A constitutive model for

polymers: Formulation
and Integration algorithm
Negar Bahramsari
Mestrado em Engenharia Matemática
Departamento de Matemática
2014

Orientador
Francisco Manuel Andrade Pires, Professor Auxiliar, FEUP

Coorientador
Sílvio Marques de Almeida Gama, Professor Associado, FCUP
Todas as correções determinadas
pelo júri, e só essas, foram efetuadas.

O Presidente do Júri,

Porto, ______/______/_________
Dedicated to my husband ...
 Acknowledgements

I would like to gratefully acknowledge Prof. Pires for accepting to be my

thesis supervisor and letting me integrate is his research group. I would
also like to thank Prof. Gama for being my thesis co-supervisor and his
support during accomplishment of this work. Eng. Mirkhalaf, a PhD
student from research group of Prof. Pires, helped me to get acquainted
with concepts of continuum mechanics and Finite Element Method and
also constitutive modelling of polymers. I am grateful for his support and
his comments while I was involved in my thesis.
Being far from my country has been difficult and my friends company has
been helping me to keep moving. I thank them all although I can not name
all of them.
At last, not the least, I am going to deeply thank my husband and my
family for their undeniable support without which my studies could be
much harder to finish.
 Abstract

The objective of this thesis is to develop an elasto-viscoplastic constitutive

model in order to characterize the deformation behaviour of polymeric ma-
terials. More importantly, it is intended to incorporate several numerical
techniques to solve the equilibrium equation obtained from kinematic and
material constitutive relations. For this purpose, the thesis starts with a
global overview of the state-of-art on the present topic in Chapter 1. In
Chapter 2, the main concepts of the Continuum Mechanics Theory and the
Finite Element Method are briefly reviewed. This chapter is followed by
development of the BPA based constitutive model in Chapter 3. In Chap-
ter 4, the most important contribution of this work is introduced, which is
derivation of the integration algorithm and finite element implementation
of the model. In order to assess the capability of the model to predict the
typical deformation behaviour of polymeric materials and the efficiency
of the integration algorithm and the code, some numerical examples are
provide in Chapter 5. The document finishes with the presentation of the
main conclusions of this work in Chapter 6 and, in addition, suggestions
for future research are made.
Contents

Nomenclature vii

1 Introduction 1
1.1 General introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Modeling of the mechanical behavior of polymers . . . . . . . . . . . . 2
1.3 Scope and outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Continuum mechanics and finite element method 4

2.1 Continuum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Kinematics of deformation . . . . . . . . . . . . . . . . . . . . . 4
2.1.1.1 Motion . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1.2 The deformation gradient . . . . . . . . . . . . . . . . 5
2.1.1.3 Strain measures . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1.4 The velocity gradient . . . . . . . . . . . . . . . . . . . 7
2.1.1.5 Stress measures . . . . . . . . . . . . . . . . . . . . . . 7
2.1.2 Fundamental laws . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2.1 Conservation of mass . . . . . . . . . . . . . . . . . . . 9
2.1.2.2 Momentum balance . . . . . . . . . . . . . . . . . . . 9
2.1.2.3 The first and second principles of thermodynamics . . 9
2.1.2.4 The Clausius-Duhem inequality . . . . . . . . . . . . . 10
2.1.3 The quasi-static initial boundary value problem . . . . . . . . . 11
2.1.3.1 The material version . . . . . . . . . . . . . . . . . . . 12
2.2 Displacement-based finite elements . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Spatial discretisation . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Temporal discretisation. The non-linear incremental finite ele-
ment procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
v CONTENTS

3 BPA based model 19

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.1 Multiplicative kinematics . . . . . . . . . . . . . . . . . . . . . . 20
3.2.2 Strain measure . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.3 Flow potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.4 Stress decomposition . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4 The integration algorithm of the model 25

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 State update . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.1 Return mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3 Consistent tangent operator . . . . . . . . . . . . . . . . . . . . . . . . 35

5 Numerical examples 42
5.1 Compression of a cube (plane strain compression) . . . . . . . . . . . . 42
5.2 Cylinder upsetting (axisymmetric compression) . . . . . . . . . . . . . 44
5.3 Notched bar compression . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6 Conclusions and suggestions for future research 48

References 52
List of Figures

3.1 Elastic-Plastic decomposition of the total deformation gradient-Reproduced

from De Souza Neto et al. (2008) . . . . . . . . . . . . . . . . . . . . . 20
3.2 Polar decomposition of the deformation gradient-Reproduced from De
Souza Neto et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5.1 The geometry and mesh of the plane strain compression test. . . . . . . 43
5.2 The stress-strain curve for the compression of the cube. . . . . . . . . . 43
5.3 The stress-strain curve for the compression of the cylinder. . . . . . . . 44
5.4 The geometry and mesh of the compression test on a notched bar . . . 45
5.5 Contour plot of effective stress at ten percent of the deformation. . . . 46
5.6 Contour plot of effective stress at twenty percent of the deformation. . 46
5.7 Contour plot of effective stress at thirty percent of the deformation. . . 46
5.8 Contour plot of effective stress at half of the deformation. . . . . . . . . 47
5.9 Contour plot of effective stress at the end of the deformation. . . . . . . 47
List of Tables

5.1 Material properties for PMMA required for the model . . . . . . . . . . 42

5.2 Local convergence table for cube compression simulation . . . . . . . . 44
5.3 Local convergence table for cylinder compression simulation . . . . . . 45
Chapter 1

Introduction

1.1 General introduction

Polymers are a group of materials which, from a microscopic point of view, contain
a large number of structural units linked together by the same kind of joints. Chain
like structures are often observed in these substances. It can be claimed that form the
beginning of time polymers have existed in the natural world. Artificial or man-made
polymers have been under study from 1832. Amazingly, the polymer industry nowa-
days is larger than the combination of aluminum, copper and steel industries.
Polymers are a group of materials which have remarkable attractive qualities form
processability to mechanical and thermal properties. Therefore, although they have
not been introduced for a long time (from the beginning of 18th century), polymers
are being widely used in structures and other applications. It is worth noting that de-
spite the widespread use of polymers, their mechanical behavior is not yet thoroughly
understood.
Polymers have a variety of applications that are far more than any other class of mate-
rials existing for mankind. The vast area of applications includes adhesives, coatings,
foams, and packaging materials to textile and industrial fibers, composites, electronic
devices, bio-medical devices, optical devices, and precursors used for high-tech ceram-
ics which have been recently developed.
2 1.2 Modeling of the mechanical behavior of polymers

1.2 Modeling of the mechanical behavior of poly-

mers
During the last few decades, many researchers have devoted their work to develop
constitutive models for different materials such as metals and polymers. A constitu-
tive model could be briefly defined as a set of mathematical relations which determine
the relation between applied forces and displacements of any single point of the body
under study. Here, we provide a very brief summary of the most well established
constitutive models proposed for polymers through the last decades.
The usage of polymers is increasing in different structures. Hence, the academic com-
munity has dedicated remarkable effort to the development of constitutive models
capable of characterizing the behavior of polymers. It can be said that the initiation
of the efforts to determine the behavior of polymers dates back to 1930s. Eyring
(1936) proposed a molecular theory for the yield stress of amorphous polymers, con-
sidering the yield behavior as a thermally activated process. Temperature and strain
rate effect is accounted for in the theory. Mooney (1940) proposed a strain energy
function for rubber elastic materials. Haward and Tackray (1968) developed a one-
dimensional constitutive model for glassy polymers. The work could be considered as
the first constitutive model proposed for predicting the deformation behavior of glassy
polymers. The three dimensional version of Haward and Tackray model was proposed
by Boyce et al. (1988). A modified version of this model was later formulated by Wu
and van der Giessen (1994). The other constitutive model which is able to predict
the typical deformation behavior of polymeric materials is the generalised compressible
Leonov model, which has been proposed by Baaijens (1991) and extended by Tervoort
et al. (1998) and Govaert et al. (2000). For a review on finite element simulation
of polymers, the reader is referred to Mackerle (1997)and Mackerle (2003). The
phenomenologiacl constitutive models developed for amorphous polymers could be, in
many cases, used for semi-crystalline polymers as well. The molecular orientation of
polymers evolves during large deformations and the corresponding strengthening is,
for some polymers, greatly dependent on strain state. Time dependent deformation
and failure of polymers have been studied using linear and non-linear visco-elastic and
visco-plastic FEM models (Mackerle (1997); Mackerle (2003)).

1.3 Scope and outline

The aims of the work can be expressed as follows:
3 1.3 Scope and outline

• To study the non-linear continuum solid mechanics theory;

• To study non-linear finite element method (FEM);

• To develop a constitutive model based on BPA(Mary C. Boyce, David M. Parks,

Ali S. Argon) model (Boyce et al. , 1988);

• To use nonlinear finite element method to solve the equilibrium equations ob-
tained from continuum mechanics relations and the constitutive model;

• To derive the integration algorithm of the model developed in the previous sec-
tion in order to be used in finite element;

• To run numerical examples with the implemented model within finite element.

In the next chapter of the work, continuum mechanics and finite element method
are explained. It should be mentioned that the objective is by no means providing
a comprehensive review of the subjects and it is intended to provide the main rela-
tions which are used in this work. In Chapter 3, the analytical formulation of the
model based on BPA model is provided. The numerical treatment of the governing
constitutive equations and the integration algorithm of the model which includes the
derivation of state update procedure and also consistent tangent operator, is given in
Chapter 4. Some numerical examples are conducted using the implemented consti-
tutive model in Chapter 5. Finally, in Chapter 6, the results and conclusions of this
work are presented.
Chapter 2

Continuum mechanics and finite

element method

In order to characterize the behaviour of polymeric materials, we need to either de-

velop new constitutive models or use some existing models capable of predicting typical
behaviour of polymers. Before developing or using any constitutive model, it is re-
quired to be familiarized with basic concepts of standard continuum mechanics. Once
the basic continuum mechanics is learned, a suitable numerical approach is needed in
order to solve the equilibrium equation obtained from continuum mechanics. In this
work, finite element method (FEM) is used for solution of equilibrium problem for
solid polymeric materials.

2.1 Continuum Mechanics

The concepts of continuum mechanics are presented in this section. In order to have
a more comprehensive description of the topic, you can refer to Shabana (2008),
Criesfield (2000), Holzapfel (2000).

2.1.1 Kinematics of deformation

In this section, motion of particle, the definition of deformation gradient, different
strain measures, the velocity gradient and different stress measures are explained.

2.1.1.1 Motion

p is a particle of a body Ω0 . The body is limited to its boundary ∂Ω0 . A motion

ϕ applies to the body and after application of the motion the particle P has a new
5 2.1 Continuum Mechanics

position x. The new position is function of time t as well.

x = ϕ(p, t). (2.1)

From the new position of the particle to the original position, a displacement filed is
given by:
u(p, t) = ϕ(p, t) − p. (2.2)

If we assume that the motion ϕ is invertible, relation (2.3) can be written:

p = ϕ−1 (x, t). (2.3)

According to above mentioned equations, two different configurations are being as-
sumed. The initial configuration (undeformed configuration or material configuration)
in which the particle p exists. The other configuration is current configuration (de-
formed configuration or spatial configuration).
In order to obtain the velocity of particle p, relation (2.4) could be used.

