Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
a r t i c l e i n f o a b s t r a c t
Article history: A quasi-static and propagating analysis is simulated for two-unequal-collinear cracks in a 2-D finite mag-
Received 19 December 2011 netoelectroelastic specimen using X-FEM. The intensity factors (IFs) are calculated using interaction inte-
Received in revised form 1 February 2012 gral in conjugation with the near tip behavior given by the Stroh formalism. In quasi-static case energy
Accepted 9 March 2012
release rate (ERR) variations are investigated with respect to inter-crack space, crack lengths and
Available online 7 April 2012
mechanical/electrical/magnetic loads. Two-collinear-unequal cracks in an infinite domain problem are
simulated, analyzed and validated. Further, effects of asymmetric orientation and symmetric orientation
Keywords:
of quasi-static two-unequal-collinear cracks vis-à-vis specimen boundaries on total energy release rate
Electric-displacement intensity factor
Extended finite element method
(TERR) and mechanical energy release rate (MERR) are investigated. The case of one edge and one internal
Magnetoelectroelastic (MEE) ceramics quasi-static crack is obtained as a corollary and the case study is presented. Next, the crack growth study
Magnetic induction intensity factor for two-unequal-collinear edge cracks is simulated using maximum modified hoop stress intensity factor
Maximum hoop stress intensity factor criterion and considering anisotropic fracture toughness behavior of polarized ceramics. The effect of vol-
Stroh formalism ume fraction, electrical and magnetic loadings are similar to the strain energy density function criterion
of crack propagation for magnetoelectroelastic ceramics. Lastly dependence of volume fraction and mate-
rial constants of magnetoelectroelastic ceramics are observed on the set of standard eight basis functions
for these ceramics. Hence, a more generalized set of basis functions is also defined here for these
ceramics.
Ó 2012 Elsevier B.V. All rights reserved.
0927-0256/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.commatsci.2012.03.013
76 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
existing analytical solutions. Wang and Mai [12] developed a finite But the most realistic crack face boundary conditions are
element code for a coupled electromagnetoelastic analysis of a considered to be the semi-permeable [7,12,35,36] which are in
crack under mode-I conditions. Further they proposed a systematic between these two ideal boundary conditions. More recently, Ro-
strategy for the analysis of electromagnetic natural boundary con- jas-Dı´ as et al. [35] implemented the different crack face bound-
dition (EM-NBC) and compared the results with those obtained ary conditions on the Griffith’s crack and multiple parallel cracks
from the traditional impermeable and permeable crack-face elec- using Dual BEM. The iterative scheme proposed by Denda [36]
tromagnetic boundary conditions. Li et al. [13] investigated the for semi-permeable crack face boundary conditions are extended
fracture behaviors of two-dimensional planar and axisymmetric to 2-D magnetoelectroelastic solids, and to implement this itera-
problems of MEE materials using meshless local Petrov–Galerkin tive scheme one may require the impermeable solution as an
(MLPG) scheme coupled with finite element method (FEM). initial solution. It is also found that the IFs/mechanical energy
Nowadays, the developments in the computational field have release rate calculated using impermeable boundary conditions
motivated many researchers to work in the field of crack propaga- are reasonable approximation to the semi-permeable boundary
tion of quasi-static crack(s). Park and Sun [14] proposed the strain conditions when the medium between the crack surfaces is ta-
energy release rate or maximal mechanical energy release rate cri- ken as air or vacuum. Further, the implementation of these con-
terion to study the crack propagation in piezoelectric materials. ditions with FEM/X-FEM requires the more computational
Kumar and Singh [15,16] applied the finite element technique to efforts/cost. Therefore, in this paper authors adopted the ideal
study the crack propagation in piezoelectric materials under com- impermeable boundary conditions for the study of collinear
bined mechanical and electrical loadings using the criterion of cracks.
maximum energy release rate. Xu and Rajapakse [17] presented a The present paper addresses the quasi-static and propagating
theoretical model to study a branched crack in piezoelectric mate- problems of two-unequal-collinear cracks in 2-D MEE specimen
rial, and to predict the crack propagation angle a modified hoop using X-FEM. On crack rims impermeable conditions are assumed
stress intensity factor based fracture criterion was implemented. to exist. Using interaction integral with near tip behavior given
Few papers are available in literature on the crack propagation by the Stroh formalism the IFs are calculated. After presenting
in MEE materials. Sih et al. [18–20] proposed the strain energy the basic equations and crack tip fields in Sections 2 and 3, the ex-
density function criterion to study the crack initiation and crack tended finite element approximation is discussed in Section 4 for
propagation in these materials under different parametric studies making the paper self-sufficient. To verify the accuracy of the mod-
such as volume fractions, piezomagnetic and piezoelectric poling els and X-FEM codes, numerical validation for Griffith’s crack and
and electric/magnetic loading. two equal collinear edge/centered cracks is done in Section 5.
