Professional Documents
Culture Documents
Concept version
1 Introduction 1
2 Fracture mechanics 9
2.1 Fracture mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Shearing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Cleavage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.3 Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.4 Crazing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.5 De-adhesion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Ductile - brittle behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.1 Charpy v-notch test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Theoretical strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Discrepancy with experimental observations . . . . . . . . . . . . . . . 16
2.3.2 Griffiths experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.3 Crack loading modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Experimental techniques 19
3.1 Surface cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Electrical resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 X-ray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4 Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5 Acoustic emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.6 Adhesion tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Fracture energy 23
4.1 Energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Griffiths energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Griffith stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3.1 Discrepancy with experimental observations . . . . . . . . . . . . . . . 26
4.4 Compliance change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4.1 Fixed grips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.2 Constant load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.3 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
I
II
5 Stress concentrations 31
5.1 Deformation and strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.3 Linear elastic material behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.4 Equilibrium equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.5 Plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.6 Plane strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.7 Displacement method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.8 Stress function method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.9 Circular hole in infinite plate . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.10 Elliptical hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
12 Fatigue 119
12.1 Crack surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
12.2 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
12.3 Fatigue load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
12.3.1 Fatigue limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
12.3.2 (S-N)-curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
12.3.3 Influence of average stress . . . . . . . . . . . . . . . . . . . . . . . . . 122
12.3.4 (P-S-N)-curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
12.3.5 High/low cycle fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
12.3.6 Basquin relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
12.3.7 Manson-Coffin relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
12.3.8 Total strain-life curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
12.4 Influence factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
12.4.1 Load spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
12.4.2 Stress concentrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
12.4.3 Stress gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.4.4 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.4.5 Surface quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
12.4.6 Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
12.5 Crack growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
12.5.1 Crack growth models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
12.5.2 Paris law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
12.5.3 Fatigue life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
12.5.4 Other crack grow laws . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
12.5.5 Crack growth at low cycle fatigue . . . . . . . . . . . . . . . . . . . . . 134
12.6 Load spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
12.6.1 Random load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
12.6.2 Tensile overload . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
12.7 Design against fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
A Laplace equation a1
Introduction
Important aspects of technological and biological structures are stiffness and strength. Re-
quirements on stiffness, being the resistance against reversible deformation, may vary over a
wide range. Strength, the resistance against irreversible deformation, is always required to
be high, because this deformation may lead to loss of functionality and even global failure.
Continuum mechanics
When material properties and associated mechanical variables can be assumed to be con-
tinuous functions of spatial coordinates, analysis of mechanical behavior can be done with
Continuum Mechanics. This may also apply to permanent deformation, although this is as-
sociated with structural changes, e.g. phase transformation, dislocation movement, molecular
slip and breaking of atomic bonds. The only requirement is that the material behavior is
studied on a scale, large enough to allow small scale discontinuities to be averaged out.
1
2
V0
A0
A
V
~u
~x0 ~e3 ~x
~e2
O
~e1
- volume / area V0 , V / A0 , A
- base vectors {~e1 , ~e2 , ~e3 }
- position vector ~x0 , ~x
- displacement vector ~u
- strains kl = 21 (uk,l + ul,k )
- compatibility relations
Material behavior
To investigate, model and characterize the material behaviour, it is necessary to do experi-
ments. The most simple experiment is the tensile/compression loading of a tensile bar.
3
Strain-time and stress-time curves reveal much information about the material behavior,
which can be concluded to be elastic : reversible, time-independent; visco-elastic : reversible,
time-dependent; elasto-plastic : irreversible, time-independent; visco-plastic : irreversible,
time-dependent; damage : irreversible, decreasing properties.
t1 t2 t t1 t2 t t1 t2 t t1 t2 t
e
p
t1 t2 t t1 t2 t t1 t2 t t1 t2 t
Stress-strain curves may show linear, nonlinear, hardening and softening behavior.
Fig. 1.5 : Stress-strain curves for hardening (left) and sofening (right) material behavior.
Fracture
When material damage like micro-cracks and voids grow in size and become localized, the
averaging procedure can no longer be applied and discontinuities must be taken into account.
This localization results in a macroscopic crack, which may grow very fast, resulting in global
failure.
4
Although early approaches have striven to predict fracture by analyzing the behavior of
atomic bonds, Griffith has shown in 1921 that attention should be given to the behavior of
an existing crack.
Fracture mechanics
What is the residual strength of a construction (part) as a function of the (initial) crack
length and the load ?
Several fields of science are involved in answering these questions : material science and
chemistry, theoretical and numerical mathematics, experimental and theoretical mechanics.
As a result, the field of fracture mechanics can be subdivided in several specializations, each
with its own concepts, theory and terminology.
5
resulting crack growth criterion is referred to as being global, because a rather large volume
of material is considered.
The crack growth criterion can also be based on the stress state at the crack tip. This
stress field can be determined analytically. It is characterized by the stress intensity factor.
The resulting crack growth criterion is referred to as local, because attention is focused at a
small material volume at the crack tip.
Assumption of linear elastic material behavior leads to infinite stresses at the crack tip.
In reality this is obviously not possible: plastic deformation will occur in the crack tip region.
Using yield criteria (Von Mises, Tresca), the crack tip plastic zone can be determined. When
this zone is small enough (Small Scale Yielding, SSY), LEFM concepts can be used.
When the plastic crack tip zone is too large, the stress and strain fields from LEFM are not
valid any more. This is also the case when the material behavior is nonlinear elastic (e.g. in
polymers and composites). Crack growth criteria can no longer be formulated with the stress
intensity factor.
In Elastic-Plastic Fracture Mechanics (EPFM) or Non-Linear Fracture Mechanics (NLFM)
criteria are derived, based on the Crack Tip Opening Displacement. Its calculation is possible
using models of Irwin or Dugdale-Barenblatt for the crack tip zone.
Another crack growth parameter, much used in NLFM, is the J-integral, which charac-
terizes the stress/deformation state in the crack tip zone.
Numerical techniques
Analytical calculation of relevant quantities is only possible for some very simple cases. For
more practical cases numerical techniques are needed. Numerical calculations are mostly done
using the Finite Element Method (FEM). Special quarter-point crack tip elements must be
used to get accurate results.
Fig. 1.11 : Loading and deformation of a cracked plate. (A quarter of the plate is
modelled and shown.)
Fatigue
When a crack is subjected to a time-dependent load, be it harmonic or random, the crack will
gradually grow, although the load amplitude is very small. This phenomenon is called Fatigue
Crack Propagation (FCP) and may unexpectedly and suddenly result in Fatigue Failure.
8
Outline
Many books and papers on fracture mechanics and fatigue exist, a lot of which are rather
intimidating for a first reader, because they are either too specialized and/or lack a proper
introduction of concepts and variables. This is partly due to the fact that Fracture Mechanics
has developed into many specialized subjects, all focused on different applications. In this
course a number of these subjects are discussed, with an emphasis on a clear introduction of
concepts and derivation of variables. The main goal is to give an overview of the whole field.
For more detail, the reader is referred to the specialized publications, some of which can be
found in the References.
Chapter 2
Fracture mechanics
In this chapter we first discuss some mechanisms, which can be recognized after fracture
has occured. They are generally related to the amount of dissipated energy during crack
growth and therefore to ductile or brittle fracture. The theoretical strength of a material
can be predicted from the maximum bond strength between atoms. The result of these
very approximating calculations appears to be not in accordance with experimental findings.
Griffith was the first to recognize that attention must be focused to imperfections like already
existing cracks. Fracture Mechanics is thus started with the experimental studies by Griffith
in 1921 and has since then developed in various directions.
shear fracture
cleavage fracture
fatigue fracture
crazing
de-adhesion
2.1.1 Shearing
When a crystalline material is loaded, dislocations will start to move through the lattice due
to local shear stresses. Also the number of dislocations will increase. Because the internal
structure is changed irreversibly, the macroscopic deformation is permanent (plastic). The
dislocations will coalesce at grain boundaries and accumulate to make a void. These voids
will grow and one or more of them will transfer in a macroscopic crack. One or more cracks
may then grow and lead to failure.
Because the origin and growth of cracks is provoked by shear stresses, this mechanism
is referred to as shearing. Plastic deformation is essential, so this mechanism will generally
be observed in FCC crystals, which have many closed-packed planes.
9
10
The fracture surface has a dough-like structure with dimples, the shape of which indicate
the loading of the crack. For some illustrative pictures, we refer to the book of Broek [10].
Fig. 2.1 : Dislocation movement and coalescence into grain boundary voids, resulting in
dimples in the crack surface [30].
2.1.2 Cleavage
When plastic deformation at the crack tip is prohibited, the crack can travel through grains by
splitting atom bonds in lattice planes. This is called intra- or trans-granular cleavage. When
the crack propagates along grain boundaries, it is referred to as inter-granular cleavage.
This cleavage fracture will prevail in materials with little or no closed-packed planes,
having HCP or BCC crystal structure. It will also be observed when plastic deformation is
11
prohibited due to low temperature or high strain rate. As will be described later, a three-
dimensional stress state may also result in this mechanism.
Inter-granular cleavage will be found in materials with weak or damaged grain bound-
aries. The latter can be caused by environmental influences like hydrogen or high temperature.
The crack surface has a shiny appearance. The discontinuity of the lattice orientations
in neighboring grains will lead to so-called cleavage steps, which resemble a river pattern.
Nice pictures can be found in [11].
intra-granulair inter-granulair
2.1.3 Fatigue
When a crack is subjected to cyclic loading, the crack tip will travel a very short distance in
each loading cycle, provided that the stress is high enough, but not too high to cause sudden
global fracture. With the naked eye we can see a clam shell structure in the crack surface.
Under a microscope striations can be seen, which mark the locations of the crack tip after
each individual loading cycle.
This mechanism is referred to as fatigue. Because crack propagation is very small in
each individual load cycle, a large number of cycles is needed before total failure occurs. The
number of cycles to failure Nf is strongly related to the stress amplitude = 21 (max min )
and the average stress m = 21 (max + min ).
Nice pictures of macroscopic and microscopic fatigue crack surfaces can again be found
in [11].
Fig. 2.3 : Clam shell fatigue crack surface and striations [30].
12
2.1.4 Crazing
In a polymer material sub-micrometer voids may initiate when a critical load level is ex-
ceeded. Sometimes, one or a limited number of these crazes grow locally to generate a large
and fatal crack. In other circumstances the crazes spread out over a larger area. This is indi-
cated as stress whitening, because the crazes refract the light, resulting in a white colored
appearance. Some nice pictures of crazes can be found in [29].
2.1.5 De-adhesion
Adhesion refers to bonding between atoms of different materials, while cohesion refers to
bonding between atoms of one and the same material. Applications include the joining of
two different entities such as structural parts and the adhesion of a generally thin layer on a
generally thicker substrate. In the latter case the thin layer is also often referred to as surface
layer, thin film or coating.
Surface layers may be metallic metals (M) and their nitrides (MNx) or oxides (MOx)
, anorganic oxides, carbides, diamond or organic polymers. Thicknesses vary from
several nanometers to hundreds of micrometers. Substrates can also be metallic metals and
alloys , anorganic ceramics, glass and organic.
The adhesion of the layer to the substrate is determined by chemical bond strength and
highly influenced by initial stresses and damage in the surface layer and the roughness of
the substrate. Initial residual stresses can be classified as thermal , due to solidification and
cooling , and mechanical , due to volume mismatch (penetration) and plastic strain (impact).
The adhesion strength can be characterized by the maximum normal strength, the maximum
13
tangential strength or the work needed for separation. Many experimental techniques are
available to determine these parameters. Modeling of the adhesion strength is often done
with cohesive zone (CZ) models, where the mentioned parameters play an important role.
2.2 Ductile - brittle behavior
When a crack propagates, new free surface is generated, having a specific surface energy ,
which for solid materials is typically 1 [Jm2 ]. This energy is provided by the external load
and is also available as stored elastic energy. Not all available energy, however, is used for the
generation of new crack surfaces. It is also transformed into other energies, like kinetic energy
or dissipative heat. When a lot of available energy is used for crack growth, the fracture is
said to be brittle. When a lot of energy is transformed into other energies, mainly due to
dissipative mechanisms, the fracture is indicated to be ductile.
Although boundary and environmental conditions are of utmost importance, it is com-
mon practice to say that a certain material is brittle or ductile. A first indication of this
follows from the stress-strain curve, registered in a tensile test up to fracture of the tensile
bar. The area under the curve is a measure for the dissipated energy before failure. Tensile
curves for a variety of materials are shown in the figure.
Because dissipation is associated with plastic deformation, shear fracture is often found
in materials which show ductile fracture. When plastic deformation and thus dissipation is
less, fracture is more brittle.
ABS, nylon, PC
PE, PTFE
0 10 100 (%)
from the height of the pendulum weight, before and after impact. The dissipated energy is
the Impact Toughness Cv [J].
The impact toughness can be determined for various specimen temperatures T . For intrinsic
brittle materials like high strength steel, the dissipated energy will be low for all T . For
intrinsic ductile materials like FCC-metals, Cv will be high for all T . A large number of
materials show a transition from brittle to ductile fracture with increasing temperature. The
transition trajectory is characterized by three temperatures :
As a rule of thumb we have for BCC alloys Tt = 0.1 a 0.2Tm and for ceramics Tt = 0.5 a 0.7Tm ,
where Tm is the melting temperature.
More on the Charpy test can be found in [29]. Other tests to determine the brittleness
are the Izod test and Drop Weight Test.
f
r
f f x
a0 1 r S
2
2x
f (x) = fmax sin ;
x = r a0
1X 2x
(x) = f (x) = max sin
S
The theoretical strength max can be determined from an energy balance. When the interac-
tion force and thus the inter-atomic distance increases, elastic energy is stored in each bond.
When the bond breaks at x = /2, the stored energy is released. The stored elastic energy
per unit of area is Ui and can be calculated by integration.
It is assumed that all bonds in the area S snap at the same time and that all the stored
energy is transformed in surface energy, which, per unit of area is Ua = 2, with the specific
surface energy of the material.
From the energy balance Ui = Ua , the wave length of the sine function can be calculated
and the stress as a function of the displacement x can be expressed in max .
x=/2 X
1
Z
Ui = f (x) dx
S x=0
16
x=/2
2x
Z
= max sin dx
x=0
= max [Nm1 ]
required surface energy
Ua = 2 [Nm1 ]
energy balance at fracture
2 x
Ui = Ua = = max sin max
max
The argument of the sine function is assumed to be small enough to allow a linear approxima-
tion. The displacement x between the atomic planes is expressed in the linear strain . The
macroscopic Youngs modulus E of the material is introduced as the derivative of the stress
w.r.t. the linear strain in the undeformed state. This leads to a relation for the theoretical
strength th = max .
linearization
x x 2
= max sin max
max
linear strain of atomic bond
x a0 2
= x = a0 =
a0 max
elastic modulus
d d 2 a0
E= = a0
= max
d x=0
dx x=0
r
E
max =
a0
theoretical strength
r
E
th =
a0
th b
11000
b
[MPa]
170
10 20 d []
Experimental techniques
To predict the behavior of a crack, it is essential to know its location, geometry and dimen-
sions. Experiments have to be done to reveal these data.
Experimental techniques have been and are still being developed. Some of these proce-
dures use physical phenomena to gather information about a crack. Other techniques strive
towards visualization of the crack.
19
20
3.3 X-ray
Direct visualization of a crack can be done using electromagnetic waves. X-rays are routinely
used to control welds.
Fig. 3.2 : X-ray robots inside and outside of a pipe searching for cracks.
(Source: Internet site University of Strathclyde (2006))
3.4 Ultrasound
Visualization is also possible with sound waves. This is based on the measurement of the
distance over which a wave propagates from its source via the reflecting crack surface to a
detector.
piezo-el. crystal
wave
sensor
S in
out
t t
bending test
scratch test
Fracture energy
4.1 Energy balance
Abandoning the influence of thermal effects, the first law of thermodynamics can be formu-
lated for a unit material volume :
the total amount of mechanical energy that is supplied to a material volume per unit of
time (Ue ) must be transferred into internal energy (Ui ), surface energy (Ua ), dissipated
energy (Ud ) and kinetic energy (Uk ).
The internal energy is the elastically stored energy. The surface energy changes, when new
free surface is generated, e.g. when a crack propagates. The kinetic energy is the result of
material velocity. The dissipation may have various characteristics, but is mostly due to
friction and plastic deformation. It results in temperature changes.
Instead of taking time derivatives () to indicate changes, we can also use another state
variable, e.g. the crack surface area A. Assuming the crack to be through the thickness of
a plate, whose thickness B is uniform and constant, we can also use the crack length a as a
state variable.
The internally stored elastic energy is considered to be an energy source and therefore
it is moved to the left-hand side of the energy balance equation.
