International Journal of Solids and Structures 49 (2012) 328–337

Contents lists available at SciVerse ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

A three-dimensional edge-crack finite element for fracture mechanics applications

Kurt J. Hall, Gabriel P. Potirniche ⇑
Department of Mechanical Engineering, University of Idaho, P.O. Box 440902, Moscow, ID 83844-0902, United States

a r t i c l e i n f o a b s t r a c t

Article history: A three-dimensional extension of a previously published two-dimensional cracked finite element
Received 29 June 2011 [Potirniche, G.P., Hearndon, J., Daniewicz, S.R., Parker, D., Cuevas, P., Wang, P.T., Horstemeyer, M.F.,
Received in revised form 13 September 2011 2008. A two-dimensional damaged finite element for fracture applications. Engineering Fracture Mechan-
Available online 14 October 2011
ics 17(13), 3895–3908] is presented in this paper. The new element has an embedded edge crack, and was
developed to model damage in three-dimensional structures using the finite element method. The
Keywords: element simulates the presence of a crack without physically inserting it in the three-dimensional finite
Edge crack
element mesh. The method involves the derivation of a modified stiffness matrix that accounts for the
Finite element
Natural frequency
change in the element flexibility due to the crack presence. The cracked element was analytically formu-
UEL lated and implemented in the finite element code ABAQUS Standard as a User-defined Element (UEL) sub-
ABAQUS routine. Tests of various cracked beam configurations were used to estimate the accuracy of the element
by comparing two models: one with a UEL and another with an embedded edge crack. Beam deflections
and natural frequencies were analyzed and compared for the two models. The results indicate that the
new element has a good potential in modeling cracks in three-dimensional parts. Moreover, the method
using this UEL computes the global response of damaged structures, in which cracks can be placed at
various locations and in an unlimited number.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction for a rectangular beam with a through-thickness crack to measure

the depth and location of a surface crack embedded in a cantilever
With the continued use of aging industrial and transportation beam. Krawczuk et al. (2003) have also used a local flexibility ma-
equipment, improved methods are needed to evaluate and monitor trix to simulate the presence of a crack in a finite spectral element
the fatigue and fracture damage in components. Experimental and for modal and elastic wave propagation. They analyzed the vibra-
computational studies (Dimarogonas, 1996; Wauer, 1990; Owolabi tion properties for a beam as a result of the embedded crack.
et al., 2003; Cacciola et al., 2003; Carneiro and Inman, 2002) have Chondros et al. (1998) modeled the presence of a crack in a beam
shown that the presence of defects such as cracks and voids leads by an added flexibility to the continuous displacement field. They
to altered vibration properties of structures. Moreover, due to the applied their theory to calculate the decrease in the natural fre-
damage presence, the original stiffness of structures is diminished. quency for cracked aluminum and steel beams with an increase
Monitoring the change of these properties over time represents a in crack length. Mahmoud et al. (1999) modeled the crack presence
common non-destructive method of evaluating the severity of in a beam as an added compliance. The beam was modeled using
damage and computing the remaining service life. one-dimensional elements. With the developed element, they ana-
The finite element method is a common numerical technique to lyzed the variation of the beam natural frequency with crack
analyze cracked components. Cracked elements for the finite ele- length. Giner et al. (2009) used the extended finite element meth-
ment method have been developed previously by several research- od (X-FEM) and a UEL in ABAQUS to analyze crack propagation in
ers. They addressed almost exclusively one dimensional cracked two-dimensional components. More recently, Kalanad and Rao
beam elements for vibration studies. Papadopoulos and Dimarogo- (2010) have developed an improved two-dimensional finite ele-
nas (1987) analyzed the influence of elliptical surface cracks on the ment with an embedded crack, based on the element formulation
local flexibility of a rotating shaft. In their model, the compliance introduced by Potirniche et al. (2008) to model more accurately
matrix was determined neglecting the influence of the shear com- long cracks, and applied the new element to damage diagnosis in
ponents, and assuming only tension and bending loading of the beams.
shaft. Gounaris et al. (1996) determined a local compliance matrix The goal of this work is to present a three-dimensional finite
element with an embedded edge crack that can model an entire
⇑ Corresponding author. Tel.: +1 208 885 4049; fax: +1 208 885 9031. cracked region in a component. The proposed three-dimensional
E-mail address: [email protected] (G.P. Potirniche).
element replaces an entire cracked region in a part by accounting

0020-7683/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijsolstr.2011.10.010
 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337 329