∂ϕ(p, t)
ẋ = , (2.4)
∂t
Using relations (2.3) and (2.4), one can write:

υ(x, t) = ẋ(ϕ−1 (x, t), t), (2.5)

which is the velocity of the material particle x at time t.

2.1.1.2 The deformation gradient

Deformation gradient is defined as the gradient of motion ϕ in order to particle p:

∂xt
F(p, t) = ∇p ϕ(p, t) = . (2.6)
∂p

By substituting (2.2) in (2.6) we can have:

F = I + ∇p u, (2.7)

where I is the second order identity tensor. It should be emphasized that in relations
(2.6) and (2.7), the deformation gradient is defined in the reference configuration and
the operator ∇p is generally a material gradient operator. If it is required to have
the deformation gradient in the spatial configuration, we should use spatial gradient
6 2.1 Continuum Mechanics

operator, ∇x . Equation (2.8) gives the deformation gradient at current configuration.

F(x, t) = [∇x ϕ−1 (x , t)]−1 = [I − ∇x u]−1 . (2.8)

The change in the volume of the body under motion is equivalent to the determinant of
the deformation gradient and represented by J and also called deformation Jacobian.

J = det F. (2.9)

Polar decomposition of the deformation gradient gives:

F = RU = VR, (2.10)

where R is a proper orthogonal tensor called rotation tensor. In relation (2.10), U

and V are symmetric positive definite tensors and called right and left stretch tensors,
respectively. The right and left stretch tensors are also represented as:
√ √
U= C; V = B, (2.11)

where C and B are, respectively, right and left Cauchy-Green strain tensors defined
by
C = U2 = FT F; B = V2 = FFT . (2.12)

2.1.1.3 Strain measures

A particle is strained when, as deformation proceeds, the relative position of the

material particle changes. There are different approaches to measure the strain of a
material particle. Some strain measures are used more frequently than the others due
to their physical meaning and mathematical convenience.
Lagrangian strain tensor is one of the most important class of strain measures:
(1
(Um − I), m 6= 0
E(m) = m (2.13)
ln[U], m=0

where, U is the right stretch tensor; I is second order identity tensor; m is a real
number and ln[•] is tensorial logarithm. Similarly, another important family of strain
7 2.1 Continuum Mechanics

tensors, called Eulerian strain tensors, are defined by using the left stretch tensor, V.
(1
(Vm − I), m 6= 0
ε(m) = m (2.14)
ln[V], m=0
In this work, Eulerian strain tensor is used and the logarithmic strain tensor (m = 0)
is considered. It is also worth noting that when the deformation is just a rotation (i.e.,
F = R), Eulerian strain tensor and Lagrangian strain tensor are null.

U = I =⇒ E(m) = 0, (2.15)

V = I =⇒ ε(m) = 0. (2.16)

2.1.1.4 The velocity gradient

Velocity gradient is defined by:

L = Ox υ, (2.17)
or, alternatively by
L = ḞF−1 . (2.18)
The velocity gradient is additively composed of symmetric and skew part:

L = D + W, (2.19)

where,
1
1

D = sym(L) = L + LT , W = skew(L) = L − LT . (2.20)
2 2
The symmetric part of the deformation gradient, D, is called rate of deformation
(stretching) tensor and the skew part, W, is called spin tensor. Stretching tensor is
associated with straining whereas spin tensor is associated with rigid velocities.

2.1.1.5 Stress measures

The Cauchy stress, σ, the Kirchhoff stress, τ , and the first Piola-Kirchhoff stress, P,
are three important stress measures. Which, Cauchy stress tensor or true stress is
defined as:
t = σn, (2.21)
where t is the surface traction and n is its associated normal vector. The Cauchy
stress tensor is additively composed of hydrostatic term and deviatoric part

σ = s + pI, (2.22)
8 2.1 Continuum Mechanics

where s is the deviatoric stress and p is hydrostatic pressure:

s = dev (σ) = Id : σ, (2.23)

where Id is the deviatoric fourth order identity tensor:

1
Id = Is − (I ⊗ I) , (2.24)
3
where Is is the symmetric fourth order identity and I is the second order identity. Is
and I in component form can be shown as:
1 
Iijkl = δik δjl + δil δjk , (2.25)
2

Iij = δij , (2.26)

where, δ is the Kronecker delta. The hydrostatic pressure is given by:

1
p = tr (σ) . (2.27)
3
The second stress measure is the Kirchhoff stress tensor denoted by τ defined as

τ = Jσ. (2.28)

Similarly to the Cauchy stress, the Kirchhoff stress tensor can also be split into two
parts, i.e.,
τ = τ d + τh I, (2.29)

where τ d = dev[τ ] and τh = 13 tr[τ ] are, respectively, the deviatoric and hydrostatic
(or spherical) parts. The last stress measure mentioned here is first Piola-Kirchhoff
stress tensor, denoted by P, also known as nominal stress. The Piola-Kichhoff stress
is defined by:
P = JσF−T = τ F−T . (2.30)

2.1.2 Fundamental laws

The quantities and definitions mentioned above are essential mathematical relations
of deformation. The fundamental laws of continuum mechanics governing the physical
phenomena of deformation provide relations between defined and expressed quantities.
9 2.1 Continuum Mechanics

2.1.2.1 Conservation of mass

The conservation of mass postulate implies:

ρ̇ + divx (ρu̇) = 0, (2.31)

where ρ is the density at the deformed configuration.

2.1.2.2 Momentum balance

The momentum balance, also referred to as strong form of the equilibrium equation,
of any given body can be expressed as

divx σ + b = ρü, (2.32)

where b denotes the body force vector in the deformed configuration. The equilibrium
equation (2.32) needs to fulfill the following boundary condition:

t = σn, (2.33)

where t is a tension vector applied on the boundary of the body.

2.1.2.3 The first and second principles of thermodynamics

The first principle of thermodynamics postulates that the energy must be conserved.
This can be mathematically expressed as

ρė = σ : D + ρr − divx q, (2.34)

where e, r and q are, respectively, the specific internal energy, the density of heat
production and the heat flux. Throughout this work, only processes with constant
temperature will be considered. In this case, the first principle reduces to

ρė = σ : D. (2.35)

The equation above states that the rate of internal energy per unit deformed volume
must equal the stress power, σ : D, per unit deformed volume. Making use of the
relation below,
ρ̄ = Jρ, (2.36)
10 2.1 Continuum Mechanics

where ρ̄ denotes the reference density, it is possible to rewrite equation (2.35) as

ρ̄ė = τ : D. (2.37)

The second principle of thermodynamics is associated with the so-called irreversibility

of entropy production, expressed by the following inequality:

ρT ṡ + divx q − ρr ≥ 0, (2.38)

where s denotes the entropy and T is the temperature. Similar to the case of the first
principle of thermodynamics, if only isothermal processes are considered, Equation
(2.38) is then given by
ρT ṡ ≥ 0. (2.39)

Pre-multiplying the equation (2.39) by J, we have

ρ̄T ṡ ≥ 0. (2.40)

2.1.2.4 The Clausius-Duhem inequality

Firstly, we introduce the Helmholtz free energy, ψ, defined by

ψ = e − Ts (2.41)

Re-arranging the equation (2.41) and differentiating with respect to time, we have

T ṡ = ė − ψ̇. (2.42)

We remark that the temperature has been assumed constant, thus, Ṫ = 0 Using the
equation (2.35), we conclude that

ρ̄T ṡ = τ : D − ρ̄ψ̇, (2.43)

with
τ : D − ρ̄ψ̇ ≥ 0. (2.44)

The fundamental inequality above is called Clausius-Duhem inequality.

11 2.1 Continuum Mechanics

2.1.3 The quasi-static initial boundary value problem

The fundamental laws presented in the last section allow us to define an initial bound-
ary value problem (IBVP) associated with the description of a given deformation
process. The solution of the IBVP delivers the prediction of how a given solid will me-
chanically behave when subjected to certain boundary conditions. Within the scope
of this work, only quasi-static problems will be addressed, hence, all inertia effects will
be neglected. Thus, the equilibrium equation (2.32), in the strong form, is re-written
to be given by
divx σ = 0. (2.45)

Multiplying the equation (2.45) by a virtual displacement, η, and integrating over the
volume, we have Z
(divx σ)T ηdV = 0. (2.46)
ϕ(Ω)

After some straightforward operations, (2.46) becomes

Z
[σ : Ox η − (divx σ · η)]dV = 0. (2.47)
ϕ(Ω)

Making use of the divergence theorem, we have

Z Z
σ : Ox ηdV − (σ · n)T ηdA = 0. (2.48)
ϕ(Ω) ϕ(∂Ω)

Finally, substituting Equation (2.33) into Equation (2.48) leads to

Z Z
σ : Ox ηdV − t · ηdA = 0. (2.49)
ϕ(Ω) ϕ(∂Ω)

Equation (2.49) is called weak form of the equilibrium equation. The use of the weak
form can significantly facilitate the use of efficient numerical methods for the solution
of the structural IBVP. With the definition of weak equilibrium at hand, we can define
the quasi-static IBVP, in the spatial description, as follows.
Problem 2.1
Given a prescribed deformation gradient history

F(t) = I + Op u(p, t) (2.50)

and the Cauchy stress, at each point of the body expressed as

σ(t) = σ(F(t, α(t)) (2.51)

12 2.2 Displacement-based finite elements

obtained from the solution of the constitutive initial boundary value problem where
α is the set of internal variables associated with the material, find a kinematically
admissible displacement function, u ∈ K such that the equation
Z Z
σ(t) : Ox ηdV − t(t) · ηdA = 0, (2.52)
ϕ(Ω) ϕ(∂Ω)

is satisfied for all t ∈ [t0 , tn ] and for all η ∈ νt . The set of kinematically admissible
displacements, K , and the space of virtual displacements at time t, νt , are respectively
given by
K = {u : Ω → U|u(p, t) = ū(p, t), t ∈ [t0 , tn ], p ∈ ∂ϕu }, (2.53)

νt = {η : ϕ → U|η = 0 ∈ ∂ϕu (t)}. (2.54)

Unfortunately, analytical solutions for the problem defined above exist only for a re-
stricted set of special cases. For accurate predictions of the mechanical behavior of
solids in the general case, the use of numerical methods is therefore indispensable.

2.1.3.1 The material version

For reference, the material version of the weak form of the equilibrium equation is also
herein provided, which reads
Z Z
P : ∇p ηdV − t̄ · ηdA = 0, (2.55)
ϕ ∂ϕ

where t̄ is the surface traction per unit reference area.

2.2 Displacement-based finite elements

In this section, the general concepts of the finite element method formulated with a
displacement-based approach will be briefly addressed. The finite element method has
been chosen in this work as the base numerical tool mainly due to its versatility and
its high effectiveness when adopted for the simulation of deformation processes.