B = thickness
a A = Ba
23
24
Ue = Ui + Ua + Ud + Uk [Js1 ]
d dA d d d
()= ( ) = A ( ) = a ( )
dt dt dA dA da
dUe dUi dUa dUd dUk
= + + + [Jm1 ]
da da da da da
According to Griffiths energy balance, a crack will grow, when the energy release rate equals
the crack resistance force.
This is illustrated with the following example, where a plate is loaded in tension and
fixed at its edges. An edge crack of length a is introduced in the plate. Because the edges are
clamped, the reaction force, due to prestraining, does not do any work, so in this case of fixed
grips we have dUe = 0. The material volume, where elastic energy is released, is indicated as
the shaded area around the crack. The released elastic energy is a quadratic function of the
crack length a. The surface energy, which is associated with the crack surface is 2a.
Both surface energy Ua and internal energy Ui are plotted as a function of the crack
length. The initial crack length a is indicated in the figure and it must be concluded that for
crack growth da, the increment in surface energy is higher than the available (decrease of)
internal energy. This crack therefore cannot growth and is stable.
When we gradually increase the crack length, the crack becomes unstable at the length
ac . In that case the crack growth criterion is just met (G = R). The increase of surface
energy equals the decrease of the internal energy.
25
dUe = 0
Ua da G, R
2
needed
ac ac
available
Ui
dUi dUa
< no crack growth
da da
dUi dUa
> unstable crack growth
da da
dUi dUa
= critical crack length
da da
We consider a crack of length 2a in an infinite plate with uniform thickness B [m]. The
crack is loaded in mode I by a nominal stress [Nm2 ], which is applied on edges at large
distances from the crack.
Because the edges with the applied stress are at a far distance from the crack, their
displacement will be very small when the crack length changes slightly. Therefore it is assumed
that dUe = 0 during crack propagation. It is further assumed that the elastic energy in the
elliptical area is released, when the crack with length a is introduced. The energy release rate
and the crack resistance force can now be calculated. According to Griffiths energy balance,
the applied stress and the crack length a are related. From this relation we can calculate
the Griffith stress gr and also the critical crack length ac .
26
y
thickness B
2a
a x
Fig. 4.3 : Elliptical region, which is unloaded due to the central crack of length 2a.
2
1
Ui = 2a2 B 2 E ; Ua = 4aB [Nm = J]
r
2E
Griffith stress gr =
a
2E
critical crack length ac =
2
1 dUe dUi
energy balance G= = R = Gc
B da da
27
Gc E
critical crack length ac =
2 2
F F
u u
P P
a a + da
F F
a a + da a a + da
dUi dUi
dUe
u u
Fig. 4.5 : Internal and external energy for fixed grips (left) and constant load (right).
28
1 dF 1 u2 dC 1 2 dC
G= u = 2
= F
2B da 2B C da 2B da
constant load
4.4.3 Experiment
It can be concluded that the energy release rate can be calculated from the change in com-
pliance and that the result for the fixed grip approach is exactly the same as that for the
constant load method.
In reality the loading of the plate may not be purely according to the extreme cases, as
is shown in the figure. From such a real experiment the energy release rate can be determined
from the force-displacement curve.
29
F
F a1 u
a2 P
a3 a
a4
shaded area 1
G=
a4 a3 B
4.4.4 Examples
As an example the energy release rate is calculated for the so-called double cantilever beam,
shown in the figure.
Using linear elastic beam theory, the opening u of the crack can be expressed in the
force F , the crack length a, beam thickness B, beam height 2h and Youngs modulus E. The
compliance is the ratio of opening and force. Differentiating the compliance w.r.t. the crack
length results in the energy release rate G, which appears to be quadratic in a.
F
u
B
2h
a
u
F
F a3 4F a3
u= =
3EI EBh3
u 2u 8a3 dC 24a2
C= = = =
F F EBh3 da EBh3
1 1 2 dC 2
12F a 2
G= 2 F = [J m2 ]
B da EB 2 h3
B 1
q
Gc = 2 Fc = Eh3
a 6
The question arises for which beam geometry, the energy release rate will be constant, so no
function of the crack length. It is assumed that the height of the beam is an exponential
function of a. Applying again formulas from linear elastic beam theory, it can be derived for
which shape G is no function of a.
u 2u 8a3 dC 24a2
C= = = =
F F EBh3 da EBh3
choice : h = h0 an
F a3 4F a3 4F a3(1n)
u= = =
3(1 n)EI (1 n)EBh3 (1 n)EBh30
2u 8a3(1n) dC 24a(23n)
C= = =
F (1 n)EBh30 da EBh30
dC 2
2
constant for n= 3 h = h0 a 3
da
Chapter 5
Stress concentrations
In this chapter attention is given to stress concentrations due to holes. Throughout this
chapter it is assumed that the material behavior is linear elastic and isotropic. The stress
state is calculated using Airy stress functions. The first sections present a summary of the
theory of linear elasticity.
Q
Q ~ ~
X + dX
~x + d~x
~u P
P
~
X ~x
~e3
~e2
~e1
31
32
xi = Xi + ui (Xi )
xi + dxi = Xi + dXi + ui (Xi + dXi ) = Xi + dXi + ui (Xi ) + ui,j dXj
The local deformation, i.e. the deformation of a small material volume in a material point, is
described by strains. When elongations and rotations are large, the Green-Lagrange strains
ij are commonly used. For small elongations and rotations, these can be linearized to the
linear strains ij .
Because three displacement components are the basis for the definition of six strain compo-
nents, the latter cannot be independent. The compatibility relations express this dependency.
5.2 Stress
When deformation is not completely unconstrained, it provokes stresses in the material. The
stress vector ~p on a plane in a material point, can be calculated from the unity normal
vector ~n on the plane and the Cauchy stress tensor. Using index notation in a coordinate
system, the components of p~ can be related to those of ~n by the Cauchy stress components
ij (i, j = 1, 2, 3). These stress components can be represented as the stress vectors on
the sides of an infinitesimal stress cube in the material point, with its sides normal to the
coordinate axes.
33
23
13
32
31 22
3 21
12
2
11
1
ij = Cijkl lk
34
For isotropic materials, Hookes law can be written in index notation, relating six strain
components to six stress components and vice versa. When these components are stored in
columns, Hookes law is represented by a six-by-six matrix.
E
ij = ij + ij kk i = 1, 2, 3
1+ 1 2
1+
ij = ij ij kk i = 1, 2, 3
E 1+
11 1 0 0 0 11
22 1 0 0 0 22
33 1 0 0 0 33
= 0
12 0 0 1 2 0 0 12
23 0 0 0 0 1 2 0 23
31 0 0 0 0 0 1 2 31
= E/[(1 + )(1 2)]
11 1 0 0 0 11
22
1 0 0 0
22
33 = 1 1 0 0 0 33
12 E
0 0 0 1+ 0 0
12
23 0 0 0 0 1+ 0 23
31 0 0 0 0 0 1+ 31
A material, which is subjected to external loads, will deform and the stresses in the deformed
state have to satisfy the equilibrium equations in each point. The three equilibrium equations
for forces are partial differential equations, which can only be solved with proper boundary
conditions. The three equilibrium equations for moments indicate that the Cauchy stress
components are symmetric.
We have assumed here that the acceleration of the material points can be neglected.
35
13 + 13,3 dx3
23 + 23,3 dx3
33 + 33,3 dx3
12
22
32 11
21
31
11 + 11,1 dx1
21 + 21,1 dx1
31 + 31,1 dx1
3 12 + 12,2 dx2
22 + 22,2 dx2
32 + 32,2 dx2
2 13
1 23
33
volume load qi
moment equilibrium ij = ji
Hookes law
E 1+
ij = ij + ij kk ; ij = ij ij kk i = 1, 2
1+ 1 E 1+
Hookes law in matrix notation
36
11 1 0 11
22 = 1 1 0 22
E
12 0 0 1+ 12
11 1 0 11
22 = E 1 0 22
1 2
12 0 0 1 12
33 = (11 + 22 ) = (11 + 22 )
E 1
13 = 23 = 0
Hookes law
1+ E
ij = (ij ij kk ) ; ij = ij + ij kk i = 1, 2
E 1+ 1 2
E
33 = (11 + 22 ) = (11 + 22 )
(1 + )(1 2)
13 = 23 = 0
37
ij,j = 0
E
ij = ij + ij kk,j
1+ 1 2
E
ij,j + ij kk,j = 0
1+ 1 2
ij = 12 (ui,j + uj,i )
E E
1
2 (ui,jj + uj,ij ) + ij uk,kj = 0
1+ (1 + )(1 2)
BCs
ui ij ij
(x1 , x2 ) ij = ,ij + ij ,kk ij,j = 0
1+
ij = (ij ij kk )
E
38
1+
ij = {,ij + (1 )ij ,kk }
E
212,12 11,22 22,11 = 0
2 (2 ) = 4 = 0
bi-harmonic equation
BCs
ij ij ui
Cylindrical coordinates
The bi-harmonic equation has been derived and expressed in Cartesian coordinates. To get
an expression in cylindrical coordinates, the base vectors of the cylindrical coordinate system
are expressed in the base vectors of the Cartesian coordinate system. It is immediately clear
that cylindrical base vectors are a function of the cylindrical coordinate .
~ez
~et
~er
~e3
~e2 y
~e1
r
x
{~er ()} = ~et () ; {~et ()} = ~er ()
Laplace operator
The Laplace operator is a scalar operator and can be written as the inner product of the
gradient operator, which is a vector operator.
Bi-harmonic equation
The bi-harmonic equation can now be written in cylindrical coordinates. Also the stress com-
ponents, which are derivatives of the stress function, can be written in cylindrical coordinates.
The latter expression can best be derived from the expression for the stress tensor and using
the expression for the gradient operator in cylindrical coordinates.
bi-harmonic equation
2 1 2 2 1 1 2
1
+ + + + =0
r 2 r r r 2 2 r 2 r r r 2 2
stress components
1 1 2
rr = + 2 2
r r r
2
tt =
r 2
1 1 2
1
rt = 2 =
r r r r r
2a
Fig. 5.5 : A circular hole in a plate, loaded with a uniform stress in x-direction.
Load transformation
At a circular cross-section at a distance r = b a, the uniform stress can be decomposed
in a radial stress rr and a shear stress rt , which can be related to using equilibrium.
The total load at the cross-section r = b is the sum of a radial load and a load that
depends on . For both loadcases, the stress field for b r a will be determined, after
which the combined solution follows from superposition.
rr rr
rt rt
2a
2b
rr (r = b, ) = 21 + 12 cos(2)
rt (r = b, ) = 12 sin(2)
41
I. rr (r = a) = rt (r = a) = 0
rr (r = b) = 21 ; rt (r = b) = 0
II. rr (r = a) = rt (r = a) = 0
rr (r = b) = 21 cos(2) ; rt (r = b) = 12 sin(2)
Load case I
In the first loadcase, the boundary at r = b is subjected to a radial load rr = 12 . Because
rr is not depending on , the problem is axisymmetric and is only a function of r. The
stress components can be expressed in this function. The bi-harmonic equation is an ordinary
differential equation in the variable r.
stress components
1 1 2 1 df
rr = + 2 2 =
r r r r dr
2 d2 f
tt = =
r 2 dr 2
1
rt = =0
r r
2 2
d 1 d d f 1 df
bi-harmonic equation + + =0
dr 2 r dr dr 2 r dr
A general solution for the differential equation has four integration constants. These have
to be determined from the boundary conditions for this loadcase. However, the boundary
conditions for rt cannot be used as this stress component is zero everywhere. Instead the
compatibility equation is used, which relates the strain components rr and tt . Also b a
is used in the calculation.
A
stresses rr = + B(1 + 2 ln r) + 2C
r2
A
tt = 2 + B(3 + 2 ln r) + 2C
r
strains (from Hookes law for plane stress)
1 A
rr = (1 + ) + B{(1 3) + 2(1 ) ln r} + 2C(1 )
E r2
11 A
tt = (1 + ) + B{(3 )r + 2(1 )r ln r} + 2C(1 )r
Er r
42
du d(r tt )
compatibility rr = = B=0
dr dr
2 BCs and b a A and C
a2 a2
rr = 21 (1 ) ; tt = 21 (1 + ) ; rt = 0
r2 r2
Load case II
The second loadcase is again defined by the stress components at r = a and r = b. At the
boundary r = b the stresses are harmonic functions of the coordinate . Confronting the
general expressions for stresses rr and rt with these prescribed values at r = b, indicates
that the Airy function (r, ) must be written as the product of a function g(r) and cos(2).
(The term cos(2) remains after two subsequent differentiations w.r.t. .) Substitution in the
bi-harmonic equation results in a differential equation for g(r) and expressions for the stress
components.
stress components
1 1 2 2
rr = + 2 2 ; tt = 2
r r r r
1 1 2
1
rt = 2 =
r r r r r
bi-harmonic equation
2
1 2
2
1 d g 1 dg 4
+ + + g cos(2) =0
r 2 r r r 2 2 dr 2 r dr r 2
2 2
d 1 d 4 d g 1 dg 4
+ + g cos(2) = 0
dr 2 r dr r 2 dr 2 r dr r 2
A general solution for the differential equation has four integration constants. These have
to be determined from the boundary conditions for this loadcase. Using b a, the stress
components rr , tt and rt can be determined as a function of r and .
1
general solution g = Ar 2 + Br 4 + C +D
r2
= Ar 2 + Br 4 + C r12 + D cos(2)
6C 4D
stresses rr = 2A + 4 + 2 cos(2)
r r
2 6C
tt = 2A + 12Br + 4 cos(2)
r
43
6C 2D
rt = 2A + 6Br 2 2 sin(2)
r4 r
a2 3a4 4a2
rr = 1 2 + 1 + 4 2 cos(2)
2 r r r
2 4
a 3a
tt = 1 + 2 1 + 4 cos(2)
2 r r
4 2
3a 2a
rt = 1 4 + 2 sin(2)
2 r r
Special points
At the inner hole boundary (r = a), the stress tt is maximum for = 2 and the stress
concentration is 3, which means that it is three times the applied stress . The stress concen-
tration factor (SCF) Kt is a ratio of the maximum stress and the applied stress and therefore
it is a dimensionless number. It appears not to depend on the hole diameter.
rr (r = a, ) = rt (r = a, ) = rt (r, = 0) = 0
tt (r = a, = 2 ) = 3
tt (r = a, = 0) =
max
stress concentration factor Kt = =3 [-]
Stress gradients
When we plot the stress tt as a function of the distance to the hole, it appears that the stress
gradient is higher when the hole diameter is smaller. For a larger hole there is more volume
of material with a higher than nominal stress level. The chance that there is a defect or flaw
in this volume is higher than in the specimen with the smaller hole. This makes a larger hole
to be more dangerous regarding the occurrence of damage than the smaller hole.
44
yy
b x
radius
a p p
yy (x = a, y = 0) = 1 + 2 = 1 + 2 a/ 2 a/
b
p
stress concentration factor Kt = 2 a/ [-]
Chapter 6
When a linear crack is loaded in Mode-I, II or III, stress components at the crack tip can be
calculated with the Airy stress function method. However, a real crack has zero radius ( = 0)
and such a crack tip represents a singularity. Due to this singularity, complex variables and
functions will enter the mathematics.
x2
r
x1
Fig. 6.1 : Line crack with local coordinate systems originating in the crack tip.
45
46
z
x2
r
~ei
~er x1
f (z) = + i = (x1 , x2 ) + i(x1 , x2 ) = f
f (z) = (x1 , x2 ) i(x1 , x2 ) = f
= 12 {f + f } ; = 21 i{f f }
2g 2g
Laplacian 2 g = +
x21 x22
derivatives
g g z g z g g
= + = +
x1 z x1 z x1 z z
2g 2g 2g 2g
= + 2 +
x21 z 2 z z z 2
g g z g z g g
= + =i i
x2 z x2 z x2 z z
2
g 2
g 2
g 2
g
2 = 2 +2 2
x2 z z z z
2g 2g 2g 2
Laplacian 2 g = + = 4 2 = 4
x21 x22 z z z z
2 2 (z, z) = 0
bi-harmonic equation
6.2.1 Stresses
From the general solution for the stress components 11 , 22 and 12 can be derived and
expressed in , , and . Derivatives w.r.t. z and z are indicated with ( ) . The detailed
derivation can be found in Appendix B.
11 = ,11 + , = ,22 = + 1
z + + z +
2
22 = ,22 + , = ,11 = + +
1
z + + z +
2
12 = ,12 = 21 i z + z
6.2.2 Displacement
The displacement vector ~u can be written in components w.r.t. a Cartesian coordinate system
{x1 , x2 } with orthonormal base vectors ~e1 and ~e2 . Considering the coordinate axes to be real
and imaginary axes in a complex plane, we have ~e1 = ~er and ~e2 = ~ei = i~er . The displacement
vector can thus be considered to be a vector in the direction ~er , having a length u, which
is a complex number. The real part is the x1 -component and the imaginary part is the
x2 -component.