for the modification of the local compliance of the material as a re- Z

Y
sult of the crack presence. The element stiffness matrix of a cracked F24 F23
element is a function of the embedded crack length. The presence X F16 F15
of the crack is accounted for in the stiffness matrix of the element 8 7
F8 F7
by means of stress-intensity factors (SIFs) resulting from the ap- α
F21 F22
plied nodal forces and the geometry of the element. The change
in local flexibility was derived from the change in the elastic strain 5 6
F5 F6
energy stored in the component between the undamaged (un-
cracked) state and the damaged (cracked) state. The analytical for- F13 F14
mulation of the element is implemented as a User Element (UEL) in
ABAQUS. Because the entire crack is embedded in a single element,
for large cracks the local stress and strain fields cannot be accu-
h F12 F11
rately estimated. The proposed method is most indicated for appli-
4 3
cations in which the changes in local and global flexibility of a F4 F3
cracked structure are of interest, and how these changes affect
b
the global displacement and vibration response of a structure.
1 F20 2 F19
The method is less indicated for situations in which one is inter- F1 F2
ested in estimating local parameters near the crack tip, such as a
stress intensity factors. F9 F10
F17 F18
Potirniche et al. (2008) developed a UEL with a crack embedded
in a two-dimensional 4-node linear element to analyze natural fre-
Fig. 1. Free body diagram for an edge crack three-dimensional finite element.
quency changes of cracked beams. The two-dimensional UEL incor-
porated the crack influence on the element stiffness matrix by
using the SIFs for mode I and mode II crack extensions. The pro- With the forces and the degrees of freedom defined according to
posed finite element in this paper is an extension of Potirniche the X, Y, and Z axes of coordinates, the force-displacement equation
et al. model to three dimensions. Their previous model was devel- for the element is written using the stiffness matrix [K]
oped for a two-dimensional element, thus it neglected the mode III
crack extension. In the present model, by extending the formula- fF 0 g ¼ ½K 0 fug; ð1Þ
tion from a two-dimensional to a three-dimensional element, all where superscript ‘‘0’’ indicates an undamaged state, fFg is the no-
three modes of crack extension are considered. A limitation of this dal force vector and fug is the nodal displacement vector. The forces
study is that the three-dimensional UEL incorporates a straight and displacements for the 8-node element are generated at each
through-thickness crack. In three-dimensional structures, cracks node in the element due to the tractions and/or body forces applied
oftentimes exhibit curved crack fronts. Future studies should ad- to the finite element mesh. The nodal forces and displacements are
dress different crack configurations by modifying the formulation defined in the vector form as
of the stress intensity factors which are used in defining the
changes in the local flexibility of the cracked element. fFg ¼ fF 1 ; F 2 ; F 3 ; . . . ; F 24 gT and fug ¼ fu1 ; u2 ; u3 ; . . . ; u24 gT : ð2Þ
A cantilever beam with a midspan crack was used to verify the
The stiffness matrix is defined assuming that an edge crack with a
correctness of the UEL formulation. Several tests considered the
length a is embedded in the element. Then, the stiffness matrix
cracked cantilever beam in pure bending, tension, bending (mode
[K] is defined as a function of a, and is different than the stiffness
I and II crack extension), and torsion (mode III out-of-plane shear-
matrix in the undamaged state [K0]
ing). The performance of the newly developed edge-crack element
was analyzed by considering the deflections and natural frequency ½K ¼ ½KðaÞ: ð3Þ
response of cracked beams. Also, the frequency responses of a
Due to the three-dimensional action of each force at the element
three-dimensional mechanical component with several crack loca-
nodes, three SIFs can be defined given the geometrical dimensions
tion scenarios are also analyzed, as an exemplification of the prac-
of the element, crack length and stress magnitudes. The three-
tical applicability of the element to more complex geometries with
dimensional configuration of the cracked element and the nodal
multiple cracks.
forces induce all three fracture modes: mode I opening, mode II slid-
ing and mode III tearing. The individual action of each of the nodal
forces in creating these fracture modes is analyzes next.
2. Element definition and stress intensity factors

An 8-node hexahedron was used as a starting point for the

newly developed three-dimensional user element with an embed-
ded edge-crack. The solid 8-node hexahedron element is one of the
most commonly used to mesh three-dimensional mechanical com-
ponents. However, the analytical procedure presented can be ap-
plied to 20-node hexandron elements, by considering the
corresponding nodal forces and displacements for this element.
In the present study, a through-thickness edge crack was inserted
in the element, and it is shown along with the applied nodal forces
in Fig. 1. Other realistic scenarios of semi-elliptical or semi-circular
surface cracks can be modeled using this method by accordingly
defining the stress intensity factors. The node numbering conven-
tion in this diagram is similar to the one used in ABAQUS. The crack
length is a, and the side lengths of the element are represented by Fig. 2. Nodal tensile force creating a mode I opening with a moment closing the
a, b and h. crack.
330 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337

The nodal tensile forces in the X-direction create a mode I SIF at

the crack tip. This is illustrated in Fig. 2 with F 6 . The application of
the force F 6 leads to the element experiencing a tensile force and a
bending moment. In this case, using the superposition principle, F 6
closes the crack resulting in the following equation:

K IF6 ¼ K fIF  K m
IF ; ð4Þ
6 6

where superscripts f and m denote force and moment, respectively.