2.2.1 Spatial discretisation

As stressed out above, the solution of Problem 2.1 often requires the use of some sort
of numerical strategy. Within a typical finite element framework, the field variables
13 2.2 Displacement-based finite elements

are discretised through the so-called interpolation or shape functions. In the case of
displacement-based finite elements, the interpolated field variable are the displace-
ments. Within a given element e, the interpolation is assumed to be
nX
node
(e)
u(x) ≡ Ni (x )ui , (2.56)
i=1

(e)
where Ni (x ) is the shape function associated with node i (evaluated at x ) and nnode is
the number of nodes of the element. In similar manner, a global interpolation function
can also be set, that is,
npoin
X
u(x) ≡ Ngi (x )ui , (2.57)
i=1

where npoin is the total number of nodes of the finite element mesh and Ngi (x ) is the
global interpolation matrix which can be represented by

Ng (x ) = [diag[Ng1 (x )] diag[Ng2 (x )] · · · diag[Ngnpoin (x )]] (2.58)

where diag[Ngi ] denotes a ndim × ndim diagonal matrix defined

 g 
Ni 0 · · · 0
 0 Ng · · · 0 
g  i 
diag[Ni ] =  .. .. . . ..  . (2.59)
 . . . . 
0 0 · · · Ngi

At this point, it is also convenient to define the global vector of nodal displacements,
given by
u = [u11 , · · · un1 dim , · · · · · · u1npoin , · · · unnpoin
dim
]T . (2.60)
With the above matrix notation at hand, equation (2.55) can be rephrased to be given
by
u(x ) = Ng (x )u, (2.61)
where the equation above represents the interpolation of the displacement field by
means of discrete functions. Analogously, we can write the field of virtual displace-
ments to be given by
η(x ) = Ng (x )η. (2.62)
We also define the global discrete symmetric gradient matrix, Bg , which in the case
of plane stress and plane strain problems assumes the form
 g 
N1,1 0 Ng2,1 ··· Ngnpoin ,1 0
g
Bg =  0 N1,2 0 ··· 0 Ngnpoin ,2  , (2.63)
Ng1,2 Ng1,1 Ng2,2 ··· Ngnpoin ,2 Ngnpoin ,1
14 2.2 Displacement-based finite elements

where use of the following notation has been made:

∂(·)i
(·)i,j = . (2.64)
∂xj

For completeness, the global discrete full gradient operator, Gg , is also provided herein,
whose format in plane stress and plane strain analyses is given by
 g 
N1,1 0 Ng2,1 0 ··· Ngnpoin ,1 0
 0 Ng1,1 0 Ng2,1 ··· 0 Ngnpoin ,1 
G =
g
Ng1,2 0 Ng2,2 0
. (2.65)
··· Ngnpoin ,2 0 
0 Ng1,2 0 Ng2,2 ··· 0 Ngnpoin ,2

2.2.2 Temporal discretisation. The non-linear incremental

finite element procedure
In practical engineering applications, it is often required the modeling of materials that
are dependent of the deformation history. Such materials are called path dependent
and, regardless whether they take strain rate effects into account or not, a suitable
temporal discretisation needs to be performed. Within the context of most of this
work, a pseudo-time discretisation between the time increments [tn , tn+1 ] will be con-
sidered for which a fully implicit scheme is adopted.
For a generic path-dependent material model, an incremental constitutive function,
σ̂, is assumed to exist, i.e.,

σ n+1 = σ̂(Fn+1 , αn ). (2.66)

In practice, the function σ̂ is associated with an integration algorithm that computes

the material behavior for a given deformation gradient Fn+1 and set of internal vari-
ables, αn , which remain constant within [tn , tn+1 ]. Accordingly, a similar incremental
function for the set of internal variables is also defined:

αn+1 = α̂(Fn+1 , αn ). (2.67)

Up to this point, it suffices to leave the incremental constitutive functions σ̂ and α̂

unspecified for the sake of generality.

The non-linear incremental finite element equation

The definition of the incremental constitutive function of Section 2.2.2, combined
with the spatial discretisation presented in Section 2.2.1, allows the definition of the
15 2.2 Displacement-based finite elements

incremental finite element equilibrium equation, obtained after some straight forward
substitutions and rearrangements from Equation (2.49):

r(un+1 ) ≡ fint (un+1 ) − fext = 0, (2.68)

where fint and fext are, respectively, the internal and external force vectors, defined for
a given element e as Z
int
f = BT σ̂(Fn+1 , α)dV, (2.69)
ϕn+1 (ϕ(e) )
Z
ext
f = NT t n+1 dA, (2.70)
ϕn+1 (∂ϕ(e) )

Within the scope of this work, Equation (2.64) will be generally non-linear and there-
fore requires an adequate (numerical) method for its solution. Considering an incre-
mental scheme, in which a given fraction of the external prescribed load is applied at
each increment, Equation (2.64) is solved as summarized in Box 1.
Remark 2.1.
In practice, the external force is computed by the expression
ext
fext
n+1 = λn+1 f̄ , (2.71)
ext
where λn+1 is the prescribed load factor at time tn+1 and f̄ is computed only once
at the first iteration of the incremental procedure through the expression
Z
ext
f̄ = NT t̄dA (2.72)
ϕn+1 (∂Ω(e) )

where t̄ is a prescribed field, which remains constant through the incremental proce-
dure.

Numerical integration of f int and f ext

One important aspect of the finite element implementation is the substitution of the
exact integrals of Equations (2.69) and (2.70) by some numerical approximations.
In this work, both the internal and external force vectors will be integrated using
standard Gaussian quadratures. For instance, in the case of the internal force vector,
f int is approximated by the following expression:
Z ngp
X
int T
f(e) = B σ̂dV ≈ wi Ji BTi σ̂ i (2.73)
ϕn+1 (Ω(e) ) i=1
16 2.2 Displacement-based finite elements

Box 1: The incremental non-linear finite element scheme - implicit solution.

1. Assemble the global external force vector,f̄ ext

2. Initialize increment counter i = 1

3. Set load factor ,λi

4. Solve the non-linear equilibrium equation

r(un+1 ) = f int (un+1 ) − λi f̄ ext = 0

5. Update increment counter i = i + 1

6. Check if prescribed number of increments has been achieved

where wi and Ji are, respectively, the Gaussian weight and the Jacobian at the ith
integration point. A similar procedure is carried out for the external force vector.

Newton-Raphson method
As stressed out before, the equilibrium equation (2.68) is generally non-linear and
demands an appropriate solution method. We will adopt herein the Newton-Raphson
method, which is particularly attractive due to its quadratic rates of convergence.
Following standard procedures of the method and particularising for the case of the
finite element framework presented in this chapter, the displacements are updated
according to:  −1
∂r

uk+1
n+1 = ukn+1 − ukn+1 r ukn+1 (2.74)
∂un+1
Equation (2.74) can be more conveniently written as

KT δu = −r(ukn+1 ), (2.75)

where
u = uk+1 k
n+1 − un+1 , (2.76)
and KT is called global tangent stiffness matrix, given by
∂r
KT = ukn+1 . (2.77)
∂un+1
The correct derivation of the tangent stiffness is crucial to guarantee the quadratic
rates of convergence of the Newton-Raphson method. In the case of finite strains
17 2.2 Displacement-based finite elements

within a spatial description, the expression for Equation (2.77) arises quite naturally
from the linearisation of the equilibrium equation in its weak form. Recalling that the
internal force vector is integrated using a Gaussian quadrature (Equation 2.73), the
element stiffness matrix is then given by:
ngp
(e)
X
KT = wi Ji BTi âi Gi , (2.78)
i=1

where ai is the spatial tangent modulus whose components are given by

1 ∂τ ij lm
aijkl = F − σ il δ jk . (2.79)
J ∂F km
For convenience, the full Newton-Raphson procedure associated with the present finite
element framework at finite strains is summarised in Box 2.

Box 2: Newton-Raphson scheme for the solution of the incremental non-linear fi-
nite element equilibrium equation implicit solution.
18 2.2 Displacement-based finite elements

1. Set k = 1, initial guess and residual function array

ukn+1 = un ; rk = f int (un ) − λn+1 f̄ ext

2. Compute the consistent spatial tangent moduli

1 ∂τ ij lm
aijkl = F − σ il δ jk .
J ∂F km

3. Assemble element tangent stiffness matrices

ngp
(e)
X
KT = wi Ji BTi âi Gi ,
i=1

4. Assemble global stiffness and solve for δuk+1

KT δuk+1 = −rk

5. Update displacements
uk+1 k
n+1 = un+1 + δu
k+1

6. Update the deformation gradient

−1
Fk+1 k+1
n+1 = I + ∇x un+1

7. Update stresses and internal variables

σ k+1 k+1 k+1 k+1

n+1 = σ̂(Fn+1 , αn ); , αn+1 = α̂(Fn+1 , αn )

8. Compute element internal force

ngp
X
int
f(e) = wi Ji BTi σ̂ k+1
n+1
i=1

9. Assemble the global internal force array and re-compute the residual function

rk+1 = f int − λn+1 f̄ ext

10. Check convergence IF krk+1 k < TOL EXIT

11. Set k = k + 1 and go to (2).

Chapter 3

BPA based model

3.1 Introduction
During the last decades, several researchers have worked in the field of constitutive
modeling of polymers and used the finite element method for analysing problems whit
polymeric materials. Academic community has followed different approaches for the
modeling of mechanical behavior of polymers. For a review on the topic the reader is
referred to Mackerle (1997), Mackerle (2003).
There is a quite vast area of applications for polymers including electronic, automotive
and medical industries. It is worth emphasizing the using polymeric based materials
in different industries is being increased due to interesting mechanical, thermal and
processing properties.
This chapter introduces a finite strain elasto-viscoplastic constitutive model. The flow
potential of the model is based on the Argon theory taken from BPA model (Boyce
et al. , 1988). The model is shown to be able to describe to typical deformation
behavior of thermoplastics namely, initial elastic, yield, post yield softening and final
hardening.
Next section describes the finite strain elasto-viscoplastic constitutive model.

3.2 Formulation
In this section, a finite strain multiplicative model for polymeric materials is described.
In order to develop a finite strain constitutive model, different issues have to be ad-
dressed. First of, the kinematics of finite strain should be explained. Dissipation
potential is probably the most important part of the constitutive model that has to
20 3.2 Formulation

be characterized.
In the following section we describe the multiplicative kinematic. Then a logarithmic
strain measure will be introduced. Free energy potential, dissipation potential and
additive decomposition of the stress are addressed.

3.2.1 Multiplicative kinematics

Multiplicative decomposition of the deformation gradient into an elastic and a plastic
contribution is adopted for this study. In fact, this elastic plastic decomposition of
the total deformation gradient is widely used, due its suitable features, by different
authors for developing constitutive models not only for polymers but also for other
class of materials such as metals,etc. According to the multiplicative kinematics, the
total deformation gradient, F, is composed by the elastic deformation gradient , Fe ,
and plastic deformation gradient, Fp . Details regarding this multiplicative kinematics
could be found in De Souza Neto et al. (2008).

