The theory of continuum and elasto plastic materials
1. The theory of continuum and
elasto-plastic materials
Written by:
Braj Bhushan Prasad
Technische UniversitΓ€t Berlin, Faculty V - Mechanical Engineering and
Transport Systems
February 11, 2018
4. The theory of continuum and elasto-plastic materials IV
4.4. Consistency condition . . . . . . . . . . . . . . . . . . . . . . . . 47
4.5. Hardening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.5.1. Isotropic hardening . . . . . . . . . . . . . . . . . . . . . . 48
4.5.2. Kinematic hardening . . . . . . . . . . . . . . . . . . . . . 51
List of Figures VI
Bibliography VII
A. Important Results i
A.1. Derivation of the Gauss Theorem . . . . . . . . . . . . . . . . . . i
A.2. Time Derivative of the Jacobian . . . . . . . . . . . . . . . . . . iii
Contents
5. The theory of continuum and elasto-plastic materials 1
1. Mathematical preliminaries
Relations between pure and applied mathematicians are based on trust
and understanding. Namely, pure mathematicians do not trust applied
mathematicians, and applied mathematicians do not understand pure
mathematicians.
β Albert Einstein (14 March 1879 β 18 April 1955)
Physical laws provide mathematical models for the physical world. These
physical laws should be invariant in nature, i.e. independent of the position and
orientation of coordinate system or observer. In order to make these physical
laws independent from the choice of coordinate system, we express these laws
in term of vector and tensor equations. Therefore, it is necessary to have
fundamental understanding of tensor analysis and vector calculus. The purpose
of this preliminary chapter is to present the basic rules and standard results of
tensor and vector analysis, which are applied throughout this thesis. For a more
detailed explanation see the standard books on vectors and tensors by (Segel and
Handelman, 1977), (Itskov, 2007), (Schade and Neemann, 2009), and (Brannon,
2003). In this work we restrict ourself to three dimensional Cartesian tensors in
Euclidian space.
1.1. Tensor analysis
Tensor analysis is an important tool to visualize the theory of continuum meΒ
chanics and is also significant in some of the theories used in Abaqus . Tensors
have been known since 1854 but the broader acceptance of its concepts began in
the 20th century. Tensors are a generalization of scalar and vector. MathematiΒ
cally tensor is visualized as a geometric object having a magnitude and one or
many associated directions. Tensors do not depend on any particular coordinate
system (invariant) and its characteristic is always linear. In more precise way,
we can define tensor as a multilinear mapping in Euclidian space 1 . One of
the important characteristics of tensor is its rank2 . On the basis of the rank of
1
Euclidian space R π
is an n-dimensional vector space defined over the field of real numbers.
2
The rank or order of a tensor provide information about the number of directions, which is
required to describe that particular tensor.
Chapter 1. Mathematical preliminaries
6. The theory of continuum and elasto-plastic materials 2
a tensor, we can say that a scalar is a tensor of rank zero and the vector is a
tensor of rank one. In general, a tensor of rank n in its Cartesian basis can be
written as:
π = ππ1,π2,Β·Β·Β· ,i π πi1 β πi2 β Β· Β· Β· β πi π . (1.1)
πi1 β πi2 represent the dyadic product between two base vectors. We use
both direct notation and index notation in this thesis. In direct notation we
use small bold Latin letters, π, π, Β· Β· Β· , for vectors and large bold Latin letters,
π΄, π΅, Β· Β· Β· , π , Β· Β· Β· , for tensors. In component form of notation a vector and a
second-order tensor are written as ππ, π΄ππ, Β· Β· Β· , respectively. The index notation is
used to store the physical quantities, like vectors, matrices, and tensors, as a set
of numbers on the computer that are used in actual computation. Throughout
this work we restrict ourself to Cartesian tensor. Therefore, there is no need
to differentiate between covariant and contravariant components. This gives
us freedom to store one form of component for computation, either we store
πi, πij...p or πi, πij...p.
Some example of tensors:
β’ The temperature, π, is a tensor of rank zero and is only described as a
magnitude.
β’ The force, πi, has magnitude as well as a direction and it is a vector,
therefore a tensor of rank one.
β’ The Cauchy stress, πij, is a tensor of rank two. It has a magnitude and
three directions consisting of nine components. It can be represented in
matrix form:
πij = [π] =
β
β
β
π11 π12 π13
π21 π22 π23
π31 π32 π33
β
β
β . (1.2)
1.1.1. The dyadic product
The dyadic product or outer product is a mathematical notation between two
vectors. Let π = (π1, π2, π3) and π = (π1, π2, π3) be two different vectors in the
three dimensional Euclidian space. Then, the dyadic product between these
two vectors are represented by π β π, and is defined as follows:
π β π = (πi πi) β (πj πj) = πi πj(πi β πj)
= π1 π1(π1 β π1) + π1 π2(π1 β π2) + π1 π3(π1 β π3)
+ π2 π1(π2 β π1) + π2 π2(π2 β π2) + π2 π3(π2 β π3)
+ π3 π1(π3 β π1) + π3 π2(π3 β π2) + π3 π3(π3 β π3). (1.3)
Chapter 1. Mathematical preliminaries β Tensor analysis
7. The theory of continuum and elasto-plastic materials 3
The symbol β represents dyadic multiplication. πi = (π1,π2,π3) and πj = (π1,π2,π3)
are the Cartesian unit base vectors that are used to express vectors and tensors
in component form. The dyadic products between the Cartesian base vectors
πi β πj is called a unit dyad.
The expression in Eq. (1.3) can be written in matrix form:
[π β π] =
β
β
β
π1 π1 π1 π2 π1 π3
π2 π1 π2 π2 π2 π3
π3 π1 π3 π2 π3 π3
β
β
β . (1.4)
The dyadic product is not commutative, the base vectors are not interchangeable:
πi β πj ΜΈ= πj β πi. (1.5)
π β π ΜΈ= π β π. (1.6)
1.1.2. Summation convention
Summation convention is also known as Einstein summation convention or
simply the Einstein notation. It was first introduced by Albert Einstein in his
general theory of relativity in 1916 (Kox, Klein, and Schulmann, 1997). In order
to understand how the summation convention works, let us consider an arbitrary
vector π in the Cartesian coordinate system, which is expressed as follows:
π = π1 π1 + π2 π2 + π3 π3 =
3βοΈ
π=1
πi πi, (1.7)
π1, π2, and π3 are components of the vector π w.r.t. three unit vectors π1 =
(1, 0, 0) , π2 = (0, 1, 0), and π3 = (0, 0, 1). Removal of the summation symbol
from Eq. (1.7), gives us the following form:
π = π1 π1 + π2 π2 + π3 π3 = πi πi. (1.8)
The expression in Eq. (1.8) is known as Einstein summation convention. AcΒ
cording to this convention, the indices are summed from one to three, if they
are occurring twice in one term. Mathematically the range of summation can be
generalized for n dimensions also, but in this work we restrict ourself to three
dimensional problems. In order to apply the Einstein summation, we need to
distinguish between dummy index and free index. If an index appears twice in a
term then it is known as a dummy index. On the other hand, an index occurring
only once in a term is known as a free index. The number of free indices available
in a term represent its rank.
Chapter 1. Mathematical preliminaries β Tensor analysis
9. The theory of continuum and elasto-plastic materials 5
1.2. Differentiation
The process of a derivative computation is known as differentiation. From a
geometrical point of view a derivative is defined as the slope of a tangent line on
a given curve. Physically it is defined as a rate of change of a dependent variable
w.r.t. an independent variable. In this section, we will discuss mainly three
different types of derivatives, which is widely used in this document, namely: (i)
partial derivatives, (ii) total derivatives, and (iii) material derivatives.