~u
u2
x2
r
~ei u1
~e2
~er x1
~e1
~u = u1~e1 + u2~e2 = u1~er + u2~ei = u1~er + u2 i~er = (u1 + iu2 )~er = u~er
u = u1 + iu2 = u1 (x1 , x2 ) + iu2 (x1 , x2 ) = u(z, z)
u = u1 iu2 = u(z, z)
Because the displacements u and u are a function of z and z, the derivatives w.r.t. z and z
can be determined. It appears that u u
z and z can be expressed in strain components only.
49
u u x1 u x2 1 u u
= + =2 +i
z x1 z x2 z x1 x2
u1 u 2 u1 u 2
= 21 +i +i = 21 (11 22 + 2i12 )
x1 x1 x2 x2
u u x1 u x2 1 u u
= + =2 i
z x1 z x2 z x1 x2
u1 u2 u1 u2 u2 u1
= 21 +i i + = 21 11 + 22 + i
x1 x1 x2 x2 x1 x2
u u x1 u x2 u u
= + = 21 i
z x1 z x2 z x1 x2
u1 u2 u1 u2
= 21 i i = 21 (11 22 2i12 )
x1 x1 x2 x2
u u x1 u x2 1 u u
= + =2 +i
z x1 z x2 z x1 x2
1 u1 u2 u1 u2 1 u2 u1
= 2 i +i + = 2 11 + 22 i
x1 x1 x2 x2 x1 x2
Using Hookes law for plane strain, the derivative of u w.r.t. z can be expressed in the stress
components and thus in the complex functions and . Integration leads to an expression
for u, comprising an unknown function M (z).
u 1+ 1
= 2 11 22 + 2i12
z E
1+
= z +
E
1+
u= z + + M
E
Using the general solution for u and the conjugate u, the derivatives of these can be determined
and added. Adding the directly derived derivatives u u
z and z leads again to an expression
in strain components and through Hookes law in the functions and . From the general
solution, these derivatives can also be calculated, whereupon the integration function M can
be determined.
1+
u = z + + M
E
u 1+
+ M
=
z E
1+
z + + M
u =
E
u 1+
+ M
=
z E
50
u u 1+
+ + M + M
+ =
z z E
u u 1+
+ = 11 + 22 = [(1 2)(11 + 22 )]
z z E
(1 + )(1 2)
2 +
=
E
M + M = (3 4) +
M = (3 4) =
1+
u= z +
E
)
= ( + i)z +1 = ( + i)r +1 ei(+1)
= ( + i)z +1 = ( + i)r +1 ei(+1)
1 +1 h
u= r ( + i)ei(+1)
2
i
( i)( + 1)ei(1) ( i)ei(+1)
E
with =
2(1 + )
1 +1 h
u= r ( + i)ei(+1)
2
i
( i)( + 1)ei(1) ( i)ei(+1)
ei = cos() + i sin()
1 +1
u= r
2
h n
cos(( + 1)) sin(( + 1))
( + 1) cos((1 )) ( + 1) sin((1 ))
o
cos(( + 1)) + sin(( + 1)) +
n
i sin(( + 1)) + cos(( + 1))
( + 1) sin((1 )) + ( + 1) cos((1 )) +
o i
sin(( + 1)) + cos(( + 1))
= u1 + iu2
6.3 Mode I
For a crack loaded in Mode I, symmetry of the normal displacement and boundary conditions
for crack face stresses, allow the determination of the integration constants.
6.3.1 Displacement
For Mode I loading of the crack, the displacement must be symmetric w.r.t. the x1 -axis, which
is the crack direction. To meet this requirement, the constants and have to be zero. This
can be concluded when studying the displacement components and using sin() = sin()
and cos() = cos(). In the functions and three unknown constants remain.
u1 ( > 0) = u1 ( < 0)
==0
u2 ( > 0) = u2 ( < 0)
= z +1 = r +1 ei(+1) ; = z +1 = r +1 ei(+1)
h n oi
11 = ( + 1) z + z 21 zz 1 + z + zz 1 + z
h n oi
22 = ( + 1) z + z + 12 zz 1 + z + zz 1 + z
h i
12 = 12 i( + 1) zz 1 + z zz 1 z
with z = rei ; z = rei
h
11 = ( + 1)r ei + ei
n oi
1 i(2) i i(2) i
2 e + e + e + e
h
22 = ( + 1)r ei + ei +
n oi
1 i(2) i i(2) i
2 e + e + e + e
h i
12 = 21 i( + 1)r ei(2) + ei ei(2) ei
with ei + ei = 2 cos() ; ei ei = 2i sin()
The boundary condition stating that the crack surface is stress-free, results in two equations
for the unknown constants and . For a non-trivial solution the equations need to be
dependent. This confines the values of to be a series of discrete numbers. For each value of
we find a ratio between and .
22 ( = ) = 12 ( = ) = 0
( 2) cos() cos() 0
=
sin() sin() 0
( 2) cos() cos()
det = sin(2) = 0 2 = n
sin() sin()
n
= 21 , , with n = 0, 1, 2, ..
2
The stress field is a series of terms, one for each value of in the range [ 12 , 0, 12 , 1, ]. The
first terms are presented here. Only one unknown constant () remains in these solution.
= 21 = 2 ; =0 = 12
= 21 = 2 ; =1 =
1
11 = 2r 2 1 3 1
cos( 2 ) 1 sin( 2 ) sin( 2 ) +
1
22 = 2r 2 cos( 12 ) 1 + sin( 23 ) sin( 21 ) +
1
12 = 2r 2 cos( 21 ) cos( 23 ) sin( 12 ) +
53
1
KI = lim 2r 22 |=0 = 2 2 [ m 2 N m2 ]
r 0
r. For plane stress and plane strain they differ because of the different versions of Hookes
law.
KI
cos( 21 ) 1 sin( 12 ) sin( 32 )
11 =
2r
KI
cos( 21 ) 1 + sin( 12 ) sin( 32 )
22 =
2r
KI
cos( 21 ) sin( 21 ) cos( 32 )
12 =
2r
r
KI r
cos( 12 ) 1 + 2 sin2 ( 12 )
u1 =
2 2
r
KI r
sin( 12 ) + 1 2 cos2 ( 12 )
u2 =
2 2
6.4 Mode II
For a crack loaded in Mode II, anti-symmetry of the tangential displacement component and
boundary conditions for the crack face stresses, allow the determination of the integration
constants.
6.4.1 Displacement
For Mode II loading of the crack, the displacement must be anti-symmetric w.r.t. the x1 -axis,
which is the crack direction. To meet this requirement, the constants and have to be
zero. This can be concluded from the displacement components, using sin() = sin()
and cos() = cos(). In the functions and three unknown constants remain.
54
u1 ( > 0) = u1 ( < 0)
==0
u2 ( > 0) = u2 ( < 0)
= iz +1 = ir +1 ei(+1) ; = iz +1 = ir +1 ei(+1)
1
KII = lim 2r 12 |=0 [ m 2 N m2 ]
r 0
r. For plane stress and plane strain they differ because of the different versions of Hookes
law.
KII
sin( 21 ) 2 + cos( 12 ) cos( 23 )
11 =
2r
KII
sin( 21 ) cos( 12 ) cos( 23 )
22 =
2r
KII
cos( 21 ) 1 sin( 12 ) sin( 23 )
12 =
2r
r
KII r
sin( 12 ) + 1 + 2 cos2 ( 12 )
u1 =
2 2
r
KII r
cos( 21 ) 1 2 sin2 ( 12 )
u2 =
2 2
31 = 12 u3,1 32 = 21 u3,2
;
Hookes law
31 = 231 = u3,1 ; 32 = 232 = u3,2
equilibrium
6.5.2 Displacement
The general solution of the Laplace equation is written as the sum of a complex function f
and its conjugate f. Choosing a specific function f with three yet unknown constants A,
B and , it can again be argumented that the exponent must be larger than 1, because
displacements must remain finite.
general solution u3 = f + f
32 ( = ) = 0
sin() cos() A 0
=
sin() cos() B 0
sin() cos()
det = sin(2) = 0 2 = n
sin() cos()
56
n
= 21 , , .. with n = 0, 1, 2, ..
2
1 1
31 = Br 2 {sin( 12 )} ; 32 = Br 2 {cos( 12 )}
KIII = lim 2r 32 |=0
r 0
KIII KIII
sin( 12 ) cos( 21 )
31 = ; 32 =
2r 2r
r
2KIII r
sin( 12 )
u3 =
2
KI KII KIII
ij = fIij () ; ij = fIIij () ; ij = fIIIij ()
2r 2r 2r
KI = I a ; KII = II a ; KIII = III a
6.6.1 K-zone
The question may arise at what distance to the crack tip, displacement and stresses are still
described accurately by the first term of the total solution. There is no clear answer to this
question. The K-zone, where the first term, whose value is determined by the stress intensity
factor K, is the only important one, depends on geometry and loading, as is illustrated in the
figure.
K-zone : D
I II DII DI
a 1/2
KI = a sec
W
a
KII = a small
2a 2a W
W W
a
KI = a 1.12 0.41 +
W
a 2 a 3
18.7 38.48 +
a W W
a 4
W 53.85
W
a
1.12 a small
W
a
KI = a 1.12 + 0.76
W
a 2 a 3
8.48 + 27.36
a a W W
W 1.12 a
For point loads and a crack loaded by internal pressure, SIFs are also known from literature.
a 1
"
PS 2
KI = 2.9
P BW 3/2 W
a 3 a 5
2 2
4.6 + 21.8
W W W
a a 7 a 9
#
2 2
P/2 S P/2 37.6 + 37.7
W W
59
a 1
"
P P 2
KI = 29.6
BW 1/2 W
a 3 a 5
a 2 2
185.5 + 655.7
W W #
W a 7 a 9
2 2
1017 + 638.9
W W
P
2a
KI = p a
p
p per unit thickness
W
6.9 Relation GK
The global crack growth criterion was formulated in terms of the energy release rate G and
derived from an energy balance. The local stress growth criterion was formulated in terms
of the stress intensity factor K and derived from crack tip stresses. It may not be surprising
that there is a relation between G and K, which will be derived in this section.
We consider a crack in an infinite plate and focus attention on normal stresses and
openings in the crack plane. First we consider a crack of half length a and then a crack of
half length a + a. To close the latter back to length a again, we know which stress is needed
and which opening has to be eliminated. Multiplying normal load and closing displacement,
60
results in the work needed to accomplish this. This work equals the energy which has been
released during opening, the energy release rate G.
y
yy
a a
crack length a
a
yy ( = 0, r = x a) = p ; uy = 0
2(x a)
crack length a + a
yy ( = , r = a + a x) = 0
(1 + )( + 1) a + a
uy = a + a x
E 2
KI2
plane stress G=
E
KI2
plane strain G = (1 2 )
E
A relation between G and K is also available for multi-mode loading and can be derived by
superposition.
61
1
c1 KI2 + c2 KII
2 2
G= + c3 KIII
E
1
plane stress G= (K 2 + KII
2
)
E I
(1 2 ) 2 2 (1 + ) 2
plane strain G= (KI + KII )+ KIII
E E
KI
KIc
2a
B
Bc B
Fig. 6.7 : Central crack in a plate, loaded with Mode I stress (left).
Dependence of critical K-value on the plate thickness B (right).
2
KIc
KIc = c a ; Bc = 2.5
y
62
Material v [MPa] KIc [MPa m ]
steel, 300 maraging 1669 93.4
steel, 350 maraging 2241 38.5
steel, D6AC 1496 66.0
steel, AISI 4340 1827 47.3
steel, A533B reactor 345 197.8
steel, carbon 241 219.8
In practical situations, a crack is mostly subjected to a combined Mode I and Mode II loading.
In the figure the load is visualized as an edge load on a square material volume element with
edges parallel and perpendicular to the crack. A tensile load parallel to the crack can be added
because it has no influence on the crack tip stress. Each random load can be transformed
to this loading situation by proper transformation of the edge loads. The crack tip stresses
can be determined as the superposition of the stress components due to separate Mode I and
Mode II loadings.
Mode I Mode II
Mode I + II
Mode I + II
Fig. 7.1 : Central crack loaded in Mode I, Mode II and a combination of Mode I and II.
63
64
base are stored in columns n and p . Using the relation between {~e1 , ~e2 } and {~e1 , ~e2 }, a
transformation matrix T can be determined
which relates the columns n to n, and p to p.
(b) p~
~n
~e2
~e2
~e1
~e1
Fig. 7.2 : Material plane with unit normal vector and stress vector.
p = T p p = T T p idem : n = T T n
With respect to the basis {~e1 , ~e2 }, the stress vector components p can be calculated from the
Cauchy stress matrix and the components n of the normal vector. Using the transformation
matrix T , the matrix w.r.t. the basis {~e1 , ~e2 } can be calculated from .
p = n
T p = T n p = T T T n = n
= T T T = T T T
11 12 c s 11 12 c s
=
21 22 s c 21 22 s c
c s c11 + s12 s11 + c12
=
s c c21 + s22 s21 + c22
c2 11 + 2cs12 + s2 22
cs11 + (c2 s2 )12 + cs22
=
cs11 + (c2 s2 )12 + cs22
2 2
s 11 2cs12 + c 22
65
Stress component transformation can also be used to determine stress components in a cylin-
drical coordinate system. These will be needed later in crack growth direction criteria. With
the figure below, it is easily understood that cylindrical stress components can be derived
from Cartesian components by transformation of the stress matrix.
yy
xy
~et xx
~er
r
~e2
~e1 rt
rr tt
Fig. 7.3 : Cartesian and cylindrical stress cubes at the crack tip.
yy
xy
xx
KI KII
xx = fIxx () + fIIxx ()
2r 2r
KI KII
yy = fIyy () + fIIyy ()
2r 2r
KI KII
xy = fIxy () + fIIxy ()
2r 2r
Substitution in the transformation relation, results in cylindrical crack tip stresses. The
functions frr , ftt and frt can be derived, using some trigonometry.
rr
rt
tt
KI KII
rr = fIrr () + fIIrr ()
2r 2r
KI KII
tt = fItt () + fIItt ()
2r 2r
KI KII
rt = fIrt () + fIIrt()
2r 2r
5 1 3
fIrr () = cos( ) cos( )
4 2 4 2
fIIrr () = 45 sin( 2 ) + 43 sin( 3
2 )
fItt () = 34 cos( 2 ) + 14 cos( 3
2 )
fIItt () = 43 sin( 2 ) 34 sin( 3
2 )
fIrt () = 41 sin( 2 ) + 41 sin( 3
2 )
fIIrt () = 14 cos( 2 ) + 34 cos( 3
2 )
67
influence on the crack tip stress state. Note that in the next cases the load is indicated
by ij and the crack tip stresses by sij .
22
22
12
12
12
11
2a 11 2a
12
~e2
~e1
22
~e2
~e1
Fig. 7.6 : Transformation of multi-mode load into Mode I and II loads.
load
2 + 2cs + s2
c 11 12 22
cs11 + (c2 s2 )12 + cs22
11 12
= cs11 + (c s2 )12 + cs22
2
21 22
2 2
s 11 2cs12 + c 22
KI KII
crack tip stresses sij = fIij () + fIIij ()
2r 2r
with KI = 22 a ; KII = 12 a
22
11
12
12
2a k 2a
load
11 = c2 11 + 2cs12 + s2 22 = c2 k + s2
22 = s2 11 2cs12 + c2 22 = s2 k + c2
12 = cs11 + (c2 s2 )12 + cs22 = cs(1 k)
KI KII
sij = fIij () + fIIij ()
2r 2r
KI = I 22 a = I (s2 k + c2 ) a
KII = II 12 a = II cs(1 k) a
2a a
22
p
12
R t
11
Fig. 7.8 : Crack in tube wall at an angle with the tube axis.
pR pR 1
t = = ; a = = 21 k=
t 2t 2
22 = s2 12 + c2 ;
12 = cs(1 12 ) = 21 cs
pR
KI = 22 a = ( 21 s2 + c2 ) a = ( 12 s2 + c2 ) a
t
pR
KII = 12 a = 21 cs = 21 cs a
t
69
For special loading cases it will be possible to calculate the crack growth direction by analytical
means. More general cases are explored with numerical techniques.
