See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/285582224

An improved method to model semi-elliptical

surface cracks using element mismatch in
ABAQUS

Conference Paper · November 2015

CITATIONS READS

0 75

2 authors, including:

Feizal Yusof
Universiti Sains Malaysia
12 PUBLICATIONS 64 CITATIONS

SEE PROFILE

Available from: Feizal Yusof

Retrieved on: 27 September 2016
 An improved method to model semi-elliptical surface cracks using
element mismatch in ABAQUS
RHA Latiff1 and F Yusof1,2
1
School of Mechanical Engineering, Universiti Sains Malaysia
2
email: mefeizal@usm.my

ABSTRACT

The purpose of this paper is to propose and investigate the feasibility of using multiple element types in a single model,
termed element mismatch, for a three-dimensional semi-elliptical crack problem. The purpose behind the exploration of
this method is to ease the meshing process in complex models. Multiple semi-elliptical surface crack in tension models
were created with single element mesh and multi-element meshes with different mesh densities. It was found that the
element mismatch models developed a consistent stress intensity factor for all the applied loading and geometries. When
compared to the single element type models, the computation times were found to be significantly lower for the element
mismatch models and the results were largely independent of mesh density.

Keywords: Fracture Mechanics, Abaqus, Element mismatch

INTRODUCTION ABAQUS has limited the region around the crack front to
The optimum use of computational resources for only certain types of elements, specifically quadrilateral
a finite element numerical analysis is an important issue in elements in two-dimensional models and hexahedral brick
computational fracture mechanics [1]. As the accuracy of elements in three-dimensional models [11]. Due to this
crack-tip deformation problem becomes critical, the crack limitation, fracture mechanics models typically used a
tip mesh nodal and elemental configurations will usually single type of element for the entire model. However, [12]
be increased, which will directly increase the time to has undertaken work to model the crack front using non-
complete the analysis due to large degrees of freedom. brick elements.
However this approach may not give an accurate outcome.
For structural problems, the sizing of the elements using Using ABAQUS/CAE, the regions outside of the
the h method [2] or p method [3] to revise finite element crack front can be meshed using other than the brick type
meshes to obtain the necessary accuracy by using as many of element. A quadratic brick element has 20 nodes while
degrees of freedom as necessary, also called adaptive a quadratic tetrahedral element has 10 nodes and a
meshing, has been introduced. However problems with quadratic wedge element has 15 nodes. The use of
geometrical discontinuities, such as cracks, may not elements with fewer nodes, and thus fewer degrees of
developed convergence of solution appropriately. freedom (DOF), in the regions outside of the crack front
would allow for models to be solved with shorter
Three-dimensional elastic-plastic crack tip computing times. Furthermore, meshing of complex
problems have evolved over 40 years from the defining bodies is relatively difficult using brick elements as
work of [4]. The development of formulation for three- meshing requires careful partitioning of the model. In
dimensional elastic-plastic crack tip problems gained fracture mechanics, one such model is the semi-elliptical
momentum after efficient finite element formulation was surface crack which requires partitioning before structured
combined with the computing capability to handle meshing using brick elements was allowed by ABAQUS,
problems with a large number of degrees of freedom in the as shown in Figure 1 below.
eighties [5, 6]. However, the results were partially
inconclusive with respect to their original hypothesis of a
plane stress field at the region where the crack front
approached the free surface. A widely accepted three-
dimensional elastic crack tip formulation breakthrough
came through the work of [7]. However, the three-
dimensional elastic-plastic crack tip formulation [8] was
not able to capture the sensitivity of the corner field effect
due to mesh inadequacy at the free surface. Three- Figure 1: Partitioning of a semi-elliptical surface crack
dimensional elliptical surface crack was discussed by [9] quarter model
but was limited in the crack configuration and material
response, while recent work by [10] was also limited to Element mismatch, henceforth EM, refers to the
unvarying crack configurations due to complexity of the condition where a mesh comprises of multiple element
problem. . types. Two types of meshes incorporating EM can be
created using ABAQUS. The first type of mesh creates a
Presently, commercially available FEA codes, single continuous mesh using different elements, while the
such as ABAQUS[11], are widely used to elucidate the second type of mesh creates separate meshes for each of
nature of crack tip stress strain behaviour quite the element types. In a continuous mesh, interface nodes
successfully. ABAQUS incorporates fracture mechanics in are shared by the neighbouring elements, as shown in
its library of functions and analysis tools. However, Figure 2, where Triangular elements share nodes with
Quadrilateral elements. The limitation to this method is
that the elements must have compatible edge and surface
node configurations. For example, the Quadrilateral and
Triangular elements in Figure 2 have compatible edge
node configurations; both elements have two nodes per
edge. Thus they can be meshed as a continuous mesh.