The right side of Eq. (4) can further be written by using SIF defini-
tions from Tada et al. (2001)
F 6 pffiffiffiffiffiffiffi f a
K fIF ¼ pa  F ; ð5Þ
6 hb b
where
a a a2 a3 a4
Ff ¼ 1:122  0:231 þ 10:55  21:71 þ 30:382 ð6Þ
b b b b b Fig. 4. Nodal force creating a mode II crack in-plane shearing.
and the SIF resulting from the applied moment is
3F 6 pffiffiffiffiffiffiffi m a Finally, the nodal forces that create mode III SIF are considered.
Km
IF ¼ pa  F ; ð7Þ
6 hb b Fig. 5 shows an example for F24 and its out-of-plane shearing effect
where on the element deformation. The mode III SIF resulting from the
a a2 a3 a4 application of F24 according again to Tada et al. (2001) is
a
Fm ¼ 1:122  1:40 þ 7:33  13:08 þ 14:0 : ð8Þ F 24 pffiffiffiffiffiffiffi
b b b b b K IIIF 24 ¼  pa: ð11Þ
hb
Similar equations can be developed for nodal forces F1, F2 and F5.
These forces result in the same effect on the crack as the one created Similar equations are developed for F17, F18, F19, F20, F21, F22 and F23
by force F6, leading to mode I SIFs. by successively replacing in the above equation F24 with each of
Until now only four of the eight nodal forces have been included these forces.
in the calculation of the mode I SIF associated with crack opening. The following relationships define the global SIFs for each defor-
Contrary to the effect of the above mentioned forces, the remaining mation mode of the cracked element using the superposition
four forces F3, F4, F7 and F8 participating in the definition of the principle
mode I SIF also create a bending moment that opens the crack. In K I ¼ K I F 1 þ K IF 4 þ K I F 5 þ K I F 8 ¼ K IF 2 þ K I F 3 þ K I F 6 þ K I F 7 ; ð12Þ
Fig. 3, an example illustrating F 7 is used to define how these forces
will contribute to the formulation of mode I SIF for the crack open-
K II ¼ K IIF9 þ K IIF 11 þ K IIF13 þ K IIF15 ¼ K IIF 10 þ K IIF12 þ K IIF 14 þ K IIF 16 ; ð13Þ
ing. Using again the superposition principle for the bending and
tension loads, the following equation results for the SIF given by
the nodal force F7 K III ¼ K IIIF 17 þ K IIIF20 þ K IIIF 22 þ K IIIF23 ¼ K IIIF18 þ K IIIF 19 þ K IIIF 21 þ K IIIF24 : ð14Þ

K IF 7 ¼ K fIF þ Km
IF : ð9Þ
7 7
3. Formulation of the modified stiffness matrix for a cracked
Similar equations can be defined for the nodal forces F3, F4 and F8. element
Next, the nodal forces that lead to mode II SIF at the crack tip are
considered. Fig. 4 shows an example for force F 13 which creates an The analytical derivation of the modified stiffness matrix em-
in-plane shearing effect on the element. According to Tada et al. ploys the formulation of the elastic strain energy release rate with
(2001), the SIF from F13 is the increase in crack growth rate. Assuming the crack area as
F 13 pffiffiffiffiffiffiffi A ¼ a  h, as illustrated in Fig. 1, the energy release rate GT is
K IIF13 ¼ 1:1215   pa: ð10Þ
hb
@U T
Similar equations can be written from the above equation by replac- GT ¼ ; ð15Þ
@A
ing F13 in the above equation successively with the nodal forces F9,
F10, F11, F12, F14, F15 and F16. where UT is the total strain energy stored in the cracked body. In the
above equation, it is assumed that the loading forces on the element

Fig. 3. Nodal tensile forces creating a mode I opening with a moment opening the
crack. Fig. 5. Nodal force creating a mode III crack out-of-plane shearing.
 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337 331

are held constant. The contribution of each propagation mode to the K o1j ¼ ð1 þ K 11 A11 ÞK 1j for any j ¼ 1; 2; . . . ; 24: ð27Þ
total energy release rate can be written as
Continuing with the example for F1 and solving the above equation
GT ¼ GI þ GII þ GIII ð16Þ it results that
or, using the SIFs for each mode,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ 1 þ 4A11 K o11
1 2 2
 1 2 K 11 ¼ ; ð28Þ
GT ¼ 0 K I þ K II þ K ; ð17Þ 2A11
E 2G III
where E0 ¼ E for plane stress, E0 ¼ 1Em2 for plane strain, while E is the 2K o1j
E K 1j ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi for any j – 1 ð29Þ
modulus of elasticity, m is the Poisson’s ratio, and G ¼ 2ð1þ mÞ is the 1 þ 1 þ 4A11 K o11
shear modulus. Although Eq. (17) is best suited for two-dimensional
cases, three-dimensional components can also be analyzed using Formulas for the following components of the stiffness matrix
this procedure when the specimen is either thin (plane stress) or K 2j ; K 5j ; and K 6j can be obtained in a similar manner.
thick (plane strain). In cases in which the component experiences The other entries in the stiffness matrix K 3j ; K 4j ; K 7j , and K 8j will
a combination of the two stress states at different points along be obtained using similar equations as above, with the difference
the crack front, a modification of Eq. (17) should be operated in that a new constant A33 must be defined to replace A11 using the
order to reflect this mixed stress state. same procedure as before
Castigliano’s first theorem for the damaged and undamaged Z a 2
states takes the form 2p
A33 ¼ 0 2
F f þ 3F m ada: ð30Þ
@U @U o E hb 0
Fi ¼ and F oi ¼ ; ð18Þ
@ui @ui The formulas for the remaining stiffness matrix entries that contrib-
where the force and corresponding displacement are defined for ute to mode I crack opening are similar to the equations for K 3j
each node i. The difference between the undamaged and damaged qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
states after substituting Eq. (17) into (18) is 1 þ 1 þ 4A33 K o33
If j ¼ 3 K 33 ¼ ; ð31Þ
Z ah  
@ 1 2  1 2 2A33
F oi  Fi ¼ 2
0 K I þ K II þ K III dA: ð19Þ
@ui 0 E 2G
2K o3j
Next, in deriving the stiffness matrix components, nodal force If j – 3 & j ¼ 1; 2; . . . 2; 4 K 3j ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð32Þ
F 1 is used as an example, while similar equations can be developed 1 þ 1 þ 4A33 K o33
for all the other nodal forces
" Z # The nodal forces that result in mode II deformation of the
@ 1 ah 2 cracked geometry are F 9 ; F 10 ; F 11 ; F 12 ; F 13 F 14 ; F 15 ; and F 16 . Eq. (19)
F o1  F 1 ¼ K dA : ð20Þ
@ui E0 0 I is then used to derive the stiffness matrix components correspond-
ing to these forces. For instance, nodal force F 9 is used as an exam-
In the above equation, KI is written using the superposition of its
ple, and similar procedures can be applied for the rest of the nodal
tension and bending components
forces. Inserting this nodal force into Eq. (19) gives
 