For both criteria, we have to use stress and displacement components in a cylindrical
coordinate system.
rr
rt
tt
tt 2 tt
=0 and <0 c
2
KIc
tt ( = c ) = tt ( = 0) = crack growth
2r
70
Because the function tt () is known, the relations can be elaborated.
tt
=0
3 KI
1 3 KII 1
4 sin( 2 ) 14 sin( 3 4 cos( 2 ) 3
cos( 3
2 2 ) + 2
4 2 ) =0
2r 2r
KI sin() + KII {3 cos() 1} = 0
2 tt
<0
2
3 KI
1 3 KII 1
4 cos( 2 ) 34 cos( 3 9 3
4
2 ) + 4
4 sin( 2 ) + 4 sin( 2 ) < 0
2r 2r
KIc
tt ( = c ) =
2r
1 KI KII
3 cos( 2c ) + cos( 32c ) + 41 3 sin( 2c ) 3 sin( 32c ) = 1
4K KIc
Ic
Mode I load
For mode I loading of the crack, it is expected that the crack tip will advance in the plane of
the crack. Indeed this obviously results from solving the relations for .
tt
= KI sin() = 0 c = 0
2 tt
<0
2 c
KIc
tt (c ) = KI = KIc
2r
Mode II load
For pure mode II loading of the crack, the crack growth direction has to be determined
from the relations for . The requirement that the first derivative of tt () is zero, leads
to an equation, which results in two candidate values for c . Only for the negative value
c = 70.6o , the second requirement is satisfied, so that we can conclude that the crack tip
will move in that direction. The critical value of the mode II stress intensity factor can be
expressed in the critical value of the mode I SIF.
tt
= KII (3 cos(c ) 1) = 0 c = arccos( 13 ) = 70.6o
2 tt
<0 c = 70.6o
2 c
KIc
q
tt (c ) = KIIc = 34 KIc
2r
71
Multi-mode load
Determining the crack growth direction for multi-mode loading is only possible using nu-
merical calculations. The relations which have to be satisfied for crack growth in a certain
KI
direction are written is terms of the parameter K Ic
.
KI [ sin( 2 ) sin( 3 3
2 )] + KII [ cos( 2 ) 3 cos( 2 )] = 0
KI [ cos( 2 ) 3 cos( 3 3
2 )] + KII [sin( 2 ) + 9 sin( 2 )] < 0
KI [3 cos( 2 ) + cos( 3 3
2 )] + KII [3 sin( 2 ) 3 sin( 2 )] = 4KIc
KI KII
f1 f2 = 0
KI f1 KII f2 = 0
K K
Ic Ic
KI KII
KI f2 + KII f3 < 0 f2 + f3 < 0
KIc KIc
KI KII
KI f4 3KII f1 = 4KIc
f4 3 f1 = 4
KIc KIc
KI
For a range of values 0 K Ic
1 a range of values for the crack growth direction angle
0 90o is evaluated to decide if the relations are satisfied. In the first one K
KIc is plotted
II
KI
as a function of KIc and it can be used to determine which combination leads to crack growth.
From the second plot the crack growth angle c can then be determined.
0 0.9
0.8
10
0.7
20
0.6
30 0.5
Ic
K /K
c
II
40 0.4
0.3
50
0.2
60
0.1
70 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
K /K K /K
I Ic I Ic
Fig. 7.11 : Crack growth direction and Mode ratio for multi-mode loading.
72
S 2S
=0 and >0 c
2
For the case of linear elastic material behavior, the strain energy density function can be
expressed in the stress components and the material parameters E and . At the crack tip
these stress components are known functions of r, , KI and KII . The strain energy density
factor is than also known and dependent on . KI and KII .
The requirements for the crack growth direction result in two relations, from which c
can be determined.
1 1 2
Ui = ( 2 + yy
2 2
+ zz ) (xx yy + yy zz + zz xx ) + 2
( + yz 2
+ zx )
2E xx E 2G xy
KI K
cos( 2 ) 1 sin( 2 ) sin( 3 II sin( 2 ) 2 + cos( 2 ) cos( 3
xx = 2 ) 2 )
2r 2r
KI K
cos( 2 ) 1 + sin( 2 ) sin( 3 II sin( 2 ) cos( 2 ) cos( 3
yy = 2 ) + 2 )
2r 2r
KI K
sin( 2 ) cos( 2 ) cos( 3 II cos( 2 ) 1 sin( 2 ) sin( 3
xy = 2 )+ 2 )
2r 2r
73
1
with a11 = 16G (1 + cos())( cos())
1
a12 = 16G sin(){2 cos() ( 1)}
1
a22 = 16G {( + 1)(1 cos()) + (1 + cos())(3 cos() 1)}
ki = Ki /
Mode I load
For Mode I loading we have KII = 0 and KI = a. The first requirement results in two
possible values of c . The second requirement is only satisfied for c = 0, because 1 3.
The obvious crack growth direction is c = 0. Crack growth will take place if the value of
S(c ) equals the critical value Sc , which is the value for plane strain.
2 a
S = a11 kI2 = {1 + cos()}{ cos()}
16G
S
= sin(){2 cos() ( 1)} = 0
1
c = 0 or arccos 2 ( 1)
2S
= 2 cos(2) ( 1) cos() > 0 c = 0
2
2a 2a
S(c ) = {2}{ 1} = ( 1)
16G 8G
(1 + )(1 2) 2
Sc = S(c , pl.strain) = KIc
2E
Mode II load
For Mode II loading we have KI = 0 and KII = a. The two requirements lead to two
possible crack growth directions. The result differs for plane stress and plane strain. For
= 0, leading to = 3, the direction from SED is the same as the one from MTS. Crack
growth will occur if the minimum S-value equals the critical value Sc , which was determined
in pure Mode I loading. From this value the critical shear stress can be derived.
2 2a
S = a22 kII = [( + 1){1 cos()} + {1 + cos()}{3 cos() 1}]
16G
S
= sin() [6 cos() + ( 1)] = 0
2S 2
= 6 cos () + ( 1) cos() > 0
2
c = arccos 16 ( 1)
2a 1
S(c ) = { (2 + 14 1)}
16G 12 r
1 192GSc
S(c ) = Sc c = 2
a + 14 1
74
Multi-mode load
It is again not possible to determine the crack growth direction for a general multi-mode
loading. Numerically it can be done, however.
10 1
=0.0.75
=0.1
0 0.9
=0.2
=0.3
10 =0.4 0.8
=0.495
20 0.7
30 0.6 =0.0.75
=0.1
Ic
=0.2
K /K
c
40 0.5
=0.3
II
=0.4
50 0.4 =0.495
60 0.3
70 0.2
80 0.1
90 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
KI/KIc K /K
I Ic
Fig. 7.12 : Crack growth direction and Mode ratio for multi-mode loading.
10 1
=0.0.75
=0.1
0 0.9
=0.2
=0.3
10 =0.4 0.8
=0.495
20 0.7
30 0.6
Ic
K /K
c
40 0.5
II
50 0.4 =0.0.75
=0.1
60 0.3 =0.2
=0.3
=0.4
70 0.2
=0.495
80 0.1
90 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
KI/KIc K /K
I Ic
Fig. 7.13 : Crack growth direction and Mode ratio for multi-mode loading.
The figure below shows a crack oriented at an angle w.r.t. the tensile load . For this
situation the scaled stress intensity factors kI and kII can be calculated, from which follows
the expression for S as a function of and . The requirements according to the SED criterion
can be used to test a sequence of c values for a sequence of values and a plot can be made.
The figure shows the result for plane strain.
kI = a sin2 () ; kII = a sin() cos()
S
= ( 1) sin(c 2) 2 sin{2(c )} sin(2c ) = 0
2S
= ( 1) cos(c 2) 4 cos{2(c )} 2 cos(2c ) > 0
2
75
90
c 80
70 = 0.5
60
50
c
40
=0
2a 30
20 = 0.1
10
0
0 10 20 30 40 50 60 70 80 90
When a crack is subjected to impact loading, the crack tip stresses and strains can only
be calculated by taking into account dynamic effects like material acceleration and solving
the equations of motion. This is the subject of Dynamic Fracture Mechanics (DFM) and
understandably involves rather sophisticated mathematics.
Another subject, studied in DFM, is the behavior of a crack, which grows at high rate
under (quasi)static loading after reaching its critical length. In this chapter an approximate
value for the crack tip speed is derived, using the global energy balance, taking into account
the kinetic energy term. This approach is referred to as the kinetic approach. In DFM a
so-called static approach is also used to study the behavior of cracks, subjected to thermal
shock loading.
To calculate the crack tip speed or crack growth rate, the general energy balance is used
as a starting point. It is elaborated for a crack with length 2a in an infinitely large plate
with uniform and constant thickness B. The crack is subjected to a Mode I load , resulting
from an external load applied at edges far away from the crack. This implies that dUe = 0
when the crack length changes. It is assumed that the presence of the crack has led to a
complete release of elastic energy Ui in the elliptical area, indicated in the figure. Dissipation
is assumed to be zero, so all available energy is going to be transformed into surface energy
Ua and kinetic energy Uk .
77
78
y
thickness B
2a
a x
Fig. 8.1 : Central crack in large plate with unloaded elliptical area.
dUe dUd
=0 ; =0
da da
dUa
Ua = 4aB = 4B
da
2 dUi 2aB 2
Ui = 2a2 B 21 =
E da E
Kinetic energy
The total kinetic energy of the plate is the integral over its volume of the specific kinetic
energy 12 (u2x + u2y ), where is the material density and ux and uy the material velocity
components.
When the crack opens, the material velocity perpendicular to the crack faces (uy ) is
much higher than the velocity in the crack plane (ux ). To allow further derivations, it is
assumed that the crack tip speed s is independent of the (half) crack length a.
The displacement uy of the crack face ( = ) is available from the analytical solution
for a static crack, here assumed to be in a state of plane stress. The kinetic energy change at
crack growth can now be calculated by integration. However, the integral cannot be calculated
analytically. Its value is indicated as ak(a) and will be considered later.
Z
Uk = 21 B (u2x + u2y ) dxdy
duy duy da duy
material velocity ux uy = = = s
dt da dt da
Z 2
1 2 duy
Uk = 2 s B dxdy
da
ds
assumption =0
da
79
2
dUk d du
Z
y
= 21 s2 B
dxdy
da da da
p duy 2a x
uy = 2 2 a2 ax = 2
E da E a2 ax
dUk 2 Z 1 x2 (x 2a) 2
= s2 B a 3 (a x)2
dxdy = s 2
B a k(a)
da E a E
Substitution of all results in the energy balance leads to an expression for the crack tip speed
s where the critical crack length ac and the sound speed c can be introduced.
The earlier assumption that s is independent of a does not appear to hold, so the result
must be an approximation. More approximation is involved in the calculation of k(a), which
is done numerically. Roberts and Wells did this in 1954 [56] and came up with k = 43.51.
As can be seen from the expression for s, the crack will grow at high speed, so that the
actual (half) crack length a will be much larger than the critical crack length ac . This leads
to the final approximation for the crack tip speed.
2a 2 2
= 4 + s2 ak
E E
1 1 1
E 2 2 2 2E 2
s= 1
k a 2
r s
2 2E E
0.38 ; ac = ; c=
2
k
1
ac 2
s = 0.38 c 1
a s 0.38 c
a ac
For some materials the table lists values for Youngs modulus E, density and calculated
sound speed c. The measured crack tip speed s is also listed. The ratio s/c takes values
between 0.2 and 0.4, so the approximation 0.38 is well within this range. Mostly the experi-
mental ratio is found to be smaller, due to the dissipation during crack growth.
s
0.2 < < 0.4
c
s
E
C0 = elongational wave speed =
r
+1
r
C1 = dilatational wave speed =
r 1
C2 = shear wave speed =
CR = Rayleigh velocity = 0.54 C0 a 0.62 C0
Corrections
More elaborate derivations of the crack tip speed have been published. Dulancy & Brace
(1960) derived an expression without assuming s to be independent of a. Freund (1972)
related s to the Rayleigh velocity CR .
ac
Dulancy & Brace (1960) s = 0.38 C0 1
a
ac
Freund (1972) s = CR 1
a
KD
Dij = fij (, r, s, E, )
2r
81
0.9
Dtt () s
tt
max
Dtt ( = 0) cR
1 0.6 0.87
0
crack branching
volgens MTS
Fig. 8.2 : Tangential stress ratio as a function of crack growth direction, illustrating crack
branching.
8.4 Experiments
In DFM experiments are done to measure crack tip speeds, using instrumented DCB speci-
mens, sometimes in combination with high speed photography (typically 106 frames/sec).
To determine the temperature dependent KDc , the Robertson Crack Arrest Temperature
(CAT) test is done, where a crack is forced to run from a low temperature region into a higher
temperature region.
Fig. 8.3 : Instrumented DCB specimen to measure crack propagation speed [39].
Chapter 9
The elastic crack tip stresses reach an infinite value, when the distance to the crack tip
decreases to zero. In reality the material will yield before the crack tip is reached and the
elastic solution is no longer valid.
In this chapter the region at the crack tip where yielding occurs is determined. The
influence of this plastic crack tip zone on the crack behavior is explained. Post-yield hardening
is not considered, so the material behaves ideal plastic.
In three-dimensional cases, the occurrence of yielding is tested by means of a yield
criterion. Here, we employ the Von Mises and the Tresca criteria. Both are formulated in
terms of principal stresses 1 , 2 and 3 , where the first two are in the plane of the plate and
3 is in the normal direction. For plane stress cases 3 = 0 and for plane strain situations
Hookes law gives 3 = (1 + 2 ), with as Poissons ratio.
83
84
For both the Von Mises and the Tresca yield criterion, the principal stresses have to be used.
They are the eigenvalues of the Cauchy stress matrix and for the plane stress case they can
be calculated analytically. For plane strain the third principal stress is not zero but related
to the first two principal stresses.
2 (xx + yy ) + (xx yy xy
2
) =0
1/2
1 = 12 (xx + yy ) + yy )2 + xy
2
1
4 (xx
1/2
2 = 12 (xx + yy ) yy )2 + xy
2
1
4 (xx
3 = 0
Elastic crack tip stress components for Mode I are used to derive the principal crack tip
stresses. The first principal stress is always the largest. For plane stress the third principal
stress is zero and the smallest. For plane strain the smallest value may be the second principal
stress, because the value of the third principal stress depends on Poissons ratio.
KI
crack tip stresses ij = fIij ()
2r
KI
1(+),2() = cos( 2 )
2r
q
1
3
2 3
2
4 2 cos( 2 ) sin( 2 ) sin( 2 ) + sin( 2 ) cos( 2 ) cos( 2 )
KI KI
1 = cos( 2 ){1 + sin( 2 )} ; 2 = cos( 2 ){1 sin( 2 )}
2r 2r
2KI
3 = 0 or 3 = cos( 2 )
2r
= 0.25 = 0.35
1000 1000
800 800
600 600
400 400
200 200
0 0
0 20 40 60 80 100 0 20 40 60 80 100
= 0.45 = 0.5
1000 1000
800 800
600 600
400 400
200 200
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Fig. 9.1 : Principal stresses for plane strain as a function of the direction angle. The third
principal stress depends on Poissons ratio.
(1 2 )2 + (2 3 )2 + (3 1 )2 = 2y2
plane stress 3 = 0
(1 2 )2 + 22 + 12 = 2y2
KI2
cos2 ( 2 ) 6 sin2 ( 2 ) + 2 = 2y2
2rv
KI2 2 2 KI2 3
sin2 ()
ry = cos ( 2 ) 1 + 3 sin ( 2 ) = 1 + cos() + 2
2y2 4y2
plane strain 3 = (1 + 2 )
86
( 2 + 1)(12 + 22 ) + (2 2 2 1)1 2 = y2
KI2
cos2 ( 2 ) 6 sin2 ( 2 ) + 2(1 2)2 = 2y2
2ry
KI2 2 3 2
ry = (1 2) {1 + cos()} + 2 sin ()
4y2
The elastic/plastic boundaries can be plotted in an xy-coordinate system with the crack tip
in the origin and the crack along the line < x < 0; y = 0. It appears that the plane stress
plastic zone is larger than the plane strain plastic zone.
0.5
0.5
1
0.5 0 0.5 1 1.5
Fig. 9.2 : Plastic zones according to the Von Mises yield criterion.
max min = y
K
p I cos( 2 ) + cos( 2 ) sin( 2 ) = y
2ry
KI2 2
cos( 2 ) + cos( 2 ) sin( 2 )
ry = 2
2y
KI2 2
(1 2) cos( 2 ) + cos( 2 ) sin( 2 )
ry = 2
2y
KI2
ry = sin2 ()
2y2
Plotting the elastic-plastic boundary reveals again that the largest plastic zone occurs for the
plane stress case. For the plane strain situation, the plastic region is defined by two curves,
one for 1 > 2 > 3 (smallest) and one for 1 > 3 > 2 (largest).
0.5
1
0.5 0 0.5 1 1.5
an in-plane load is applied. Over the whole plate thickness a plane stress state exists with
3 = 0. The plane stress plastic crack tip zone develops over the whole plate thickness.
For the thick plate, only the surface layers are free to contract in normal direction and
are thus in a plane stress state. The crack tip material at greater depth from the surface
is prohibited to contract by the surrounding material, which is at a much lower stress and
has no intention at all to contract. The inner crack tip material is thus in a state of plane
strain. In the surface layers we find plane stress plastic zones and in the deeper bulk material
the plastic crack tip zone is smaller as it is associated with plane strain. A typical dog-bone
plastic zone develops at the crack tip.