Figure 4: Tie constrained elements

This paper investigates the feasibility of using

EM within a single continuous mesh for fracture
mechanics purposes. Emphasis was given to accuracy of
the results and also the computing time required in
performing the analysis. The scope of this paper is limited
Figure 2: Elements of differing types sharing nodes. to an elastic stationary semi-elliptical surface crack under
tension.
The requirement of having matching node
configurations on edges also applies to 3D modelling. The METHODOLOGY
surfaces of the elements must match in order for A semi-elliptical surface crack in tension was
ABAQUS to generate a continuous mesh. Thus, not all modelled using ABAQUS/CAE. To investigate the
elements are compatible. For example, the surfaces of a feasibility of using EM in fracture mechanics, models
brick element and a tetrahedral element have no matching incorporating EM and standard single element type
surfaces, as shown in Figure 3. The brick element has 8 meshes were used. The stress intensity factor K of the
nodes per surface while the tetrahedral element only has 6 models were determined and compared to the equation
nodes per surface. However, wedge elements share provided by [14]:
compatible surfaces with brick elements, as highlighted in
Figure 3. √ (1)

Where St is the remote tensile stress, and F and Q

are functions of the geometry of the specimen.

The models created share a single geometry, as

shown in Table 1 and Figure 5. Figure 5(a) also indicates
the lines of symmetry. Figure 5(b) shows the angular
system used to discuss the results in the Discussion
section. The depth of surface crack a, the half-width of the
cracked plate b, the half-length of surface crack c, the
plate thickness t, and the parametric angle of the ellipse φ
are indicated in Figure 5.

Figure 3: Linear Brick and Tetrahedral Element

Configurations

As previously discussed, another method for

incorporating EM into a finite element model is the
creation of separate meshes using different elements. a)
These meshes can be configured to behave as a continuous
mesh via the Tie Constraint provided in ABAQUS/CAE.
The elements in this configuration do not share nodes, as
shown in Figure 4. The authors of [13] have done some
work using the Tie Constraint and XFEM for fracture b)
mechanics. The use of the Tie Constraint is not discussed
in this paper, as it generates a discontinuous mesh.

Figure 5: Semi-elliptical surface cracked plate [14]

 Table-1. Geometry configurations based on Figure 5 Table 2 below indicates the configurations for
Geometric Ratios Value each of the models used. The naming convention for the
models is such that the brick models are denoted with ‘B’
a/t 0.5 in their names, while the EM models are denoted with
‘M’. The numbering indicates the relative density of the
a/c 0.4
mesh at the crack front, with ‘4’ indicating the finest mesh
c/b 0.325 at the crack front. Thus, B1 and M1 have the same crack
front mesh, as do B2 and M2, and so on.
Wedge elements were chosen for the for the EM
Table-2. Mesh Configurations
mesh as they have a compatible surface with brick
elements, as show in Figure 3. The element types used in Name Elements used No. DOF
the analyses are reduced-integration quadratic brick B1 C3D20R 98,151
elements(C3D20R) and reduced integration quadratic B2 C3D20R 160,188
wedge elements(C3D15). The K values of the models B3 C3D20R 510,171
were benchmarked to (1). A comparison of the accuracy of B4 C3D20R 433,173
the results and computation time were made. M1 C3D20R, C3D15 73,953
M2 C3D20R, C3D15 87,138
Finite Element Models M3 C3D20R, C3D15 217,950
Due to symmetry, as shown in Figure 5(a), the M4 C3D20R, C3D15 249,087
analysis was conducted based on a quarter model The meshes outside of the crack front were
framework. A tensile load was applied to the models using intentionally designed to be coarse in comparison to the
displacement, as shown in Figure 6 while symmetrical crack front. The mesh seeding for the models were
conditions were enforced for the x and z axes. manually adjusted until the elements within the mesh did
not exceed the built-in aspect ratio limits. The models
using solely brick elements have a greater number of
DOF. This occurred as it is relatively easy to transition
from a fine mesh to a rough mesh using a wedge element
when compared to a brick element without exceeding the
built in aspect ratio in ABAQUS, as shown in Figure 8.
This was done to indicate that the roughness of the mesh
outside the crack front region did adversely affect the
results, as presented in the following section.

