Investigation and Comparison of Cohesive Zone

Models for Simulation of Crack Propagation

Sara Eliasson and Alexander Lundberg
2015-01-13

Abstract

Looking at crack propagation modeling today there are a few different methods
that enables the possibility to simulate an advancing crack. One of these methods are cohesive zones. Cohesive zones are modeled as an interface between two
continuum surfaces and are described by a constitutive law that is to represent
the crack propagation.
The work carried out in this paper is supposed to give a deeper knowledge and a
better understanding of the cohesive zone model to be able to further implement
this method within a larger problem.
The first step is looking into linear elastic fracture mechanics to understand
the theory behind cohesive zones. Understanding the theory and trying to
implement it by using different traction-separation laws proposed by different
authors, several difficulties were encountered. The commercial FEM-program
used to implement cohesive zones, ABAQUS CAE, gives the possibility to implement linear and exponential softening in the traction-separation law. For
further implementation of the traction-separation laws presented in this paper,
user-defined subroutine ought to be used.
Using a double cantilever beam the traction-separation law with a linear initial
loading and a linear or exponential softening is implemented in ABAQUS CAE.
Controlling the penalty stiffness, the maximum traction and the fracture energy
or the fracture displacement several results were produced and compared.
Looking at the results, when using the same variables for the exponential and
linear softening, there is no big difference in the results of the behavior of the
beam. The curves for both implementations coincide very well. There are
differences in the convergence of the calculations, both methods converge well
but the linear softening not as well as the exponential softening. When looking at different variables that vary fracture energy and maximum traction, the
traction-separation law and the beam is behaving very well according to theory.
An explanation of the cohesive zone model and how it can be implemented
it is proposed. There is no new theory presented, only a summary of various
different theories developed by other authors. With the knowledge gathered
throughout this project it is not possible to apply the theory of cohesive zones
on more advanced problems. The next step in the process will be to examine
phase transformation in steel in an advancing crack.

Preface

This report is a bachelor thesis written at the Solid Mechanics Department at

LTH. When working with another topic where cohesive zones were proposed
as a method suitable for crack propagation analysis, not much thought was
given to the complexity of the cohesive zones. When it was clear that more
information was needed regarding these so called cohesive zones to get an understanding of what was happening and why, it was chosen to pursue a new
problem formulation before continuing on with the other issue.
We want to thank H
akan Hallberg, our supervisor, for the great support he has
given and for the commitment he has shown. We also want to thank Mathias
Wallin, our examiner, and the people at the Solid Mechanics Department for
helping us out when we encountered different problems and for showing such a
big interest in our work.

2.1

Individual Contributions

Working with the report both Alexander Lundberg and Sara Eliasson has tried
to distribute the work evenly among them. They have collaborated, trying to
discuss problems and make sure both have equal knowledge about the problem
formulation. Both find it important that they can answer questions that may
rise.
To make the work easier and to make sure one doesnt miss anything, Sara has
had the main responsibility for the report and Alexander the main responsibility
that the correct results are produced.

2.2

Ethnical considerations

Ethnical aspects are not relevant to the subject at hand.

Contents
1 Abstract

2 Preface
2.1 Individual Contributions . . . . . . . . . . . . . . . . . . . . . . .
2.2 Ethnical considerations . . . . . . . . . . . . . . . . . . . . . . . .

2
2
2

3 Introduction

4 Method

5 Linear Elastic Fracture Mechanics

6 Cohesive Zone Modeling of Fracture

6.1 Traction-Separation Laws . . . . . . .
6.1.1 Tvergaard and Hutchinson . .
6.1.2 Needleman . . . . . . . . . . .
6.1.3 Cornec, Scheider and Schwalbe
6.1.4 Schwalbe and Cornec . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

10
13
15
16
19
20

7 Cohesive Zone Modeling in Abaqus

23
7.1 Cohesive Zone Element . . . . . . . . . . . . . . . . . . . . . . . 23
8 Numerical Examples
27
8.1 Double Cantilever Beam Model . . . . . . . . . . . . . . . . . . . 27
8.2 Comparison of Traction-Separation Laws . . . . . . . . . . . . . . 28
8.3 Implementation Issues . . . . . . . . . . . . . . . . . . . . . . . . 28
9 Results
9.1 Choice of Mesh for the DCB . . . . . . . . .
9.2 Element Convergence for the Cohesive Zone
9.3 Results for Traction-Separation Laws . . . .
9.4 Results for Varying Fracture Energy . . . .
9.5 Results for Varying Maximum Traction . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

30
31
32
33
34
35

10 Discussion and Conclusions

37

11 Future Work

39

12 References

40

13 References - Images

42

Introduction

The information given to the reader in this report is supposed to help to get a
better understanding about what a cohesive zone is and what it is useful for.
Why do we use cohesive zones? Why is the choice of parameters so important to
make the calculation run smoothly? How do we model them in a FE-calculation?
The goal with the report is to be able to answer the questions above and to be
able to use all information gathered and implement the cohesive zone model in
other problem formulations. The work is also intended to be an aid in choosing
a suitable traction-separation formulation for a specific problem given.
The report is limited to a discussion and review on cohesive zones. No basic
Finite Element (FE) theory will be given. It is assumed that the reader already
has understanding and basic knowledge in this area. The calculations made in
ABAQUS will not be explained from a FE perspective except for the cohesive
elements.
Another limitation are the analytical results. It would provide additional value
to have a comparison with an analytical solution. But the time limit prevented
us from doing a good analytical solution worth comparing the numerical results
to. The calculations that should have been done for this require more literature
study in another area. This might be something to consider in future work.

Method

Looking into cohesive zones there is at first a lot of information to be found in

the literature. Before starting to model a cohesive zone the theoretical aspects
of cohesive zones are studied. At first understanding is built on why we use
cohesive zones and then also understanding on how they work. Learning more
and getting a better understanding is pivotal before implementing it in the FEprogram. So the first step is to find useful information about cohesive zones in
articles and books.
Modeling of the cohesive zones is done in a commercial FEM-program called
ABAQUS CAE. ABAQUS is a very powerful tool for FE-calculations. ABAQUS
gives the opportunity to use user-defined subroutines which makes it possible to
write your own calculation or specifications for the cohesive zone in, for example,
FORTRAN code. This is not used for this report, instead functionality already
available in the software was used to investigate the possibilities and limitations
of ABAQUS.
As a point of departure, the ABAQUS manual [30] was used to learn more about
how the cohesive elements was modeled in the software. Gradually, making the
models more advanced, it was possible to create a model that was suitable for
this report.
Cohesive zones are for most parts implemented in fracture analyses and used to
simulate crack growth. For this report the cohesive zones are used for studying
crack tip conditions at a mode I crack. Crack modes are discussed further in
section 5. For the crack tip it is assumed that the nonlinear effects are confined
to a very small region around the crack tip. This makes it still possible to use
linear elastic theory for the calculations of the crack [18].
A single numerical model of a double cantilever beam is used to gather simulation results in a consistent way. Model variations are also studied, for example,
in terms of varying mesh density in the cohesive zone and by using different socalled traction-separation laws to describe the cohesive behavior. These aspects
are discussed further in subsequent sections.

