Cauchy’s equations of motion. In case of static equilibrium,
v˙ is zero, and the equations of motion are reduced to the equilibrium equa-
the acceleration ~
tions:
∇·σT + ρ~
~ q =~0 (4.11)

4.2.2 Translational equilibrium

Let’s consider a three-dimensional parallelepiped that interconnects parallel facets in two
neighbouring material points P and Q. The infinitesimal volume is depicted in figure 4.3

79
4. M ECHANICAL EQUILIBRIUM

¶s zz
s zz + dz
r
ez ¶z

¶s yz
s yz + dz
¶s xz ¶z
s xz + dz
¶z

s xx
Q ¶s zy
¶s zx s yx s zy + dy
s zx + dx ¶y
s xy ¶x ¶s yy
s yy s yy + dy
s zx ¶y

¶s yx ¶s xy
s yx + dx s xy + dy
¶s xx ¶x ¶y
s zy s xx + dx
¶x
P
s xz r
s yz
ey

s zz
r
ex

Figure 4.3: Three-dimensional stress parallelepiped

along with the Cartesian stress components that act on all the six facets. The back, left and
bottom facets are representative for the first material point , while the other remaining facets
belong to a material point that is located d x, d y and d z further in space. The stress compo-
nents differ through this spatial difference as indicated in the figure.
The translational or force equilibrium of the stress parallelepiped with respect to the
three Cartesian axes is here equivalent to the balance of momentum. Let’s solve the equi-
librium along the y-axis. To this purpose, the stress components oriented along the y-axis
are sketched in a separate figure, see figure 4.4. Taking into account these stress components
and the surfaces on which they act, leads to the following force equilibrium along the y-axis

³ ´ ³ ´ ³ ´
∂σ y y ∂σ y z ∂σ y x
σy y + ∂y d y d xd z + σ yz + ∂z d z d xd y + σ yx + ∂x d x d yd z
(4.12)
− σ y y d xd z − σ yz d xd y − σ yx d yd z + ρq y d xd yd z = 0

where ρq y is the volume force in the y − d i r ect i on. Simplifying this equation, and dividing
by d xd yd z yields
∂σ yx ∂σ y y ∂σ yz
+ + + ρq y = 0 (4.13)
∂x ∂y ∂z

The equilibrium along the x-axis and the z-axis delivers similar equations, which results in

80
 4.2. Balance of momentum

r
ez

¶s yz
s yz + dz
¶z

s yx ¶s yy
s yy s yy + dy
¶y

¶s yx
dz s yx + dx
¶x

r
s yz ey
dx
dy
r
ex

Figure 4.4: Contributing components for the translational equilibrium along the y-axis

a set of partial differential equations

∂σxx ∂σx y ∂σxz
+ + + ρq x = 0 (4.14)
∂x ∂y ∂z
∂σ yx ∂σ y y ∂σ yz
+ + + ρq y = 0 (4.15)
∂x ∂y ∂z
∂σzx ∂σz y ∂σzz
+ + + ρq z = 0 (4.16)
∂x ∂y ∂z
which can be written in index notation as

σi j ,j + ρq i = 0 (4.17)

or in its tensorial equivalent

∇·σT + ρ~
~ q = ~0 (4.18)
The latter equation is the general equation for static equilibrium. Note that equation (4.18)
is independent of the coordinate system used, which means that is can also be applied to
e.g., cylindrical coordinate problems.

4.2.3 Equilibrium in a Cartesian and cylindrical basis

The equilibrium equation can be written in components w.r.t. a Cartesian vector basis. This
results in three partial differential equations, one for each coordinate direction.

81
4. M ECHANICAL EQUILIBRIUM

σxx,x + σx y,y + σxz,z + ρq x = 0

σ yx,x + σ y y,y + σ yz,z + ρq y = 0
σzx,x + σz y,y + σzz,z + ρq z = 0
Writing tensor and vectors in components w.r.t. a cylindrical vector basis is more elab-
orative because the cylindrical base vectors ~
e r and ~
e t are a function of the coordinate θ, so
they have to be differentiated, when expanding the divergence term.
1 1
σr r,r + σr t,t + (σr r − σt t ) + σr z,z + ρq r = 0
r r
1 1
σtr,r + σt t,t + (σtr + σr t ) + σt z,z + ρq t = 0
r r
1 1
σzr,r + σzt,t + σzr + σzz,z + ρq z = 0
r r

4.3 Balance of moment of momentum

4.3.1 General derivation
The balance of moment of momentum states that the total moment about a fixed point of
~ o ), equals the change
all forces working on a randomly chosen volume of material points (M
of the total moment of momentum of the material points inside the volume, taken w.r.t. the
same fixed point (~L o ).

~
p

~
q Ā
V̄ A V̄

V
~
x

Figure 4.5: Moment of forces on a random section of a material body

Z
d~Lo d
~o =
M = ~
x × ρ~
v dV ∀ V̄ (4.19)
dt dt
V̄
Making use of the mass balance and mathematical manipulations, it can be shown that
this reduces to the following condition:
   
σ32 − σ23 0
 σ13 − σ31  =  0 
σ21 − σ12 0

82
 4.3. Balance of moment of momentum

The balance of moment of momentum therefore implies that the Cauchy stress tensor has
to be symmetric

σT = σ ∀ ~
x ∈ V (t ) (4.20)

As a result of the balance of moment of momentum, the symmetric stress tensor simpli-
fies the balance of momentum to

~ v˙
q = ρ~
∇·σ + ρ~ (4.21)

4.3.2 Moment or rotational equilibrium

The equilibrium of a material element also requires a zero resulting moment. This is the mo-
ment or rotational equilibrium. The derivation is here performed with respect to the x-axis,
other axes lead to similar results (cyclic permutation of indices). The infinitesimal paral-
lelepiped is reconsidered, but only with the components which cause a non-zero moment
around the x-axis, see figure 4.6. Properly taking care of the moment arm around the x-axis

¶s zz
r s zz + dz
ez ¶z

¶s yz
s yz + dz
¶z

¶s zy
s zy + dy
s zx +
¶s zx
dx
s yx ¶y
¶x
s yy ¶s yy
s zx s yy + dy
¶y

dz s yx +
¶s yx
dx
¶x

r
ey
dx
dy
s zz
r
ex

Figure 4.6: Contributing components for the moment around the x-axis

83
4. M ECHANICAL EQUILIBRIUM

for each component leads to

³
∂σ y y
´ dz dz
− σ y y + ∂y d y d xd z + σ y y d xd z
³ ´ 2 ³ 2 ´
∂σz y ∂σ y z
+ σz y + ∂y d y d xd zd y − σ yz + ∂z d z d xd yd z
³ ´ dy dy
+ σzz + ∂σ zz
∂z d z d xd y − σzz d xd y
2 2
³ ´ d y d y (4.22)
∂σ
+ σzx + ∂xzx d x d yd z − σzx d yd z
2 2
³ ´ d z d z
∂σ y x
− σ yx + ∂x d x d yd z + σ yx d yd z
2 2
dz dy
− ρq y d xd yd z + −ρq z d xd yd z =0
2 2
Dividing this equation by d xd yd z and taking the limit d x → 0, d y → 0 and d z → 0 finally
gives the straightforward equation
σ yz = σz y (4.23)
Completing this with the rotational equilibrium around the y and z-axis gives

σx y = σ yx
σ yz = σz y
σxz = σzx

The rotational equilibrium leads to the fundamental conclusion that the stress tensor is sym-
metric, a property which was already exploited earlier in this chapter. In the tensorial format
this simply reads
σ = σT (4.24)
Note that the symmetry of σ permits to rewrite the translational equilibrium as

~ q = ~0
∇·σ + ρ~ (4.25)

which is the commonly used format for static equilibrium.

4.3.3 Rotational equilibrium in a Cartesian and cylindrical basis

With respect to a Cartesian or cylindrical basis the symmetry of the stress tensor results in
three equations.

σ = σT →

Cartesian : σx y = σ yx ; σ yz = σz y ; σzx = σxz

cylindrical : σr t = σtr ; σt z = σzt ; σzr = σr z

4.4 Special deformation states

The three-dimensional equilibrium equations can be simplified for special deformation or
stress states, such as plane strain, plane stress and axisymmetric cases.

84
 4.4. Special deformation states

4.4.1 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
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

σxx,x + σx y,y + ρq x = 0
σ yx,x + σ y y,y + ρq y = 0
σx y = σ yx

cylindrical components

1 1
σr r,r + σr t,t + (σr r − σt t ) + ρq r = 0
r r
1 1
σtr,r + σt t,t + (σtr + σr t ) + ρq t = 0
r r

σr t = σtr

4.4.2 Axisymmetric deformation

In many cases the geometry, boundary conditions and material behavior is such that no
∂
state variable depends on the circumferential coordinate θ : ∂θ = 0. For such axisymmetric
deformations, the equilibrium equations can be simplified considerably.
In many axisymmetric deformations the boundary conditions are such that there is no
displacement in the circumferential direction : u t = 0. In these cases there are only four
relevant strain and stress components and only three equilibrium equations.
When boundary conditions and material behavior are such that displacement of material
points are only in the r θ-plane, the deformation is referred to as plane strain in the r θ-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 θ-plane.

1
σr r,r + (σr r − σt t ) + σr z,z + ρq r = 0
r
2
σtr,r + (σtr ) + σt z,z + ρq t = 0 (if u t 6= 0)
r
1
σzr,r + σzr + σzz,z + ρq z = 0
r
σr t = σtr ; σt z = σzt (if u t 6= 0)
σzr = σr z

planar
1
σr r,r + (σr r − σt t ) + ρq r = 0
r
2
σtr,r + (σtr ) + ρq t = 0 (if u t 6= 0)
r
σr t = σtr (if u t 6= 0)

85
4. M ECHANICAL EQUILIBRIUM

4.5 Mechanical work and internal energy

Let’s consider a volume V of material that is enclosed by a closed surface S. The considered
body is deformed under the influence of external surface forces and body forces and it is now
in an equilibrium state. The total work performed by the external forces is denoted We , while
the work performed by the internal forces is written Wi . The change of the internal work or
internal energy due to a change in deformations is denoted δWi . Evidently, such a change is
due to a change of the external forces, with a variation in external work that equals δWe . We
limit ourselves here to adiabatic processes, and hence according to the law of conservation
of energy it must hold that
δWe = δWi (4.26)

The internal work or energy can be computed by taking the integral over the volume V of the
so-called energy density Wi , which is the energy per unit of volume. One can write
 
Ñ
δWi = δ  Wi dV  = δWe (4.27)
V

The change in external work is equal to the contribution of the body forces and a contribu-
tion of the surface forces
Ñ Ï
δWe = ρ~
q ·δ~
u dV + ~·δ~
p u dS (4.28)
V S

where δ~u is the infinitesimal change in the displacement vector of a material point. The
contribution of the surface forces may be written as
Ï Ï
~·δ~
p u dS = (σ·~
n )·δ~
u dS (4.29)
S S
Ï
= (~
n ·σ)·δ~
u dS (4.30)
S
Ï
= ~ u) d S
n ·(σ·δ~ (4.31)
S

Making use of the divergence theorem for a vector ~

a = σ·δ~
u
Ñ Ï
~ a dV =
∇·~ ~
n ·~
a dS (4.32)
V S

where ~
n is the normal on the surface element d S, permits to rewrite the contribution of the
surface forces as
Ï Ñ
~
p ·δ~
u dS = ~ u ) dV
∇·(σ·δ~ (4.33)
S V
Ñ
£ ¤
= (~ u + σ : (~
∇·σ)·δ~ u )T dV
∇δ~ (4.34)
V

86
 4.5. Mechanical work and internal energy

Combining this term with the contribution of the body forces,

Ñ Ñ
£ ¤
δWe = (ρ~ ~
q + ∇·σ)·δ~
u dV + σ : (~ u )T dV
∇δ~ (4.35)
V V

Substituting the equilibrium equation (4.18) makes the first integral vanish. Furthermore,
the symmetry of σ permits to write
£ ¤
σ : (~ u )T = σ : (~
∇δ~ u) = σ :
∇δ~ 1
2
~ u + (~
∇δ~ u )T = σ : δε
∇δ~ (4.36)

Finally the external work simplifies to the internal work by

Ñ
δWe = σ : δε dV = δWi (4.37)
V

The variation of the internal energy density δWi is thus simply equal to

δWi = σ : δε (4.38)

In rate format, this equation is written as

Ẇi = σ : ε̇ (4.39)

87
4. M ECHANICAL EQUILIBRIUM

4.6 Summary

Equilibrium

Rotational (Moment) σ = σT

Translational (Force) ~ q =~0

∇·σ + ρ~

External/internal work

Ð
Infinitesimal change δWe = σ : δε dV = δWi
V

Internal energy density rate Ẇi = σ : ε̇

88
 CHAPTER
5
T HREE - DIMENSIONAL LINEAR ELASTICITY

This chapter establishes the relation between the kinematics and the forces or stresses acting
on linear elastic solids. Attention is given to the three-dimensional constitutive relationship
for isotropic and several anisotropic materials. Various elastic constants are introduced and
their mutual relationships are established. The special case of isotropy is analyzed in its
various aspects, and simplifications towards plane strain and plane stress are treated. Finally,
an example for an anisotropic fibre-reinforced composite material is given.

5.1 General aspects of constitutive equations

The equations of equilibrium are not sufficient to determine the stress and displacement
fields in a material. In the more general case, one could add other equations of physics, i.e.,
conservation of mass and energy, but even then physical laws for the constituent material
are missing. In this course, we will only focus on the constitutive behaviour that establishes
the relation between stresses and kinematics.

5.2 The concept and nature of linear elasticity

One of the essential requirements for the application of the theory of linear elasticity, is the
restriction to small displacements, as shown in the previous chapters. If this condition is
satisfied, almost each material presents a linear elastic behaviour up to a certain deforma-
tion. This is due to the nature of linear elasticity. It finds its roots in the very small de-
formation of the existing microstructure, which does not alter its structure during defor-
mation. Hence, the geometry of crystals (crystalline materials, e.g., metals), of molecular
networks (polymers), of granular structures (ceramics), fibre-matrix arrangements (com-
posites) is not affected in linear elasticity. It is this essential property that distinguishes
elastic behaviour from the nonlinear dissipative deformation mechanisms, during which
microstructural changes will always take place. This also means that any process which
changes the microstructure by another treatment (e.g., heat treatment) will also affect the
elastic properties of the material.
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

The elastic properties of a crystalline material are thus attributed to its sensitivity to
small changes in inter-atomic distances due to the inter-atomic binding energy. The bind-
ing forces which strongly vary with distance may be of a different nature, e.g., ionic (electro-
static), covalent (sharing electrons), metallic (metal ion interaction with free-electron gas),
van der Waals (intermolecular attraction), hydrogen (dipole moments).

5.2.1 The elastic response of materials

In the case of perfect elasticity, the material fully recovers its initial structure. In the case of
infinitesimal displacements, it has been shown that the strain tensor ε suffices to character-
ize the deformation of the material, while the stress tensor σ gives a representation of the
forces that should act on an infinitesimal volume of material. In the most general case, a
stress tensor σ(~ x ) in a material point may depend on the strain tensors ε(~ ξ|~
ξ ∈ V (~
x )) of all
surrounding points. In this case, the constitutive relation F is said to be nonlocal. In most
cases, one assumes that the macroscopic scale is large enough to postulate the principle of
local action, according to which the stress tensor σ(~ x ) is only related to the local strain tensor
x ). In the case of elasticity, the load does not depend on the loading history but only on the
ε(~
actual deformed state of the material. Hence, there is no need to include the history with in-
ternal history variables that are related to the changes in the underlying microstructure. This
considerably simplifies the resulting constitutive law. In the case of small displacements, one
may write that elastic behaviour is described by some function F according to

x ) = F (ε(~
σ(~ x )) (5.1)

In the specific case of linear elasticity, which is of interest in this course, the function
F will be linear as well. This is a first-order approximation that applies to any continuous
function, since each continuous function can be approximated by a linear relationship in a
sufficiently small range of its variables.

5.2.2 The elastic stored energy function

Since elastic material is characterized by the reversibility of deformations and by the fact
that the underlying microstructure remains structurally unchanged, a function exists which
measures the amount of energy that is stored in the material. Assuming a local constitutive
behaviour σ = F (ε) and no history dependence, permits to integrate equation (4.39) in time,
which yields the stored energy density
Z Z
W = Ẇi d t = σ : ε̇ d t (5.2)
t t
1
= 2
σ :ε (5.3)

This functions quantifies the amount of energy per unit of volume that is stored in a material
point.
It can be noticed that the stress tensor σ is related to the stored energy density by

∂W
σ= (5.4)
∂ε
This property is inherent to the reversibility of the elastic deformations. In terms of the indi-
vidual components of the stress and strain tensor with respect to a Cartesian basis {~e x ,~
e y ,~
e z },

90
 5.3. General 3D linear elastic solid

one may write from equation (5.4)

δW = σ : δε
= σxx δεxx + σ y y δε y y + σzz δεzz
σx y δε yx + σxz δεzx + σ yz δεz y
σ yx δεx y + σzx δεxz + σz y δε yz
= σxx δεxx + σ y y δε y y + σzz δεzz
τx y δγx y + τxz δγxz + τ yz δγ yz (5.5)

and hence

∂W ∂W ∂W
σxx = σy y = σzz = (5.6)
∂εxx ∂ε y y ∂εzz
∂W ∂W ∂W
τx y = τxz = τ yz = (5.7)
∂γx y ∂γxz ∂γ yz

Note that the engineering shear strain appear naturally in this relation.

5.3 General 3D linear elastic solid

Uniaxial linear elastic behaviour has already been treated in previous courses, for which it is
known that the uniaxial stress σ is related to the uniaxial strain ε by means of only one elastic
constant, i.e., Young’s modulus E
σ = Eε (5.8)

This is the well-known Hooke’s law (Robert Hooke, 17th century). The modulus E inherits
its name from Thomas Young (early 19th century). This linear relation between stress and
strain will next be generalized to three dimensions.

5.3.1 Generalized Hooke’s law for anisotropic solids

As shown in the introductory chapter on tensors, the linear mapping of a second-order ten-
sor (i.c. the strain tensor ε onto another second-order tensor (i.c. the stress tensor σ) is per-
4
formed with a fourth-order tensor, which is here denoted the elasticity tensor C . The three-
dimensional equivalent of (5.8) thus reads

4
σ = C :ε (5.9)

In general, a fourth-order tensor has 3 × 3 × 3 × 3 = 81 components. The resulting tensor (i.c.

4
σ) must be symmetric and hence C must be left symmetric. Furthermore, since both the
strain tensor ε and the stress tensor σ are symmetric tensors, 36 distinct components are yet
sufficient to uniquely determine equation (5.9). This conclusion is easily made by reconsid-
ering the matrix condensation of the tensor equation (5.9). This example was already treated
in equation (1.210) of the chapter on tensors. If we replace in this equation the symmetric
4 4
tensor C by the stress tensor σ, the fourth-order tensor A by the elasticity tensor C and the

91
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

symmetric tensor B by the strain tensor ε, one finds

σ
~ = C~ε (5.10)
 
 σ11  C 1111 C 1122 C 1133 C 1112 +C 1121 C 1113 +C 1131 C 1123 +C 1132  ε 
11
σ22 C 2211 C 2222 C 2233 C 2212 +C 2221 C 2213 +C 2231 C 2223 +C 2232 ε22
 
 στ 33  =  C 3311 C 3322 C 3333 C 3312 +C 3321 C 3313 +C 3331 C 3323 +C 3332   ε33 
(5.11)
12  C 1211 C 1222 C 1233 C 1212 +C 1221 C 1213 +C 1231 C 1223 +C 1232  ε12
τ
13 C 1311 C 1322 C 1333 C 1312 +C 1321 C 1313 +C 1331 C 1323 +C 1332 ε13
τ23 ε23
C 2311 C 2322 C 2333 C 2312 +C 2321 C 2313 +C 2331 C 2323 +C 2332

This equation clearly shows that since only 36 numbers are involved, and one can thus al-
4
ways choose the tensor C right symmetric as well, i.e., one chooses C i j 12 = C i j 21 . This is
what is generally done, and hence the elasticity tensor is left and right symmetric
4 4 LT
C = C =⇒ C i j kl = C j i kl (5.12)
4 4 RT
C = C =⇒ C i j kl = C i j l k (5.13)

The number of components has to be reduced further to correctly account for the path
independence of elastic deformations (only the current deformation state determines the
stored energy and not the path that was followed towards that state). For a linear elastic
material, equation (5.3) simplifies to

W = 21 σ : ε (5.14)

Path independence implies that a stored energy density function exists, and the energy in
the current deformed state is obtained by substitution of (5.9) in (5.14) leading to
4
W = 21 ( C : ε) : ε
4
= 21 ε : C : ε (5.15)

which is a symmetric quadratic form in terms of the strain tensor ε. The energy density is a
scalar, which means that taking the transpose gives exactly the same result. Performing this
operation and making use of the symmetry of the strain tensor yields
h 4 iT
W = 21 ε : C : ε (5.16)
4 T
= 21 ε : C : ε (5.17)
4
Hence, the elasticity tensor C must be middle symmetric as well, which limits the total num-
ber of independent elastic constants for a generally anisotropic material to 21. Note that the
individual elasticity constants can be obtained by the partial derivatives of the stored energy
density, e.g., it holds that
∂W
= σ11 = C 1111 ε11 +C 1122 ε22 +C 1133 ε33 +
∂ε11
2C 1112 ε12 + 2C 1113 ε13 + 2C 1123 ε23

and hence
∂2 W
C 1111 = (5.18)
∂ε211
∂2 W ∂2 W
C 1122 = = = C 2211 (5.19)
∂ε11 ∂ε22 ∂ε22 ∂ε11

92
 5.4. Material symmetry and anisotropy

which again illustrates the middle symmetry. In tensor format, the latter set of equations is
generalized as
4 ∂2 W
C = (5.20)
∂ε∂ε
The generalization of 1D to 3D thus involves 21 material constants instead of 1 single
value. The matrix condensation given in equation (5.11) can be rewritten as

σ ε
~ = C~ (5.21)
    
σ11 C 1111 C 1122 C 1133 2C 1112 2C 1113 2C 1123 ε11
σ  C C 2222 C 2233 2C 2212 2C 2213  ε22 
2C 2223   
 22   1122
    
σ33  C 1133 C 2233 C 3333 2C 3312 2C 3313 2C 3323  ε33 
 =   (5.22)
 τ12  C 1112 C 2212 C 3312 2C 1212 2C 1213 2C 1223  ε12 
    
 τ13  C 1113 C 2213 C 3313 2C 1213 2C 1313 2C 1323  ε13 
τ23 C 1123 C 2223 C 3323 2C 1223 2C 1323 2C 2323 ε23

or by using the engineering shear strains instead

σ ε
~ = C~
    
σ11 C 1111 C 1122 C 1133 C 1112 C 1113 C 1123 ε11
σ  C C 2222 C 2233 C 2212 C 2213 C 2223   ε22 
 
 22   1122 
    
σ33  C 1133 C 2233 C 3333 C 3312 C 3313 C 3323   ε33 
 =   (5.23)
 τ12  C 1112 C 2212 C 3312 C 1212 C 1213 C 1223  γ12 
    
 τ13  C 1113 C 2213 C 3313 C 1213 C 1313 C 1323  γ13 
τ23 C 1123 C 2223 C 3323 C 1223 C 1323 C 2323 γ23

in which the matrix C is the general anisotropic Hookean stiffness or elasticity matrix. The
4 4
inverse of the elasticity tensor C is called the compliance tensor S. It satisfies the relations
4
ε = S :σ (5.24)
4
W = 21 σ : S : σ (5.25)

and
4 4 LT 4 RT 4 T
S= S = S = S (5.26)

Like the stiffness tensor, the compliance tensor counts 21 independent components only.

5.4 Material symmetry and anisotropy

The general constitutive equation for linear elasticity, will now be simplified systematically
in order to account for various types of symmetry in the material. This symmetry always
originates from the microstructure of the material. Fortunately, most engineering materials
do present symmetry properties, which allows to reduce the independent elastic constants
considerably. It is then convenient to formulate the components of the elasticity tensor with
respect to a basis that is attached to the symmetry axes or planes in the material. Note that
we systematically neglect the dependence of the material parameters on other state vari-
ables that influence the microstructure, e.g., the temperature.

93
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

5.4.1 Orthotropic materials

Orthotropic materials present two orthogonal planes of symmetry (e.g., orthorhombic crys-
tals, Fe 3C cementite). If the normals to these two planes are taken as the base vectors ~ e 2 and
~
e 3 , we will first consider the symmetry plane with normal ~ e 3.
⋆
Reverting the ~ e 3 base vector to ~e 3 therefore does not alter the constitutive relation. If this
vector is reverted, then clearly certain shear strains (i.e., the change of angles) change sign,
e.g., the angle changes between ~ e 1 and ~e 3 , as well as ~
e 2 and ~
e3
γ⋆
13 = −γ13
γ⋆
23 = −γ23

Likewise, the shear stresses in these directions change sign

τ⋆
13 = −τ13
τ⋆
23 = −τ23

while all other stress components remain unchanged. Since the elasticity tensor must be
identical in both cases, it must hold that
σ11 = σ⋆
11
C 1111 ε11 +C 1122 ε22 +C 1133 ε33 +C 1112 γ12 +C 1113 γ13 +C 1123 γ23 =
C 1111 ε11 +C 1122 ε22 +C 1133 ε33 +C 1112 γ12 −C 1113 γ13 −C 1123 γ23
for each strain state ε, which is only possible if C 1113 = 0 and C 1123 = 0. Repeating this for all
components yields
C 1113 = C 1123 = C 2213 = C 2223 = 0
(5.27)
C 3313 = C 3323 = C 1213 = C 1223 = 0
We can repeat this reasoning for the second symmetry plane, with ~ e 2 as the normal vec-
tor. For a reflection with respect to this second symmetry plane, elastic coefficients may not
be affected neither, i.e., one has to add the following conditions
γ⋆
12 = −γ12
γ⋆
23 = −γ23
τ⋆
12 = −τ12
τ⋆
23 = −τ23

which is only possible if additional elastic constants vanish, i.e.,

C 1112 = C 2212 = C 3312 = C 1323 = 0 (5.28)
Orthotropic elasticity is thus characterized by 9 independent elastic constants. The corre-
sponding matrix format attached to the anisotropic base vectors reads
σ ε
~ = C~
    
σ11 C 1111 C 1122 C 1133 0 0 0 ε11
σ  C C C 0 0 0  ε 
 22   1122 2222 2233   22 
    
σ33  C 1133 C 2233 C 3333 0 0 0   ε33 
 =   (5.29)
 τ12   0 0 0 C 1212 0 0  γ12 
    
 τ13   0 0 0 0 C 1313 0  γ13 
τ23 0 0 0 0 0 C 2323 γ23

94
 5.4. Material symmetry and anisotropy

Note that the symmetry with respect to the planes normal to~ e 2 and ~
e 3 automatically involves
the symmetry with respect to the plane normal to ~ e 1 as well. This is typical for orthotropic
symmetry, e.g., for orthorhombic crystal structures. The stiffness matrix C in equation (5.29)
can also be written in a different format, which will be used further in the course:
 
A Q R 0 0 0
Q B S 0 0 0 
 
 
R S C 0 0 0 
C =  (5.30)
0 0 0 K 0 0 
 
0 0 0 0 L 0 
0 0 0 0 0 M

A more detailed treatment of orthotropic elasticity is given in Appendix A.

