X-FEM for Abaqus (XFA) Toolkit for Automated

Crack Onset and Growth Simulations

Jay Shi1, Jim Lua2, Haim Waisman3, Phillip Liu4,
Ted Belytschko5, and N. Sukumar6
Global Engineering and Materials, Inc.
9 Goldfinch Terrace
East Lyme, CT 06333
1
[email protected]
2
[email protected]
3
[email protected]
4
[email protected]
5

Northwestern University
[email protected]
6

University of California at Davis

[email protected]

Abstract: A software tool for automated crack onset and growth simulations based on the
eXtended Finite Element Method (X-FEM) has been developed. For the first time, this tool is able
to simulate arbitrary crack growth and composite delamination without remeshing. The
automated tool is integrated with Abaqus/Standard and Abaqus/CAE via the customization
interfaces. It seamlessly works with the Commercial, Off-The-Shelf (COTS) Abaqus suite. Its
unique features include: 1) CAE-based insertion of 2D or 3D multiple cracks with arbitrary shape
of crack front independent of an existing mesh; 2) simulation of crack growth inside or between
solid elements, and potentially along a shell/solid interface or along a shell/shell interface; 3)
simulation of non self-similar crack growth along an arbitrary path or a user-specified interface;
4) extraction of near tip strain energy release rates via the modified VCCT; and 5) CAE-based
data processing and visualization. The levelset method coupled with X-FEM is used to enrich the
displacement field with jump and asymptotic near-tip solutions and track the crack geometry as it
grows. A penalty-based formulation is developed within the UEL framework to simulate crack
closure and frictional contact. To account for energy dissipation associated with the frictional
contact, a modified VCCT approach is employed for an arbitrary crack front element with a
partial contact zone. A fracture front tracking and levelset update module is used for either a userspecified growth size or a Paris-type fatigue law. Both the validity and applicability of the toolkit
have been demonstrated via numerical examples.

Keywords: X-FEM, VCCT, Fatigue, Fracture & Failure, Crack Growth, Delamination,
Abaqus/Standard, UEL

1.

Introduction

2008 Abaqus Users Conference

The fracture failure mode is particularly significant in the damage tolerance of advanced
composites or metallic joints, since manufacturing flaws and in-service damage most often
manifest themselves as initial cracks. To assess the criticality of an initial flaw and its impact on
the residual strength and life, a finite element analysis is performed, typically for various crack
configuration, location, shape, and size. Nevertheless, conventional approaches are severely
limited as any change in the geometrical characteristics of cracks requires re-meshing the finite
element domain. This is a sever restriction and burdensome both for 2D and 3D analysis,
especially for crack growth problems with complex geometries. Moreover, the convergence rate of
conventional FEM is dominated by the nature of the solution near the point of singularity, which
requires appropriate crack-tip singular elements to accurately represent cracks in FEM
approximations.
To alleviate the computational burden associated with insertion of arbitrary cracks into an FEM
model without conforming to its mesh, the extended finite element method (X-FEM) has recently
been developed by Ted Belytschko and his research group (Belytschko, 1999, Moes, 1999, and
Sukumar 2000). The X-FEM enables the modeling of stationary and growing cracks in solids
subjected to static, dynamic, and fatigue loading. In essence, the idea is to enrich the finite element
solution space with a Heaviside step function for displacement discontinuity and asymptotic
branch functions for near tip singular fields. This has been shown to greatly improve convergence
rates and in particular separate the geometry of the cracks from its finite element mesh.
We employ the levelset method coupled with X-FEM to represent the jump and asymptotic
displacement enrichments and track the crack geometry as it grows. For example, a 3D crack can
be defined by two almost-orthogonal levelsets that are derived from the signed distance functions.
One function describes the crack surface (plane) in a 3D space and the other is used to describe the
crack front. With this implicit description of the crack front, arbitrary evolving crack can be
modeled without explicit geometric representation of the crack. Once the crack starts to propagate,
the levelset values are updated, by either solving the governing hyperbolic equations or reinitializing the levelsets by calculating the signed distance functions for the new crack location.
Under the Air Force funded SBIR program for residual strength and delamination failure
prediction of composite structure GEM has successfully developed an add-on X-FEM software
toolkit that works seamlessly with the common commercial, off-the-shelf (COTS) version of
Abaqus/Standard and Abaqus/CAE for an automated crack onset and growth. Our partner LM
Aero is currently using a 2D version of the toolkit for damage tolerance analysis of composite
joints. We will release the 3D version to public by summer of 2009. The main XFA tool features
are: 1) arbitrary insertion of multiple initial cracks that are independent of the finite element mesh;
2) characterization of a growing crack without remeshing; 3) accurate extraction of Strain Energy
Release Rate (SERR) using the enriched near-tip displacement field via the modified Virtual
Crack Closure Technique (VCCT); and 4) characterization of crack closure via a penalty contact
algorithm.
The results of this R&D will have significant benefits in military and commercial industries.
Specific applications of the software include structural integrity and damage tolerance design,
evaluation of structural integrity and damage tolerance requirements to reduce reliance on testing,
optimization of design concepts and reduction of cost associated with multiple configuration
component testing, exploration of innovative manufacturing and/or composite bonding
technologies for composite or metallic joints, and specification of damage tolerance design
guidelines, inspection and repair requirements.
2