The extended finite element method (X-FEM) proposed by Moës The parametric studies of IFs and ERR with respect to inter-crack
et al. [21] and Belytchko and Black [22] proved to be a very efficient space, crack lengths, electrical/mechanical/magnetic loadings
tool for the numerical modeling of crack propagation in LEFM. Vec- along with symmetrical and asymmetrical situated cracks are pre-
tor level sets defined by Ventura et al. [23] in conjugation with X- sented for two-unequal-collinear cracks in MEE ceramic in Section
FEM efficiently handle the crack propagation problems with less 6. A problem of an edge crack situated with an internal crack is also
computational cost. Huang et al. [24] implemented the X-FEM to derived as a special case of two-unequal-collinear cracks in this
study the propagation of different types of cracks such as center section. The crack propagation paths of two unequal/equal-collin-
crack, inclined crack, arc shaped crack and multiple cracks of ear edge cracks have been simulated using X-FEM and Level Set
mud-cracks-pattern in isotropic and layered materials. Be’chet Method in Section 7. Modified hoop stress intensity factor criterion
et al. [25] and Laborde et al. [26] proposed the geometrical/fixed in conjugation with anisotropic fracture toughness parameter for
area enrichment scheme to the conventional X-FEM approximation polarized ceramic is used for these investigations. A parametric
in order to improve the convergence rates in the classical X-FEM. study of crack propagation is carried with respect to mesh discret-
The main idea behind this strategy is to enrich a whole fixed area ization, crack propagation length, domain radius, volume fraction
around the crack tip. Yazid et al. [27] and Belytschko et al. [28] pre- and prescribed mechanical, electrical and magnetic loadings. For
sented a review of X-FEM applied to various engineering problems. both the cases considered the results obtained by X-FEM are vali-
Recently, Rojas-Dı ´ az et al. [29,30] extended the idea of X-FEM in dated with the existing results available in literature. In Section
piezoelectric materials (proposed by Be’chet et al. [31]) to the 8, effect of volume fraction and material constants are investigated
MEE materials. on the set of standard eight basis functions defined for MEE mate-
In the past, only few researchers [14–20] have contributed to rials. And the last section reports the conclusions of the work done
the study of crack propagation on piezoelectric/MEE materials in the paper.
and employed maximum hoop stress, maximum energy release
rate, strain energy density function and maximum hoop stress
intensity factor criterion. But most of these criterions are based 2. Basic equations for magnetoelectroelastic media
on isotropic fracture toughness. It has been found that in aniso-
tropic materials the theoretical predictions obtained from maxi- The fundamental equations and the boundary conditions for
mum modified hoop stress intensity factor and considering linear MEE media are.
anisotropic fracture toughness [32,33] are in good agreement with
experimental results. Recently, Janski’s et al. [34] applied the same 2.1. Field equations
criteria to study the crack propagation in piezoelectric media using
adaptive finite element technique. In a fixed rectangular coordinate system xj (j = 1, 2, 3), the field
In magnetoelectroelastic fracture problems another important equations for a linear MEE medium are:
issue is the consideration of the crack face boundary conditions.
Most of the researchers [5,8–11,13,18–20,29,30] implemented Constitutive equations
the impermeable crack face boundary conditions, in which crack rij ¼ C ijks eks esij Es hsij Hs ; ð1Þ
faces are electrically and magnetically insulated. The other ideal
Di ¼ eiks eks þ jis Es þ bis Hs ; ð2Þ
crack face boundary conditions are permeable one in which the
crack faces are electrically and magnetically conductive [1,4,5]. Bi ¼ hiks eks þ bis Es þ cis Hs : ð3Þ
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 77
rffiffiffiffiffi
Kinematic equations 2r P N
ui ðr; hÞ ¼ K d ðhÞ; ð17Þ
1
eij ¼ ðui;j þ uj;i Þ; Ei ¼ /;i ; Hi ¼ u;i : ð4Þ
pN N i
rffiffiffiffiffi
2 2r P
/ðr; hÞ ¼ K mN ðhÞ; ð18Þ
Equilibrium equations for stresses, electric displacements and pN N
magnetic inductions in the absence of body forces, free electric rffiffiffiffiffi
2r P
charges and free magnetic currents, respectively, may be writ- uðr; hÞ ¼ K N wN ðhÞ; ð19Þ
ten as
p N
rij;j ¼ 0; Di;i ¼ 0; Bi;i ¼ 0; ð5Þ where i, j = 1, 2, and the summation over N = {II, I, III, IV, V} com-
prises the mechanical stress intensity factors (SIFs) mode-I, mode-
where rij, eij, Di, Ei, Bi and Hi denote the components of the stress,
II, mode-III cases, and are denoted by KI, KII, KIII, respectively. And
strain, electric-displacement, electric field, magnetic induction and
KIV and KV denote the electric displacement intensity factor (EDIF)
magnetic field, respectively; Cijks, eiks, hiks and bis denote the elastic,
and magnetic induction intensity factor (MIIF), respectively. In case
piezoelectric, piezomagnetic and electromagnetic constants,
of two-dimensional MEE problem, KIII = 0.The functions
respectively; jis and cis denote the dielectric permittivities and N
fijN ðhÞ; g Ni ðhÞ; qNi ðhÞ; di ðhÞ; v N ðhÞ and wN(h), are the standard angular
magnetic permeabilities, respectively. Comma denotes partial dif-
functions for a crack in a homogeneous MEE medium depending
ferentiation with respect to argument following it; u is the elastic
on the material properties. These expressed in terms of complex
displacement vector; / is the electric potential; and u is the mag-
material eigenvalues pa, eigenvectors AMa and matrices MMa and
netic potential; where i, j, k and s = 1, 2, 3.
