You are on page 1of 21

An analysis of crack growth in thin-sheet metal via a

cohesive zone model


Weizhou Li, Thomas Siegmund
*
School of Mechanical Engineering, Purdue University, West Lafayette, IN 47907, USA
Received 13 September 2001; received in revised form 21 December 2001; accepted 31 December 2001
Abstract
A cohesive zone model (CZM) is applied to crack growth in thin sheet metal. CZM parameters are determined from
results of global measurements and micromechanical damage models. Crack propagation in constrained center-cracked
panels is analyzed to verify the choice of CZM parameters. Special attention is paid to the interaction between buckling
and crack growth and to crack link-up in multi-site damaged specimens. The good agreement found between the
predicted and experimental data demonstrates that the approach is attractive in investigation of structural integrity of
thin-walled structures and does not require assumptions regarding the geometry and size dependence of crack growth
parameters.
2002 Elsevier Science Ltd. All rights reserved.
Keywords: Crack growth; Buckling; Aluminum; Cohesive zone model; FEM; Crack tip opening angle; Multi-site-damage
1. Introduction
The present work aims at investigations of the mechanical integrity of structures built of thin-walled
members. These structures are often made of high strength sheet metal and are frequently used if light-
weight and high stiness construction is required. While the high strength characteristics are attractive in an
array of engineering applications, these thin-walled shell structures are also known to be very sensitive to
instability if imperfections are present. In the case of the occurrence of through thickness cracks, the
competition and interaction between global failure due to buckling and local failure due to crack growth
determine the ultimate behavior of the structure.
Failure of cracked thin-walled shell structures due to buckling has attracted considerable attention in the
past [16], including investigation about critical loads and buckling modes which depend on the geometry
and boundary conditions [1,2], the eect of buckling deformations on the stress intensity [3] and energy
release rate [4], and the inuence of buckling on failure loads [5,6]. In addition to the global deformation
instability behavior, the structural behavior of a thin-walled shell structure is also aected by the growth of
Engineering Fracture Mechanics 69 (2002) 20732093
www.elsevier.com/locate/engfracmech
*
Corresponding author. Fax: +1-765-494-0539/496-7536.
E-mail address: siegmund@ecn.purdue.edu (T. Siegmund).
0013-7944/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S0013- 7944( 02) 00013- 9
Nomenclature
2a
i
initial length of lead crack
Da crack extension
b
n
length of nth-ligament
c
n
length of nth-minor crack
COD/CTOD crack opening displacement/crack tip opening displacement
CTOA crack tip opening angle
CTOA
I;II;III
crack tip opening angle for mode I, II, III
e exp(1)
E elastic modulus
F deformation gradient
h size of cohesive element
J J-integral
J
IC
mode I critical J-integral
J
ICi
mode I critical J-integral for plane strain crack initiation
K
IC
mode I plane strain fracture toughness
2L specimen height
n surface normal
n hardening exponent
s nominal stress tensor
S
ext
, S
int
external and internal surfaces
t specimen thickness
T
CZ
crack surface traction vector
T
e
traction vector on the external surface of the body
T
n
normal crack surface traction
T
t1
, T
t2
tangential crack surface traction
u displacement vector
Du displacement jump vector across crack surface
Du
n
normal displacement jump across crack surface
Du
t1
, Du
t2
tangential displacement jump components across crack surface
V specimen volume
2w specimen width
x
i
coordinate axes
e uniaxial true strain
e strain tensor
d cohesive length
/ cohesive potential function
/
c
cohesive energy
k nondimensional separation measure
m Poissons ratio
r uniaxial true stress
r Cauchy stress tensor
r
0
tensile yield stress
r
max
cohesive strength parameter
R
22
applied stress
s Kirchho stress tensor
2074 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
cracks. Long cracks can arise from accidental, unexpected damage during use, and fatigue crack growth.
Situations are especially critical if several smaller cracks along a row of rivets might have linked up. Within
the concept of the damage tolerant design approach, it is assumed that such critical cracks can be present
within the structure at any time. Consequently, in an assessment of e.g. the structural integrity of aircraft
structures [7], it is of vital importance to determine the residual strength of a cracked structure, and to
investigate whether the structure under consideration can contain a crack without failing in an unstable
manner. Ideally, in a design of a thin-walled shell structure, the presence of cracks should thus be taken
explicitly into account. To obtain an even more fundamental insight, a framework is required for analysis
capabilities that allow for crack initiation and crack propagation. For shell structures such an analysis
concept should allow for the understanding of the nonlinear interaction between crack growth and the
deformation instabilities of bulging, buckling and collapse [5,6].
Methods being applied to analyze crack growth in thin-walled components include global approaches
and local approaches. The global approaches include the use of stress intensity factor [8], the J-integral
[9] or the T

-integral [10]. The usefulness of linear elastic fracture mechanics based fracture criteria [8] is
limited due to the inherent elasticplastic character of metallic materials. Elasticplastic fracture mechanics
concepts based on the J-integral or its two-parameter (JQ) extension are appropriate for prediction of
loads prior to crack extension and provide a good representation of the crack tip elds. The J-integral,
however, loses its predictive capabilities if a large amount of crack extension has to be taken into account
[9]. Consequently, the J-integral can be applied to the predictions of the residual strength of real structures
only if certain geometry requirements are fullled. The resistance curves determined in a laboratory test
program are required to stem from specimens of similar size and geometry as those encountered in the
structure to be investigated. This issue of lack of transferability of fracture data from one specimen ge-
ometry to another imposes severe restrictions on fracture resistance prediction based on the J-integral. In
terms of the thin sheet metal considered here, the consequence is the necessity to use very large specimens in
the fundamental laboratory fracture tests. The T

-integral was proposed to circumvent these problems.