Fig. 9.4 : Plastic crack tip zone for a thin and a thick plate.
In a thin plate the relative volume of yielding material is much larger than in the thick plate.
This explains that measured KIc values are higher when determined with thin test specimens.
For thicker specimens, the value is lower and becomes independent of the thickness, when
this is larger than a threshold value Bc , which is empirically related to KIc and y .
2 2
25 KIc KIc
Bc > > 2.5
3 y y
Plastic deformation takes place by sliding of crystallographic planes induced by shear stresses.
The planes with the maximum shear stresses can be found at an angle of 45o between the
planes with the maximum and minimum principal stresses, which are perpendicular. Em-
ploying an etching procedure, the shear planes can be visualized in some cutting planes.
89
Fig. 9.5 : Shear planes at the crack tip: plane stress (left), plane strain (right) [39, 25, 30].
In a plane stress state with m = 0, we can consider the influence of the pcf on yielding. For
a range of n-ratios, the pcf is calulated and the corresponding {1 , 2 }-values constitute the
plane stress Von Mises yield contour, as expected.
2
1.5
1
y
/ = n /
1
0.5
y
2
0.5
1
1 0.5 0 0.5 1 1.5 2
/
1 y
In the crack tip region the principal stresses are known as a function of the angle around
and the distance r to the crack tip. The ratio of 2 and 1 , n, can be calculated. The ratio
of 3 and 1 , m, is zero for plane stress and a function of Poissons ratio for plane strain.
n = 1 sin( 2 ) / 1 + sin( 2 )
plane stress ; m=0
In the plane of the crack where = 0 we have n = 1. For plane stress we already saw that
m = 0 and for plane strain m = 2. The plastic constraint factor can then be determined as
a function of Poissons ratio . For = 31 its value is 3, but experiments indicate that it is
p
smaller. Irwin [36] suggests an over the thickness averaged value pcf = 2 2 = 1.68.
1
plane stress pcf = 1 ; plane strain pcf =
1 4 + 4 2
xx xx
yy yy
y y
a ry r a ry r
rp
KI
=0 xx = yy =
2r
2
1 KI
yield xx = yy = y ry =
2 y
ry ry
KI 2KI
Z Z
1
y rp = yy (r) dr = r 2 dr = ry
0 2 0 2
2
2KI ry
1 KI
rp = rp = = 2 ry
2 y y
y
y
x
a
rp
singular term = 0 KI () = KI (y )
KI2
a
= cos rp =
a + rp 2y 8y2
ry |rp
criterion state ry or rp
(KI /y )2
KI 2
1
Von Mises plane stress 0.1592
2 y
KI 2
1
Von Mises plane strain 0.0177
18 y
1 KI 2
Tresca plane stress 0.1592
2 y
KI 2
1
Tresca plane strain 1 > 2 > 3 0.0177
18 y
The accuracy of the elastic stress field in SSY can be enhanced by using an effective
crack length, which takes the plastic zone into account, and is calculated by adding ry or rp
to a. Calculation of K has to be done iteratively, as the plastic zone size depends on K and
vice versa.
aef f = a + (ry |rp ) KI = I (aef f ) aef f
94
Chapter 10
When Small Scale Yielding cannot be assumed or when the material behavior is intrinsically
nonlinear, concepts from LEFM loose their meaning. For such cases NonLinear Fracture
Mechanics (NLFM) or Elasto-Plastic Fracture Mechanics provides new theories and concepts
to analye the behavior of cracks.
(1 + )( + 1) p
(x) = 2uy (x) = 2a(a x)
E
Crack Tip Opening Displacement (CTOD)
t = (x = a) = 0
Wells (1963) [68] suggested that this CTOD can be used in a crack growth criterion, when
plasticity at the crack tip is taken into account and the actual crack length is replaced by the
effective crack length.
95
96
xx xx
yy yy
y y
a ry r a ry r
rp
2
1 KI
aef f = a + ry = a +
2 y
When the effective crack length of Irwin is used, the crack tip opening can be calculated for
plane stress and for plane strain.
(1 + )( + 1)
q
(x) = 2aef f (aef f x)
E
(1 + )( + 1)
q
= 2(a + ry )(a + ry x)
E
(1 + )( + 1)
q
t = (x = a) = 2(a + ry )ry
E
(1 + )( + 1) q
= 2ary + 2ry2
E
(1 + )( + 1) p
2ary
E
4 KI2 4G
plane stress : t = =
Ey y
4(1 2 ) KI2
1
plane strain : t =
3 Ey
97
y y
x
a
rp
2
KI
aef f = a + rp = a +
8 y
The requirement that the singular term in the stresses at the crack tip must be zero, results
in a crack tip opening which is not zero. It can be calculated for plane stress and for plane
strain.
KI2 G
plane stress : t = =
Ey y
K2
1
plane strain : t = (1 2 ) I
2 Ey
98
t = tc (, T )
10.2 J-integral
In LEFM, crack growth can be predicted after calculation of the energy release rate G or
the stress intensity factor K. In NLFM, the behavior of a crack can be described with
the J-integral, which has been introduced by Rice in fracture mechanics in 1968 [54]. The
J-integral is a vector, J~, with three components in a three-dimensional (here Cartesian)
coordinate system. Its value results from an integration along a trajectory . In each point
of this trajectory, the specific elastic energy must be calculated from stresses and strains.
Also the unit normal vector ~n in that point, the stress vector ~t and the spatial derivatives of
the displacement ~u play a role.
What is generally called the J-integral in fracture mechanics, is the first component of
~ so the component in x1 -direction.
the vector J,
~n
~t
x2
V
S
~e2
positive
~e1 x1
Fig. 10.3 : Closed curve in a plane with unit normal vector and stress vector.
Z Epq
ui
Z
Jk = W n k ti d ; W = ij dij
xk 0
Z
ui N
J = J1 = W n 1 ti d
x1 m
99
Z
Jk = W jk ij ui,k nj d
W
homogeneous hyper-elastic mn =
mn
linear strain mn = 21 (um,n + un,m )
x2
~n
~e2
+ ~n
~e1 x1
B
A
Z Z Z Z
fk d + fk d + fk d + fk d = 0
A B +
Z Z Z Z
f1 d + f1 d + f1 d + f1 d = 0
A B +
10.2.3 Relation J K
In LEFM, the stress and displacement components at the crack tip are known as a function
of the position relative to the crack tip. For multi-mode loading, they are characterized by
the stress intensity factors KI , KII and KIII .
Because the J-integral is path-independent, the integration path can be chosen to be
a circle with the crack tip as its center. Integration over this circular path reveals that the
J-integral is related to the stress intensity factors.
Mode I + II + III
1
ij = [KI fIij + KII fIIij + KIII fIIIij ]
2r
ui = uIi + uIIi + uIIIi
( + 1)(1 + ) (1 + ) 2
J1 = KI2 + KII2
+ KIII
4E E
( + 1)(1 + )
J2 = KI KII
2E
For Mode I loading of the crack, it follows immediately that the J-integral is equivalent to
the energy release rate G. This means that the J-integral can be used in the crack growth
criteria of LEFM as a replacement for K and G.
( + 1)(1 + ) 2
Mode I J1 = J = KI = G
4E
3 1+ 4 1 2
plane stress +1 = + = J= K
1+ 1+ 1+ E I
(1 2 ) 2
plane strain + 1 = 4 4 J= KI
E
Obviously, J-integral is now also related to the crack tip opening displacement t .
plane stress Irwin J = y t
4
Dugbale J = y t
plane strain Irwin J= 3 y t
4
Dugbale J = 2y t
J = m y t
a u
m = 0.111 + 0.817 + 1.36
W y
n
= +
y0 y0 y0
4 n=3
3
/y0
n=5
2 n=7
n = 13
0
0 1 2 3 4 5 6
/y0
10.3.2 HRR-solution
The HRR-solution comprises a parameter , which is directly related to the J-integral. This
means that the stress and deformation state at the crack tip is completely characterized by
J, just as it is by K in LEFM. The parameter also depends on a constant In , which value
is determined numerically as a function of the exponent n, both for plane stress and plane
strain.
The solution reveals that the dependency of the distance to the crack tip is determined
by the exponent n. For n = 1 the singularity of the stresses is 1/ r, the same as we had in
LEFM. For n > 1 the singularity is smaller.
1 1
n+1
ij = y0 r ij () ; ui = y0 n r n+1 ui ()
1
J n+1
with : = (In from num. anal.)
y0 y0 In
103
In
5 plane strain
plane stress
2.5
0
5 10 15
n
Fig. 10.6 : Value of the variable In for plane strain and plane stress.
J = Jc
104
Chapter 11
In the crack growth criteria of LEFM and NLFM, a crack growth parameter has to be calcu-
lated and compared to a critical value, which is known from experiments. Crack parameters,
G, K, t and J, can be calculated analytically, but this is only possible with great mathemat-
ical effort and even then for rather simple cases. When such solutions exist, they can often
be found in literature. Using numerical methods, it is possible to calculate their values rather
straightforwardly, also for practical problems.
Finite element programs are widely available and can be used for the numerical calcula-
tion of crack growth parameters. Although special procedures are developed and implemented,
the use of standard elements is possible after a minor adaptation of the element mesh. Cal-
culation of the J-integral is often implemented as a standard option.
With the finite element method it is also possible to simulate crack propagation. Several
techniques are used and new ones are still under development.
Besides the finite element method, the boundary element method is used in numerical
fracture mechanics. Commercial programs for BEM are available although not so widespread
as for FEM.
For two-dimensional finite element analyses, the 8-node plane stress or plane strain element
gives accurate results for most mechanical problems. The displacement of an internal element
point is interpolated quadratically between the nodal displacements. Although initially the
element edges are straight lines, after deformation they may become of parabolic shape.
Interpolation or shape functions are used for this purpose, which are a function of two
local so-called isoparametric coordinates 1 and 2 , which have values between -1 and +1.
The 1 , 2 -coordinate system is not orthonormal, because the axes are not perpendicular.
Strains, which are derivatives of the displacements, are more or less linear in the local
coordinates and this is also the case for the stresses, which are proportional to the strains. It
has to be said that these statements are not valid when non-linear strain-displacement and
stress-strain relations are used.
105
106
2
2
3
7
1 4
1
6
8
1 5 2
3
3 7
7 4
4 6
6 8
8 2
5
1 5 2 1
4 7 3
3 3 3
6 5 6
8 6 7
1 5 2 1 4 2 4 5 2 4 2
8
1 1
p 3p p 3p
The crack tip mesh is often made as a crack tip rozet, with quarter point elements at the
crack tip. Lim advised in 1993 [43], to use transition elements between the quarter point
elements and the regular mid point elements for better results.
= 1 =0 =1
1 3 2
= 1 =0 =1
1 3 2
dx
x = 21 ( + 1)L ( 2 1) 12 L = 21 ( + 1)L = 21 L
d
du
du d du 2 3 1
= 1 x=0
= 2 u1 + 2 u2 + 2u3
dx 2L
dx =1
L
= 1 =0 =1
1 3 2
r
1 2 2 4x
x= 2 ( + 1)L ( 1) 41 L = 1
4 ( + 1) L +1=
L
dx
= 12 ( + 1)L = xL
d
du
du d du 1
= x=0
= singularity
dx xL dx =1 x
110
u u
a a + a
I II
dUi C
BG= = 12 uT u with C = C(a + a) C(a)
da a
Calculation of the stress intensity factor is also possible through its definition, which relates it
to the stress field. Stress components have to be calculated accurately upon which K results
from extrapolation to the crack tip. Because stresses are calculated in integration points and
the crack tip is located in an element node, this calculation is not exactly according to the
definition of K. It always remains to be investigated how the interpolation has to be done to
get the best results.
KI = lim 2r 22 |=0 ; KII = lim 2r 12 |=0
r 0 r 0
r K
p4
p3 Kp1
p2 Kp2
p1 Kp3
Kp4
r1 r2 r3 r4 r
Stresses in element integration points are calculated by differentiation, which makes them less
accurate then the calculated nodal displacements. It seems logical therefore to calculate the
stress intensity factor not from the stress field but from the displacement field.
" r #
4(1 2 )
r
r E 2
uy = KI gij () KI = lim uy ( = 0)
E 2 r 0 4(1 2 ) r
112
11.7 J-integral
The J-integral can be calculated according to its definition, i.e. by integration along a path
around the crack tip. Because stresses and strains are known in integration points, the
integration path must preferably pass through these points. In a general crack tip mesh such
a path would be difficult to define during post-processing of the results.
In a regular crack tip mesh, an integration path could be defined as indicated in the figure.
It consists of lines in coordinate directions. When symmetry can be considered, only half of
the path has to be taken into account. When such a path is taken at a certain distance from
the crack tip, there is no need to use quarter point crack tip elements.
x2
~n
~e2 q=0
+ ~n
~e1 x1
B
A q=1
q ui
Z
J= ij W 1j d
xj x1
q ui
Z
J= i1 W 1j d
xj x1
ui ui oij
Z Z Z
qpi d q(qi ui ) d + qij d
x1 x1 x1
s
this CTM can follow the crack tip. The example shows the crack propagation in the interface
between fiber and matrix for a plane strain simulation in the transversal plane, where the
loading is in the vertical direction. The crack opening is shown with enlarged displacements.
It was possible to simulate the crack propagation, which initially was in the fiber-matrix
interface but from some point moved into the matrix material. The CTM was moved by hand
for each crack advancement, which was a very laborious operation.
Fig. 11.16 : Crack tip mesh moves along fiber-matrix interface [42].
Instead of advancing a crack along element edges, propagation through elements is also an
option. It implies the subdivision of an element in two or more smaller elements, which
changes the mesh topology. Results are more accurate than those of the node release method.
Mesh refinement can be realized in the region of the crack tip.
116
The smeared crack approach can be seen as an intermediate between fracture mechanics and
continuum damage mechanics. Attention is not focused on one large crack, but the material
damage is modeled with small cracks in element integration points. These cracks may initiate
and lengthen according to certain loading criteria. The cracks are not really modeled, but
simulated with anisotropic material behavior, where in one direction the material stiffness is
reduced.
117
~e2
~e1
~e2
~n2
2
1 ~n1
~e1
~n2
~n1
Fatigue
Many structures are subjected to cyclic loading, with an amplitude so low that failure due to
yielding or crack growth, is never to be expected. However, experience tells a different story.
After a large number of load reversals, structural parts fail unexpectedly, often leading to
dramatic and overall collapse. This very dangerous phenomenon is the result of fatigue crack
growth.
Fatigue failure became an issue when carriages and trains with iron or steel axles began
traveling frequently over long distances in the mid 19th century. Axles broke mostly at
diameter-jumps and this was explained first as the result of changes in the materials structure.
Wohler was among the first to investigate these phenomena systematically, all of course
experimentally. Over many years, a tremendous amount of data was gathered, but theoretical
explanation had to wait until Griffith saw the importance of cracks for the failure of materials.
Cyclic loading of structures does not only occur as the result of deliberately applied
load reversals. Vibrations (machine(s) (parts)), repeated pressurization en depressurization
(airplanes), thermal cycling (switching of electronic devices), random forces (ships, vehicles,
planes) occur always and everywhere.
When fatigue failure has occurred, the crack surface shows two distinct markings on two
different scales.
At a macroscopic scale, so-called clam shell markings also called beach marks can be
seen. They are the result of irregularities in the growth of the fatigue crack, due to changes
in loading conditions.
When the crack surface is observed with a microscope, the crack displacement during each
individual load cycle is seen as a pattern of striations. During one cycle, plastic deformation
at the crack tip results in sequential blunting and sharpening of the crack tip. This crack
growth is very regular.
119
120
Fig. 12.1 : Clam shell fatigue crack surface and striations [29].
12.2 Experiments
Experiments can be done on full-scale structures subjected to real-life loading. This, however,
is often not feasible, due to costs and the extremely long times involved before the detection
of fatigue damage. Experiments are therefore mostly done on laboratory scale, using small
test specimens subjected to high loads varying at high frequencies. Results must be carefully
evaluated, because size effects and up-scaling of load amplitude and frequencies may lead to
inconsistencies with real-life applications.
max
m
min
0 t
0 N
i i+1
= max min ; a = 12
a 1R
m = 12 (max + min ) ; R = min /max ; =
m 1+R
th
12.3.2 (S-N)-curve
The experimental data are represented as a (S-N)- or Wohler curve, where the maximum
stress S = max is indicated along the vertical axis and the logarithm of the number of cycles
to failure or fatigue life Nf along the horizontal axis. Determination and representation of
these data is prescribed in normalization sheet B.S. 3518 part I 1984.
These curves are mostly recorded for zero average stress. For materials with a fatigue
limit, the (S-N)-curve will advance towards a horizontal asymptote at the level = th . When
a fatigue limit does not exist, the fatigue strength or endurance limit is defined as the stress
variation for failure after a specified high typically 106 number of cycles. The ratio of
fatigue limit or endurance limit to tensile strength is referred to as endurance ratio.