NaN using extended Stroh formalism in conjugation with semi-ana-
lytical calculations (Ref. Rao and Kuna [37]).
P 8
P k
þ NI ðx; yÞ F k r; h; are im
k ; ak F k xI ; yI ; are im
k ; ak bI
In case of impermeable conditions on the upper and lower faces I2NTIP k¼1
Sc, of the crack
ð20Þ
Dj nj ¼ 0; on Sc ; ð12Þ
P P
Bj nj ¼ 0; on Sc : ð13Þ /h ðx; yÞ ¼ NI ðx; yÞ/I þ NI ðx; yÞðHðf h ðx; yÞÞ HðfI ÞÞcI
I2N I2Ncr
P 8
P k
3. Crack tip fields in homogeneous magnetoelectroelastic media þ NI ðx; yÞ F k r; h; are im
k ; ak F k xI ; yI ; are im
k ; ak dI
I2NTIP k¼1
ð22Þ
1 P
Di ðr; hÞ ¼ pffiffiffiffiffiffiffiffiffi K N g Ni ðhÞ; ð15Þ
2pr N where H(f(x, y)) is a modified Heaviside step function
1 P 1 if z1 < 0
Bi ðr; hÞ ¼ pffiffiffiffiffiffiffiffiffi K N qNi ðhÞ; ð16Þ Hðz1 Þ ¼ ð23Þ
2pr N þ1 if z1 > 0
and the near tip displacement field, electrical potential and mag- And the shape functions, NI(x, y), are isoparametric linear quadrilat-
netic potential may be written as eral element shape functions. The column matrices uI, /I and uI are
78 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
the nodal displacements, electric potential and magnetic potential, 5. Numerical validation
k k
respectively, and aI ; bI ; cI ; dI ; sI and t kI are the additional parameters.
k
re im
F r; h; ak ; ak is the basis for the crack tip defined in [29,30] as This section mainly focused on the numerical validation of the
k
pffiffiffi pffiffiffi pffiffiffi pffiffiffi pffiffiffi pffiffiffi pffiffiffi pffiffiffi
IFs/energy release rate evaluated here using X-FEM to the estab-
F r;h; are im
k ; ak ¼ rf1 ðhÞ; r f2 ðhÞ; rf3 ðhÞ; rf4 ðhÞ; rf5 ðhÞ; r f6 ðhÞ; rf7 ðhÞ; rf8 ðhÞ
lished analytical/numerical results. The following analyses have
ð24Þ
been made to justify the models and numerical approach consid-
where ered in this paper:
8
> re im wm ðh;are
m ;am Þ
im
1. Numerical validation for Griffith–Irwin’s crack in an infinite
>
> q h; a ; a cos if aim
m > 0
>
< m m m 2
domain.
fm ðhÞ ¼ ð25Þ 2. Comparison with results of Li & Lee [10] for center crack
>
>
>
> wm ðh;are
m ;am Þ
im
problem
: qm h; are im
m ; am sin 2
if aim
m >0
3. Numerical validation for double edge collinear cracks in a finite
im specimen
The complex numbers am ¼ are m þ iam are the eight roots of the char- 4. Numerical validation for two equal centered collinear cracks in
acteristic Eq. (A.9) of Appendix A. The modified angle and radius are an infinite domain
of the form
p h To achieve the numerical results, the X-FEM models are de-
wm h; are
m ; aim
m ¼ þ p int signed with bilinear quadrilateral elements on a structured mesh
2 p
h ! and in conjugation with the conventional topological/geometrical
cos h pint ph þ are m sin h pint p
arctan
; ð26Þ enrichment scheme. A 2 2 Gaussian quadrature rule is employed
aim
sin h pint h
m p for every non-enriched element, while more integration points, e.g.
re im
qm h; am ; am 4 4, is applied to non-partitioned enriched elements, especially
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i ffi for enriched elements partitioned into sub-triangles, even higher-
1 4 re im 2 re 2 2
¼ pffiffiffi am þ am þ am sin 2h are m þ a im
m 1 cos 2h ; order Gaussian rules, e.g. 6 6, can be utilizable for each sub-trian-
2 gle. The domain radius for the interaction integral is taken 0.8a.
ð27Þ The total energy release rate (TERR), GT, and mechanical energy re-
lease rate (MERR), GM, are calculated using
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
r¼ ðx xtip Þ2 þ ðy ytip Þ2 ; GT ¼ Y 22 K2I þ 2Y 24 KI KIV þ 2Y 45 KIV KV þ Y 44 K2IV þ Y 55 K 2V ;
2
1
h = atan2(y ytip, is the four-quadrant inverse tangent function; GM ¼ Y 22 K 2I þ Y 24 KI KIV þ Y 25 KI KV ; ð29Þ
2
x xtip)
N is the set of all nodes in the discretization; where ymn(m, n = 2, 4, 5) evaluated from the Irwin matrix Y [37].