Recent applications of the T

-integral to crack growth simulations [10] indicate that this quantity might in
fact be a geometry independent fracture parameter. Still, as an integral quantity, the T

-integral does not


provide fundamental understanding of the interplay between the two fundamental contributions to ma-
terial toughness, namely material separation and remote plasticity. For thin-walled structures, the issues
associated with those fracture parameters are aggravated if the out-of-plane displacements due to local
buckling and bulging are taken into account [3]. Stress intensity factor for bending and membrane stresses
cannot be obtained by linear superposition; additional components of the J-integral are needed to fulll the
path independence of J-integral due to the nonlinearty of the problem [3]. Currently, these nonlinear eects
and their inuence on the failure behavior are taken into account mainly by the use of correction factors
[11]. Considering the advances in nonlinear structural mechanics analysis tools, e.g. the nite element
method, such correction factors become increasingly outdated.
A more fundamental understanding of crack growth resistance prediction can be obtained by application
of the so-called local approach to fracture. Within this approach, fracture is described by a failure cri-
terion locally at the crack tip and the interplay of this criterion with the description of the surrounding
material determines the macroscopic toughness of the structure under investigation. This approach is very
well established in e.g. the analysis of pressure vessel steels, adhesives and composite structures [12].
Most previous analyses of ductile tearing in sheet metal reported in the past used a local deformation
criterion to describe material separation. Mainly, two methods have been applied with success. One is a
criterion based on a critical crack tip opening angle, CTOA, [13]. The other approach uses a crack (tip)
opening displacement, COD/CTOD, at a dened location relative to the crack tip [14]. These two criteria can
either be used in combination with full-eld nite element solutions [1517] or within solutions based on the
Dugdale-model [18,19]. Recently, COD/CTOD based criteria have proven advantageous especially in mixed
mode fracture prediction including the predictions of crack growth directions [20]. CTOA, COD/CTOD
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2075
criteria are based on local deformation measures and thus have the disadvantage that their critical values
cannot be linked to the micromechanical failure processes of void nucleation, growth and coalescence un-
derlying crack growth. The commonly made choice of a constant CTOA criterion implies a constant slope of
the resulting crack growth resistance curve (JDa) and cannot account accurately for the transient eects at
crack initiation or crack link-up in multiple-site damaged specimens.
To take account of the combined eects occurring due to material failure and structural response, such as
instability, a fracture description based on physically meaningful and controllable parameters is needed
[21]. Such a framework can be established if a constitutive description of material failure based on physical
parameters is used in the structural analysis.
The use of the dilatant plasticity models of Gurson [22] or Rousselier [23] has been a successful method to
analyze crack growth in thick sections of high strength steels [24]. Recently, several attempts were made to
apply these models also to thin sheet metal [2528], and void growth models for low triaxility situations
have been developed [29,30]. While this approach provides detailed solutions of the 3D damage evolution at
the crack tip, its incorporation into large scale structural models remains challenging. The application of
continuum mechanics models for void growth in the nite element method requires use of element size of
the order of the physical dimple distance. In high strength aluminum alloys, the typical dimple size is found
to be of the order of 10 lm [31]. With this restriction on element size, the numerical analysis of typical crack
extensions up to Da 100 mm out of reach even today.
Within the present study, a more practical approach is presented. Material separation is described by the
use of a cohesive zone model (CZM), which represents the mechanical processes in the fracture process zone
in front of the crack tip. The present application of a CZM diers from those undertaken in the past since
the cohesive elements are embedded into a surrounding of shell elements. This method represents a con-
venient approach to the analyses of specimens and structures in which cracks are not only subjected to
tensile loading due to membrane stresses, but also inuenced by potentially large out-of-plane deformation.
In principle, such problems can be solved using full 3D solid models, however, shell elements present
themselves as a cost eective alternative.
In the present paper, nite element simulations of crack growth in sheet metal specimens of Al 2024-T3
Alclad and Al 2024-T3 are presented. Special consideration is given to the determination of the cohesive
zone model properties. Numerical results are compared to experimentally determined data as published in
literature [18,33]. Once the cohesive zone material property selections are veried on at, constraint
specimens with dierent sizes and initial crack length, symmetric and anti-symmetric buckled specimens, as
well as multi-site damaged specimens are investigated.
2. Problem description
2.1. Cohesive zone model formulation
In a CZM, fracture surfaces in front of a crack tip can separate under loading with the process of
material separation being described by a softening constitutive equation relating the crack surface tractions,
T
CZ
, to the displacement jump, Du, across the crack. This constitutive equation introduces the cohesive
energy, strength, and length as additional material parameters, in addition to the elasticplastic properties.
The simulations reported here are performed using the nite element method. The cohesive zone con-
stitutive relation is embedded into so-called cohesive elements, which are then surrounded by continuum
elements [37]. The mechanical equilibrium statement written as the principle of virtual work is given by:
_
V
s:dFdV
_
S
int
T
CZ
dDudS
_
S
ext
T
e
dudS 1
2076 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
Here, the standard parts of the equation (including the terms for specimen volume, V, and the external
specimen surface, S
ext
) are described by the nominal stress tensor, s, the deformation gradient, F, the dis-
placement vector, u, as well as by the traction vector, T
e
, on the external surface of the body. A:B denotes
A
ij
B
ji
, and traction vectors are related to s by T ns, with n being the surface normal. Also, s F
1
s,
where s is the Kirchho stress, s detFr, with r being the Cauchy stress. The cohesive surface contri-
butions, representing the crack and the process zone in front of the crack tip, are described by the integral
over the internal surface, S
int
. This framework is attractive since it allows for the simulation of crack growth
without the need for analyst to control the solution or to specify any articial node release algorithms.
To achieve the success of proposed CZM-shell element approach, past developments in crack growth
modeling with cohesive zone elements are extended, and three main model developments are presented
here.
Initial developments of CZMs for crack growth simulation are due to [3436]. In the past, cohesive ele-
ment formulations were developed for use with axis-symmetric [37], as line elements for use in combination
with 2D plane strain element [3842], or as surface elements for use with 3D tetrahedral or brick elements,
e.g. [4346]. Within 3D surface elements, material separation in three fracture modes is incorporated. The
dierence between the present study and previous cohesive zone formulations is that the cohesive elements
used here are combined with a continuum environment of shell elements. Such a formulation which rep-
resents structures containing through cracks hasto the present knowledge of the authorsnot yet been
explored. To be compatible with large deformation of shell structures, an extension of the kinematics of the
cohesive element formulation, [3842], was undertaken. Even considering the fact that the cohesive ele-
ments remain a line elements, the traction vector, T
CZ
, and the displacement jump Du, possess three
components each and material separation in mode I, II and III can be taken into account within a cohesive
line element. Normal separation is described in terms of the normal crack opening stress, T
n
, as related to
the normal displacement jump across the crack surfaces, Du
n
. Furthermore, two shear separation com-
ponents for modes II and III are described in terms of tangential tractions, T
t1
, T
t2
and tangential dis-
placement jumps, Du
t1
, Du
t2
, across the crack. Based on a previous relation [37] the cohesive constitutive
equation is based on a potential, /:
/ /
c
/
c
1 k expk 2
The quantity k designates an eective normalized displacement jump across the crack [43].
k