122
th
0
0 log(Nf )
Fatigue data may also be presented in a (Sa -N)-curve, where the vertical axis indicates the
stress amplitude : Sa = a = 12 .
For some real materials the (S-N)-curves are shown in the figure.
600
550
500
steelT1
450
400
max [MPa]
350
300
250
steel1020
200
150
Mgalloy Al2024T4
100
4 5 6 7 8 9
10 10 10 10 10 10
N
f
When the average stress is positive, a crack will grow faster and the curve will be located at
lower stress values, indicating that the fatigue life Nf is reached earlier for the same stress
amplitude a . It can also be stated that for the same Nf , a is reduced to a .
123
th
0
0 log(Nf )
Fig. 12.6 : Influence of the average stress on the Sa -N-curve.
Corrections have been proposed to take into account the decrease in a due to m 6= 0 and to
relate it to the tensile strength u or the initial yield stress y0 . Much used are the emperical
formulas of Gerber, Goodman and Soderberg.
a m 2
Gerber (1874) =1
a u
a m
Goodman (1899) =1
a u
a m
Soderberg (1939) =1
a y0
12.3.4 (P-S-N)-curve
Variations in experimental fatigue data can be represented in a (P-S-N)-curve, where proba-
bility (P) of failure is incorporated. The center curve represents the stress for which 50 % of
all specimens will fail after the associated Nf . When the upper curve is reached, 99 % of all
specimens will have failed. The lower curve represents the weakest specimens. For the given
stress, 1 % of all specimens fail for Nf on this curve.
550
500
450
400
350
max [MPa]
300
90% prob.failure
250
50% prob.failure
200
150
10% prob.failure
100
4 5 6 7 8
10 10 10 10 10
N
f
Sa
m = 0
0
0 4 5
log(Nf )
LCF HCF
1
2 = a = f (2Nf )b Nfb = constant
log 2
log(2Nf )
1
2
p
= f (2Nf )c p Nfc = constant
p
log 2
log(2Nf )
log(
2 )
log(Nf )
e p 1
= + = f (2Nf )b + f (2Nf )c
2 2 2 E
126
1
th (notched) = th (unnotched) ; 1 < Kf < Kt
Kf
127
Kf = 1 + q(Kt 1)
1
Peterson : q= with a = material parameter
1 + a
1
Neuber : q= q with b = grain size parameter
1 + b
10m
12.4.6 Environment
Environmental influences on fatigue life are mostly due to temperature effects and chemical
attack.
Low temperatures lead to embrittlement, due to the prevention of plastic deformation.
High temperature (T > 0.5Tm ) facilitates the creep deformation due to diffusion of atoms to
and from grain boundaries, dislocation movement and migration of vacancies. Weakening of
the grain boundaries leads to loss of inter-granular adhesion. Fatigue cracks will travel more
easily between grains or cross grain boundaries. This phenomenon is referred to as creep
fatigue.
Grain boundaries may also be weakened by chemical influence. Hydrogen and other
gasses can diffuse to the grain boundaries and diminish bond strength. Protective surface
films may be cracked. This process is called corrosion fatigue.
fatigue life, where a relative short number of cycles will result in fast crack growth leading
to failure. The initial fatigue crack length ai seems to be a very important parameter for the
fatigue life Nf .
For an initially undamaged material, it takes Ni cycles to initiate a crack by dislocation
movement and void coalescence. At this fatigue crack initiation life the initial crack has been
formed, but in most cases it is so small that it cannot be detected. In this stage I, crack
growth is provoked by shear stresses and involves slip on a single crystallographic slip plane.
The crack propagation rate is very low, typically < 0.25 nm/cycle.
After Ni cycles, in stage II of crack growth, crack propagation is faster, typically ms
per cycle. The crack growth is provoked by tensile stresses and involves plastic slip on multiple
slip planes at the crack tip, resulting in striations. After a large number of cycles the crack
reaches a length a1 , which can be detected by non-destructive techniques. The crack growth
is now much faster and after the fatigue life Nf its length is af and after a few cycles ac the
critical crack length is reached and failure occurs. For higher loading amplitudes, the crack
growth will be faster.
After N cycles, the cycles to go until failure at Nf , is indicated as Nr . The rest-life is
the ratio of Nr and Nf .
a I II III
ac ac
af
a1
ai
Ni Nf N
Nr N
=1
Nf Nf
da
crack length a. This is generally written as a relation between dN and K, where scaling
da
with E improves the generality of the fitted results. However, in LEFM, dN is mostly related
to K by the so-called Paris law, using two parameters : a coefficient and an exponent,
da
= C(K)m
dN
da
log dN
[mm/c]
-1
-2
-3
-4
-5
0 1 2 3 4 log(K)
-6
-7 [MPa m]
-8
-9 log C = 8.7
da da
= C(K)m log = log(C) + m log(K)
dN dN
da [mm]
log(K) = 0 log(C) = log = 8.7 C = 2 109 m
dN [MPa m]
(2) (4)
m= =4
(2) (1.5)
da
log( dN )
R
rapid crack growth
power law growth
Kth Kc log(K)
slow crack growth
Conversion
Because the dimension of C depends on the value of m, the conversion of units must be done
with care. The example shows the conversion from BS to SI units and vice versa.
da
da
= C ( a)m C= dN
dN ( a)m
[in] and [ksi] [m] and [MPa]
[ in ] 0.0254 [ m ] 0.0254 [m]
1 = =
(1.09)m [ MPa m ]m
p
[ ksi in ]m {6.86 [ MPa ] 0.0254 [ m ] }m
[m] and [MPa] [mm] and [MPa]
!
[m] 103 [ mm ] 103 [ mm ]
1 m = =
[ MPa m ] {[ MPa ] 103 [ mm ]}m { 103 }m [ MPa mm ]m
132
" (1 m ) #
()m
(1 m ) ai 2
Nf Ni = m m m af 2 1
C( ) (1 2) af
Paris law can also be integrated numerically, here, using an explicit integration scheme. The
initial crack length is indicated to be a0 and the final crack length is the critical crack length
ac .
set , N, ac
initialize N = 0, a = a0
while a < ac
K = a
da da
= C (K)m a = N
dN dN
a = a + a
N = N + N
end
80 80
a [mm]
a [mm]
60 60
40 40
20 20
0 0
0 1 2 3 4 5 0 2 4 6 8 10 12 14
N [c] 6 N [c] 6
x 10 x 10
aluminum ; = 50 [MPa] mild steel ; = 50 [MPa]
Fig. 12.18 : Fatigue crack length in aluminum and mild steel for two values of the initial
crack length.
133
Fatigue load
The fatigue life is reached, when the crack length becomes critical (a = ac = 2 E
2 ). For
aluminum, the number of cycles to reach this length has been calculated for various stress
amplitudes.
[MPa] 25 50 75 100
a0 [mm] 0.1 0.1 0.1 0.1
ac [mm] 56 28 12.5 7
Nf [c] 35070000 4610000 1366000 572000
60
50
40
a [mm]
30
20
10
0
0 1 2 3 4
N [c] 7
x 10
Fig. 12.19 : Fatigue crack length in aluminum for various load amplitudes.
Erdogan (1963)
da C(K)n Kmin
= with R=
dN (1 R)Kc K Kmax
Donahue (1972)
da
= C(K Kth )m with Kth = (1 R) Kth (R = 0)
dN
134
Walker (1970)
m
da K
=C with m = 0.4 ; n = 0.5
dN (1 R)n
Priddle (1976)
K Kth m
da
=C
dN KIc Kmax
with Kth = A(1 R) and 1
2 1 [Schijve (1979)]
slipline
3 sin2 () cos2 ( 2 ) K 2
da 1 Kmax
= 1 2
dN 9 sin() Ev {1 (1 ) max
v }
= cos1 13
= 0.5 + 0.1R + 0.4R2
da 7 K 2
Kmax
= (1 0.2R 0.8R2 )
dN 64 2 Ev {1 (1 ) max
v }
For high values of crack tip stress and consequently a large plastic zone, the Paris law may
be used with J instead of K. Note that J 6= Jmax Jmin over a cycle.
da
= C ( J)m
dN
Z pq
Zmax
ui
with J = W n1 ti d ; W = ij dij
x1
pqmin
Until now, we only looked at harmonic loads with constant frequency and amplitude. In prac-
tical situations, such loads will almost never be seen. Both loading frequency and amplitude
will generally vary over time. Cyclic loading with such a time dependent character is referred
to as a load spectrum. It is assumed here that the direction of the load remains the same.
A load spectrum is shown in the figure below. It consists of four (i = 1..4) harmonic
loads with different frequencies and amplitudes. In the subsequent loads the number of cycles
is ni . For each individual harmonic load the life time Nif is known.
Miner has found in 1945 that the fatigue life for the whole spectrum is independent on
the actual sequence of the individual loadings [47]. This means that the individual harmonic
loadings do not influence the damage growth in the following loading stage with a different
character. The final life time is reached, when the summation of Nnifi becomes one. This
so-called Miners rule is approximately valid for a wide range of materials. In reality the
ultimate sum of all the ratios is between 0.6 and 2.0, which indicates that there is actual
some interaction between the different load cycles.
136
0
N
n1 n2 n3 n4
L
X ni
=1
Nif
i=1
Miners rule is clearly illustrated by regarding the rest life after each load cycle. When it
reaches zero, the summation of the ratio Nnifi must reach the value one.
n1
1 1
N1f
n1 n2
2 1
N1f N2f
n1 n2 n3
3 1
N1f N2f N3f
n1 n2 n3 n4
4 1 =0
N1f N2f N3f N4f
0
t
For some applications, random load sequences are measured and used in design. During
service life the load can be monitored and the expected fatigue life time can be updated.
Typical loadings for a manufacturing machine (a), a sea vessel (b) and an airplane (c) are
shown in the figure.
138
Fig. 12.23 : Random loads for machine (a), sea vessel (b) and airplane (c).
Known random loadings are used to calculate the fatigue life during the design of structures.
During service life extreme overloads can occur, which must of course be low enough to cause
no direct damage. The question is how they affect the fatigue life.
Although the intuitive feeling is that overload must be bad, practical experience shows
that the crack propagation is slowed down by such an overload, at least during a limited
number of cycles. This is illustrated in the next figure, where (Kmax )OL is the maximum SIF
during the overload.
139
(Kmax )OL
Kmax
a
b2
b1
The crack retardation is also illustrated quantitatively by the values in the next table (from
[29]), which apply to the Mode I loading of specimen made of Al 2024-T3 alloy.
The crack retardation can be explained by regarding the stress state in a region in front of
the crack tip. The figure shows the stress yy and the plastic limit stress. In a certain time
period, the loading stress is increased to a maximum value 1 , representing the overload, and
140
decreased again. In two points A and B the stress yy is plotted in the figure as a function of
the strain yy . In point A, the material is plastically deformed, while the deformation in point
B remains elastic. After unloading, the strain in A and B is more or less the same, so that it
becomes clear from the stress-strain curves that a compressive residual stress remains in point
A. This compressive stress inhibits the opening of the crack and thus the crack elongation.
When the crack tip has moved through the compressive zone, the normal crack grow rate is
again established.
1
yy
v
0
t
A B
yy
A1 yy
B1
=0
B2
yy
A2
Fig. 12.25 : Residual stress at the crack tip resulting from overload.
Willenborg (1971)
" "s # #
a
KR = (Kmax )OL 1 Kmax ; a < ry
ry
KR = residual SIF ; KR = 0 delay distance
1
= [1 (Kth /Kmax )](S 1) ; S = shut-off ratio
(Kmax )OL
Kmax
a a
ry
Johnson (1981)
1 (Kmax )OL 2
ef f Kmin KR
R = ; ry =
Kmax KR v
= plastic constraint factor
Crack retardation due to crack closure, caused by the compressive stress zone behind the
crack tip, have also been devised and published.
Schijve (1981) [58] U = 0.55 + 0.33R + 0.12R2 with 1.0 < R < 0.54
Polymer materials are widely used in functional and structural applications. External loads
and other influences lead of course to damage, cracking and failure. Systematic study of these
phenomena has brought insight in their behavior and, even more than in the case of metal
alloys, has given rise to the development of new materials with better damage resistance.
There are a huge number of different polymer materials, some of which are listed in the
table.
ABS acrylonitrilbutadieenstyreen EM
HIPS high-impact polystyreen TP
LDPE low-density polyetheen TP
Nylon
PC polycarbonaat TP
PMMA polymethylmethacrylaat (plexiglas) TP
PP polypropeen TP
PPO polyfenyleenoxyde TP
PS polystyreen TP
PSF PolySulFon TP
PTFE polytetrafluoretheen (teflon) TP
PVC polyvinylchloride TP
PVF polyvinylfluoride TP
PVF2 polyvinylidieenfluoride TP
143
144
13.1.1 Damage
Damage in polymers may occur as shearbands or crazes. These can occur both in the same
material in regions with different stress states.
Shear bands are associated to shear yielding. In this case no change in density is observed.
Crazes are connected to normal yielding. The local density may decrease with 40 - 60 %. In
a craze, fibrils are connecting the two craze surfaces. When the craze opens, the fibrils are
elongated due to extrusion of macromolecules from the bulk polymer. When the maximum
elongation is reached, the fibril snaps.
The next table shows some properties of engeneering plastics, which are related to their
damage behavior. The crystallinity is a very important item, because crystals can stop cracks
in an early stage. The occurence of crazes is also relevant, because crazing is generally
associated to brittle behavior.
145
AC CZ KIc
PMMA a + 13.2
PS a + 17.6
PSF a - low
PC a - high main chain segmental motions
energy dissipation
Nylon 66 cr - main chain segmental motions
energy dissipation
crystalline regions crack retardation
PVF2 sc
PET sc amorphous
strain induced crystallization at crack tip
CPLS - cross-linked suppressed crazing
HIPS + -sized rubber spheres
enhanced crazing
ABS blending
AC : a = amorphous
AC : c = crystalline
AC : sc = semicrystalline
AC : cr = crystalline regions
CZ = crazing
KIc = fracture toughness in MPa m
13.2 Fatigue
Fatigue crack propagation in polymers is strongly affected by properties like molecular weight,
cross linking, main chain mobility and crystallinity. Blending one polymer with another
polymer or particles to improve ductility and fatigue life is referred to as toughening. Data
for the next figures are taken from [29].
146
10 2
da
4
dN
[mm/cycle]
3 5
3 103
2
1
103
3 104
104
0.5 1 2 5
K [MPa m]
1 : PMMA 5 Hz crazing
2 : LDPE 1 Hz
3 : ABS 10 Hz
4 : PC 10 Hz no crazing
5 : Nylon 10 Hz cristalline regions
101
da
4 5
dN 3
[mm/cycle] 2
102 6
7
103
8
1
104 9
105
1 : PS 5 : PC
2 : PMMA 6 : Nylon 6.6
3 : PSF 7 : PVF2
4 : PPO 8 : PET
9 : PVC
147
da
dN
[mm/cycle]
2
3
102 1
4
103
104
Cohesive zones
With ongoing miniaturization, interfaces get more important concerning the (mechanical)
behavior of material systems. The degradation of interfaces e.g. delamination can be
studied and analyzed with cohesive zone models.
These models are used to investigate the behavior of interfaces between polymer coating
and steel sheet substrate, a material system used more and more for all kind of applications.
Also the peel-off lid of food containers is relying on the proper delamination of a polymer
interface layer. As a final example, the interfaces in solder joints for microelectronic compo-
nents are studied, with special attention for their fatigue life.
The next sections are based on the research work of Marco van den Bosch (PhD), Sil
Bijker (MSc) and Muge Erinc (PhD), whos papers and reports are listed in the reference
section [].
149
150
traction = f(opening) T = f ()
normal / tangential Tn = fn (n ) ; Tt = ft (t )
coupling Tn = fn (n , t ) ; Tt = ft (t , n )
A large variety of cohesive zone laws has been described in literature. Most of them can be
categorized into the following groups: polynomial, piece-wise linear, rigid-linear and expo-
nential. These cohesive zone laws are described in the next sections. The maximum normal
traction and the maximum tangential traction are indicated by Tn,max and Tt,max , respec-
tively. Characteristic openings are n and t . The areas below the traction-opening curves
represent the normal and tangential work-of-separation Wn and Wt .