1.2.1. The partial derivatives
Let π(π₯1, π₯2, Β· Β· Β· , π₯n) be a multivariable real valued function in R π. Then, the
process of differentiating the function π w.r.t. a single variable while holding all
other independent variables constant, is known as partial derivatives:
ππ
ππ₯ π
= lim
ββ0
π(π₯1, Β· Β· Β· , π₯ π + β, Β· Β· Β· , π₯ π) β π(π₯1, Β· Β· Β· , π₯ π, Β· Β· Β· , π₯ π)
β
. (1.15)
If the limit in Eq. (1.15) does not exist, then we say that the partial derivative is
not defined.
1.2.2. The total derivatives
Total derivative of a function, π(π₯1, π₯2, Β· Β· Β· , π₯n), express a change in a function
due to a change in each dependent variable. It is expressed over all sum of the
change in the dependent variable time the rate of change of the function, π, w.r.t.
that dependent variable. Mathematically it is written as:
dπ =
ππ
ππ₯1
dπ₯1 +
ππ
ππ₯2
dπ₯2 + Β· Β· Β· +
ππ
ππ₯ π
dπ₯ π. (1.16)
If the variables (π₯1, π₯2, Β· Β· Β· , π₯n) are the function of scalar variable π‘, then the
total derivative is given by:
dπ
dπ‘
=
dπ
dπ₯1
dπ₯1
dπ‘
+
dπ
dπ₯2
dπ₯2
dπ‘
+ Β· Β· Β· +
dπ
dπ₯ π
dπ₯ π
dπ‘
. (1.17)
1.2.3. The material derivative
The time derivative of a physical or kinematic property of a material particle is
known as the material time derivative. In other words, we can say that a time
derivative of a given quantity holding reference configuration fixed, is known as
material derivative. There are two different ways to express the material time
Chapter 1. Mathematical preliminaries β Differentiation
10. The theory of continuum and elasto-plastic materials 6
derivative, namely: Lagrangean description and Eulerian description. From
a mathematical point of view both are equivalent.
The Lagrangean description
Let, π (π, π‘) be a scalar or a vector field, which is a differential w.r.t. time. Then,
the material derivative or the Lagrangean derivative is defined as:
Λπ =
(οΈ
dπ
dπ‘
)οΈ
π=constant
:=
ππ (π, π‘)
ππ‘
. (1.18)
The Eulerian description
Let, π (π₯, π‘) be a scalar or a vector field. In order to find the material derivative
of the given function , π, we write the position vector, π₯, as:
π = π (π₯, π‘) = π (π₯(π,π‘),π‘) . (1.19)
Then, the material derivative, according to Eq. (1.18), is given by:
dπ
dπ‘
=
(οΈ
π(π₯(π, π‘), π‘)
ππ‘
)οΈ
πβfixed
. (1.20)
If the vector π₯ in Cartesian coordinate system have three components, (π₯1, π₯2, π₯3),
then the Eq. (1.20) can be written as:
(οΈ
ππ(π₯(π, π‘), π‘)
ππ‘
)οΈ
πβfixed
=
ππ(π₯(π, π‘), π‘)
ππ₯1
ππ₯1(π,π‘)
ππ‘
+
ππ(π₯(π, π‘), π‘)
ππ₯2
ππ₯2(π,π‘)
ππ‘
+
ππ(π₯(π, π‘), π‘)
ππ₯3
ππ₯3(π,π‘)
ππ‘
+
ππ(π₯(π, π‘), π‘)
ππ‘
=
ππ(π₯(π, π‘), π‘)
ππ₯π
ππ₯π(π,π‘)
ππ‘
+
ππ(π₯(π, π‘), π‘)
ππ‘
. (1.21)
Substituting π = π(π₯, π‘) in the above equation, leads us to the result:
(οΈ
dπ(π₯(π, π‘), π‘)
dπ‘
)οΈ
π=π(π₯,π‘)
=
ππ(π₯,π‘)
ππ₯π
π£π(π₯,π‘) +
ππ(π₯,π‘)
ππ‘
= π£ Β· (βπ) +
ππ(π₯,π‘)
ππ‘
. (1.22)
Here, π£ is the Eulerian velocity. The term ππ(π₯,π‘)
ππ‘ , is known as Eulerian rate
of change and the term, π£ Β· (βπ), is called convective rate of change. β is a del
or nabla operator, discussed in the next section.
Chapter 1. Mathematical preliminaries β Differentiation
11. The theory of continuum and elasto-plastic materials 7
1.3. The differential operators
Five different differential operators, del, gradient, divergence, curl and the
Laplace operator, in the Cartesian coordinate system, will be discussed in this
section. All quantities used for these operators are assumed to be continuous
and differentiable.
Del operator
A del operator also known as nabla operator is used as a shorthand notation to
make many mathematical expression simple, and is defined as:
β = π1
π
ππ₯1
+ π2
π
ππ₯2
+ Β· Β· Β· + π π
π
ππ₯ π
= ππ
π
ππ₯π
. (1.23)
A nabla operator is not commutative in nature with the quantity on which it is
applied:
βπ ΜΈ= πβ. (1.24)
Here, π is a scalar function. With the help of the del operator we can construct
all main operators of vector calculus, namely: gradient, divergence, and curl.
Gradient of a tensor field
When the nabla symbol is used on a scalar quantity, then we get a new term
known as the gradient. If π(π₯1, π₯2, Β· Β· Β· , π₯n) is a scalar function, then the gradient
of π represents its directional derivative and is a vector quantity. In the Cartesian
coordinate system, the gradient of a function π(π₯1, π₯2, Β· Β· Β· , π₯n) is defined as:
grad π =
ππ
ππ₯1
π1 + Β· Β· Β· +
ππ
ππ₯ π
π π =
ππ
ππ₯π
ππ = βπ. (1.25)
Geometrically, we can say that the gradient of a function is always orthogonal
to the level curves (in two dimension) or normal to the level surfaces (in three
dimension). The magnitude of βπ represent the rate of change of the function π
in the direction of grad π, see Fig. 1.1. The gradient of a vector field π is defined
as:
grad π =
π(ππ ππ)
ππ₯ π
=
πππ
ππ₯ π
π π β ππ. (1.26)
The gradient of π is the second order tensor. Likewise we can define the gradient
of a second order tensor π :
βπ =
π(πππ ππ β π π)
ππ₯ π
β π π =
ππππ
ππ₯ π
π π β ππ β π π. (1.27)
Chapter 1. Mathematical preliminaries β The differential operators
12. The theory of continuum and elasto-plastic materials 8
Figure 1.1.: Geometrical representation of the gradient of a function, π(π₯), to the level
surface π(π₯) = constant (Lubarda, 2001).
The gradient of a second order tensor is a tensor of third order. In general we
can say that the gradient of the πth order tensor is a tensor of order π + 1.
Divergence of a tensor field
The divergence of a tensor field tells us that how much a tensor field diverges
from a given point. The divergence is positive if the field is spreading out and is
negative if the tensors or vectors are coming closer together. Mathematically it
is defined as the dot product between the nabla operator and vectors or tensors:
div = β Β· (β) =
π
ππ₯i
πi Β· (β). (1.28)
Applying the above definition of the divergence on a vector π£ gives:
div π£ = β Β· π£ =
ππ£i
ππ₯j
πi Β· πj =
ππ£i
ππ₯j
πΏij =
ππ£i
ππ₯i
= π£i,i. (1.29)
The divergence of a vector field gives the scalar. Likewise, the divergence of a
second-order tensor is a vector and is defined as:
div π = β Β· π =
ππij
ππ₯π
π π. (1.30)
Chapter 1. Mathematical preliminaries β The differential operators
13. The theory of continuum and elasto-plastic materials 9
We can not define the divergence of a scalar because tensors smaller than 0th-order
do not exists.