NTIP the set of all nodes that are connected to The plain-strain case is considered with tensile loading ryy = r1
elements containing crack tip (s); in-plane electric-displacement Dy = D1 and in-plane magnetic
Ncr the set of nodes that are connected to induction By = B1 prescribed on the remote boundary of a 2-D
elements containing the crack but not in NTIP. MEE specimen. The material is assumed to be a composite consist-
ing of piezoelectric phase BaTiO3 and piezomagnetic phase Co-
Fe2O4, and material constants presented in Table 1 are taken
from Li and Lee [10].
The set of nodes selected on the basis of conventional topolog-
ical and the geometrical enrichment are shown in Fig. 1. Substitut- 5.1. Numerical validation for Griffith–Irwin’s crack in an infinite
ing approximate displacement from Eq. (20), electric potential domain
from Eq. (21) and magnetic potential from Eq. (22) into the weak
´ az [30], the standard discrete system
form illustrated by Rojas-Dı To analyze the fracture parameters, Griffith’s crack in an infinite
of equations are obtained as: domain is considered with the applied loading conditions
ryy = r1 = 5 MPa, Dy = D1 2(e33/c33)ryyC/m2 and By = B1 = 2(h33/
K s d ¼ f ext ð28Þ
c33)ryy NA1 m1 at infinity as shown in Fig. 2. The polarization
s ext
where j is the vector of external nodal forces and f the stiffness direction is taken along Y-axis. The numerical computations are
matrix. made on a finite specimen with dimensions 4 4 and a crack
Table 1
Material constants for BaTiO3–CoFe2O4.
1
10
errW
0
10
Topological-Standard(slope~0.5)
Topological-Four basis(slope~0.5)
Without enrichment(slope~0.48)
Geometrical enrichment(rg=0.2a)-slope~1.0
-1
10
1 2
10 10
Fig. 2. Griffith’s crack normal to polarization direction in an infinite domain.
1/he
Fig. 3. Convergence of error in energy norm with respect to 1/he for Griffith’s crack.
length, 2a = 1. As the computational domain is finite, thus the exact
solutions for Griffith’s crack in an infinite domain are applied as the
boundary conditions for such finite computational domain. The ex-
act solution for the symmetric MEE Griffith problem without shear -1
10
at infinity with poling direction perpendicular to the crack is ex-
tended here by the authors and presented in Appendix A.
(KI(exact) - KI(X-FEM)) / KI(exact)
ex ex -4
where eex
j ; Ej and Hj are the strain, electric field and magnetic field
10
1 2
components obtained from the exact solution given in Appendix A. 10 10
The error in the total energy norm with respect to the number 1/he
of elements per unit length is plotted on a log–log scale in Fig. 3.
Fig. 4. Relative error in KI with respect to 1/he for Griffith’s crack.
The results in the total energy norm are analyzed using the topo-
logical and the geometrical enrichment strategies. A radius of
0.2a is applied to the geometrical enrichment and both are com- Furthermore, the results using the X-FEM converge better than
pared with the cases of topological four basis and without enrich- that using without enrichment functions.
ment. A convergence rate of 0.5 is obtained for the both topological
enrichment cases whereas the geometrical one achieves the higher
convergence rate than the classical X-FEM, i.e. 1.0. It is also 5.1.2. Convergence study in the intensity factors
observed that the results of four basis functions are in good Figs. 4–6 present the convergence study of the IF KI, KIV & KV
approximation to the standard eight basis functions defined for versus the number of elements per unit length in each direction,
MEE materials and this is in agreement to Rojás-Diaz et al. [30]. respectively. An excellent agreement of the relative errors in the
80 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
-1 7
10 K I (X-FEM)
K I (10 5 Nm -3/2 )
Topological-Standard K I (Exact/Li&Lee)
Fracture Parameters
K V (Exact/Li&Lee)
G T (N/m)
GT (X-FEM)
4 GT (Exact/Li&Lee)
-2
10
3
-3
10 0
10
1
10
2 0.5 1 1.5 2 2.5 3
1/he half-crack length
Fig. 5. Relative error in KIV with respect to 1/he for Griffith’s crack. Fig. 7. Variation of IFs and TERR with respect to half-crack length for center crack.