14.1.1 Polynomial
In the polynomial cohesive zone law, coupling is achieved with an effective opening displace-
ment and a coupling parameter , which relates the maximum tractions. The function f ()
is such that the tractions are non-zero for 0 < 1 and remain zero for > 1. The figure
below illustrates this cohesive zone behavior with the function f () taken from [64]. The
coupling is incorporated well. The plots show that the tractions in one direction deminish,
when the cohesive zone is opened in the other direction. The largest tractions are associated
with the opening in one direction or, which is the same, with an uncoupled formulation. Note
that in this cohesive zone n and t indicate the opening, when the traction is zero.
n
Tn = Tn,max f (; 0 < 1)
n
t t
Tt = Tt,max f (; 0 < 1) = Tn,max f (; 0 < 1)
t t
s
n 2
2
t
effective displacement = +
n t
27
1 2 + 2
function from [64] f () =
4
151
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
Tt,max = (n /t )Tn,max
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
14.1.3 Rigid-linear
Cohesive zones may be modelled to have a very high stiffness, until the maximum traction
is reached. In such a rigid-linear cohesive zone law, as proposed in [12], the normal traction
is not a function of the tangential opening t , while the tangential traction is controlled
completely by the normal opening n . This implies that the interface can only fail due
to decohesion in normal direction. The plot shows a rigid-linear law without this coupling
( = 1). The characteristic openings n and t are the openings at zero traction.
152
n
Tn = Tn,max 1 n > 0
n
t n
= coupling
t
t n
Tt = Tt,max 1 t > 0
t
=1 no coupling
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
14.1.4 Exponential
The exponential cohesive zone law is used very much. It has some advantages compared to
other laws. First of all, a phenomenological description of contact is automatically achieved
in normal compression. Secondly, the tractions and their derivatives are continuous, which is
attractive from a computational point of view. The exponential cohesive zone law originates
from the universal relationship between binding energies and atomic separation of interfaces.
Two exponential cz-laws are described in the following, both based on a potential ,
which is a function of both the normal and the tangential opening. The first law, proposed by
Xu & Needleman is much used, but appears to have some deficiancies in describing coupling.
n
(n , t ) = n + n exp
n
2t
n 1q r q n
1r+ q+ exp 2
n r1 r 1 n t
Tractions in normal (Tn ) and tangential (Tt ) direction are derived from the potential by
differentiation w.r.t. n and t respectively.
153
2t 2t
n n n 1q n
Tn = = exp exp 2 + 1 exp 2 r
n n n n t r1 t n
n t
r q n
n
2
Tt = =2 q+ exp exp 2t
t t t r 1 n n t
For pure normal and tangential loading, the traction-opening relations are simplified. In-
troducing the parameter t , the coupling parameter q is defined as q = t /n . The nor-
mal and tangential traction-opening relations are plotted for the parameter values n =
100 Jm2 , t = 80 Jm2 , Tn,max = 10 N , Tt,max = 10 N .
n n n
Tn = Tn (t = 0) = exp
n n n
2t 2t
n t t t
Tt = Tt (n = 0) = 2 q exp 2 = 2 exp 2
t t t t t t
n n
Tn,max = Tn (n = n ) = n =
n exp(1) Tn,max exp(1)
t
q
1
Tt,max = Tt (t = 2 t ) = q t = qt
1
t 2 exp(1) Tt,max 12 exp(1)
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
Fig. 14.5 : Uncoupled normal and tangential traction-opening for the Xu&Needleman
cohesive zone law.
When the cohesive zone is loaded simultaneously (or subsequently) in normal and tangential
direction, the coupling is taken into account. The results are dependent of the value of r. For
r = 0 the normal traction does not become zero, when the interface is failed in tangential
direction. For r = q and other values of r, the results are even more unrealistic, as the normal
traction is not zero for zero normal opening.
154
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
Fig. 14.6 : Coupled normal and tangential traction-opening for the Xu&Needleman
cohesive zone law for r = 0.
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
Fig. 14.7 : Coupled normal and tangential traction-opening for the Xu&Needleman
cohesive zone law for r = q.
2t
n n
(n , t ) = n|t 1 1 + exp exp 2
n n t
2t
n n n
Tn = = exp exp
n n n n n t 2
2t
t t n n
Tt = =2 1+ exp 2 exp
t t t t n t n
155
n
n = ; t = qt
Tn,max exp(1) Tt,max 1
exp(1)
2
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
Fig. 14.8 : left: The uncoupled normal response of the exponential cohesive zone law with
the maximum traction Tn,max at n /n = 1. right: The uncoupled tangential response of
the exponential cohesive zone law with the maximum traction Tt,max at t /t = 1/ 2.
Coupling
Adequate coupling between the normal and tangential directions is required in a cohesive
zone law to describe the physically occurring interface behavior realistically. If complete
loss of interfacial integrity is important (e.g. in the case of a moving delamination front)
and a cohesive zone completely fails in shear, its load-carrying capacity in normal traction
should completely vanish as well and vise versa. That this behavior is described adequately,
is illustrated by the traction-opening plots below.
1 1
0.5 0.5
Tn /Tn,max
Tt /Tt,max
0 0
0.5 0.5
1 1
1 0 1 2 3 4 2 1 0 1 2
n /n t /t
Fig. 14.9 : The coupled normal and tangential response of the exponential cohesive zone
law.
156
To investigate this behavior in more detail, the interface is first loaded in normal direction
until a maximum opening n,max . After that it is broken in shear : t , as is illustrated
in the figure below. The shaded areas under the curves are the work of separation for normal
and tangential opening.
Z
n,max Z
Wn = Tn (n )|t =0 dn ; Wt = Tt (t )|n =n,max dt
0 0
Several loading sequences are evaluated. In the left figure below the evolution of the maximum
shear traction is shown as a function of the normal separation. As can be seen in this figure, the
maximum shear traction decreases to zero for increasing normal separations. Theoretically,
this can also happen in compression. However, these negative values of n cannot be reached
in practice since the normal compressive traction increases very fast in this regime.
In the right figure below the maximum normal traction as a function of the tangential
separation is shown. The maximum normal traction decreases to zero for increasing tangential
separation.
0.8 0.8
/Tn,max
Tt /Tt,max
0.6 0.6
Tn
0.4 0.4
0.2 0.2
0 0
1 0 1 2 3 4 2 1 0 1 2
n,max /n t,max /t
Fig. 14.11 : left: The normalized maximum shear traction as a function of the normalized
normal separation of the alternative exponential law.
right: The normalized maximum normal traction as a function of the normalized tangential
separation of the alternative exponential law.
157
Coupling should also be realistic, when considering the total work-of-separation Wtot = Wn +
Wt in case of a sequential loading, in normal and tangential direction or vice-versa..
The behavior is as expected and shown in the figures below : in the case where there is
first normal separation until n,max and then complete tangential separation, Wtot increases
monotonically from the value t to the value of n .
In the other case where there is first tangential separation until t,max and then complete
normal separation, Wtot smoothly decreases from the value of n to t .
100 100
80 80
60 60
W[Jm ]
W[Jm ]
2
2
40 40
W W
n n
W W
20 t 20 t
W W
tot tot
0 0
0 2 4 6 8 0 2 4 6 8
n,max /n t,max /t
Fig. 14.12 : left: Work-of-separation when the cohesive zone is first loaded in normal
direction until n,max and then broken in shear. right: Work-of-separation when the
cohesive zone is first loaded in tangential direction until t,max and then broken in tension.
Mode-mixity
The influence of mixed mode loading on the total work-of-separation is studied next. To this
purpose, a cohesive zone is loaded under an angle w.r.t. the normal direction until complete
separation occurs, and the corresponding value of Wtot is calculated.
Wtot is shown as a function of the loading angle . Again, values n = 100 Jm2 and
t = 80 Jm2 are used in these calculations. In the figure below it can be seen that Wtot = n
when the cohesive zone is only loaded in normal direction (mode-I: = 0o ) and Wtot = t
when it is only loaded in tangential direction (mode-II: = 90o ).
q = ~v
~ + ~ ~x V
eq. of equilibrium ~ q = ~0
+ ~ ~x V
158
Analytical solutions are seldomly reached for practical problems and we have to be satisfied
with approximate solutions, which can be determined with the finite element method. To this
purpose the differential equation is transformed into a weighted residual integral by use of a
vectorial weighting function w(~
~ x). Partial integration of the first term leads to a weak version.
The two integrals with the external loads are abbreviated as the external load integral fe .
The remaining integral is the internal load integral fi .
Z h i
~ + ~
q = ~0 ~x V w
~ ~ + ~
q dV = 0 w(~ ~ x)
V
Z Z Z
~ w)
( ~ c : dV = w~ ~
q dV + ~ ~t dA w(~
w ~ x)
V V A
fi (w, ~ ~t, ~
~ ) = fe (w, q) w(~
~ x)
The internal load integral can be written as the sum of an integral over bulk material (volume
Vb ) and an integral over cohesive zones (volume Vcz ).
Z Z Z
fi = ~ w)
( ~ c : dV = ~ w)
( ~ c : dV + ~ w)
( ~ c : dV = fib + ficz
V Vb Vcz
When the weighting function w ~ is chosen to be a virtual displacement ~u, the weighted
residual integral becomes the virtual work equation in which the logarithmic strain ln is
introduced. We only consider the integral over the cohesive zones. Furthermore, we consider
a two-dimensional case, where the cz has uniform depth d. The unity vector ~e is defined along
the line between two associated points of the interface. The deformed opening of the cohesive
zone is h.
Z
ficz = ~ w)
( ~ c : dV
Vcz
w
~ = ~u ~ w)
( ~ c : = (~~ u)c : = ln :
Z Z
ficz = ln : dV = d ln : dA
Vcz Acz
ln = ln~e~e = ~ln~e
ln : = ~ln ~
= ~e~e = ~~e
~u (~u)
Z Z Z
ficz =d ~ln ~ dA = d ~ dA = d ~ dA
Acz Acz h Acz h
d
Z
= w ~ ~ dA
h Acz
159
When large deformations occur, the integral has to be transformed to a known state. In this
case we choose the initial undeformed state. In this transformation the deformation tensor F
is used. The transformed integral contains the first Piola-Kirchhoff stress tensor T .
Z
~ w)
( ~ c : dV = fe (w,
~ ~t, ~q) w(~
~ x)
V
~ = F c
~0 ~ w)
( ~ 0 w)
~ c = ( ~ c F 1
dV = det(F ) dV0 = J dV0
Z
~ 0 w)
( ~ c F 1 : J dV0 = fe0 w(~
~ x)
V0
Z
~ 0 w)
( ~ c : (F 1 J) dV0 = fe0 w(~
~ x)
V0
Z
~ 0 w)
( ~ c : T dV0 = fe0 w(~
~ x)
V0
fi0 (w, ~ ~t0 , ~q0 )
~ T ) = fe0 (w, w(~
~ x)
Again the internal load integral is written as sum of an integral over the bulk volume and an
integral over the cohesive zones.
Z Z Z
fi0 = ~ 0 w)
( ~ c : T dV0 = ~ 0 w)
( ~ c : T dV0 + ~ 0 w)
( ~ c : T dV0 = fi0b + fi0cz
V0 V0b V0cz
When the weighting function w ~ is chosen to be a virtual displacement ~u, the weighted
residual integral becomes the virtual work equation in which the linear strain is introduced.
We only consider the integral over the cohesive zones and take the cz to have a uniform initial
depth d0 . The unity vector ~e0 is defined along the line between two associated points of the
interface. The undeformed opening of the cohesive zone is h0 .
Z
fi0cz = ~ 0 w)
( ~ c : T dV0
V0cz
w
~ = ~u (~ 0 w)
~ c : T = (~ 0 ~u)c : T = l : T
Z Z
fi0cz = l : T dV0 = d0 l : T dA0
V0cz A0cz
l = l ~e0~e0 = ~l ~e0
l : T = ~l T~
T = T ~e0~e0 = T~ ~e0
~u ~ (~u) ~
Z Z Z
fi0cz = d0 ~l T~ dA0 = d0 T dA0 = d0 T dA0
A0cz A0cz h0 A0cz h0
d0
Z
= w ~ T~ dA0
h0 A0cz
160
3
Q
4 T~ B
1
A 0
-1
2
1 P
1 1
d0 d0 h0 l0
Z Z Z
fi0cz = ~ T~ dA0 =
w ~ T~
w dd
h0 A0cz h0 =1 =1 2 2
1 1 1
d0 l0 d0 l0
Z Z Z
= ~ T~ dd =
w ~ T~ d
w
2 =1 =1 2 2 =1
T~
4
T~ ~en B
A ~et
2
1
~et
w ~ = wt~et + wn~en = [wt wn ] = wT ~e
~en
~et
T~ = Tt~et + Tn~en = Tt Tn = T T ~e
~en
d0 l0 1
Z
fi0cz = wT ()T () d
2 =1
161
Interpolation
The weighting function components are interpolated between their values in point A and
point B. The interpolation functions are linear in the local coordinate.
1
2 (1 ) 0
1
0 2 (1 )
wT () = [wt wn ] == wtA wnA wtB wnB
1
(1 + )
2 0
1
0 2 (1 + )
T
= w AB N T ()
The values w in points A and B are written as the differences of the weighting function in
the nodal points 4 and 1 and 3 and 2 respectively. The index l indicates that components are
taken w.r.t. the local vector basis {~et , ~en }.
1
wt
w1
n
wtA 1 0 0 0 0 0 1 0 wt2
wA 0 1 0 0 0 0 0 1 wn2
AB n
w = = 3 = P wl
wtB 0 0 1 0 1 0 0 0 wt
wnB 0 0 0 1 0 1 0 0 wn
3
4
wt
wn4
The column with weighting functions is transformed from the local to the global coordinate
system, which is done with a rotation matrix R. It transforms the local base vectors {~et , ~en }
to the global base vectors {~ex , ~ey }. The only non-zero components in the rotation matrix
are cosine (c) and sine (s) functions of the angle between the line AB and the global x-axis.
The cosine and sine values can be easily calculated from the coordinates of points A and B :
c = (xB xA )/ and s = (y B y A )/, where is the deformed length of line AB. Coordinates
of A and B can be expressed in the nodal point coordinates : x|y A|B = 12 (x|y 1|2 + x|y 4|3 ).
wt1 c s 0 0 0 0 0 0 w11
wn1 s c 0 0 0 0 0 0 w21
wt2 w12
0 0 c s 0 0 0 0
wn2 w22
0 0 s c 0 0 0 0
wl = = = Rw
wt3 w13
0 0 0 0 c s 0 0
wn3
0 0 0 0 s c 0 0
w23
wt4 0 0 0 0 0 0 c s w14
wn4 0 0 0 0 0 0 s c w24
162
So we can write :
wT = wTl P T N T () = wT RT P T N T ()
and
1
d0 0 T T
Z
fi0cz = w T R P N T ()T () d = w T f i0 (R, T )
2 =1 cz
After assembling all elements, the global weighted residual equation has to be satisfied for all
nodal values of the weighting function. This means that the nodal point internal forces have
to be in equilibrium with the external nodal forces.
wT f i0 + w T f i0 = wT f e0 w
b cz
f i0 (, F ) + f i0 (R, T ) = f e0
b cz
T = T + T ; R = R + R R f i = f i + f i
1
d0 0 T T
Z
f i0 = R P N T () T () d
cz 2 =1
The relation between the iterative traction and the iterative opening represents the stiffness
of the cohesive zone.
Tt Tt
T t
t n
T () = =
= M () ()
Tn Tn n
t n
The iterative opening is now interpolated analoguously to the interpolation of the weighting
function. Also the rotation of the local to the global coordinate system is applied. The
iterative opening is expressed in the iterative nodal displacements in the global directions.
The iterative tractions are then expressed in the iterative nodal displacements. The iterative
internal load column can now be expressed in the iterative opening, using the stiffness matrix
K cz of the cohesive zone element.
163
Examples
The cohesive zone element is placed between two bulk elements, which have linear elastic
material behavior. The figure below shows the opening and sliding mode loading, where the
displacements of the upper nodes are prescribed. The plots show the reacton forces in x
and ydirection in the upper nodes. The forces are the same as the traction opening relaton
of the cohesive zone.
7 8
1 1
np7 np7
np8 np8
fy /Tn,max
fx /Tt,max
5 6
0.5 0.5
7 8
3 4 5 6
0 0
1 2
3 4
0.5 0.5
1 2
1 1
1 0 1 2 3 4 3 2 1 0 1 2 3
uy /n ux /t
In the next example the cohesive zone is opened in normal and tangential direction simul-
taneously without rotation. The plots show reaction forces in the upper nodes and the
traction-opening behavior in the integration points of the cohesive zone element. Due to the
coupling in the traction-opening law, the normal and tangential tractions do not reach their
maximum values.
7 8
2 1 1 0.7
np7 np7 ip1 ip1
np8 np8 ip2 ip2
0.6
Tn /Tn,max
0.8 0.8
fy /Tn,max
5 6
1.5
fx /Tt,max
Tt /Tt,max
0.5
1 0.6 0.6
0.4
0.5 0 0 0
0 1 2 3 4 5 0 0.5 1 1.5 2 2.5 0 1 2 3 4 5 0 0.5 1 1.5 2 2.5
uy /n ux /t n /n t /t
Fig. 14.14 : Mixed mode opening of a cohesive zone element without rotation.