F = Fe Fp (3.1)

In order to understand the multiplicative kinematics better, Figure (3.1) explains this
decomposition.
FINITE STRAIN ELASTOPLASTICITY 579

initial
configuration current
configuration

p
x = ϕ( p)
F = F e Fp

Fp
Fe

local
intermediate
configuration

Figure 14.1. Multiplicative decomposition of the deformation gradient.

Figure 3.1: Elastic-Plastic decomposition of the total deformation gradient-
Reproduced from De Souza Neto et al. (2008)
Remark 14.1. Rather than being just a convenient mathematical extension of the standard
additive decomposition of the strain tensor adopted in the infinitesimal theory, the multiplica-
It tive
candecomposition
be said thatofusing
F findsthis multiplicative
a solid and consistentkinematics, relationin(3.1),
physical justification implies
the slip theory the
of crystals (Asaro, 1983; Peirce et al., 1982, 1983; Rice, 1971). Essentially, the macroscopic
plastic deformation of metallic crystals is the result of microscopic sliding between blocks
of crystals along certain crystallographic planes (for more details on the plastic deformation
of crystals, we refer to Chapter 16, where the computational treatment of an elastoplastic
single crystal model is addressed). This mechanism is schematically illustrated in Figure 14.2.
21 3.2 Formulation

existence of an intermediate configuration which is unstressed and is obtained by

elastic unloading from the final deformed configuration.
Both elastic and plastic deformation gradients, could be polar decomposed as provided
by relations (3.2) and (3.3).

Fe = Re Ue = Ve Re , (3.2)

Fp = Rp Up = Vp Rp , (3.3)

polar decomposition of the deformation gradient is explained graphically is Figure

(3.2).
50 COMPUTATIONAL METHODS FOR PLASTICITY: THEORY AND APPLICATIONS

U dp
R
U deformed
configuration

reference
configuration dx=F dp
F x
dp i2*E
i2 :
p M:
D iiD1*

i1 R
V
R dp

Figure 3.8. Polar decomposition of the deformation gradient. Stretches and rotation.

Figure 3.2: Polar decomposition of the deformation gradient-Reproduced from De

Souza
TheNeto ettensors
stretch al. (2008)
U and V can be expressed as
√ √
U= C, V= e B, (3.42)
The components of the polar decomposition, R , Ue and Ve are the elastic rotation
where
tensor, C andright
elastic B – named, respectively,
stretch tensor and the right and left
elastic left Cauchy–Green strain
stretch tensor, tensors – are The
respectively.
defined by
terms in relation (3.3) are the2 corresponding
T plastic
2 terms.
T The definition of the
C = U = F F, B=V =F F . (3.43)
velocity gradient is provided below.
Example 3.1.2 (A simple plane deformation). To illustrate the meaning of the polar
= ḞF−1of
decomposition of F , a simple exampleLconsisting . a body subjected to a homogeneous (3.4)
deformation, i.e. with F independent of p, is given in what follows. Consider the rectangular
body
Using of Figure 3.8 subjected
multiplicative to homogeneous
kinematics, stretching/compression
relation (3.1), in the directions
in velocity gradient, of its(3.4),
equation
longitudinal and transversal axes (respectively, the directions of i1 and i2 in the reference
gives:
configuration) with a superimposed rigid rotation of angle α. With pi and xi denoting
coordinates of p and x in the Cartesian
L = system L (Fe )−1
Le + Fassociated
e p
, the orthonormal basis {i1 , i2 }, (3.5)
with
the deformation map is defined as

x1 = p1 λ1 cos α − p2 λ2 sin α
ϕ: (3.44)
x2 = p1 λ1 sin α + p2 λ2 cos α,
22 3.2 Formulation

where, Le is the elastic velocity gradient and Lp is the plastic deformation gradient
defined by following relations.

Le ≡ Ḟe (Fe )−1 , (3.6)

Lp ≡ Ḟp (Fp )−1 . (3.7)

Symmetric and skew part of the plastic velocity gradient are given below:
1h p i
Dp ≡ sym (Lp ) = L + (Lp )T , (3.8)
2

1h p i
Wp ≡ skew (Lp ) = L − (Lp )T , (3.9)
2
where, the superscript T stands for transpose. Dp is called plastic stretching tensor
or the rate of plastic deformation and Wp is called plastic spin tensor. It should be
mentioned that the plastic spin tensor is assumed to be zero.

Wp = 0 (3.10)

Boyce et al. (1989) and Timmermans (1998) explained that in case of polymeric
materials, having different plastic spin tensors is not of great effect on the overall
deformation behaviour. Consequently, assuming plastic spin tensor null, does not
considerably changes the stress-strain curve predicted by the model.

3.2.2 Strain measure

In this chapter, in order to measure the elastic deformations, natural (logarithmic)
spatial (Eulerian) strain is chosen. The reason for choosing the strain measure is
because of its physical meaning. Moreover, adopting logarithmic strain measure helps
to simplify the stress integration algorithm in a sense that the small strain elastic
predictor/return mapping algorithm is easily extended to finite strains.
The definition of logarithmic elastic strain tensor is given in:
1
εe ≡ ln Ve = ln Be (3.11)
2
In relation (3.11), the logarithm operator, ln(•), is tensorial logarithm of •.
Left elastic Cauchy-Green deformation tensor is defined by relation (3.12)

Be ≡ Fe [Fe ]T = [Ve ]2 . (3.12)

23 3.2 Formulation

3.2.3 Flow potential

The dissipation potential and flow rule of the model are defined using the Argon
theory. The plastic stretching tensor at the current configuration is given by:
∂Ψ
dp = γ̇ p = γ̇ p N, (3.13)
∂σ
where, Ψ is a dissipation potential, N is a flow vector and dp is spatial plastic stretching
tensor:
dp ≡ Re Dp ReT . (3.14)
Different terms in relation (3.14) could be explained and interpreted as follows:
• Dp is plastic stretching tensor in the relaxed configuration.
• dp is plastic stretching tensor in the current configuration (rotated from intermediate
configuration using elastic rotation).
As stated above, the flow vector N is the derivative of dissipation potential in order to
stress. Hence, the dissipation potential should be defined in order to obtain the flow
vector. The dissipation potential is assumed to be:
r
1
Ψ= s : s. (3.15)
2
As a result, the flow vector is obtained:
r
∂Ψ 1 s
N= = , (3.16)
∂σ 2 ||s||
where, ||s|| is the norm of s defined by:
√
||s|| = s : s. (3.17)

Combining relations (3.13) and (3.16), the multi-dimensional plastic flow rule of the
model is obtained as: r
p p 1 s
d = γ̇ . (3.18)
2 ||s||
Plastic shear strain rate, γ̇ p , is given by:
 
p ∆G∗
γ̇ = γ̇0 exp − , (3.19)
kT
where, γ̇0 is pre-exponential shear shear strain rate factor (one of the material prop-
erties); k is Boltzmanns constant; T is the absolute temperature and ∆G∗ is the free
energy change. The free energy change is given by:
 !(5/6) 
2 3
3πµω α  τ ,
∆G∗ = 1 − 0.077µ (3.20)
16(1 − ν) 1−ν
24 3.3 Conclusions

where, µ is shear modulus; ν is poissons ratio; ω is the net angle of rotation of the
molecular segment between the initial configuration and the activated configuration
and α is the mean molecular radius. Relation (3.19) can be rewritten as follows:
   τ (5/6) 
p Ae
s
γ̇ = γ̇0 exp − 1− , (3.21)
T se
where, τ is the shear stress. The parameter se is given by:

se = s + αp, (3.22)

where, p is hydrostatic pressure. The evolution of s is characterized by:

 
s
ṡ = h 1 − γ̇ p , (3.23)
sss

s = s0 + αp (3.24)
A and s0 are defined by relations (3.25) and (3.26), respectively.
39πω 2 α3
A≡ , (3.25)
16K

0.077µ
s0 ≡ . (3.26)
1−ν

3.2.4 Stress decomposition

The total stress, σ total , of the model, is additively composed of driving stress and
hardening stress:
σ total = σ driving + σ hardening . (3.27)
In this study, the hardening stress is not taken into account. Since the main objec-
tive of this thesis is to use numerical methods, Finite Element Method and Newton-
Raphson method, to solve the nonlinear system of equations of equilibrium, the total
stress is assumed to be the driving component of the stress.

3.3 Conclusions
In this chapter a finite strain elasto-viscoplastic constitutive model was introduced.
The kinematic and material contribution of the constitutive model were introduced.
The description of the flow potential, which is in fact the core of the model, was
introduced by Boyce et al. (1988) and the model is known as BPA model.
Chapter 4

The integration algorithm of the

model

4.1 Introduction
This chapter deals with the integration algorithm of the model introduced in chapter
(3). Stress integration algorithm and consistent tangent operator are explained. The
numerical implementation of of stress integration algorithm and tangent operator is
demonstrated. (Ortiz et al. (1983); Zienkiewicz and Taylor (1991); Owen and Hinton
(1980); De Souza Neto et al. (2008); (Simo and Hughes , 1998); Simo et al. (1998))
studied numerical solution of constitutive models within finite element approach which
basically is solution of a set of evolutionary equations in an iterative fashion.
The solution pursued in this study has a strain driven structure and is performed at
each Gauss point of the finite element mesh. Stress and internal variables are updated
according to the the level of current strain and the values of the internal variables in
the previous increment.
In order to have the solution of the global boundary value problem, using Newton-
Raphson method, efficiently converged, the consistent linearization of the time dis-
critized constitutive equations is very important.
Using operator split algorithms are nowadays standard procedure for numerical in-
tegration of elasto-plastic and elasto-viscoplastic constitutive equations (Criesfield
(1997); Simo and Hughes (1998)). In order to implement the constitutive model into
a finite element code, it is required to derive the state update equations and also con-
sistent tangent operator. The state update procedure is derived using operator split
algorithm and as a result an elastic predictor/return mapping algorithm is obtained
for the state update procedure.
26 4.2 State update

It is worth emphasizing that state update and consistent tangent are derived on the
small strain format of the constitutive equations. Then they are extended to finite
strain as explained (De Souza Neto et al., 2008).

4.2 State update

In this section, the constitutive equations of the model introduced in chapter (3) are
time discretized and operator split algorithm is applied to the equations in order to
derive the elastic predictor/return mapping algorithm of the state update procedure.
Initially, the total strain at the beginning of state update is assumed elastic (this stage
is called elastic predictor) and then using the return mapping equations (derived form
the flow potential and other constitutive equations), the inelastic contribution of the
total strain would be determined through an iterative procedure (this stage is called
plastic corrector or return mapping).

4.2.1 Return mapping

In order to update the deformation gradient and obtain the deformation gradient at
time step tn+1 , we use:
Fn+1 = F∆ Fn . (4.1)

where, Fn+1 is the deformation gradient at tn+1 , Fn is the deformation gradient at tn

and F∆ is the incremental deformation gradient. Considering incremental displace-
ment known, the incremental deformation gradient is obtained by:

F∆ = I + 5n (∆u), (4.2)

where, 5n (∆u) is the gradient of the incremental displacement.