Linear Elastic Fracture Mechanics

There are three different types of pure crack modes, see figure 1. Mode I is
when the loading is applied normal to the crack plane and tends to open the
crack. Mode II is in-plane shear loading and tends to slide the crack faces over
each other. Mode III is out-of-plane shear, as seen in figure 1 the crack edges
are moved across each other in opposite directions [2]. Also mixed modes may
occur as combinations of any of the three basic modes.

Figure 1: Illustration of the three different crack modes, taken from [1].

Early work in the field of fracture mechanics regarding stress concentrations

around elliptical holes was developed by C.E. Inglis in 1913 [12]. His theory predicted that the stresses at a perfectly sharp crack tip approach infinity. In other
words; that the material would have zero strength [20]. Instead of analysing
the stress state in the vicinity of the crack tip, A.A. Griffith developed a theory
based on energy-balance [9].
The energy criterion is based on the principle that crack extension can only
occur when the energy available for crack growth is sufficient to overcome the
resistance of the material. This resistance includes all types of energy dissipation
related to crack extension, for example surface energy and plastic work [2].
The early work by Griffith was later developed further by G.R. Irwin[13]. Irwin
defined the energy release rate, G, as the rate of change in potential energy with
crack area for a linear elastic material. Crack extension occurs when the energy
release rate reaches a critical value, G = Gc , which is a measure of fracture
toughness.
On the atomic scale, fracture occurs when the applied stress and work are
sufficient to sever the atomic bonds. The bond between atoms is caused by the
attractive electrostatic force between opposite charges. The cohesive strength
can be derived in terms of the cohesive stress. This is shown in figure 2 and

provides the theoretical strength as

c

(5.1)

Experiments have, however, shown that the actual fracture strength is several
orders of magnitude lower than the theoretical value. This discrepancy is related
to stress concentrations around flaws in the material which magnify the stresses
locally.

Figure 2: Potential energy and force as a function of atomic separation, taken from [2].

Griffiths energy balance is based on the first law of thermodynamics which

states that a system that goes from a state of nonequilibrium to a state of
equilibrium will have a net drecrease in energy. The interpretation of the application on cracks is that a crack can only grow if the fracture process results

in constant or decreased total energy [9]. Hence, the critical condition for crack
extension can be defined as the point where crack growth occurs under equilibrium conditions, with zero net change in total energy [2]. The balance equation
under equilibrium conditions for an incremental increase in crack area, dA, can
therefore be expressed as
d dWs
dE
=
+
=0
dA
dA
dA

(5.2)

which is equivalent to

d
dWs
=
dA
dA

where E is the total energy, Ws is the work needed to create two new surfaces
and is the potential energy in the form of strain energy and work done by
external forces. For an edge crack, two new surfaces are created when a crack
is formed. Thus the expression for Ws takes the form
dWs
= 2s
dA

(5.3)

where s is the material specific surface energy.

Griffith first developed his theory while investigating cracks in glass. It has since
been shown that his equation is valid for ideally brittle solids. The theory does,
however, greatly underestimate the fracture strength of ductile materials such
as metals. A modification to Griffiths approach was later developed by Irwin
and Orowan [20]. The modified version of the energy-balance equation accounts
for plastic flow. In brittle materials cracks form simply by breaking the atomic
bonds, and the surface energy represents the total energy of the broken bonds
in a unit area. Crack propagation in ductile metals, on the other hand, include
a plastic zone in the vicinity of the crack tip. In this plastic region dislocation
motion occurs. The process of plastic flow around the crack tip contributes to
additional energy dissipation. The modified expression takes the form
W f = s + p

(5.4)

where p is the plastic work per unit area of surface created. It has been shown
that fracture in ductile metals under highly plastic conditions involve dissipation
mostly due to plastic work, rather than separation. In other words s is small
compared to p as discussed by Siegmund and Brocks, cf. [24].
Irwin later developed an energy criterion for fracture that is, in essence, equivalent to Griffiths criterion. The concept is based on the energy release rate, G,

defined as a measure of energy available for an increment of crack extension [2]

G=

d
dA

(5.5)

Crack extension occurs when the energy release rate reaches a critical value, i.e
when
G = Gc =

dWs
= 2Wf
dA

(5.6)

Important to note is that these theories are derived under the assumption that
the material response is strictly linear elastic. This means that inaccurate results
may be produced when applying the model on nonlinear problems. The criterion
for this theory to be valid is that the global behavior of the structure must be
linear elastic, while plasticity must be confined to small regions around the crack
tip [2].

Cohesive Zone Modeling of Fracture

Failure and fracture is a big part of many fields in engineering. It is therefore

important to understand and to be able to do calculations on failure processes
and fractures. To analyze these phenomena efficiently in arbitrary geometries,
a general numerical method is needed that can describe the initiation and evolution of a crack. The method should include and be able to simulate the initial
loading, the damage initiation with initial debonding, and the damage evolution
until complete separation and failure has occurred. A method used for these
kind of problems is modeling with cohesive zones. Cohesive zone models have
been proved useful in many different varieties of fracture issues in homogeneous
solids as well as in analyzing interface crack problems.
For this report a cohesive zone model will be used to analyze crack propagation.
For the analysis a numerical FE model will be implemented. The cohesive
zone is defined as an interface in between the structure faces where the crack
advance will take place. The interface consists of cohesive elements which are
set to behave in the same way as a crack would when it propagates. So for a
cohesive zone model no elements but the cohesive elements are damaged. The
crack will only propagate where cohesive elements are modeled. This requires a
pre-defined crack path.
Cohesive zone models are based on theory from Barenblatt and Dugdale [3, 8].
Their method for using cohesive zones to represent a crack propagation path
is very similar to Griffiths theory based on a surface energy that measures the
resistance against crack advance.
Barenblatt and Dugdales theory on cohesive zones is based on a method trying
to explain how a crack advances. The cohesive zone is formulated to imitate
the behavior ahead of the crack tip. Both Barenblatt and Dugdale divide the
crack surfaces into two separate regions, one part that is stress free and another
part which is loaded by cohesive stresses. Dugdale assumed that there is a
plastic zone near the crack tip. Within this plastic zone there is a stress acting
across the crack that is equal to the yield strength, . Dugdale examined the
yielding of steel sheets and investigated yielding in the sheets at the end of
slits. He obtained a relation between plastic yielding and external applied load
and found that the influence of yielding was approximately represented by a
long crack extending into the region that had a stress equal to the yield stress.
Dugdales theory holds for plane stress but the crack opening stresses can be
greater than the equivalent stress in a multiaxial stress state.
Barenblatt investigated brittle materials and instead of a constant stress in the
cohesive zone he let the stress vary with deformation in the zone ahead of the
crack tip. The stress of the cohesive zones works as a restraining stress that
keeps the separating surfaces together. It corresponds to atomic or molecular
attractions. The restraining stress, the stress in the cohesive zone, can be seen
as a function of the separation distance, = (), see figure 3.