Curl of a tensor field
The result of the cross product between nabla symbol and vector or tensor is
known as curl. For a given vector π in a three dimensional Cartesian coordinate
system, the curl is defined as:
rot π = det
β
β
β
β
β
β
β
π1 π2 π3
π
ππ₯1
π
ππ₯2
π
ππ₯3
π1 π2 π3
β
β
β
β
β
β
β
. (1.31)
The tensor notation of the curl for the πth component is more compact and is
given as:
β Γ π =
(οΈ
πi
π
ππ₯i
)οΈ
Γ (πj πj) =
ππj
ππ₯i
πi Γ πj = πkij
ππj
ππ₯i
πk. (1.32)
The curl of a vector field results in a vector. Similarly we can define the curl of
a second-order tensor:
β Γ π = πijk
ππππ
ππ₯π
πk β πl. (1.33)
The Laplace operator of a tensor field
The Laplace operator is named after the French mathematician and astronomer
Pierre Simon marquis de Laplace (1749-1827) (Selvadurai, 2000). The Laplace
operator is given as:
β Β· β(β) :=
π2(β)
ππ₯π ππ₯π
= β2
(β). (1.34)
The Laplacian of a scalar π is given by:
div grad π = β2
π =
π2 π
ππ₯π ππ₯π
. (1.35)
The Laplacian of a vector π is given by:
β2
π =
π2 πk
ππ₯π ππ₯π
πk. (1.36)
An alternative way to write the Laplacian of a vector π is:
β2
π = β (β Β· π) β β Γ (β Γ π) . (1.37)
Chapter 1. Mathematical preliminaries β The differential operators
17. The theory of continuum and elasto-plastic materials 13
2. Fundamental of continuum
mechanics
The continuum is that which is divisible into indivisible that are infinitely
divisible Physics.
β Aristotle (384β322 BCE)
Continuum mechanics is a special branch of physics that is based on the hypothesis
of continuum. According to this hypothesis a material is assumed to be continuous
if its distribution fills the space completely which is occupied by it, without
leaving any void. In other words, we can say that the mathematical functions that
are applied in the theory of classical continuum mechanics should be continuous
and differentiable. The aim of this chapter is to develop geometrical concept of
motion and deformation for a continuum body, i.e. kinematics of the continuous
medium. In the second part of this chapter we derive the fundamental governing
laws that are valid for all continuous medium, i.e. conservation laws of classical
physics, namely: balance of mass, linear momentum, and energy. We will develop
global as well as local form of the balance laws. These balance laws will be given
in both Lagrangean- and Eulerian description.
2.1. Kinematics of the continuum bodies
Kinematics is the geometrical description of motion, where forces are not considΒ
ered to be the cause of motion. In continuum mechanics we use two different
ways to describe the motion, namely: Lagrangean- and Eulerian description,
respectively. Motion in both the descriptions remains same, it is just a matter
of defining the motion from a different point of view. The idea of the LaΒ
grangean description also known as the referential description, was developed
in the mid-eighteenth century by Euler. In this description we analyze motion
of a individual particle4. Therefore, the Lagrangean description of motion is
very well suited for the theory of classical elasticity and continuum mechanics.
Abaqus also uses the Lagrangean description for the mechanical modeling
capabilities (Simulia, 2015b).
4
With particle we mean an arbitrary material point of a body.
Chapter 2. Continuum mechanics
21. The theory of continuum and elasto-plastic materials 17
line element) and a line element vector in the current configuration, dπ₯, (spatial
line element), see Fig. 2.3. In other words we can say that, the deformation
gradient build a relation between the distances in the reference- and current
configurations. Mathematically, it is understood as the linear approximation to
the mapping π. Applying Taylor series for a first order approximation for the
material- and spatial line element leads us to the following relation:
π₯π + dπ₯π = πi (π π½ + dπ π½ ) = ππ (π π½ ) +
(οΈ
ππi
ππJ
)οΈ
πJ
dπJ = π₯i + πΉiJ dπJ. (2.7)
Simplifying the Eq. (2.7), gives the following:
Figure 2.3.: Mapping of a line element vector dπJ in the reference configuration to a
line element vector dπ₯i in the current configuration (Mase and Mase, 1999).
dπ₯π = πΉππ½ dπ π½ . (2.8)
Here, πΉiJ is known as the deformation gradient. Substituting Eq. (2.3) in Eq. (2.8),
leads us to a another definition of the deformation gradient:
πΉiJ =
ππi
ππJ
=
ππ₯i
ππJ
= π₯i,J =
β
β
β
β
ππ₯1
ππ1
ππ₯1
ππ2
ππ₯1
ππ3
ππ₯2
ππ1
ππ₯2
ππ2
ππ₯2
ππ3
ππ₯3
ππ1
ππ₯3
ππ2
ππ₯3
ππ3
.
β
β
β
β . (2.9)
From Eq. (2.9), we can say that the deformation gradient is the derivative of the
function, π₯ = π(π,π‘), w.r.t. π. The inverse of the deformation gradient is given
as:
πΉβ1
π½π =
ππβ1
π½
ππ₯π
=
ππ π½
ππ₯π
. (2.10)
Chapter 2. Continuum mechanics β Kinematics of the continuum bodies
22. The theory of continuum and elasto-plastic materials 18
If no motion is taking place, i.e. π₯i = πJ, then the deformation gradient is equal
to unit tensor:
πΉππ½ = πΏππ½ . (2.11)
2.1.3. Transformation of volume elements
In this section, we develop the relationship between the infinitesimal volume
elements (Parallelepiped)6 in the reference- and current configurations. The
edges of the parallelepiped in the reference configuration is given by the vectors
dπ1, dπ2 and dπ3, respectively, see Fig. 2.4. The components of these vectors
w.r.t. the orthonormal basis can be written as:
dπ1 = dπ1 πΈ1, dπ2 = dπ2 πΈ2, dπ3 = dπ3 πΈ3. (2.12)
Here, πΈ1, πΈ2 and πΈ3 are the Cartesian basis vectors for the undeformed conΒ
figuration. In order to compute volume of the parallelepiped in the reference
Figure 2.4.: Schematic representation of a volume element in the reference configuration
and the deformed configuration (Martinec, 2000).