We need to obtain the elastic trial state so that we will then apply return mapping
procedure in order to determine which part of the strain has been elastic and which
part has been plastic strain. As we know the total elastic strain at tn which is εen , the
elastic left Cauchy-Green deformation tensor is obtained by:

Ben = exp[2 εen ]. (4.3)

The elastic trial left Cauchy-Green deformation tensor at tn+1 is given by:

Ben+1
trial
= F∆ Ben (F∆ )T . (4.4)
27 4.2 State update

Having elastic trial left Cauchy-Green deformation tensor computed, the driving pa-
rameter of the study which is the elastic trial strain at tn+1 i.e. εen+1
trial
is given by:

1
εen+1
trial e trial
= ln [Vn+1 ln [Ben+1
]= trial
]. (4.5)
2
It is worth noting that so far, everything is independent of the material models i.e.
everything is done at the kinematic level and for whatever constitutive model could
be used. At this stage, when we have the elastic trial strain determined, the stresses
and state variables should be updated using the constitutive relations.
In order to derive the integration algorithm of the model, the small strain counterpart
of relation (3.18) is used: r
p p 1 s
ε̇ = γ̇ . (4.6)
2 ||s||
where, s is the deviatoric part of the Kirchhof stress.

s = Id : τ , (4.7)

and ||s|| is the norm of the deviatoric stress.

√
||s|| = s : s. (4.8)

Integration of relation (4.6) over the time step [tn , tn+1 ] gives:
r
1 ∆t
εpn+1 − εpn = γ̇n+1
p
sn+1 (4.9)
2 ||sn+1 ||

The time discretized form of relation (3.21) is given by:

"  (5/6) !#
p Aesn+1 τ
γ̇n+1 = γ̇0 exp − 1− , (4.10)
T sen+1

where,
sen+1 = sn+1 + αpn+1 (4.11)

Integrating relation (3.23) over time step [tn , tn+1 ] gives relation (4.12):
 
sn+1 p
sn+1 − sn − h 1 − γ̇n+1 ∆t = 0 (4.12)
sss
28 4.2 State update

From relation (4.9), the following system of four non-linear equations are obtained:
 q

 ∆ε P
(1) = γ̇ p 1 ∆t
s (1)

 n+1 n+1 2 ||sn+1 || n+1



 q



 ∆ε P
(2) = γ̇ p 1 ∆t
s (2)
 n+1 n+1 2 ||sn+1 || n+1

q (4.13)



 P p 1 ∆t
∆εn+1 (3) = γ̇n+1 2 ||sn+1 || sn+1 (3)





 q


 ∆εP (4) = γ̇ p 1 ∆t
s (4)
n+1 n+1 2 ||sn+1 || n+1

Considering relations (4.10)-(4.12), and the system of equations (4.13), the following
system of six algebraic equations are obtained, in the two dimensional space, for the
finite element implementation of the model.
 q

 1
R := ∆ε P
(1) − γ̇ p 1 ∆t
s (1) = 0

 n+1 n+1 2 ||sn+1 || n+1



 q



 R := ∆ε P
(2) − γ̇ p 1 ∆t
s (2) = 0

 2 n+1 n+1 2 ||sn+1 || n+1



 q



 P p
R := ∆εn+1 (3) − γ̇n+1 12 ||sn+1 ∆t
s (3) = 0

 3 || n+1

q (4.14)

 P p

 R4 := ∆εn+1 (4) − γ̇n+1 12 ||sn+1 ∆t
s (4) = 0
|| n+1





    (5/6) 



 p
R5 := γ̇n+1 − γ̇0 exp − T Ae
sn+1
1 − sen+1τ
=0







  

 R := s
6 − s − h 1 − sn+1 γ̇ p ∆t = 0
n+1 n sss n+1

The unknowns of the system of equations (4.14) are ∆εPn+1 (1), ∆εPn+1 (2), ∆εPn+1 (3),
p
∆εPn+1 (4), γ̇n+1 and sn+1 . The system of equation will be solved suing the well-known
iterative Newton-Raphson method. The unknowns are called u1 , u2 , u3 , u4 , u5 and
u6 , respectively. In order to solve the system of equations using the Newton-Raphson
method in an iterative fashion, the derivatives of all equations in order to all unknowns
should be calculated:
 k  k−1
 ∂R1 ∂R1 ∂R1 ∂R1
 δu 1 R1
· · · δu2   
∂u1 ∂u2 ∂u3 ∂u6
 ∂R2 ∂R2 ∂R2 · · · ∂R2   R2 
 ∂u1 ∂u2 ∂u3  
∂u6  δu3 
R3 
 .. .. . .    = − 
R4  . (4.15)
 . . .. ..  δu4   
∂R6 ∂R6 ∂R6
· · · ∂R 6 δu5  R5 
∂u1 ∂u2 ∂u3 ∂u6
δu6 R6
29 4.2 State update

Where, superscripts k −1 and k stand for two consecutive Newton-Raphson iterations.

The required derivatives in relation (4.15) are provided below.
Derivatives of the first residual equation
Derivative of the first residual equation in order to first unknown:
r
∂R1 p 1
= 1 − γ̇n+1 ∆t C1 , (4.16)
∂u1 2
where,
||sn+1 ||C2 − sn+1 (1)C3
C1 = , (4.17)
(||sn+1 ||)2
where, C2 and C3 are given by:

−2Gsn+1 (1)
C2 = −2G , C3 = . (4.18)
||sn+1 ||

Derivative of the first residual equation in order to second unknown:

r
∂R1 p 1
= −γ̇n+1 ∆t C4 , (4.19)
∂u2 2
where, we have:
sn+1 (1)C5 sn+1 (2)
C4 = − , C5 = −2G . (4.20)
(||sn+1 ||)2 ||sn+1 ||
Derivative of the first residual equation in order to third unknown:
r
∂R1 p 1
= −γ̇n+1 ∆t C6 , (4.21)
∂u3 2
where,
sn+1 (1)C7 sn+1 (3)
C6 = − , C7 = −4G . (4.22)
(||sn+1 ||)2 ||sn+1 ||
Derivative of the first residual equation in order to fourth unknown:
r
∂R1 p 1
= −γ̇n+1 ∆t C8 , (4.23)
∂u4 2
where,
sn+1 (1)C9 sn+1 (4)
C9 = − , C9 = −2G . (4.24)
(||sn+1 ||)2 ||sn+1 ||
Derivative of the first residual equation in order to fifth unknown:
r
∂R1 1 sn+1 (1)
=− ∆t . (4.25)
∂u5 2 ||sn+1 ||
30 4.2 State update

Derivative of the first residual equation in order to sixth unknown:

∂R1
= 0. (4.26)
∂u6

Derivatives of the second residual equation

Derivative of the second residual equation in order to first unknown:
r
∂R2 p 1
= −γ̇n+1 ∆t C10 , (4.27)
∂u1 2
where,
sn+1 (2)C3
C10 = − . (4.28)
(||sn+1 ||)2
Derivative of the second residual equation in order to second unknown:
r
∂R2 p 1
= 1 − γ̇n+1 ∆t C11 , (4.29)
∂u2 2
where,
||sn+1 ||C12 − sn+1 (2)C5
C11 = , (4.30)
(||sn+1 ||)2
where, C12 and is given by:
C12 = −2G. (4.31)

Derivative of the second residual equation in order to third unknown:

r
∂R2 p 1
= −γ̇n+1 ∆t C13 , (4.32)
∂u3 2
where,
sn+1 (2)C7
C13 = − . (4.33)
(||sn+1 ||)2
Derivative of the second residual equation in order to fourth unknown:
r
∂R2 p 1
= −γ̇n+1 ∆t C14 , (4.34)
∂u4 2
where,
sn+1 (2)C9
C14 = − . (4.35)
(||sn+1 ||)2
Derivative of the second residual equation in order to fifth unknown:
r
∂R2 1 sn+1 (2)
=− ∆t . (4.36)
∂u5 2 ||sn+1 ||
31 4.2 State update

Derivative of the second residual equation in order to sixth unknown:

∂R2
= 0. (4.37)
∂u6

Derivatives of the third residual equation

Derivative of the third residual equation in order to first unknown:
r
∂R3 p 1
= −γ̇n+1 ∆t C15 , (4.38)
∂u1 2
where,
sn+1 (3)C3
C15 = − . (4.39)
(||sn+1 ||)2
Derivative of the third residual equation in order to second unknown:
r
∂R3 p 1
= −γ̇n+1 ∆t C16 , (4.40)
∂u2 2
where,
sn+1 (3)C5
C16 = − . (4.41)
(||sn+1 ||)2
Derivative of the third residual equation in order to third unknown:
r
∂R3 p 1
= 1 − γ̇n+1 ∆t C17 , (4.42)
∂u3 2
where,
||sn+1 ||C18 − sn+1 (3)C7
C17 = , (4.43)
(||sn+1 ||)2
where, C18 and is given by:
C18 = −2G. (4.44)
Derivative of the third residual equation in order to fourth unknown:
r  
∂R3 p 1 sn+1 (3)C9
= −γ̇n+1 ∆t − . (4.45)
∂u4 2 (||sn+1 ||)2
Derivative of the third residual equation in order to fifth unknown:
r
∂R3 1 sn+1 (3)
=− ∆t . (4.46)
∂u5 2 ||sn+1 ||
Derivative of the third residual equation in order to sixth unknown:
∂R3
= 0. (4.47)
∂u6
32 4.2 State update

Derivatives of the fourth residual equation

Derivative of the fourth residual equation in order to first unknown:
r
∂R4 p 1
= −γ̇n+1 ∆t C19 , (4.48)
∂u1 2
where,
sn+1 (4)C3
C19 = − . (4.49)
(||sn+1 ||)2
Derivative of the fourth residual equation in order to second unknown:
r
∂R4 p 1
= −γ̇n+1 ∆t C20 , (4.50)
∂u2 2
where,
sn+1 (4)C5
C20 = − . (4.51)
(||sn+1 ||)2
Derivative of the fourth residual equation in order to third unknown:
r
∂R4 p 1
= −γ̇n+1 ∆t C21 , (4.52)
∂u3 2
where,
sn+1 (4)C7
C21 = − . (4.53)
(||sn+1 ||)2
Derivative of the fourth residual equation in order to fourth unknown:
r
∂R4 p 1
= 1 − γ̇n+1 ∆t C22 , (4.54)
∂u4 2
where,
||sn+1 ||C23 − sn+1 (4)C9
C22 = . (4.55)
(||sn+1 ||)2
where, C18 and is given by:
C23 = −2G (4.56)

Derivative of the third residual equation in order to fifth unknown:

r
∂R4 1 sn+1 (4)
=− ∆t . (4.57)
∂u5 2 ||sn+1 ||

Derivative of the fourth residual equation in order to sixth unknown:

∂R4
= 0. (4.58)
∂u6
33 4.2 State update

Derivatives of the fifth residual equation

Derivative of the fifth residual equation in order to first unknown:
∂R5
= −γ̇0 exp (C24 ) C25 , (4.59)
∂u1
where,
 (5/6) !
Ae
sn+1 τn+1
C24 =− 1− . (4.60)
T sen+1
and    (−1/6)
5 Ae
sn+1 τn+1
C25 = C26 , (4.61)
6 T sen+1
where, C26 is given by:
1 C3
C26 = C27 , C27 = √ . (4.62)
sen+1 2
Derivative of the fifth residual equation in order to second unknown:
∂R5
= −γ̇0 exp (C24 ) C28 , (4.63)
∂u2
where,
   (−1/6)
5 Ae
sn+1 τn+1
C28 = C29 , (4.64)
6 T sen+1
where, C29 is given by:
1 C5
C26 = C30 , C30 = √ . (4.65)
sen+1 2
Derivative of the fifth residual equation in order to third unknown:
∂R5
= −γ̇0 exp (C24 ) C31 , (4.66)
∂u3
where,
   (−1/6)
5 Ae
sn+1 τn+1
C31 = C32 , (4.67)
6 T sen+1
where, C29 is given by:
1 C7
C32 = C33 , C33 = √ . (4.68)
sen+1 2
Derivative of the fifth residual equation in order to fourth unknown:
   (−1/6)
∂R5 5 Ae sn+1 τn+1
= −γ̇0 exp (C24 ) C34 , (4.69)
∂u3 6 T sen+1
34 4.2 State update

where, C34 is given by:

1 C9
C34 = C35 , C35 = √ . (4.70)
sen+1 2
Derivative of the fifth residual equation in order to fifth unknown:
∂R5
= 1 − γ̇0 exp (C24 ) C36 , (4.71)
∂u5
where, C36 is equal to:
 (5/6) !  
τn+1 Ae
sn+1
C36 = 1− C37 − C38 , (4.72)
sen+1 T

where, C37 and C38 are given by:

A
C37 = − C39 , (4.73)
T

 (1/6)
5 sen+1 τn+1
C38 = C39 , (4.74)
6 τn+1 sn+1 )2
(e
where, C39 , is given by the following relation:
p

p



sss h∆t sss + hγ̇n+1 ∆t − h∆t sss sn + hγ̇n+1 ∆t
C39 = p

2 . (4.75)
sss + hγ̇n+1 ∆t

Derivative of the fifth residual equation in order to sixth unknown:

∂R5
= −γ̇0 exp (C24 ) C40 , (4.76)
∂u6
where, C40 is given by:
 (5/6) !    (−1/6)
A τn+1 5 Ae
sn+1 τn+1
C40 =− 1− + C41 , (4.77)
T sen+1 6 T sen+1

where,
τn+1
C41 = − . (4.78)
sn+1 )2
(e

Derivatives of the sixth residual equation

Derivatives of the sixth residual equation in order to first four unknowns are equal to
zero.
∂R6 ∂R6 ∂R6 ∂R6
= = = = 0. (4.79)
∂u1 ∂u2 ∂u3 ∂u4
35 4.3 Consistent tangent operator

Derivative of the sixth residual equation in order to fifth unknown:

 
∂R6 sn+1
= −h∆t 1 − . (4.80)
∂u5 sss
Derivative of the sixth residual equation in order to sixth unknown:
∂R6 hγ̇ p ∆t
= 1 + n+1 . (4.81)
∂u6 sss
Having all the derivatives computed,it is possible to introduce the relation (4.16)-
(4.81), in to the system of equations (4.15) and solve it using Newton-Raphson method
in an iterative manner. After solving the system of equations, we have:
k
uk1 = ∆εP,k P,k−1
n+1 (1) = ∆εn+1 (1) + (δu1 ) , (4.82)

k
uk2 = ∆εP,k P,k−1
n+1 (2) = ∆εn+1 (2) + (δu2 ) , (4.83)

k
uk3 = ∆εP,k P,k−1
n+1 (3) = ∆εn+1 (3) + (δu3 ) , (4.84)

k
uk4 = ∆εP,k P,k−1
n+1 (4) = ∆εn+1 (4) + (δu4 ) , (4.85)

p,k
uk5 = γ̇n+1 p,k−1
= γ̇n+1 + (δu5 )k , (4.86)

uk6 = skn+1 = sn+1

k−1
+ (δu6 )k . (4.87)

4.3 Consistent tangent operator

For the sake of completing the finite element implementation of the model, it is re-
quired to derive the tangent operator due to the fact that the tangent operator is the
only material related term in the global tangent stiffness matrix. De Souza Neto et
al. (2008) introduced the spatial global tangent stiffness by:
1
aijkl = [D : L : B]ijkl − σ il δjk , (4.88)
2J
where, δ is the Kronecker delta, D is the small strain consistent tangent operator. The
fourth order tensor L is defined by:
∂ ln [Ben+1
trial
]
L= e trial
, (4.89)
∂[Bn+1 ]
36 4.3 Consistent tangent operator

and the fourth order tensor B is defined by the cartesian components:

Bijkl = δik Ben+1
trial
jl
+ δjk Ben+1
trial
il
. (4.90)

In the spatial global tangent stiffness, relation (4.88), the fourth order tensor D is
material dependent. It means that consistent tangent operator, D, is dependent on
the material model under study but the other components of spatial tangent stiffness
are totally kinematic components, i.e. independent of the material model.
We know that stress is composed of deviatoric stress and hydrostatic pressure:

σ n+1 = sn+1 − pn+1 I, (4.91)

where, the deviatoric stress is given by:

sn+1 = 2Gεedn+1 = 2GId : (εen+1 ). (4.92)

The stress deviator could be represented as:

sn+1 = 2GId : (εen+1

trial
− (∆ε)Pn+1 ) = 2GId : εetrial P
n+1 − 2GId : (∆ε)n+1 .

The hydrostatic pressure defined with the following relation:

1
pn+1 = − tr(σ n+1 ). (4.93)
3
The tangent operator is given by:
dσ n+1 dσ n+1
D= = e trial . (4.94)
dεn+1 dεn+1
Using relations (4.91) and (4.94), the tangent operator could be written as:
dσ n+1 dsn+1 dpn+1
= − I. (4.95)
dεn+1 dεn+1 dεn+1
The first term in the tangent operator which is derivative of the stress deviator in
order to strain is given by:
dsn+1 d
= [2GId : εetrial P
n+1 − 2GId : (∆ε)n+1 ], (4.96)
dεn+1 dεn+1
or equivalently,
dsn+1 d
= 2GId − [2GId : (∆ε)Pn+1 ],
dεn+1 dεn+1
which can be rewritten in the following from:
dsn+1 d(∆ε)Pn+1
= 2GId − 2G . (4.97)
dεn+1 dεn+1
37 4.3 Consistent tangent operator

d(∆ε)P
n+1
Considering equation (4.97), we need to determine the calculate the term dεn+1
which is basically derivative of a second order tensor, (∆ε)Pn+1 , in order to strain
which is a second order, εn+1 , which results in a fourth order tensor.
d(∆ε)P
n+1
According to relation (4.9), the derivative dεn+1 could be written as follows:
r !
d(∆ε)Pn+1 d p 1 ∆t
= γ̇n+1 sn+1 (4.98)
dεn+1 dεn+1 2 ||sn+1 ||

which could be rewritten as follows:

r  p  r  
d(∆ε)Pn+1 1 dγ̇n+1 sn+1 1 p d sn+1
= ∆t ⊗ + ∆tγ̇n+1 (4.99)
dεn+1 2 dεn+1 ||sn+1 || 2 dεn+1 ||sn+1 ||
The second term in the right hand side of equation (4.99) could be expanded by making
the following:
  dsn+1 d||sn+1 ||
d sn+1 ||sn+1 || dεn+1
− dεn+1 n+1
s
= 2 (4.100)
dεn+1 ||sn+1 || (||sn+1 ||)
The derivative of the norm of stress deviator in order to strain is given by:
 
d||sn+1 || 1 dsn+1
= : sn+1 (4.101)
dεn+1 ||sn+1 || dεn+1
Using relations (4.97), (4.99), (4.100), (4.101) the following relation is obtained:
r  
dsn+1 1 p 1 dsn+1 1 dsn+1
+ ∆tγ̇n+1 − : (sn+1 ⊗ sn+1 ) =
dεn+1 2 ||sn+1 || dεn+1 (||sn+1 ||)3 dεn+1
r
 p 
1 dγ̇n+1 sn+1
2GId − ∆t ⊗ (4.102)
2 dεn+1 ||sn+1 ||
According to relation (4.95), we need to have the derivative of the deviatoric stress,
sn+1 , and hydrostatic stress, pn+1 , in order to strain, εn+1 . Considering relation
(4.102), in order to have the derivative of stress deviator in order to strain, it is
dγ̇ p
required to compute dεn+1
n+1
. In order to obtain the required terms, for the tangent
operator, we will proceed as presented.
For the sake of notational convenience, the unknowns we have, ∆εPn+1 (1), ∆εPn+1 (2),
∆εPn+1 (3), ∆εPn+1 (4), are named as shown below:
 

 ∆εPn+1 (1) → u(1) 


 


 


 

 ∆εPn+1 (2) → u(2) 
(4.103)

 


 ∆εPn+1 (3) → u(3) 


 


 

 
∆εPn+1 (4) → u(4)
38 4.3 Consistent tangent operator

Matrix relation (4.104), could be written:

 
du1
   ∂R1 
∂R1  du2 
− ∂ε∂R 1
: dεn+1 ∂u1
∂R1
∂u2
∂R1
∂u3
··· ∂u7  
∂R2   du3 
n+1
− ∂R2 : dεn+1   ∂R2 ∂R2 ∂R2
···  
 ∂εn+1   ∂u1 ∂u7  
  du4 
∂u2 ∂u3
 .  =  .. .. ... .. . (4.104)
 .
.   . . .  P  
∂R7 ∂R7 ∂R7 ∂R7 
dγ̇n+1 
∂R7
− ∂εn+1 : dεn+1 ∂u1 ∂u2 ∂u3
··· ∂u7 dsn+1 
dpn+1
Where, the seventh residual equation, R7 , is given by:

R7 := pn+1 + KI : εn+1 = 0. (4.105)

Relation (4.104), is rewritten by :

 
du1
 du2     ∂R1
− ∂εn+1 : dεn+1

  C1,1 C1,2 · · · C1,7
 du3  
  C2,1 C2,2 · · · C2,7   ∂R2 
 − ∂εn+1 : dεn+1 
 du4  = 
 P   . .. .. ..   .. , (4.106)
dγ̇n+1   .. . . .  . 
  C7,1 C7,2 · · · C7,7 ∂R
dsn+1  − ∂εn+1 : dεn+1
7

dpn+1
where,
   ∂R1 ∂R1 ∂R1

∂R1 −1
C1,1 C1,2 · · · C1,7 ∂u1 ∂u2 ∂u3
··· ∂u7
C2,1 C2,2 · · · C2,7   ∂R2 ∂R2 ∂R2
··· ∂R2 
   ∂u1 ∂u2 ∂u3 ∂u7 
 .. .. . . ..  =  .. .. ... ..  . (4.107)
 . . . .   . . . 
∂R7 ∂R7 ∂R7 ∂R7
C7,1 C7,2 · · · C7,7 ∂u ∂u2 ∂u3
··· ∂u7
1