-1
10
Topological-Standard
Topological-Four basis
(KV(exact) - KV(X-FEM)) / KV(exact)
Without enrichment
Geometrical enrichment(rg=0.2a)
-2
10
-3
10
1 2
10 10
1/he
KI, KIV, and KV is observed for standard eight basis functions with Fig. 8. Model geometry of the double edge collinear cracks.
respect to 1/he. In this investigation, we again found the same as
accounted for the total energy norm errors. The geometrical
5.3. Numerical validation for double edge collinear cracks in a finite
enrichment is better than the topological one and further both
specimen
these enrichment approaches are significantly better than the
one without enrichment case. A slightly higher accuracy is found
A double edge collinear crack problem in a finite MEE specimen
for the standard eight basis functions than the four basis functions.
having dimensions 4 8 is considered. The length of each edge
crack is taken as a = 0.5 such that a/w = 0.25. The applied loadings
5.2. Comparison with results of Li and Lee [10] for center crack on the boundary of a finite specimen are ryy = 5 MPa, Dy =
problem 1010ryy C/m2 and By = 109ryy NA1 m1. The model geometry,
loadings and the symmetric boundary conditions considered are
A finite specimen of width 2w = 30 mm, length 2h = 30 mm is shown in Fig. 8. The results obtained using X-FEM are compared
taken for the analysis. Since, the computational domain is finite with the results of BEM given by Rojás-Diaz et al. [30] for a finite
so the applied loadings are considered from the exact solution as MEE specimen. The relative errors of IFs corresponding to BEM
in Section 5.1. The domain radius for the interaction integral is ta- results are plotted on a log–log scale with respect to number of ele-
ken 0.8a. Results of KI, KIV, KV and TERR are calculated for different ments per unit length (1/h e) and shown in Figs. 9–11. An excellent
crack lengths, and are shown in Fig. 7. A structured mesh with agreement of the X-FEM and BEM results are obtained with respect
99 99 elements in conjugation with topological and standard to number of elements per unit length. Here too, we observe that
eight basis functions is considered for the numerical calculations. the geometrical enrichment is better than the topological one and
It is observed that the results of IFs and TERR obtained using X- further X-FEM results are significantly better than the without
FEM are in close agreement with the results of Li and Lee [10]/exact enrichment case. A good approximation of four and eight standard
results. basis functions is found for KI and KIV but in case of KV a slightly
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 81
10
-1 5.4. Numerical validation for two equal centered collinear cracks in an
infinite domain
Fig. 9. Relative error in KI with respect to 1/he for double edge cracks.
-1
10
(KIV(BEM) - KIV(X-FEM)) / KIV(BEM)
-2
10
Topological-Standard
Topological-Four basis
Without enrichment
Geometrical enrichment(rg=0.2a)
-3
10
1 2
10 10
1/he
Fig. 10. Relative error in KIV with respect to 1/he for double edge cracks.
-1 Geometrical enrichment(rg=0.2a)
10 GM (By =10-6σ yy )
GM (By =-10-6σ yy )
(GM(X-FEM) - GM(Tian)) / GM(Tian)
GM (By =0)
-2
10
-3
10
σyy=5MPa
-8
Dy = 10 σyy
-4
10
1 2
10 10
1/he
-2
10
0 1 2
Fig. 11. Relative error in KV with respect to 1/he for double edge cracks. 10 10 10
a/he
higher accuracy can be seen for standard eight basis functions than
the four basis functions. Fig. 13. Relative error in GM with respect to a/he for two equal collinear cracks.
82 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
induction show better convergence rate than the other two cases rd/d0 Tip a1 Tip b2
of applied magnetic induction loadings. Kffiffiffiffiffiffiffi
p I ffi K IV
p ffiffiffiffiffiffiffiffi Kffiffiffiffiffiffiffi
p V ffi Kffiffiffiffiffiffiffi
p I ffi K ffiffiffiffiffiffiffi
p IV ffi Kffiffiffiffiffiffiffi
p V ffi
ryy pa01 Dy pa01 By pa01 ryy pa02 Dy pa02 By pa01
2a02 = 4.0 mm and inter-crack space d0 = 1.5 mm is considered. rd/d0 Tip a1 Tip b2
The offset distances of left crack and the right crack from the edge Kffiffiffiffiffiffiffi
p I ffi K ffiffiffiffiffiffiffi
p IV ffi Kffiffiffiffiffiffiffi
p V ffi Kffiffiffiffiffiffiffi
p I ffi K ffiffiffiffiffiffiffi
p IV ffi Kffiffiffiffiffiffiffi
p V ffi
ryy pa01 Dy pa01 By pa01 ryy pa02 Dy pa02 By pa01
of specimen are b11 and b12, respectively. Boundaries of specimen
are prescribed in-plane loading ryy = r1 = 5 MPa, an electric dis- 0.47 1.191 1.189 1.195 1.305 1.306 1.317
0.53 1.317 1.315 1.300 1.450 1.451 1.437
placement, Dy = D1 = 2(e33/c33)ryy C/m2 and magnetic induction
0.67 1.319 1.316 1.300 1.453 1.452 1.437
By = B1 = 2(h33/c33)ryy NA1 m1. The entire configuration is sche- 0.75 1.310 1.312 1.300 1.451 1.450 1.434
matically shown in Fig. 14. MEE material, BaTiO3–CoFe2O4 with
volume fraction, Vf = 0.5 is chosen for study. A structured mesh
with 99 99 bilinear quadrilateral elements and topological
6.2.1. Effect of aspect ratio
enrichment approach is considered for the numerical solution.