6
Parallelepiped is used as the elementary geometry to calculate any arbitrary volume in R3
(Dreyer, 2014)
Chapter 2. Continuum mechanics β Kinematics of the continuum bodies
23. The theory of continuum and elasto-plastic materials 19
configuration (material volume element), we use triple product7:
dπ0 = dπ1 Β· (dπ2 Γ dπ3) ,
= dπ1 dπ2 dπ3 πΈ1 Β· (πΈ2 Γ πΈ3) ,
= πΈ1 Β· (πΈ2 Γ πΈ3)
β β
=1
dπ1 dπ2 dπ3,
= dπ1 dπ2 dπ3. (2.13)
Likewise, we compute volume of the parallelepiped in the current configuration
(spatial volume element):
dπ = dπ₯1 Β· (dπ₯2 Γ dπ₯3) . (2.14)
Here, dπ₯1, dπ₯2 and dπ₯3 are edges of the spatial volume element, and is computed
by using Eq. (2.8):
dπ₯1 = πΉ dπ1 =
ππ₯
ππ
dπ1 = dπ1
ππ₯
ππ
πΈ1, (2.15a)
dπ₯2 = πΉ dπ2 =
ππ₯
ππ
dπ2 = dπ2
ππ₯
ππ
πΈ2, (2.15b)
dπ₯3 = πΉ dπ3 =
ππ₯
ππ
dπ3 = dπ3
ππ₯
ππ
πΈ3. (2.15c)
Substituting Eq. (2.15) in Eq. (2.14), gives us volume of the deformed volume
element:
dπ = πΈ1 Β· (πΈ2 Γ πΈ3)
β β
=1
dπ1 dπ2 dπ3
ππ₯
ππ1
Β·
(οΈ
ππ₯
ππ2
Γ
ππ₯
ππ3
)οΈ
,
= dπ1 dπ2 dπ3
ππ₯
ππ1
Β·
(οΈ
ππ₯
ππ2
Γ
ππ₯
ππ3
)οΈ
. (2.16)
The term,
ππ₯
ππ1
Β·
(οΈ
ππ₯
ππ2
Γ
ππ₯
ππ3
)οΈ
, is known as the determinant of the deformation
gradient πΉ , which is nothing but Jacobian:
π½ = det (πΉ ) = det
(οΈ
ππ₯
ππ
)οΈ
=
ππ₯
ππ1
Β·
(οΈ
ππ₯
ππ2
Γ
ππ₯
ππ3
)οΈ
. (2.17)
Hence, the change in the volume of parallelepiped from the reference configuration
to the current configuration is given by the following relationship:
dπ = π½ dπ0. (2.18)
7
It is defined as the dot product of one of the vectors with the cross product of other two.
Chapter 2. Continuum mechanics β Kinematics of the continuum bodies
24. The theory of continuum and elasto-plastic materials 20
The Jacobian is the measure of the volume change. The volume of an element
can not be negative. It means, the physical interpretation of the volume elements
are only possible, if its volumes are positive. Mathematically, it means that J
must be always positive definite. If there is no motion, then the determinant of
the deformation gradient is equal to the determinant of unit tensor:
π½ = det πΉ = det πΌ = 1, (2.19)
2.1.4. Transformation of surface area elements
In this section, we develop a relationship between the infinitesimal small surface
elements in the reference-, dπ΄0, and deformed, dπ΄, configurations. The surface
element dπ΄0 (material surface element) is constructed by using two infinitesimal
material line elements, dπ1 and dπ2, respectively, see Fig. 2.5. Then, the area
Figure 2.5.: Schematic representation of a area element in the reference configuration
and the deformed configuration (Abeyaratne, 2012).
can be computed by using cross product:
dπ΄0 = dπ1 Γ dπ2 = π0 dπ΄0. (2.20)
Here, π0 is the unit normal vector to the material surface element. The area
of the surface element dπ΄ (spatial surface element), is defined in the deformed
configuration by the following relation:
dπ΄ = dπ₯1 Γ dπ₯2 = π dπ΄. (2.21)
Here, π is the unit normal to the spatial surface element. dπ₯1, dπ₯2 are the
two material line elements in the deformed configuration, see Fig. 2.5. Using
Eq. (2.18), we get:
dπ₯3 Β· (dπ₯1 Γ dπ₯2) = π½ dπ3 Β· (dπ1 Γ dπ2) . (2.22)
Chapter 2. Continuum mechanics β Kinematics of the continuum bodies
25. The theory of continuum and elasto-plastic materials 21
Substituting Eq. (2.20) and Eq. (2.21) in Eq. (2.22) leads us to the following
relation:
dπ₯3 Β· π dπ΄ = dπ3 Β· π0 dπ΄0. (2.23)
Application of Eq. (2.8) in Eq. (2.23) gives us the relationship between the
material- and spatial volume element:
(πΉ dπ3) Β· π dπ΄ = π½ dπ3 Β· π0 dπ΄0. (2.24)
An alternative form of the above equation is:
π Β· dπ΄ = π½πΉ βπ
Β· π0 dπ΄0. (2.25)
The method which is used in this section is valid for any surface element of
a continuum body, but the direction of the unit normal vector may differ.
Eq. (2.25) is widely known as Nansonβs relation, named after Edward J. Nanson
(1850β1936) (Bertram, 2012).
2.1.5. Measures of deformation
In order to do the physical and geometrical interpretation of deformations, we
need to understand the concept of different strain tensors. There exist various
definitions and names of strain tensors, but in this work we limit ourself to
the most common strain tensors, which are widely used in classical continuum
mechanics. The strain tensors, which are discussed in this section, are either
related to the reference configuration or the current configuration.
Let us begin our discussion with a material line element dπΏ. The square of this
material line element gives:
dπΏ2
= dπ πΌ dπ πΌ. (2.26)
Similarly, the square of the spatial line element can be given as:
dπ2
= dπ₯π dπ₯π. (2.27)
Application of Eq. (2.8) in Eq. (2.27) gives us the following relation:
dπ2
= (πΉππΌ dπ πΌ) (πΉππ½ dπ π½ ) = (πΉππΌ πΉππ½ ) dπ πΌ dπ π½ = πΆ πΌπ½ dπ πΌ dπ π½ . (2.28)
Here, πΆIJ is known as right Cauchy-Green deformation tensor, which is defined
as:
πΆ πΌπ½ = πΉππΌ πΉππ½ =
ππ₯π
ππ πΌ
ππ₯π
ππ π½
. (2.29)
Right Cauchy-Green deformation tensor is named after the French matheΒ
matician Augustin-Louis Cauchy (1789 - 1857) and the British mathematical
Chapter 2. Continuum mechanics β Kinematics of the continuum bodies
26. The theory of continuum and elasto-plastic materials 22
physicist George Green (1793 - 1841). It is defined only in the reference conΒ
figuration and is symmetric and positive definite tensor. From Eq. (2.29) we
can conclude that if all the nine components of deformation gradient, πΉiI, are
known then we can compute the six components of the right Cauchy-Green
deformation tensor. But, the reverse is not possible, i.e. if we know πΆIJ, then it
is not possible to compute πΉiI (Holzapfel, 2000).
The second most commonly used strain tensor in continuum mechanics is the
left Cauchy-Green deformation tensor. It is defined as:
π΅ππ = πΉππΎ πΉ ππΎ =
ππ₯π
ππ πΎ
ππ₯ π
ππ πΎ
. (2.30)
The tensor π΅ij is only defined in the current configuration. It is also known as
Finger deformation tensor, named after the Austrian physicist and mathematiΒ
cian Josef Finger (1841 - 1925). Like the right Cauchy-Green strain tensor,
π΅ij is also a symmetric and positive definite second-order tensor.
Another most commonly used strain tensor is the Lagrangean strain tensor.
In order to define this tensor, let us subtract Eq. (2.28) from Eq. (2.26):
dπ2
β dπΏ2
= (πΆIJ β πΏIJ) dπ πΌ dπ π½ = 2 πΈ πΌπ½ dπ πΌ dπ π½ . (2.31)
Here, πΈIJ is known as Lagrangean strain tensor or Greenβ St-Venant strain
tensor, which is associated with the reference configuration, and is defined as
follows (Bertram, 2012):
πΈIJ =
1
2
(πΆIJ β πΏ πΌπ½ ) =
1
2
(πΉππΌ πΉiJ β πΏ πΌπ½ ). (2.32)
Since the right Cauchy-Green strain tensor, πΆIJ, and the Kronecker delta,
πΏIJ, are symmetric in nature, then from Eq. (2.32) we can conclude that πΈIJ is
also symmetric in nature.
The last most commonly used strain tensor is associated with the current
configuration and is known as the Eulerian - Almansi strain tensor. Like other
strain tensors, this is also symmetric in nature and is defined as:
πππ =
1
2
(οΈ
πΏππ β πΉβ1
πΎπ πΉβ1
πΎπ
)οΈ
. (2.33)
In general, πΆIJ and πΈIJ are also called material strain tensors because they are
associated with the reference configuration. Tensors, which are associated with
the current configuration, like π΅ij and πij are also known as spatial strain tensors.