Expanding matrix relation (4.106) gives the following system of equations.

du1 = C1,1 (− ∂ε∂R 1
n+1
: dεn+1 ) + C1,2 (− ∂ε∂R 2
n+1
: dεn+1 )
∂R7
+ · · · + C1,7 (− ∂εn+1 : dεn+1 )
du2 = C2,1 (− ∂ε∂R 1
n+1
: dεn+1 ) + C2,2 (− ∂ε∂R 2
n+1
: dεn+1 )
∂R7
+ · · · + C2,7 (− ∂εn+1 : dεn+1 )
du3 = C3,1 (− ∂ε∂R 1
n+1
: dεn+1 ) + C3,2 (− ∂ε∂R 2
n+1
: dεn+1 )
∂R7
+ · · · + C3,7 (− ∂εn+1 : dεn+1 )
du4 = C4,1 (− ∂ε∂R 1
: dεn+1 ) + C4,2 (− ∂ε∂R 2
: dεn+1 )
n+1
∂R7
n+1
(4.108)
+ · · · + C4,7 (− ∂εn+1 : dεn+1 )
dγ̇n+1 = C5,1 (− ∂ε∂R
P 1
n+1
: dεn+1 ) + C5,2 (− ∂ε∂R 2
n+1
: dεn+1 )
∂R7
+ · · · + C5,7 (− ∂εn+1 : dεn+1 )
dsn+1 = C6,1 (− ∂ε∂R 1
n+1
: dεn+1 ) + C6,2 (− ∂ε∂R 2
n+1
: dεn+1 )
∂R7
+ · · · + C6,7 (− ∂εn+1 : dεn+1 )
dpn+1 = C7,1 (− ∂ε∂R 1
n+1
: dεn+1 ) + C7,2 (− ∂ε∂R 2
n+1
: dεn+1 )
∂R7
+ · · · + C7,7 (− ∂εn+1 : dεn+1 )
39 4.3 Consistent tangent operator

In order to compute the required unknowns for the tangent operator, D, we need to
have the derivatives of the residual equations, R1 , R2 , R3 , R4 , R5 , R6 , R7 , in order to
p
unknowns,∆εPn+1 (1), ∆εPn+1 (2), ∆εPn+1 (3), ∆εPn+1 (4), γ̇n+1 , sn+1 and pn+1 and also in
order to strain εn+1 . For the state update stage of the solution the derivatives of
the first six residual equations, R1 , R2 , R3 , R4 , R5 , R6 , in order to first six unknowns,
p
∆εPn+1 (1), ∆εPn+1 (2), ∆εPn+1 (3), ∆εPn+1 (4), γ̇n+1 , sn+1 , are computed and provided in
relations (4.16)-(4.81).
We need to Compute:

• The derivatives of the the first six residual equations in order to hydrostatic
pressure, pn+1 :

• The derivatives of seventh equation, R7 , in order to all unknowns:

• The derivatives of all equations in order to strain to strain.

The first set of required calculations are the derivatives in order to pressure:
∂R1 ∂R2 ∂R3 ∂R4
= = = = 0, (4.109)
∂pn+1 ∂pn+1 ∂pn+1 ∂pn+1

∂R5
= −γ̇0 exp (C24 ) K1 , (4.110)
∂pn+1
where,
∂C24
K1 = , (4.111)
∂pn+1
which is given by:
 (5/6) !  (−1/6) !
−Aα τn+1 Ae
sn+1 5 τn+1
K1 = 1− − K2 , (4.112)
T sen+1 T 6 sen+1

where,
ατn+1
K2 = −
. (4.113)
sen+1
The other derivatives in order to pressure is given below:
∂R6 ∂R7
=0 , = 1. (4.114)
∂pn+1 ∂pn+1
The last equation, R7 , is only dependent on pressure, pn+1 , and consequently the
derivatives of the last residual equation in order to all unknowns but pressure, pn+1 ,
are zero.
∂R7 ∂R7 ∂R7 ∂R7 ∂R7 ∂R7
= = = = = = 0. (4.115)
∂u1 ∂u2 ∂u3 ∂u4 ∂u5 ∂u6
40 4.3 Consistent tangent operator

The last set of required relations to be determined are the derivatives of the all residual
equations in order to strain. It should be emphasized that since strain, εn+1 , is a second
order tensor and the equations are scalar quantities, the derivatives of each residual
equation in order to strain result in a second order tensor.
∂R1
Now, we proceed with the derivatives R1 , R2 , R3 , R4 in order to strain: ∂εn+1
=?
r
∂R1 P 1 ∂ sn+1 (1)
= γ̇n+1 ∆t ( ), (4.116)
∂εn+1 2 ∂εn+1 ||sn+1 ||
where, 2

 3 

∂sn+1 (1)  −1 
= 2G 3 , (4.117)
∂εetrial 
 0
n+1  −1 

3
∂R2
and the next one ∂εn+1
=?,
r
∂R2 P 1 ∂ sn+1 (2)
= γ̇n+1 ∆t ( ), (4.118)
∂εn+1 2 ∂εn+1 ||sn+1 ||
where, we have:  −1 

 32 

∂sn+1 (2)  
= 2G 3 . (4.119)
∂εetrial 
 0
n+1  −1 

3
∂R3
Now, it is the turn of ∂εn+1
=?
r
∂R3 P 1 ∂ sn+1 (3)
= γ̇n+1 ∆t ( ), (4.120)
∂εn+1 2 ∂εn+1 ||sn+1 ||
where,  

 0
∂sn+1 (3)   
0
= 2G 1 . (4.121)
∂εetrial 
 

n+1  
2
0
∂R4
The derivative ∂εn+1
is given in relation (4.122):
r
∂R4 P 1 ∂ sn+1 (4)
= γ̇n+1 ∆t ( ), (4.122)
∂εn+1 2 ∂εn+1 ||sn+1 ||
where,  −1 
 3 
 −1 
∂sn+1 (4)  
= 2G 3 . (4.123)
∂εetrial 
 0
n+1 2
3
41 4.3 Consistent tangent operator

The norm of stress deviator is given by:

p
||sn+1 || = sn+1 (1)sn+1 (1) + sn+1 (2)sn+1 (2) + 2sn+1 (3)sn+1 (3) + sn+1 (4)sn+1 (4).
(4.124)
The derivative of the fifth equation in order to strain is given in:
 
∂R5 −Aesn+1 τn+1 65
= −γ̇0 exp (1 − ( ) ) ×,
∂εn+1 θ sen+1

∂ −Ae sn+1 τn+1 56

[ (1 − ( ) )]. (4.125)
∂εn+1 θ sen+1
Relation (4.125) could be rewritten as:
 
∂R5 −Ae
sn+1 τn+1 65
= −γ̇0 exp (1 − ( ) ) ×,
∂εn+1 θ sen+1

Ae
sn+1 ∂ τn+1 65
− × [−( ) ], (4.126)
θ pεn+1 sen+1
or in a more simplified form:
 
∂R5 −Ae
sn+1 τn+1 65
= −γ̇0 exp (1 − ( ) ) ×,
∂εn+1 θ sen+1

Ae
sn+1 ∂ τn+1 56
+ × [( ) ], (4.127)
θ ∂εn+1 sen+1
where τn+1 is an equivalent stress defined by:
r
1
τn+1 = sn+1 : sn+1 . (4.128)
2
Using relations (4.127) and (4.128), we can have:
√
∂R5 5 2 γ̇0 AG τn+1 − 61 −Ae
sn+1 τn+1 56
=− ( ) sn+1 exp{ (1 − ( ) )}. (4.129)
∂εn+1 6 θ||sn+1 || sen+1 θ sen+1

The other derivatives are provided below:

∂R6
= 0, (4.130)
∂εn+1

∂R7
= KI. (4.131)
∂εn+1
Chapter 5

Numerical examples

This chapter presents some numerical examples through which the capability of the
model to characterize the typical deformation behaviour of polymers is assessed and
also the efficiency of the derived and implemented algorithm is shown.
The material under study is polymethylmethacrylate (PMMA). The material prop-
erties required for performing the simulations are taken from (Boyce et al. , 1988).
The material properties are tabulated in Table (5.1). Different examples including

Table 5.1: Material properties for PMMA required for the model
E ν s0 α γ̇0 A h sss R
PMMA 2300 0.37 88E+6 0.2 1.13E+11 167E-06 900E+06 77E+06 8.3143

compression on a cube, cylinder compression, notched bar compression and necking

of a round bar are selected to be analysed.

5.1 Compression of a cube (plane strain compres-

sion)
In this section, a compression test on a cube is performed which can be approximated
in 2D as plane strain compression. The geometry and mesh of the problem is shown in
Figure (5.1). The specimen is spatially discretized with 128 quadrilateral eight noded
elements with reduced four integration Gauss points. The stress strain curve for the
plane strain compression test is given in Figure (5.2). In Figure (5.2), the typical
deformation behaviour of glassy polymers which includes initial linear elastic, yield,
post-yield softening could be observed. It should be emphasized that the reason for
not-seeing the final hardening regime in the deformation is that the hardening stress
43 5.1 Compression of a cube (plane strain compression)

8 mm

8 mm

Figure 5.1: The geometry and mesh of the plane strain compression test.

7,00E+07

6,00E+07

5,00E+07
True compressive stress

4,00E+07

3,00E+07

2,00E+07

1,00E+07

0,00E+00
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4

True compressive strain

Figure 5.2: The stress-strain curve for the compression of the cube.

is not considered in this study.

In order to assess the performance and numerical efficiency of the derived and imple-
mented algorithm, the convergences are going to to be checked. In Table (5.2), the
local convergence of the problem at two different load increments is shown.
From Table (5.2), it can be concluded that the state update algorithm is derived
44 5.2 Cylinder upsetting (axisymmetric compression)

Table 5.2: Local convergence table for cube compression simulation

The value of the residual at local level
Iteration number Increment (6) Increment (81)
1 0.383915E-05 1.885185
2 0.295615E-10 0.194465
3 2.044682E-03
4 4.130582E-06
5 8.780845E-09

properly.