2008 ABAQUS Users Conference

2.

XFA Tool Architecture

The XFA tool includes three major components: an interactive crack definition via Abaqus/CAE
and X-FEM pre-processing, a solver via Abaqus/Standard, and data processing and visualization
via Abaqus/Viewer. The integration with Abaqus suite is illustrated in Figure 1. The pre-processor
is composed of the following modules: 1. the GUI module, which provides graphic user interfaces
for the user to define initial cracks in an existing FEM model; 2. the X-FEM initialization module,
which calculates the X-FEM attributes based on the user-specified crack geometry and fracture
mechanics parameters; and 3. the X-FEM update module, which updates X-FEM attributes
including levelsets and enrichments and is based on the crack configuration at the previous and
current increment.

ABAQUS/CAE
ABAQUS/CAE
kernel
kernel

Python
Python
interpreter
interpreter

Commands

Update messages

WIP messages
ODB files
Input
files

ABAQUS
ABAQUS
analysis
analysis

Python
Python
Embedded
Embedded
Data
DataConverter
Converter

Kernel
Kernel
Customization
Customization
Scripts
Scripts

Solver
Solver
Customization
Customization
Subroutines
Subroutines

GUI
GUI
Customization
Customization
Scripts
Scripts

ABAQUS
ABAQUS
GUI
GUI
Command
Command
line
line
interpreter
interpreter
Python
Python
interpreter
interpreter

Figure 1. XFA Tool Integration with Abaqus Suite via Its Customization Interfaces
The XFA solver is built on a series of Abaqus user subroutines: UEL, UMAT, UEXTERNALDB,
and SIGINI. The X-FEM parameters, such as crack geometric data, levelsets, and nodal
enrichment types are passed from Abaqus/CAE using additional input file for the XFA solver. The
element operator is implemented in UEL and UMAT. Some post-processing modules, such as the
construction of sub-elements for visualization and VCCT for SERRs extraction are also in the
UEL. The X-FEM update module is implemented within UEXTERNALDB. In order to visualize
X-FEM subdomain data, a post-processing module is developed within UEXTERNALDB to
generate Abaqus output database using its ODB API.
As shown in Figure 2, at the beginning of each load increment, the X-FEM update module is
called to process crack growth information and update the X-FEM parameters. The process then
goes into the element loop and enters UEL by checking the element type ID. Upon convergence of
iterations both the VCCT module and the crack growth module are called from within
UEXTERNALDB.

2008 Abaqus Users Conference

Start of analysis
XFEM updates

Growth?

No

ABAQUS built-in
element types

Load Increment

End of analysis

No
XFEM?

XFEM element integration

Crack growth criterion

Material point operation

No

Stress recovery &VCCT

Contact?

Penalty contact

Figure 2. XFEM Flowchart for ABAQUS/Standard

For the 2D case the linear plane stress/strain quadrilaterals (CPS4, CPE4) are implemented and for
the 3D case linear hex and tetra solids (C3D8, C3D4) will be implemented. For shell-shell or
shell-solid interfaces an enriched interface element will also be developed to allow arbitrary
shapes of crack front. Currently we support isotropic and orthotropic elastic materials and elastoplastic materials with isotropic hardening and eventually most materials for solid elements will be
supported. For the sub-domain outside of the X-FEM region, any default element types and
materials can be used.