Figure 6: Boundary conditions for all models

To investigate the effects of EM, two types of a)

models were created and compared, with one
incorporating EM and another model using only brick
elements. Both models share the same crack front
configuration in terms of element type and element sizes.
In both models, the crack fronts were meshed with brick
elements, as per the limitation set by ABAQUS. The
remaining sections of the models were meshed differently.
In the ‘Brick models,’ the entire model was meshed with
brick elements while in the ‘EM models’, the remaining b)
sections of the models were meshed with wedge elements.
This is shown in Figure 7. Four crack fronts with varying
mesh densities were also modelled to further investigate
the effects of EM. Thus, a total of 8 models were used for
the study.

Figure 8: Comparison of transitioning mesh densities

RESULTS AND DISCUSSION

a) b)
Figures 9 and 10 show K of the brick element
models and models incorporating EM compared to the
benchmark results by [14]. The angles indicated in the
Figure 7: a)Brick element and b)Element mismatch figures are based of the notation used in Figure 5(b).
meshes
 Table-3. DOF Comparison
Brick DOF EM DOF %
Model Model Reduction
of DOF
B1 98,151 M1 73,953 24.7
B2 160,188 M2 87,138 45.6
B3 510,171 M3 217,950 57.3
B4 433,173 M4 249,087 42.5

Comparing the K values between the brick

element models and EM models in Figures 9 and 10, the
Figure 9: SIF of B1-B4 results do not vary by a significant margin, approximately
by 2%. Here, we can infer that K is largely unaffected by
the use of EM. The results of the analysis can be further
improved to fit the benchmark by increasing the mesh
density around the crack front.

Thus, EM appears to be a feasible alternative to

meshing in fracture mechanics analyses as K was more
heavily influenced by the mesh around the crack tip as
opposed to the mesh outside of the crack tip.

Computation time
Referring to Table 3, the number of DOF for the
Figure 10: SIF of M1-M4 EM models was fewer than their counterpart brick element
models. As the meshes around the crack front compared
Benchmarking were the same, the reduction in number of DOF was a
Referring to Figures 9 and 10, K of the models result of the mesh outside of the crack front. This is due to
were generally in good agreement with the work by [14]. the ease in which wedge elements are able to transition
The percentage differences for all models when compared from a dense mesh to a course mesh when compared to
to [14] were found to remain below 3% between the 20º to brick elements, as previously shown in Figure 8. Despite
90º range. However, the percentage difference increased the fewer DOF, the results were not adversely affected, as
below the 20º angle. The percentage difference reached a previously discussed.
maximum of 15% at the 0º angle for the B1 and M1
models. A comparison of the computation times shown in
Table 4 confirms that the analysis of the EM models
It should be noted that the meshing density required less time when compared to the brick element
around the crack tip appears to have affected the results of models. In the most accurate of the models, the B4 and
the analysis. The maximum percentage difference, M4 models, the reduction of computation time from the
occurring at the 0º angle in all cases, showed a marked B4 model to the M4 model was 37%. This reduction
decrease with increasing crack tip mesh density. The corresponds to a 42.5% reduction in the number of DOF.
models with the densest mesh at the crack front were B4 This was to be expected as the computation time is linked
and M4. Both exhibited a percentage difference of 5.6%. to the number of degrees of freedom. This is a good
Thus, the higher percentage difference at the 0º angle can indication that computing times for fracture mechanics
be reduced via a finer mesh around the crack front. In both models can be significantly reduced via the
the brick element and EM models, more accurate K values implementation of EM, at least within the elastic
were achieved with increasing mesh density around the paradigm.
crack front.
Table-4. Computation Time Comparison
Referring to Table 3, the number of DOF for M4 Brick CPU EM CPU Time %
was much lower than the number of DOF for B3. It must Model Time Model Reduction
be noted that the values in Table 3 refer to the DOF of the of CPU
entire model. Although experience suggests that a denser time
mesh would return more accurate results, the K values for B1 51.2 M1 25.9 49.4
M4 are more accurate than B3 at the 0º angle. This further B2 129.4 M2 36.4 71.9
proves that the mesh around the crack front is more crucial B3 1363.8 M3 264.1 80.6
to the accuracy of the results than the mesh in other B4 1291.8 M4 813.6 37
regions.