The Lagrangean- and Eulerian - Almansi strain tensors can also be expressed
in term of displacement vector gradient. In order to express πΈIJ and πij by using
displacement vector gradient, let us start by computing the displacement field of
a typical particle. The displacement field is a vector field, which is defined as the
Chapter 2. Continuum mechanics β Kinematics of the continuum bodies
28. The theory of continuum and elasto-plastic materials 24
Lagrangean strain tensor:
πΈ πΌπ½ =
1
2
(οΈ
πΏ ππ
(οΈ
ππ π
ππ πΌ
+ πΏ π πΌ
)οΈ
πΏ ππ
(οΈ
ππ π
ππ π½
+ πΏ π πΏ
)οΈ
β πΏ πΌπ½
)οΈ
,
=
1
2
(οΈ
πΏ π π
(οΈ
ππ π
ππ πΌ
+ πΏ π πΌ
)οΈ (οΈ
ππ π
ππ π½
+ πΏ π π½
)οΈ
β πΏ πΌπ½
)οΈ
,
=
1
2
(οΈ(οΈ
ππ π
ππ πΌ
+ πΏ π πΌ
)οΈ (οΈ
ππ π
ππ π½
+ πΏ π π½
)οΈ
β πΏ πΌπ½
)οΈ
,
=
1
2
(οΈ
ππ πΌ
ππ π½
+
ππ π½
ππ πΌ
+
ππ π
ππ πΌ
ππ π
ππ π½
)οΈ
. (2.37)
Similarly, we can express the Eulerian - Almansi strain tensor in the following
form:
πππ =
1
2
(οΈ
ππ’π
ππ₯ π
+
ππ’ π
ππ₯π
β
ππ’ π
ππ₯π
ππ’ π
ππ₯ π
)οΈ
. (2.38)
Here, π’i is a function of the current position vector π₯ and time π‘ and is known
as displacement field in the Eulerian form (spatial description):
π’π (π₯, π‘) = π₯π β π πΌ. (2.39)
Displacement fields in the Lagrangean and the Eulerian form have the same
values, which are connected by the following relationship:
ππ’π
ππ₯ π
β
ππ’π
ππ₯π
(οΈ
πΏ πΎπ½ +
ππ πΎ
ππ π½
)οΈ
=
ππ’π
ππ₯ π
ππ π
ππ π½
=
ππ πΌ
ππ π½
. (2.40)
Strain tensors defined in Eq. (2.37) and Eq. (2.38) are valid for any arbitrary
deformation. In this work we restrict ourself only to small deformations. ThereΒ
fore, nonlinear terms in the Lagrangean- and Eulerian - Almansi strain
tensor will be neglected. In the case of small deformation, we do not differentiate
between the reference- and current configurations, so that we get the simple
form of linear strain tensor πij:
πΈIJ β πij β πij =
1
2
(οΈ
ππ’i
ππ₯j
+
ππ’j
ππ₯i
)οΈ
. (2.41)
2.2. Field equations of continuum mechanics
Balance laws of mass, momentum, and energy are the heart of the continuum
physics theories. These balance laws are used to interpret the effect of the
surroundings on the motion of a material body.
In this work, we concentrate only on the balance laws of those fields, which are
additive in nature and also continuous and differentiable. The system in which
Chapter 2. Continuum mechanics β Field equations of continuum mechanics
38. The theory of continuum and elasto-plastic materials 34
3. Constitutive models for elastic
materials
There is nothing that can be said by mathematical symbols and relations
which cannot also be said by words. The converse, however, is false.
Much that can be and is said by words cannot successfully be put into
equations, because it is nonsense.
β Clifford Truesdell (February 18, 1919 β January 14, 2000)
In the previous chapter we discussed the governing equations of continuum meΒ
chanics, which are valid for any continuous medium. These governing equations,
namely: balance of mass, linear momentum, and energy provide five partial
differential equations with sixteen unknowns (stress tensor is assumed to be
symmetric) for a continuous thermo-mechanical system. In order to compute
all these unknowns we need to construct these missing equations, which are
known as constitutive relations. Using constitutive relations, we can describe
behavior of the material under the thermal and mechanical loading. In this
chapter we discuss the Hookeβs law for anisotropic and isotropic solid material.
Constitutive relations should follow the principle of objectivity, which means
they should be independent of the observer.
3.1. Generalized Hookeβs law
The original law of Robert Hooke, an English physicist, was published in the
year of 1676. This law states βIf the force applied on the per unit area of any
elastic material, then it is linearly proportional to the amount of elongation and
compressionβ. According to (Truesdell, 1968), James Bernoulli (1704) was the
one, who had given the first instance of a true stress-strain relation. For uniaxial
loading the Hookeβs law can be given as the following:
π βΌ π, (3.1)
π = πΈ π. (3.2)
The proportional factor πΈ is known as elasticity modulus or Youngβs modulus,
named after the English polymath Thomas Young. It is a material parameter
Chapter 3. Constitutive models for elastic materials
39. The theory of continuum and elasto-plastic materials 35
that tells how much a material is going to deform under tension and compresΒ
sion. π represent axial stress and π is strain produced due to this axial stress.
In general, the properties of many materials vary according to their crystalloΒ
graphic orientations. In such cases Eq. (3.2) is not more valid. Therefore, it is
necessary to develop the generalized form of Hookeβs law. The Hookeβs law
for three dimensional case can be derived from the strain energy density8 and
mathematically it is given as:
πij = πΆijkl πkl. (3.3)
πij and πkl are the second order stress and strain tensors, respectively. πΆijkl is
elastic stiffness of the material. It is a fourth order tensor with 81 components.
The law stated in Eq. (3.3) is only valid for small deformations. Abaqus
also uses Eq. (3.3) to model elasticity for small-strain problems and also for an
elastic-plastic model in which elastic strains are considered to be small (Simulia,
2015b).
Inverting Eq. (3.3) allows us to relate strain to stress:
πππ = π ππππ π ππ. (3.4)
Here, π ijkl is known as compliance tensor.
Computation of a matrix, πΆijkl, with 81 components is a bit time consuming
process. Therefore, we use the symmetry property to reduce the number of
elements in the stiffness matrix. For i = 1, j = 2 and for i = 2, j = 1, Eq. (3.3)
can be written as the following:
π12 = πΆ12kl πkl, π21 = πΆ21kl πkl. (3.5)
According to Cauchyβs second law of motion or Boltzmannβs axiom, the
Cauchyβs stress tensor is considered to be symmetric, if the medium is non-polar
(Bertram, 2012). In this work we assume a non-polar medium. Therefore, πij is
symmetric, which leads us to the following relation:
π12 = π21 β (πΆ12kl β πΆ21kl) πkl = 0 β πΆ12kl = πΆ21kl. (3.6)
The stiffness matrix in Eq. (3.6) shows minor symmetry, due to which the number
of its components reduced from 81 to 54. The number of elements in πΆijkl can be
8
For complete derivation please refer (Hahn, 1985)
Chapter 3. Constitutive models for elastic materials β Generalized Hookeβs law
40. The theory of continuum and elasto-plastic materials 36
reduced further from 54 to 36 by considering the symmetry of the strain tensor:
β
β
β
β
β
β
β
β
β
β
β
β
πΆ1111 πΆ1122 πΆ1133 πΆ1112 πΆ1113 πΆ1123
πΆ2211 πΆ2222 πΆ2233 πΆ2212 πΆ2213 πΆ2223
πΆ3311 πΆ3322 πΆ3333 πΆ3312 πΆ3313 πΆ3323
πΆ1211 πΆ1222 πΆ1233 πΆ1212 πΆ1213 πΆ1223
πΆ1311 πΆ1322 πΆ1333 πΆ1312 πΆ1313 πΆ1323
πΆ2311 πΆ2322 πΆ2333 πΆ2312 πΆ2313 πΆ2323
β
β
β
β
β
β
β
β
β
β
β
β
. (3.7)
Use of Voigt notation permits us to write the stiffness matrix in compact form,
and at the same time this compact form will give more clearance to analyze πΆijkl.