3.

Penalty Contact in X-FEM Elements

A penalty-based formulation is developed inside the element operator to simulate crack closure
and frictional contact. Considering these effects is critical in load cases involving shear modes.
One approach to simulate crack closure behavior is to apply the so-called global-local iterative
algorithm proposed by (Dolbow, 2001). The method may exhibit slow convergence rate when the
number of contact pairs become large. This issue is mainly due to the displacement jump and
contact force that are solved in a local system and the unknowns do not appear in the global
system of equations, ie, the contact constraints are decoupled. To enhance the solution accuracy
and numerical efficiency, we have developed the direct approach where displacement jumps of
contact surfaces are naturally interpolated by the extended displacement representation via XFEM. Furthermore, a penalty-based formulation is developed so there is no Lagrange multiplier to
take extra degrees of freedom and enter the system of equations via user elements. Following this
approach, contact discontinuity can be resolved during the equilibrium iterations.
The contact surface is defined based on nodal levelsets. The discretization of the contact surface is
achieved by assigning the accumulation points along the surface segments. For each crack
4

2008 ABAQUS Users Conference

segment the contact condition with a frictional law is examined, which gives the additional
stiffness terms and RHS. The element stiffness matrix and the residual force vector are updated for
the next iteration. It should be noted that the condition of zero relative motion is approximated by
a stiff elastic model. The model stiffness for the frictional condition is chosen such that the relative
motion from the position of zero shear stress is bounded by a value, or, the allowable maximum
elastic slip.

4.

Modified VCCT

The modified VCCT is implemented to estimate the energy release rates. For the case that the
crack fronts (tips) coincide with the element nodes, the algorithm can be summarized as the
following:
1. Extend jump enrichments to the node in front of the tip. The elements connecting the
tip node and the newly-enriched node become cut-through type. The tip node now
contains displacement jump. The artificial crack surface between the tip node and the
newly-enriched node can be treated as crack processing zone or cohesive zone.
2. Apply bonding constraint to the cohesive zone such that the displacement jump at the
tip is minimized. This constraint is enforced using a penalty method. When the
displacement jump is ignorably small, the penalty force can be regarded as the traction
force in front of tip. The resultant force at the tip node is the equivalent to the reaction
force in the conventional VCCTs spring double-node model.
3. Multiply the nodal force with the crack opening displacement behind the crack and at
the distance equivalent to the size of cohesive zone to obtain the total energy released.
A variant of this method can also be applied to solve internal crack problems. The actual
implementation of the VCCT is analogous to the contact mechanics algorithm mentioned
previously.

5.

XFA Tool Validation

5.1
Validation 1: Edge-cracked plate under surface traction (Liu, 2004) to verify
the accuracy of stress intensity factors estimation using the modified VCCT in XFA
toolkit for an internal crack
An edge-cracked plate, as shown schematically in Figure 3, is fixed at the bottom and subjected to
a uniform shear stress of magnitude 1.0 unit along the top edge. The plate has a length (L) of 16
units, width (W) of 7 units, and a crack length (a) of 3.5 units. A plane strain condition is assumed
with a Poissons ratio of 0.3 and elastic modulus of 100 units. A uniform mesh consisting of
2347 quad elements is created. The resulting respective diagonal length of an element is 0.457
units.

2008 Abaqus Users Conference

(a)
(b)
(c)
Figure 3. Edge-cracked plate under shear: (a) geometry and loads, (b) FEM model
with the 2347 mesh, and (c) XFA simulation result
The reference solution provided in the paper is K1=34.0 and K2 = 4.55. Our estimation using
VCCT gives K1 and K2 of 34.3 and 4.51, respectively. The relative error as compared with the
reference solution is about 0.88%.
5.2
Validation 2: Bi-material interface central crack in a plate under tensile load
(Nagashima, 2003) to verify the VCCT procedure for bi-material interface
In this test we consider the following model with bi-orthotropic interface crack: a rectangular plate
with a bi-material center crack subjected to a surface traction, as shown in Figure 4. The width 2W
and length 2H of the plate is given by 100 and 200 mm, respectively.