In this section length, 2 h, of the specimen is varied from 10 mm
to 40 mm, whereas other geometric parameters and the loadings
6.1. Domain independence
remain the same as defined in Section 6. The number of elements
in the mesh is taken according to the dimensions of the specimen
Tables 2 and 3 show the normalized values of IFs at the outer
such that length of the element is approximately equal to 0.25 mm.
tips and inner tips of the cracks with respect to the domain radius
Fig. 15 shows the variation of TERR with respect to the aspect ratio,
rd/d0 (in case of d0 < a02), respectively. It is concluded from these ta-
h/w, at inner and outer tips of both the cracks. It is observed that
bles that the domain independence at the outer and inner tips of
TERR stabilizes for aspect ratio h/w P 1. Hence, a square specimen
both the cracks are observed for the same value of rd P 0.53d0.
is sufficient to consider the effects of far field loadings. The devia-
Similarly, domain independence is also observed for rd = 0.5a02
tion at the low aspect ratio is because of the applied loads at the
(in case of d0 P a02). Therefore, in further subsections, rd = 0.53d0
finite specimen boundaries still affect the TERR.
(for d0 < a02) and rd = 0.6a02 (for d0 P a02) are fixed for the domain
radius at outer and inner tips.
6.2.2. Effect of offset distances
6.2. Effect of finite specimen Now, the effect of finiteness of specimen on TERR is observed by
varying the dimensions of specimen 2w = 2h from 18 mm to
To study the finite specimen effects the following analyses have 38 mm. The cracks lengths and the inter-crack space are fixed,
been investigated: 2a01 = 6.0 mm, 2a02 = 4.0 mm, and d0 = 1.5 mm, respectively. For
numerical simulation number of elements is chosen in such a
way that the length of an element is approximately equal to
0.25 mm. Fig. 16 depicts variation of TERR with increasing values
Fig. 15. Variations of TERR versus aspect ratio analysis for two-unequal-collinear
Fig. 14. Model geometry of the two-unequal-collinear cracks. cracks.
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 83
Fig. 18 depicts the variation at the four tips of the cracks with
respect to the increase in smaller crack length when the bigger
crack length is kept fixed. TERR increases at all the tips of the
cracks when length of the smaller crack is increased. It may also
be noted that TERR is higher at inner tips of the cracks than those
at outer tips, as expected. For two equal collinear cracks the values
of TERR become equal at inner/outer tips of bigger crack with the
corresponding values of inner/outer tips of the smaller crack. The
behavior of TERR at inner and outer tips of both the cracks are sim-
ilar to that in Ref. [10] Li and Lee.
Fig. 17 shows the behavior of total energy release rate with re-
spect to increase in inter-crack space. The analysis is done for two
different crack lengths taken for smaller crack, a02 = 2.0 mm and
1.0 mm, respectively. It is observed that the TERR is more at the in-
ner tips of the cracks than that at the outer tips of the cracks. It is
further noted that increasing inter-crack space d0; the mutual
influence of cracks on each other decreases. Consequently, TERR
at the inner and outer tips becomes equal for both lengths of short Fig. 18. Effect of crack length on TERR for two-unequal-collinear cracks.
84 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
The results of TERR obtained for asymmetric orientation case of Fig. 20. Model geometry of the special case of two-unequal-collinear cracks.
two-unequal-collinear cracks with respect to inter- crack space d0
is depicted in Fig. 19. These results for symmetric orientation
(d1 = d2) case are also plotted in the same figure. Fig. 19 shows that
the TERR decreases with increase in inter-crack distance for both
symmetric as well as asymmetric orientation cases (except at outer
tip a1, of bigger crack where TERR slightly increases). It may be
noted that in case of a bigger crack, the difference in TERR for sym-
metric and asymmetric orientation cases is not significant for
smaller values of d0 but as the inter-crack space, d0 is increased,
asymmetric orientation case gives energy release rate higher than
that in the symmetric one. This is the case when bigger crack is clo-
ser to the boundary of the finite specimen. TERR does not show a
significant difference at the tips of the smaller crack with respect
to inter-crack space considered here for symmetric and asymmet-
ric orientations. A similar variation of TERR is seen for smaller
crack in asymmetric orientation case when d1 is fixed and d2 is
varied.
Anisotropic K c (X-FEM)
0.08 Isotropic K c (Exact)
Anisotropic K c (Exact)
0.06
0.04
0.02
-3 -2 -1 0 1 2 3
θ
Fig. 29. Behavior of modified hoop stress intensity factor with respect to h.
Fig. 30. Model geometry of two-unequal-collinear edge cracks in MEE material for
quasi-static crack propagation.
Table 4
Comparison of critical crack propagation angle.
crack. And crack propagation rate for bigger edge crack is less than
hc (anisotropic) hc (isotropic)
the smaller crack for all the incremental steps. It may be due to the
Exact 7.90° 28.20° fact of asymmetry of problem investigated. Further, it is seen that
X-FEM 7.99° 28.35°
the behavior of crack propagation paths for both bigger and smal-
Janski’s et al. [34] 7.0° 26.0°
ler cracks is not mesh sensitive in MEE ceramic and these approach
to a smoother curve with increase in number of elements.