5.4.2 Transverse isotropy

While the preceding cases of symmetry are typically directionally symmetric materials, trans-
verse isotropic materials present a rotational symmetry around a specific axis. Assume that
this symmetry axis is oriented along the base vector ~ e 3 . This means that there is no material
distinction between the vector ~ e 1 and ~
e 2 , which can be oriented in any direction perpen-
dicular to the axis of rotational symmetry. The elastic constants involving one of these two
directions thus should be strictly identical, i.e.,

C 1111 = C 2222 = A (5.31)

C 1122 = Q (5.32)
C 1133 = C 2233 = R (5.33)
C 3333 = C (5.34)
C 1111 −C 1122
C 1212 = = 12 (A −Q) = K (5.35)
2
C 1313 = C 2323 = L (5.36)

while angle changes between any direction with ~

e 1 and ~
e 2 may produce according shear
stresses only

C 1112 = C 1113 = C 1123 = C 2212 = C 2213 = 0

(5.37)
C 2223 = C 3312 = C 3313 = C 3323 = C 1213 = C 1323 = 0

The full proof is obtained by rotating the axes around the symmetry axis ~ e 3 , which may not
alter the components of the elasticity tensor (the elasticity coefficients remain invariant un-
der a rotation). The constitutive behaviour for a transverse isotropic materials is thus fully
determined from 5 constants. This is typically the case for hexagonal crystals (HCP, e.g., Z n,
M g , T i ) or honeycomb composite structures. The constitutive relation in matrix form then
reads     
σ11 A Q R 0 0 0 ε11
σ  Q A R 0 0 0   ε 
 22     22 
    
σ33   R R C 0 0 0   ε33 
 =   (5.38)
 τ12   0 0 0 K 0 0  γ12 
    
 τ13   0 0 0 0 L 0  γ13 
τ23 0 0 0 0 0 L γ23

95
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

Note that these 5 elastic constants are often related to engineering elasticity coefficients, like
Young’s moduli E i , Poisson’s ratios νi j and shear moduli G i j . The following identities are
then used
E 1 (1 − ν31 ν13 )
A = (5.39)
(1 + ν12 )(1 − ν12 − 2ν31 ν13 )
E 1 (ν12 + ν31 ν13 )
Q = (5.40)
(1 + ν12 )(1 − ν12 − 2ν31 ν13 )
E 1 ν31
R = (5.41)
(1 − ν12 − 2ν31 ν13 )
E 3 (1 − ν12 )
C = (5.42)
(1 − ν12 − 2ν31 ν13 )
E1
L = G 13 = G 23 = (5.43)
2(1 + ν12 )

where
E1
ν13 = ν31 (5.44)
E3
More comments on the physical meaning of Young’s moduli and Poisson’s ratios are given
in the next section. A detailed treatment of transversely isotropic elasticity is given in Ap-
pendix A.

5.5 Linear elastic isotropic materials

If a material has a microstructure which is sufficiently randomly oriented and distributed
on a very small scale, the macroscopic material properties will be identical in all directions.
This is typically the case for metals with a randomly oriented polycrystalline structure, ran-
dom granular structures in ceramics, random fibre reinforced composites. Such a material is
called isotropic. This means that the orientation of the material in a material point does not
influence its mechanical behaviour. The properties of the material are said to be invariant
under a rotation of the material axes. This means that the constitutive relation becomes fully
independent of the basis, since any particular choice for the orientation of the base vectors
must give the same elasticity tensor.
In the symmetry cases presented above, the material properties depend on the consid-
ered direction. That is a typical characteristic of an anisotropic material. If the material
properties are constant from point to point in the material, the material is said to be ho-
mogeneous. A continuum may be anisotropic and non-homogeneous (materials with large
anisotropic grains with different orientations), anisotropic and homogeneous (single crys-
talline materials, directionally fibre reinforced composites) or isotropic and homogeneous
(randomly oriented microstructures from polycrystals). Clearly, the notion isotropy and ho-
mogeneity depend on the scale of observation and application of the material. Metals are
often macroscopically homogeneous and isotropic, but on a much smaller scale of the in-
dividual grains they are highly anisotropic and non-homogeneous. In the remainder of this
course, we will limit ourselves to the macroscopic scale.
Since isotropic materials are by definition independent of the orientation of the material
(the undeformed reference configuration), the stored energy must be a function that does
not depend on the orientation of the material. In equation (5.15), a quadratic form in terms
of the strain tensor ε was found. In order to have a stored energy that is independent of the

96
 5.5. Linear elastic isotropic materials

orientation (the choice of the basis), expression (5.15) must simplify into an equation that
only depends on the invariants of the strain tensor ε, since these are the only scalar tensor
functions that are independent of the basis. Hence, it will hold that

W = f (J 1 (ε), J 2 (ε), J 3 (ε)) (5.45)

5.5.1 The isotropic elasticity tensor

An isotropic tensor is a tensor of which the components do not change by orthogonal coor-
dinate transformations (rotations). Evidently all scalars are isotropic tensors (of rank zero).
Vectors, which are tensors of order 1, can never be isotropic, since they will always have dif-
ferent components if the base vectors are rotated. Isotropic second-order tensors are nec-
essarily constructed on the second-order unit tensor I . This tensor always takes the same
components, irrespective of the orientation of the base vectors. A general isotropic second-
order tensor is obtained through the multiplication of I with a scalar, i.e.,

kI or k δi j (5.46)

Dyadic products of the second-order unit tensor with itself, i.e., I I , yields a fourth-order
isotropic and total symmetric tensor. Apart from this one, only the symmetric fourth-order
4 s
unit tensor I is isotropic and totally symmetric. Hence, the isotropic fourth-order elasticity
tensor is necessarily a linear combination of these two isotropic fourth-order total symmet-
ric tensors, i.e.,
4 4 s
C = λ I I + 2µ I (5.47)

or in index notation
C i j kl = λ δi j δkl + µ(δi l δ j k + δi k δ j l ) (5.48)

This is the general format of the isotropic elasticity tensor, in which the elastic constants λ
and µ are called Lamé’s constants (from Gabriel Lamé, middle 19th century). Lamé’s second
constant µ equals the more frequently used shear modulus G = µ.
The constitutive relation that ensues from the linear mapping of ε on σ simplifies to

4
σ = C :ε (5.49)
4 s
= (λ I I + 2G I ) : ε (5.50)
4 s
= λ I (I : ε) + 2G ( I : ε) (5.51)
= λ tr(ε)I + 2G ε (5.52)

or in index format
σi j = λεkk δi j + 2G εi j (5.53)

This constitutive equation is often rewritten to clearly separate the pressure-dilatation re-

97
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

sponse from the deviatoric response (i.e., relating σh to εh and likewise σd to εd )

σ = λ I tr(ε) + 2G ε (5.54)
³ ´
= λtr(ε)I + 2G εd + 13 tr(ε)I
= (λ + 32 G)tr(ε)I + 2G εd
= (3 λ + 2G) 13 tr(ε)I + 2G εd
3 λ + 2G
=3 εv + 2G εd
3
= 3 K εv + 2G εd (5.55)
where the elastic constant K is called the bulk modulus or compressibility modulus. The
bulk modulus, given by
3 λ + 2G
K = (5.56)
3
relates the hydrostatic pressure p to the volumetric strain e by e = tr(ε), i.e.,
p = −K e (5.57)
1
3
tr(σ) = K tr(ε)
σm = 3 K εm (5.58)

In fact, the latter equation clearly illustrated that the volumetric parts (σh = 13 tr(σ), εv =
1 d h d
3 tr(ε)) and deviatoric parts (σ = σ − σ , ε = ε − εv ) of the stress and strain tensors are
constitutively related in a separate way

σh = 3 K εv (5.59)
σd = 2G εd (5.60)
In matrix representation, and making use of the engineering shear strains instead of the
tensorial shear strains, the constitutive equation reads
σ ε
~ = C~ (5.61)
    
σ11 λ + 2G λ λ 0 0 0 ε11
σ   λ λ + 2G λ 0 0 0   ε22 
 
 22   
    
σ
 33   λ λ λ + 2G 0 0 0 ε
  33 
 =   (5.62)
 τ12   0 0 0 G 0 0  γ12 
    
 τ13   0 0 0 0 G 0  γ13 
τ23 0 0 0 0 0 G γ23
From a mathematical point of view, Lamé’s constant are perfectly suited to characterize the
linear elastic material behaviour. In practice however, engineering moduli which can be
related to classical tensile tests, are frequently used instead.
Note that the volumetric-deviatoric split of the constitutive relation, permits to find very
easily the inverse relation, i.e., the so-called compliance relation. From the equations (5.59)
and (5.60), one immediately notices
1 h
εv = σ (5.63)
3K
1 d
εd = σ (5.64)
2G

98
 5.5. Linear elastic isotropic materials

4 4
The compliance tensor S is like the elasticity tensor C an isotropic fourth-order tensor,
which thus also takes the format
4 4 s
S = αII +β I (5.65)

Using equations (5.63) and (5.64), this tensor is easily reconstructed

ε = ε v + εd (5.66)
1 h 1 d
= σ + σ (5.67)
3K 2G
1 1 d
= tr(σ)I + σ (5.68)
·9 K 2G ¸
1 1 4 s 1
= II + ( I − 3II) : σ (5.69)
9K 2G
· ¸
1 1 1 4 s
= ( − )II + I :σ (5.70)
9 K 6G 2G
4
= S :σ (5.71)

The compliance tensor is thus given in terms of K and G or in terms of Lamé’s constants λ
and µ = G by
µ ¶
4 1 1 1 4 s
S = − II + I (5.72)
9 K 6G 2G
µ ¶
λ 1 4 s
= − II + I (5.73)
2G(3λ + 2G) 2G

5.5.2 Engineering elastic moduli

The introduction of the classical engineering elasticity entirely relies on special states of
stress which are attained with simple mechanical tests. The first test is simple tension.

Simple tension

In a simple tensile test, a force is applied into one direction of a specimen only, while all
other stress components are equal to zero, e.g.,

σ11 = σ (5.74)
σ22 = σ33 = τ12 = τ13 = τ23 = 0 (5.75)

The first three constitutive equations from (5.62) then simplify to

σ11 = (λ + 2G)ε11 + λε22 + λε33 (5.76)

0 = λε11 + (λ + 2G)ε22 + λε33 (5.77)
0 = λε11 + λε22 + (λ + 2G)ε33 (5.78)

Solving ε22 and ε33 from the equations (5.77) and (5.78) in terms of ε11 , followed by a substi-
tution in (5.76) gives the relation between σ11 and ε11

(2G + 3λ)G
σ11 = ε11 = E ε11 (5.79)
λ +G

99
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

from which Young’s modulus is defined as the linear coefficient between the applied axial
stress and the axial strain in the tensile bar. The strain components obtained in terms of the
externally applied stress σ read
λ +G
ε11 = σ (5.80)
(2G + 3λ)G
λσ
ε22 = ε33 = − (5.81)
2G(2G + 3λ)
The fractional contraction relates the contraction of the tensile bar perpendicular to the
loading axis with the axial tensile strain, i.e., the relative lateral contraction.
ε22 ε33 λ
− =− = =ν (5.82)
ε11 ε11 2(λ +G)
which is quantified by Poisson’s ratio ν.
Two coefficients fully characterize the isotropic constitutive law in linear elasticity. This
means that either {E , ν} or {λ,G} can be chosen as the set of elastic constants. Recomputing
λ and G in terms of the Young’s modulus and Poisson’s ratio gives
Eν E
λ= G = (5.83)
(1 + ν)(1 − 2ν) 2(1 + ν)

Simple shear

The physical significance of the shear modulus becomes most apparent in a simple shear
test. In that case, one shear stress component is constant and equal to τ, while all other
stress components equal zero, e.g.,

τ12 = τ21 = τ (5.84)

σ11 = σ22 = σ33 = τ13 = τ23 = 0 (5.85)

The matrix form (5.62) is then simplified to

λtr(ε) + 2Gε11 = 0
λtr(ε) + 2Gε22 = 0
λtr(ε) + 2Gε33 = 0
γ13 = γ23 = 0
τ = Gγx y (5.86)

which leads to the following solution for the strain components

ε11 = ε22 = ε33 = γ13 = γ23 = 0 (5.87)

τ
γ12 = (5.88)
G
The shear modulus thus gives the proper linear relation between the shear strain and shear
stress in the case of simple shear.
Note that Poisson’s ratio is limited by −1 < ν < 0.5, which is easily concluded from (5.83),
in order to keep the value of λ finite. Materials with a negative Poisson’s ratio are difficult to
imagine, but advanced microstructures nowadays permit to construct such materials. The
upper limit for ν = 0.5 is typical for incompressible behaviour, since then tr(ε) = 0. A com-
plete conversion table for all elastic moduli is given in table 5.1.

100
 E, ν λ, G K, G E, G E, K

↓ ↓ ↓ ↓ ↓

E → E =E E = (2G+3λ)G
λ+G
9K G
E = 3K +G
E =E E =E
λ 3K −2G
ν→ ν=ν ν = 2(λ+G) ν = 2(3K +G) ν = E −2G
2G ν = 3K6K−E
E 3K E
G → G = 2(1+ν) G =G G =G G =G G = 9K −E

E
K → K = 3(1−2ν) K = 3λ+2G
3 K =K EG
K = 3(3G−E ) K =K
Eν
λ → λ = (1+ν)(1−2ν) λ=λ λ = 3K −2G
3 λ = G(E −2G)
3G−E λ = 3K9K
(3K −E )
−E

E, λ G, ν λ, ν λ, K K, ν

↓ ↓ ↓ ↓ ↓

5.5. Linear elastic isotropic materials

E → E =E E = 2G(1 + ν) E = λ(1+ν)(1−2ν)
ν E = 9K3K(K−λ
−λ)
E = 3K (1 − 2ν)
p
−E −λ+ (E +λ)2 +8λ2
ν→ ν= 4λ
ν=ν ν=ν ν = 3Kλ−λ ν=ν
p
−3λ+E + (3λ−E )2 +8λE
G → G= 4 G =G G = λ(1−2ν)
2ν G = 3(K2−λ) G = 3K2(1+ν)
(1−2ν)
p
E −3λ+ (E −3λ)2 −12λE
K → K= 6
K = 2G(1+ν)
3(1−2ν)
K = λ(1+ν)
3ν
K =K K =K
2Gν ν
λ→ λ=λ λ = 1−2ν λ=λ λ=λ λ = 3K
1+ν

Table 5.1: Conversion tables of elastic moduli

101
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

Constitutive relation in terms of the engineering moduli

The constitutive equation (5.52) is easily transformed with the equation (5.83) to

Eν E
σ= tr(ε)I + ε (5.89)
(1 + ν)(1 − 2ν) 1+ν

which in a vector basis is written in the following matrix format

    
σ11 1−ν ν ν 0 0 0 ε11
σ   ν 1 − ν ν 0 0 0 ε 
 22     22 
  E   
σ33   ν ν 1−ν 0 0 0  ε33 
 =  1−2ν   (5.90)
 τ12  (1 + ν)(1 − 2ν)  0 0 0 2 0 0 γ12 
   1−2ν  
 τ13   0 0 0 0 2 0 γ13 
1−2ν
τ23 0 0 0 0 0 2
γ23

The compliance tensor (5.73) can be rewritten in terms of the engineering moduli by
making use equation (5.83), which leads to

4 ν 1+ν 4 s
S =− II + I (5.91)
E E

The associated stress-strain constitutive relation in matrix format (with the engineering shear
strains!), with the compliances written in terms of the elastic constants given above, then
reads

σ Sε
~ =~ (5.92)

1
− Eν − Eν 0 0 0
   E  
ε11 − ν 1
− Eν 0 0 0  σ11
ε   E E 
 22    σ22 
  
 ε33   − Eν − Eν 1
E
0 0 0 
 σ33 
 = 
2(1+ν)

  τ12  (5.93)
γ12   0 0 0 0 0  
   E  τ 
γ13   2(1+ν)  13 
 0 0 0 0 0 
γ23  E  τ23
2(1+ν)
0 0 0 0 0 E

In index notation (with the tensorial shear strains!) these relations would read

1£ ¤
εi j = −ν σkk δi j + (1 + ν) σi j (5.94)
E

or in tensor format
1
ε= [−ν tr(σ) I + (1 + ν) σ] (5.95)
E
Note that the bulk modulus K is formulated in terms of the engineering moduli by

E
K = (5.96)
3(1 − 2ν)

102
 5.5. Linear elastic isotropic materials

5.5.3 Stored energy density for isotropic linear elastic materials

At the start of this section, it was remarked that the stored energy density for isotropic ma-
terials can only depend on the invariants of the strain tensor ε. If not, material orientation
plays a role, which is by definition of isotropy excluded. For the elasticity tensor elaborated
in the previous sections, one finds
4
W = 21 ε : C : ε
³ 4 s
´
= 21 ε : λ I I + 2G I : ε
= 21 λtr(ε)2 +G ε : ε
= ( 12 λ +G)[J 1 (ε)]2 − 2G J 2 (ε) (5.97)
Clearly, W only depends on invariants of ε, which is characteristic for an isotropic material.

5.5.4 Coincidence of principal axes of stress and strain

An important consequence of the use of an isotropic elasticity tensor, is that it does not
change the principal directions of the second-order tensor to which it is applied. Hence,
principal directions of strain will be preserved as principal directions of stress. This becomes
clear if the isotropic constitutive behaviour is substituted in the eigenvalue problem for the
stresses
~ i = ~0
[σ − σi I ] ·M
~ i = ~0
[λtr(ε)I + 2G ε − σi I ] ·M
· ¸
σi − λtr(ε)
2G ε − ( )I ·M ~ i = ~0 (5.98)
2G
which resembles exactly the eigenvalue problem for the strain tensor
~ i = ~0
[ε − ǫi I ] ·N (5.99)
~ i are simply equal to the principal directions of
Hence the principal directions of stress M
~i
strain N
~i = N
M ~i (5.100)
The principal stresses are evidently different from the principal strains. Their mutual relation
is given by
σi − λtr(ε)
ǫi = (5.101)
2G
These relations also permit to establish the relations that relate the invariants of stress to the
invariants of strain, i.e.,
J 1 (σ) = (3λ + 2G)J 1 (ε) (5.102)
J 2 (σ) = λ(3λ + 4G)J 1 (ε)2 + 4G 2 J 2 (ε) (5.103)
J 3 (σ) = λ2 (λ + 2G)J 1 (ε)3 + 4λG 2 J 1 (ε)J 2 (ε) + 8G 3 J 3 (ε) (5.104)
The equality of principal directions can be noticed directly, if one realizes that the shear
strains along principal directions are equal to zero, i.e., γ12 = γ13 = γ23 = 0. Making use of the
constitutive relation for isotropic linear elasticity then immediately gives τ12 = τ13 = τ23 = 0,
which means that the same directions remain principal for the stresses as well. These axes
are simply called the principal axes, without making any further distinction between stresses
and strains.

103
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

5.5.5 Plane strain and plane stress

This subsection presents a brief treatment of plane deformation states, limited to isotropic
elasticity. The more general case is given in Appendix A.

Plane strain

As shown in previous chapters, plane strain is characterized by

ε33 = γ13 = γ23 = 0 (5.105)

and the 3D problem reduces to a 2D strain problem for which the matrix format of the gov-
erning constitutive equation reads

σ ε
~ = C~
    
σ11 λ + 2G λ 0 ε11
σ22  =  λ λ + 2G 0   ε22  (5.106)
τ12 0 0 G γ12

The stress component σ33 for plane strain (in-plane loaded thick structures) does not equal
zero and is given by
σ33 = λ(ε11 + ε22 ) (5.107)
which can be computed easily. The stress-strain relation (5.106) can also be written in terms
of the engineering elasticity constants E and ν, which leads to

σ ε
~ = C~
    
σ11 1−ν ν 0 ε11
E  
σ22  =  ν 1−ν 0   ε22  (5.108)
(1 + ν)(1 − 2ν)
τ12 0 0 1−2ν γ12
2

Plane stress

In the case of plane stress (in-plane loaded thin plane structures), the stress components
with respect to the thickness directions vanish

σ33 = τ13 = τ23 = 0 (5.109)

which makes the stresses two-dimensional. The strain ε33 is now different from zero, and
the constitutive relations read

σ ε
~ = C~
    
σ11 λ + 2G λ λ 0 ε11
    
σ22   λ λ + 2G λ 0   ε22 
 =   (5.110)
 0   λ λ λ + 2G 0   ε33 
τ12 0 0 0 G γ12

The third equation for τ33 permits to express the strain component ε33 in terms of ε11 and
ε22
λ λ
ε33 = − ε11 − ε22 (5.111)
λ + 2G λ + 2G

104
 5.6. The superposition principle

Back substitution in (5.110) then yields

 
  λ2 λ 2
σ11 λ + 2G − λ+2G λ − λ+2G 0  ε11 
 
σ22  =  λ − λ2 λ 2
0 ε22 (5.112)
λ + 2G − λ+2G  
 λ+2G
τ12 0 0 G γ12

which is easily simplified to

 
  (λ+G)4G 2λG
σ11 0  ε11 
 λ+2G λ+2G 
σ22  =  2λG (λ+G)4G
0 ε22 (5.113)
 
 λ+2G λ+2G
τ12 0 0 G γ12

or by substitution of the engineering elasticity constants E and ν

    
σ11 1 ν 0 ε11
E ν 1
σ22  = 0   ε22  (5.114)
1 − ν2 1−ν
τ12 0 0 2
γ12

5.6 The superposition principle

Linear elastic material behaviour has two major characteristics which constitute the basis of
the superposition principle, i.e.,

• The path independence: only the final loading state determines the deformed config-
uration, stresses and strains in the material. It does not matter what loading path was
followed to reach that state.

• The linearity of the problem: the constitutive stress-strain relationship is entirely lin-
ear, and the strain-displacement relations (i.e., the linear strain tensor) are linear as
well.

This path independent and linear character permits to treat any loading case as an arbi-
trary subdivision of elementary loading cases, for which the solution can be summed up in
a trivial way. In other words, if the solution of a mechanical problem in linear elasto-statics
is known for two specific loading cases, then it is also known for any linear combination of
these two loading cases. This is called the superposition principle. It permits to superimpose
the results for specific loading arrangements, in order to investigate their combined effect.
Note however, that the principle is inherent to the theory of linear elasticity for small
displacements, and that it may by no means be generalized to the large deformation case.

105
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

5.7 Summary

Elastic stored energy function

Stored energy density W = 12 σ : ε

Stress tensor σ = ∂∂ε

W

∂W
σxx = ∂ε xx
∂W
σ y y = ∂ε
yy
∂W
σzz = ∂ε
zz
∂W
τx y = ∂γx y
∂W
τxz = ∂γ xz
∂W
τ yz = ∂γ
yz

3D linear elasticity

1D Hooke’s law σ=Eε

4
3D Hooke’s law σ= C :ε

4 4 LT
The elasticity tensor C = C
4 4 RT
C = C
4 4 T
C = C
C i j kl = C j i kl = C i j l k = C l k j i
4
C has 21 independent components

4
Stored energy density W = 12 ε : C : ε
 
 σ11  C 1111 C 1122 C 1133 C 1112 C 1113 C 1123  ε11 
σ22  CC 1122 C 2222 C 2233 C 2212 C 2213 C 2223
 ε22
C 3323   ε33 
Matrix format  στ 33  =  1133 C 2233 C 3333 C 3312 C 3313
12  C 1112 C 2212 C 3312 C 1212 C 1213 C 1223  γ12
τ13 C 1113 C 2213 C 3313 C 1213 C 1313 C 1323 γ13
τ23 γ23
C 1123 C 2223 C 3323 C 1223 C 1323 C 2323

4 4
Compliance tensor S ε= S :σ

106
 5.7. Summary

Material symmetry and anisotropy

Orthotropy Two (and hence three) symmetry planes

9 elastic constants
Transverse isotropy Rotational symmetry around one axis
5 elastic constants
Isotropy Omni-directional symmetry
2 elastic constants

Isotropic linear elasticity

4 4 s
Isotropic elasticity tensor C = λI I + 2µ I
C i j kl = λδi j δkl + µ(δi l δ j k + δi k δ j l )

Lamé’s constants λ and G = µ

Constitutive response σ = λtr(ε)I + 2G ε

Index notation σi j = λεkk δi j + 2G εi j

Uncoupled response σ = 3 K εv + 2G εd
σh = 3 K εv
σd = 2G εd
Bulk modulus K
   
σ11 ε11
σ  
0 0  ε22 
 
 22  λ+2G λ λ 0
σ33   λλ λ+2G λ 0 0 0 ε 
 
Matrix format  = 0 λ λ+2G 0 0 0   33 
 τ12  0 0 G 0 0 γ 

  0 0 0 0 G 0  12 
 τ13  0 0 0 0 0 G γ13 

τ23 γ23

4 λ 1 4 s
Compliance tensor S = (− 2G(3λ+2G) ) I I + 2G I

107
5. T HREE - DIMENSIONAL LINEAR ELASTICITY

Elastic moduli

Engineering moduli
Young’s modulus σ11 = E ε11 in simple tension
E = (2G+3λ)G
λ+G

Poisson’s ratio − εε22 = ν in simple tension

11
λ
ν = 2(λ+G)

Eν
Mutual relations λ = (1+ν)(1−2ν)

E
G = 2(1+ν)

K = 3λ+2G
3
E
= 3(1−2ν)

Limitations 0 < E < ∞ and −1 < ν < 0.5

Incompressibility ν = 1/2

Stored energy density W = ( 21 λ +G)[J 1 (ε)]2 − 2G J 2 (ε)

Principal axes, plane strain and plane stress for isotropic linear elasticity

Principal axes ~i = M
Coincide −→ N ~i

    
σ11 1−ν ν 0 ε11
E  
Plane strain (matrix) σ22  =
(1+ν)(1−2ν) 
ν 1−ν 0   ε22 
τ12 0 0 1−2ν γ12
2

    
σ11 1 ν 0 ε11
Plane stress (matrix) σ22  = E 2 ν 1 0   ε22 
1−ν
1−ν
τ12 0 0 2
γ12

Superposition principle

Superposition principle ~1 + p
q +p
σ(ρ~ ~2 ) = σ(ρ~
q ) + σ(~
p 1 ) + σ(~
p2)

108
 CHAPTER
6
L IMIT DESIGN CRITERIA
IN LINEAR ELASTICITY

This chapter is devoted to the design criteria which ensue from the limits of the elastic ma-
terial behaviour. The loss of elasticity is assumed if the material’s microstructure changes
in an irreversible way, leading to permanent deformations. A large class of criteria is pre-
sented, which are applicable to different types of materials. Particular attention is given to
the widely used von Mises criterion, where a clear connection with the physics of the de-
forming microstructure is established.

6.1 Failure modes

Proper design rules are based on the adequate knowledge of failure modes, as well the cri-
teria which play a role in their initiation. Materials are intended to remain within the linear
elastic domain for a vast majority of applications. This ensures that no permanent defor-
mations will occur, so that the designed component is capable to fulfill its tasks with the
initial geometry used for the design. Different types of failure may occur and hence limit
the applicable load to a structure or require an improved geometry by means of a coupled
CAD-mechanical design.

• plastic yielding; this failure mode is characterized by the appearance of irreversible

deformations in the material. For metals this is due to crystallographic slip. Hence,
the microstructure starts to change locally.

• brittle fracture; many brittle (or quasi-brittle) materials present a sudden crack ini-
tiation and crack growth without the appearance of prior permanent deformations.
This is mainly due to local stress concentrations which constitute an overload for the
microstructure, in which bonds are suddenly broken. The theory of elastic fracture
mechanics is devoted to this subject (see the corresponding optional course on this
subject).
6. L IMIT DESIGN CRITERIA

• progressive damage; microstructural damage may also evolve progressively through

the appearance of micro-cracks, which grow, accumulate, coalesce, etc.. This phe-
nomenon may take place without meaningful permanent deformations and it gener-
ally leads to a progressive reduction of the stiffness of the material (i.e., of the Young’s
modulus for the perfectly isotropic case). This failure mode is the subject of the course
on continuum damage mechanics.

• fatigue; fatigue damage is a highly complex phenomenon that depends on the cyclic
character of the external load, geometrical aspects of the specimen, external physical
and chemical influences and the surface conditions of the material. If fatigue damage
is a relevant design criterion for the application of interest (i.e., a cyclic loading exists),
one commonly resorts to the so-called fatigue limit, which is the maximum stress am-
plitude under which the fatigue phenomenon will not occur. For specific applications
(low cycle fatigue), design rules up to a dedicated number of cycles may be used as
well.

• dynamic failure; vibrations due to dynamic loadings may constitute a dangerous fail-
ure mode since resonance may occur if the spectrum of the external loading presents
frequencies that are close to the eigenfrequencies of the structure.

• thermal failure; temperature may have a large influence on the physical, geometrical
and hence mechanical properties of a material’s microstructure. Temperature changes
in combination with a non-constant (possibly cyclic) loading may lead to a significant
reduction of the strength of the material.