0.10 inch

Mat 1

2a
0.10 inch

Mat 2

0.10 inch

Plane strain (CPE4) small deformation

Crack Size 2a = 0.040 inch
Mesh Size 50 x 100
Surface traction x = 100 psi (Mode II)
(additional y = 200 psi for Mixed Mode)
Mat 1:
E-glass Vinylester Composite (Plain Weave)
E1= 3.693 Msi, E3 = 3.237 Msi, E2 = 1.662 Msi
G13 = 0.705 Msi, G12 = 0.528 Msi, G23 = 0.528 Msi
v13 = 0.150, v12 = 0.400, v23 = 0.400
Mat 2:
Carbon fiber and epoxy resin (UD)
E1=23.35 Msi, E3=1.65 Msi, E2=1.65 Msi
G13=0.75 Msi, G23=0.57 Msi, G12=0.75 Msi
v13=0.325, v12=0.325, v23=0.45

Figure 4. Rectangular plate with a bi-material center crack

The XFA result is compared CAIs formulation (provided by LM Aero) and excellent agreement
is found as in Table 1.

2008 ABAQUS Users Conference

XFAs VCCT
CAIs VCCT
Left tip:
Displacement n 2x2.36208e-006
Displacement s 2x5.47384e-007
2.49299e+000
2.480
Traction n
1.29666e+000
1.283
Traction s
2.94432e-003
0.002934
GI
3.54884e-004
0.000353
GII
Right tip:
Displacement n 2x8.81282e-007
Displacement s 2x3.94411e-007
7.47925e-001
0.7436
Traction n
9.04461e-001
0.8945
Traction s
3.29566e-004
0.000328
GI
1.78365e-004
0.000178
GII
Table 1. Results using XFAs VCCT and CAIs VCCT.
5.3

Validation 3: DCB/ENF tests to verify VCCT for composite delamination

Both DCB/ENF analyses and tests have been performed extensively by research communities and
government labs. Both the material system and component geometry of DCB and ENF specimen
were defined by the Navy Lab in Carderock (NSWCCD). The composite material system is a
plain weave laminated composite with E-glass fiber and vinyl ester resin. Both the VCCT in
Abaqus and the cohesive model for LS-DYNA have been used by NSWCCD and GEM,
respectively, for response and delamination onset prediction.
The 3D orthotropic material properties and model parameters are defined in Figure 5 below. Using
the same material and model parameters, the X-FEM based delamination tool is used to compute
the SERRs at the onset of delamination growth. The X-FEM results are compared with the VCCT
and cohesive model predictions.
Orthotropic Material Model
E1=3.693e+06 psi; E2=3.237e+06 psi; E3=1.662e+06 psi; G12=0.705e+06 psi
G13=0.528e+06 psi; G23=0.528e+06 psi, 12=0.15; 13=23=0.4
DCB - Specimen

BCs (ux=0.0; uz=0.338 in)

P

Z(3)

Uz=0.338 in

0.231 in

X (1)
2.151 in
5in
BCs (ux=uz=0.0)

Figure 5. Problem statement of DCB specimen

2008 Abaqus Users Conference

The mesh size used in the XFEM model is the same as used by NSWCCD for their Abaqus +
VCCT model with the mesh size of 0.012 inch. The GI prediction from Abaqus VCCT and the
LS-DYNAs cohesive model is 4.77 in lb/in2 and 4.785 in lb/in2, respectively. The applied load
measured from tests is P = 20 lb at the onset of delamination growth.
The response prediction from the XFA tool is given below. The following Figure 6 show the
Mises stresses of the DCB.

Figure 6. Mises stress of the DCB specimen using the XFA tool
The X-FEM predicts GI = 4.765 in lb/in2 and GII = 2.33e-04 in lb/in2 with the reaction force of
21.67 lb. An excellent prediction in GI has been achieved from the XFEM tool simulation of the
DCB specimen. Given the dominant Mode I characteristic in the DCB specimen, the predicted GII
is extremely small.
The setup and modeling parameters of the ENF test coupon is the shown in Fgure 7. The measured
center deflection at the onset of Mode II delamination growth is 0.168 in. The corresponding load
at the delamination growth is 205 lb.
(uz=d; ux=0.0)