Z a  @ K fI  K m  Z 
2h IF @ 1 a 2
F o1  F1 ¼ K fIF  Km
IF
F 1 1
da: ð21Þ F o9  F 9 ¼ 0 K II da : ð33Þ
E0 0 1 1 @u1 @ui E 0

Substituting the formulas introduced above for the two SIFs and The components that make up mode II of the stiffness matrix
simplifying, Eq. (21) becomes and their corresponding equations will include a constant A99 de-
Z a 2 fined below. This is a result from Eq. (9), and can be derived
2p @F 1
F o1  F 1 ¼ 0 2
F f  3F m aF 1 da: ð22Þ through the same procedure as the one demonstrated for A11 . Thus,
E hb 0 @u1 A99 is
The previous equation can further be simplified by defining the fol-
ð1:1215Þ2 p  a2
lowing constant A99 ¼ 2
: ð34Þ
Z a 2 E0  h  b
2p
A11 ¼ 2
F f  3F m ada: ð23Þ The stiffness matrix components for F 9 are
E0 hb 0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The first entry in the stiffness matrix, K11 is defined from the stiff- 1 þ 1 þ 4A99 K o99
If j ¼ 9 K 99 ¼ : ð35Þ
ness matrix presented in Eq. (1) 2A99
@F 1
K 11 ¼ : ð24Þ 2K o9j
@u1 If j – 9 & j ¼ 1; 2; . . . ; 24 K 9j ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð36Þ
Separating the two forces in the damaged and undamaged states in 1 þ 1 þ 4A99 K o99
Eq. (22), one obtains
The nodal forces that result in a mode III deformation of the
F o1 ¼ ð1 þ K 11  A11 ÞF 1 : ð25Þ cracked element are F 17 ; F 18 ; F 19 ; F 20 ; F 21 F 22 ; F 23 ; and F 24 . Nodal
force F 17 is used as an example, and the rest of the nodal forces
From Eq. (18), which defines the relationship between nodal forces
are treated by analogy. Inserting the nodal force into Eq. (19) gives
and nodal displacements, Eq. (25) becomes  Z a 
@ 1
X
24 X
24 F o17  F 17 ¼ K 2III da : ð37Þ
K o1j uj ¼ ð1 þ K 11 A11 ÞK 1j uj : ð26Þ @ui 2G 0
j¼1 j¼1
The stiffness matrix components that include the mode III SIF
Because displacements uj are independent variables, Eq. (26) is va- will include the constant A1717 defined below. This constant can
lid only if the coefficients multiplying uj for each j are the same on be derived using the same procedure as the one demonstrated
both sides of Eq. (26). Thus, by identification for A11
332 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337

Fig. 6. Five-element beam with an inserted UEL.

Fig. 8. (a) Boundary conditions for a cantilever beam in tension, (b) boundary
conditions for a cantilever beam in a mixed mode bending and (c) boundary
conditions for a cantilever beam in torsion.

Fig. 7. (a) Cantilever beam with and embedded physical crack and mesh refinement
and (b) cracked cantilever beam with the node that was used for result extraction.
Formulas for all remaining entries in the 24  24 damaged
stiffness matrix can be derived in a similar manner. Once all com-
ponents of the modified stiffness matrix were defined, the analyt-
p  a2 ical formulation was implemented in the ABAQUS finite element
A1717 ¼ 2
: ð38Þ
2G  h  b
The stiffness matrix components for F17 are
Table 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Comparison of the X-displacement at the node of interest for the UEL model and
1 þ 1 þ 4A1717 K o1717 physical-crack model under tension loading (mode I crack opening).
If j ¼ 17 K 1717 ¼ : ð39Þ
2A1717 Tension Displacement X (m)
Crack length a/H Embedded crack model UEL model Difference (%)
2K o17j
If j – 17 & j ¼ 1; 2; . . . ; 24 K 17j ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð40Þ 0.00 1.151E06 1.142E06 0.78
1 þ 1 þ 4A1717 K o1717 0.10 1.210E06 1.236E06 2.10
0.20 1.415E06 1.442E06 1.85
Similar formulas can be developed for the following components of 0.25 1.597E06 1.570E06 1.73
0.30 1.849E06 1.714E06 7.30
the stiffness matrix
0.40 2.706E06 2.068E06 23.58
K 18j ; K 19j ; K 20j ; K 21j ; K 22j ; K 23j ; and K 24j ; where j ¼ 1; 2; . . . ; 24: 0.50 4.420E06 2.554E06 42.22
 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337 333