Du
n
d
_ _
2

Du
t1
d
_ _
2

Du
t2
d
_ _
2

3
Tractions are obtained as derivative of the cohesive zone potential with respect to the displacement jump:
T
n

o/
oDu
n
r
max
Du
n
d
_ _
exp1 k
T
t1

o/
oDu
t1
r
max
Du
t1
d
_ _
exp1 k
T
t2

o/
oDu
t2
r
max
Du
t2
d
_ _
exp1 k
4
In this formulation, r
max
denotes the cohesive strength, i.e. the maximum traction value that can be
sustained within the cohesive zone. The cohesive length, d, is the value of the displacement jump across the
crack surfaces at which the stress carrying capacity of the cohesive elements reach maximum value. The
work required for the creation of new crack surface, i.e. the cohesive energy, /
c
, is given by:
/
c
er
max
d 5
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2077
with e exp(1). The current location of crack tip is dened by the condition that k 16e=9 [38]. Fig. 1
depicts the normal traction in dependence of the normal displacement jump across the crack faces for three
dierent value of tangential displacement jump Du
t
.
Before any computation can be performed, the determination procedure of the cohesive material pa-
rameters for thin sheet metal is to be addressed. In many applications of the CZM, the values of the
constitutive parameters were determined by a calibration of numerical results on crack growth data ob-
tained by experiments on well dened test specimens. Here, a dierent approach is followed and an attempt
is made to provide cohesive zone parameter values from global measurement results and micromechanical
damage model considerations, thus avoiding the comprehensive numerical calibration. Present estimate for
the cohesive zone parameters is based on results from micromechanical investigations of void growth [47],
and previous studies of crack initiation and growth [4856].
For the description of material separation in ductile materials, when the crack growth is the material
separation mechanism, it is well established that the value of the cohesive strength is dependent on the yield
strength of the material under consideration and the state of mechanical constraint present at the crack tip.
Constraint is thereby commonly expressed by the stress triaxiality, i.e. the ratio of hydrostatic stress to
equivalent stress. Void growth and nucleation processes in specimens under nominal plane strain, a rather
highly constrained state, are generally characterized by values of r
max
=r
0
between three and four [39]. In the
thin sheet material considered here, the level of triaxiality is considerably lower. Moreover, crack propa-
gation commonly occurs as slant fracture [50]. In [47], a study on interaction between void growth and
microvoid cavitation was presented for a material with a ratio of r
0
=E characteristic of the alloys con-
sidered here. In the material studied in [47], the peak stress was found to be approximately 3.5r
0
if the
loading was at high strain biaxiality and void coalescence occurred in a plane perpendicular to the principal
loading direction. However, for low strain biaxiality, void coalescence took place under 45 as in slant
fracture and the value of the peak stress dropped to 2r
0
. Similar results were obtained in a study on dy-
namic crack growth in a thin sheet [25]. There, the Gurson model was applied to simulate crack growth. If
the specimen was unconstrained at the surface (a nominal plane stress situation), fracture was slant and the
average normal opening stress was low. If the specimen was constrained (a nominal plane strain situation),
the fracture surface remained perpendicular to the specimen surface and the average normal opening stress
Fig. 1. The tractionseparation law employed in the cohesive zone model: normal traction under dierent shear separations levels.
2078 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
increased. Finite element studies of the stress distribution at crack tips in thin specimens were also reported
in [48,49]. While these studies considered only elasticplastic material properties, they again showed that
the peak normal opening stress at the crack tip is approximately 2r
0
for thin sheet. Little change in this
average peak normal opening stress with crack extension was reported in [49]. Based on these past studies
on void growth and crack extension, the value of the cohesive strength was taken to be r
max
=r
0
2:0 for all
computations. It is recognized that the processes of ductile fracture at crack initiation and propagation
in thin sheet can be substantially dierent. However, since the focus of the study was on relatively large
amounts of crack extension, a single and constant value for r
max
was used in the present study.
For the instance of crack initiation under small scale yielding, the mode I cohesive energy can be
identied with the value of the J-integral at crack initiation, /
i
J
ICi
. The values of plane strain fracture are
then chosen as the values of the initiation toughness in present simulations. This choice is motivated by
experimental results on high-resolution detection of crack tip damage and crack initiation in thin aluminum
sheet [51,52]. These experimental investigations demonstrated that crack initiation in thin sheet specimens
took place in the center of the sheet at a value of the applied energy release rate equal to J
IC
. To estimate the
cohesive energy for the propagation stage where slant fracture has developed, the mixed mode fracture
mechanism maps presented in [5356] are used. These papers summarized a large number of experiments
concerned with initiation of crack growth in specimens with the initial crack tilting relatively to the di-
rection perpendicular to the sheet surface. Charts are provided the ratio of mixed mode initiation J-integral
to mode I initiation J-integral vs. a function of the ratio of mode I initiation J-integral to yield strength. The
value of the mixed mode critical J-integral is then identied as the value of the cohesive energy for the crack
propagation under slant fracture.
Finally, in the present approach, it is assumed that all specimens are under nominal plane stress con-
ditions, i.e., the fracture is slant and the plane stress plastic zone fully developed through the specimen
thickness.
2.2. Numerical implementation
The cohesive surface contributions are incorporated into a nite element code by the use of cohesive
elements. The commercial general-purpose code ABAQUS [58] allows such a numerical setup by means of
user-dened elements (UEL).
Special attention is given to the modeling issues associated with the thickness change of the thin sheet
specimens. The thickness change of the continuum elements near the crack tip is determined from com-
putations without crack growth to be up to 15%. Two dierent types of shell elements are considered here.
One type of elements (e.g. S4R5 in ABAQUS) considers in-plane rotations but small strains only. This type
of elements is intended for using in thin shell applications, and the thickness changes are ignored. The other
type of elements (e.g. S4R or S4 in ABAQUS) considers large in-plane rotations and nite membrane
strains, and allows for the thickness to change. When using the nite membrane strain elements, the co-
hesive element must also take into account out-of-plane deformation. The thickness of the cohesive zone
element must be updated continuously. A special numerical setup is developed to accomplish this. The
thickness dependence of the cohesive zone elements is achieved by coupling of the cohesive elements to the
adjacent shell elements. For each four-node shell element adjacent to a cohesive zone element, the values of
the in-plane strain components at the integration point are transported into the corresponding cohesive
element. This is accomplished by use of the UVARM subroutine in ABAQUS. The current thickness of the
cohesive element is then computed under the assumption of constant volume during plastic deformation.
The numerical solution of the crack growth problem for the buckled specimens is performed in two steps.
Initially, an eigenvalue buckling analysis is performed without the cohesive elements employed. Based on
the solutions from this step, initial imperfections are imposed on the perfect at specimens. The maxi-
mum imposed out-of-plane disturbance was 5% of the panel thickness. Thereby, the discontinuous response
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2079
at the buckling bifurcation is turned into a continuous problem. The modied Riks method, which can
provide a convergent solution for complex unstable problems, is then used to predict the specimen behavior
including buckling and crack growth.
The numerical implementation uses an adaptive load control. Such a strategy is implemented so that a
single large load increase is subdivided appropriately. Such that, a single load step does not cause several
crack growth steps. This adaptive load control strategy enforces the increase in material separation in a
single load step to be at maximum 0.2d. A violation of this requirement results in a 50% reduction of the
increment size in the next solution step.
2.3. Specimen denition
The present investigation is concerned with crack growth simulation in two dierent aluminum alloys,
namely Al 2024-T3 Alclad and Al 2024-T3. The constitutive relation used to describe the deformation
behavior of these two alloys is an isotropic elasticplastic constitutive relation using the Mises yield
function. In this paper, the following representation of uniaxial true stress, r, vs. uniaxial true strain, e,
curve is used:
e
r
E
for r < r
0
r
0
E
_ _
r
r
0
_ _
n
for r > r
0
_