-1 -5
69x139 69x139
-2 79x159 -10 79x159
89x179 89x179
99x199 99x199
-4 -20
-5 -25
-6 -30 Δa = 2.5*(5/79)
rd = = 2.0*(5/79)
-7 -35
-8 -40
1.5 2 2.5 3 2.8 3 3.2 3.4 3.6 3.8 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 31. Influence of mesh size on the crack propagation path of two-unequal-collinear edge cracks.
0.1 -5
Δ a = 3.5h e
-10
Δ a = 3.0h e
0.05
Δ a = 2.5h e -15
Crack propagation angle
Crack propagation angle
Δ a = 2.0h e
-20
0
he = 5/79
rd = 2.0he -25
-0.05
-30 Δ a = 3.5h e
Δ a = 3.0h e
-35
-0.1 Δ a = 2.5h e
-40 Δ a = 2.0h e
-0.15 -45
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 32. Influence of crack growth increment on the crack propagation path of two-unequal-collinear edge cracks.
-1 -5
rd = 2.0he rd = 2.0he
-20
-4
-25
-5
he = 5/79
-30
Δ a = 2.5he
-6
-35
-7 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 33. Influence of domain radius on the crack propagation path of two-unequal-collinear edge cracks.
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 89
20 10
Vf =0.9996 Vf =0.9996
5
15
Vf =0.5 Vf =0.5
0
Vf =0.0724 Vf =0.0724
he = 5/99 -10
5 Δ a = 2.5he -15
rd = 2.0he
-20
0
-25
-30
-5
-35
-10 -40
1.5 2 2.5 3 3.5 2 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 34. Influence of volume fraction on the crack propagation path of two-unequal-collinear edge cracks.
Table 5
Crack propagation angles at different incremental steps for two-unequal-collinear edge cracks in MEE ceramic with different volume fractions.
0 -5
he = 5/79
σ xx = 0.25 σ yy
Δ a = 2.5he
rd = 2.0he
-10 σ xx = 0.5 σ yy
-5
σ xx = 0.75 σ yy
-25
-15
-30
-20
-35
-25 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 36. Influence of mechanical loading on the crack propagation path of two-unequal-collinear edge cracks.
Crack propagation angle
-8 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Crack propagation angle
0 0
-4 -20
-6 -30
-8 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Crack propagation angle
-2 0
-8 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 37. Influence of electrical loading on the crack propagation path of two-unequal-collinear edge cracks.
crack propagation growth rate. And it is observed for both bigger loading. Therefore, more specific anisotropic fracture toughness
and smaller edge cracks. Further, this effect is observed more in needs to be considered which depends upon not only h but also on
case of higher volume fraction whereas it is almost independent applied mechanical/electrical and magnetic loadings. This requires
of applied electrical loading for Vf = 0.0724. It may have also the more experimental evidence to be performed on these ceramics.
same reason that a lower value of volume fraction has less
piezoelectric coupling effect and consequently, it is independent 7.2.7. Influence of magnetic loading
of applied electrical loading. The observation of electrical loading The effect of magnetic loading is demonstrated here for differ-
on crack propagation growth rate in MEE ceramic is similar to ent volume fractions on crack propagation paths of two-unequal-
the results given by Sih et al. [18]. collinear edge cracks in MEE ceramic. The factor k = 0, 1, 2, 3 and
It is also observed that even the anisotropic fracture toughness 4, respectively has been varied only for magnetic flux loadings
does not highlight the influence of direction of applied electrical By ¼ kðh33 =c33 Þryy NA1 m1 .
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 91
Fig. 38 shows that there is not a significant difference in the loading, discussed in Section 7.2.6, here too, crack propagation
simulated crack propagation paths of both edge cracks under the growth rate is independent of the direction of applied magnetic
influence of different applied magnetic flux loading, and for all load.
the volume fractions Vf = 0.9996, 0.5 and 0.0724. But, it is also no- Consequently, this study also emphasizes the need of more suit-
ticed from numerical values that crack propagation growth rate de- able anisotropic fracture toughness parameter for crack propaga-
creases with increase in magnetic flux loading. Similar to electrical tion criterion in MEE ceramics. The effect of magnetic loading on
Crack propagation angle
-8 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Crack propagation angle
-8 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Crack propagation angle
0 0
-4 -20
-6 -30
-8 -40
1.5 2 2.5 3 2.5 3 3.5 4
Extended length of a bigger crack Extended length of a smaller crack
Fig. 38. Influence of magnetic loading on the crack propagation path of two-unequal-collinear edge cracks.
5 5
Δ a = 2.5he Δ a = 2.5he
0 0
-5 he = 5/99 -5
a = b = 1.5m
Crack propagation angle
rd = 2.0he
-10 -10
-15 -15
-20 -20
-25 -25
-30 -30
-35 -35
1.5 2 2.5 3 3.5 1.5 2 2.5 3 3.5
Extended length of a left crack Extended length of a right crack
Fig. 39. Crack propagation path of two-equal-collinear edge cracks in MEE ceramics.