10

Figure 3: The cohesive zone ahead of a crack tip, taken from [3].

Figure 4: To the left is Dugdales crack model and to the right is Barenblatts crack model
showing the separation of a crack connected by a cohesive interface, taken from [4].

In figure 4 it is seen in each end of a crack the difference between Dugdales and
Barenblatts theories. Looking at the right end we see Barenblatts crack model
where the cohesive stress varies with the crack tip displacement. To the left we
se Dugdales crack model where the cohesive zone has a constant stress equal
to the yield stress.
When using cohesive elements a constitutive equation describes the behavior of
the failing cohesive material elements in front of the crack tip. The constitutive
relation for a cohesive interface is such that the traction across the interface
will vary depending on the separation of the crack. With increasing separation,
the traction will reach a maximum, start to decrease and eventually be reduced
to zero. When the traction is zero a complete decohesion is allowed. A typical
cohesive stress-displacement diagram is shown in figure 5. The constitutive
behavior of the cohesive elements is described by a traction-separation law. The
parameters that set the properties of the cohesive zone is the cohesive strength
(a peak stress required for separation) and a cohesive energy (separation work

11

per unit area) [25, 27]. The traction-separation laws will be discussed in greater
detail in the next section.
Looking at figure 3, represents the contour tracing an arbitrary path surrounding the crack tip, and t is the separation distance at the crack tip. To
connect cohesive theory with Griffiths work we evaluate the J-integral for the
cohesive zone as
Z
(W dy T

J=

u
ds)
x

(6.1)

Looking at only the cohesive zone, using path independence and shrinking the
contour down to the lower and upper surface of the cohesive zone, this results
in dy = 0 on . The J-integral then becomes
Z
J =
CZ

d
() dx =
dx

d
dx

 Z

Zt

()d dx =
0

CZ

()d

(6.2)

When a cracked structure is exposed to some external loads the crack surfaces
are subjected to forces which restrain the surfaces from separating. These forces
can be seen as cohesive forces. The cohesive stress is a function of the relative
displacement between the crack surfaces, = (). The external loads will
increase until it reaches , see figure 5. When is reached the bond between
the crack faces breaks and new free surfaces are created.

Figure 5: A typical stress-displacement diagram for a cohesive element, taken from [5].

When two new free surfaces are created, the atoms can be considered to be
pulled apart. They are slowly moving out of range from their neighbours. For
12

the process where new free surfaces are created, the cohesive stresses perform
some amount of work. The work is written as
Z
W =

()d

(6.3)

According to equation (6.2) this relation is equal to the J-integral. To propagate

a crack through the distance a, a surface energy is needed which corresponds
to
ZaZ
Us =

Z
()ddx = a

()d

(6.4)

The area under the traction-separation curve is by definition twice the surface
energy. Recalling equation (5.3), s is the surface energy for one new free surface
created and for a crack we have two new free surfaces. This gives us
Z

()d = 2s + p

(6.5)

If the cohesive zone is negligible in size compared to the characteristic lengths

of the structure around the crack it can be concluded that Griffiths theory and
the theory of atomic cohesive forces are identical, cf. equation (5.3).
The information above on cohesive zone models is collected from several authors,
see references [1, 8, 15, 17, 19, 21, 29]

6.1

Traction-Separation Laws

The cohesive elements are initially made of a damage-free bulk material which
will eventually describe the damage and failure of the structure. The damage
behavior is typically described by a traction-separation law, as discussed in the
previous section, also cf. [22].
Depending on the material, one has to choose a suitable traction-separation
law which correctly describes the fracture behavior of the particular material.
The cohesive elements model the initial loading, the damage initiation and the
damage evolution. These three parts can be identified in a curve for the tractionseparation law [30].
A basic bilinear traction-separation law, frequently used in calculations, can be
seen in figure 6, where the softening after damage initiation is linear. Another

13

model frequently used is the one seen in figure 7, where the softening after
damage initiation is described by an exponential function. Crack propagation
can be simulated using different parameters that control the advance of the crack
front for cohesive zone models. It can be based on either the local energy release
or on the separation of the crack surfaces which corresponds to the displacement
of the cohesive elements [4].

Figure 6: Illustration of a simple bilinear traction-separation law, taken from [6].

Figure 7: Illustration of exponential damage evolution in a traction-separation law, taken

from [7].

The bilinear model is uniquely defined by the set of parameters that describes
the top point, (n0 ,t0n ), and the end point, (nt ,0), of the triangle. For both the
bilinear and the exponential model the maximum traction sustainable by the
cohesive element, tn , is required. This threshold value is important since it
governs when initiation of damage occurs. The penalty stiffness, K0 , is also
14

an important parameter in ensuring realisitic pre-crack conditions. K0 is the

slope of the first part of the curve seen in figure 6 and 7. If the penalty stiffness
is insufficient, large displacements will occur in the interface, which alters the
behavior of the structure.
The second part of the curve is related to damage evolution. When assumed to
be linear, it can be defined by either prescribing a failure displacement, nf , or
the critical energy release rate, 0 . When the damage evolution is assumed to be
exponential the curve can be defined by prescribing a failure displacement, nf ,
and a exponential law parameter to define the flexion of the curve. It can also
be based on the energy, 0 . When choosing to prescribe failure displacement,
the critical energy release rate can be calculated as the area under the curve,
and vice versa.
Z0
0 =

T ()d

(6.6)

When modeling fracture in ductile materials, such as steel, the deformation

process involves the processes of elasticity, plasticity and damage. This means
that the shape of the traction-separation law cannot easily be determined experimentally, and would have to be assumed [22]. The cohesive model is a
phenomenological model. The cohesive model does not model the real physical
fracture process and there is no evidence for which traction-separation law is the
right one to use. In the literature there are several different approaches found.
Since there are many different assumptions of the traction-separation law a few
will be described with reference to the authors. They all build on the same
fundamental ideas, but include a few differences.
6.1.1