Using this notation, we can write the stiffness matrix πΆijkl as the following:
β
β
β
β
β
β
β
β
β
β
β
β
πΆ11 πΆ12 πΆ13 πΆ14 πΆ15 πΆ16
πΆ21 πΆ22 πΆ23 πΆ24 πΆ25 πΆ26
πΆ31 πΆ32 πΆ33 πΆ34 πΆ35 πΆ36
πΆ41 πΆ42 πΆ43 πΆ44 πΆ45 πΆ46
πΆ51 πΆ52 πΆ53 πΆ54 πΆ55 πΆ56
πΆ61 πΆ62 πΆ63 πΆ64 πΆ65 πΆ66
β
β
β
β
β
β
β
β
β
β
β
β
. (3.8)
The number of elements of the stiffness matrix πΆijkl can be further reduced to
21, if we use the existence of equivalence of the mixed partials:
πΆijkl =
π2 π
ππkl ππij
=
π2 π
ππij ππkl
= πΆklij. (3.9)
Using Eq. (3.6), Eq. (3.9) and the strain symmetry, we can establish the following
identity for the stiffness matrix, πΆijkl:
πΆijkl = πΆjikl, πΆijkl = πΆijlk, and πΆijkl = πΆklij. (3.10)
Using symmetrical property of the stiffness matrix πΆijkl, as shown in Eq. (3.10),
and the Voigt notation, Eq. (3.3) can be written as the following:
β
β
β
β
β
β
β
β
β
π11
π22
π33
π23
π31
π12
β
β
β
β
β
β
β
β
β
=
β
β
β
β
β
β
β
β
β
πΆ11 πΆ12 πΆ13 πΆ14 πΆ15 πΆ16
πΆ22 πΆ23 πΆ24 πΆ25 πΆ26
πΆ33 πΆ34 πΆ35 πΆ36
πΆ44 πΆ45 πΆ46
πΆ55 πΆ56
sym. πΆ66
β
β
β
β
β
β
β
β
β
β
β
β
β
β
β
β
β
β
π11
π22
π33
2π23
2π13
2π12
β
β
β
β
β
β
β
β
β
. (3.11)
Here, the stiffness matrix, πΆijkl, has 21 independent components.
Chapter 3. Constitutive models for elastic materials β Generalized Hookeβs law
43. The theory of continuum and elasto-plastic materials 39
Substituting Eq. (3.22) in Eq. (3.3), gives the Hookeβs law for isotropic material
due to thermoelastic effect:
πij = πΆijkl πkl
= πΆijkl
(οΈ
πkl β πth
kl
)οΈ
= πΆijkl (πkl β πΌ Ξπ πΏkl)
= πΆijkl πkl β πΆijkl πΌ Ξπ πΏkl. (3.23)
Substituting the isotropic stiffness matrix from Eq. (3.14) in Eq. (3.23), gives us
the following relation:
πij = π πkk πΏij + 2π πij β (π πΏij πΏkl + π (πΏik πΏjl + πΏil πΏjk)) πΌ Ξπ πΏkl
= π πkk πΏij + 2π πij β π πΌ Ξπ πΏij πΏkl πΏkl β π πΌ Ξπ πΏik πΏjl πΏkl β π Ξπ πΏil πΏjk πΏkl.
(3.24)
In order to simplify Eq. (3.24), we need to compute the product of Kronecker
delta and use its substitution property:
πΏkl πΏkl = 3, πΏik πΏjl πΏkl = πΏij, πΏil πΏjk πΏkl = πΏij. (3.25)
Substituting Eq. (3.25) in Eq. (3.24), gives:
πij = π πkk πΏij + 2π πij β 3π πΌ Ξπ πΏij β 2π πΌ Ξπ πΏij. (3.26)
Then, the Hookeβs law for an isotropic material taking thermoelastic effect
under consideration is given by the following relation:
πij = π (πkk β 3πΌ Ξπ) πΏππ + 2π (πij β πΌ Ξπ πΏij) . (3.27)
Chapter 3. Constitutive models for elastic materials β Generalized Hookeβs law
44. The theory of continuum and elasto-plastic materials 40
4. Constitutive models for
elasto-plastic materials
A scientific truth does not triumph by convincing its opponents and
making them see the light, but rather because its opponents eventually
die and a new generation grows up that is familiar with it.
β Max Karl Ernst Ludwig Planck (April 23, 1858 β October 4, 1947)
The theory of elasto-plasticity belongs to the branch of mechanics in which we
study the behavior of such material, that initially goes under elastic deformation
but upon reaching certain criteria, i.e. an initial yield stress πY0 it shows
permanent (irreversible) deformation. The aim of this chapter is to give a
brief introduction to time-independent plasticity theory and develop constitutive
equations.
Continuum plasticity should fulfill important requirements, which arise from
the theory of micro-plasticity. The first requirement is the condition of incomΒ
pressibility. According to this condition, the volume of a body does not change
during and after deformation because plastic slip, occurring during deformation,
does not lead to volume change. Second important requirement states that the
material response should be independent of rate effects. The fact that plastic
slip is a shearing process, tells us that plastic deformation of metal at macroΒ
scopic scale is independent of hydrostatic stress. This statement can be proved
mathematically by using crystal plasticity constitutive equation.
When the problem with small deformation is concerned, then the results provided
by the classical plasticity theory is satisfactory. But this theory gives absurd
result if it is applied for the large deformations. In order to study the large
plastic deformations, the theory of finite plasticity is popular (Bertram, 2012).
4.1. Stress-strain relationship
If the load applied on a material body9 is small, then the strain, occurring due
to the force applied per unit area, is directly proportional to the applied stress
9
Not all the material body obey this rule, but for this work we will consider only those material
body, which obey this relationship
Chapter 4. Constitutive models for elasto-plastic materials
45. The theory of continuum and elasto-plastic materials 41
at room temperature, see sec. 3.1. If the applied load reaches the elastic limit or
the yield point10 of the material, then the stress-strain curve shows nonlinear
behavior. Therefore, application of Hookeβs law is not valid beyond this elastic
limit. In other words, we can say that the material starts yielding or begins to
flow.
Removal of load beyond the initial yielding11 causes permanent deformation.
This behavior of the material body is known as plasticity. In order to compute the
plastic strain, a fundamental approach of additive decomposition is used (MΓΌller,
2014a). According to this approch, the total strain observed can be separated in
two parts, namely: reversible elastic strain, πe
ij, and irreversible plastic strain,
πp
ij, see Fig. 4.1. This assumption is only applicable, when the body undergoes
small deformation. For large deformation, the assumption of multiplicative
decomposition is used. Mathematically the strain additive decomposition is
Figure 4.1.: Elasto-plastic stress-strain curve for a uniaxial loading, with the initial
yield stress and strain decomposition (Kossa, 2011).
given as:
πij = πe
ij + πp
ij. (4.1)
Using Eq. (4.1), the plastic strain in the material body is defined as:
πe
ij = πij β πp
ij. (4.2)
10
The limit beyond which permanent deformation occurs during a tension test.
11
Generally it is not possible to differentiate between the initial- and upper yield point for all
materials.