Table 2 4. Results
Comparison of the X-displacement at the node of interest for the UEL model and
physical-crack model under bending and shear loading (combined mode I + II crack
opening and in-plane shear). 4.1. Cracked beam analysis

Bending and shear Displacement X (m)

The equations presented in the previous section have been
Crack length a/H Embedded crack model UEL (m) Difference (%) implemented in ABAQUS Standard as a Fortran subroutine UEL.f.
0.00 2.420E05 2.759E05 12.29 A five-element beam model shown in Fig. 6 was built consisting
0.10 2.505E05 2.758E05 9.15 of four C3D8 elements and one UEL with an edge crack. Through-
0.20 2.731E05 2.793E05 2.22 out this paper, empty elements (represented only by their edges)
0.25 2.961E05 2.830E05 4.43
0.30 3.224E05 2.879E05 10.72
in three-dimensional meshed models are UELs. A graphical limita-
0.40 4.036E05 3.015E05 25.29 tion of ABAQUS is that UELs do not show as solid elements in the
0.50 5.477E05 3.215E05 41.30 mesh. The five-element beam has a length X = 0.5 m, a height
Y = 0.1 m and a width Z = 0.05 m. For the following cracked-beam
studies, the overall beam dimensions were held constant. Fig. 7a
shows the same beam as in Fig. 6 modeled with a physical edge-
crack embedded at midspan. A physical crack is inserted in the
Table 3
Comparison of the Y-displacement at the node of interest for the UEL model and model in Fig. 7, and a mesh refinement is performed for the
physical-crack model under bending and shear loading (combined mode I + II crack crack-tip region at the top edge of the beam. Embedding an edge
opening and in-plane shear). crack in an ABAQUS solid model was achieved by using the seam
Bending and shear Displacement Y (m) crack interactive command. When a seam crack is inserted in a
model, a partition needs to be created in the crack plane. After cre-
Crack length a/H Embedded crack model UEL model Difference (%)
ating the partition, ABAQUS creates two sets of nodes in the crack
0.00 8.595E05 1.007E04 14.72 plane. These nodes move apart freely when the beam experiences
0.10 8.667E05 9.952E05 12.91
0.20 8.737E05 9.799E05 10.84
various loads. One can observe the significant difference in the
0.25 8.991E05 9.754E05 7.82 number of elements between the two meshes shown in Figs. 6
0.30 9.170E05 9.730E05 5.75 and 7, and the potential in computational time savings that the
0.40 9.707E05 9.737E05 0.31 UEL can provide.
0.50 1.063E04 9.817E05 7.66
Fig. 7b presents an enlarged view of the circled area in Fig. 7a.
This figure shows the embedded crack opening in pure bending.
In this example, the crack length is 0.03 m. In addition, Fig. 7b

Table 4
Comparison of the Y-displacement at the node of interest for the UEL model and (a)
physical-crack model under torsion loading (mode III out-of-plane shear).

Pure torsion Displacement Y (m)

Crack length a/H Embedded crack model UEL model Difference (%)
H
0.00 6.285E06 5.338E06 15.07 α
0.10 6.536E06 5.370E06 17.83
0.20 6.482E06 5.457E06 15.81
0.25 6.576E06 5.516E06 16.11
0.30 6.734E06 5.583E06 17.09
0.40 7.193E06 5.729E06 20.35
0.50 7.826E06 5.891E06 24.72

(b)
1
code as a user defined element (ABAQUS, 2008). In order to prop-
erly define the stiffness matrix for the UEL, the undamaged stiff-
ness matrix must also be defined. The undamaged stiffness
Normalized natural frequency (f /f )
0 d

0.9
matrix is computed using selective reduced integration, as pre-
sented by Hughes (2000) and applied by Hall (2009).
0.8

1nd mode − UEL model

Table 5 0.7
1st mode − Physical crack model
Comparison of the Z-displacement at the node of interest for the UEL model and 2nd mode − UEL model
physical-crack model under torsion loading (mode III out-of-plane shear). 2nd mode − Physical crack model
0.6 3rd mode − UEL model
Pure torsion Displacement Z (m)
3rd mode − Physical crack model
Crack length a/H Embedded crack model UEL model Difference (%)
0.00 1.257E05 1.069E05 14.97 0.5
0.10 1.312E05 1.072E05 18.31 0 0.1 0.2 0.3 0.4 0.5 0.6
0.20 1.316E05 1.078E05 18.08 Normalized crack length (α/H)
0.25 1.346E05 1.084E05 19.46
0.30 1.394E05 1.087E05 22.02 Fig. 9. Normalized frequency response for a cantilever beam with a crack at 50%
0.40 1.538E05 1.088E05 29.25 of the beam length: (a) beam model with boundary conditions, (b) frequency
0.50 1.750E05 1.083E05 38.11 response comparison between the damaged and undamaged models (d – damaged,
0 – undamaged).
334 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337