Z(3)

L=5 in; t=0.235 in; a1= 1.25 in

Center Deflection d=0.168 in
X(1)
Uz=0.0
Ux=uz=0.0

Figure 7. Problem Statement of ENF Specimen

The GII predicted from NSWCCDs VCCT and GEMs cohesive model is 6.80 in lb/in2, and 6.90
in lb/in2, respectively. The GII prediction from the XFA tool is 6.61 in lb/in2, which is slightly
lower than the VCCT prediction (3%). In addition, the failure load predicted from the XFA tool is
211.5 lb, which slightly high than the measured failure load of 205 lb. Again the excellent
agreement in load and SEER prediction has demonstrated the validity and capability of the XFA
tool in characterizing Mode I and II. The Mises stress contour of the ENF specimen under zoomout and zoom-in configurations is given in Figure 8 below.

2008 ABAQUS Users Conference

Figure 8. Bending stress distribution of the ENF specimen

5.4
Validation 4: 3D cube under mixed mode loading to verify VCCT compared
to the conventional spring double-node model in Abaqus
A simple model that contains a 1x1x1 unit cube and 800 brick elements with elastic material is
considered, as illustrated in Figure 9. The cube is loaded with surface traction in all directions over
the top and bottom surfaces. The side surface in front of the crack front is completely fixed.

Figure 9. Illustration of mixed mode loading and its corresponding displacement

In Table 2 the displacement jumps and the concentrated forces at all 9 tip nodes are compared with
the conventional, spring double-node model using Abaqus. Again the excellent agreement
between the results is found.
Node

Fx
(Abaqus)

1
2
3
4
5

-3.431
-5.316
-3.944
-2.850
-1.884

Fx
(XFA)
3.4309e+000
5.3161e+000
3.9441e+000
2.8501e+000
1.8843e+000

2008 Abaqus Users Conference

Fy
(Abaqus)
2.544
5.456
5.781
5.853
5.875

Fy
(XFA)
2.5437e+000
5.4564e+000
5.7808e+000
5.8534e+000
5.8754e+000

Fz
(Abaqus)
0.7420
1.179
1.312
1.427
1.474

Fz
(XFA)
7.4199e-001
1.1785e+000
1.3119e+000
1.4274e+000
1.4743e+000
9

6
7
8
9

-0.9440
9.4396e-001
5.853
5.8534e+000
1.370
1.3701e+000
5.664E-02 5.6647e-002
5.781
5.7808e+000
1.144
1.1439e+000
1.212
1.2117e+000 5.456
5.4564e+000
0.8245
8.2452e-001
1.145
1.1454e+000 2.544
2.5437e+000
0.4071
4.0714e-001
Table 2. Forces extracted from XFA and the Abaqus double-node model

5.5
Validation 5: Non-planar growth of an inclined penny-shape crack to verify
the capability of nonplanar crack growth
Finally, to illustrate our 3D crack growth capability, a penny shaped crack in a 3D elastic body is
analyzed as shown in Figure 10. We consider an inclined penny crack ( = 45 degrees) that is
under uniaxial tension. Since the loading at the crack front is of mixed-mode, the crack kinks. In
this model a 17-thousand-node mesh is used. The snapshots of crack propagation at t=0.00, 0.33,
0.67, and 1.00 are plotted, respectively, in Figure 10. The simulations reveal the non-planar
character of the growth and the results illustrate that our crack update algorithm is able to capture
the representation of the crack at each step.

Figure 10. Nonplanar growth of the penny-shape crack under simple tension

6.
XFA tool application: Fatigue crack growth of aluminum T-joint
and fatigue life prediction
T-weldments are widely used components in various engineering structures. Its general structure
contains three regions: base, web and welding. Typically the base and web are connected by the
welding while each component is made of different materials. The dimensions of our T-weldment
model are based on (Latorre, 2002), which is a T-stiffner used in hull panels of ship structures.
Material heterogeneity is an important concern in the T-weldment fatigue analysis. The welding
process causes hot spots at the weldment toe which sometimes reaches a temperature of over 800
K. Often the temperature ramps down to the room temperature on the other side of the panel due
to high heat conductivity. Phase transformations in the weldment and the Heat Affected Zone
(HAZ), and the selection of different base, web, and weldment materials are eventually the factors
leading to a significant amount of residual stresses near the welding zone. The size of the HAZ
can only be determined empirically. The final geometry of the simplified T-joint model and the
mesh is shown in Figure 11. The component is simply supported with the concentrated load, P
equal to 55N, which is applied on a reference point. This reference point is then rigidly coupled to
the top-end surface of the web.