_
6
The elasticplastic material properties for both the alloy 2024-T3 Alclad and the alloy 2024-T3 used in
Eq. (4) are summarized in Table 1. Based on the procedure outlined in Section 2.1, the cohesive strength is
r
max
570 MPa for Al 2024-T3 Alclad and r
max
690 MPa for Al 2024-T3. For both the alloys a true
crack initiation value of K
IC
35 MPa m
1=2
was used [51]. The mixed mode fracture mechanism maps of
[5357] provide the ratio between the I/III mixed mode and mode I crack initiation toughness values. With
J
ICi
K
2
IC
=E, for the current materials the ratios of J
ICi
=r
0
are 0.06 and 0.05, respectively. In this case, the
ratios between the I/III mixed mode initiation toughness and the pure mode I toughness are close to one for
both materials and also independent of the mode mixity up to very high degrees of mode mixity. As a
consequence, the cohesive energy for crack initiation and for crack propagation under slant fracture are
estimated to be of identical magnitude, i.e. /
c
17 kJ/m
2
for both alloys. Finally, with Eq. (2), the values of
the cohesive lengths are d 0:01 and 0.009 mm, respectively.
Investigations are carried out also for center-cracked panel specimens, MT, either containing only a
single crack, or for multi-site damaged, MSD, specimens containing a center crack plus additional minor
cracks. The dimensions of the specimens are the height 2L, the width 2w. The initial length of the center
crack is 2a
i
, the length of the nth-ligament between the additional cracks in front of the main crack is b
n
,
and the length of the minor cracks is c
n
. The geometries of the center-cracked and the multi-site-damaged
panel are given in Fig. 2. In the nite element simulation, one-quarter models with appropriate boundary
conditions are analyzed for constraint specimen, and the symmetric buckling case. For the anti-symmetric
buckling case, one-half models were used.
Table 1
Elasticplastic material properties of the alloys under investigation
Material E (MPa) m r
0
(MPa) n Ref.
Al 2024-T3 Alcad 71,400 0.3 285 8 [18]
Al 2024-T3 71,300 0.3 345 10 [33]
2080 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
3. Results
3.1. Flat panelsmodel validation
The rst set of results presented here is on results of crack growth in center-cracked panels of Al 2024-T3
Alclad. Experimental data on crack growth in specimens of this material were summarized in [18]. The
specimens under investigation possessed width of 2w 508 mm, thickness of t 1:0 mm, and three initial
crack lengths of 2a
i
101:6, 177.8 and 280 mm, respectively. The specimens are loaded under constraint to
avoid buckling. As an example, for the specimen with 2a
i
177:8 mm, the numerical simulations use 2652
continuum elements surrounding 200 cohesive elements of length h 0:25 mm. Results from previous
investigations [59,60] demonstrated that the resulting value of h=d 5 lead to convergent predictions of
fracture within the cohesive element approach. In the experiments, initial cracks were introduced as saw
cuts instead of by fatigue. The present investigation neglects this inuence on the specimen behavior. Shell
elements without thickness change are used in these simulations.
Numerical predictions of applied stress, R
22
, far eld tension in x
2
direction, vs. crack extension, Da, for
the three dierent crack lengths are depicted in Fig. 3, and compared with experimentally obtained data.
The numerical simulation results are in good agreement with the experimentally determined data of [18].
They clearly predict an increase of applied stress level after crack initiation and subsequently a decrease of
applied stress occurs as the crack propagate further. The computations are used to determine the crack tip
Fig. 2. The geometry of the specimens under consideration (a) center-crack panel, (b) multi-site-damage specimen.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2081
opening angle, CTOA, in dependence of the crack extensions. For the calculation of CTOA, the distance
behind the current crack tip used to calculate CTOA is equal to the length of the cohesive elements. Results
for the prediction of the evolution of the mode I crack tip opening angle, CTOA
I
, in dependence of the
crack extension, Da are depicted in Fig. 4 for the three center-crack panels of Al 2024-T3 Alclad. The
evolution of CTOA
I
is nearly identical for all three initial crack lengths. At crack initiation, the values of
CTOA
I
are large but decrease substantially during the rst 5 mm of crack extension. After that, only a
small further reduction of CTOA
I
occurs on further crack growth. The results on the evolution of CTOA
I
with the crack extension as predicted here are in very good agreement with the CTOA
I
measurements in
Fig. 3. Applied stress vs. crack extension for the MT specimens of Al 2024-T3 Alclad. Three dierent crack lengths are considered.
Comparison of numerically determined values to the experimental data.
Fig. 4. The evolution of the CTOA vs. the crack extension in the three center-crack panels of Al 2024-T3 Alclad.
2082 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
[50] for the same material. Also, CTOA
I
values of 8.2 for initiation and of 4.8 for propagation were used
in the numerical investigation of Ref. [16]. These CTOA values are in good agreement with the results in
Fig. 4.
The second set of results on constrained center-cracked panel specimens is concerned with crack growth
experiments in Al 2024-T3 as reported in [33]. Three specimen geometries are discussed here, namely