highlights the node used to extract displacement values, which of the beam, as indicated in Fig. 8b. Mode I + II is a combination of
were compared for both models. This node is located at a distance the crack opening and in-plane shear sliding. Each node load is
X = 0.3 m from the left edge of the model. All elements in both 1000 N. The fixities applied at each node on the left side of the
beam models analyzed, except the UEL, are C3D8 standard 8-node beam restrict the movement in all X-, Y- and Z-directions, thus sim-
hexahedral solid elements from the ABAQUS library. The modulus ulating a cantilever beam boundary condition. Tables 2 and 3 show
of elasticity and Poisson’s ratio are the same for both beams. The the differences in deflections at the node of interest between the
values of these elastic constants are E = 207 GPa and m = 0.292. five-element beam with the UEL and the embedded edge-crack
beam. In this study of bending and in-plane shear, only displace-
ments in the X- and Y-direction are analyzed. The displacement
4.1.1. Mode I
in the Z-direction is negligible and not shown. From Table 2, it
A tension load was applied to achieve a mode I crack opening of
can be observed that the difference in the X-displacement between
the crack. Fig. 8a shows the loading and boundary conditions for
the two models is less than 10% for crack lengths up to 0.25 of the
this analysis. The applied loads are 1000 N at each node. After
beam height, while for larger cracks the difference becomes larger.
the loading was applied, the displacement in the X-direction was
In Table 3, the displacement in the Y-direction is compared be-
recorded for both models at the node of interest indicated in
tween the two models. In this loading case, the Y-displacement is
Fig. 7b. Table 1 shows the deflections recorded in the five-element
more meaningful, and it can be observed that it is much larger than
beam with the UEL and the embedded edge-crack beam. In this
the X-displacement. The UEL performs satisfactorily in this mixed
study only displacements in the X-direction are significant. The
mode I + II opening and in-plane shear loading. The difference be-
displacements in the Y- and Z-directions are negligible and have
tween the two models is relatively small including for large crack
not been analyzed.
of 0.5 of the beam height.
From Table 1, it can be observed that the five-element UEL
beam performs within 10% of the embedded edge-crack beam in
the X-direction of displacement, up to a normalized crack length 4.1.3. Mode III
of 0.3 with respect to the beam height. The displacement in the Fig. 8c shows the cantilever beam loading with a torsion loading
X-direction for crack lengths of 0.4 and 0.5 had a larger difference to induce a mode III out-of-plane shearing on the crack plane. The
between the five-element beam and the embedded edge-crack same fixities are applied as in the case of the cantilever beam in
beam. case of Fig. 8b. Each applied node loading has a value of 1000 N.
The displacements at the node of interest in the Y- and Z-directions
are reported in Tables 4 and 5. The difference in the Y-displace-
4.1.2. Mode I + II ment between the two models is less than 18% for crack lengths
To analyze mode I + II crack extension, a bending test was used
with point loads applied in the negative Y-direction at the free end
(a)
(a)

α H
H α

(b) (b)
1 1
Normalized natural frequency (f0 /f )

0.9
d

0.9
Normalized natural frequency (f /f )
0 d

0.8 0.8

1st mode − UEL model

1st mode − UEL model
0.7 0.7 1st mode − Physical crack model
1st mode − Physical crack model
2nd mode − UEL model
2nd mode − UEL model
2nd mode − Physical crack model
2nd mode − Physical crack model
3rd mode − UEL model
0.6 3rd mode − UEL model 0.6 3rd mode − Physical crack model
3rd mode − Physical crack model

0.5 0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Normalized crack length (α/H) Normalized crack length (α/H)

Fig. 10. Normalized frequency response for a cantilever beam with a crack at 20% of Fig. 11. Normalized frequency response for a simply supported beam with a crack
the beam length: (a) beam model with boundary conditions, (b) frequency at 50% of the beam length: (a) beam model with boundary conditions, (b) frequency
response comparison between the damaged and undamaged models (d – damaged, response comparison between the damaged and undamaged models (d – damaged,
0 – undamaged). 0 – undamaged).
 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337 335

up to 0.3 of the beam height, while for larger crack lengths, the er-
ror reaches 25%. The large difference is in part explained by the dif-
ference between the two models at a crack length of a = 0, which
means that the mesh refinement plays an important role in these
differences for the two models with and without a crack. This
inherent difference due to the mesh refinement is propagated for
all crack models at every crack length. The differences in the Z-dis-
placement for mode III shearing is less than 20% for crack lengths
of up to 0.25 of the beam height, and they become larger to up
to 39% for larger crack lengths of 0.5 of the beam height.

Fig. 13. Normalized frequency response of the brake pedal: (a) first natural
Fig. 12. Edge cracked UEL inserted into a brake pedal: (a) model with one UEL, (b) frequency response in the YZ-plane, (b) second natural frequency response in the
model with two UELs and (c) model with six UELs. YZ-plane and (c) first natural frequency response in the XZ-plane.
336 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337