10

2008 ABAQUS Users Conference

Load P
6.25

Unit: mm

Simplification

35.0
15.63
BC 1
Fix U,V

7.813

BC 2
Fix U

6.25
37.5
100

Figure 11. The XFA model of aluminum T-joint specimen

In this example, the base and the web are aluminum alloy 5083 and the weldment is aluminum
alloy 5183. The properties of both alloys as well as the HAZ are adopted from (Galatolo, 1997).
A Ramberg-Osgood model, typically used for aluminum, is used for each material component
with material parameters obtained from experimental data. The material parameters for each
region are listed in the Table 3.

= +
E
0.2

E (MPa)

0.2 (Mpa)

Base/Web(A5083)
Welding(A5183)

70000
70000

11.8
11.8

0.000278
0.000491

183.4
183

HAZ

70000

7.17

0.014

189.5

base/web

welding

HAZ

Table 3. Material parameters for T-joint base, welding, and HAZ regions
2008 Abaqus Users Conference

11

The static analysis of the uncracked model suggests that it is a high-cycle fatigue problem with a
maximum equivalent plastic strain (PEEQ) less than 0.2%; therefore, a simple, elasto-plastic with
isotropic hardening UMAT subroutine is developed for the user element domain and the
equivalent material model (*PLASTIC) is specified for the rest of the model using the built-in
element type: CPE4. It should be noted that for X-FEM elements, both the location and number of
Gauss points between load increments may be changed due to the propagating crack and hence
material state variables must be updated consistently. For all analyses small deformation theory is
assumed.
The residual stress distribution is affected by the material heterogeneity, the component geometry,
and the temperature field. Generally it is very difficult to obtain accurate distribution by
simulation that considers all influential factors. Nevertheless, it is common to use measured
distribution from experiments. For instance, in (Mann, 2006) the residual stress components of a
similar aluminum structure are measured. The tensile stresses are found near the weldment toe
region, which then turn to compressive stresses away from the toe, and rapidly attenuate to zero
further away. To simulate a similar residual stress distribution, the user subroutine SIGINI is
developed to calculate the initial stress distribution at Gauss points. Since such a superimposed
stress distribution is unbalanced, a static step is first performed to obtain self-balanced
distribution, which is then followed by the actual, fatigue analysis. The self-balanced residual
stress distribution is shown in Figure 12. In the same figure, the simulated stress profile starting
from the weldment toe and along the base panel is plotted. The general profile matches well with
the experimental result in (Mann, 2006).

weldment
toe

weldment
toe

Figure 12. Residual stress profile introduced at the beginning of fatigue analysis
It is usually assumed that the fatigue crack is initiated where the maximum Mises stress, or the
hot spot stress, is located. This assumption is supported by experimental observation such as in
(Latorre, 2002). More interestingly, as shown in Figure 13, the fatigue crack starts at the weldment
tip not only because the stress is magnified but also because the weld filler tends not to properly
fuse with the base metal and introduces the cold lap. The direction of the initial crack is assumed

12

2008 ABAQUS Users Conference

to normal to the maximum principal stress at this location. An initial crack size of 0.5mm is
assumed.
Initial 0.5mm crack

29.50

Smax
Experimental Observation on
the Crack Initiation Site

Initial crack location

Figure 13. The fatigue crack starts at the weldment tip not only because the stress
is magnified but also because the weld filler tends not to properly fuse with the
base metal and introduces the cold lap
To consider the effect of residual stresses, a variant of the Paris fatigue law, is used. The total
time-to-failure in the fatigue analysis step is treated as the number of cycles, dN . At the end of
each increment the range of stress intensity factor (SIF), K is computed at the crack tip. The
stress ratio, R , is then calculated as the ratio of the residual stress component at the crack growth
normal direction to the stress component along the same direction and under the full load
condition. Once the modified SIF range and the stress ratio are obtained, the crack growth size,