• elastic instabilities; this failure mode may occur prior to all other types of failure. It is
essentially a structure and geometry dependent failure mode, which depends on the
deformed geometry of the considered structural member or component. The simplest
and well-known example is column buckling.

6.2 Standard tests

6.2.1 Tension
A standard uniaxial tensile test, where the load is applied in one direction of a cylindrical
specimen only (with a corresponding stress σt in that direction), is characterized by a stress
state for which the principal stresses are given by

σ1 = σ t
σ2 = 0 (6.1)
σ3 = 0

The stress tensor is thus trivially given by σ = σt ~ e x~

e x , where ~
e x is the direction of the applied
uniaxial load. Note that the solution of the static equilibrium equation (4.25) now simply
points out that the axial stress must be constant along the axis of the tensile bar (since the
deformation of the cross-section is neglected in linear elasticity). Assuming that ~ e y and ~
ez
are two vectors perpendicular to ~ e x , one easily finds the hydrostatic and deviatoric part of σ

110
 6.2. Standard tests

as
σt
σh = 13 tr(σ)I = I (6.2)
3
σd = 32 σt ~ e x − 13 σt ~
e x~ e y − 31 σt ~
e y~ e z~
ez (6.3)

The tensile stress in tension, further denoted σt is used in the following paragraphs to
identify the onset of yielding, i.e., the loss of elastic behaviour.

6.2.2 Torsion
Torsion is a particular loading mode, which is often used in laboratory tests to characterize
materials. A member with a circular cross section is subjected to a twisting moment (torque)
around its axis of symmetry (see also course 4A130 on mechanics). Consider a cylindrical
basis, with the base vector ~ e z aligned with the axis of the shaft. In the case of a shaft with a
cylindrical cross-section, the solution is easy.
Under the influence of the applied torque the shaft will twist. The shaft has a length L and
a radius R. Assume that the shaft is clamped in z = 0. The applied loading and the geometry
of the shaft are homogeneous along the axis ~ e z , and hence the twist will be homogeneous
as well. If one denotes the rotation of the cross-section per unit length of the shaft by α,
one notices that α is constant along z. The circular cross-section remains circular, and axial
displacements (along ~ e z ) of material points are impossible (plane sections remain plane be-
cause of symmetry). The only deformation is a relative rotation of cross-sections, and hence
the corresponding displacement components (u, v, w) with respect to the cylindrical basis
{~
e r ,~
e t ,~
e z } are given by
u=0 v = αz r w =0 (6.4)

Using the strain-displacement equations (2.55), one easily finds the six corresponding rele-
vant strain components that automatically satisfy the compatibility equations

εr r = εt t = εzz = εr t = εr z = 0
µ ¶
1 ∂v 1 ∂w (6.5)
εt z = 2 + = 21 α r
∂z r ∂θ

Evidently, the three remaining components are obtained through symmetry. Hence, only
one of the six relevant shear strain component differs from zero and the entire strain tensor
is given by ε = 12 α r~ e z + 21 α r~
e t~ e z~
e t By means of the constitutive equation for isotropic linear
elasticity, the stress tensor is obtained

4
σ = C :ε
h 4 s
i
= K I I + 2G ( I − 13 I I ) : ε
e t~
= G α r~ e z~
e z +G α r~ et (6.6)

Note that this stress tensor satisfies static equilibrium (substitute the divergence operator (1.165)
with this stress tensor in (4.25)).
Boundary conditions at the boundaries z = 0 (u, v, w = 0) and r = R (σr r = 0) are satisfied.
The torque moment applied at the end of the bar (z = L) equals M t and must equal the

111
6. L IMIT DESIGN CRITERIA

resultant torque of all shear stresses acting on that boundary (the normal to that boundary
is ~
e z , the concerned component thus is τt z ), i.e.,
ZR Z2π
Mt = r τt z d θ d r (6.7)
0 0
ZR Z2π
= Gα r 2 d θ d r = G α Jt (6.8)
0 0
ZR
R3
= 2πG α r 2 d r = 2πG α (6.9)
0 3

where J t is the polar moment of inertia of the cross-section S, defined by

Ï
Jt = r 2 dS (6.10)
S

The shear stress in torsion, further denoted τθ = τt z is used in the following paragraphs
to identify the onset of yielding in the particular case of torsion, i.e., the loss of elastic be-
haviour. The principal stresses for torsion are given by

σ1 = τ θ
σ2 = 0 (6.11)
σ3 = −τθ

6.3 Elastic limit criteria for isotropic materials

The state of stress in a material point is generally characterized by the stress tensor σ. A
general stress criterion for failure, i.c. the limit of the elastic domain, is thus obtained through
some scalar function of the stress tensor, or an invariant of σ:

f (σ) = 0 (6.12)

The most frequently used limit criteria are those that distinguish elastic from elasto-plastic
behaviour in theories of plasticity.
An important scalar function, which is often connected to the occurrence of inelastic
deformations, is based on the stored energy inside the material. It has been shown that
deformation can be interpreted in terms of volume changes and shape changes of an in-
finitesimal piece of material. As shown in chapter (5), the constitutive relation for 3D linear
elasticity can be split in its hydrostatic part and its deviatoric part. Consequently, the strain
4
energy density W can be split accordingly. Making use of the compliance tensor S defined
by (5.24), one finds

W = 21 σ : ε
4
= 21 σ : S : σ
4
= 21 (σh + σd ) : S : (σh + σd )
³ 4 4 4 4
´
= 12 σh : S : σh + σh : S : σd + σd : S : σh + σd : S : σd (6.13)

112
 6.3. Elastic limit criteria for isotropic materials

4
Writing S in terms of the engineering moduli with expression (5.91) and (5.96) yields
4 4
σh : S : σh = 91 [tr(σ)]2 I : S : I
ν 1+ν 4 s
= 91 [tr(σ)]2 I : (− I I + I ):I
E E
9ν 3(1 + ν)
= 91 [tr(σ)]2 (− + )
E E
3(1 − 2ν)
= 19 [tr(σ)]2
E
3(1 − 2ν)
= 19 [tr(σ)]2
E
1 p2
= [tr(σ)]2 = (6.14)
9K K
while
4 4
σh : S : σd = 31 tr(σ) I : S : σd
ν 1+ν 4 s
= 31 tr(σ) I : (− I I + I ) : σd
E E
1 1+ν d
= 3 tr(σ) I : ( σ )
E
=0 (6.15)
4
= σd : S : σh (6.16)
Finally, the last term of equation (6.13) is elaborated to
µ ¶
d 4 d d ν 1+ν 4 s
σ : S : σ = σ : − II + I : σd
E E
1+ν d
= (σ : σd )
E
1
= (σd : σd )
2G
1
= − J 2 (σd ) (6.17)
G
Putting all terms together in (6.13) leads to the additive split of the stored energy density
W = Wv + Wd (6.18)
2
p
Wv = (6.19)
2K
1 1
Wd = − J 2 (σd ) = (σd : σd ) (6.20)
2G 4G
The volumetric (or hydrostatic) part Wv of the stored energy only involves the bulk modulus
K as elasticity constant, while the deviatoric part Wd only involves the shear modulus G.

6.3.1 Rankine criterion

A typical stress-based limit criterion is the Rankine criterion, which simply states that the
limit of the elastic domain is reached as soon as the absolute value of one of the principal
stresses exceeds a material dependent value, i.e.,
max |σi | = σR (6.21)
i

113
6. L IMIT DESIGN CRITERIA

where σR is called the Rankine equivalent stress. This limit theory inherits its name from W.
Rankine (19th century), and is often referred to as the maximum principal stress theory. The
limit value σR is determined from the stress in a tensile test at the onset of yielding

σR = σY = σ t tensile test, onset yielding (6.22)

The Rankine limit surface is depicted in the three-dimensional space of the principal stresses
in figure 6.1. In the case of plane stress, the third principal stress σ3 equals zero, and the cor-

Figure 6.1: The Rankine limit surface in the principal stress space.

responding limit surface simplifies to the curve shown in figure 6.2. The criterion is clearly
sensitive to hydrostatic stresses, but it only involves the stresses on a single plane (on which
the largest principal stress acts) while the stress state on the two other principal planes is
completely neglected. In the case of torsion, the maximum principal stress would simply
equal τθ and the Rankine criterion thus implies that the ultimate elastic stress in tension
would equal the ultimate elastic stress in torsion, i.e.,

σ t = τθ (6.23)

This is not experimentally observed for ductile materials, where the ultimate elastic stress in
torsion τθ is much less than in tension. The Rankine criterion is therefore only applicable to
brittle materials.

6.3.2 Tresca yield criterion

Another stress-based limit criterion is the Tresca yield criterion, which states that the limit
of elastic behaviour is reached once the maximum shear stress in a material point reaches a

114
 6.3. Elastic limit criteria for isotropic materials

σ
2

σ
1

Figure 6.2: The Rankine limit surface for plane stress.

particular value that corresponds to the maximum shear stress in a tensile specimen at the
onset of yielding, i.e.,
τns max = 21 (σ1 − σ3 ) = 21 σt = 12 σY (6.24)
For a general state of stress, yielding occurs if one of the maximum shear stresses reaches
σY , i.e.,

τns max,1 = 12 σY =⇒ ±(σ1 − σ2 ) = σY or

τns max,2 = 12 σY =⇒ ±(σ1 − σ3 ) = σY or
1
τns max,3 = 2 σY =⇒ ±(σ2 − σ3 ) = σY

The corresponding limit surface, depicted in the principal stress space in figure 6.3, is a reg-
ular hexagonal prismatic surface, that is strictly insensitive to the hydrostatic stress in the
material. The prism has an infinite length in the direction of the mean stress σ1 = σ2 = σ3 ,
i.e., one of the octahedral directions. This means that if one cuts the surface by the octahe-
dral plane corresponding to the mean stress (three equal direction cosines with respect to
the principal directions), one always retrieves the same curve (the so-called C -curve), which
is a regular hexagon for the Tresca criterion. Note that this octahedral direction is the diag-
onal in the principal space for which σ1 = σ2 = σ3 . This particular octahedral plane is also
called the π-plane. The hexagon is shown further on, in figure 6.7.
In the case of plane stress, the prism is cut by a plane that is not normal to the octahe-
dral axis of the prism, an one finds a criterion that simplifies to an irregular two-dimensional
hexagon, shown in figure 6.4. Like the von Mises criterion given hereafter, the Tresca crite-
rion is particularly suited for ductile materials in which high shears occur upon yielding.
For the torsion test, substitution of the principal stresses in (6.24) gives

1
τθ = 2 σY (6.25)

and hence the limit shear stress in torsion is indeed lower than the limit tensile stress.

115
6. L IMIT DESIGN CRITERIA

Figure 6.3: The Tresca yield surface in the principal stress space.

σ
2

σ
1

Figure 6.4: The Tresca limit surface for plane stress.

116
 6.3. Elastic limit criteria for isotropic materials

6.3.3 Von Mises yield criterion

The von Mises criterion is a widely used limit criterion, that is essentially based on the defor-
mation mechanisms in crystalline materials, which makes it a suitable criterion for the elas-
tic limit analysis of most metals. As stated earlier, linear elasticity implies that the underlying
microstructure is not modified upon deformation. In crystalline materials, the microstruc-
ture starts to change as soon as slip occurs on crystallographic planes, which is the onset of
plastic deformation. The essential difference between elastic and plastic deformation of a
crystal is illustrated in figure 6.5. Clearly, a slip mechanism in an ideal crystal takes place

n
a tio
f orm
e de
ttic
ta l la
Crys
Elastic deformation
(the lattice deforms)

Deviatoric or distortional
Cr deformation
ys
tal
log
rap
hic
sli
p

Plastic deformation
(the lattice planes slip)

Figure 6.5: Elastic and plastic deformation of a crystalline material.

without any volume change (it is a local state of pure shear, on which the maximum shear
stress criterion of Tresca is based), which means that the scalar measure that should be used
to characterize its initiation should depend on the deviatoric part or distortional part of the
elastically stored energy or the deviatoric part of the stress tensor. The criterion which was
proposed by Huber, von Mises and Hencky was therefore based on the strain energy density
of distortion. The criterion associates the elastic limit behaviour with the energy absorbed
in changing shape, since the hydrostatic stresses, responsible for volume changes, are not
present in this criterion.

The loss of elastic behaviour occurs as soon as the deviatoric strain energy density in
an arbitrary point equals the limit value that occurs in a classical tensile test at the onset of

117
6. L IMIT DESIGN CRITERIA

yielding. From equation (6.20) and (6.3), it appears that this happens if
h i
J 2 (σd ) = J 2 (σd ) (6.26)
tensile test
1 2
= − 2 ( 3 σt ~ e x~e x − 13 σt ~
e y~e y − 13 σt ~
e z~
ez ) :
2 1 1
( 3 σt ~ e x − 3 σt ~
e x~ e y~e y − 3 σt ~e z~
ez ) (6.27)
= − 12 ( 94 σ2t + 19 σ2t + 19 σ2t ) (6.28)
= − 13 σ2t (6.29)

If the yield stress σY is defined as the tensile stress at the onset of yield (i.e., σY = σt ), one
can defined the so-called von Mises equivalent stress σVM for a complex state of stress as
q
σVM = −3 J 2 (σd ) (6.30)
q
= 32 σd : σd (6.31)

If the von Mises equivalent stresses exceeds the value σY (where σY is determined from σVM
with a tensile test), the limit of the elastic domain is reached. The fact that σVM can also be
expressed in terms of the second invariant of σd explains why the name J 2 -plasticity is often
used starting with this limit surface. The von Mises equivalent stress can also be expressed
with respect to a given basis {~e x ,~
e y ,~
e z } or in terms of the principal stresses.
q
3 d
σVM = 2
σ : σd
q
3
= 2
[σ − 31 tr(σ)I ] : [σ − 13 tr(σ)I ]
q
3
= 2
[σ : σ − 32 tr(σ)I : σ + 19 [tr(σ)]2 I : I ]
q
3
= 2 (σ : σ − 31 [tr(σ)]2 ) (6.32)

which in terms of the principal stresses yields

q
σVM = 32 (σ21 + σ22 + σ23 − 13 [σ1 + σ2 + σ3 ]2 )
q
= 32 (σ21 + σ22 + σ23 − 13 [σ21 + σ22 + σ23 + 2σ1 σ2 + 2σ1 σ3 + 2σ2 σ3 ])
q
= 12 ([σ21 − 2σ1 σ2 + σ22 ] + [σ21 − 2σ1 σ3 + σ23 ] + [σ22 − 2σ2 σ3 + σ23 ])
q
= 12 ([σ1 − σ2 ]2 + [σ1 − σ3 ]2 + [σ2 − σ3 ]2 ) (6.33)

or with respect to a Cartesian basis {~

e x ,~
e y ,~
ez }
q
3
σVM = 2
2 (σxx + σ2y y + σ2zz − +2τ2x y + 2τ2xz + 2τ2y z − 31 [σxx + σ y y + σzz ]2 )
q
1
= 2 ([σxx − σ y y ]2 + [σxx − σzz ]2 + [σ y y − σzz ]2 ) + 3τ2x y + 3τ2xz + 3τ2y z (6.34)

Note that this criterion has been established from the distortional strain energy only, and
hence it is strictly insensitive to the hydrostatic pressure or volumetric deformations. The
corresponding limit surface (i.e., the yield surface) is depicted in figure 6.6, in terms of the
three principal stresses σ1 , σ2 and σ3 by means of equation (6.33). It is a cylindrical surface

118
 6.3. Elastic limit criteria for isotropic materials

Figure 6.6: The von Mises yield surface in the principal stress space.

that extends itself to infinity in the direction of the mean stress (i.e., it is not sensitive to
hydrostatic stress states). If one cuts the von Mises cylinder by the octahedral plane normal
to the mean stresses, one finds the same curve independent of the hydrostatic pressure, i.e.,
a circle. This circle fully circumvents the corresponding hexagon of the Tresca criterion, see
figure 6.7. This curve shows that the von Mises criterion constitutes a smooth envelope of the
Tresca criterion. In the particular case of plane stress (σ3 = 0), the 3D cylindrical limit surface
cuts the σ1 − σ2 plane (not normal to the octahedral axis of the cylinder) in an ellipse, which
is shown in figure 6.8.
Comparing equation (6.33) for the equivalent von Mises stress with the octahedral shear
stress obtained in equation (3.37) of chapter 3, one notices that σVM is directly related to the
octahedral shear stress by
3
σVM = p τoct (6.35)
2
This means that the von Mises yield criterion predicts the limit of the elastic domain at a
point where the octahedral shear stress reaches a particular limit value. Hence, the use of the
distortional energy or the octahedral shear stress leads to an identical criterion for a linear
elastic material.
Since the limit value of the von Mises stress is determined in simple tension, it is interest-
ing to evaluate the equivalent von Mises stress in the case of torsion. From equation (6.33)
for σVM and the principal stresses in torsion (6.11), it appears that yielding in the torsion test
would occur if p
σVM = 3 τθ = σY = σt (6.36)
Hence, the shear stress in torsion upon yield is clearly less than the tensile yield stress
1
τθ = p σ t (6.37)
3
which is also observed for ductile materials.

119
6. L IMIT DESIGN CRITERIA

von Mises
Tresca

σ3

σ1 σ2

Figure 6.7: The regular von Mises and Tresca C -curves in the π-plane

The differences between the 3D limit criteria given so far is shown in figure 6.9, where the
plane stress limit surface for the Rankine, von Mises and Tresca criterion are depicted for the
same value of σY .

6.3.4 Mohr-Coulomb limit criterion

The Mohr-Coulomb criterion distinguishes itself from all previous criteria in the sense that
it behaves different in traction or compression. It is physically motivated on the basis of
the existence of two physical parameters, i.e., an internal friction angle φ and an internal
cohesion c. The friction angle measures the resistance to frictional slip on individual planes
(a sort of internal Coulomb friction), while the cohesion measures the adhesion between two
planes. This limit criterion is typically used for cohesive-frictional materials, e.g., granular
materials, ceramics, etc. In chapter 3, normal and shear strains on individual planes in a
material point have been elaborated. The Mohr-Coulomb criterion states that the material
loses elasticity (i.e., changes in microstructure) as soon as the frictional slip or decohesion

120
 6.3. Elastic limit criteria for isotropic materials

σ2

σ1

Figure 6.8: The von Mises limit surface for plane stress.

σ2

Rankine
von Mises
Tresca

σ1

Figure 6.9: Comparison of three limit surfaces for plane stress.

121
6. L IMIT DESIGN CRITERIA

occurs. The resistance to these two phenomena on an arbitrary plane is expressed by

|τns | < − tan(φ) σnn + c (6.38)

The limit state is reached if (6.38) is violated on one single plane, which is then called the
critical plane. On the basis of the equations elaborated for τns and σnn in terms of the di-
rection cosines of the normal ~ n on the plane, it is possible to compute this critical plane.
Here however, use will be made of Mohr’s circles for 3D, since it points directly to the answer
we are looking for. The three circles are sketched in figure 6.10. The limit criterion (6.38) is

c
tg ( f)

s3 s2 s1 c
f s nn

t ns
Figure 6.10: The Mohr-Coulomb limit state reached on Mohr’s circles

sketched in this τns − σnn figure as well, such that only one single point on the circles (i.e.,
the critical orientation) reaches that limit state. The limit criterion, which is a straight line,
is tangent to the outer circle spanned by the largest and smallest principal stress σ1 and σ3
respectively. The radius of this outer circle is perpendicular to the limit line. Goniometric
considerations on this figure therefore lead to

− 12 (σ3 − σ1 )
sin(φ) = (6.39)
− 12 (σ3 + σ1 ) + tan(φ)
c

which is further elaborated to the 3D Mohr-Coulomb criterion

· ¸
1 1 c
2 (σ3 − σ1 ) = 2 (σ3 + σ1 ) − tan(φ) sin(φ)
1
2 (σ3 − σ1 ) = sin(φ) 12 (σ3 + σ1 ) − cos(φ) c (6.40)

Elastic behaviour implies

σ1 − σ3 σ1 + σ3
+ sin(φ) < cos(φ) c (6.41)
2 2
Once the internal friction angle φ and the cohesion c have been determined (experimen-
tal set-ups exist for this purpose), a graph of the corresponding 3D limit surface can be made,

122
 6.3. Elastic limit criteria for isotropic materials

Figure 6.11: The Mohr-Coulomb yield surface in the principal stress space.

see figure 6.11. One clearly notices from this limit surface, that it is a so-called pressure sen-
sitive limit surface. If the hydrostatic pressure increases the elastic domain extends, which
gives a pyramidal shape to the surface with a hexagonal basis, which is apparent from fig-
ure 6.11. The hexagonal basis stems from the fact that the maximum shear stress entered the
limit equation, similar to the Tresca yield surface but different in shape. Cutting the pyramid
by the octahedral plane normal to the pyramid axis gives an irregular hexagon, further shown
in figure 6.14. The shape of this hexagon is not affected by the hydrostatic pressure, but the
size definitely is. The larger the pressure (negative mean stress), the larger the hexagon will
be. At some positive mean stress, the hexagon shrinks to a point and there is no elastic stress
state beyond that point.
The resemblance and differences with the Tresca criterion can also be noticed from the
plane stress limit curve depicted in figure 6.12. This curve is obtained by cutting the pyramid
with the σ1 − σ2 plane, which is not normal to the octahedral pyramid axis. For the Mohr-
Coulomb limit curve, the elasticity limit is different for biaxial compression (σ1 = σ2 < 0)
compared to biaxial tension (σ1 = σ2 > 0). This is due to the pyramidal shape, which de-
scribes the weakness of the material if the hydrostatic pressure diminishes or reverts to hy-
drostatic tension. This type of behaviour is characteristic for brittle materials with a granular
structure. Note that if the internal friction angle φ equals zero and the cohesion equals the
yield stress σY , than the Tresca criterion is recovered.

6.3.5 Drucker-Prager limit criterion

Since the Mohr-Coulomb criterion was expressed in terms of the maximum shear stress
1
2 (σ3 − σ1 ) in analogy with the Tresca criterion, it seems convenient to formulate a cohesive-
frictional criterion that presents an analogy with the von Mises criterion. The latter two have
the advantage that they are smooth in the three-dimensional principal stress space, which is
physically and numerically more appealing. Furthermore, a sound physical motivation has
been given for the von Mises criterion. The von Mises criterion can be obtained directly by

123
6. L IMIT DESIGN CRITERIA

σ2

σ1

Figure 6.12: The Mohr-Coulomb limit surface in 2D plane stress.

substituting the maximum shear stress by a scalar multiplication with the octahedral shear
stress, see equation (6.35). In the same way, the term 12 (σ3 + σ1 ) in the Mohr-Coulomb crite-
rion can be generalized to the hydrostatic stress 31 (σ3 + σ2 + σ1 ). Drucker and Prager made
this generalization under the additional constraint that the obtained limit surface consti-
tutes and envelope of the Mohr-Coulomb surface, like the von Mises surface was an envelope
of the Tresca limit surface. This leads to the Drucker-Prager limit criterion, given by
q
3 d 6 sin(φ) 6 cos(φ)
2σ : σd + p < c (6.42)
3 − sin(φ) 3 − sin(φ)

The corresponding 3D limit surface is again shown in figure 6.13. The Drucker-Prager limit
surface is conical with a circular base curve instead of the irregular hexagon of the Mohr-
Coulomb pyramid. This base curve, along with the irregular hexagon of the Mohr-Coulomb
criterion, is visible in the cross-section of the cone and the pyramid by the octahedral plane
given in figure 6.14.
The Drucker-Prager cone fully circumvents the Mohr-Coulomb pyramid, but does not
pass through all angular points of the Mohr-Coulomb hexagon (which is different from von
Mises compared to Tresca). Like for the Mohr-Coulomb criterion, the size of the circle de-
pends on the hydrostatic stress.
The corresponding plane stress curve for Drucker-Prager is again obtained by cutting the
cone with the σ1 − σ2 plane, which is not normal to the cone axis. Hence, an ellipse is ob-
tained, shown in figure 6.15. The ellipse again illustrates the differences in compression and
tension. The elastic domain in compression is much larger compared to tension. Combining
the plane stress limit curves for Mohr-Coulomb and Drucker-Prager gives figure 6.16, which
shows that the differences between these two criteria are considerably larger than for non-
frictional materials. Again it should be noted that both limit curves lead to the Tresca and
von Mises limit curves, once the internal friction angle equals zero and the cohesion equals
the yield stress. In that sense, the cohesive-frictional criteria constitute a generalization of
the more ’classical’ yield criteria.

124
 6.3. Elastic limit criteria for isotropic materials

Figure 6.13: The Drucker-Prager yield surface in the principal stress space.

Drucker−Prager
Mohr−Coulomb

σ
3

σ σ
2 1

Figure 6.14: The C -curves for the Mohr-Coulomb and Drucker-Prager limit cones in the π-
plane.

125
6. L IMIT DESIGN CRITERIA

σ2

σ1

Figure 6.15: The Drucker-Prager limit surface for plane stress.

σ
2
Mohr−Coulomb
Drucker−Prager

σ1

Figure 6.16: Comparison of Mohr-Coulomb and Drucker-Prager for plane stress.

126
 6.4. Limit criteria for anisotropic materials

6.3.6 General criteria for isotropic solids

On the basis of equation (6.12), many more criteria can be established. The major motivation
for doing so, is that the available criteria insufficiently characterize the true limit behaviour
of the particular material of interest. Other ellipsoidal, conical, pyramidal, paraboloidal sur-
faces may be constructed. We limit ourselves here to one additional example, where the limit
surface is formulated with respect to the octahedral planes in a material point. It was shown
that the octahedral normal and shear stress are equal for all eight octahedral planes.

f [J 1 (σ)] + g [J 2 (σ)] τoct + h[J 3 (σ)] τ2oct = 0 (6.43)

Both the von Mises and Tresca yield surfaces can be rephrased in this format with particular
functions f , g and h.

6.4 Limit criteria for anisotropic materials

For anisotropic materials, it is not possible to develop a limit criterion in the principal stress
space, and the anisotropy directions inevitably enter the limit criterion. A lot of research has
been devoted to this subject, and many criteria are nowadays available for specific materials
or specific anisotropic arrangements (composites).
Like for the von Mises and Tresca criterion, a special family of anisotropic criteria can
be specified which is also independent with respect to the hydrostatic pressure and which is
also expresses as a quadratic form of the deviatoric stress tensor.While the von Mises crite-
rion is formulated in terms of σd : σd , an equivalent anisotropic form could be formulated
as the equivalent stress
4
σ2eq = σd : A : σd (6.44)
4
The fourth-order tensor A is the anisotropy tensor, that accounts for all symmetries present
in the material. The obvious symmetries of this tensor immediately reduces the number of
relevant components to 21 (like for the elasticity tensor). The independence of the hydro-
static pressure (A j j kl = 0) further reduces the number to 15 for the case of general anisotropy.
4
Special cases of symmetry reduces the number further (like for C ). A well-known example
is Hill’s criterion for metals. Hill considered the special case of orthotropic symmetry for
thin rolled sheets having orthotropic symmetries with respect to the rolling direction(~ e 1),
the transverse (~ e 2 ) and the normal direction (~
e 3 ). Taking into account the independence of
the hydrostatic pressure, Hill’s orthotropic criterion depends on 6 coefficients. The quadratic
form reads
4
σd : A : σd = F (σ22 − σ33 )2 +G (σ33 − σ11 )2 + H (σ11 − σ22 )2 +
(6.45)
2L σ223 + 2M σ213 + 2N σ212

Hill’s criterion simplifies again to von Mises in the case of full isotropy, and it is in this sense
the anisotropic equivalent of it.
Note that more general criteria taking into account different limit values in tension or
compression (like for Mohr-Coulomb and Drucker-Prager) are also available, e.g., Tsai’s cri-
terion in which a linear term in σd is added as well. This type of criteria has proven to be
particularly efficient for anisotropic composites.