Tvergaard and Hutchinson

Tvergaard and Hutchinson developed a model for an idealized traction-separation

law specified on the crack plane to characterize the fracture process in 1992 [27].
The relation can be seen in figure 8. The relation is built on an increase of traction in the cohesive element until it reaches the peak traction,
, at 1 . The
traction will stay the same for further separation until 2 is reached. Then
damage evolution is modeled until reaching c where fracture occurs.
The energy disspated by the cohesive elements at failure is defined by the area
underneath the curve. For this specific curve, the expression for the work of
separation per unit area in equation (6.6) takes the following form
Z1
0 =

d =

[c + 2 1 ]
2

15

(6.7)

To fully specify Tvergaard and Hutchinsons separation law the parameters

needed to describe the fracture process are the work of separation per unit
area, 0 , and the peak traction,
. Shape parameters are 1c and 2c [27]. We
also need a set of parameters to describe the continuum behavior of the solid.
The elastic-plastic solid is characterized by the following expression
(
/E
=
1
(y /E)(/y ) N

y
y

(6.8)

The solid is specified by its Youngs modulus, E, the Poissons ratio, , the
initial tensile yield stress, y , and the strain hardening exponent, N .

Figure 8: Graphical illustration of Tvergaard and Hutchinsons traction-separation law, taken

from [8].

Tvergaard and Hutchinsons suggestion is shown mostly to be suitable for ductile

fracture.
6.1.2

Needleman

A cohesive zone model which takes finite geometry changes into account proposed by Needleman [16], is used to provide a description of the behavior of void
nucleation. The model is aimed to describe the evolution from initial debonding
through to complete decohesion which refers to subsequent void growth in a material. The decohesion occurring can occur either in a brittle or ductile manner.
This is dependent on the ratio of a characteristic length to the inclusion radius.
The characteristic length is introduced by the dimensions of the surrounding
material e.g. dimensions around a crack tip.
16

The characteristic length that is introduced comes from the fact that the model
is to take finite geometry changes into account. The mechanical response of
the interface is based on both a critical interfacial strength and the work of
separation per unit area. Based on this, the characteristic length is introduced.
The interfacial traction is only dependent on the displacement difference across
the interface. Consider two points A and B, placed on opposite sides of the
interface. The interfacial traction is then seen as dependent on the displacement
difference, uAB , between the points. For each point of the interface Needleman
defined
un = n uAB ,

ut = t uAB ,

ub = b uAB

(6.9)

and
Tn = n T,

Tt = t T,

Tb = b T

(6.10)

For the equations (6.9) and (6.10), n, t, b form a right-handed coordinate system
where un represents the interfacial separation. A positive un corresponds to
increasing separation and a negative un corresponds to decreasing separation.
The constitutive relation for the mechanical response of the interface gives a
dependence of the tractions (Tn , Tt , Tb ) on the separations (un , ut , ub ). The
response can be specified as a potential (un , ut , ub ), being defined as
Zu
(un , ut , ub ) =

[Tn dun + Tt dut + Tb dub ]

(6.11)

Needlemans model resembles the other traction-separation laws in the way that
with increasing separation of the interfaces, the traction reaches a maximum
value, then start to decrease to eventually vanish. When the interfacial traction
has vanished complete decohesion occurs. The specific potential function that
is used is the following

(un , ut , ub ) =
(   
 
 2 
2
1 un
4 un
1 un
27
max
1
+
4
2
3
2
 2 
   2 
ut
un
un
1
12
+
+
2

 2 
   2 )
1
ub
un
un
+
12
+
2

17

(6.12)

The interface is undergoing a purely normal separation, ut ub 0, is the

characteristic length and specifies the ratio of shear to normal stiffness of
the interface. By differentiating equation (6.12) we can obtain the tractions,

.
T = u
( 
   2 
27
un
un
un
Tn = max
12
+
4

 2  

 2  
)
ut
un
ub
un
+
1 +
1

(6.13)

(  
   2 )
ub
un
un
27
12
+
Tt = max
4

(6.14)

(  
   2 )
27
ut
un
un
Tb = max
12
+
4

(6.15)

This is the case for when un and for un > we get Tn Tt Tb 0.

The response of the potential of equation (6.12) is shown in figure 9. Here
the traction Tn is plotted as a function of un with ut ub 0. The work of
separation is the area under the curve.

Figure 9: Normal traction across the cohesive interface as a function of the interface separation un with ut ub 0, taken from [9].

Needlemans interface description is a phenomenological one and is characterized

by the three parameters; max , and .
18

6.1.3

Cornec, Scheider and Schwalbe

Cornec, Scheider and Schwalbe proposed a traction-separation law, adressing

metallic materials. The method is to simulate a ductile tearing process, which
consists of initiation, growth and coalescence of voids. The method used for the
practical application of the cohesive model consists of a defined shape of the
traction-separation law and determination of the parameters for this law [6].

Figure 10: Graphical illustration of Cornec, Scheider and Schwalbes traction-separation

law, taken from [10].

The suggested shape function for the traction-separation law can be seen in
figure 10. The three sections visible are expressed by following analytical expression

    2 

2 1 1
0

T = T0

 
3

2


2
2

3
+
1
T0 2
0 2
0 2

for < 1
for 1 2

(6.16)

for 2 0

With 1 = 0.010 and 2 = 0.750 . The traction-separation law will be determined from three material parameters, comprising the cohesive stress, T0 , the
cohesive energy, 0 and the separation at material decohesion, 0 . There are
only two independent parameters since the cohesive energy can be determined
from the area underneath the traction-separation curve as
Zc
0 =


T d = T0

19

1 1 1
1 2

+
2 3 0
2 0


(6.17)

Which in this particular case can be simplified to

0 = 0.87T0 0

(6.18)

According to Cornec, Scheider and Schwalbe a very simple traction-separation

law would be chosen just as
0 = T0 0

(6.19)

But the slopes are chosen to avoid sharp edges. The slope in the beginning helps
out with numerical problems between the cohesive elements and the surrounding
continuum elements. The slope in after 2 down to 0 models the rapid softening
during void growth and coalescence.
6.1.4

Schwalbe and Cornec

Schwalbe and Cornec [23] suggested a crack growth resistance curve, at fracture,
for a material constitutive model that is defined by the parameters 0 , E, and
n, where n is a strain hardening exponent. They also concluded that the total
work to fracture can be broken down into three parts; work to separation, S ,
plastic energy, P and elastic energy, E . This results in a total energy on the
form
J = S + P + E

(6.20)

The work of separation and the plastic energy can be allocated to the process
zone and the plastic zone, see figure 11. Geometry effects are caused by the
plastic energy whilst the separation work is more of a material specific variable.
In the model there are two constants representing the porosity in the material;
f0 representing the initial porosity and fc which is the final porosity at fracture. These quantities are related to the materials microstructure. Since the
microstructure of the material is a fundamental part of how the crack will advance these parameters are important and is a fundamental basis of the fracture
model. Since Schwalbe and Cornec have been looking at a porous model, this
is a brittle fracture process determined by void growth and coalescence.