Table 4 shows that the reduction of computing

time is significant without much loss from the results, as
shown in Figures 9 and 10. This indicates that
incorporating EM is a feasible approach for fracture [3] O. Zienkiewicz, C. Emson, and P. Bettess, "A
mechanics simulation, within an elastic material response. novel boundary infinite element," International
Journal for Numerical Methods in Engineering,
vol. 19, pp. 393-404, 1983.
CONCLUSION [4] N. Levy, P. Marcal, and J. Rice, "Progress in
Models of elastic semi-elliptical surface cracks were three-dimensional elastic-plastic stress analysis
built and meshed in various configurations. It was found for fracture mechanics," Nuclear Engineering
that the results of the EM models and [14] are in good and Design, vol. 17, pp. 64-75, 1971.
agreement. The computation times for the mismatch [5] W. Brocks and J. Olschewski, "On J-dominance
models were significantly decreased, when compared to of crack-tip fields in largely yielded 3D
the brick element models. Thus, the implementation of structures," International Journal of Solids and
Element Mismatch (EM) is viable in the simulation of Structures, vol. 22, pp. 693-708, 1986.
stationary cracks. [6] G. W. Wellman, S. T. Rolfe, and R. H. Dodds,
"Three-dimensional elastic-plastic finite element
analysis of three-point bend specimens," ASTM
STP, vol. 868, pp. 214-237, 1985.
[7] T. Nakamura and D. Parks, "Three-dimensional
FURTHER WORK stress field near the crack front of a thin elastic
1. To investigate effects of Tetrahedral elements in plate," Journal of applied Mechanics, vol. 55, pp.
element mismatch on fracture mechanics model. 805-813, 1988.
2. To update models to include an elastic-plastic [8] T. Nakamura and D. Parks, "Three-dimensional
response. crack front fields in a thin ductile plate," Journal
3. To determine the effects of element mismatch & of the Mechanics and Physics of Solids, vol. 38,
tie constraint on both elastic and elastic-plastic pp. 787-812, 1990.
fracture mechanics model. [9] Y. Wang, "A two-parameter characterization of
elastic-plastic crack tip fields and applications to
cleavage fracture," Massachusetts Institute of
ACKNOWLEDGEMENTS Technology, 1991.
The authors would like to acknowledge the [10] O. Terfas, "Quantification of Constraint in Three-
MOHE grant (FRGS/1/2014/TK01/USM/02/5) which Dimensional Fracture Mechanics," Doctor of
funded this research project. The ABAQUS finite element Philosphy, Department of Mechanical
code that was used to perform the analyses was made Engineering, University of Glasgow, 2010.
available under an academic license from Dassualt [11] A. Version, "6.13," Analysis User’s Guide,
Systemes, K. K. Japan.. Dassault Systems, 2013.
[12] M. Nejati, A. Paluszny, and R. W. Zimmerman,
"On the use of quarter-point tetrahedral finite
elements in linear elastic fracture mechanics,"
REFERENCES Engineering Fracture Mechanics, vol. 144, pp.
194-221, 2015.
[1] M. Kuna, "Finite Elements in Fracture [13] M. Levén and R. Daniel, "Stationary 3D crack
Mechanics," Theory–Numerics–Applications. analysis with Abaqus XFEM for integrity
Solid Mechanics and Its Applications, vol. 201, assessment of subsea equipment," 2012.
2013. [14] J. Newman and I. Raju, "An empirical stress-
[2] L. Demkowicz, P. Devloo, and J. T. Oden, "On intensity factor equation for the surface crack,"
an h-type mesh-refinement strategy based on Engineering Fracture Mechanics, vol. 15, pp.
minimization of interpolation errors," Computer 185-192, 1981.
Methods in Applied Mechanics and Engineering,
vol. 53, pp. 67-89, 1985.