1.1 change in crack length. Again, the five-element UEL beam per-
formed well up to a crack length of 0.2. After the normalized crack
Normalized natural frequency (f0 /fd )

length of 0.2, the frequencies measured using the UEL model did
not decrease as steeply as those measured using the embedded
1
edge-crack beam. This response is opposite to the one recorded
in the XY-plane. A possibility for this discrepancy is the large differ-
ence in the number of elements between the two beams. In
0.9 Fig. 10b it was also plotted the variation of the third fundamental
frequency, corresponding to displacements in the YZ-plane loading
Natural frequency 1 − YZ plane the beam in a mode III crack extension. The first fundamental
Natural frequency 2 − YZ plane frequency in the YZ-plane produces better results than the
0.8 Natural frequency 3 − XZ plane frequencies in the XY-plane for this beam configuration. The five-
element UEL beam responded smoothly to the change in crack
length through its crack length range. Contrary to the second fun-
0.7 damental frequency for the XY-plane, the five-element UEL beam
0 1 2 6 did not record as great of a frequency change as the crack length
Number of elements increased. For crack configurations and loading cases in which
the accuracy of the present method is not satisfactory, an increased
Fig. 14. Normalized natural frequency response to increasing UEL elements
inserted into the brake pedal. mesh refinement should be used. Also, models with inserted cracks
using standard discretization or the X-FEM technique can also be
considered.
4.2. Natural frequency analysis for cracked beams Lastly, a simply supported beam was analyzed with a crack in-
serted at midspan. Fig. 11a shows the boundary conditions and
Frequency analysis for the two models, with and without the crack orientation. The variations of the first three natural frequen-
UEL was also performed and the results were compared. Cracks cies as the crack becomes larger are shown in Fig. 11b. The first
were inserted in two cantilevers at a distance of 20% and 50% of fundamental frequency in the XY-plane of the simply supported
the beam length from the fixed end, and in a simply supported UEL beam follows the frequency change of the embedded edge-
beam at 50% of its length. Natural frequencies of these models crack beam smoothly. The second fundamental frequency in the
were computed, and the UEL model was compared with the phys- XY-plane for the UEL model matches closely the edge-crack beam
ical crack model. The vibration modes analyzed for each type of results throughout its vibration response with the change in crack
beam were chosen such that they subjected the crack to a mode length. The variation response of the third fundamental frequency
I, II or III crack opening/shearing. The beam material was assumed in the YZ-plane for the five-element simply supported beam had a
to be steel with the modulus of elasticity and Poison’s ratio 5% difference from the embedded edge crack beam at its extreme.
E = 207 GPa, and m = 0.292, respectively and the mass density The five-element UEL beam produced accurate results for the nat-
q = 7850 kg/m3. Figs. 9–11 present the variation of the natural fre- ural frequency corresponding to the vibration in the YZ-plane.
quency for a damaged (cracked) beam with respect to the crack Overall, the UEL beam model produced results close to the
length. In these plots, the frequency of the damaged beam fd is nor- physical crack model for all three study cases. The only large dis-
malized with the frequency of the undamaged beam f0, and the crepancies were observed from the cantilevered beam with a
crack length a is normalized with the beam height H. mid-span crack. This seems to have been caused by the location
First, a cantilevered beam with a crack located at 50% of the of the crack in the cantilever beam. The UEL beam with a greatly
beam length from the fixed end was analyzed, as shown in reduced number of elements helped verify the validity of the ana-
Fig. 9a. The variations of the normalized natural frequency with lytical formulation for the UEL.
crack length for each of the first three vibration modes are shown
in Fig. 9b. The first fundamental frequency for the XY-plane de- 4.3. Frequency analysis of a three-dimensional mechanical structure
creases with increasing crack length, and the UEL predicts well
the variation for all crack lengths up to 0.5 of the beam height. In the section, the resourcefulness of the newly developed UEL
The variations of the second fundamental frequency mode for the for implementing cracks in three-dimensional mechanical struc-
XY-plane and the third fundamental frequency in the YZ-plane tures is demonstrated. The frequency response of a three dimen-
are also shown in Fig. 9b. It can be observed that the UEL slightly sional brake pedal with several cracks placed at arbitrary
overpredicts the values of the frequency for the damaged beam locations was analyzed. For this analysis, the brake pedal was
compared with the values from the physical crack beam model. tested first without any cracks, as a base line study. UELs were then
Next, the case of the cantilever beam with a crack located at 20% added to the brake pedal to simulate the presence of cracks at var-
of its length with respect to the fixed end is analyzed in Fig. 10. ious locations. Three studies were analyzed with the brake pedal
Fig. 10a shows the beam and crack orientation. Fig. 10b shows having: (i) one, (ii) two, and (iii) six cracks. The material properties
the change in the normalized frequency as the crack becomes lar- assumed for the pedal rail are those for 6061-T6 aluminum, with
ger for the first three vibration modes. For the first fundamental the modulus of elasticity E = 68.9 GPa, Poisson’s ratio m = 0.33 and
frequency in the XY-plane, the five-element UEL beam results are mass density q = 890.5 kg/m3. The mesh used a combination of
close to the embedded edge-crack beam data for a normalized C3D4 and C3D8 linear elements from the ABAQUS element library.
crack length up to 0.20. After a normalized crack of 0.20, the Fig. 12a–c shows the area in the brake pedal where the UEL ele-
five-element UEL beam has a steeper decline in the variation of ments were inserted for all three tests.
the frequency up to a crack of 0.50. This error in the analysis could Fig. 13 shows the different vibration modes, and Fig. 14 plots
be due to the large difference in the number of elements between the variation of the frequency change for the increase in the num-
the two beams. Fig. 10b also shows the second fundamental fre- ber of cracks embedded in the brake pedal. The brake pedal pro-
quency in the XY-plane and the change in frequency as the crack duced interesting results for the natural frequency variations in
becomes larger. The second fundamental frequency in the XY-plane the YZ-plane. The frequency variations increased as more cracks
was smoother throughout its entire vibration response due to the were added to the brake pedal rail. The natural frequency in the
 K.J. Hall, G.P. Potirniche / International Journal of Solids and Structures 49 (2012) 328–337 337