20

Figure 11: The process zone and the plastic zone allocating the work to fracture and the
plastic energy, respectively, taken from [11].

Figure 12: The traction-separation law of the cohesive zone model by Schwalbe and Cornec,
taken from [12].

The cohesive zone model is based on the fact that the cohesive energy, 0 , is constant during crack growth. The cohesive energy is presented as a material property. As in previous traction-separation laws, the energy is based on the area
underneath the curve of the cohesive force law. To simplify the cohesive stress,
T0 is taken as a constant. Figure 12 illustrates how the traction-separation law
of the cohesive zone model is defined.

21

The total work equals the cohesive energy and can be expressed as
Z
Ji =

T (S )dS
= T0 S,f racture = 0

(6.21)

Schwalbe and Cornec used this model to predict crack growth resistance curves
for various geometries and load versus displacement relationships for structural
parts.

22

Cohesive Zone Modeling in Abaqus

Different cohesive zone models can be defined in ABAQUS CAE, which is the
pre- and post processor of ABAQUS. In the present study, the cohesive zone
model is implemented in a double cantilever beam (DCB). A DCB is a good
model to be able to investigate the basic behavior of the cohesive elements
and find a traction-separation law that is suitable for predicting crack growth.
DCB structures in combination with cohesive zones have been employed in, for
example, [1] and [5]. For the analysis a plane stress condition is implemented
for conditions of small-scale yielding where LEFM applies. Plane stress is used
since we are looking at a DCB which is small in the thickness direction compared
to the in-plane dimensions.
Arbitrary traction-separation laws are not easily implemented through the CAE
interface. For more general formulations one needs to resort to writing userdefined subroutines.
Since the concept of user-defined subroutine will not be pursued in this report
the results will be those that are possible to create by just using ABAQUSs
own interface. ABAQUS provides the possibility to define the initial tractionseparation stiffness, combined with linear or exponential softening during evolution of damage, cf. figure 6 and 7. The damage evolution can be defined by
displacement or energy.

7.1

Cohesive Zone Element

The cohesive elements are modeled as an own separate part of the model. The
cohesive layer is one element layer thick in between the regions where the crack
is to advance. For two-dimensional modeling, the four-node cohesive element
COH2D4 is used. The cohesive element has a linear displacement formulation
and the stress in the third direction does not affect the element behavior and
thus there is no difference between plane stress and plane strain condition.
The separations and stresses in the cohesive elements are calculated in each
increment at the integration points according to the traction-separation law.
When the critical separation energy is reached, the element has failed. The
integration point which contributes to the stiffness obtains the status failed.
Once a integration point has lost its stiffness, it will never obtain another status
[6].
To understand the calculations done by ABAQUS it is interesting to know how
the stiffness matrix of a cohesive element is formulated [21]. It can be derived
from the principle of virtual work, which for the cohesive stresses is defined by
Z
i =

t [u]dA

23

(7.1)

Where the separation is denoted as [u] and t is the vector of the cohesive stresses
related to the separations.
The coordinates and separations are written, using a FE approximation. The
displacements [u] are replaced by
u = Vu ue

(7.2)

Where Vu is a matrix with the shapefunctions and ue is the displacement for

each node. Vu is defined as

Vij =

f1
0

0
f1

f2
0

0
f2

... fn
... 0

0
fn


(7.3)

Using equation (7.2) and inserting it into equation (7.1) leads to

Z
i = [ue ]

VTu tdA

(7.4)

The cohesive stresses t in equation (7.1) are nonlinear functions of the displacement. This makes it impossible to extract the nodal displacement from the
integral, and a linear system of equations can not be setup. This means we
will need to find a tangential stiffness matrix. The tangential stiffness matrix
is the change of the internal forces corresponding to infinitesimal changes in
displacements [14].
To find the tangential stiffness matrix, the integral part of equation (7.4) is
differentiated. The integral in this equation can be seen as the internal forces.
a = [ue ]f

(7.5)

Where
Z
f=

VTu tdA

(7.6)

When differentiating the internal forces the total differential of the traction is
used as
dt =

t
du
[u]

(7.7)

For du the difference u is used. Looking at the total displacement at the

time t + t we get ut+t = ut + u. The next step is to write the differential
formulation of the integral in equation (7.4) so that we can find the tangential
24

stiffness matrix K. Looking at the differential of only the integral it can be

written as
Z

VTu dtdA =

VTu

t
Vu dA [ue ] = K [ue ]
[u]

(7.8)

The integration is evaluated numerically using the Gauss integration algorithm,

whereby the tangential stiffness matrix can be written as
Z
K=

VTu


X

t
t
T
Vu dA =
Vu
Vu  i Ji

[u]
[u]
i
i

(7.9)

The weight function for the Gauss integration is i , Ji is the Jacobian determinant of the cohesive element at the specific Gauss point. The Jacobian can be
calcuated from the matrix

J=

!
(7.10)

ABAQUS uses the definition for the tangent stiffness matrix shown in equation
(7.9) in the FE-calculations for the cohesive zone. In ABAQUS the cohesive
zone is modeled between the two potential crack faces and tied to the upper
and lower face by tie-constraints, c.f figure 13 [7].

Figure 13: Illustration of the cohesive elements and how they are modeled, taken from [13].

The softening part of the cohesive law often give rise to convergence issues when
running the calculations. The analysis tends to exhibit mesh sensitivity, lack of
25

convergence, computing inefficiency, sensitivity to element aspect ratio and so

on. These difficulties need to be adressed when using cohesive zones and it is
also very important to pick the right traction-separation law with the correct
parameters [28].

26

Numerical Examples

Having summarized the basic theory of a cohesive zone model and tractionseparation laws attention is now turned to implementation in ABAQUS. For
the results a linear damage evolution and an exponential damage evolution will
be compared since these are easily accessible in ABAQUS CAE without using a
user-defined subroutine to describe the material behavior of the cohesive zone.

8.1

Double Cantilever Beam Model

The cohesive zone model is implemented on a double cantilever beam, see figure
14. The material properties for the beam can be seen in table 8.1. The material
is linear-elastic.