Chapter 4. Constitutive models for elasto-plastic materials β Stress-strain relationship
46. The theory of continuum and elasto-plastic materials 42
In general, the macroscopic continuum approach is based on the usage of an
incremental strain or strain rate. We use the approach based on the strain rates
because in plasticity the study of deformation is path dependent. Therefore, the
incremental form of Eq. (4.2) can be given as:
Λπe
ij = Λπij β Λπp
ij. (4.3)
Unloading process do not obey the Hookeβs law as in Eq. (3.3), if the material
is subjected to plastic flow. Therefore, we need to modify the Hookeβs law.
In order to do that we substitute Eq. (4.3) in Eq. (3.3), which gives us the
incremental form of the Hookeβs law:
Λπππ = πΆππππ Λπe
ππ = πΆππππ
(οΈ
Λπij β Λπp
ij
)οΈ
. (4.4)
Eq. (4.3) is also used in Abaqus for most of the inelastic constitutive models.
In many engineering example it had been seen that the elastic strain πe
ij is much
smaller than the plastic strain πp
ij (Dunne and Petrinic, 2005).
If we also consider deformation due to thermal effect then Eq. (4.3) can be written
as the following:
Λπe
ij = Λπij β Λπp
ij β Λπth
ij . (4.5)
Using the above relationship the incremental form of the Hookeβs law as in
Eq. (4.4) can be written as the following:
Λπππ = πΆππππ Λπe
ππ = πΆππππ
(οΈ
Λπij β Λπp
ij β Λπth
ij
)οΈ
. (4.6)
4.2. The yield criterion
The yield criterion (condition) is used to determine the transition of a material
behavior from elastic to plastic. In other words we can say that it defines the
stress state for which the material shows plastic behavior. For a uniaxial stress
test, the yield condition is easy to determine. In uniaxial yield condition, if the
stress, π, is smaller than the initial yield stress, πY0, then the material exhibits
elastic behavior otherwise plastic. Mathematically we can say that the material
shows plastic behavior only if the following yield criterion is fulfilled (JirΓ‘sek
and Bazant, 2002):
|π| β πY0 = 0. (4.7)
In a multiaxial stress state, finding the yield condition is not that straightforward.
In this case the yield criterion is defined by a scalar function known as yield
function (Prager and Hodge, 1968):
π (πππ) = 0. (4.8)
Chapter 4. Constitutive models for elasto-plastic materials β The yield criterion
47. The theory of continuum and elasto-plastic materials 43
Here, πij is a symmetrical stress tensor. Therefore, Eq. (4.8) can be written in
the following form:
π (π11, π22, π33, π23, π31, π12) = 0. (4.9)
If the yield function is equal to zero, π (πij) = 0, then the material shows plastic
behavior and when π (πij) < 0, then the material exhibit elastic behavior.
If the material is considered to be isotropic, then the yield condition is free from
the orientation of the coordinate system. It means, the expression in Eq. (4.8)
can be expressed in the term of the invariants of the stress tensor (Hill, 1986):
π (π½1, π½2, π½3) = 0, (4.10)
Here, π½1, π½2 and π½3 are the first, second, and third invariants of the stress tensor,
respectively. These are defined by the following relations (Hill, 1986):
π½1 = π11 + π22 + π33,
π½2 = β (π11 π22 + π22 π33 + π33 π11) ,
π½3 = π11 π22 π33. (4.11)
Here π11, π22, and π33 are principal components of the stress tensor.
There exist many different yield criteria for the metals, which assume that a
hydrostatic stress always influences the yielding of a metal. This assumption
is a contradiction to the present days experimental fact. Therefore, these yield
conditions are only for historic interest. But Tresca12 and Von Mises13 yield
criterion do not have this fault and are used more often to build mathematical
model of plasticity (Hill, 1986).
In this work we limit our discussion to von Mises yield criterion only, because it
is widely used in the engineering application, mainly for computational analysis.
Von Mises yield criterion considered all three principal shear stresses to predict
the yielding of metal. The use of all three principal shear stresses gives a accurate
prediction of yielding.
4.2.1. Von Mises yield criterion
The defining equation for the Von Mises stress was independently proposed by
the Polish mechanical engineer Tytus Maksymilan Huber and Richard Von
Mises in 1904 and 1913, respectively. The physical interpretation of this equation
was given by the German engineer Heinrich Hencky in 1924. According to
12
Named after the French mechanical engineer Henri Tresca. According to this yielding
criterion, the yielding begins when the maximum shear stress reaches a critical value.
13
Named after the famous scientist and mathematician Richard von Mises
Chapter 4. Constitutive models for elasto-plastic materials β The yield criterion
48. The theory of continuum and elasto-plastic materials 44
Von Mises yield criterion, material starts flowing after reaching the initial
yield stress, πY0 (MΓΌller, 2014a). In other words, materials start yielding when
the second deviatoric stress invariant, π½2, reaches a critical value (Hill, 1986).
Mathematically the Von Mises yield criterion is expressed as:
π½2 = π2
. (4.12)
Here, π is a material parameter and can be understood as yield stress of a given
material under simple shear. The value of π is
β
3 times smaller than the tensile
yield stress in simple tension case (Prager and Hodge, 1968):
π =
πY
β
3
. (4.13)
An alternative way to express Von Mises yield criterion is given below:
πMises = πY =
βοΈ
3 π½2 . (4.14)
Here, πMises is known as Von Mises equivalent stress, and it predict the yielding
of material under multiaxial loading. Abaqus also use the Von Mises equivalent
stress to model isotropic behavior in the metal plasticity, which depends on the
deviatoric stress. In order to write Eq. (4.14) in terms of the deviatoric stress,
let us define the second deviatoric stress invariant, π½2, in an alternative form
(JirΓ‘sek and Bazant, 2002):
π½2 =
1
2
π ij π ij, (4.15)
where π ij is a deviatoric stress tensor, which is defined as the difference between
the stress tensor πij and the hydrostatic stress tensor π πΏππ (MΓΌller, 2014a). The
hydrostatic stress tensor can be obtained by taking average of the three normal
stress components of any stress tensor. Therefore, the deviatoric stress tensor π ij
can be given as:
π ij = πππ β π πΏππ = πij β
1
3
π ππ πΏij. (4.16)
Substituting Eq. (4.16) in Eq. (4.15) gives us π½2, which we substitute in Eq. (4.14)
to get Von Mises equivalent stress in term of deviatoric stress tensor:
πMises =
βοΈ
3
2
π ij π ij . (4.17)
Von Mises yield criterion as a flow function can be written as:
π (πij, πY) =
βοΈ
3
2
π ij π ij β πY. (4.18)
Chapter 4. Constitutive models for elasto-plastic materials β The yield criterion
49. The theory of continuum and elasto-plastic materials 45
Here, π π is the current yield stress. Von Mises yield criterion shows great
results for many ductile metals, like copper, nickel, aluminum, etc. This fact is
established by experiments done for these metals (Hill, 1986).
4.3. Flow rule
As we have seen that the total strain in an elasto-plastic model can be decomposed
in elastic strain, πe
ij, and plastic strain, πp
ij. All components of the elastic strain
can be determined by using the Hookeβs law. In oder to determine the plastic
strain we need to important concepts, i.e. yield conditions, which we have already
discussed, and flow rule. Flow rule is used to establish a relationship between
stress and plastic strain under multi-axial loading. Flow rule also defines the
direction of plastic flow.
In other words we can say that, the derivative of plastic potential π, which is a
function of stress, w.r.t. stress is equal to the plastic strain. Mathematically we
can write this as follows:
dπp
ij = dπ
ππ
ππij
. (4.19)
The form in Eq. (4.19) is known as non-associated flow rule. Using non-associated
flow rule, we can model a wide class of material behavior, like soil, rock, etc.