5.2 Cylinder upsetting (axisymmetric compression)

This section presents results of a compression test on a cylinder. The Geometry is
the same as the one shown in Figure (5.1) and the same material properties but
under axisymmetric condition is used for the simulation. Figure (5.3), shows stress-
strain curve for the cylinder upsetting simulation. In order the check the numerical

6,00E+07

5,00E+07

4,00E+07
True compressive stress

3,00E+07

2,00E+07

1,00E+07

0,00E+00
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4

True compressive strain

Figure 5.3: The stress-strain curve for the compression of the cylinder.

efficiency, the convergence of the cylinder upsetting simulation are investigated. Table
(5.3) presents local convergence (state update) for two different load increments.
45 5.3 Notched bar compression

Table 5.3: Local convergence table for cylinder compression simulation

The value of the residual at local level
Iteration number Increment (6) Increment (35)
1 0.779533 0.532699
2 0.010266E-04 0.880138E-02
3 0.312095E-08 0.059476E-04
4 0.740346E-05
5 0.737990E-07

5.3 Notched bar compression

The next example is a notched bar under uniaxial compression. The geometry and
mesh of the problem is shown in Figure (5.4). The specimen is spatially discretized

Figure 5.4: The geometry and mesh of the compression test on a notched bar

with 350 eight noded quadrilateral elements with reduced four integration Gauss
points. Figure (5.5) depicts contour plot of the effective stress in whole specimen
when ten percent of the deformation is applied. Figure (5.6) shows contour plot of
the effective stress in whole specimen at twenty percent of the deformation. Effective
stress in the specimen when 30 percent of the deformation is applied is shown in Figure
(5.7). When half of the deformation has proceed, the effective stress is given in Figure
(5.8). The final contour plot ,depicted in Figure (5.9), shows the effective stress at the
end of the deformation.
 46 5.3 Notched bar compression

Equivalent Stress (0.1)

3.87e+004 3.45e+007 6.89e+007

Figure 5.5: Contour plot of effective stress at ten percent of the deformation.

Equivalent Stress (0.2)

3.87e+004 3.45e+007 6.89e+007

Figure 5.6: Contour plot of effective stress at twenty percent of the deformation.

Equivalent Stress (0.3)

3.87e+004 3.45e+007 6.89e+007

Figure 5.7: Contour plot of effective stress at thirty percent of the deformation.
 47 5.3 Notched bar compression

Equivalent Stress (0.5)

3.87e+004 3.45e+007 6.89e+007

Figure 5.8: Contour plot of effective stress at half of the deformation.

Equivalent Stress (1)

3.87e+004 3.45e+007 6.89e+007

Figure 5.9: Contour plot of effective stress at the end of the deformation.
Chapter 6

Conclusions and suggestions for

future research

In this thesis, a constitutive model was developed based on BPA model, which is one
of the most well-known constitutive models proposed for polymers. Finite Element
Method was used to solve the equilibrium equation. The integration algorithm of
the model was presented for finite element implementation. The well-known Newton-
Raphson method is used at state update (local) level and also equilibrium (global) level
to solve the non-linear system of equations. Hence, in this thesis we have used two
important numerical methods, namely finite element method and Newton-Raphson
method, to deal with equilibrium equations of solid polymers. It is worth emphasizing
that, there might be a possibility to improve the integration algorithm of the model by
reducing the system of equations to another system of equations with reduced number
of equations or even to a single scalar equation.
According to the results presented in chapter five of this work, it can be concluded that
the BPA based model can predict the typical polymeric materials behavior. In order
to have an idea of the capabilities and drawbacks of the model predictions, it is re-
quired to have a wide range of experimental results under different deformation modes.
Comparing simulations and experimental results will help to realize the accuracy of
the model predictions and also the necessity of required changes and modifications to
the model. Considering all the remarks provided above, the suggestions for the future
research can be summarized as follows:
• Checking the possibility of modification of the integration algorithm

• Making some comparisons with existing experimental results

• Proposing some applicable modifications to the constitutive relations in order to

improve model predictions
References

Shabana, A.A., 2008. Computational continuum mechanics. Cambridge University

Press. 4

Bonet, J. and Wood, R.D., 2008. Nonlinear continuum mechanics for finite element
analysis. Cambridge University Press, 2 edition.

Bertram, A., 2005. Elasticity and plasticity of large deformations. An introduction.

Springer-Verlag, Berlin, Germany.

Holzapfel, G.A., 2000. Nonlinear solid mechanics: a continuum approach for engineer-
ing. John Wiley and Sons, Chichester. 4

Parks, D.M., Azhi, S., 1990. Polycrystalline plastic deformation and texture evolution
for crystals lacking five independent slip systems. Journal of Mechanics and Physics
of Solids 38, 701724.

Dahoun, A., 1992. Comportement plastique et textures de deformation des polyme ‘res
semi-cristallins en traction uniaxiale et en cisaillement simple. Ph.D. Dissertation,
Institut National Polytechnique de Lorraine, France.

Mooney, M., 1940. A theory of large elastic deformation. Journal of Applied Physics
11, 582. 2

Eyring, H., 1936. Viscosity, plasticity, and diffusion as examples of absolute reaction
rates. Journal of Chemical Physics 4, 283295. 2

van Breemen, L.C.A., Engels, T.A.P., Klompen, E.T.J., Senden, D.J.A., Govaert,
L.E., 2012. Rate- and temperature-dependent strain softening in solid polymers.
Journal of Polymer Science Part B: Polymer Physics 50, 1757-1771.
50 REFERENCES

van Melick, H.G.H., Govaert, L.E., Meijer, H.E.H., 2003. Localisation phenomena in
glassy polymers: influence of thermal and mechanical history. Polymer 44, 3579-
3591.

Arruda, E.M., Boyce, M.C., 1993. Evolution of plastic anisotropy in amorphous poly-
mers during finite straining. International Journal of Plasticity 9, 697-720.

Baaijens, F.P.T., 1991. Calculation of residual stresses in injection molded products.

Rheological Acta 30, 284-299 2

Boyce, M.C., Weber, G.G., Parks, D.M., 1989. On the kinematics of finite strain
plasticity. Journal of the Mechanics and Physics of Solids 37, 647-665. 22

Boyce, M.C., and Arruda, E.M., 1990. An experimental and analytical investigation
of the large strain compressive and tensile response of glassy polymers. Polymer
Engineering and Science 30, 12881298.

Criesfield, M.A., 2000. Non-linear finite element analysis of solids and structures, Vol
1, Essentials. John Wiley, UK. 25

Criesfield, M.A., 1997. Non-linear finite element analysis of solids and structures, Vol
2, Advanced topic. John Wiley, UK. 4

De Souza Neto, E.A., Peric, D.J., Owen, D.R.J., 2008. Computational methods for
plasticity: Theory and applications. John Wiley, UK. vi, 20, 21, 25, 26, 35

De Souza Neto, E.A., Peric, D.J., Owen, D.R.J., 1994. A model for elastoplastic dam-
age at finite strains: algorithmic issues and applications. Engineering computations
11, 257-281.

Diez, J.M., 2010. Thermoviscoplasticity of Glassy Polymers: Experimental Character-

ization, Parameter Identification and Model Validation. Ph.D. thesis, University of
Stuttgart, Stuttgart, Germany.

Govaert, L., Timmermans, P., Brekelmans, W., 2000. The influence of intrinsic strain
softening on strain localization in polycarbonate: modelling and experimental vali-
dation. Journal of Engineering Materials and Technology 122, 177-185. 2

Hasan O.A., Boyce M.C., Li X.S., Berko, S., 1993. An investigation of the yield
and postyield behaviour and corresponding structure of poly(methyl methacrylate).
Journal of Polymer Science Part B: Polymer Physics 31, 185197.
51 REFERENCES

Leonov, A., 1976. Non-equilibrium thermodynamics and rheology of viscoelastic poly-

mer media. Rheological Acta 15, 85-98.

Mackerle, J., 1997. Finite element analysis and simulation of polymers: a bibliography
(1976-1996). Modelling and Simulation in Materials Science and Engineering 5, 615-
650. 2, 19

Mackerle, J., 2003. Finite element anlysis and simulation of polymers an addendum:
a bibliography (1996-2002). Modelling and Simulation in Materials Science and En-
gineering 11, 195-231. 2, 19

Ortiz, M., Pinsky, P.M., Taylor, R.L., 1983. Operator split methods for the numer-
ical solution of the elastoplastic dynamic problem. Computer Methods in Applied
Mechanics and Engineering 39, 137-157. 25

Peric, D.J., Owen, D.R.J., Honnor, M.E., 1992. A model for finite strain elasto-
plasticity based on logarithmic strains: Computational issues. Computer Methods
in Applied Mechanics and Engineering 94, 35-61.

Owen, D.R.J., Hinton. E., 1980. Finite Elements in Plasticity: Theory and Practice.
Pineridge Press, Swansea, UK. 25

Rouinia, M., Peric, D.J., 1998. A computational model for elasto-viscoplatic solids
at finite strain with reference to thin shell applications. International Journal for
Numerical Methods in Engineering 42, 289-311.

Simo, J.C., Hughes, T.J.R., 1998. Computational Inelasticity. Springer-Verlag, New

York, US. 25

Simo, J.C., Kennedy, J.G., Govindjee, S. 1998. Non-smooth multisurface plasticity

and viscoplasticity. Loading/unloading conditions and numerical algorithms. Inter-
national Journal for Numerical Methods in Engineering 26, 2161-2185. 25

Tervoort, T., 1996. Constitutive modelling of polymer glasses: finite, nonlinear vis-
coelastic Behaviour of Polycarbonate. Ph.D. thesis, Eindhoven University of Tech-
nology, Eindhoven, The Netherlands.

Tervoort, T., Smit, R., Brekelmans, W., Govaert, L., 1998. A constitutive equa-
tion for the elasto-viscoplastic deformation of glassy polymers. Mechanics of Time-
Dependent Materials 1, 269-291. 2
52 REFERENCES

Timmermans, P.H.M., 1997. Evaluation of a constitutive model for solid polymeric

materials: model selection and parameter quantification. Ph.D. thesis, Eindhoven
University of Technology, Eindhoven, The Netherlands. 22

Van der Aa, M.A.H., 1999. Wall ironing of polymer coated sheet metal. Ph.D. thesis,
Eindhoven University of Technology, Eindhoven, The Netherlands.

Van Melic, H.G.H., Govaert, L.E., Meijer, H.E.H., 2003. Localisation phenomena in
glassy polymers: influence of thermal and mechanical history. Polymer 44, 3579-
3591.

Zienkiewicz, O.C., Taylor, R.L., 1991. The Finite Element Method, Vol 2: Solid and
Fluid Mechanics, Dynamics and Non-Linearity. McGraw-Hill, London, UK. 25

Argon, A.S., 1973. A theory for the low temperature plastic deformation of glassy
polymers. Philosophical Magazine 28, 839-865.

Boyce, M.C., Parks, D.M., Argon, A.S., 1988. Large inelastic deformation of glassy
polymers, Part I: Rate dependent constitutive model. Mechanics of Materials 7,
15-33. 2, 3, 19, 24, 42

Haward, R.N., Thackray, G., 1968. The use of mathematical model to describe isother-
mal stress-strain curves in glassy polymers, Proceedings of the Royal Society, A 302,
453-472. 2

Haward, R.N., 1987. The application of simplified model for the stress-strain curves
of polymers. Polymer 28, 1485-1488.

Wu, P.D., van der Giessen, E., 1994. Analysis of shear band propagation in amorphous
glassy polymers. International Journal of Solids and Structures 31, 1493-1517. 2

Hall, C., 1989. Polymer Materials, an introduction for technologists and scientists.
Wiley.