127
6. L IMIT DESIGN CRITERIA

6.5 Examples
Equivalent Von Mises stress

The stress state in a point is represented by the following Cauchy stress tensor :

e 1~
σ = 3σ~ e 2~
e 1 − σ~ e 3~
e 2 − 2σ~ e 1~
e 3 + σ(~ e 2~
e 2 +~ e 1)
The corresponding Cauchy stress matrix is
 
3σ σ 0
σ =  σ −σ 0 
0 0 −2σ
The Von Mises equivalent stress is defined as
q q q
3 d 3 3
σ̄V M = 2
σ : σd = 2
tr(σd ·σd ) = 2
tr(σd σd )

This equivalent stress expanded in terms of a Cartesian basis was given by equation
(6.34), and reads
q
1
σVM = 2
([σxx − σ y y ]2 + [σxx − σzz ]2 + [σ y y − σzz ]2 ) + 3τ2x y + 3τ2xz + 3τ2y z

Substitution of the stress components σxx = 3σ, σ y y = −σ, σzz = −2σ and σx y = σ yx =
σgives for the equivalent von Mises stress
p
σVM = 2 6 σ

Equivalent Von Mises and Tresca stresses

The stress state in a point is represented by the following Cauchy stress tensor :

e 1~
σ = σ~ e 2~
e 1 + σ~ e 3~
e 2 + σ~ e 1~
e 3 + τ(~ e 2~
e 2 +~ e1)
with all component values positive.
The Tresca yield criterion states that yielding will occur when the maximum shear stress
reaches a limit value, which is determined in a tensile experiment. The equivalent Tresca
stress is two times the maximum shear stress.

σ̄T R = 2τmax = σmax − σmi n

The limit value is the one-dimensional yield stress σ y0 . To calculate σ̄T R , we need the
principal stresses, which can be determined by requiring the matrix σ − s I to be singular.
 
σ−s τ 0
det(σ − s I ) = det  τ σ−s 0 =0 →
0 0 σ−s
(σ − s)3 − τ2 (σ − s) = 0 → (σ − s){(σ − s)2 − τ2 } = 0 →
(σ − s)(σ − s + τ)(σ − s − τ) = 0 →
σ1 = σmax = σ + τ ; σ2 = σ ; σ3 = σmi n = σ − τ

128
 6.5. Examples

The equivalent Tresca stress is

σ̄T R = 2τ
so yielding according to Tresca will occur when

τ = 21 σ y0
The equivalent Von Mises stress is expressed in the principal stresses :
q
1 2 2 2
σ̄V M = 2 {(σ1 − σ2 ) + (σ2 − σ3 ) + (σ3 − σ1 ) }

and can be calculated by substitution,

p p
σ̄V M = 3τ2 = 3τ
Yielding according to Von Mises will occur when the equivalent stress reaches a limit
value, the one-dimensional yield stress σ y0 , which results in
p
τ = 31 3 σ y0

129
6. L IMIT DESIGN CRITERIA

6.6 Elastic limit behaviour: Summary

Elastic limits and failure

Plastic yielding Onset of permanent deformations

Brittle fracture Sudden crack growth

Continuum damage Progressive loss of stiffness

Fatigue Failure under cyclic loading

Dynamic Failure Resonance

Thermal Failure Temperature dependent degradation

Elastic instabilities Geometrical and structural phenomenon

Standard tests

Tension σ1 = σ t
σ2 = σ3 = 0
σt
σh = 3 I
2
σd = ~~
3 σt e x e x − 13 σt ~ e y − 31 σt ~
e y~ e z~
ez

Torsion e t~
σ = G α r~ e z~
e z +G α r~ et
3
Circular M t = 2πG α R3
σ1 = τ θ
σ2 = 0
σ3 = −τθ

130
 6.6. Elastic limit behaviour: Summary

Limit criteria for isotropic materials

General f (σ) = 0

p2
Strain energy density split Wv = 12 K
and Wd = 12 1
2G
(σd : σd )

Rankine Maximum principal stress criterion

max |σi | = σY
i
τθ = σ t

Tresca Maximum shear stress criterion

σ1 − σ3 = σY
Insensitive to hydrostatic pressure
τθ = 12 σY

von Mises Distortional strain energy density criterion

p
σVM = −3 J 2 (σd )
q
σVM = 32 σd : σd
q
σVM = 12 ([σ1 − σ2 ]2 + [σ1 − σ3 ]2 + [σ2 − σ3 ]2 )
σVM = [ 21 ([σxx − σ y y ]2 + [σxx − σzz ]2 + [σ y y − σzz ]2 )
+3τ2x y + 3τ2xz + 3τ2yz ]1/2
Insensitive to hydrostatic pressure
Equivalent to octahedral shear stress criterion
σVM = p3 τoct
2
σ
τθ = pt
3

Mohr-Coulomb Hydrostatic sensitive cohesive-frictional

1 σ1 +σ3
2 (σ1 − σ3 ) + sin(φ) 2 < cos(φ) c
Pyramidal limit surface

Drucker-Prager Hydrostatic sensitive smooth cohesive-frictional

q
3 d d 6sin(φ) 6cos(φ)
2 σ : σ + 3−sin(φ) p < 3−sin(φ) c
Conical limit surface

131
6. L IMIT DESIGN CRITERIA

General criteria

Isotropic f (J 1 (σ) + g (J 2 (σ) τoct + h(J 3 (σ) τ2oct = 0

4
Hill orthotropic σ2eq = σd : A : σd

4
σd : A : σd = F (σ22 − σ33 )2 +G (σ33 − σ11 )2 +
H (σ11 − σ22 )2 + 2L σ223 + 2M σ213 + 2N σ212

132
 CHAPTER
7
M ECHANICAL DESIGN PROBLEMS :
GOVERNING EQUATIONS AND
SOLUTION METHODS

In this chapter we will recall the equations, which have to be solved to determine the defor-
mation of a three-dimensional linear elastic material body under the influence of an external
load. The equations will be written in component notation w.r.t. a Cartesian and a cylindri-
cal vector base and simplified for plane strain, plane stress and axisymmetry. The material
behavior is assumed to be isotropic.

7.1 Vector/tensor equations

The deformed (current) state is determined by 12 state variables : 3 displacement compo-
nents and 9 stress components. These unknown quantities must be solved from 12 equa-
tions : 6 equilibrium equations and 6 constitutive equations.
With proper boundary (and initial) conditions the equations can be solved, which, for
practical problems, must generally be done numerically. The compatibility equations are
generally satisfied for the chosen strain-displacement relation. In some solution approaches
they are used instead of the equilibrium equations.

~ T
gradient operator : ~ ~
∇=∇ e
~
T
position : ~ ~ ~
x=x e
~
T
displacement : ~
u =u ~n ~
e
~
¡ ¢T ¡ ¢o
strain : ε = 21 ~ u + ~
∇~ u =~
∇~ e T ε~
~ e
~
compatibility : ∇ {tr(ε)} − ~
2
∇·(~ T
∇·ε) = 0
stress : e T σ~
σ =~
~ e
~
balance laws : ~ T
∇·σ + ρ~ u¨
q = ρ~ ; σ = σT
−1
material law : σ = 4C : ε ; ε = 4C : σ = 4S : σ

133
7. M ECHANICAL DESIGN : SOLUTIONS

7.2 Three-dimensional scalar equations

The vectors and tensors can be written in components with respect to a three-dimensional
vector basis. For various problems in mechanics, it will be suitable to choose either a Carte-
sian coordinate system or a cylindrical coordinate system.

7.2.1 Cartesian components

The governing equations are written in components w.r.t. a Cartesian vector base {~
e x ,~
e y ,~
e z }.
The stresses can be represented with a Cartesian stress cube, see figure 7.1.

~
ez
~
u σzz

~
ex σ yz
~
ey
σxz
σz y
~
ez ~
x σy y
σzx
~
ez σ yx
~
ey y σx y
~
ey
~
ex
σxx
~
ex
x

Figure 7.1: Cartesian coordinate system and stress cube

£ ¤ h i £ ¤
T T ∂ ∂ ∂ T
x
~ = x y z ; ∇
~ = ∂x ∂y ∂z ; u
~ = ux uy uz

 
2u x,x u x,y + u y,x u x,z + u z,x
ε = 12  · · · 2u y,y u y,z + u z,y 
··· ··· 2u z,z

2εx y,x y − εxx,y y − ε y y,xx = 0 → cyc. 2x

εxx,yz + ε yz,xx − εzx,x y − εx y,xz = 0 → cyc. 2x
£ ¤
εT = £ εxx ε y y εzz εx y ε yz εzx
~ ¤
T σ σ σ σ σ σ
σ
~ = xx y y zz x y yz zx

σxx,x + σx y,y + σxz,z + ρq x = ρ ü x (σx y = σ yx )

σ yx,x + σ y y,y + σ yz,z + ρq y = ρ ü y (σ yz = σz y )
σzx,x + σz y,y + σzz,z + ρq z = ρ ü z (σzx = σxz )

σ ε
~ = C~ ; ε = Sσ
~ ~

134
 7.2. Three-dimensional scalar equations

7.2.2 Cylindrical components

The governing equations are written in components w.r.t. a cylindrical vector base {~
e r (θ),~
e t (θ),~
e z },
with :

d~
er d~
et
et
=~ and er
= −~
dθ dθ
The stress components can be represented with a cylindrical stress cube, see figure 7.2.

z
~
u σzz

~
ez σr z

~et z σt z
~
er
σtr
~
ez ~
x σr r
~
ey σr t r
y σzr
σzt
~
ex ~er
σt t
θ
r θ
x

Figure 7.2: Cylindrical coordinate system and stress cube

T
£ ¤ T
£ ∂ 1 ∂ ∂
¤ T
£ ¤
x
~ = r θ z ; ∇
~ = ∂r r ∂θ ∂z
; u
~ = ur ut uz
 1 
2u r,r r (u r,t − u t ) + u t,r u r,z + u z,r
1
ε= ··· 2 1r (u r + u t,t ) 1
u + u t,z 
r z,t
2
··· ··· 2u z,z

2εr t,r t − εr r,t t − εt t,r r = 0 → cyc. 2x

εr r,t z + εt z,r r − εzr,r t − εr t,r z = 0 → cyc. 2x

T
£ ¤
ε
~ = εr r ε t t ε zz εr t ε t z ε zr
T
£ ¤
~ = σr r σt t σzz σr t σt z σzr
σ

σr r,r + r1 σr t,t + r1 (σr r − σt t ) + σr z,z + ρq r = ρ ü r (σr t = σtr )

σtr,r + r1 σt t,t + r1 (σtr + σr t ) + σt z,z + ρq t = ρ ü t (σt z = σzt )
σzr,r + r1 σr t,t + r1 σzr + σr z,z + ρq z = ρ ü z (σzr = σr z )

σ ε
~ = C~ ; ε = Sσ
~ ~

135
7. M ECHANICAL DESIGN : SOLUTIONS

7.2.3 Material law

When deformations are small, every material will show linear elastic behavior. For orthotropic
material there are 9 independent material constants. When there is more material symme-
try, this number decreases. Finally, isotropic material can be characterized with only two
material constants.
Be aware that we use now the strain components εi j and not the shear components γi j .
In an earlier chapter, the parameters for orthotropic, transversally isotropic and isotropic
material were rewritten in terms of engineering parameters: Young’s moduli and Poisson’s
ratio’s.

   
A Q R 0 0 0 a q r 0 0 0

 Q B S 0 0 0 


 q b s 0 0 0 

   r s c 0 0 0 
R S C 0 0 0
 → S = C −1 = 
   
C = 1 
 0 0 0 2K 0 0   0 0 0 2k 0 0 
   1 
 0 0 0 0 2L 0   0 0 0 0 2l 0 
1
0 0 0 0 0 2M 0 0 0 0 0 2
m

quadratic B = A ; S = R ; M = L;
transversal isotropic B = A ; S = R ; M = L ; K = 21 (A −Q)
cubic C = B = A ; S = R = Q ; M = L = K 6= 12 (A −Q)
isotropic C = B = A ; S = R = Q ; M = L = K = 21 (A −Q)

7.3 Planar deformation

In many applications the loading and deformation is in one plane. The result is that the
material body is in a state of plane strain or plane stress. The governing equations can than
be simplified considerably.

7.3.1 Cartesian

In a plane strain situation, deformation in one direction – here the z-direction – is sup-
pressed. In a plane stress situation, stresses on one plane – here the plane with normal in
z-direction – are zero.
Eliminating σzz for plane strain and εzz for plane stress leads to a simplified Hooke’s law.
Also the equilibrium equation in the z-direction is automatically satisfied and has become
obsolete.

136
 7.3. Planar deformation

¾ ½
plane strain : εzz = εxz = ε yz = 0 u x = u x (x, y)
plane stress : σzz = σxz = σ yz = 0 u y = u y (x, y)
£ ¤ £ ¤
εT = εxx ε y y εx y = u x,x
~ u y,y 1
2 (u x,y + u y,x )
2εx y,x y − εxx,y y − ε y y,xx = 0

T
£ ¤
~ = σxx σ y y σx y
σ
σxx,x + σx y,y + ρq x = ρ ü x (σx y = σ yx )
σ yx,x + σ y y,y + ρq y = ρ ü y
   
Ap Qp 0 ap qp 0
C = Qp Bp
 0  ; S = qp
 bp 0 
p p 1
0 0 2K 0 0 2k

7.3.2 Cylindrical
In a plane strain situation, deformation in one direction – here the z-direction – is sup-
pressed. In a plane stress situation, stresses on one plane – here the plane with normal in
z-direction – are zero.
Eliminating σzz for plane strain and εzz for plane stress leads to a simplified Hooke’s law.
Also the equilibrium equation in the z-direction is automatically satisfied and has become
obsolete.

¾ ½
plane strain : εzz = εr z = εt z = 0 u r = u r (r, θ)
plane stress : σzz = σr z = σt z = 0 u t = u t (r, θ)

T
£ ¤ £ 1
¡
1 1
¢ ¤
ε
~ = εr r ε t t ε r t = u r,r r (u r + u t,t ) 2 r (u r,t − u t ) + u t,r
2εr t,r t − εr r,t t − εt t,r r = 0

T
£ ¤
σ
~ = σr r σt t σr t

σr r,r + 1r σr t,t + r1 (σr r − σt t ) + ρq r = ρ ü r (σr t = σtr )

σtr,r + r1 σt t,t + r1 (σtr + σr t ) + ρq t = ρ ü t

   
Ap Qp 0 ap qp 0
C = Qp Bp
 0  ; S = qp
 bp 0 
p p 1
0 0 2K 0 0 2k

7.3.3 Cylindrical : axisymmetric + u t = 0

∂( )
When geometry and boundary conditions are such that we have = ( )t = 0 the situation
∂θ
is referred to as being axisymmetric.
In many cases boundary conditions are such that there is no displacement of material
points in tangential direction (u t = 0). In that case we have εr t = 0 → σr t = 0

137
7. M ECHANICAL DESIGN : SOLUTIONS

¾ ½
plane strain : εzz = εr z = εt z = 0 u r = u r (r )
plane stress : σzz = σr z = σt z = 0 ut = 0

T
£ ¤ £ 1
¤
ε
~ = ε r r ε t t = u r,r r (u r )
εr r = u r,r = (r εt t ),r = εt t + r εt t,r

T
£ ¤
σ
~ = σr r σ t t
1
σr r,r + (σr r − σt t ) + ρq r = ρ ü r
r
· ¸ · ¸
Ap Qp ap qp
C = ; S =
p Qp Bp p qp bp

7.4 Stress inconsistency under plane stress

Although for plane stress the out-of-plane shear stresses must be zero, they are not, when
calculated afterwards from the strains. This inconsistency is inherent to the plane stress
assumption. Deviations must be small to render the assumption of plane stress valid.

σxz = 2K εxz = 2K u z,x 6= 0

σ yz = 2K ε yz = 2K u z,y 6= 0

7.5 Solution strategies

7.5.1 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 σ. For static problems, this ten-
sor has to satisfy the equilibrium equations. Solving stresses from these equations is gener-
ally not possible and additional equations are needed, which relate stresses to deformation.
These constitutive equations, which describe the material behavior, relate the stress tensor
σ to the strain tensor ε, which is a function of the displacement gradient tensor (~
∇~u ). Com-
ponents of this strain tensor cannot be independent and are related by the compatibility
equations.

unknown variables
¡ ¢T
displacements : ~
u =~
u (~
x) → F= ~∇0~
x → E,ε
stresses : σ
equations
compatibility : ∇2 {tr(ε)} − ~
∇·(~
∇·ε)T = 0
equilibrium : ~
∇·σT + ρ~ q = ρ~u¨ ; σ = σT
material law : σ = σ(F ) → σ = C : ε → ε = 4 S : σ
4

limit criteria
yield surface : g (σ) = g (σ1 , σ2 , σ3 ) = g t

138
 7.5. Solution strategies

7.5.2 Boundary conditions

Some of the governing equations are partial differential equations, where differentiation is
done w.r.t. the spatial coordinates. These differential equations can only be solved when
proper boundary conditions are specified. In each boundary point of the material body,
either the displacement or the load must be prescribed. It is also possible to specify a relation
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 ρ~ u¨ as its right-hand term. In that case a solution
can only be determined when proper initial conditions are prescribed, i.e. initial displace-
ment, velocity or acceleration. In this section we will assume ~ u¨ =~0.

displacement : ~
u =~up ∀ ~
x ∈ Au
edge load : ~ n=~
p = σ·~ pp ∀ ~ x ∈ Ap

Superposition principle

The superposition principle was introduced at the end of chapter 5, and is based on the path
independence and the linearity of the problem. This path independent and linear character

u1

F1
F2 u2

F2 u1 + u2

F1

Figure 7.3: Principle of superposition

permits to treat any loading case as an arbitrary subdivision of elementary loading cases, for
which the solution can be summed up in a trivial way. It permits to superimpose the results
for specific loading arrangements, in order to investigate their combined effect.
The principle of superposition states that the solution S for a given combined load L =
L 1 + L 2 is the sum of the solution S 1 for load L 1 and the solution S 2 for L 2 , so : S = S 1 + S 2 .

139
7. M ECHANICAL DESIGN : SOLUTIONS

7.6 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 definition in terms of the displacement
gradient. This results in a differential equation in the displacement ~u , which can be solved
when proper boundary conditions are specified.
In a Cartesian coordinate system the vector/tensor formulation can be replaced by index
notation.

¡4 ¢T 
¾ ~
∇· q =~0
C : ε + ρ~ 
∇·σT + ρ~
~ q =~0

4 → →
σ= C :ε n¡ ¢T ¡ ¢o 
ε = 21 ~

∇~u + ~ u
∇~ 

©4 ¡ ¢ªT
~
∇· C: ~ u
∇~ q =~0
+ ρ~ → ~
u → ε → σ

Cartesian index notation

¾ ¾
σi j ,j + ρq i = 0i C i j kl εl k,j
¡ + ρq i = ¢0i
→ 1 →
σi j = C i j l k εl k εl k = 2 u l ,k + u k,l

C i j kl u l ,k j + ρq i = 0i → ui → εi j → σi j

7.6.1 Planar, Cartesian : Navier equations

The displacement method is elaborated for planar deformation in a Cartesian coordinate
system. Linear deformation and linear elastic material behavior is assumed. Elimination
and substitution results in two partial differential equations for the two displacement com-
ponents.

σxx,x + σx y,y + ρq x = ρ ü x ; σ yx,x + σ y y,y + ρq y = ρ ü y 





σxx = A p εxx +Q p ε y y

σ y y = Q p εxx + B p ε y y 




σx y = 2K εx y
¾
A p εxx,x +Q p ε y y,x + 2K εx y,y + ρq x = ρ ü x
2K εx y,x +Q p εxx,y + B p ε y y,y + ρq y = ρ ü y
¾
A p u x,xx +Q p u y,yx + K (u x,y y + u y,x y ) + ρq x = ρ ü x
K (u x,yx + u y,xx ) +Q p u x,x y + B p u y,y y + ρq y = ρ ü y
¾
A p u x,xx + K u x,y y + (Q p + K )u y,yx + ρq x = ρ ü x
K u y,xx + B p u y,y y + (Q p + K )u x,x y + ρq y = ρ ü y

7.6.2 Planar, axisymmetric with u t = 0

Many engineering problems present a rotational symmetry w.r.t. an axis. They are axisym-
metric. In many cases the tangential displacement is zero : u t = 0. This implies that there are
no shear strains and stresses.

140
 7.7. Weighted residual formulation

The radial and tangential stresses are related to the radial and tangential strains by the
planar material law. Material parameters are indicated as A p , B p and Q p and can later
be specified 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 differential equa-
tion for the radial displacement u r

displacements : u r = u r (r ) ; u z = u z (r, z)
strains : εr r = u r,r ; εt t = r1 u r ; εzz = u z,z
stresses : σt z = 0 ; σr z ≈ 0 ; σtr = 0
eq. of motion : σr r,r + r1 (σr r − σt t ) + ρq r = ρ ü r
¾
σr r = A p εr r +Q p εt t
→ eq. of motion →
σt t = Q p εr r + B p εt t
1 1
u r,r r + u r,r − ζ2 2 u r = f (r )
r sr
Bp
with ζ=
Ap
ρ ¡ ¢
and f (r ) = ü r − q r
Ap

7.6.3 Planar, axisymmetric with u t = 0, isotropic

For isotropic material the coefficients A p and B p are the same, which implies that ζ = 1 and
also Θp1 = Θp2 .
¾
σr r = A p εr r +Q p εt t
→ eq. of motion →
σt t = Q p εr r + A p εt t
1 1
u r,r r + u r,r − 2 u r = f (r )
r r
ρ ¡ ¢
with f (r ) = ü r − q r
Ap

7.7 Weighted residual formulation

Unknown variables have to be solved from the combined set of equilibrium equations and
constitutive equations. Some of the equilibrium equations are partial differential equations.
For the general case of large deformations and nonlinear material behavior, the equations
are nonlinear. It is obvious that only for academic and very simple cases, analytic solutions
exist. For more practical problems, approximate solutions must be determined with a nu-
merical technique, of which the finite element method is widely used and will be considered
here.
Application of the finite element method in continuum mechanics requires the reformu-
lation of the equilibrium equations. They are transformed from differential equations to an
integral equation, the so-called weighted residual integral.
First, we formulate the weighted residual integral for linear problems, so for small defor-
mation and linear elastic material behavior. The finite element method is then explained for
this case. Examples of plane stress, plane strain and axisymmetric problems will be calcu-
lated with a Matlab program.

141
7. M ECHANICAL DESIGN : SOLUTIONS

Subsequently, we formulate the weighted residual integral for nonlinear problems, where
the iterative solution procedure has to be applied. Finite element analyses can be done again
with a Matlab program.

7.7.1 Weighted residual formulation for 3D deformation

For an approximation, the equilibrium equation is not satisfied exactly in each material
point. The error can be ”smeared out” over the material volume, using a weighting func-
~ (~
tion w x ).

equilibrium equation ∇·σT + ρ~

~ q =~0 ∀~
x ∈V
approximation → residual ~
∇·σT + ρ~ q =~ x ) 6=~0
∆(~ ∀~x ∈V
Z Z
£ ¤
weighted residual w x )·~
~ (~ ∆(~x ) dV = w ∇·σT + ρ~
~· ~ q dV
V V

~,
When the weighted residual integral is satisfied for each allowable weighting function w
the equilibrium equation is satisfied in each point of the material.
Z
£ ¤
~· ~
w ∇·σT + ρ~
q dV = 0 ∀ ~ (~
w x) ↔ ∇·σT + ρ~
~ q =~0 ∀~
x ∈V
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
continuous 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
formulation of the weighted residual integral.
Gauss theorem is used to transfer the volume integral with the term ~ ∇.( ) to a surface
T T
integral. Also ~ n =~
p = σ·~ n ·σ and σ = σ is used.
Z 
£ ¤
w ∇·σT + ρ~
~· ~ q dV = 0 

V
→
T T c T

~
∇·(σ ·w ~
~ ) = (∇w~) : σ +w ~
~ ·(∇·σ )

Z 
£ ¤
∇·(σT ·w
~ ~ ) − (~
∇w~ )T : σT + w
~ ·ρ~
q dV = 0 ~
∀w 




V
Z Z Z →
∇·(σT ·w
~ ~)= ~ T
~ dA=
n ·σ ·w ~ ·~
w pdA 




V V A
Z Z Z
T
(~
∇w~ ) : σ dV = w~ ·ρ~
q dV + w~ ·~ ~)
p d A = f e (w ~
∀w
V V A

7.7.2 Weighted residual formulation : linear

When deformation and rotations are small, the deformation is geometrically linear. The
deformed state is almost equal to the undeformed state. This implies that integration can be
carried out over the undeformed volume V0 and the undeformed area A 0 .

142
 7.7. Weighted residual formulation

The material behavior is described by Hooke’s law, which can be substituted in the weighted
residual integral, according to the displacement solution method.
The weighted residual integral is now completely expressed in the displacement ~ u . Ap-
proximate solutions can be determined with the finite element method.
Z Z Z
~ T
(∇w~ ) : σ dV0 = w ~ ·ρ~
q dV0 + w~ ·~ ~)
p d A 0 = f e0 (w ∀w~
V0 V0 A0
© ª
σ = 4C : ε = 4C : 21 (~ u ) + (~
∇0 ~ u )T = 4C : (~
∇0 ~ ∇0 ~
u)

Z Z Z
(~ ~ )T : 4C : (~
∇0 w ∇0 ~
u ) dV0 = ~ ·ρ~
w q dV0 + ~ ·~
w ~)
p d A 0 = f e0 (w ~
∀w
V0 V0 A0

7.7.3 Finite element method for 3D deformation

Discretisation

The integral over the 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. Sub-
dividing the volume implies that also the surface with area A is subdivided in element sur-
faces (faces) with area A e .

Figure 7.4: Finite element discretisation

XZ XZ XZ
(~
∇w~ )T : 4C : (~
∇~u ) dV e = ~ ·ρ~
w q dV + e
w p d Ae
~ ·~ ~
∀w
e e eA
Ve Ve Ae

Isoparametric elements

Each point of a three-dimensional element can be identified with three local coordinates
{ξ1 , ξ2 , ξ3 }. 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 determinant
of this matrix relates two infinitesimal volumes, areas or lengths of both element represen-
tations.
isoparametric (local) coordinates (ξ1 , ξ2 , ξ3 ) ; − 1 ≤ ξi ≤ 1 i = 1, 2, 3

T T
¡ ¢
Jacobian matrix J= ∇~ ξx
~ ; dV e = det(J ) d ξ1 d ξ2 d ξ3

143
7. M ECHANICAL DESIGN : SOLUTIONS

ξ2
ξ2

ξ1 ξ1
ξ1

ξ3

Figure 7.5: Isoparametric elements

Interpolation : 4-node linear element

When the elements have a simple shape, e.g. an six-sided volume, the shape and thus vol-
ume is known when the position of a discrete number of edge points is known. These points
are the element nodal points. For a cube with plane faces, eight corner points are needed.
In two dimensions quadrilaterals with straight edges can be used, where four corner nodes
describe the shape. The position of an internal point of the element can be expressed in the
position of the nodal points. This interpolation is done with so-called shape- or interpola-
tion functions, which are a function of local element coordinates ξi , i = 1, 2, 3, which assume
values between -1 and +1.

3
4 ξ2
4 3

ξ1
1 2 1 2

Figure 7.6: Quadrilateral element with four corner nodes

x = N 1 (ξ)~
~ x 1 + N 2 (ξ)~
x 2 + N 3 (ξ)~
x 3 + N 4 (ξ)~
x4
~ ~ ~ ~
with

N 1 = 41 (ξ1 − 1)(ξ2 − 1) ; N 2 = − 14 (ξ2 + 1)(ξ2 − 1)

N 3 = 41 (ξ1 + 1)(ξ2 + 1) ; N 4 = − 14 (ξ1 − 1)(ξ2 + 1)

The value of the unknown quantity – here the displacement vector ~ u – in an arbitrary
point of the element, can also be interpolated between the values of that quantity in the
element nodes.

144
 7.7. Weighted residual formulation

~u = N 1 (ξ)~
u 1 + N 2 (ξ)~
u 2 + N 3 (ξ)~
u 3 + N 4 (ξ)~
u4 = N T
~ (ξ ue
)~
~ ~ ~ ~ ~~
Besides ~
x and ~u , 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.