w 75, 300 and 600 mm, respectively. The initial crack lengths are a
i
w=3 in all specimens, the sheet
thickness is t 2:3 mm. Results from the numerical simulations together with experimental data are shown
in Fig. 5. It should be noted here that in this plot, Da is normalized by the corresponding initial crack
length, a
i
, in order to merge the results into a single diagram. Again, a good agreement between the nu-
merical predictions and the experimentally measured data is obtained. Specically, the crack initiation and
peak loads are very well predicted. Only for large amount of crack extensions, the numerical simulation
predicts a smaller load than that observed experimentally. The predicted values of CTOA
I
in dependence of
crack extension for the Al 2024-T3 specimens are shown in Fig. 6. These results display a very similar trend
as already shown in Fig. 4. It shall, however, be noticed that the steady state CTOA
I
now is approximately
3.4 which is 0.4 lower than for the Al 2024-T3 Alclad. In comparison to previous investigation, a CTOA
I
value of 5.1 was used in numerical study of [33] as a crack growth criterion to best t the experimental
results.
3.2. Choice of type of nite elements
To obtain a better understanding of the proposed numerical approach, computations for the MT
specimen of Al 2024-T3 Alclad with a
i
88:9 mm were performed with several continuum elements and
cohesive zone elements combinations. The results of these simulations are depicted in Fig. 7. The predic-
tions obtained from the models using nite membrane strain (S4R) elements without corresponding
thickness change in the cohesive elements deviate considerably from those obtained by the small-strain shell
(S4R5) element, which are in good agreement with experimental data, see Fig. 3. Considering the thickness
changes only in the continuum elements, but not in the cohesive elements leads to considerable errors in the
numerical predictions. Only if the thickness changes of the nite membrane strain shell elements are
Fig. 5. Remote applied stress vs. normalized crack extension for the MT specimens of Al 2024-T3; w 75, 300 and 600 mm,
a
i
=w 0:33.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2083
considered in the cohesive zone elements, the predictions obtained by using this element also match the
experimental data.
3.3. Multi-site-damage specimen
An issue of major interest with respect to residual strength analyses of metal sheet specimens is the
prediction of the residual strength of multi-site-damage structures, see [5,32,61] for overviews and [16,18,19]
Fig. 6. The evolution of the CTOA
I
vs. the crack extension for the MT specimens of Al 2024-T3; w 75, 300 and 600 mm,
a
i
=w 0:33.
Fig. 7. Remote applied stress vs. crack extension for the Al 2024-T3 Alclad using dierent continuum and cohesive zone element
combination; w 508 mm, a
i
88:9 mm.
2084 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
for some recent developments. With the cohesive zone material parameter selection validated and the
numerical setup tested, the predictive capabilities of the present crack growth model are demonstrated by
analyzing one of the multi-site-damage specimens of [32]. The specimen (P6 in the report [32], Al 2024-T3
Alclad) possesses a width of 2w 508 mm and a thickness of t 1:0 mm. In addition to the lead crack of
initial length 2a
i
193 mm, two additional minor cracks in front of each initial lead crack tip were present
such that the ligament lengths b
1
6:35 mm, b
2
12:70 mm and the minor crack lengths c
1
19:05 mm,
c
2
31:75 mm. Results of the analysis of this specimen are depicted in Fig. 8; together with the specimen
containing the lead crack only, in form of the applied stress, R
22
, vs. the extension of the lead crack. The
applied stress increases monotonically rst as the main crack grows across the rst ligament. As the main
crack tip interacts with the rst minor crack, the applied stress drops about 20 MPa. After the link-up of the
lead crack with the rst minor crack, the crack then jumps across the rst minor crack, and subsequently
crack growth is restarted within the second ligament. Again, as the main crack starts to interact with the
second minor crack, a decrease in the applied stress is predicted of approximately 30 MPa. Once the re-
maining ligament is reached, a peak in the applied stress is predicted again as the crack growth further. The
experimental report in [32] contains data on the applied peak stresses as indicated in Fig. 8. Again, the
present crack growth model predicts the experimentally determined data reasonably well. When the crack
growth data of the MSD specimen are compared with the specimen including the lead crack only, a
considerable reduction in the residual strength can be observed.
The application of the cohesive zone approach in this study allows a detailed understanding of the
fracture processes in the MSD specimen because events during crack initiation, link-up and renucleation
are captured in their transient nature without having imposed any analyst-dened crack growth criteria.
The evolution of CTOA
I
in the MSD specimens is given in Fig. 9, where CTOA
I
is given in dependence of
the initial location of a cohesive zone element in the ligament. All CTOA
I
values are assigned positive
numbers, regardless of the crack growth direction. As expected, again, no constant CTOA
I
is observed.
CTOA
I
has large values at each of the initial crack tips. Subsequently, a rapid reduction from these large
initial values is observed for all crack tips. For cracks growing in x