da , can be estimated using the Walkers model: da / dN = C (KWal ker ) , where

K
KWal ker =
and the fatigue strength parameters: C = 1.24 E 14 , m = 4.01 , and
(1 R)1
= 0.6 , are adopted from (Brant, 1998). Similar to the initial crack orientation, the crack growth
m

direction is assumed normal to the maximum principal stress and is estimated at the end of each
increment. The component is considered failure once the total crack size has reached the critical
size of 2.0 mm. The fatigue life as the number of cycles is equivalent to the step time when the
critical size is reached. It is noted that the X-FEM elements are used only in the region around the
initial crack. In Figure 14 a few snapshots of the Mises stress in the X-FEM region and
corresponding nodal enrichment type are shown.

2008 Abaqus Users Conference

13

Figure 14. Snapshots of the Mises stress in the X-FEM region and corresponding
nodal enrichment types
It is obvious that by using X-FEM, the crack growth path is completely independent on the mesh.
The tip can move arbitrarily in the region and no remeshing is needed. For this particular model
the component fails with final crack size equal to 2.0194 mm and the stress ratio equal to 0.4808.
The number of cycles during fatigue crack growth is 4.58E+06.
To visualize the crack opening in its deformed shape, a post-processing module is developed as
part of the XFA toolkit to slice the crack-intersected elements based on the levelset information
and the enriched degrees of freedom. This kind of element slicing is for visualization purpose
only. The process has nothing to do with local remeshing because the resulting subelements do not
exist during the analysis phase. The final snapshot of Mises stress and the crack opening are
shown in figure 15.

Figure 15. Mises stress and crack opening at failure

14

2008 ABAQUS Users Conference

7.

Conclusion

Global Engineering & Materials, Inc has developed an add-on X-FEM toolkit for Abaqus (XFA)
that works seamlessly with the common commercial, off-the-shelf (COTS) version of Abaqus
software suite for an automated crack onset and growth prediction analysis. During the
development process, rigorous validation has been made to ensure the tool is robust, accurate and
generates high fidelity predictions in the fracture and fatigue analysis The tool has great potential
to be used in various designs and analysis practices such as the one illustrated in the last example.
We have scheduled to release this tool by summer 2009. Support from the Air Force SBIR
program and our industry partners, including SIMULA of Dassault Systems, LM Aero, and Bell
Helicopter are highly acknowledged.

8.

References

1. Belytschko T, Black T, Elastic crack growth in finite elements with minimal remeshing,
International Journal for Numerical Methods in Engineering, 45, 601-620, 1999.
2. Brandt, Lawrence, and Sonsino, The fatigue behavior of 5 and 25 mm 5083 AluminumAlloy Weldments, Research Report, IIW Document XIII 1718 98.
3. Dolbow, J., N. Moes, and T. Belytschko, An Extended Finite Element Method for
Modeling Crack Growth with Frictional Contact, Comput. Methods Appl. Mech. Engrg.
190, 6825-6846, 2001.
4. Galatolo R. Fatigue crack propagation in residual stress fields of welded plates, Int. J.
Fatigue, 19, 43-49, 1997.
5. Latorre R. G., Stress analysis of a transversely loaded aluminium weldment, Marine
Structures, 15, 175191, 2002.
6. Liu, et al, XFEM for direct evaluation of mixed mode SIFs in homogenous and bimaterials, International Journal for Numerical Methods in Engineering, 59:1103, 2004.
7. Mann, Tveiten, and Harkegard, Fatigue crack growth analysis of welded aluminum RHS
T-joints with manipulated residual stress level, Fatigue Fract Engng Mater Struct, 29,
113122,2006.
8. Mos, N, Dolbow, J, Belytschko, T., A finite element method for crack growth without
remeshing, International Journal for Numerical Methods in Engineering, 46, 131-150,
1999.
9. Nagashima, et al, Stress intensity factor analysis of interface cracks using X-FEM,
International Journal for Numerical Methods in Engineering, 56:1153, 2003.
10. Sukumar, N, Mos, N, Moran, B, Belytschko T., Extended finite element method for
three-dimensional crack modeling, International Journal for Numerical Methods in
Engineering, 48, 1549-1570, 2000.
2008 Abaqus Users Conference

15