~ = N 1 (ξ) w
w ~ 1 + N 2 (ξ) w
~ 2 + N 3 (ξ) w
~ 3 + N 4 (ξ) w
~4 =N T
~ (ξ ~e
)w
~ ~ ~ ~ ~ ~
The gradient of ~
u and w~ also has to be elaborated. and can be written as the product of a
column which contains the derivatives of the interpolation functions, and the column with
nodal components of ~ u and ~u.

~ u = (~ T e ~T ~ e ~ ~ = (~ T e ~T w e
∇~ ∇N~) ~
u
~ =B
~ u~ ; ∇w ∇N ~
~) w ~ ~
~ =B ~
Finally, everything is substituted in the weighted residual integral. The volume integral
in the left hand side is the element stiffness matrix K e . The integrals in the right hand side
e
represent the external load and are summarized in the column ~ f .
~e
Z Z Z
T e T 4 T e e eT e eT
~
(B
~ w~
~ ) : C : (B
~
~ ~~ ) dV = w
u ~
~ N q dV + w
~ ·ρ~ ~~ N p d Ae
~ ·~
Ve Ve Ae
     
Z Z Z
~ T dV e  ·~
T T
e  ~ · 4C ·B e e  eT 
~
w
~ · B
~ ~ u
~ =w~
~ · N q dV e  + w
~ ρ~ ~
~ · ~~
N p d Ae  →
Ve Ve Ae

eT e e e T ~e
~
w u ~
~ =w
~ ·K ·~ ~ ·~ f
e

Integration

Calculating the element stiffness matrix K e and the external loads f e implies the evaluation
~e
of an integral over the element volume V e and the element surface A e . This integration is
done numerically, using a fixed set of ni p Gauss-points, which have s specific location in the
element. The value of the integrand is calculated in each Gauss-point and multiplied with a
Gauss-point-specific weighting factor c i p and added.

Z Z1 Z1 Z1 ni
Xp
e ip ip ip
g (x1 , x2 , x3 ) dV = f (ξ1 , ξ2 , ξ3 ) d ξ1 d ξ2 d ξ3 = c i p f (ξ1 , ξ2 , ξ3 )
i p=1
Ve ξ1 =−1 ξ2 =−1 ξ3 =−1

Z Z1 Z1 4
X
e ip ip
g (x1 , x2 ) dV = f (ξ1 , ξ2 ) d ξ1 d ξ2 = c i p f (ξ1 , ξ2 )
i p=1
Ve ξ1 =−1 ξ2 =−1

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 appro-
priate locations into the structural or global stiffness matrix K and the load column f .
~e
Because the resulting equation has to be satisfied for all w ~ , the nodal displacements u
~
have to satisfy a set of equations.

145
7. M ECHANICAL DESIGN : SOLUTIONS

3
4 ξ2
4
3 p
1
3 3

1 2 ξ1
1 2 1
p
3 3

Figure 7.7: Integration points in 4-node quadrilateral element

e T ~e
X T
e e e
X T T ~
~
w u
~ ·K ·~
~ = ~
w
~ ·~ f → ~
w u ~
~=w
~ ·K ·~ ~ ·~f ~
∀w
~ →
e e
e e

K ·~
u ~
f ~
u −1 ~
~=~ e
→ ~ = K ·~f
e

Boundary conditions

The initial governing equations were differential 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 finite element discretisation, displacements
and forces can be applied in nodal points.
The set of nodal equations K u
~=~ f cannot be solved yet, because the structural stiffness
e
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.

146
 CHAPTER
8
M ECHANICAL DESIGN PROBLEMS :
ANALYTICAL SOLUTIONS

In this chapter, a number of classical mechanical design problems are presented, which al-
low for an analytical solution. The equations are presented and the solution is given without
extensive derivations. Many problems involve the calculation of integration constants from
boundary conditions. For such problems these integration constants can be found in ap-
pendix B. Examples with numerical values for parameters, are presented. More examples
can be found in the above-mentioned appendix.

8.1 Homogeneous plates

For planar problems in a Cartesian coordinate system, two partial differential equations for
the displacement components u x and u y , the so-called Navier equations, have to be solved,
using specific boundary conditions. Only for very simply cases, this can be done analyti-
cally. For practical problems, approximate solutions have to be determined with numerical
solution procedures. The Navier equations have been derived in section 7.6.1 and are re-
peated below for the static case, where no material acceleration is considered. The material
parameters A p , B p , Q p and K have to be specified for plane stress or plane strain and for the
material model concerned.

A p u x,xx + K u x,y y + (Q p + K )u y,yx + ρq x = 0

K u y,xx + B p u y,y y + (Q p + K )u x,x y + ρq y = 0

Tensile test

When a square plate (length a) of homogeneous material is loaded uniaxially by a uniform

tensile edge load p, this load constitutes an equilibrium system, i.e. the stresses satisfy the
equilibrium equations : σxx = p and σ y y = σx y = σzz = 0. The deformation can be calculated
directly from Hooke’s law.

147
8. M ECHANICAL DESIGN : ANALYTICAL

y y
σxx = p σxx = p

a x x
a

Figure 8.1: Uniaxial tensile test

1 p p
εxx = σxx = → u x = x + c ; u x (x = 0) = 0 → c = 0
E E E
p p
u x = x → u x (x = a) = a
E E
p p
ε y y = −νεxx = −ν → u y = −ν y + c ; u y (y = 0) = 0 → c =0
E E
p pa
u y = −ν y ; u y (y = a/2) = −ν
E E2

8.2 Axisymmetric cylinders and discs

The differential equation for the radial displacement u r is derived in chapter 7 by substitu-
tion of the stress-strain relation (material law) and the strain-displacement relation in the
equilibrium equation w.r.t. the radial direction. It is repeated here for isotropic material be-
havior. Material parameters A p and Q p have to be specified for plane stress and plane strain.

1 1
u r,r r + u r,r − 2 u r = f (r )
r r
ρ ¡ ¢
with f (r ) = ü r − q r
Ap

A general solution for the differential equation can be determined as the addition of the
homogeneous solution û r and the particular solution ū r , which depends on the specific
loading f (r ). From the general solution the radial and tangential strains can be calculated
according to their definitions.

û r = r λ → û r,r = λ r λ−1 → û r,r r = λ(λ − 1) r λ−2 →

λ−2
[λ(λ − 1) + λ − 1] r =0 →
2
λ =1 → λ = ±1 → û r = c1 r + c2 r −1

148
 8.2. Axisymmetric cylinders and discs

The general solution for an isotropic material then becomes

c2
general solution u r = c1 r + + ū r
r

εr r = c1 − c2 r −2 + ū r,r
ū r
εt t = c1 + c2 r −2 +
r

c2 ū r
σr r = (A p +Q p )c1 − (A p −Q p ) + A p ū r,r +Q p
r2 r
c2 ū r
σt t = (Q p + A p )c1 − (Q p − A p ) 2 +Q p ū r,r + A p
r r
When there is no right-hand loading term f (r ) in the differential equation, the particulate
part ū r will be zero. Then, for isotropic material, the radial and tangential strains are uni-
form, i.e. no function of the radius r . For a state of plane stress, the axial strain is calculated
as a weighted summation of the in-plane strains, so also εzz will be uniform. The thickness
of the axisymmetric object will remain uniform. For non-isotropic material behavior this is
not the case, however.

Loading and boundary conditions

In the following subsections, different geometries and loading conditions will be considered.
The external load determines the right-hand side f (r ) of the differential equation and as a
consequence the particulate part ū r of the general solution. Boundary conditions must be
used subsequently to determine the integration constants c1 and c2 . Finally the parameters
A p , B p and Q p must be chosen in accordance with the material behavior and specified for
plane stress or plane strain.
The algebra, which is involved with these calculations, is not very difficult, but rather
cumbersome. In appendix B a number of examples is presented. When numerical values
are provided, displacements, strains and stresses can be calculated and plotted with a Matlab
program, which is available on the website of this course. Based on the input, it selects the
proper formulas for the calculation. Instructions for its use can be found in the program
source file. The figures in the next subsections are made with this program.

8.2.1 Prescribed edge displacement

The outer edge of a disc with a central hole is given a prescribed displacement u(r = b) = u b .
The inner edge is stress-free. With these boundary conditions, the integration constants in
the general solution can be determined. They can be found in appendix B.
For the parameter values listed below, the radial displacement u r and the stresses are
calculated and plotted as a function of the radius r .

| u b = 0.01 m | a = 0.25 m | b = 0.5 m | h = 0.05 m | E = 250 GPa | ν = 0.33 |

8.2.2 Edge load

A cylinder has inner radius r = a and outer radius r = b. It is loaded with an internal (p i )
and/or an external (p e ) pressure.

149
8. M ECHANICAL DESIGN : ANALYTICAL

r
ub
f (r ) = 0 → ū r = 0
b
a u r (r = b) = u b
z r
σr r (r = a) = 0
c1 , c2 : App. B

Figure 8.2: Edge displacement of circular disc

9
−3
x 10 x 10
10 10
σ
rr
σ
tt
9.9 8 σ
zz

9.8
6

σ [Pa]
u [m]

9.7
r

4
9.6

2
9.5

9.4 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.3: Displacement and stresses for plane stress (σzz = 0)

The general solution to the equilibrium equation has two integration constants, which
have to be determined from boundary conditions. In appendix B they are determined for the
case that an open cylinder is subjected to an internal pressure p i and an external pressure p e .
For plane stress (σzz = 0) the cylinder is free to deform in axial direction. The solution was
first derived by Lamé in 1833 and therefore this solution is referred to as Lamé’s equations.
When these integration constants for isotropic material are substituted in the stress so-
lution, it appears that the stresses are independent of the material parameters. This implies
that radial and tangential stresses are the same for plane stress and plane strain. For the
plane strain case, the axial stress σzz can be calculated directly from the radial and tangen-
tial stresses.
The Tresca and Von Mises limit criteria for a pressurized cylinder can be calculated ac-
cording to their definitions (see chapter 6).

p i a 2 − p e b 2 a 2 b 2 (p i − p e ) 1 p i a 2 − p e b 2 a 2 b 2 (p i − p e ) 1
σr r = − ; σt t = +
b2 − a2 b2 − a2 r2 b2 − a2 b2 − a2 r2

σT R = 2τmax = max [ |σr r − σt t |, |σt t − σzz |, |σzz − σr r | ]

q © ª
σV M = 12 (σr r − σt t )2 + (σt t − σzz )2 + (σzz − σr r )2

150
 8.2. Axisymmetric cylinders and discs

r
pe
f (r ) = 0 → ū r = 0
b
a σr r (r = a) = −p i
z r
pi σr r (r = b) = −p e
c1 , c2 : App. B

Figure 8.4: Cross-section of a thick-walled circular cylinder

Open cylinder

An open cylinder is analyzed with the parameters from the table below. Stresses are plotted
as a function of the radius.

| p i = 100 MPa | a = 0.25 m | b = 0.5 m | h = 0.5 m | E = 250 GPa | ν = 0.33 |

8
8
x 10 x 10
2 3
σ σ
rr
TR

σ σVM
1.5 tt
2.5
σzz

1
2
σ [Pa]
σ [Pa]

0.5
1.5
0

1
−0.5

−1 0.5
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.5: Stresses in a thick-walled pressurized cylinder for plane stress (σzz = 0)

That the inner material is under much higher tangential stress than the outer material,
can be derived by reasoning, when we only consider an internal pressure. This pressure
will result in enlargement of the diameter for each value of r , but it will also compress the
material and result in reduction of the wall thickness. The inner diameter will thus increase
more than the outer diameter – which is also calculated and plotted in figure 8.6 – and the
tangential stress will be much higher at the inner edge.

Closed cylinder

A closed cylinder is loaded in axial direction by the internal and the external pressure. This
load leads to an axial stress σzz , which is uniform over the wall thickness. It can be deter-
mined from axial equilibrium and can be considered as an The radial and tangential stress
are not influenced by this axial load.

151
8. M ECHANICAL DESIGN : ANALYTICAL

−4
x 10
2
pss
1.9 psn

1.8

1.7

1.6

u [m]
1.5

1.4

1.3

1.2

1.1
0 0.1 0.2 0.3 0.4 0.5
r [m]

Figure 8.6: Radial displacement in a thick-walled pressurized cylinder for plane stress (pss)
and plane strain (psn)

The resulting radial displacement due to the contraction caused by the axial load, u r a ,
can be calculated from Hooke’s law.

pi a2 − pe b2 ν
axial equilibrium σzz = → u r a = εt t a r = − σzz r
b2 − a2 E

8.3 Circular hole in infinite medium

When a circular hole is located in an infinite medium, we can derive the stresses from Lame’s
equations by taking b → ∞. For the general case this leads to limit values of the integration
constants c1 and c2 . These can then be substituted in the general solutions for displacement
and stresses.

8.3.1 Pressurized hole in an infinite medium

For a pressurized hole in an infinite medium the external pressure p e is zero. In that case the
absolute values of radial and tangential stresses are equal. The radial displacement can also
be calculated.

pa 2 pa 2
b→∞ ; pi = p ; pe = 0 → σr r =− 2 ; σt t = 2
r r
For a plane stress state and with parameter values listed below, the radial displacement
and the stresses are calculated and plotted as a function of the radius in figure 8.7. Note that
we take a large but finite value of for b.

| p i = 100 MPa | a = 0.2 m | b = 20 m | h = 0.5 m | E = 200 GPa | ν = 0.3 |

8.3.2 Stress-free hole in bi-axially loaded infinite medium

We consider the case of a stress-free hole of radius a in an infinite medium, which is bi-
axially loaded at infinity by a uniform load T , equal in x- and y-direction. Because the load

152
 8.3. Circular hole in infinite medium

−4
x 10 8
1.4 x 10
1.5
σrr
1.2 σ
1 tt
σzz
1
0.5
0.8

σ [Pa]
u [m]

0
0.6

−0.5
0.4

0.2 −1

0 −1.5
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
r [m] r [m]

Figure 8.7: Displacement and stresses in a pressurized circular hole in an infinite medium

is applied at boundaries which are at infinite distance from the hole center, the bi-axial load
is equivalent to an externally applied radial edge load p e = −T .
Radial and tangential stresses are different in this case. The tangential stress is maximum
for r = a and equals 2T .
µ
¶ µ ¶
a2 a2
b → ∞ ; p i = 0 ; p e = −T → σr r = T 1 − 2 ; σt t = T 1 + 2
r r

σmax σt t (r = a) 2T
stress concentration factor = Kt = = =2
T T T
For a plane stress state and with parameter values listed below, the radial displacement
and the stresses are calculated and plotted as a function of the radius in figure 8.8.

| p e = −100 MPa | a = 0.2 m | b = 20 m | h = 0.5 m | E = 200 GPa | ν = 0.3 |

−3
x 10 8
x 10
2.5
σrr

1 σtt
2
σ
zz
0.8 1.5
σ [Pa]
u [m]

0.6 1

0.4 0.5

0.2 0

0 −0.5
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
r [m] r [m]

Figure 8.8: Displacement and stresses in a stress-free hole in a biaxially loaded infinite
medium

153
8. M ECHANICAL DESIGN : ANALYTICAL

8.4 Rotating disc

A circular disc, made of isotropic material, rotates with angular velocity ω [rad/s]. The outer
radius of the disc is taken to be b. The disc may have a central circular hole with radius a.
When a = 0 there is no hole and the disc is called ”solid”. Boundary conditions for a disc
with a central hole are rather different than those for a ”solid” disc, which results in different
solutions for radial displacement and stresses. The external load f (r ) is the result of the
radial acceleration of the material (see sections 8.2
ρ ρ 2
ü r = −ω2 r → f (r ) = ü r = − ω r
Ap Ap
The particular solution ū r can be computed to be
1 1
ū r = − ρω2 r 3
8 Ap

8.4.1 Solid disc

In a disc without a central hole (solid disc) there are material points at radius r = 0. To pre-

r
ω
u r (r = 0) 6= ∞
b
σr r (r = b) = 0
z r
c1 , c2 : App. B

Figure 8.9: A rotating solid disc

vent infinite displacements for r →0 the second integration constant c2 must be zero. At the
outer edge the radial stress σr r must be zero, because this edge is unloaded. With these
boundary conditions the integration constants in the general solution can be calculated (see
appendix B).
For a plane stress state and with the listed parameter values, the stresses are calculated
and plotted as a function of the radius.

| ω = 6 c/s | a = 0 m | b = 0.5 m | t = 0.05 m | ρ = 7500 kg/m3 |

| E = 200 GPa | ν = 0.3 |
In a rotating solid disc the radial and tangential stresses are equal in the center of the
disc. They both decrease with increasing radius, where of course the radial stress reduces to
zero at the outer radius. The equivalent Tresca and Von Mises stresses are not very different
and also decrease with increasing radius. In the example the disc is assumed to be in a state
of plane stress.
When the same disc is fixed between two rigid plates, a plane strain state must be mod-
elled. In that case the axial stress is not zero. As can be seen in figure 8.11, the axial stress
influences the Tresca and Von Mises equivalent stresses.

154
 8.4. Rotating disc

5 5
x 10 x 10
12 11
σrr σTR

10 σ 10 σVM
tt
σ
zz
8 9

6 8
σ [Pa]

σ [Pa]
4 7

2 6

0 5

−2 4
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.10: Stresses in a rotating solid disc in plane stress

5 5
x 10 x 10
12 4.6
σ σ
rr TR
σ 4.4 σ
10 tt VM
σ 4.2
zz

8 4
σ [Pa]

σ [Pa]

3.8
6
3.6

4 3.4

3.2
2
3

0 2.8
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.11: Stresses in a rotating solid disc in plane strain

8.4.2 Disc with central hole

When the disc has a central circular hole, the radial stress at the inner edge and at the outer
edge must both be zero, which provides two equations to solve the two integration constants.
For a plane stress state and with parameter values listed below, the radial displacement
and the stresses are calculated and plotted as a function of the radius in figure 8.13.

| ω = 6 c/s | a = 0.2 m | b = 0.5 m | t = 0.05 m | ρ = 7500 kg/m3 |

| E = 200 GPa | ν = 0.3 |

8.4.3 Disc fixed on rigid axis

When the disc is fixed on an axis and the axis is assumed to be rigid (see figure 8.14), the
displacement of the inner edge is suppressed. The radial stress at the outer edge is obviously
zero.

155
8. M ECHANICAL DESIGN : ANALYTICAL

r
ω
σr r (r = a) = 0
b
a σr r (r = b) = 0
z r
c1 , c2 : App. B

Figure 8.12: A rotating disc with central hole

6
−6 x 10
x 10 2.5
2.35 σrr
σ
2.3 tt
2

2.25
1.5
σ [Pa]

2.2
ur [m]

2.15 1

2.1
0.5
2.05

2 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.13: Displacement and stresses in a rotating disc with a central hole

r
ω
u r (r = a) = 0
b
a σr r (r = b) = 0
z r
c1 , c2 : App. B

Figure 8.14: Disc fixed on rigid axis

156
 8.4. Rotating disc

For a plane stress state and with parameter values listed below the stresses and the radial
displacement are calculated and plotted in figures 8.15 and 8.16.

| ω = 6 c/s | a = 0.2 m | b = 0.5 m | t = 0.05 m | ρ = 7500 kg/m3 | E = 200 GPa | ν = 0.3 |

5 5
x 10 x 10
14 14
σ σ
rr TR
12 σ σ
tt 12 VM

10
10
8
σ [Pa]

σ [Pa]
8
6
6
4

2 4

0 2
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.15: Stresses in a rotating disc, fixed on an axis

−7
x 10
8

5
ur [m]

0
0 0.1 0.2 0.3 0.4 0.5
r [m]

Figure 8.16: Displacement in a rotating disc

8.4.4 Rotating disc with variable thickness

For a rotating disc with variable thickness t (r ) the equation of motion in radial direction can
be derived. For a disc with inner and outer radius a and b, respectively, and a thickness
distribution t (r ) = t2a ar , a general solution for the stresses can be derived. The integration
constants can be determined from the boundary conditions, e.g. σr r (r = a) = σr r (r = b) = 0.

157
8. M ECHANICAL DESIGN : ANALYTICAL

∂(t (r )r σr r ) ta a
equilibrium − t (r )σt t = −ρω2 t (r )r 2 with t (r ) =
∂r 2 r
general solution stresses
2c1 d1 2c2 d2 3 + ν 2 2 2c1 2c2 1 + 3ν 2 2
σr r = r + r − ρω r ; σt t = d 1 r d1 + d 2 r d2 − ρω r
at a at a 5+ν at a at a 5+ν
q q
1 5
with d1 = − 2 + 4 + ν ; d 2 = − 2 − 54 + ν
1

boundary conditions σr r (r = a) = σr r (r = b) = 0 →
· µ 2 ¶¸
2c1 3+ν 2 −d1 2 d2 b − a −d1 b d1 a 2
= ρω a a −a
at a 5+ν b d2 − a d2 a −d1 b d1
µ 2 ¶
2c2 3+ν 2 b − a −d1 b d1 a 2
= ρω
at a 5+ν b d2 − a d2 a −d1 b d1
A disc with a central hole and a variable thickness rotates with an angular velocity of 6
cycles per second. The stresses are plotted as a function of the radius in figure 8.17.

| isotropic | plane stress | ω = 6 c/s |

| a = 0.2 m | b = 0.5 m | t a = 0.05 m | ρ = 7500 kg/m3 | E = 200 GPa | ν = 0.3 |

6
x 10
5 1.8
18
x 10 σTR
σ
rr σVM
16 σ 1.6
tt
14 σ
zz
1.4
12
σ [Pa]
σ [Pa]

10 1.2
8

6 1

4
0.8
2

0 0.6
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.17: Stresses in a rotating disc with variable thickness

8.5 Large thin plate with a central hole

A large rectangular plate is loaded with a uniform stress σxx = σ. In the center of the plate is
a hole with radius a, much smaller than the dimensions of the plate.
The stresses around the hole can be determined, using an Airy stress function approach.
The relevant stresses are expressed as components in a cylindrical coordinate system, with
coordinates r , measured from the center of the hole, and θ in the circumferential direction.

158
 8.5. Large thin plate with a central hole

σ σ
r
θ x

Figure 8.18: Large thin plate with a central hole

·µ ¶ µ ¶ ¸
σ a2 a4 a2
σr r = 1 − 2 + 1 + 3 4 − 4 2 cos(2θ)
2 r r r
·µ 2¶ µ 4¶ ¸
σ a a
σt t = 1 + 2 − 1 + 3 4 cos(2θ)
2 r r
· 4 2¸
σ a a
σr t =− 1 − 3 4 + 2 2 sin(2θ)
2 r r

At the inner edge of the hole, the tangential stress reaches a maximum value of 3σ for
θ = 90o . For θ = 0o a compressive tangential stress occurs. The stress concentration factor
K t is independent of material parameters and the hole diameter.
At a large distance from the hole, so for r ≫ a, the stress components are a function of
the angle θ only.

stress concentrations

σt t (r = a, θ = π2 ) = 3σ σt t (r = a, θ = 0) = −σ
;
σmax
stress concentration factor Kt = =3
σ
stress at larger r
σ
σr r = [1 + cos(2θ)] = σ cos2 (θ)
2
σ £ ¤
σt t = [1 − cos(2θ)] = σ 1 − cos2 (θ) = σ sin2 (θ)
2
σ
σr t = − sin(2θ) = −σ sin(θ) cos(θ)
2
For parameters values listed below stress components are calculated and plotted for θ = 0
and for θ = π2 as a function of the radial distance r .

| a = 0.05 m | σ = 1000 Pa |

159
8. M ECHANICAL DESIGN : ANALYTICAL

1000 3000
σrr
σtt
2500
σzz
500
2000
σ [Pa]
σ [Pa]

0 1500

1000
−500 σrr
σtt 500
σzz
−1000 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 8.19: Stresses in plate for θ = 0 and θ = π2

160
 CHAPTER
9
M ECHANICAL DESIGN PROBLEMS :
NUMERICAL SOLUTIONS

In this chapter, the numerical solution of some selected problems is presented. These solu-
tions are determined with the MSC.Marc/Mentat FE-package. The numerical solutions can
be compared with the analytical solutions, described in the previous section.

9.1 MSC.Marc/Mentat
The MSC.Mentat program is used to model the structure, which is subsequently analyzed
by the FE-program MSC.Marc. Modeling the geometry – shape and dimensions – is the
first step in this procedure. Dimensional units have to be chosen and consistently used in
the entire analysis. In this stage it is already needed to decide whether the model is three-
dimensional, planar or axisymmetric. The finite element mesh is generated according to
procedures which are described in the tutorial. In the examples discussed in this chap-
ter, only linear elements are used, i.e. elements where the displacement is interpolated bi-
linearly between the nodal point displacements. Quadratic elements will lead to more accu-
rate results in most cases.
After defining the geometry, the material properties can be specified. Only linear elastic
material behavior is considered, both isotropic and orthotropic. Boundary conditions can
be : prescribed displacements, edge loads, gravitational loads and centrifugal loading due to
rotation. Thermal loading is not shown here, but can be applied straightforwardly.
When the model is complete, it can be analyzed and the results can be observed and
plotted. Contour bands of variables can be superposed on the geometry and variables can
be plotted. In the next sections these plots will be presented.

9.2 Cartesian, planar

The most simple analysis, which can be done is the calculation of stress and strain in a ten-
sile test and a shear test. Boundary conditions must be prescribed to ensure homogeneous

161
9. M ECHANICAL DESIGN : NUMERICAL

deformation. As expected, the exact solution is reproduced with only one element.

9.2.1 Tensile test

A uniaxial tensile test, resulting in homogeneous deformation, can be modeled and analyzed
with only one linear element, resulting in the exact solution. In the next example a square

y y

p p

x x

Figure 9.1: Tensile test with different boundary conditions

plate (length a, thickness h) is modeled with four equally sized elements, because the central
node on the left edge must be fixed to prevent rigid body movement. The right-hand edge is
loaded with an edge load p. The deformation is shown in figure 9.2 with a magnification of
500.
Inc: 0 Inc: 0
Time: 0.000e+00 Time: 0.000e+00

Y Y

Z X Z X

job1 job1

1 1

Figure 9.2: Undeformed and deformed element mesh at 500 × magnification

When the left edge is clamped, the deformation and stress state is no longer homoge-
neous. An analytical solution does not exist for this case. An approximate solution can be de-
termined rather easily. To model the inhomogeneous deformation, we need more elements,
especially in the neighborhood of the clamped edge. Using more elements improves the ac-
curacy of the result. Equal accuracy can be realized with fewer, but higher-order (quadratic)
elements, as such elements interpolate the displacement field better than a linear element.
When subsequent analyses are done with decreasing element sizes, we will notice that
at some point, further mesh refinement will not change the solution any more. This con-
vergence upon mesh refinement is essential for good finite element modeling and analysis.
When it does not occur, the results are always dependent on the element mesh and such a
mesh dependency is not allowed. It may be found to occur when singularities are involved or
when the (non-linear) material shows softening, i.e. decrease of stress at increasing strain.

| p = 100 MPa | a = 0.5 m | h = 0.05 m | E = 200 GPa | ν = 0.25 |

162
 9.2. Cartesian, planar

u x (x = a) = 0.25 × 10−3 m ; u y (y = a/2) = −0.3125 × 10−4

9.2.2 Shear test

The true shear test is another example of a homogeneous deformation. It is done here with
a 20 × 20 mesh of square linear elements, but could have been done with only one element,
leading to the same exact result. To prevent rigid body translation, the left-bottom node

y y
p p

p p

x x
p

Figure 9.3: Shear test with different boundary conditions

is fixed, i.e. its displacement is prevented. To prevent rigid body rotation, the nodes at the
bottom are only allowed to move horizontally. Edges are loaded with a shear load p, leading
to the deformation, which is shown in figure 9.4, again with a magnification of 500.