Figure 14: The abaqus model of the DCB with dimensions.

Table 8.1: Material parameters for the DCB, representative for steel.

Youngs modulus, E [N/mm2 ]

Poissions ratio

205 000

0.3

For the DCB analysis displacement controlled loading is used. The top and
bottom left nodes of the DCB are the nodes which are subjected to a vertical
displacement of 8 mm in opposite directions.
Different mesh densities and element sizes are tested to find a mesh which provides an appropriate result for the beam calculation. If the mesh is too coarse
true behavior will not be captured and a too fine mesh consumes unnecessary
time on calculations for results that are provided also with a coarser mesh. To
identify a suitable mesh disctretization, only one half of the DCB geometry is
used and the simulation results are checked against the analytical expression,
found from elementary beam theory, c.f [26], as

P l3
3EI

27

(8.1)

8.2

Comparison of Traction-Separation Laws

For the traction-separations laws, that are used in the cohesive zone model, to
be compared in a consistent way some parameters have to be held constant. The
constant parameters are the penalty stiffness, K0 , and the maximum traction,
tn . Further, it is chosen for the traction-separation laws to depend on the critical
energy release rate, 0 , also set as a constant. In table 8.2 the values for the
parameters can be seen.
Table 8.2: Parameters for the traction-separation law.

K0
tn
0

10 000
100
2

[N/mm3 ]
[N/mm2 ]
[J/mm2 ]

When comparing the traction-separation laws for the cohesive zone a forcedisplacement curve will be analysed. The curve will show how the beam responds when the crack advances. By analysing the curve it is then possible
to see if the choice of traction-separation law will make the crack propagation
behave differently.
The mesh for the DCB is tested to find one that is possible to use for all analyses
of the cohesive zone. However, the element size in the cohesive zone is also varied
to investigate how the results differ with different mesh densities.
The maximum traction chosen will be used for all analyses except one where
different maximum tractions are investigated to see how the results are affected.

8.3

Implementation Issues

There are two different ways to implement the cohesive behavior in ABAQUS.
The first includes modeling the cohesive zone as a surface interaction. The other
method, which is used in this paper, is based on implementing actual cohesive
elements between the two surfaces along the crack path. The surfaces are then
connected to the cohesive elements using tie constraints. These constraints tie
the nodes on the slave surface, in this case the surface of the cohesive zone, to
the elements of the master surface, i.e. the adjacent solid material.
When modelling the tie constraints there are different options for the discretization method. One can choose from either surface to surface or node to surface
discretization. For this application it is important to choose the latter to ensure
that all the nodes on the cohesive zone are properly connected to the ligament.
The characteristics of the traction-separation law have a great influence on the
convergence. If the penalty stiffness is too high, sharp increases in stiffness
will occur in elements which are suddenly loaded. Another parameter which
has great influence on the convergence is the fracture energy. The fracture

28

energy determines, in the case of the basic triangular traction-separation law,

the decrease in stiffness. As with most non-linearities, numerical issues can
arise when the material softens rapidly. Lowering of the fracture energy greatly
reduces the stiffness once damage is initiated and can cause convergence issues
due to sharp decreases in stiffness.
Trying to implement all the traction-separation laws described in this report
was very difficult. The possibility of completely defining your own tractionseparation law is not accessible unless using a user-defined subroutine to define
the cohesive behavior. In ABAQUS CAE it was possible to define the linear
and exponential damage evolution and compare these two. For other tractionseparation laws as for example Tvergaard and Hutchinsons, it is recommended
to use a user subroutine to define the cohesive zone behavior.

29

Results

In this section the results from ABAQUS are presented. Running the DCB
calculation with different traction-separation laws for the cohesive zone, a deformation of the beam similiar to the on seen in figure 15 should be obtained.
The cohesive elements in figure 15 are seen as just black, this is because the elements are too small to distinguish. The magnified cut-out gives a closer view,
see figure 16.

Figure 15: The deformation of the DCB after displacement controlled loading. Seen is also
the deformed cohesive elements modeled inbetween the top and bottom cantilever beams.

The force-deflection curves are represented by values from the top node in the
DCB which was subjected to a prescribed displacement. This node represents
the behavior of the beam structure. The graphical illustrations of the tractionseparation laws are constructed of values from a node in a cohesive element
which is known to have collapsed.
Running the simulation in ABAQUS, a value for the damage variable can be
retrieved, varying between zero and one. This variable is a good way to see
when and if a cohesive element has failed. In figure 17 the red elements indicate
a failed element with a damage variable equal to one.

30

Figure 16: A magnified cut out of the DCB when deformed to get a better view of the
cohesive elements. The cohesive elements are represented by the blue elements inbetween the
beam parts.

Figure 17: The damage of the cohesive elements with red indicating failure.

9.1

Choice of Mesh for the DCB

The DCB will have an unchanged mesh throughout the analyses. But the choice
of the mesh is determined by analysing the convergence of the result compared
to the element density. Figure 18 is showing the results of testing different
element densities. The amount of elements are chosen to be 100, 400, 800 and
2400. The results obtained with different meshes are compared to the analytical
31

solution provided by equation (8.1).

Figure 18: The element convergence for the DCB showing results provided by testing 100,
400, 800 and 2400 elements.

The final choice of mesh for the DCB came to 800 elements, 8x100.

9.2

Element Convergence for the Cohesive Zone

Looking at the influence of the mesh density of the cohesive zone on the results, three different densities were chosen. The elements for the cohesive zone
is always one layer thick and was varied between 100, 300 and 500 cohesive
elements. It appears as if the results are quite insensitive to the number of
cohesive elements that are used.

32

Figure 19: The element convergence for the cohesive zone showing results provided by testing
100, 300 and 500 cohesive elements.

9.3

Results for Traction-Separation Laws

Two traction-separation laws are analysed. One which is bilinear and one that
is linear until damage initiation and then has a exponential damage evolution.
The results are seen in figure 20, illustrated by a force-deflection curve. The
traction-displacement curve for the models are shown in figure 21.

Figure 20: Showing the force-displacement curve for the linear and exponential tractionseparation law.

33

Figure 21: Showing traction-separation laws for the exponential and linear model.

Looking at figure 20, it appears as both traction-separation models provide very

similar results, based on the current model set-up.

9.4

Results for Varying Fracture Energy

Seen in figures 22 and 23 are the force-deflection curve and the traction-separation
curves from the simulation where the maximum traction is kept constant and
the fracture energy varies. The fracture energy varies with 2, 3 and 4 J/mm2

Figure 22: Force-deflection curves for the response of the DCB when varying the fracture
energy.