1
, including the lead crack, the reduction
of CTOA
I
is slightly slower than that for the cracks growing in x

1
. Once two crack tips start to interact, a
Fig. 8. Remote applied stress vs. crack extension for the multi-site-damaged specimen of Al 2024-T3 Alclad, comparison to experi-
mental data, a specimen containing the lead crack only, and the symmetrically buckled specimen.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2085
strong reduction in CTOA
I
is observed, and CTOA
I
nearly falls to zero during link-up. A similar type of
crack interaction with the occurrence of an instability at the instance of crack link-up was also observed for
double cantilever beam specimens with secondary cracks [62]. There, instability during link-up is mani-
fested as a drop in the apparent fracture toughness. During this stage of strong crack tip interaction, the
ratio of the length of the remaining ligament to the cohesive length is small. From Fig. 9, the distance over
which CTOA
I
drops signicantly is determined to approximately 5 mm during link-up of the lead crack
with the two minor cracks. The ratio between this remaining ligament length and d is approximately 500.
For this ratio of length scales, the CZM predicts a transition from a crack like behavior to a uniform
debonding behavior along the remaining ligament. A similar transition behavior was observed in [43] in a
study of debonding of a plain strain block from a rigid surface, where the size of the block was changed to
the cohesive length.
3.4. Crack growth buckling interaction
During loading of the thin sheet specimens under a remote tensile load in x
2
direction, compressive
stresses in x
1
develop in the area surrounding the crack. These stresses can cause local buckling, and
amplication of the loading of the crack tip. The nonlinear interaction between the local buckling and crack
growth via the coupling of in-plane and out-of-plane deformation will aect the structural integrity of the
specimens. The material used in buckling analyses is Al 2024-T3 Alclad. Both MT specimens with a range
of a
i
=w ratios and the MSD panel are investigated with the out-of-plane displacements are unconstraint.
Both symmetric and anti-symmetric buckling modes are considered in the present analysis.
The overall specimen behavior of MT specimen with a
i
=w 0:33 under the consideration of buckling is
shown in Fig. 10, in terms of a plot of applied load vs. out-of-plane displacement, u
3
. The critical buck-
ling load is considerably smaller than the load required to initiate crack growth. So, for the present ma-
terial and specimen geometry, buckling always precedes crack growth independent of whether symmetric or
Fig. 9. The evolution of the CTOA
I
in MSD specimens, in dependence of the initial location of the cohesive element. Comparison to
the evolution of CTOA
I
in the specimen containing the lead crack only, and the symmetrically buckled specimens.
2086 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
anti-symmetric buckling modes are considered. For the symmetric buckling mode, the out-of-plane dis-
placement is identical for two corresponding locations on either side of the crack, thus fracture still occurs
as a nominal mode I case. For the anti-symmetric buckling mode, the out-of-plane displacements for
corresponding location on either side of the crack are of identical magnitude but have opposite sign. Thus,
fracture occurs under mixed mode I/III. Fig. 11 depicts the deformed specimen shapes for the symmetric
and anti-symmetric buckled specimen after some amount of crack growth. The results describing the crack
growth behavior under the inuence of buckling are given in Figs. 12 and 13 in the form of applied load and
the crack tip opening angles vs. crack extension, respectively. The applied load at crack initiation in the
symmetrically and the anti-symmetrically buckled specimen is only slightly lower than in the constraint
specimen. However, a considerable reduction in the peak load is present for the buckled specimen. The
drop in peak stress is about 30% for the symmetrically and 28% for the anti-symmetrically buckled spec-
imen. A better understanding of the development of the fracture process is obtained from the plot of the
crack tip opening angles for mode I, CTOA
I
, and mode III, CTOA
III
, vs. crack extension, Da, in Fig. 13.
For the symmetrically buckled specimen in which only mode I separation takes place, CTOA
I
is again
large at crack initiation and drops subsequently. The CTOA
I
value of steady state for the symmetrically
buckled specimen is approximately 3.4, and 0.4 smaller than that for the constraint specimen. For the
anti-symmetrically buckled specimen, the evolution of CTOA
I
is similar to the one for the symmetri-
cally buckled specimen. However, now CTOA
III
is nonzero. CTOA
III
at crack initiation is approximately
3.6. CTOA
III
subsequently increases and reaches a steady state value of 5.0. It shall be noticed here
that the predicted value of the steady state CTOA
III
is about 1.6 larger than the corresponding CTOA
I
value.
Within the CZM approach to fracture, changes in crack growth resistance are linked not only to vari-
ations in the cohesive zone properties but also to the stress and strain distribution in the adjacent solid. For
sheet metal specimen, it is known that the stress and strain components perpendicular to the crack only
show little sensitivity to the local mechanical constraint [9]. The strain component parallel to the crack,
however, is a good constraint indicator. In Fig. 14, the distribution of e
11
, is depicted for locations just
above the cohesive zone at the instance of peak load. It can be seen that e
11
is signicantly aected by the
Fig. 10. Remote applied stress vs. the out-of-plane displacement for the MT specimens of Al 2024-T3 Alclad, w 508 mm, a
i
88:9
mm. (1) Anti-symmetric buckling without crack growth. (2) Symmetric buckling without crack growth. (3) Anti-symmetric buckling
with crack growth. (4) Symmetric buckling with crack growth.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2087
buckling event. e
11
under consideration for symmetric buckling is 2.5 times larger than that of the constraint
specimen and 1.8 times larger for anti-symmetric buckling. The larger biaxial loading in the buckling cases
results in a reduced sizes of the plastic zone, thus decreasing the apparent toughness.
For the MSD specimen, the predictions of the crack growth behavior under the inuence of buckling are
shown in Figs. 8 and 9, on plots of the applied load and CTOA
I
vs. crack extension, respectively. The load
necessary for initiation of the major crack drops about 16% and the peak loads drop about 25% if the
specimen buckles symmetrically. Also, in the CTOA
I
Da plot, the crack tip interaction region slightly
moves towards large Da value.
Fig. 11. Deformed shape of MT specimens. (a) Symmetric buckling shape. (b) Anti-symmetric buckling shape.
2088 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
Finally, Fig. 15 presents a summary of the inuence of buckling eect on the load capacity. Results from
ve MT specimens (w 508 mm) with various crack lengths are given together with the MSD panel
results, in terms of the ratio of applied remote stress with buckling to applied remote stress without
buckling in dependence of the ratio of initial crack length to sheet thickness. For comparison, the linear
approximation from Kuhn and Figge [63] is shown, too. As the ratio a
i
=t increases, the inuence of
Fig. 12. Remote applied stress vs. crack extension for the Al 2024-T3 Alclad MT specimen for symmetric buckling, anti-symmetric