Inc: 0 Inc: 0
Time: 0.000e+00 Time: 0.000e+00

Y Y

Z X Z X

job1 job1

1 1

Figure 9.4: Undeformed and deformed element mesh at 500 and 250 × magnification

Instead of this homogeneous shear test, often a so-called simple shear test is done exper-
imentally. In that case the shear load p is only applied at the upper edge. The left- and right-
edge is stress-free. Moreover, the displacement in y-direction of the upper edge is prevented,
as is the case for the bottom-edge, which is clamped. The result is shown in the right-hand
of figure 9.4 with a magnification of 250. It is immediately clear that the deformation is no
longer homogeneous.

| isotropic | plane stress | p = 100 MPa | a = 0.5 m | h = 0.05 m|

| E = 200 GPa | ν = 0.25 |

163
9. M ECHANICAL DESIGN : NUMERICAL

9.2.3 Orthotropic plate

The uniaxial tensile test and the real shear test are now carried out on a plate with orthotropic
material behavior. For the tensile test, the center node on the left-edge is fixed, while the
other nodes on this edge are restricted to move in y-direction. For the shear test, the center
node is fixed and again the upper-right and lower-left corner nodes are restricted to move
along the diagonal. The material coordinate system is oriented at an angle α w.r.t. the global
x-axis. Material parameters are listed in the table. Because a plane stress state is assumed,
the surface of the plate decreases and also the thickness will change.

y y
p

x p p

x
p

Figure 9.5: Tensile test and shear test for orthotropic plate

| orthotropic | plane stress | p = 100 MPa | a = 0.5 m | h = 0.05 m | α = 20o |

| E 11 = 200 GPa | E 22 = 50 GPa | E 33 = 50 GPa|
| ν12 = 0.4 | ν23 = 0.25 | ν31 = 0.25|
| G 12 = 100 GPa | G 23 = 20 GPa | G 31 = 20 GPa |

Inc: 0 Inc: 0
Time: 0.000e+00 Time: 0.000e+00

Y Y

Z X Z X

job1 job1

1 1

Figure 9.6: Undeformed and deformed element mesh at 500 × magnification

9.3 Axisymmetric, u t = 0
A tensile test on a cylindrical bar can be analyzed analytically when the material is isotropic.
For orthotropic material, with principal material directions in radial, axial and tangential di-
rection, the problem is analyzed numerically. Although the loading is uniaxially and uniform
over the cross-section, the strain and stress distribution is not homogeneous.
The cylindrical tensile bar of length 0.5 m is modelled with axisymmetric elements. The
radius of the bar is 0.0892 m. The material coordinate system is {1, 2, 3} = {r, z, t }. Material

164
 9.4. Axisymmetric, planar, u t = 0

parameters are listed in the table. The bar is loaded with an axial edge load p. The axial
displacement is then about 0.001 m. The figure shows the stresses as a function of the radius.

| E 11 = 200 GPa | E 22 = 50 GPa | E 33 = 50 GPa|

| ν12 = 0.4 | ν23 = 0.25 | ν31 = 0.25|
| G 12 = 100 GPa | G 23 = 20 GPa | G 31 = 20 GPa | p = 100 MPa |

z 7
p x 10
1.5

0.5
σ [Pa]

−0.5
σrr
−1 σtt
r σzz
−1.5
0 0.02 0.04 0.06 0.08
r [m]

Figure 9.7: Stresses in an orthotropic tensile bar

9.4 Axisymmetric, planar, u t = 0

Axisymmetric problems can be analyzed with planar elements – plane stress or plane strain
– but also with axisymmetric elements. In the latter case, the model is made in the zr -plane
for r > 0.

9.4.1 Prescribed edge displacement

The model is made in the zr -plane with axisymmetric elements, see figure 9.8 A plane stress
state is modeled by choosing the proper boundary conditions.

| u b = 0.01 m | a = 0.25 m | b = 0.5 m | h = 0.05 m | E = 250 GPa | ν = 0.33 |

Displacements and stresses are shown in figure 9.9.

9.4.2 Edge load

A thick-walled cylinder is loaded with an internal pressure p i , see figure 9.10. When the
cylinder is open and its elongation unconfined, each cross-section over the axis is in a state
of plane stress: σzz = 0.
The cylinder is open (plane stress) and made of isotropic material. Dimensions and ma-
terial properties are listed in the table. The plots in figure 9.11 show the stresses as a function
of the radius. There values coincide with the analytical solution, except near the edges. The

165
9. M ECHANICAL DESIGN : NUMERICAL

r
ub
b
a z r

Figure 9.8: Edge displacement of circular disc

9
−3
x 10 x 10
10 12
σrr
σ
10 tt
9.9
σzz

9.8 8
σ [Pa]
ur [m]

9.7 6

9.6 4

9.5 2

9.4 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 9.9: Displacement and stresses for plane stress (σzz = 0)

r
pe

b
a z r
pi

Figure 9.10: Cross-section of a thick-walled circular cylinder

166
 9.4. Axisymmetric, planar, u t = 0

reason is that stresses (and strains) are calculated in the integration points, which are lo-
cated inside the element, and edge values are extrapolated. When more elements are used,
the deviation will decrease.

| p i = 100 MPa | a = 0.25 m | b = 0.5 m | h = 0.5 m | E = 250 GPa | ν = 0.33 |

8
8
x 10 x 10
2 2.4
σrr σ
VM
2.2
σtt
1.5
σ 2
zz

1 1.8

σ [Pa]
1.6
σ [Pa]

0.5
1.4

0 1.2

1
−0.5
0.8

−1 0.6
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 9.11: Stresses in a thick-walled pressurized cylinder for plane stress (σzz = 0)

For plane strain (εzz = 0) the length of the cylinder is kept constant. This will obviously
lead to an axial stress σzz . The radial displacement in figure 9.12 is obviously larger for the
plane stress situation, because the material has more ”freedom” to deform in radial direc-
tion.
−4
x 10
2
pss
1.9 psn

1.8

1.7

1.6
u [m]

1.5

1.4

1.3

1.2

1.1
0 0.1 0.2 0.3 0.4 0.5
r [m]

Figure 9.12: Radial displacement in a thick-walled pressurized cylinder for plane stress (pss)
and plane strain (psn)

9.4.3 Centrifugal load

A centrifugal load is modeled to analyze rotating discs. The analysis can be done with an
axisymmetric model or with a plane stress model. Except for the location near the edges, the

167
9. M ECHANICAL DESIGN : NUMERICAL

numerical solution coincides with the analytical solution, presented in chapter 8.

9.4.4 Large thin plate with a central hole

A rectangular plate with central hole is loaded uniaxially. The analytical solution for the
stress distribution is known and given in chapter 8. To get a numerical solution, the plate
is modeled with linear elements. Geometric and material data are listed in the table. The
figures show the stresses as a function of the radial coordinate for the 0o - and 90o -direction.

σ σ
r
θ x

Figure 9.13: Large thin plate with a central hole

| a = 0.05 m | σ = 1000 Pa |

1000 3500
σrr
3000 σtt
500
2500

2000
0
σ [Pa]
σ [Pa]

1500
−500 1000

500
−1000 σrr
0
σ
tt
−1500 −500
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Figure 9.14: Stresses in plate for θ = 0 and θ = π2

168
 CHAPTER
10
L ARGE DEFORMATIONS AND NONLINEAR
MATERIAL BEHAVIOUR : BASIC CONCEPTS
AND NONLINEAR ELASTICITY

This chapter provides an outlook to the mechanics of large deformations and rotations, as
well complex nonlinear material behaviour. A one-dimensional truss-like description will be
used to explain basic concepts. The governing equations are scalar. The extension to three
dimensions is obviously tensorial, but beyond the scope of this bachelor course.

10.1 Nonlinear axial deformation of trusses

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 surroundings with hinges.
A truss can transfer only axial forces along its axis, so bending is not possible, and the axis
must remain straight.
When deformation and/or rotation of the truss are large, various strains and stresses can
be defined and related by material laws. The material behavior can be expected to be no
longer linearly elastic.

10.1.1 Strains for large elongation

The scalar nonlinear strain measures were already introduced in chapter 2. In this subsec-
tion, these strain measures will be elaborated for a truss. 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 stretch
λ. The change in cross-sectional area is described by the contraction µ. It is assumed that
the load, which provokes the deformation, is such that the deformation is homogeneous. For
a homogeneous truss, λ and µ are the same in each point of the truss. The volume change is
described by the volume ratio J .
The strain measures are expressed in terms of the stretch ration λ, see chapter 2.

169
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

Linear strain

The linear strain definition 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 pos-
sible.

∆l
linear strain ε = εl = λ − 1 =
l0
contraction strain εd = µ − 1 = −νεl = −ν(λ − 1)

change of cross-sectional area

s
A
µ= = 1 − ν(λ − 1) → A = A 0 {1 − ν(λ − 1)}2
A0

restriction of elongation

1 1+ν
1 − ν(λ − 1) > 0 → λ−1 < → λ<
ν ν

Logarithmic strain

The logarithmic strain definition does not lead to unrealistic values for the contraction. There-
fore it is very suitable to describe large deformations.

logarithmic strain ε = εl n = ln(λ)

contraction strain εd = ln(µ) = −νεl n = −ν ln λ

change of cross-sectional area

s
A h i−ν
µ= = e −νεl n = e −ν ln(λ) = e ln(λ) = λ−ν → A = A 0 λ−2ν
A0

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.

0 1 2

01 12

Figure 10.1: Two-step deformation process

170
 10.1. Nonlinear axial deformation of trusses

l 0 →l 1 εl (01) = l 1l−l
0
0

εl n (01) = ln( ll 1 )
0

l 1 →l 2 εl (12) = l 2l−l
1
1

εl n (12) = ln( ll 21 )

l 0 →l 2 εl (02) = l 2l−l 0 6= εl (01) + εl (12)

0
εl n (02) = ln( ll 20 ) = ln( ll 12 ll 10 ) = εl n (01) + εl n (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 ε = εg l = 21 (λ2 − 1)

contraction strain εd = 21 (µ2 − 1) = −νεl n = −ν 12 (λ2 − 1)

change of cross-sectional area

r
2 1+ν
1 − ν(λ − 1) > 0 → λ<
ν

10.1.2 Mechanical power and stress definitions

The figure shows a tensile bar which is elongated due to the action of an axial force F . At
constant force F an infinitesimal small increase in length is associated with a change in me-
chanical energy per unit of time (power) : P = F l˙. The elongation rate l˙ can be expressed in
various strain rates.

~
e2 l 0, A0

~
e3 ~
e1 l, A F

Figure 10.2: Axial elongation of homogeneous truss

l˙
linear strain εl = λ − 1 → ε̇l = λ̇ =
l0
l˙
logarithmic strain εl n = ln(λ) → ε̇l n = λ̇λ−1 =
l
l˙ l˙
Green-Lagrange strain εg l = 21 (λ2 − 1) → ε̇g l = λ̇λ = λ = λ2
l0 l

171
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

F F
P = F ℓ̇ = F ℓ0 ε̇l A 0 ℓ0 ε̇l =
= V0 ε̇l
A0 A0
F F
P = F ℓ̇ = F ℓε̇l n = Aℓε̇l n = V ε̇l n
A A
F ℓ0 F
P = F ℓ̇ = F ℓ0 ε̇l = Aℓ ε̇l = V λ−1 ε̇l
A ℓ A
F F
P = F ℓ̇ = F ℓλ−2 ε̇g l = Aℓλ−2 ε̇g l = V λ−2 ε̇g l
A A
Various stress definitions automatically emerge when the mechanical power is consid-
ered in the undeformed volume V0 = A 0 l 0 or the current volume V = Al of the tensile bar.
The stresses are :

σ : Cauchy or true stress

σn : engineering or nominal stress
σp1 : 1st Piola-Kirchhoff stress = σn
σκ : Kirchhoff stress
σp2 : 2nd Piola-Kirchhoff stress

The resulting mechanical power in terms of these stresses reads

P = = = V0 σn ε̇l
P = V σε̇l n = V0 (J σ)ε̇l n = V0 σκ ε̇l n
P = V (σλ−1 )ε̇l = V0 (J σλ−1 )ε̇l = V0 σp1 ε̇l
P = V (σλ−2 )ε̇g l = V0 (J σλ−2 )ε̇g l = V0 σp2 ε̇g l
specific mechanical power : P = V0Ẇ0 = V Ẇ

Ẇ0 = σn ε̇l = σκ ε̇l n = σp1 ε̇l = σp2 ε̇g l

Ẇ = = σε̇l n = σλ−1 ε̇l = σλ−2 ε̇g l

10.1.3 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 f i is a nonlinear function of the
elongation u.
Nonlinear material behavior may also result in a nonlinear function f i (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.

external force fe
internal force f i = σA = f i (u)
equilibrium of point P f i (u) = f e

172
 10.1. Nonlinear axial deformation of trusses

f i (u)

fe

u exact u

Figure 10.3: Nonlinear internal load and constant external load

Iterative solution procedure

It is assumed that an approximate solution u ∗ for the unknown exact solution u exact exists.
(Initially u ∗ = 0 is chosen.)
The residual load r ∗ is the difference between f (u ∗ ) and f e . 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).

f i (u)

fe
∗
r

f i∗

u∗ u exact u

Figure 10.4: Approximation of exact solution

analytical solution f i (u exact ) = f e → f e − f i (u exact ) = 0

approximation u ∗ f e − f i (u ∗ ) = r (u ∗ ) 6= 0
residual r ∗ = r (u ∗ )

The unknown exact solution is written as the sum of the approximation and an unknown
error δu. The internal force f i (u exact ) is then written as a Taylor series expansion around
u ∗ and linearized with respect to δu. The first derivative of f i with respect to u is called the
tangential stiffness K ∗ . Subsequently δu is solved from the linear iterative equation. The
solution is called the iterative displacement.

173
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

f i (u)

K∗ fe

r∗

f i∗
δu

u∗ u

Figure 10.5: Tangential stiffness and iterative solution

¾
f i (u exact ) = f e
→ f i (u ∗ + δu) = f e
u exact = u ∗ + δu
¯
∗ d f i ¯¯
f i (u ) + δu = f e → f i∗ + K ∗ δu = f e
d u ¯u∗
1 ∗
K ∗ δu = f e − f i∗ = r ∗ → δu = r
K∗

With the iterative displacement δu 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 calcu-
late the deviation of the approximation directly. There are several methods to quantify the
convergence.

f i (u)

K∗ fe

r∗

f i∗
δu

u∗ u ∗∗ u exact u

Figure 10.6: New approximation of the exact solution

new approximation u ∗∗ = u ∗ + δu
error u exact − u ∗∗
error smaller → convergence

174
 10.1. Nonlinear axial deformation of trusses

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 satisfied 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 δu with a
convergence criterion cu . If δu < cu it is assumed that the exact solution is determined close
enough.
When the convergence criterion is satisfied, 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 difference between f i and f e is added to the load in the next incre-
ment, which is known as residual load correction.

f i (u)

r ∗∗ fe

f i∗∗

δu

u∗ u ∗∗ u

Figure 10.7: New residual for approximate solution

residual force |r ∗∗ | ≤ cr → stop iteration

iterative displacement |δu| ≤ cu → stop iteration

f i (u)
fe

Figure 10.8: Converging iteration process

175
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

Residual and tangential stiffness

The residual and the tangential stiffness 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 σ and the elongation factor or stretch ratio λ = ll0 : σ = σ(λ). It is also necessary to now
the relation between the cross-sectional area A and λ.
internal nodal force f i∗ = N (λ∗ ) = A ∗ σ∗
¯ ¯ ¯
∗ ∂ f i ¯¯ ∂N (λ) ¯¯ d N ¯¯ d λ
tangential stiffness K = = =
∂u ¯u∗ ∂u ¯u∗ d λ ¯λ∗ d u
∆l 1 dλ 1
geometry λ = 1+ = 1+ u → =
l0 l0 d u l0
¯ ¯ ¯ ¯∗
∗ d N ¯¯ ∂λ d N ¯¯ 1 d N ¯¯∗ 1 1 d ¯
tangential stiffness K = ¯ = ¯ = ¯ = (σA)¯¯
d λ ∗ ∂u
λ dλ ∗ l
λ 0 dλ l 0l dλ
0

¯ ¯
∗ 1 d σ ¯¯∗ ∗ 1 ∗ d A ¯¯∗
K = A + σ
l0 d λ ¯ l0 dλ ¯

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, see figure 10.9. A time interval between two

fi
fe f fe

∆ fe

0 t un u n+1 u
tn ∆t tn+1

Figure 10.9: Incremental loading

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 fictitious or virtual time.

176
 10.2. Nonlinear material models

Non-converging solution process

The iteration process is not always converging. Some illustrative examples are shown in
figure 10.10.

fe
f i (u)

fe

f i (u)

fe
f i (u)

Figure 10.10: Non-converging solution processes

Modified Newton-Raphson procedure

Sometimes, it is possible to reach the exact solution by modifying the Newton-Raphson iter-
ation process. The tangential stiffness is then not updated in every iteration step. Its initial
value is used throughout the iterative procedure.
Figure 10.11 shows a so-called ”snap-through” problem, where no convergence can be
reached due to a cycling full Newton-Raphson iteration process. With modified Newton-
Raphson, iteration proceeds to the equilibrium f i = f e .

10.2 Nonlinear material models

10.2.1 Material behaviour
The 1D material behaviour is modeled, using a discrete mechanical model of springs, dash-
pots and friction sliders. The axial stress is related to the axial strain by one or more (dif-

177
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

fe

f i (u)

fe

f i (u)

Figure 10.11: Modified Newton-Raphson procedure

ferential) equation(s) from which the stress response must be calculated when the strain
excitation is prescribed.
The various material models are incorporated in a finite 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 stiffness plays
an essential role and must be derived from the material law.
Characterization of the mechanical behavior of an unknown material almost always be-
gins with performing a tensile experiment. A stepwise change in the axial stress σ may be
prescribed and the strain ε 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 behav-
ior.
For elastic material behavior the strain follows the stress immediately and becomes zero
after stress release. For elasto-plastic material behavior the strain also follows the stress im-
mediately, 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 defor-
mation, the behavior is referred to as viscoplastic.
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. Different behavior in tensile and compression may
be observed.

178
 10.2. Nonlinear material models

σ ε

t1 t2 t t1 t2 t

εe
σ ε
εp

t1 t2 t t1 t2 t

σ ε

t1 t2 t t1 t2 t

σ ε

t1 t2 t t1 t2 t

Figure 10.12: Strain response for a stress-step for a) elastic b) elasto-plastic, c) viscoelastic
and d ) viscoplastic material behavior

Tensile curve : elastic behavior

When elastic behavior is well described by a linear relation between stress and strain, the
elastic behavior is referred to as linear.

σ
σ σ σ

ε ε ε
ε

Figure 10.13: Tensile curves for elastic material behavior

Tensile curve : viscoelastic behavior

Viscoelastic behavior is time-dependent. The stress is a function of the strain rate. There
is a phase difference between stress and strain, which results in a hysteresis loop when the
loading is periodic.

179
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

σ
σ σ0

ε̇

ε0 ε

Figure 10.14: Tensile curve and hysteresis loop for viscoelastic material behavior

Tensile curve : elasto-plastic 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
referred to as elasto-plastic. Stress-strain curves may indicate different characteristics, espe-
cially when the loading is reversed from tensile to compressive.

σ σ

ε ε

σ σ

ε ε

Figure 10.15: Tensile curves for elasto-plastic material behavior

Tensile curve : viscoplastic behavior

A combination of plasticity and time-dependency is called viscoplastic behavior. This be-

havior 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.

180
 10.2. Nonlinear material models

Figure 10.16: Tensile curves for viscoplastic material behavior

Figure 10.17: Tensile curve for damaging material with different behavior in tension and
compression

Tensile curve : damage

Structural damage influences 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 different for tensile and compressive
loading.

10.2.2 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 cou-
pled (differential) equations. In this course, nonlinear material models will be presented in
a one-dimensional setting only. 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 σ and the axial strain ε can be derived.

181
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

σ
ε

σ
ε

σ
ε

Figure 10.18: Discrete elements : spring, dashpot and friction slider

10.3 Large deformation elasticity

10.3.1 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 energy dissipation. For the one-dimensional case
of an axially loaded truss the elastic behavior is described by a relation between the stress σ
and the elongation factor λ or the strain ε.

loading

unloading
0
1 λ

Figure 10.19: Non-linear elastic material behavior

Small strain elastic behavior

For small elongations, all strain definitions are the same, as are all stress definitions. The re-
lation between stress and strain is linear and the constant material parameter is the Young’s

182
 10.3. Large deformation elasticity

modulus.

strain ε = εg l = εl n = εl = λ − 1

F F
stress σ= = = σn
A A0
linear elastic behavior σ = E ε = E (λ − 1)

dσ dσ
modulus E = lim = lim
λ→1 d λ ε→ 0 dε

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 λ. 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 defini-
tion is associated with a certain strain definition, 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 rota-
tions of the material, which is known as the requirement of objectivity.

σ σ

0 1 λ 0 1 λ

Figure 10.20: Non-linear stress-strain relations for an elastomeric material and for an atomic
bond

Elasticity models

The discrete one-dimensional model for elastic material behavior is a spring. The behavior
is modeled with a relation between the stress σ and the elongation factor λ or a strain ε. The
material stiffness C λ is the derivative of σ w.r.t. the stretch ratio λ. The derivative w.r.t. the
strain ε results in the stiffness C ε .

183
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

Consideration of the stored elastic energy per unit of material volume (see 10.1.2) learns
that, in a material model, true stress σ should be combined with logarithmic strain εl n , engi-
neering stress σn with linear strain εl or 2nd-Piola-Kirchhoff stress σp2 with Green-Lagrange
strain εg l . Experimentally observed tensile behavior can often be described with a linear re-
lation between a certain stress and its associated strain.

σ
ε

Figure 10.21: Spring

constitutive equation σ = σ(λ)

dσ dσ dε dε
stiffness Cλ = = = Cε
dλ dε dλ dλ

10.3.2 Hyper-elastic models

Elastomeric materials (rubbers) show very large elastic deformations (elongation up to λ =
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 three-dimensional versions of these so-called Rivlin or Mooney models are expressed in
the principal elongation factors λi , i = 1, 2, 3. Experimental observations indicate that elas-
tomeric materials are incompressible, so that we have λ1 λ2 λ3 = 1.
n X
X m
W= C i j (I 1 − 3)i (I 2 − 3) j with C 00 = 0
i j

I 1 = λ21 + λ22 + λ23

1 1 1
I 2 = λ21 λ22 + λ22 λ23 + λ23 λ21 = + +
λ23 λ21 λ22

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 = σ1 d εl n1 + σ2 d εl n2 + σ3 d εl n3

Mooney models

For incompressible materials like elastomer’s (rubber) the stored elastic energy per unit of
deformed volume is specified and fitted onto experimental data. Several specific energy
functions are used.

Neo-Hookean W = C 10 (I 1 − 3)

184
 10.3. Large deformation elasticity

Mooney-Rivlin W = C 10 (I 1 − 3) +C 01 (I 2 − 3)

Signiorini W = C 10 (I 1 − 3) +C 01 (I 2 − 3) +C 20 (I 1 − 3)2

Yeoh W = C 10 (I 1 − 3) +C 20 (I 1 − 3)2 +C 30 (I 1 − 3)3

Klosner-Segal W = C 10 (I 1 − 3) +C 01 (I 2 − 3) +C 20 (I 1 − 3)2 +C 03 (I 2 − 3)3

2-order invariant W = C 10 (I 1 − 3) +C 01 (I 2 − 3) +C 11 (I 1 − 3)(I 2 − 3) +C 20 (I 1 − 3)2

Third-order model of James, Green and Simpson

W = C 10 (I 1 − 3) +C 01 (I 2 − 3) +C 11 (I 1 − 3)(I 2 − 3) +
C 20 (I 1 − 3)2 +C 02 (I 2 − 3)2 +C 21 (I 1 − 3)2 (I 2 − 3) +
C 30 (I 1 − 3)3 +C 03 (I 2 − 3)3 +C 12 (I 1 − 3)(I 2 − 3)2

Ogden models

For ’slightly’ compressible materials the Ogden specific energy functions are used. Because
the volume change is not zero, these functions depend on the volume change factor J =
λ1 λ2 λ3 . The second part of the energy function accounts for the volumetric deformation.
Because the volumetric behavior is characterized by a constant bulk modulus K , the model
is confined to slightly compressible deformation.
For ’highly’ compressible materials like foams, specific energy functions also exist. The
first part of the energy function also describes volume change.

N a · b ³ ´ ¸ µ
1
¶2
− i
X i bi bi bi
slightly compressible W= J 3 λ1 + λ2 + λ3 − 3 + 4.5K J − 1 3
i =1 b i
X N a ³ ´ X N a ¡ ¢
i b b b i
highly compressible W= λ1 i + λ2 i + λ3 i − 3 + 1 − J ci
i =1 b i i =1 c i

Uniaxial behaviour

For tensile (or compressive) loading of a homogeneous and isotropic truss,pwhere the axial
direction is taken to be the 1-direction, we have : λ1 = λ and λ2 = λ3 = 1/ λ. In this case
there is only an axial stress σ1 = σ, so that we have

dW dW d λ dW
dW = σ d εl n → σ= = = λ
d εl n d λ d εl n dλ
The Neo-Hookean model is the simplest model as it contains only one material parame-
ter. Axial stress σ and axial force F can be calculated easily. From statistical mechanics it is
known that for an ideal rubber material the stress is :
µ ¶
ρRT 2 1
σ= λ − with ρ : density
M λ
R : gas constant = 8.314 JK−1 mol−1
T : absolute temperature
M : average molecular weight

185
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

¡ ¢
W = C 10 λ2 + λ2 − 3
µ ¶ µ ¶
2 2 1
σ = C 10 2λ − 2 λ = 2C 10 λ −
λ λ
µ ¶
∂σ 1 ∂σ
Cλ = = 2C 10 2λ + 2 ; E = lim = 6C 10
∂λ λ λ→1 ∂λ
³ ´
F = σA = σ λ1 A 0 = 2C 10 A 0 λ − λ12

Most rubber materials cannot be characterized well with the Neo-Hookean model. The
more complex Mooney-Rivlin model yields better results. The stiffness C λ is a function of
the elongation factor λ. The initial stiffness E is often referred to as the modulus.

µ ¶ µ ¶
2 2 1
W = C 10 λ + − 3 +C 01 2 + 2λ − 3
λ λ
µ ¶ µ ¶
1 1 1
σ = 2C 10 λ2 − + 2C 01 λ2 −
λ λ λ
µ ¶ µ ¶
∂σ 1 2 ∂σ
Cλ = = 2C 10 2λ + 2 + 2C 01 1 + 3 ; E = lim = 6(C 10 +C 01 )
∂λ λ λ λ→1 ∂λ
· µ ¶ µ ¶ ¸
1 1 2 1 2 1 1
F = σA = σ A 0 = A 0 2C 10 λ − + 2C 01 λ −
λ λ λ λ λ

10.4 Examples
10.4.1 Example 1: limitations of a linear relation between the Cauchy
stress and the Green-Lagrange strain
A cylindrical tensile bar with initial length l 0 and initial cross-sectional area A 0 is loaded with
an axial force F . The elongation is described by the stretch ratio λ = ll . The contraction is
q 0
A
described by the stretch ratio µ = A 0 .

The material is homogeneous and the elastic behavior is described by a linear relation
between the Cauchy stress σ and the Green-Lagrange strain εg l = 21 (λ2 − 1).

σ = C (λ2 − 1) with C = constant > 0

In each direction the same strain definition must be used, so the contraction strain is

εd = 21 (µ2 − 1)
When the contraction strain is related to the axial strain with Poisson’s ratio ν, we have

εd = 21 (µ2 − 1) = −νεg l = −ν 21 (λ2 − 1) → µ2 = 1 − ν(λ2 − 1)

When Poisson’s ration is assumed to be constant, the axial elongation is limited because
the cross-sectional area obviously cannot become zero.

186
 10.4. Examples

r
2 2 1 1+ν 1+ν
1 − ν(λ − 1) = 0 → ν(λ − 1) = 1 → 2
λ − 1 = → λ2 = →λ=
ν ν ν
When the cross-sectional area is plotted as a function of λ with the value ν = 0.25, we
clearly see the limit value for λ where A = 0.

1.5

0.5
A

−0.5

−1
0 0.5 1 1.5 2 2.5 3
λ

Figure 10.22: Cross-sectional area versus stretch ratio.

The axial force F can be calculated and expressed

© as function
ª of λ.
F = σA = σµ2 A 0 = σ 1 − ν(λ2 − 1) A 0
= C A 0 {1 − ν(λ2 − 1)}(λ2 − 1)

When we plot this relation for values ν = 0.25 and A 0 = 1, it becomes clear that the pro-
posed material law has some physical inconsistencies.