34

Figure 23: The traction-separation curves for a linear model when varying the fracture
energy.

9.5

Results for Varying Maximum Traction

Crack propagation in the DCB model is also studied when varying the maximum traction but keeping the energy release rate constant. The resulting
force-deflection curve can be seen in figure 24. The maximum traction is set as
100, 80, 50 and 10 MPa, respectively.

Figure 24: Showing the force-displacement curve for varying maximum traction. The maximum traction is varied between 100, 80, 50 and 10 MPa.

35

Figure 25: Showing the force-displacement curve for varying maximum traction. The maximum traction is varied between 100, 80, 50 and 10 MPa.

36

10

Discussion and Conclusions

The double cantilever beam is chosen to have a mesh of 8 elements across the
height and 100 elements along the width. This choice comes from the fact
that the amount of elements slowly converges to the result obtained with the
elementary beam theory. Looking at figure 18, it is clearly seen that 100 elements
is a too coarse a mesh. The result from ABAQUS with this mesh shows a huge
deviation from the analytical result. Also for 400 elements the difference in the
result is significant. But comparing 800 elements to the analytical result we are
close enough to the correct solution to have an acceptable error. Since there
is very little difference in a solution with 2400 elements and the solution with
800 elements, it is concluded that 800 elements gives a satisfying result without
degrading the accuracy of the calculations.
The mesh convergence study for the cohesive zone shows that no significant
difference in the results are obtained when using different mesh densities, see
figure 19. The curves for the different counts of cohesive elements coincide.
This shows that the number of cohesive elements will not significantly affect the
results. However, the general experience from working with cohesive elements
during this project is that choosing too few elements can result in instability
and convergence difficulties. Using a very fine mesh in the cohesive zone can, on
the other hand, lead to very long computation times. The mesh density must
therefore be chosen carefully with these two points in mind.
Using an exponential softening model the calculations was perceived to converge
faster. The linear softening has a good convergence as well but not as good. The
behavior of the two different models does not show differences in the resulting
behavior of the simulated beam. The crack extends in the same way, which is
controlled by the fracture energy, and initiates at the same time, due to the
choice of penalty stiffness and maximum traction. Most inequalities are seen
after damage initiation which is when the damage evolution behaves according
to either of the two different models. The results, however, follow the same
slope and converge to about the same value.
The appearance of the traction-separation laws are seen in figure 21. There is
an obvious difference between the shapes of the linear and exponential softening
models. The first part of the curve, which is controlled by the penalty stiffness
and the maximum traction, coincides.
The penalty stiffness is mostly set to a very high value since this choice resembles
reality. If the penalty stiffness is too low, the displacement of a cohesive element
will be quite large before any damage occurs. In reality a crack does not open
unless the crack is advancing and damage is evolving. Since a lower penalty
stiffness also greatly increases the energy absorbed by the elastic loading part
of the curve it reduces the amount of energy available for the damage part of
the curve for a fix fracture energy. This leads to very steep softening which,
in turn, often lead to convergence difficulties. It can be concluded that a high

37

penalty stiffness is needed to ensure realistic pre-crack extension behavior.

The fracture energy, i.e. the area underneath the two curves in figure 21, is the
same for both models. Increasing the fracture energy entails that the failure
displacement is increased, since the area under the curve must increase. The
damage initiation will still occur in the same place and at the same time, but
the damage evolution will be a prolonged process. For the force-deflection curve
in figure 20, the lowering of the stiffness in the structure is caused by damage. A
higher fracture energy essentially means that more energy is needed at the crack
tip to create new surfaces. Linear elastic fracture mechanics predicts that the
fracture energy is proportional to the stress intensity factor squared. Since the
stress intensity factor is related to the stress, it is intuitive that a higher load is
needed to advance the crack. This corresponds well with the results presented
in figure 22.
Looking at results from when the maximum traction is varied there is a more
significant difference in the results from the beam analysis. The maximum
traction varied with a constant energy implies that the failure displacement
must be different. This is seen in figure 25. Since the same amount of energy
is needed for the crack to advance, the crack propagation will behave in the
same way. As seen in figure 24, the curves eventually coincides for all choices
of maximum traction. The main thing the maximum traction affects is the precrack extension behavior, equal to the softening during the damage evolution.

38

11

Future Work

Writing this report gave us a perspective of how useful cohesive zone models
can be. Even though it is not yet big in the area of crack propagation in steel it
has huge potential. Most applications today are for brittle materials and very
ductile materials, like polymers. The theory on the subject is endless and there
are still much to look into even further. There are a lot of traction-separation
models for the cohesive zone proposed by different authors and still more to
come. The work on cohesive zones is a long way from being finished.
The review on the cohesive zone theory in this report can be used in many
applications. Being able to use cohesive zones in FE-programs such as ABAQUS
and understanding the approach makes it possible to apply the knowledge on
various FE-problems involving crack growth.
A proposal on further work with crack advance is to consider Hallbergs article
from 2007 [11]. In this article, a constitutive model for martensite transformation in austenitic stainless steel is derived. In a subsequent article, also by
Hallberg, from 2011 [10], results are shown from a stationary crack where the
martensite transformation at the crack tip is included. Using this constitutive model with a cohesive zone model on an advancing crack would make it
possible to investigate how the martensite transformation influences the crack
propagation.
Using a cohesive zone model for this problem one has to consider that the
fracture toughness is different for the martensitic and austenitic phases. This
means that the material response in the vicinity of the crack tip will change and
the traction separation law will have to account for these changes. An example
of possible alterations that could be made to the cohesive zone model to account
for the phase transformation could be by varying fracture energy and maximum
traction to account for changes in fracture toughness and yield stress.
A good idea to look into is also the comparison of the results in this report
against further analytical results.
All traction-separation laws decribed in this paper were not implemented in
ABAQUS. The models presented by Tvergaard and Hutchinson and Scheider
and Cornec proved to be difficult to implement using ABAQUS CAE. In order
to implement these models user-defined subroutines are called for. This was not
done in this project due to lack of time, but it would be of great interest to do
this and compare the results with the ones presented in this paper.