The cohesive zone is now opened in normal direction with a different displacement in the upper
nodes, resulting in rotation. Reaction forces in x-direction are zero. Tangential tractions in
the integration points of the cz-element differ also.
8
7 1.4 1 1 0.25
np7 np7 ip1 ip1
6
1.2 np8 np8 ip2 ip2
Tn /Tn,max
0.8
fy /Tn,max
fx /Tt,max
0.2
Tt /Tt,max
1 0.5
5
0.2 0.5
1 2
0.2 0.05
0
0.2 1 0 0
0 0.5 1 1.5 2 2.5 0.5 0.4 0.3 0.2 0.1 0 0 0.5 1 1.5 2 2.5 0 0.05 0.1 0.15 0.2
uy /n ux /t n /n t /t
7 7
7 0.25 0.4 0.04
6 6
ip1 ip1 ip1
6
ip2 0.35 ip2 0.03 ip2
0.2
Tt /Tt,max
Tt /Tt,max
Tt /Tt,max
0.3 0.02
5 5 5
0.1 0.02
1 2 1 2 1 2
0.05
0.05 0.03
0 0 0.04
0 0.05 0.1 0.15 0.2 0 0.1 0.2 0.3 0.4 0.06 0.04 0.02 0 0.02 0.04
t /t t /t t /t
40
30
T [MPa]
20
10
0
0 1 2 3 4 5 6
[m]
~ = ~e ; T~ = T ()~e
; T~ = f~()
~ ; T = exp
Z
= T () d ; Tmax =
=0 exp(1)
165
Upon unloading, the cohesive zone law may show an irreversible response. Two types of
irreversible behavior are considered: linear elastic unloading to the origin, i.e. elasticity-
based damage (i) and unloading with the initial stiffness of the cohesive zone, i.e. plastically
damaged (ii). Both cases are shown in the figure below.
Fig. 14.18 : Large deformation cohesive zone response for elastic and history dependent
behavior.
Mode-mixity
The opening mode of the cohesive zone is quantified by a mode-mixity parameter d. In a
two-dimensional cohesive zone, d~1 and d~2 are the components of the normals ~n1 and ~n2 of the
two cohesive zone edges perpendicular to .~ Parameter d has a value between 0 (mode-I) and
2 (mode-II). Intermediate values of d represent a mixed-mode opening. The traction-opening
relation is extended with this mode-mixity d, where a parameter controls the influence of
mode-mixity behavior: the interface is stronger ( > 0) or weaker ( < 0) in mode-II than
in mode-I.
d = ||d~1 d~2 ||
d
T = exp exp
2
166
To investigate the mode-mixity behavior, a single cohesive zone is opened under an angle
and the total dissipated energy is quantified. The value of the potential was taken to be
= 100 Jm2 . The two faces are kept parallel to each other and is defined as the angle
between one of the faces and the vector . ~
The results are shown in the figure below. Obviously, the energy dissipated in mode-I
( = 90o ) is independent of and equals . In the case of mixed-mode or mode-II ( = 0o )
opening of the cohesive zone, the total dissipated energy depends on . In the right figure,
the traction is plotted for three values of in mode-II, where the total dissipated energy (area
under the curve) depends on the parameter .
50
120 = 0.2 = 0.2
= 0.0 = 0.0
40
110 = 0.2 = 0.2
W [Jm2]
30
T [MPa]
100
20
90
10
80
0 30 60 90 0
0 2 4 6 8 10
[o] [m]
14.5 Applications
Cohesive zones are used to study de-adhesion between a wide range of materials under a large
scope of loading conditions. Most applications comprise small deformations and quasi-static
loading.
Large deformations occur in the next two examples : the delamination of the polymer
coating on a steel sustrate, as is used in beverage cans, and, the opening of a food-can, where
the lid is opened by peeling off a sealing foil.
Repeated cyclic loading with small deformations can lead to fatigue damage and de-
adhesion between two entities. In electronic devises, the repeated switching causes cyclic
thermal loading and leads to de-adhesion between the solder and the connection pad. This
phenomenon is also studied with cohesive zone models, combined with a damage evolution
law.
can be predicted, the processing routes, parameters and tooling can be adjusted to prevent
it.
This research is described in detail in the PhD-thesis of M.J. van den Bosch [8] and
papers, which can be found in the References.
Thermal cycling due to repeated power switching evokes heat related phenomena: the
mismatch in the coefficient of thermal expansion between the package components causes
cyclic mechanical strains.
As a result of the multi-phase nature of Sn based solder alloys and the thermal anisotropy
of -Sn, internal stresses build up in the solder.
Bending of the board induces shear and tensile stresses in the solder joints.
The figure below shows a typical bump between two pads, where thermal loading of the
system leads to deformation and results in plastic strains. Cyclic loading leads to fatigue
damage and finally to failure. Instead of using phenomenological laws like Manson-Coffin,
the fatigue damage is assessed with a cohesive zone approach.
This research is described in detail in the MSc-thesis of M.E. Erinc [23] and related
papers, which can be found in the References.
Fig. 14.24 : Ball grid array (BGA) and FE-model of one solder ball with cohesive zones.
Bibliography
[1] Barenblatt, G.I. The mathematical theory of equilibrium cracks in brittle fracture. Ad-
vances in Applied Mechanics, Vol 7, 1962, pp 55-129.
[2] Barsoum, R.S. On the use of isoparametric Finite Elements in Linear Fracture Mechan-
ics. Int. J. Numerical Methods in Engineering, Vol 10, 1976.
[3] Bates, R.C.; Clark, W.G. Fractography and fracture mechanics. Trans. ATM, Vol 62,
1969, pp 380-388.
[4] Bijker, S.Study of the opening behavior of food containers with Easy Peel-Off Lid. MSc
Thesis, MT 08.11, Eindhoven University of Technology, 2008.
[5] Bosch van den, M.J.; Schreurs, P.J.G.; Geers, M.G.D.An improved description of the
exponential Xu and Needleman cohesive zone law for mixed-mode decohesion. Engineering
Fracture Mechanics, Vol 73, 2006, pp 1220-1234.
[6] Bosch van den, M.J.; Schreurs, P.J.G.; Geers, M.G.D.A cohesive zone model with a
large displacement formulation accounting for interfacial fibrillation. Eur.J.Mech., Vol
26, 2007, pp 1-19.
[7] Bosch van den, M.J.; Schreurs, P.J.G.; Geers, M.G.D.On the development of a 3D cohe-
sive zone element in the presence of large deformations. Comp.Mech., Vol 42, 2008, pp
171-180.
[8] Bosch van den, M.J.Interfacial delamination in polymer coated metal sheet. PhD Thesis,
Eindhoven University of Technology, 2007.
[9] Bosch van den, M.J.; Schreurs, P.J.G.; Geers, M.G.D.; Van Maris, M.P.H.F.L.Interfacial
characterization of pre-strained polymer coated steel by a numerical-experimental ap-
proach. Mech.Mat., Vol 40, 2008, pp 302-314.
[10] Broek, D. Elementary engineering fracture mechanics; 4thed.. Martinus Nijhoff Publish-
ers, 1986.
[11] Broek, D. The practical use of Fracture Mechanics. Kluwer Academic Publishers, 1989.
[12] Camacho, G.T.; Ortiz, M. Computational modelling of impact damage in brittle materi-
als. Int. J. of Solids and Structures, Vol 33, 1996, pp 2899-2938.
[13] Carlson, R.L.; Kardomateas G.A. An Introduction to Fatigue in Metals and Composites.
Chapman and Hall, 1996.
169
170
[14] Donahue, R.J.; Clark H.M.; Atanmo P.; Kumble R.; McEvily A.J. Crack opening dis-
placement and the rate of fatigue crack growth. Int. J. of Fracture Mechanics, Vol 8, 1972,
pp 209-219.
[15] Dugdale, D.S. Yielding of steel sheets containing slit. J. of the Mechanics and Physics of
Solids, Vol 8, 1960, pp 100-108.
[16] Elber, W. Fatigue crack closure under cyclic tension. Engineering Fracture Mechanics,
Vol 2, 1970, pp 37-45.
[17] Elber, W. The significance of fatigue crack closure. Damage Tolerance in Aircraft Struc-
tures, ASTM STP 486, 1971, pp 230-242.
[18] Erdogan, F,; Sih, G.C. On the crack extension in plates under plane loading and trans-
verse shear. J. Basic Eng., Vol 85, 1963, pp. 519-527.
[19] Erdogan, F.; Ratwani M. Fatigue and fracture of cylindrical shells containing a circum-
ferential crack. Int. J. of Fracture Mechanics, Vol 6, 1970, pp 379-392.
[20] Erinc, M.E.; Schreurs, P.J.G.; Zhang, G.Q.; Geers, M.G.D.Microstructural dam-
age analysis of SnAgCu solder joints and an assessment on indentation procedures.
J.Mat.Sc.:Mat.Electr., Vol 16, 2005, pp 693-700.
[21] Erinc, M.; Schreurs, P.J.G.; Zhang, G.Q.; Geers, M.G.D.Characterization and fatigue
damage simulation in SAC solder joints. Microelectronics Reliability, Vol 44, 2004, pp
1287-1292.
[23] Erinc, M.E.Thermo-mechanical fatigue failure of interfaces in lead-free solders. PhD The-
sis, Eindhoven University of Technology, 2007.
[24] Forman, R.G.; Kearney V.E.; Engle R.M. Numerical analysis of crack propagation in
cyclic-loaded structures. J. of Basic Engineering; Trans. ASME, Vol 89, 1967, pp 459-
464.
[25] Gdoutos, E.E. Fracture Mechanics ; an introduction. Kluwer Academic Publishers, 1993.
[26] Gordon, J.E. The new science of strong materials. Penguin Books, 1976.
[28] Griffith, A.A. Phil. Trans. Roy. Soc., Vol 221, 1921, pp 163.
[31] Hutchinson, J.W. Singular behavior at the end of a tensile crack in a hardening material.
J. of the Mechanics and Physics of Solids, Vol 16, 1968, pp 13-31.
[32] Inglis, C.E. Stresses in a plate due to the presence of cracks and sharp corners. Transac-
tions of the Institute of Naval Architects, Vol 55, 1913, pp 219-241.
[33] Irwin, G.R.; Kies, J.A. Welding J. Res. Suppl., Vol 33, 1954, pp 1935.
[34] Irwin, G.R.; Kies, J.A.; Smith, H.L. Fracture strength relative to onset and arrest of
crack propagation. Proceedings of the American Society for Testing Materials, Vol 58,
1958, pp 640-657.
[36] Irwin, G.R. Plastic zone near a crack and fracture toughness. Proc. 7th Sagamore Conf.,
1960, pp IV-63.
[37] Johnson, Gordon R.; Cook, William H. Fracture characteristics of three metals subjected
to various strains, strain rates, temperatures and pressures. Engineering Fracture Me-
chanics, Vol 21, 1985, pp 31-48.
[38] Kanchanomai, C.; et al. Low cycle fatigue test for solders using non-contact digital image
measurement system. Int.J. of fatigue, Vol 24, 2002, pp 57-67.
[39] Kanninen, M.F.; Popelar, C.H. Advanced fracture mechanics. Oxford University TTss,
New York, 1985.
[40] Klingbeil, N.W. A total dissipated energy theory of fatigue crack growth in ductile solids.,
2003.
[42] Kurz, Wolfgang Debonding along a fiber/matrix interface in a composite. WFW 93.109,
1993(jul), pp 34.
[43] Lim, I.L.; Johnston, I.W.; Choi, S.K. Application of singular quadratic distorted isopara-
metric elements in linear fracture mechanics. Int. J. for Numerical Methods in Engineer-
ing, Vol 36, 1993, pp 2473-2499.
[44] Matous, K.; Geubelle, P. Multiscale modeling of particle debonding in reinforced elas-
tomers subjected to finite deformations. International Journal for Numerical Methods in
Engineering, Vol 65, 2006, pp 190-223.
[46] Mediavilla, J. Continuous and discontinuous modelling of ductile fracture. PhD Thesis
TU/e, 2005.
[47] Miner, M.A. Cumulative damage in fatigue. Journal of Applied Mechanics, Vol 12, 1945,
pp A159-A164.
[48] Murakami, Y. Stress intensity factors handbook, Vol.1 and Vol.2. Pergamon, N.Y., 1987.
172
[49] Murakami, Y. Stress intensity factors handbook, Vol.3. Pergamon, N.Y., 1992.
[50] Neuber, H. Theory of stress concentrations for shear-strained prismatical bodies with
arbitrary non-linear stress-strain law. J. Appl. Mech. ASME Trans., Vol 28, 1961, pp
544-550.
[51] Paris, P.; Erdogan, F. A critical analysis of crack propagation laws. J. of Basic Engineer-
ing; Trans. ASME, Vol 85, 1963, pp 528-534.
[52] Peterson, R.E. Stress concentration design factors. John Wiley, 1953.
[53] Peterson, R.E. Stress concentration factors. Wiley, New York, 1974.
[54] Rice, J.R. A path independent integral and the approximate analysis of strain concentra-
tions by notches and cracks. J. of Applied Mechanics, Vol 35, 1968, pp 379-386.
[55] Rice, J.R.; Rosengren, G.F. Plane strain deformation near a crack tip in a power-law
hardening material. J. of the Mechanics and Physics of Solids, Vol 16, 1968, pp 1-12.
[56] Roberts, D.K.; Wells, A.A. The velocity of brittle fracture. Engineering, Vol 171, 1954,
pp 820-821.
[57] Rooke, D.P.; Cartwright, D.J. Compendium of stress intensity factors. Her Majestys
Stationary Office, London, 1976.
[58] Schijve, J. Some formulas for the crack opening stress level. Engineering Fracture Me-
chanics, Vol 14, 1981, pp 461-465.
[59] Sih, G.C.; Chen, E.P. Fracture analysis of unidirectional composites. J. of composite
materials, Vol 7, 1973, pp 230-244.
[60] Sih, G.C. Handbook of stress intensity factors. Lehig University, Bethlehem, Pennsylvania
18055, 1973.
[61] Sih, G.C. Strain energy density factor applied to mixed mode crack problems. Int. J. of
Fracture, Vol 10, 1974, pp 305-322.
[62] Tada, H.; Paris, P.C.; Irwin, G.R. The stress analysis of cracks handbook. Del Research,
1973.
[63] Tada, H.; Paris, P.C.; Irwin, G.R. The stress analysis of cracks handbook; 2nd Ed. Paris
Productions, Inc., St. Louis, 1985.
[65] Tvergaard, V.; Hutchinson, J.W. The relation between crack growth resistance and frac-
ture process parameters in elastic-plastic solids. J. Mech. Phys. Solids, Vol 40, 1992, pp
1377-1397.
[66] Vinals, J. High performance fracture approach to fatigue crack analysis and life prediction.
In: WCCM-5, 2002, 2002.
173
[67] Wasiluk, B.; Hoshide, T. The fracture process in elastic-plastic materials under biaxial
cyclic loading., 2003.
[68] Wells, A.A. Application of fracture mechanics at and beyond general yielding. British
Welding Research Ass. Rep., MI3, 1963.
[69] Xu, X.P.; Needleman, A. Void nucleation by inclusions debonding in a crystal matrix.
Model Simul Mater Sci Engng, VOL 1, 1993, pp 111-132.
[70] Yoffe, E.H. The moving Griffith crack. Philosophical Magazine, Vol 42, 1951, pp 739-750.
External sources of figures and other data
Source
Laplace equation
The derivatives of the complex function f w.r.t. x1 and x2 are calculated first. After using
tha Cauchy-Riemann equations and a second derivation w.r.t. x1 and x2 , it follows that the
real part of the complex function f satisfies the Laplace equations. The same can be deribed
for the imaginary part of f .
f f z f
= = = f (z)
x1 z x1 z
f f
f (z) = = i
f f z f x1 x2
= if (z)
= =i
x2 z x2 z
+i = i +i = i
x1 x1 x2 x2 x2 x2
Cauchy-Riemann equations
= and = =
x1 x2 x1 x2 x2 x1
2 2 2 2
2 = ; 2 =
x1 x1 x2 x2 x1 x2
2 2
+ 2 =0
x21 x2
a1
a2
Appendix B
The general solution of the complex bi-harmonic equation is an expression for the Airy func-
tion in terms of complex functions.
1
= 2 z + z + +
= 21 z + +
z
= 21 + z +
z
2
= 12 +
z z
2
= 21 z +
z 2
2
= 21 z +
z 2
z z
,1 = + = +
z x1 z x1 z z
2 2 2
z z
,11 = + + + =2 + +
z z z x1 z z z x1 z z z 2 z 2
2
2
z z
,12 = + + + =i
z z z x2 z z z x2 z 2 z 2
z z
,2 = + = i i
z x2 z x2 z z
2 2 2
z z
,22 = i i + i i =2 2
z z z x2 z z z x2 z z z 2 z
a3
a4