buckling and full in-plane constraint.
Fig. 13. The evolution of the CTOA vs. the crack extension for the Al 2024-T3 Alclad MT specimen for symmetric buckling, anti-
symmetric buckling and full in-plane constraint. (1) No buckling, CTOA
I
. (2) Symmetrical buckling, CTOA
I
. (3) Anti-symmetrical
buckling, CTOA
I
. (4) Anti-symmetrical buckling, CTOA
III
.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2089
buckling both on crack initiation and the maximum load carrying capacity increases. The linear prediction
of [63] is in good agreement with the present computational results for crack initiation. The peak stress
carrying capacity, however, decreases faster. For a
i
=t > 75, the initial trend is reversed and the reduction of
the peak load become small as a
i
=t further increases. This behavior is due to the crack growth that occurs
with the peak load, and the result of the interaction of a large crack with the specimen boundaries. Without
including crack growth in the analysis, such eects could not be accounted for.
Fig. 14. Distribution of e
11
in front of the crack at the peak load for the MT specimens for Al 2024-T3 Alclad with w 508 mm,
a
i
88:9 mm, for no buckling, anti-symmetric buckling and symmetric buckling.
Fig. 15. Normalized remote initiation and peak stress vs. normalized crack initial length, a
i
=t, for symmetric, anti-symmetric buckled
MT specimens and MSD specimens.
2090 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
4. Conclusion
Crack growth in thin sheet specimen of high strength aluminum alloys has been analyzed by a com-
putational method in which crack growth is determined within the deformation analysis of the specimens
under consideration. This approach becomes possible since a cohesive constitutive relation considers the
processes of material separation directly within the computations. The materials under investigation are
thus described by their elasticplastic properties, and in addition, by the parameters describing material
separation at the crack tip. The current approach diers from other approaches to the analysis of crack
growth in thin sheet metal since it describes fracture by physical material separation quantities.
In the present investigation, an attempt is made to choose the cohesive material parameters from global
measurement results and micromechanical damage model considerations to the analysis. Without em-
ploying a tting procedure, the current approach to the cohesive zone parameter selection generally results
in a good agreement between numerically predicted and experimentally determined data. Discrepancies
between computations and experiments were observed only for small amounts of crack extensions in the
constrained Al 2024-T3 Alclad panels, Fig. 3. An even closer t between experiment and prediction could
be obtained by further adjustment of the cohesive zone material parameters.
The current investigation predicts the experimentally determined evolution of crack-tip-opening angles in
good agreement with experiments. Since a CTOA-criterion is more readily available in commercial software
packages, the present approach could provide values of the critical CTOA being used in subsequent struc-
tural analyses. However, the assumption of a constant CTOA does not seem valid, especially in the case
where a multi-site-damaged specimen is investigated. In the regime of crack tip interaction via a short re-
maining ligament, the present approach predicts a change from a crack-like behavior to a situation of ho-
mogeneous debonding in agreement with results in [43]. In this case the crack-tip-opening angle drops to zero.
The use of a CZM diers from the application of the Dugdale-model since it does not lump the eects of
plasticity and material separation into a single strip. In comparison with the results of [18], this is expressed
in the dierent values of the parameters characterizing the maximum stress carrying capacity within the two
models. While the eective yield stress for the present material used in the Dugdale model in [18] is ap-
proximately 1.2r
0
, the cohesive strength here is considerably larger, namely 2r
0
.
The inuence of constraint eects due to strain hardening or specimen thickness has not been addressed
so far, but could be considered by the use of CZM in which the cohesive strength and energy are dependent
on constraint [64].
To further improve the predictive capabilities of the approach presented here, more validation with
experimental results is necessary. These further studies will especially have to take account of the eects of
dierent specimen sizes and congurations. It is strongly believed that the application of cohesive zone
elements to these problems holds great promise to nd solutions to problems of structural integrity in thin-
walled structure.
Acknowledgements
Support by Purdue Research Foundation via fellowship to W.L. and by a grant (award number:
MSS000003N) from NCSA at UIUC is gratefully acknowledged. The authors thank Prof. R.H. Dodds for
supplying experimental data and discussion.
References
[1] Riks E, Rankin CC, Brogan FA. The buckling behavior of a central crack in a plate under tension. Eng Fract Mech
1992;43(4):52948.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2091
[2] Markstrom K, Storakers B. Buckling of cracked members under tension. Int J Solids Struct 1980;16:2179.
[3] Barut A, Madenci E, Britt VO, Starnes JH. Buckling of thin, tension-loaded, composite plate with an inclined crack. Eng Fract
Mech 1997;58(3):23348.
[4] Seshadri BR, Newman JC. Analyses of buckling and stable tearing in thin-sheet materials. NASA/TM-1998-208428.
[5] Harris CE, Newman JC, Piascik RS, Starnes JH. Analytical methodology for predicting widespread fatigue damage onset in
fuselage structures. J Aircraft 1998;35:30717.
[6] Starnes JH, Rose CA. Buckling and stable tearing response of unstiened aluminum shells with long cracks. AIAA paper 98-1991.
[7] Proceedings of the FAA-NASA Symposium on Continued Airworthiness of Aircraft Structures. 1997.
[8] ASTM E561-98. Standard practice for R-curve determination. Annual Book of Standards. 1998. p. 494506.
[9] May GB, Kobayashi AS. Plane stress stable crack growth and J-integral/HRR eld. Int J Solids Struct 1995;32(67):85781.
[10] Omori Y, Kobayashi AS, Okada H, Atluri SN, Tan PW. T-epsilon* integral as a crack growth criterion. Mech Mater 1998;28(1
4):14754.
[11] Rose CA, Young RD, Starnes JH, Nonlinear local bending response and bulging factors for longitudinal cracks in pressurized
cylindrical shells. AIAA paper 99-1421.
[12] Hutchinson JW, Evans AG. Mechanics of materials: top-down approaches to fracture. Acta Mater 2000;48(1):12535.
[13] Wells AA. Application of fracture mechanics at and beyond general yielding. Br Welding J 1963;11:56370.
[14] Hellmann D, Schwalbe KH. Geometry and size eects on JR and dR curves under plane stress conditions. In: Sanford RJ,
editor. Proceedings 15th National Symposium on Fracture Mechanics ASTM STP 833. Philadelphia: American Society for
Testing of Materials; 1984. p. 577605.
[15] Dawicke DS, Newman JC, Bigelow CA. Three dimensional CTOA and constraint eects during stable crack tearing in a thin-sheet
material. In: Reuter WG, Underwood JH, Newman JC, editors. Fracture mechanics, vol. 26 ASTM STP 1256. Philadelphia:
American Society for Testing of Materials; 1995. p. 22342.
[16] Seshardri BR, Newman JC. Analyses of buckling and stable tearing in thin-sheet materials. In: Panontin TL, Sheppard SD,
editors. Fatigue and fracture mechanics, ASTM STP 1332. Philadelphia: American Society for Testing of Materials; 1998.