1000

−1000

−2000

−3000
F

−4000

−5000

−6000

−7000

−8000
0 0.5 1 1.5 2 2.5 3
λ

Figure 10.23: Axial force versus stretch ratio.

The volume change ratio is :

© ª
J = λµ2 = λ 1 − ν(λ2 − 1)

When the material is assumed to be incompressible, J = 1 and Poisson’s ratio ν cannot

be constant any more but is a function of λ :

187
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

1
µ2 = = 1 − ν(λ2 − 1) →
λ
1 λ−1 λ−1 1
ν(λ2 − 1) = 1 − = → ν= =
λ λ λ(λ2 − 1) λ(λ + 1)

For very small elongations, the value of ν becomes 21 , which is already known from three-
dimensional Hooke’s law for linear elasticity.

10.4.2 Example 2: instability & localisation

A tensile test has revealed that the elastic behavior of a material is described very well by a
linear relation between the engineering stress σn and a nonlinear function of the elongation
factor λ.

1
σn = C ln(λ) with σn = engineering stress
λ
C = elasticity constant > 0
λ = axial elongation factor
The undeformed cross-sectional area of the truss is A 0. Poisson’s ratio is ν and is assumed
to be constant.

The relation between the axial force F and the axial stretch factor λ is

F = C A 0 λ−1 ln(λ)
For values C = 1000 and A 0 = 1, this relation is shown in the figure below.

400

300

200

100
F

−100

−200

−300
0 1 2 3 4 5 6
λ

Figure 10.24: Axial force versus stretch ratio.

The value for λ for which the axial force reaches the maximum value can be determined
by differentiation.
½ ¾
dF 1 1 1
= C A 0 − 2 ln(λ) + 2 = C A 0 2 (1 − ln(λ))
dλ λ λ λ
dF
= 0 → ln(λ) = 1 → λ = e = ±2.7
dλ

188
 10.4. Examples

The maximum value of the force is

1 1
Fmax = F (λ = e) = C A 0 ln(e) = C A 0
e e
When this maximum is reached, the deformation increases while the load diminishes.
This instability is called a snap-through. Because in a real material there will always be some
inhomogeneity, e.g. a cross-section with a slightly smaller area, or with a slightly lower value
of C , The maximum will be reached there first. All the elongation will then be concentrated
in this weakest cross-section, a phenomenon which is referred to as localisation.

10.4.3 Example 3: inflating a spherical balloon

A spherical balloon has an inner diameter D 0 and a uniform wall thickness w 0 ≪ D 0 in the
undeformed state. The balloon is loaded with an internal pressure p whereupon it deformes
homogeneously to have an inner diameter D and wall thickness w.
To describe the mechanics of the balloon, we use three coordinate axes : two perpendic-
ular tangential directions t and the radial direction r , as is shown in the figure below for the
deformed state.

t
p
t

r σt

Figure 10.25: Balloon in deformed state.

The wall of the balloon is made from elastomeric material, the behavior of which is char-
acterised by the Neo-Hookean elastic energy function, expressed in the principal stretch ra-
tios λ1 , λ2 and λ3 :
© ª
W = C 10 λ21 + λ22 + λ23 − 3
where C 10 is a positive valued material constant. The material is incompressible.

In the case of the pressurized balloon, the principal directions of deformation are the
perpendicular tangential directions and the radial direction in each point of the balloon wall,
so we have :

D w
λ1 = λ2 = λt = =λ and λ3 = λr =
D0 w0
With the knowledge that the material is incompressible, the elastic energy function can
be expressed in the tangential elongation factor λ.
1
J = λ1 λ2 λ3 = λ2t λr = 1 → λr = →
λ2t
µ ¶
¡ 2 2 2
¢ 2 1
W = C 10 λt + λt + λr − 3 = C 10 2λ + 4 − 3
λ

189
10. L ARGE DEFORMATIONS & NONLINEAR BEHAVIOUR

The principal stress directions coincide with the principal strain directions and are :

σ1 = σ2 = σ t and σ3 = σr ≈ 0
For the isotropic hyperelastic model, the incremental specific elastic energy can be writ-
ten as

dW
dW = σ1 d εl n 1 + σ2 d εl n 2 + σ3 d εl n 3 = 2σt d εl n t → σt = 21
d εl n t
The tangential stress σt is then expressed in λ.
µ ¶
1 dW 1
2
σt = 2 dλ λ = 2C 10 λ − 4
λ
From equilibrium in the deformed state, the relation between the internal pressure p and
the tangential stress σt is derived.
w
p = 4σt
D
The internal pressure can then be expressed in λ and the initial dimensions D 0 and w 0 .
µ ¶ µ ¶
w w0 2 1 1 w0 1 1
p = 4σt = 8C 10 λ − 4 3 = 8C 10 −
D D0 λ λ D 0 λ λ7
The plot of p versus λ for 1 ≤ λ ≤ 6 in figure 10.26 shows a clearly nonlinear relation.
This nonlinearity comes from the large reduction of the load-carrying wall thickness and
also from the nonlinear material behavior.

0.7

0.6
p / [ 8 (C10)∗(w0/D0) ]

0.5

0.4

0.3

0.2

0.1

0
1 2 3 4 5 6
λ = D/D0

Figure 10.26: Pressure versus diameter change.

190
 CHAPTER
11
P LASTICITY: BASIC CONCEPTS

11.1 Elasto-plastic 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 elon-
gation 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 harden-
ing.
Reversed loading will first 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 elasto-plastic material behavior is history depen-
dent: 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.

11.1.1 Uniaxial test

To investigate the characteristics of elasto-plastic material behavior, a uniaxial tensile/com-
pression test is carried out.

Tensile test

When a tensile bar, with undeformed length l 0 and cross-sectional area A 0 , is subjected to
a tensile test, the axial force F and the length l can be measured. The axial strain ε can be
calculated from the elongation factor λ. To calculate the true stress σ = FA , the actual cross-
sectional area of the tensile bar must be measured during the experiment. The nominal
stress σn = AF0 can be calculated straightforwardly.
The nominal stress σn can be plotted against the linear strain εl = λ−1 = l −l
l0
0
= ∆l
l0
result-
ing in the σn − εl stress-strain curve.
Until the proportionality limit σn = σP is reached, the material behavior is assumed to
be linear elastic : σn = E ε, where E is Young’s modulus. When the stress exceeds the initial

191
11. P LASTICITY

σ σ

ε ε

σ σ

ε ε

Figure 11.1: Stress-strain curves for elasto-plastic material behavior

yield stress σ y0 > σP , unloading will reveal permanent (= plastic) deformation of the bar. The
exact value of σ y0 cannot be determined so in practice σ y0 is taken to be the stress where a
plastic strain of 0.2 % remains. In the following however, we will assume that σ y0 is exactly
known and that σ y0 = σP .
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 σT . In forming processes strains can be much higher than in a tensile test, because
of the compression in certain directions.
After reaching the tensile strength the nominal stress will decrease while the strain is still
increasing. Fracture occurs at the fracture stress σn = σF . The fracture strain εF is for metals
and metal alloys about 10% = 0.1. This is a rather small elongation which means that for
these materials we can assume σ = σn and also that all strain definitions are approximately
equivalent, so ε = εl .
Plastic deformation in metals is typically carried by crystallographic slip in the crystal
lattice. This physical deformation process does not entail any volume changes, as confirmed
by experiments. Plastic deformation in metals is therefore incompressible.

Compression test

For metal alloys a compression test instead of a tensile test will reveal that first yield will
occur at σ = σn = −σ y0 . The initial material behavior is the same in tension and compres-
sion. In general terms the transition from purely elastic behavior to elasto-plastic behavior
is determined by a yield criterion. For the one-dimensional case this criterion says that first
yielding will occur when :

f = σ2 − σ2y0 = 0

The function f is the yield function.

192
 11.1. Elasto-plastic material behavior

σn
T

σ y0 F
σP

ε = 0.002 εℓ

Figure 11.2: Stress-strain curve during tensile test

σ
σ y0

− σ y0

Figure 11.3: Stress-strain curve during tensile or compression test

Interrupted tensile test

When the axial load is released at σ A (see figure below) with σ y0 < σ A < σT , the unloading
stress-strain path is elastic and characterized by the initial Young’s modulus E . The perma-
nent or plastic elongation is represented by the plastic strain εp . The difference between the
total strain in point A and the plastic strain is the elastic strain εe = ε A − εp = σEA .

Resumed tensile test

When after unloading, the bar is again loaded with a tensile force, the elastic line B A will
be followed where ∆σ = E ∆ε = E ∆εe holds. For σ ≥ σ A (ε ≥ ε A ) further elasto-plastic de-
formation takes place and the stress-strain curve will be followed as if unloading were not
occurred.

193
11. P LASTICITY

εe
σ ε
εp

t1 t2 t t1 t2 t

σ
A
σy
σ y0

B
εA
ε
εp εe

Figure 11.4: Stress-strain curve after interrupted tensile test

The stress σ A is the current yield stress σ y , which is generally larger then the initial yield
stress σ y0 . The increase, referred to as hardening, is related to the plastic strain by a harden-
ing law.

11.1.2 Hardening

To study the hardening phenomenon, the tensile bar is not reloaded in tension but in com-
pression. Two extreme observations may be made, illustrated in the figure below.
In the first case the elastic trajectory increases in length due to plastic deformation :
A A ′ > Y0 Y0′ . The elastic trajectory is symmetric about σ = 0 (B A = B A ′ ). What we observe is
isotropic hardening.
In the second case the elastic trajectory remains of constant length : A A ′ = Y0 Y0′ . It is
symmetric about the line OC (C A = C A ′). After unloading the yield stress under compression
is different 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 σ = q. This phenomenon is
also referred to as the Bauschinger effect.
Real materials will show a combination of isotropic and kinematic hardening.

194
 11.1. Elasto-plastic material behavior

σ A σ A
Y0 Y0 σ y0

C
0 B 0 B q
ε ε
′
A
Y0′ Y0′

A′

Figure 11.5: Isotropic and kinematic hardening

isotropic hardening : elastic area larger & symmetric w.r.t. σ = 0

¾
tensile : σ = σy
→ f = σ2 − σ2y = 0
compression : σ = −σ y
kinematic hardening : elastic area constant & symmetric w.r.t. σ = q
¾
tensile : σ = q + σ y0
→ f = (σ − q)2 − σ2y0 = 0
compression : σ = q − σ y0
combined isotropic/kinematic hardening
¾
tensile : σ = q + σy
→ f = (σ − q)2 − σ2y = 0
compression : σ = q − σy

Effective plastic strain

Isotropic hardening could be described by relating the yield stress σ y to the plastic strain εp .
However, as the figure below shows, this would lead to the unrealistic conclusion that the
yield stress increases while the plastic strain decreases. To prevent this problem, the effective
plastic strain ε̄p 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.

τ=t Zt Zt
X X |∆εp |
ε̄p = |∆εp | = ∆t = |ε̇p | d τ = ε̄˙ p d τ
ε τ=0 ∆t τ=0 τ=0

Hardening laws

For one-dimensional stress states encountered in the axial loading of a truss, several hard-
ening laws are formulated, based on experimental observations. They can be generalized
to three-dimensional stress-strain states. For isotropic hardening the current yield stress is
related to the effective plastic strain and the initial yield stress. The isotropic hardening pa-
dσ y
rameter is H = d ε̄p
. For kinematic hardening the shift stress q is related to the plastic strain
dq
εp . The kinematic hardening parameter is K = dε .
p

195
11. P LASTICITY

D D
ε σ
C
A
E A
C
B
t ε

B
E

Figure 11.6: Increasing yield stress at decreasing plastic strain

Linear and power law hardening laws

Many hardening laws represent a linear or exponential relationship between stress and
strain.

linear hardening σ y = σ y0 + H ε̄p

µ ¶n
ε̄p
Ludwik (1909) σ y = σ y0 + σ y0 (0 ≤ n ≤ 1) →
ε y0
µ ¶ µ ¶n−1
σ y0 ε̄p n−1 ε̄p
H =n = nE
ε y0 ε y0 ε y0
³ ´
mod. Ludwik σ y = σ y0 1 + m ε̄np → H = σ y0 mn ε̄n−1
p

σ y0
Swift (1952) σ y = C (m + ε̄p )n with C=
mn
¡ ¢n−1
H = C n m + ε̄p
· µ ¶ ¸
σy σ y m−1
Ramberg-Osgood (1943) ε̄p = 1+α (m ≥ 0; α ≈ 73 )
E σ y0
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.

ideal plastic σ y = σ y0

196
 11.2. Elasto-plastic model

µ ¶
E ε̄p
Prager (1938) σ y = σ y0 tanh
σ y0
· µ ¶¸ · µ ¶¸2
σ y0 ε̄p 2 ε̄p
H= sech = E sech
ε y0 ε y0 ε y0
E ε̄p m 1/m
· µ ¶ ¸
Betten I (1975) σ y = σ y0 tanh (m > 1)
σ y0
¶ ¸1 ·
ε̄p m m −1 ε̄p m 2
µ ¶m−1 · µ µ ¶ ¸
ε̄p
H =E tanh sech
ε y0 ε y0 ε y0
¡ ¢ σ y0
Voce (1949) σ y = C 1 − ne −m ε̄p with C = (m > 1)
1−n

H = C nme −m ε̄p

ε̄p m −1/m
· µ ¶ ¸
Betten II (1975) σ y = σ y0 + (E ε̄p ) 1 +
ε y0
1 " µ ¶ ¾ #
ε̄p m − m ε̄p m −1
· µ ¶ ¸ ¶ ½ µ
ε̄p m
H = E 1+ 1− 1+
ε y0 ε y0 ε y0

Cyclic load for linear hardening

A truss can be loaded with a prescribed strain −εm ≤ ε ≤ εm . It is assumed that the stress will
reach values above the initial yield stress σ y0 and that linear hardening occurs.
For purely isotropic hardening the stress will increase after each load reversal and finally
no further plastic deformation will take place. For purely kinematic hardening the stress-
strain path will be one single hysteresis loop, where the stress cycles, as does the strain, be-
tween two constant values −σm ≤ σ ≤ σm .

11.2 Elasto-plastic model

The elasto-plastic deformation characteristics can be represented by a discrete mechanical
model. A friction element represents the yield limit and a hardening spring – stiffness H
(H > 0) – provides the stiffness reduction after reaching the yield limit. The elasto-plastic
model describes rate-independent plasticity – there is no dashpot in the discrete model –, so
the time is fictitious and ”rate” is just referring to momentary change.
The yield criterion is used to decide at which stress level a purely elastic deformation
will be followed by elasto-plastic deformation. During elasto-plastic deformation the total
strain rate (ε̇) is additively decomposed in an elastic (ε̇e ) and a plastic (ε̇p ) part. The plastic
∂f
strain rate ε̇p is related to ∂σ by the rate of the plastic multiplier λ, the so-called consistency
parameter λ̇. During ongoing plastic deformation the consistency equation f˙ = 0 must be
satisfied, because f must remain zero.
The hardening law relates the current yield stress σ y to the initial yield stress σ y0 and the
effective plastic strain ε̄p . The shift stress q is related to the plastic strain εp .

197
11. P LASTICITY

σ1
σ y0
σ y0 σm

−εm ε y0 εm −εm
ε y0 εm

−σm

Figure 11.7: Stress-strain curve during cyclic loading for isotropic and for kinematic harden-
ing

σy
E
σ
H

εp εe

Figure 11.8: Discrete mechanical model for elasto-plastic material behavior

198
 11.2. Elasto-plastic model

• f = (σ − q)2 − σ2y with f < 0 | f = 0 ∧ f˙ < 0 → elastic

f = 0 ∧ f˙ = 0 → elasto-plastic
• σ y = σ y (σ y0 , ε̄p ) ; q = q(εp )
• ε̇ = ε̇e + ε̇p
1
• σ = E εe → ε̇e = σ̇
E
∂f
• ε̇p = λ̇ = 2λ̇(σ − q) ; ε̄˙ p = |ε̇p | = 2λ̇ |σ − q|
Z∂σt X
• ε̄p = ε̄˙ p d τ = |∆εp |
τ=0 t

Constitutive equations

From the constitutive relations a set of constitutive equations can be derived.


σ̇ = E ε̇e = E (ε̇ − ε̇p ) = E {ε̇ − 2λ̇(σ − q)} 
→
f =0 

σ̇ + 2E (σ − q)λ̇ − E ε̇ = 0 

f =0 

Integration of the constitutive equations

The current stress has to be determined from these constitutive equations. The first one is
a differential equation in pseudo-time. To solve it, we need an initial condition, which is
not the stress in the undeformed state, which would obviously be zero, but the stress at the
beginning of the current increment. Values at the beginning if the current increment are
indicated with an index n. A value at the end of this increment, which is the current state, is
often indicated with the index n + 1, but is skipped here.
There are many procedures which can be followed to solve the differential equation for
the stress. They can be classified as implicit or explicit. The implicit methods are more ac-
curate and more stable then the explicit methods.
We assume that the begin-increment state resides on the yield trajectory, so f n = 0. In
reality this is of course not always the case : the begin-increment case may be elastic ( f n < 0)
and plastic deformation will develop during the increment. The implicit procedures can well
cope with this phenomenon. Explicit procedures will need some correction.

Explicit solution procedure

An explicit procedure starts from the known state at the beginning of the increment and
calculates incremental changes directly, assuming that values of some variables remain the
same during the increment. Obviously, this is not through, so these procedures are not very
accurate. The final solution may not satisfy the yield criterion f = 0 exactly, which calls for
a correction, where the final state is projected onto the yield trajectory. Here we ignore this
inaccuracy and calculate only the material stiffness C ε = ∆σ
∆ε
during the increment. Because
the begin-increment state is assumed to be on the yield-trajectory, we know that |σn − q n | =
σ yn .

199
11. P LASTICITY


∆σ + 2E (σn − q n )∆λ = E ∆ε 

¯ ¯
∂ f ¯¯ ∂ f ¯¯
∆f = 0 → ∆σ + ∆λ = 0 

∂σ ¯n ∂λ ¯n



∆σ + 2E (σn − q n )∆λ = E ∆ε 





2(σn − q n )∆σ − 4K n (σn − q n )2 ∆λ − 4Hn σ yn |σn − q n |∆λ = 0 → →




(σn − q n ) 1 

∆λ = 2
∆σ = ∆σ 

2K n (σn − q n ) + 2Hn σ yn |σn − q n | 2K n (σn − q n ) + 2Hn (σn − q n )

E (K n + Hn ) E
∆σ = ∆ε ; ∆εp = 2(σn − q n )∆λ = ∆ε
E + K n + Hn E + K n + Hn

Linear isotropic hardening

For linear isotropic hardening with constant hardening parameter H, the above stiffness re-
duces to a simpler relation.

EH 1 E
∆σ = ∆ε ; ∆εp = ∆σ = ∆ε
E +H H E +H
These relations can also easily be derived by using the figure below. In a point of the
elastic trajectory we know that ∆σ = E ∆ε holds. In a point of the elasto-plastic trajectory we
can write ∆σ = S∆ε, where the material stiffness S = C ε will depend on E and H.

σ B
∆σ
A

ε
∆ε ∆εe

∆εp

Figure 11.9: Stress-strain curve for monotonic tensile loading

µ ¶ µ ¶
∆σ y ∆σ
∆σ = E ∆εe = E (∆ε − ∆εp ) = E ∆ε − = E ∆ε − →
H H
EH ∆σ E
∆σ = ∆ε = S∆ε ; ∆εp = = ∆ε
E +H H E +H

200
 11.3. Examples

Kinematic hardening

For linear kinematic hardening, the result is similarly derived.

EK 1 E
∆σ = ∆ε ; ∆εp = ∆σ = ∆ε
E +K K E +K
Again these relations can be derived straightforwardly from the figure.
µ ¶ µ ¶
∆q ∆σ
∆σ = E ∆εe = E (∆ε − ∆εp ) = E ∆ε − = E ∆ε − →
K K
EK ∆σ E
∆σ = ∆ε = S∆ε ; ∆εp = = ∆ε
E +K K E +K
Note that the stiffness equals Young’s modulus when H (or K ) approaches infinity.

EH E
lim = lim E
=E
H →∞ E + H H →∞ +1
H

11.3 Examples
11.3.1 Example 1: Reversed plasticity in a tensile bar
A cylindrical tensile bar is loaded with an axial stress σ, which is applied as a function of the
(pseudo)time t as indicated in the figure below.

σ(t )

σ1

0 t1 2t1 t

Figure 11.10: Prescribed stress as a function of (pseudo)time.

When the load is increased from t = 0, the material behaves linearly elastic – Young’s
modulus is E – until the initial yield stress σ y0 is reached. The maximum stress σ1 is such
that σ1 > σ y0 . After reaching this maximum value the stress is reduced to zero and compres-
sive loading is applied. After yielding, the material shows linear, isotropic hardening with
hardening constant H.

First the stress-strain diagram is plotted and relevant stress-strain points are indicated.

201
11. P LASTICITY

σ
σ1 1

0
σ y0

ε2

p
ε y0 ε1 ε1 ε

σ2
2
3 σ3

Figure 11.11: Stress-strain diagram.

When the stress reaches the maximum value σ1 at t = t1 the strain ε1 can be calculated
and expressed in σ1 , E , σ y0 and H.

E +H σ y0 E + H
ε1 = ε y0 + (σ1 − σ y0 ) = + (σ1 − σ y0 )
µ EH ¶ E EH
σ y0 E +H E +H
= 1− + σ1
E H EH
σ y0 E + H
=− + σ1
H EH

p
The plastic strain ε1 at maximum stress σ1 is
µ ¶
p σ1 σ y0 E +H σ1
ε1 = ε1 − =− + −1
E H H E
σ1 − σ y0
=
H

From t = t1 the stress is reduced to zero after which compression occurs. The strain ε2 at
yield under compressive loading is :

σ1
ε2 = ε1 − 2
E
The stress σ3 for which the plastic strain is zero, is :

p p
ε3 = 0 → − (σ3 − σ2 ) = σ1 − σ y0 = Hε1 → σ3 = −2σ1 + σ y0

202
 11.3. Examples

L A

a F
L 2A

Figure 11.12: Parallel truss structure.

11.3.2 Example 2: parallel truss structure

Figure 11.12 shows two parallel bars a and b, which are connected to a rigid wall and to a
rigid block, which can only translate in horizontal direction. The bars have the same initial
length L. The cross-sectional area of a and b is A and 2A, respectively. Both a and b have
the same Young’s modulus E and initial yield stress σ y0 . Upon yielding bar a is ideal plastic,
while bar b shows isotropic linear hardening with hardening constant H.
The axial deformation is provoked by the force F , which increases to a maximum value
after which it will be reduced to zero. The maximum is such that both trusses will undergo
elasto-plastic deformation. The displacement of the rigid block is δ.

For all deformation states the strain in both trusses will be equal : εa = εb , due to the
fact that their initial length and their elongation is the same. With this in mind we draw the
stress-strain diagram of both bars below each other in figure 11.13.
The force F0 at which yielding occurs for the first time, is

F0 = σ y0 A + 2σ y0 A = 3σ y0 A

The strain ε0 and the displacement δ0 is

σ y0 σ y0
ε0 = ε y0 = → δ0 = L
E E
The force F is increased to F = F1 = 53 F0 . Because truss a doesn’t show any hardening,
the axial stress does not change : σ1a = σ0a = σ y0 , and thus F1a = σ y0 A. The axial force and the
stress in truss b is then :

F1b = F1 − F1a = 5σ y0 A − σ y0 A = 4σ y0 A → σb1 = 2σ y0

The strain εb1 in truss b and the displacement δ1 can now be calculated.
µ ¶ µ ¶
b
σ y0 E + H σ y0 E +H σ y0 E + 2H
ε1 = + σ y0 = 1+ =
E EH E H E H
µ ¶
σ y0 E + 2H
δ1 = Lεb1 = L
E H

203
11. P LASTICITY

σ
σ y0 0 1
a

ε
2
2σ y0 1

σ y0 0
b
2

Figure 11.13: Stress-strain diagrams of both trusses.

The plastic strain of the trusses is

σ y0 E +H
εap1 = ε1a − = σ y0
E EH µ ¶
2σ y0 σ y0 E + H σ y0 E +H σ y0
εbp1 = εb1 − =− + σ y0 = −1 + =
E E EH E H H
The force F is now reduced to zero. The residual stresses in both trusses can be calcu-
lated.

a a 1¡ a a
¢ 

ε2 − ε1 = σ2 − σ1 


E 


1 ³ ´
b b
ε2 − ε1 = b
σ2 − σ1b → σ2a − σ1a = σb2 − σb1 →
E 




σ1a = σ y0 ; σb1 = 2σ y0
a b a b 

ε =ε
2 ; ε =ε
2 1
 1

σ2a − σ y0 = σb2 − 2σ y0 → σb2 − σ2a = σ y0 
→
equilibrium → σ2a A + σb2 2A = 0 → σ2a = −2σb2 

3σb2 = σ y0 → σb2 = 31 σ y0 ; σ2a = − 32 σ y0

11.3.3 Example 3: serial truss structure

A cylindrical truss is divided in two parts a and b as shown in figure 11.14. In the unde-
formed state, part a has length L and cross-sectional area 2A, part b has length 2L and cross-
sectional area A. In point R an axial force F ia applied. Point P is fixed and point R will show
a displacement δ.
For each state in the deformation process, we will have :

204
 11.3. Examples

P 2A
Q A R
F

b
a

L 2L

Figure 11.14: Series configuration of two trusses.

Na = Nb → σa = 21 σb
When the load is increased from F = 0 at t = 0 the material is linearly elastic – Young’s
modulus E – as long as the axial stress is below the initial yield stress σ y0 . After yielding the
material will show linear isotropic hardening with hardening constant H, see figure 11.15.

σ
2b
12
5 σ y0
2σ y0 1b
6 2a
5 σ y0 0b
σ y0 1a
0a

εap εbp ε

Figure 11.15: Stress-strain diagram of the two trusses.

When yielding occurs for the first time the force F0 is part b will yield first at a force

F0 = σ y0 A

The displacement δ0 at that moment is


σ y0 σ y0 
σb0 = σ y0 → εb0 = → ∆L b0 = 2L 
 σ y0
E E → δ0 = ∆L a0 + ∆L b0 = 52 L
σ y0 σ y0  E
σ0a = 12 σ y0 ε0a = ∆L a0 = L

→ → 
2E 2E

The load is increased to F = F1 > F0 when part a will yield for the first time.

205
11. P LASTICITY

σ1a = σ y0 → σb1 = 2σ y0 → F1 = 2σ y0 A
The displacement δ1 can than be calculated :
σ y0 σ y0 
σ1a = σ y0 → ε1a
= → ∆L a1
=L 

E E µ ¶ µ ¶ →
b b E +H σ y0 E + H σ y0 E + 2H b
σ y0 E + 2H 
ε1 = ε y0 + σ y0 = + σ y0 = → ∆L 1 = 2L 
EH E EH E H E H
µ ¶ µ ¶
σ y0 σ y0 E + 2H σ y0 2E + 5H
δ1 = ∆L a1 + ∆L b1 = L + 2L =L
E E H E H

The force F is increased further to F = F2 = 56 F1 = 125

σ y0 A and is subsequently reduced
to zero : F = F3 = 0. The residual plastic strain in each part can be determined.

F2 12 E +H b σ y0
σb2 = = 5 σ y0 → εb2 − εb1 = (σ2 − σb1 ) → εb2 = 51 (7E + 12H)
A EH EH
F2 6 E +H a σ
1 y0
σ2a = = 5 σ y0 → ε2a − ε1a = (σ2 − σ1a ) → ε2a = 5 (E + 6H)
2A EH EH
σ y0 σ y0 σ y0 σ y0
εbp2 = εb3 = εb2 − 12
5
= 75 ; a
ε2p = ε3a = ε2a − 65 = 15
E H E H
The total displacement after unloading is
σ y0 σ y0 σ y0
δ3 = ∆L a3 + ∆L b3 = L 15 + 2L 75 = 3L
H H H

206
 NDIX
A

APPE
S TIFFNESS AND COMPLIANCE MATRICES

In chapter 5 the three-dimensional stiffness and compliance matrices have been derived for
various materials. Increasing microstructural lattice symmetry gave rise to a reduction of
the number of material constants. Starting from triclinic with no symmetry and character-
ized by 21 material constants, increased symmetry was seen for monoclinic (13 constants),
orthotropic (9), quadratic (6), transversal isotropic (5), cubic (3) and finally, isotropic, with
only 2 material constants.