But in this work, we consider only those sub-class of materials, whose plastic
potential can be taken as the yield function, i.e. π = π. Substituting π instead
of π in Eq. (4.19), gives us a different form of flow rule, which is known as an
associated flow rule:
dπp
ij = dπ
ππ
ππij
or Λπp
ij = Λπ
ππ
ππij
. (4.20)
Here, Λπ is a non-negative hardening parameter that determine the magnitude of
the plastic strain. The direction of the plastic strain increment is given by the
following term:
ππ
ππij
, (4.21)
which is normal to the yield surface. This is the reason that an associative
flow rule is also known as normality hypothesis of plasticity. Next we apply the
associated flow rule for the Von Mises yield condition. In order to do that we
substituting the Von Mises function:
π = π½2 β
π2
Y
3
, (4.22)
Chapter 4. Constitutive models for elasto-plastic materials β Flow rule
50. The theory of continuum and elasto-plastic materials 46
in Eq. (4.20), which gives the following:
ππ
ππij
=
π
ππij
(οΈ
π½2 β
π2
Y
3
)οΈ
. (4.23)
For a given material, πY is constant, which means derivative of the term
π2
Y
3 can
be considered to be zero. This consideration simplify Eq. (4.23):
ππ
ππij
=
ππ½2
ππij
. (4.24)
Substituting Eq. (4.15) in Eq. (4.24) and then applying chain rule of differentiation
gives the following:
ππ
ππij
=
1
2
(οΈ
ππ kl
ππij
π kl + π kl
ππ kl
ππij
)οΈ
= π kl
ππ kl
ππij
. (4.25)
Substitution of Eq. (4.16) in Eq. (4.25), gives:
ππ
ππij
= π kl
π
ππij
(οΈ
πkl β
1
3
πmm πΏkl
)οΈ
= π kl
ππkl
ππij
β
1
3
π kl
π
ππij
(πmm πΏkl)
= π kl
(οΈ
ππkl
ππij
β
1
3
ππmm
ππij
πΏkl
)οΈ
= π kl
(οΈ
πΏil πΏkj β
1
3
πΏim πΏjm πΏkl
)οΈ
= π ki π kj β
1
3
π kk πΏij
= π ij β
1
3
π kk πΏij. (4.26)
According to the definition of deviatoric stress tensor:
π kk = 0. (4.27)
Substituting Eq. (4.27) in Eq. (4.26) gives us the partial derivative of the yield
function, π, w.r.t. the stress tensor, πij, which is equal to the deviatoric stress
tensor, π ij:
ππ
ππij
= π ij. (4.28)
Substituting Eq. (4.28) in Eq. (4.20), gives the associated flow rule in terms of
the deviatoric stress tensor:
Λπp
ij = Λπ π ij. (4.29)
Chapter 4. Constitutive models for elasto-plastic materials β Flow rule
51. The theory of continuum and elasto-plastic materials 47
4.4. Consistency condition
The condition that the stress should remains on the yield surface, which is the
graphical representation of the yield function π(πij), during plastic flow is known
as consistency condition. In order to visualize this concept, let us consider a
material on which we apply an external load. Because of the applied load, the
material will first deform elastically and after reaching its yield point, it starts
showing plastic behavior. If we assume that the material shows plastic behavior
with no hardening then the material will deform further plastically under the
constant stress, i.e. the load point remains on the yield surface, see Fig. 4.2.
Mathematically it means that the yield function, π, remains equal to zero, see
Fig. 4.2 and as a result the time derivative of π will vanish:
Λπ = 0. (4.30)
The above equation is only valid for plastic yielding and can not be applied
during elastic deformation. Consistency condition is useful for computing the
plastic multiplier. If π is the stress state and πY is the current yield stress. Then,
Figure 4.2.: Graphical representation of the Von Mises yield surface for plane stress
and the corresponding stress-strain curve (Dunne and Petrinic, 2005).
according to Von Mises yield criterion we can write the following relation:
π (π, πY) =
1
2
π ij π ij β
1
3
π2
Y = 0. (4.31)
According to consistency condition:
Λπ (π, πY) =
ππ
ππij
Λπij +
ππ
ππY
ΛπY = 0. (4.32)
Chapter 4. Constitutive models for elasto-plastic materials β Consistency condition
52. The theory of continuum and elasto-plastic materials 48
From Eq. (4.31), we can calculate the partial differentiation of yield function
w.r.t. current yield stress, which is equal to:
ππ
ππY
= β
2
3
πY. (4.33)
Substituting Eq. (4.28) and Eq. (4.33) in Eq. (4.32), gives us the following of the
consistency condition:
Λπ (π, πY) = π ij Λπij β
2
3
πY ΛπY = 0. (4.34)
4.5. Hardening
In the previous sections we have discussed only about materials, whose yield
surfaces have not changed their size or position during loading process, i.e.
perfectly elasto-plastic materials. In practice, we can observe the changes in
elastic domains or yield surfaces during loading or unloading, due to the changes
in the microstructure of the material as plastic flows continue. In order to
understand these changes we need to modify the initial yield surface, which gives
rise to the concept of hardening. An increase of the yield stress during plastic
deformation is known as hardening. In this section we discuss mainly two types
of hardening, namely: isotropic hardening and kinematic hardening, respectively.
4.5.1. Isotropic hardening
A hardening process, in which the yield strength or expansion in the yield surface
in all directions are uniform, is called isotropic hardening, see Fig. 4.3. For
uniaxial loading, the isotropic hardening law, which is linear in nature, is given
as:
πY = πY0 + π» πp
11. (4.35)
Here, π» is known as plastic modulus. It is defined as the slope of the non-linear
part of the stress-strain curve, obtained through uniaxial tensile test. πp
11 is
the plastic part of the normal strain, π11. Whether the material will exhibit
true hardening, perfect plasticity or softening, depends on the value of plastic
modulus. If the plastic modulus is positive then we say that the yield stress will
increase. If π» = 0, then it corresponds to perfect plasticity and if the material
has a negative value of plastic modulus, then it shows softening character.
In order to extend Eq. (4.35) for a general multiaxial case, we need to introduce
definition of the cumulative plastic strain also known as effective plastic strain
or the equivalent plastic strain πp
eq. The rate equation of the equivalent plastic
Chapter 4. Constitutive models for elasto-plastic materials β Hardening
53. The theory of continuum and elasto-plastic materials 49
Figure 4.3.: Isotropic Hardening: Left: Uniform expansion of the yield surface in stress
space with plastic deformation. Right: Stress (π) strain (π) curve, representing increase
in the yield strength due to strain hardening (Dunne and Petrinic, 2005).
strain is given as (MΓΌller, 2014a):
Λπ
p
=
βοΈ
2
3
Λπp
ππ Λπp
ππ . (4.36)
The above equation was first proposed by Odqvist without the factor
βοΈ
2
3 .
integration of Eq. (4.36) w.r.t. time gives us the cumulative plastic strain:
πp
eq =
Λ βοΈ
2
3
Λπp
ππ Λπp
ππ dπ‘. (4.37)
Substituting πp
eq instead of πp
11 in the linear isotropic hardening law we get the
hardening law for multiaxial load:
πY = πY0 + π» πp
eq. (4.38)
Derivative of current yield stress can be given as:
ΛπY = π» Λπ
p
. (4.39)
Substituting the derivative of the current yield stress in Eq. (4.34), gives us the
following relation:
Λπ
p
=
3
2
π ππ Λπππ
π» πY
. (4.40)
Applying the incremental form of the Hookeβs law as in Eq. (4.4), gives us an
Chapter 4. Constitutive models for elasto-plastic materials β Hardening