XZ-plane created torsion on the pedal rail resulting in a variation in shapes by modifying the stress intensity factors used in the defini-
the natural frequency that decreased with the increase in the num- tion of the stiffness of the cracked element.
ber of embedded cracks in the brake pedal rail.
References
5. Conclusions
ABAQUS, Documentation Version 6.8, 2008. Hibbit, Karlson and Sorensen, Inc.
Cacciola, P., Impollonia, N., Muscolino, G., 2003. Crack detection and location in a
A new three-dimensional finite element to account for the pres- damaged beam vibrating under white noise. Computers and Structures 81,
ence of edge cracks was developed and implemented in the finite 1773–1782.
Carneiro, S.H.S., Inman, D.J., 2002. Continuous model for transverse vibration of
element program ABAQUS. The crack presence is simulated by cracked Timoshenko beams. Journal of Vibration and Acoustics Transactions of
reducing the stiffness matrix of the element derived from Castigli- the ASME 124, 310–320.
ano’s first theorem. The element was tested for various cracked Chondros, T.G., Dimarogonas, A.D., Yao, J., 1998. A continuous cracked beam
vibration theory. Journal of Sound and Vibration 215, 17–34.
beam configurations and the results were compared with those
Dimarogonas, A.D., 1996. Vibration of cracked structures: a state of the art review.
from a typical beam model embedded with a physical crack. The Engineering Fracture Mechanics 55, 831–857.
new element is well suited for frequency analysis commonly used Giner, E., Sukumar, N., Tarancón, J.E., Fuenmayor, F.J., 2009. An ABAQUS
in damage assessment studies. The frequency analysis data indi- implementation of the extended finite element method. Engineering Fracture
Mechanics 76, 347–368.
cates that the five-element beam with the UEL closely follows Gounaris, G.D., Papadopoulos, C.A., Dimarogonas, A.D., 1996. Crack identification in
the trend of the embedded edge-crack beam. beams by coupled response measurements. Computers & Structures 58, 299–
The new three-dimensional edge cracked element was inserted 305.
Hall, K.J., 2009. The Development of a Three-Dimensional Edge Cracked Finite
into a complex three-dimensional structure. The analysis indicated Element. M.S. Thesis, The University of Idaho.
that the dominant natural frequency mode resulting from an ap- Hughes, T.J.R., 2000. The Finite Element Method; Linear Static and Dynamic Finite
plied torque on the pedal rail was of greater importance than the Element Analysis. Mineola, Dover Publications Inc.
Kalanad, A., Rao, B.N., 2010. Detection of crack location and size in structures using
vibration mode that caused mode I and II crack extension. The improved damaged finite element. 2010 IOP Conf. Ser.: Mater. Sci. Eng. 10,
new finite element has a great potential in predicting and evaluat- 012054 doi:10.1088/1757-899X/10/1/012054.
ing the damage and life expectancy of three-dimensional parts. It Krawczuk, M., Palacz, M., Ostachowicz, W., 2003. The dynamic analysis of a cracked
Timoshenko beam by the spectral element method. Journal of Sound and
helps to reduce the number of elements needed to analyze the re- Vibration 264, 1139–1153.
sponse of damaged structures with cracks at single or multiple Mahmoud, M.A., Abu Zaid, M., Al Harashani, S., 1999. Numerical frequency analysis
locations. With the new element instead of physically embedding of uniform beams with a transverse crack. Communications in Numerical
Methods in Engineering 15, 709–715.
cracks in parts, the UEL can readily substitute cracked regions in
Owolabi, G.M., Swanidas, A.S.J., Seshadri, R., 2003. Crack detection in beams using
the model, keeping the number of elements in the mesh low, while changes in frequencies and amplitudes of frequency response functions. Journal
maintaining the same degree of accuracy. of Sound and Vibration 265, 1–22.
The proposed method is recommended for studies of the Papadopoulos, C.A., Dimarogonas, A.D., 1987. Coupled longitudinal and bending
vibrations of a rotating shaft with an open crack. Journal of Sound and Vibration
changes in local and global flexibility as they affect the global dis- 117, 81–93.
placement and vibration response of a cracked structure. The Potirniche, G.P., Hearndon, J., Daniewicz, S.R., Parker, D., Cuevas, P., Wang, P.T.,
method cannot accurately compute the local stress and strain Horstemeyer, M.F., 2008. A two-dimensional damaged finite element for
fracture applications. Engineering Fracture Mechanics 17 (13), 3895–3908.
fields, thus is not indicated for the computation of crack growth Tada, H., Paris, P.C., Irwin, G.R., 2001. The Stress Analysis of Cracks Handbook. ASME
rates and local parameters near the crack tip, such as stress inten- Press, New York.
sity factors. The proposed UEL incorporates a straight through- Wauer, J., 1990. Cracked rotor dynamics: a state of the art survey. Applied
Mechanics Reviews 42, 13–17.
thickness crack. Future studies should consider more realistic crack