39

12

References

[1] G. Alfano, M. A. Crisfield, Finite Element Interface Models for Delaminaion Analysis of Laminated Composites: Mechanical and Computational
Issues. 2001
[2] T.L. Andersson, Fracture Mechanics - Fundamentals and Applications.
Department of Mechanical Engineering, Texas A&M University. CRC
Press, 1991.
[3] G.I. Barenblatt, The Mathematical Theory of Equilibrium Cracks in Brittle Fracture. Advances in Applied Mechanics vol. 7, pp 55-129, 1962
[4] J. Chen M. Crisfield, Predicting Progressive Delamination of Composite
Material Specimens via Interface Elements. Mechanics of Composite Materials and Structures 6:301-317, 1999
[5] M.A. Crisfield, G.A.O. Davies, H.B. Hellweg, Y. Mi, Progressive Delamination Using Interface Elements. Journal of Composite Materials, 1998
[6] A. Cornec, I. Scheider, K-H. Schwalbe, On the Practical Application of the
Cohesive Model. Institute for Materials Research, GKSS Research Centre
Geesthacht, 2003
[7] T. Diehl, Modeling Surface-Bonded Structures with ABAQUS Cohesive
Elements: Beam-Type Solutions.
[8] D.S. Dugdale, Yielding of Steel Sheets Containing Slits, Engineering Department, Univeristy College ov Swansea, 1959
[9] A.A.Griffith, The Phenomena of Rupture and Flow in Solids. 1920
[10] H. Hallberg, L. Bank-Sills, M. Ristinmaa, Crack Tip Transformation
Zones in Austenitic Stainless Steel. Division of Solid Mechanics, Lund
University, 2011
[11] H. Hallberg, P. H
akansson, M. Ristinmaa, A Consitutive Model for the
Formation of Martensite in Austenitic Steels under Large Strain Plasticity.
Division of Solid Mechanics, Lund University, 2007
[12] C.E Inglis, Stresses In a Plate Due to the Presence of Cracks and Sharp
Corners. Cambridge, 1913
[13] G.R Irwin, Analysis of Stresses and Strains Near the End of a Crack
Traversing a Plate. Journal of Applied Mechanics 24, 361364, 1957
[14] S. Krenk, Non-Linear Modeling and Analysis of Solids and Structures.
Cambridge University Press, 2009
[15] N. Moes, T. Belytschko, Extended Finite Element Method for Cohesive
Crack Growth. Department of Mechanical Engineering, Northwestern University, USA, 2001

40

[16] A. Needleman, A Continuum Model for Void Nucleation by Inclusion

Debonding. Division of Engineering, Brown University, 1987
[17] A. Needleman, An Analysis of Decohesion Along an Imperfect Interface 1988. International Journal of Fracture vol.42 pp.21-40, 1990
[18] F. Nilsson, Fracture Mechanics - From theory to applications. Royal Institute of Technology, 2001
[19] J.R. Rice, A Path Independent Integral and the Approximate Analysis of
Strain Concentration by Notches and Cracks. Journal of Applied Mechanics Vol. 35, pp. 379-386, 1968
[20] D. Roylance, Introduction to Fracture Mechanics. 2001
[21] I. Scheider, Cohesive Model for Crack Propagation Analyses of Structures with Elastic-Plastic Material Behavior. GKSS Research Center
Geesthacht, 2001
[22] I. Scheider, W. Brocks, The Effect of the Traction Separation Law on the
Results of Cohesive Zone Crack Propagation Analyses, 2003
[23] K-H. Schwalbe, A. Cornec, Modelling Crack Growth Using Local Process
Zones. GKSS Research Centre Geesthacht, Germany, 1994
[24] T. Siegmund, W. Brocks A Numerical Study on the Correlation Between
the Work of Separation and the Dissipation Rate in Ductile Fracture. 2000
[25] T. Siegmund, W. Brocks, The Role of Cohesive Strength and Separation
Energy for Modeling of Ductile Fracture. Fatigue and Fracture Mechanics:
30th Volume, 2000, pp. 139-151.
[26] B. Sundstr
om and other writers, Handbok och Formelsamling i
H
allfasthetsl
ara. Institutionen for h
allfasthetslara KTH, 1998
[27] V. Tvergaard, J.W. Hutchinson, The Relation Between Crack Growth Resistance and Fracture Process Parameters in Elastic-Plastic Solids. 1992
[28] D. Xie, A.M. Waas, Discrete Cohesive Zone Model for Mixed-Mode Fracture Using Finite Element Analysis. Department of Aerospace Engineering, The University of Michigan, 2006
[29] A. Yavari, Generalization of Barenblatts Cohesive Fracture Theory for
Fractal Cracks. California Institute of Technology, 2001
[30] Abaqus Analysis, Users Manual: Volume IV. Dassault Syst`emes, 2011

41

13

References - Images

[1] Date that collected: 2014-12-09

http://upload.wikimedia.org/wikipedia/commons/thumb/e/e7/
Fracture_modes_v2.svg/2000px-Fracture_modes_v2.svg.png
[2] Image from figure 2.1 in
T.L. Andersson, Fracture Mechanics - Fundamentals and Applications.
Department of Mechanical Engineering, Texas A&M University. CRC
Press, 1991.
[3] Image from figure 7a in
J.R. Rice, A Path Independent Integral and the Approximate Analysis of
Strain Concentration by Notches and Cracks. Journal of Applied Mechanics Vol. 35, pp. 379-386, 1968
[4] Image from figure 1.1 in
I. Scheider, Cohesive Model for Crack Propagation Analyses of Structures with Elastic-Plastic Material Behavior. GKSS Research Center
Geesthacht, 2001
[5] Image from figure 3b in
A. Yavari, Generalization of Barenblatts Cohesive Fracture Theory for
Fractal Cracks. California Institute of Technology, 2001
[6] Image from figure 31.5.6-1 in
Abaqus Analysis, Users Manual: Volume IV. Dassault Syst`emes, 2011
[7] Image from figure 31.5.6-5 in
Abaqus Analysis, Users Manual: Volume IV. Dassault Syst`emes, 2011
[8] Image from figure 1 in
V. Tvergaard, J.W. Hutchinson, The Relation Between Crack Growth Resistance and Fracture Process Parameters in Elastic-Plastic Solids. 1992
[9] Image from figure 1 in
A. Needleman, A Continuum Model for Void Nucleation by Inclusion
Debonding. Division of Engineering, Brown University, 1987
[10] Image from figure 2 in
A. Cornec, I. Scheider, K-H. Schwalbe, On the Practical Application of the
Cohesive Model. Institute for Materials REsearch, GKSS Research Centre
Geesthacht, 2003
[11] Image from figure 2 in
K-H. Schwalbe, A. Cornec, Modelling Crack Growth Using Local Process
Zones. GKSS Research Centre Geesthacht, Germany, 1994
[12] Image from figure 3b in
K-H. Schwalbe, A. Cornec, Modelling Crack Growth Using Local Process
Zones. GKSS Research Centre Geesthacht, Germany, 1994
42

[13] Image from figure 31.5.3-2 in

Abaqus Analysis, Users Manual: Volume IV. Dassault Syst`emes, 2011

43