In this appendix, we again present the material matrices for orthotropic, transversal isotropic
and fully isotropic material. The material constants will be expressed in engineering con-
stants, where we choose Young’s moduli, Poisson’s ratios and shear moduli.

In many engineering problems, the state of strain or stress is planar. Both for plane strain
and plane stress, only the strain and stress components in a plane have to be related through
a material law. Here we assume that this plane is the 12-plane. For plane strain we than
have ε33 = γ23 = γ31 = 0, and for plane stress σ33 = σ23 = σ31 = 0. The material law for these
planar situations can be derived from the linear elastic three-dimensional stress-strain rela-
tion. This is done, first for the general orthotropic material law. The result is subsequently
specified in engineering parameters for orthotropic, transversal isotropic and fully isotropic
material.

A.1 General orthotropic material law

The general orthotropic material law is expressed by the stiffness matrix C and/or its inverse,
the compliance matrix S.

207
A. S TIFFNESS AND COMPLIANCE MATRICES

    
σ11 A Q R 0 0 0 ε11

 σ22  
  Q B S 0 0 0   ε22 
 
    
 σ33   R S C 0 0 0   ε33 
σ =
~  =  =C~ ε
 σ12   0 0 0 K 0 0   γ12 
    
 σ23   0 0 0 0 L 0   γ23 
σ31 0 0 0 0 0 M γ31
    
ε11 a q r 0 0 0 σ11

 ε22  
  q b s 0   σ22 
0 0   
    
ε33 r s c 0 0 0   σ33 
 = C −1 σ
  
ε=
~ =  ~ =Sσ
~
 γ12   0 0 0 k 0 0   σ12 
    
 γ23   0 0 0 0 l 0   σ23 
γ31 0 0 0 0 0 m σ31

The inverse of C can be expresses in its components.

 
BC − S 2 −QC + RS QS − BR 0 0 0

 −QC + RS AC − R 2 −AS +QR 0 0 0 

1  QS − BR −AS +QR AB −Q 2

−1  0 0 0 
C =  
∆c  0 0 0 ∆c (1/K ) 0 0 
 
 0 0 0 0 ∆c (1/L) 0 
0 0 0 0 0 ∆c (1/M)

with ∆c = ABC − AS 2 − BR 2 −CQ 2 + 2QRS

As will be clear later, it will mostly be easier to start with the compliance matrix and cal-
culate the stiffness matrix by inversion.
 
bc − s 2 −qc + r s q s − br 0 0 0

 −qc + r s ac − r 2 −as + qr 0 0 0 

ab − q 2
1  
q s − br −as + qr 0 0 0
S −1 =
 
 
∆s  0 0 0 ∆s (1/k) 0 0 
 
 0 0 0 0 ∆s (1/l ) 0 
0 0 0 0 0 ∆s (1/m)

with ∆s = abc − as 2 − br 2 − cq 2 + 2qr s

Increasing material symmetry leads to a reduction in material parameters.

quadratic B = A ; S = R ; M = L;
transversal isotropic B = A ; S = R ; M = L ; K = 21 (A −Q)
cubic C = B = A ; S = R = Q ; M = L = K 6= 12 (A −Q)
isotropic C = B = A ; S = R = Q ; M = L = K = 21 (A −Q)

The planar stress-strain laws can be derived either from the stiffness matrix C or from
the compliance matrix S. The plane strain state will be denoted by the index ε and the plane
stress state will be indicated with the index σ.

208
 A.1. General orthotropic material law

A.1.1 Plane strain

For a plane strain state with ε33 = γ23 = γ31 = 0, the stress σ33 can be expressed in the planar
strains ε11 and ε22 . The material stiffness matrix C can be extracted directly from C . The
ε
material compliance matrix S has to be derived by inversion.
ε

ε33 = γ23 = γ31 = 0 → σ33 = Rε11 + Sε22

       
σ11 A Q 0 ε11 Aε Qε 0 ε11
~ = σ22 = Q B 0
σ      ε22  =  Q ε B ε 0   ε22  = C ~ε
ε
σ12 0 0 K γ12 0 0 K γ12
 
  B −Q 0     
ε11 σ11 aε qε 0 σ11
1  −Q A 0 
ε = ε22
~
 =    σ22  =  q ε b ε 0   σ22  = S σ
AB −Q 2  AB −Q 2  ε~
γ12 0 0 σ12 0 0 k σ12
K
Because the components of the three-dimensional compliance matrix S are most con-
veniently expressed in Young’s moduli, Poisson’s ratios and shear moduli, this matrix is a
good starting point to derive the planar matrices for specific cases. The plane strain stiffness
matrix C must then be determined by inversion.
ε
r s
ε33 = 0 = r σ11 + sσ22 + cσ33 → σ33 = − σ11 − σ22
c c

        
ε11 a q 0 σ11 r h σ11
r s i
ε
~ =  ε22
 =  q b 0   σ 22
 −  s  0  σ22 
c c
γ12 0 0 k σ12 0 σ12
     
ac − r 2 qc − r s 0 σ11 aε qε 0 σ11
1
= qc − sr bc − s 2 0   σ22  =  q ε b ε 0   σ22  = S σ
c ε~
0 0 kc σ12 0 0 k σ12
 

σ11
 
aε qε 0
−1 
ε11
 bc − s 2 −qc + r s 0 
ε11

1  −qc + r s ac − r 2
 
0  ε 
~ = σ22 = q ε b ε 0
σ      ε22 = 22
∆s  ∆s 
σ12 0 0 k γ12 0 0 ε12
k

with ∆s = abc − as 2 − br 2 − cq 2 + 2qr s

  
Aε Qε 0 ε11
=  Q ε B ε 0   ε22  = C σ
ε~
0 0 K ε12

We can now derive by substitution :

1 £ ¤
σ33 = − (br − q s)ε11 + (as − qr )ε22
∆s

209
A. S TIFFNESS AND COMPLIANCE MATRICES

A.1.2 Plane stress

For the plane stress state, with σ33 = σ23 = σ31 = 0, the two-dimensional material law can be
easily derived from the three-dimensional compliance matrix S . The strain ε33 can be di-
ε
rectly expressed in σ11 and σ22 . The material stiffness matrix has to be derived by inversion.

σ33 = σ23 = σ31 = 0 → ε33 = r σ11 + sσ22

       
ε11 a q 0 σ11 aσ qσ 0 σ11
ε =  ε22  =  q b 0   σ22 
~ =  q σ b σ 0   σ22  = S σ
σ~
γ12 0 0 k σ12 0 0 k σ12
 
  b −q 0     
σ11 ε11 Aσ Qσ 0 ε11
 σ22 1  −q a 0 
  ε22  =  Q σ B σ 0   ε22  = C ε
σ
~ = = 
ab − q 2  ab − q 2  σ~
σ12 0 0 γ12 0 0 K γ12
k
We can derive by substitution :

1 £ ¤
ε33 = (br − q s)ε11 + (as − qr )ε22
ab − q 2
The same relations can be derived from the three-dimensional stiffness matrix C .

R S
σ33 = 0 = Rε11 + Sε22 +C ε33 → ε33 = − ε11 − ε22
C C

        
σ11 A Q 0 ε11 R · ¸ ε11
σ =  σ22  =  Q B 0   ε22  −  S  R S
0  ε22 
~
~ C C
σ12 0 0 K γ12 0 γ12
     
AC − R 2 QC − RS 0 ε11 Aσ Qσ 0 ε11
1
=  QC − SR BC − S 2 0   ε22  =  Q σ B σ 0   ε22  = C ~ε
C σ
0 0 KC γ12 0 0 K γ12

   −1     
ε11 Aσ Qσ 0 σ11 aσ qσ 0 σ11
ε =  ε22  =  Q σ B σ 0   σ22  =  q σ b σ 0   σ22  = S σ
~ σ~
γ12 0 0 K σ12 0 0 k σ12

A.1.3 Plane strain/stress

In general we can write the stiffness and compliance matrix for planar deformation as a 3 × 3
matrix with components, which are specified for plane strain (p = ε) or plane stress (p = σ).
   
Ap Qp 0 ap qp 0
C =  Qp Bp 0  ; S =  qp bp 0 
p p
0 0 K 0 0 k
The general relations presented before can be used to calculate the components of C
p
and/or S when components of the three-dimensional matrices C and/or S are known.
p

210
 A.2. Linear elastic orthotropic material

In the next sections the three-dimensional and planar material matrices are presented for
orthonormal, transversal isotropic and fully isotropic material.

A.2 Linear elastic orthotropic material

For an orthotropic material 9 material parameters are needed to characterize its mechanical
behavior. Their names and formal definitions are :

∂σi i
Young’s moduli : Ei =
∂εi i
∂ε j j
Poisson’s ratios : νi j = −
∂εi i
∂σi j
shear moduli : Gi j =
∂γi j
The introduction of these parameters is easily accomplished in the compliance matrix S.
Due to the symmetry of the compliance matrix S, the material parameters must obey the
three Maxwell relations.
 
E 1−1 −ν21 E 2−1 −ν31 E 3−1 0 0 0
−1 −1 −1

 −ν12 E 1 E2 −ν32 E 3 0 0 0 

 −1 −1 −1 
 −ν13 E 1 −ν23 E 2 E3 0 0 0 
S= −1 
 0 0 0 G 12 0 0 
 −1 
 0 0 0 0 G 23 0 
−1
0 0 0 0 0 G 31
ν12 ν21 ν23 ν32 ν31 ν13
with = ; = ; = (Maxwell relations)
E1 E2 E2 E3 E3 E1

The stiffness matrix C can then be derived by inversion of S.

 1−ν32 ν23 ν31 ν23 +ν21 ν21 ν32 +ν31

E2E3 E2E3 E2E3 0 0 0
 ν13 ν32 +ν12 1−ν31 ν13 ν12 ν31 +ν32 
 E1E3 E1E3 E1E3
0 0 0 
 ν12 ν23 +ν13 ν21 ν13 +ν23 1−ν12 ν21

1  0 0 0 
C =  E1E2 E1E2 E1E2 
∆s 
 0 0 0 ∆s G 12 0 0


 
 0 0 0 0 ∆s G 23 0 
0 0 0 0 0 ∆s G 31
1 − ν12 ν21 − ν23 ν32 − ν31 ν13 − ν12 ν23 ν31 − ν21 ν32 ν13
with ∆s =
E 1E 2E 3

A.2.1 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 stiffness) 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.

211
A. S TIFFNESS AND COMPLIANCE MATRICES

T
σ
~ = [σ11 σ22 σ33 σ12 σ23 σ31 ] = [σ1 σ2 σ3 σ6 σ4 σ5 ]
εT = [ε11 ε22 ε33 γ12 γ23 γ31 ] = [ε1 ε2 ε3 ε6 ε4 ε5 ]
~

    
ε1 S 11 S 12 S 13 0 0 0 σ1

 ε2  
  S 21 S 22 S 23 0 0 0 
 σ2 

    
 ε3   S 31 S 32 S 33 0 0 0  σ3 
 =  
 ε4   0 0 0 S 44 0 0  σ4 
    
 ε5   0 0 0 0 S 55 0  σ5 
ε6 0 0 0 0 0 S 66 σ6

A.2.2 Plane strain

For plane strain the stiffness matrix can be extracted from the three-dimensional stiffness
matrix. The inverse of this 3x3 matrix is the plane strain compliance matrix.

E3 E3
σ33 = ν13 σ11 + ν23 σ22
E1 E2
 1−ν31 ν13 ν31 ν23 +ν21 
E1 − E2 0
 ν ν +ν 1−ν32 ν23 
S = − 13 E321 12 E2 0 
ε 1
0 0 G 12
 1−ν32 ν23

ν31 ν23 +ν21
E2E3 0
E2E3
1 
C = S −1 = ν13 ν32 +ν12 1−ν31 ν13

 E1E3 0
E1E3 
ε ε ∆s
0 0 ∆s G 12
1 − ν12 ν21 − ν23 ν32 − ν31 ν13 − ν12 ν23 ν31 − ν21 ν32 ν13
with ∆s =
E 1E 2E 3
½ ¾
1 ν12 ν32 + ν13 ν21 ν13 + ν23
σ33 = ε11 + ε22
∆s E 1E 2 E 1E 2

A.2.3 Plane stress

For plane stress the compliance matrix can be extracted from the three-dimensional com-
pliance matrix. The inverse of this 3x3 matrix is the plane strain stiffness matrix.

ε33 = − ν13 E 1−1 σ11 − ν23 E 2−1 σ22

 
E 1−1 −ν21 E 2−1 0
S =  −ν12 E 1−1 E 2−1 0 
σ −1
0 0 G 12
 
E1 ν21 E 1 0
1
C = S −1 =  ν12 E 2 E2 0 
σ σ 1 − ν21 ν12
0 0 (1 − ν21 ν12 )G 12
1
ε33 = − {(ν12 ν23 + ν13 )ε11 + (ν21 ν13 + ν23 )ε22 }
1 − ν12 ν21

212
 A.3. Linear elastic transversal isotropic material

A.3 Linear elastic transversal isotropic material

Considering an transversally isotropic material with the 12-plane isotropic, the Young’s mod-
ulus E p and the Poisson’s ratio νp in this plane can be measured. The associated shear mod-
Ep
ulus is related by G p = . In the perpendicular direction we have the Young’s modu-
2(1 + νp )
lus E 3 , the shear moduli G 3p = G p3 and two Poisson ratios, which are related by symmetry :
νp3 E 3 = ν3p E p .
 
E p−1 −νp E p−1 −ν3p E 3−1 0 0 0
 −νp E p−1 E p−1 −ν3p E 3−1
 
0 0 0 
 
 −νp3 E p−1 −νp3 E p−1 E 3−1 0 0 0 
S=  
 0 0 0 G p−1 0 0  
 −1 
 0 0 0 0 G p3 0 
−1
0 0 0 0 0 G 3p
νp3 ν3p
with =
Ep E3
 1−ν3p νp3 ν3p νp3 +νp νp ν3p +ν3p 
Ep E3 Ep E3 Ep E3 0 0 0
 νp3 ν3p +νp 1−ν3p νp3 νp ν3p +ν3p 

 Ep E3 Ep E3 Ep E3
0 0 0 

1  νp νp3 +νp3 νp νp3 +νp3 1−νp νp 
C = S −1 = 0 0 0
 
 Ep Ep Ep Ep Ep Ep 
∆s  

 0 0 0 ∆s G p 0 0 

 0 0 0 0 ∆s G p3 0 
0 0 0 0 0 ∆s G 3p
1 − νp νp − νp3 ν3p − ν3p νp3 − νp νp3 ν3p − νp ν3p νp3
with ∆s =
Ep Ep E3

A.3.1 Plane strain

The plane strain stiffness matrix can be extracted from the three-dimensional stiffness ma-
trix. The inverse of this 3x3 matrix is the plane strain compliance matrix.
E 3 νp3
σ33 = (σ11 + σ22 ) = ν3p (σ11 + σ22 )
Ep
 1−ν3p νp3 ν3p νp3 +νp 
Ep
− Ep
0
 νp3 ν3p +νp 1−ν3p νp3 
S =
 − Ep Ep
0 

ε
1
0 0 Gp
 1−ν3p νp3 ν3p νp3 +νp

Ep E3 Ep E3 0
1  νp3 ν3p +νp 1−ν3p νp3

C = S −1 =  0 
ε ε ∆s  Ep E3 Ep E3 
0 0 ∆s G p
1 − νp νp − νp3 ν3p − ν3p νp3 − νp νp3 ν3p − νp ν3p νp3
with ∆s =
Ep Ep E3

1 νp3 (νp + 1)
σ33 = (ε11 + ε22 )
∆s E p2

213
A. S TIFFNESS AND COMPLIANCE MATRICES

A.3.2 Plane stress

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 stiffness matrix.

νp3
ε33 = − (σ11 + σ22 )
Ep
 
E p−1 −νp E p−1 0
S =  −νp E p−1 E p−1
 
0 
σ
0 0 G p−1
 
Ep νp E p 0
−1 1
C =S =  νp E p Ep 0 
σ σ 1 − νp νp
0 0 (1 − νp νp )G p
νp3
ε33 = − (ε11 + ε22 )
1 − νp

A.4 Linear elastic isotropic material

The linear elastic material behavior can be described with the material stiffness matrix C or
the material compliance matrix S. These matrices can be written in terms of the engineering
elasticity parameters E and ν.

 
1 −ν −ν 0 0 0

 −ν 1 −ν 0 0 0 

1
 −ν −ν 1 0 0 0


S=  
E 0 0 0 2(1 + ν) 0 0 
 
 0 0 0 0 2(1 + ν) 0 
0 0 0 0 0 2(1 + ν)
E
C = S −1 =
(1 + ν)(1 − 2ν)
 
1−ν ν ν 0 0 0
 ν 1 − ν ν 0 0 0 
 
 ν ν 1−ν 0 0 0 
 
 1 
 0 0 0 2
(1 − 2ν) 0 0 
 1 
 0 0 0 0 2 (1 − 2ν) 0 
1
0 0 0 0 0 2 (1 − 2ν)

214
 A.4. Linear elastic isotropic material

A.4.1 Plane strain

σ33 = ν(σ11 + σ22 )

 
1 − ν −ν 0
1+ν 
S = −ν 1 − ν 0 
ε E
0 0 2
 
1−ν ν 0
−1 E
C =S =  ν 1−ν 0 
ε ε (1 + ν)(1 − 2ν) 1
0 0 2 (1 − 2ν)
E
σ33 = ν(ε11 + ε22 )
(1 + ν)(1 − 2ν)

It is immediately clear that problems will occur for ν = 0.5, which is the value for incom-
pressible material behavior.

A.4.2 Plane stress

ν
ε33 = − (σ11 + σ22 )
E
 
1 −ν 0
1 
S = −ν 1 0 
σ E
0 0 2(1 + ν)
 
1 ν 0
E 
C = S −1 = ν 1 0 
σ σ 1 − ν2 1
0 0 2 (1 − ν)
ν
ε33 = − (ε11 + ε22 )
1−ν

215
 APPENDIX
B
M ECHANICAL DESIGN PROBLEMS

In this appendix, details of the examples presented in Chapter 8 are shown. The general
relations for the analytical solutions can be found in chapter 8 and are specified here. The
integration constants are calculated for the specific boundary conditions and loading, and
are given here.

217
B. M ECHANICAL DESIGN PROBLEMS

B.1 Governing equations and general solution

1 1
u r,r r + u r,r − ζ2 2 u r = f (r )
r r
s
Bp
with ζ=
Ap
ρ ¡ ¢ A p +Q p Ap − Bp 1
and f (r ) = ü r − q r + α(∆T ),r + α∆T
Ap Ap Ap r

orthotropic material :

general solution u r = c1 r ζ + c2 r −ζ + ū r
ū r
εr r = c1 ζr ζ−1 − c2 ζr −ζ−1 + ū r,r ; εt t = c1 r ζ−1 + c2 r −ζ−1 +
r
ζ−1 −ζ−1 ū r
σr r = (A p ζ +Q p )c1 r − (A p ζ −Q p )c2 r + A p ū r,r +Q p − (A p +Q p )α∆T
r
ū r
σt t = (Q p ζ + B p )c1 r ζ−1 − (Q p ζ − B p )c2 r −ζ−1 +Q p ū r,r + B p − (B p +Q p )α∆T
r

isotropic material :

c2
general solution u r = c1 r + + ū r
r
ū r
εr r = c1 − c2 r −2 + ū r,r ; εt t = c1 + c2 r −2 +
r
c2 ū r
σr r = (A p +Q p )c1 − (A p −Q p ) 2 + A p ū r,r +Q p − (A p +Q p )α∆T
r r
c2 ū r
σt t = (Q p + A p )c1 − (Q p − A p ) 2 +Q p ū r,r + A p − (A p +Q p )α∆T
r r

218
 B.2. Disc, edge displacement

B.2 Disc, edge displacement

r
ub
f (r ) = 0 → ū r = 0
b
a u r (r = b) = u b
z r
σr r (r = a) = 0;

Figure B.1: Edge displacement of circular disc

orthotropic material :

general solution u r = c1 r ζ + c2 r −ζ

εr r = c1 ζr ζ−1 − c2 ζr −ζ−1 ; εt t = c1 r ζ−1 + c2 r −ζ−1

σr r = (A p ζ +Q p )c1 r ζ−1 − (A p ζ −Q p )c2 r −ζ−1

σt t = (Q p ζ + B p )c1 r ζ−1 − (Q p ζ − B p )c2 r −ζ−1

(A p ζ −Q p )b ζ u b (A p ζ +Q p )b ζ a 2ζ u b
c1 = ; c2 =
(A p ζ +Q p )a 2ζ + (A p ζ −Q p )b 2ζ (A p ζ +Q p )a 2ζ + (A p ζ −Q p )b 2ζ

isotropic material :

c2
general solution u r = c1 r +
r

εr r = c1 − c2 r −2 ; εt t = c1 + c2 r −2

c2
σr r = (A p +Q p )c1 − (A p −Q p )
r2
c2
σt t = (Q p + A p )c1 − (Q p − A p ) 2
r

(A p −Q p )b (A p +Q p )ba 2
c1 = u
2 b
; c2 = ub
(A p +Q p )a 2 + (A p −Q p )b (A p +Q p )a 2 + (A p −Q p )b 2

219
B. M ECHANICAL DESIGN PROBLEMS

B.3 Disc/cylinder, edge load

r
pe
f (r ) = 0 → ū r = 0
b
a σr r (r = a) = −p i
z r
pi σr r (r = b) = −p e

Figure B.2: *
Cross-section of a thick-walled circular cylinder

orthotropic material :

general solution u r = c1 r ζ + c2 r −ζ

εr r = c1 ζr ζ−1 − c2 ζr −ζ−1 ; εt t = c1 r ζ−1 + c2 r −ζ−1

σr r = (A p ζ +Q p )c1 r ζ−1 − (A p ζ −Q p )c2 r −ζ−1

σt t = (Q p ζ + B p )c1 r ζ−1 − (Q p ζ − B p )c2 r −ζ−1

1 a ζ+1 p i − b ζ+1 p e 1 a ζ+1 b 2ζ p i − b ζ+1 a 2ζ p e

c1 = ; c2 =
A p ζ +Q p b 2ζ − a 2ζ A p ζ −Q p b 2ζ − a 2ζ

isotropic material :

c2
general solution u r = c1 r +
r

εr r = c1 − c2 r −2 ; εt t = c1 + c2 r −2

c2
σr r = (A p +Q p )c1 − (A p −Q p )
r2
c2
σt t = (Q p + A p )c1 − (Q p − A p ) 2
r

1 1 1 a2b2
c1 = 2 2
(p i a 2 − p e b 2 ) ; c2 = (p i − p e )
A p +Q p b − a A p −Q p b 2 − a 2

220
 B.4. Rotating solid disc

B.4 Rotating solid disc

r
ω ρ 2
f (r ) = − ω r
b Ap
z r u r (r = 0) 6= ∞
σr r (r = b) = 0

Figure B.3: A rotating solid disc

orthotropic material :

1 1
u r = c1 r ζ + c2 r −ζ − βr 3 with β= ρω2
Ap 9−ζ

3A p +Q p
σr r = (A p ζ +Q p )c1 r ζ−1 − (A p ζ −Q p )c2 r −ζ−1 − βr 2
Ap
3Q p + B p
σt t = (Q p ζ + B p )c1 r ζ−1 − (Q p ζ − B p )c2 r −ζ−1 − βr 2
Ap

3A p +Q p
c2 = 0 ; c1 = βb −ζ+3
A p (A p ζ +Q p )

isotropic material :

c2 1 ρ 2 3 c2 1 1
u r = c1 r + − ω r = c1 r + − βr 3 with β= ρω2
r 8 Ap r Ap 8

c2 (3A p +Q p ) 2
σr r = (A p +Q p )c1 − (A p −Q p ) − βr
r2 Ap
c2 (A p + 3Q p ) 2
σt t = (A p +Q p )c1 + (A p −Q p ) 2 − βr
r Ap

(3A p +Q p )
c2 = 0 ; c1 = βb 2
A p (A p +Q p )

221
B. M ECHANICAL DESIGN PROBLEMS

B.5 Rotating disc with central hole

r
ω ρ 2
f (r ) = − ω r
b Ap
a z r σr r (r = a) = 0
σr r (r = b) = 0

Figure B.4: A rotating disc with central hole

orthotropic material :

1 1
u r = c1 r ζ + c2 r −ζ − βr 3 with β= ρω2
Ap 9−ζ

3A p +Q p
σr r = (A p ζ +Q p )c1 r ζ−1 − (A p ζ −Q p )c2 r −ζ−1 − βr 2
Ap
3Q p + B p
σt t = (Q p ζ + B p )c1 r ζ−1 − (Q p ζ − B p )c2 r −ζ−1 − βr 2
Ap
µ ¶
3A p +Q p b ζ+3 − a ζ+3
c1 = β
A p (A p ζ +Q p ) b 2ζ − a 2ζ
µ 2ζ−2 ζ+1 ¶
3A p +Q p a b − a ζ+1 b 2ζ−2
c2 = (a 2 b 2 )β
A p (A p ζ −Q p ) b 2ζ − a 2ζ

isotropic material :

c2 1 ρ 2 3 c2 1 1
u r = c1 r + − ω r = c1 r + − βr 3 with β= ρω2
r 8 Ap r Ap 8

c2 (3A p +Q p ) 2
σr r = (A p +Q p )c1 − (A p −Q p ) − βr
r2 Ap
c2 (A p + 3Q p ) 2
σt t = (A p +Q p )c1 + (A p −Q p ) 2 − βr
r Ap

(3A p +Q p ) (3A p +Q p )
c1 = (a 2 + b 2 )β ; c2 = (a 2 b 2 )β
A p (A p +Q p ) A p (A p −Q p )

222
 B.6. Rotating disc fixed on rigid axis

r
ω
b ρ 2
f (r ) = − ω r
a r Ap
z
u r (r = a) = 0
σr r (r = b) = 0

Figure B.5: Disc fixed on rigid axis

B.6 Rotating disc fixed on rigid axis

orthotropic material :
1 1
u r = c1 r ζ + c2 r −ζ − βr 3 with β= ρω2
Ap 9−ζ

3A p +Q p
σr r = (A p ζ +Q p )c1 r ζ−1 − (A p ζ −Q p )c2 r −ζ−1 − βr 2
Ap
3Q p + B p
σt t = (Q p ζ + B p )c1 r ζ−1 − (Q p ζ − B p )c2 r −ζ−1 − βr 2
Ap

β
c1 =
(A p ζ +Q p )b ζ+1 a −ζ+1 + (A
½ p ζ −Q p )b
−ζ+1 a ζ+1
¾
3A p +Q p 4 −ζ+1 A p ζ −Q p −ζ+1 4
b a + b a
Ap Ap
β
c2 =
(A p ζ +Q p )b ζ+1½a −ζ+1 + (A p ζ −Q p )b −ζ+1 a ζ+1 ¾
A p ζ +Q p ζ+1 4 3A p +Q p 4 ζ+1
b a − b a
Ap Ap

isotropic material :
c2 1 ρ 2 3 c2 1 1
u r = c1 r + − ω r = c1 r + − βr 3 with β= ρω2
r 8 Ap r Ap 8

c2 (3A p +Q p ) 2
σr r = (A p +Q p )c1 − (A p −Q p ) − βr
r2 Ap
c2 (A p + 3Q p ) 2
σt t = (A p +Q p )c1 + (A p −Q p ) 2 − βr
r Ap
½ ¾
β 3A p +Q p 4 A p −Q p 4
c1 = b + a
(A p +Q p )b 2 + (A p −Q p )a 2 ½ Ap Ap ¾
β A p +Q p 4 2 3A p +Q p 2 4
c2 = a b − a b
(A p +Q p )b 2 + (A p −Q p )a 2 Ap Ap

223