[17] Chen CS, Wawrzynek PA, Ingraea AR. A methodology for fatigue crack growth and residual strength prediction with
applications to aircraft fuselages. Comput Mech 1997;19(6):52732.
[18] Deng X, Hutchinson JW. Approximate methods for analyzing the growth of large cracks in fatigue damaged aircraft sheet
material and lap joints. Finite Elem Anal Des 1996;23:10114.
[19] Nilsson KF, Hutchinson JW. Interaction between a major crack and small crack damage in aircraft sheet material. Int J Solids
Struct 1994;31(17):233146.
[20] Sutton MA, Deng X, Ma F, Newman Jr JC, James M. Development and application of crack tip opening displacement-based
mixed mode fracture criterion. Int J Solids Struct 2000;37(26):3591618.
[21] Noor AK. Computational structures technology: leap frogging into the twenty-rst century. Comput Struct 1999;73(15):1
31.
[22] Gurson AL. Continuum theory of ductile rupture by void nucleation and growth, Part IYield criteria and ow rules for porous
ductile media. J Eng Mater Tech 1977;99:215.
[23] Rousselier G. Ductile fracture models and their potential in local approach of fracture. Nucl Eng Des 1987;105:97111.
[24] Xia L, Shih CF, Hutchinson JW. A computational approach to ductile crack growth under large scale yielding conditions. J Mech
Phys Solids 1995;43(3):389413.
[25] Mathur KK, Needleman A, Tvergaard V. Three dimensional analysis of dynamic ductile drack growth in a thin plate. J Mech
Phys Solids 1996;44(3):43964.
[26] Eckstein A, Basar Y. Ductile damage analysis of elasto-plastic shells at large inelastic strains. Int J Numer Meth Eng
2000;47:166387.
[27] Baaser H, Gross D. Crack analysis in ductile cylindrical shells using Gursons model. Int J Solids Struct 2000;37:7093104.
[28] Brocks W, Besson J, Chabanet O, Steglich D. Ductile rupture of aluminum sheets. In: Ravi-Chandar K et al., editors. Proceeding
of ICF 10. 10th International Conference of Fracture. Honolulu: 2001.
[29] Gologanu M, Leblond JB, Devaux J. Approximate models for ductile metals containing nonspherical void-case of axisymme-
terical prolate ellipsoidal cavities. J Mech Phys Solids 1993;41(11):172354.
[30] Pardoen T, Hutchinson JW. An extended model for void growth and coalescence. J Mech Phys Solids 2000;48(12):2467512.
[31] Kikuchi M. Study on the eect of the crack length on the J
IC
value. Nucl Eng Des 1997;174(1):419.
[32] Broek D, Thomson D, Jeong DY. Testing and analysis of at and curved panels with multiple cracks. FAA-NASA International
Symposium on Advanced Structural Integrity Methods for Airframe Durability and Damage Tolerance, NASA Conference
Publication 3274, Part I, September 1994. p. 8598.
[33] Gullerud AS, Dodds RH. Three-dimensional modeling of ductile crack growth in thin sheet metals: computational aspects and
validtion. Eng Fract Mech 1999;63:34774.
[34] Dugdale DS. Yielding of steel sheets containing slits. J Mech Phys Solids 1960;8:1004.
[35] Barenblatt GI. The mathematical theory of equilibrium cracks in brittle fracture. Adv Appl Mech 1962;7:55129.
2092 W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093
[36] Rice JR. Mathematical analysis in the mechanics of fracture. In: Liebowitz H, editor. Fracture. New York: Academic Press; 1968.
p. 191311.
[37] Needleman A. A continuum model for void nucleation by inclusion debonding. J Appl Mech 1987;54:52531.
[38] Needleman A. An analysis of decohesion along an imperfect interface. Int J Fract 1990;42:2140.
[39] Tvergaard V, Hutchinson JW. The relation between crack growth resistance and fracture process parameters in elasticplastic
solids. J Mech Phys Solids 1992;40:137797.
[40] Xu X-P, Needleman A. Numerical simulation of fast crack growth in brittle solids. J Mech Phys Solids 1994;42:1397434.
[41] Siegmund T, Fleck NA, Needleman A. Dynamic crack growth normal to a bimaterial interface. Int J Fract 1997;85:381402.
[42] Siegmund T, Needleman A. A numerical study on dynamic crack growth in elasticviscoplastic solids. Int J Solids Struct
1997;43(7):76987.
[43] Pandol A, Ortiz M. Solid modeling aspects of three-dimensional fragmentation. Eng Comput 1998;14(4):287308.
[44] Lin G, Cornec A, Schwalbe K-H. Three-dimensional nite element simulation of crack extension in aluminum alloy 2024 FC.
Fatigue and Fracture of Eng Mater Struct 1998;21:115973.
[45] Rahul-Kumar P, Jagota A, Bennison SJ, Saigal S, Muralidhar S. Polymer interfacial fracture simulations using cohesive elements.
Acta Mater 1999;47(1516):41619.
[46] Roy YA, Dodds RH. Simulation of ductile crack growth in thin aluminum panel using 3-D surface cohesive elements. Int J Fract
2001;111(1):2145.
[47] Faleskog J, Shih CF. Micromechanics of coalescenceI. Synergistic eects of elasticity, plastic yielding and multi-size-scale voids.
J Mech Phys Solids 1997;45(1):2150.
[48] Hom CL, McMeeking RM. Large crack tip opening in thin elasticplastic sheets. Int J Fract 1990;45:10322.
[49] Newman JC, Crews JH, Bigelow CA, Dawicke DS. Variations of a global constraint factor in cracked bodies under tension and
bending loads. In: Kirk M, Bakker A, editors. Constraint eects in fracture, theory and applications, vol. 2 ASTM STP 1244.
Philadelphia: American Society for Testing of Materials; 1995. p. 2142.
[50] Lloyd WR, Piascik RS. Three-dimensional crack growth assessment by microtopographic examination. In: Reuter WG,
Underwood JH, Newman JC, editors. Fracture mechanics, vol. 26 ASTM STP 1256. Philadelphia: American Society for Testing
and Materials; 1995. p. 30318.
[51] Haynes MJ, Ganglo RP. High resolution R-curve characterization of the fracture toughness of thin sheet aluminum alloys.
J Testing and Evaluation 1997;25:8298.
[52] Schwalbe K-H. Inuence of stress state on static crack growth in AlZnMgCuO5. Eng Fract Mech 1977;9:55783.
[53] Kamat SV, Hirth JP. Mixed mode I/III fracture toughness correlation. Scr Metall Mater 1994;30(2):1458.
[54] Jones RH, Li H, Hirth JP. Eects of mixed-mode I/III loading on environment-induced cracking. Eng Fract Mech 2001;68(6):789
801.
[55] Camplin J, Zakrajsek J, Srivatsan TS, Lam PC, Manoharan M. Inuence of notch severity on the impact fracture behavior of
aluminum alloy 7055. Mater Des 1999;20:33141.
[56] Manoharan M. Development of a mixed-mode fracture mechanism map and its extension to mixed-mode impact fracture. In:
Madhidhara RK et al., editors. Recent Advances in Fracture. The Minerals, Metals & Materials Society, 1997. p. 37384.
[57] Feng X, Kumar AM, Hirth JP. Mixed mode I/III fracture toughness of 2034 aluminum alloys. Acta Metal Mater 1993;41(9):2755
64.
[58] ABAQUS Users Manual, Version 5.8, Hibbitt, Karlson and Sorensen, Inc. 1998.
[59] Needleman A. Numerical modeling of crack growth under dynamic loading conditions. Comput Mech 1997;19:4639.
[60] Siegmund T, Brocks W. The role of cohesive strength and separation energy for modeling of ductile fracture. In: Jerina KL, Paris
PC, editors. Fatigue and fracture mechanics, ASTM STP 1360. Philadelphia: American Society for Testing of Materials; 1999.
[61] Schijve J. Multiple-site damage in aircraft fuselage structures. Fatig Fract Eng Mater Struct 1995;18(3):32944.
[62] Hill JC, Bennison SJ, Klein PA, Jagota A, Saigal S. Co-planar crack interaction in cleaved mica. In: Ravi-Chandar K, Karihaloo
BL, et al., editors. Proceedings of ICF10. 10th International Conference on Fracture. Honolulu: 2001.
[63] Kuhn P, Figge IE. Unied notch-strength analysis for wrought aluminum alloys, NASA TN D-1259, 1962.
[64] Siegmund T, Brocks W. A numerical study on the correlation between the work of separation and the dissipation rate in ductile
fracture. Eng Fract Mech 2000;67:13954.
W. Li, T. Siegmund / Engineering Fracture Mechanics 69 (2002) 20732093 2093

You might also like