92 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
10m
5m
Fig. 42. Anomaly associated with the basis functions at the marked Gauss points.
Fig. 40. Simulated crack propagation paths of two-equal-collinear edge cracks in
MEE.
that of R’ethor’e et al. [42]. It concludes that the behavior of crack
propagation paths of two-equal-collinear edge cracks in MEE
crack propagation growth rate observed is similar to the findings of
ceramics is similar to the crack propagation paths for two-equal-
Sih et al. [18].
collinear edge cracks in isotropic materials. The behavior of
mechanical/electrical/magnetic IFs for the left end edge crack are
7.2.8. Crack propagation study of two-equal-collinear edge cracks shown in Fig. 41 with respect to extended crack length. It is ob-
In this section, a particular case of two equal collinear edge served that except KII all other IFs increase with respect to increase
cracks is simulated by considering a = b=1.5 m. The applied in crack length. Intensity factors for right end edge crack are not
mechanical loads are rxx = ryy = 5 MPa whereas the other applied significantly different from the IFs for the left end edge crack at
loadings remain the same as described in subsection 7.2. A mesh all incremental steps.
with 99 199 elements is considered for the simulation of crack
paths of both left and right edge cracks. Fig. 39 clearly depicts
the symmetrical crack propagation path of left edge crack to the 8. Anomaly associated with standard eight basis functions for
right edge crack, as expected. A better visual representation of MEE material
the crack propagation paths of both the left and right edge cracks
of two-equal-edge cracks in MEE ceramic is shown in Fig. 40. An From Eq. (27), modified radii of the basis functions are defined
excellent agreement of crack propagation paths is observed with as
x 107 x 107
15 1
KI K II
0
10
KII
KI
-1
5
-2
0 -3
1.5 2 2.5 3 3.5 1.5 2 2.5 3 3.5
Extended length of a left crack Extended length of a left crack
x 10-4
0.03 8
K IV KV
6
0.02
KIV
KV
0.01
2
0 0
1.5 2 2.5 3 3.5 1.5 2 2.5 3 3.5
Extended length of a left crack Extended length of a left crack
Fig. 41. Intensity factors for left crack of two-equal-collinear edge cracks in MEE ceramic.
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 93
Table 6
Roots of the characteristic Eq. (A.9) with positive imaginary part for different MEE ceramics and volume fractions.
Vf 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
!
BaTiO3/CoFe2O4 0.69449i 0.71216i 0.73114i 0.75144i 0.77339i 0.79756i 0.82500i 0.85823i 0.91005i 0.93246i 0.69350i
0.93154i 0.94464i 0.94357i 0.94197i 0.94002i 0.93758i 0.93427i 0.92863i 0.91005i 0.93246i 0.97657i
1.45761i 1.43424i 1.40967i 1.38328i 1.35476i 1.32364i 1.28923i 1.25037i 1.20479i 1.14640i 0.01466 + 1.01095i
1.96434i 1.95111i 1.93678i 1.92079i 1.90228i 1.87976i 1.85040i 1.80813i 1.73712i 1.57890i 0.01466 + 1.01095i
PZT-PIC151/CoFe2O4 0.69449i 0.70453i 0.71585i 0.72851i 0.74281i 0.75915i 0.77811i 0.80058i 0.82807i 0.86368i 0.64152i
0.93154i 1.12382i 1.12990i 1.12855i 1.12455i 1.11871i 1.11103i 1.10102i 1.08761i 1.06818i 0.91557i
1.45761i 1.45556i 1.45523i 1.45484i 1.45424i 1.45329i 1.45176i 1.44917i 1.44373i 1.41407i 1.03478i
1.96434i 1.94686i 1.92608i 1.90256i 1.87542i 1.84314i 1.80300i 1.74939i 1.66884i 1.52789i 1.45747i
PZT-6B/CoFe2O4 0.69449i 0.66941i 0.64707i 0.62664i 0.60775i 0.59018i 0.57375i 0.55835i 0.54389i 0.53033i 0.51760i
0.93154i 1.01272i 1.02085i 1.02299i 1.02331i 1.02279i 1.02175i 1.02035i 1.01861i 1.01654i 0.70402i
1.45761i 1.51419i 1.56905i 1.62205i 1.67251i 1.71826i 1.75353i 1.76512i 1.73027i 1.59141i 1.01453i
1.96434i 1.96209i 1.95943i 1.95637i 1.95315i 1.95074i 1.95227i 1.96494i 1.99556i 2.04226i 2.10146i
1 -1
10 10
-2
10
0 -3
10 10
1 2 1 2
10 10 10 10
1/he 1/he
-1 -1
10 10
(KIV(exact) - KIV(X-FEM)) / KIV(exact)
Topological-Standard Topological-Standard
Topological-Four basis Topological-Four basis
-2 -2
10 10
-3 -3
10 10
1 2 1 2
10 10 10 10
1/he 1/he
Fig. 43. Convergence of error in energy norm and relative error in IFs with respect to 1/he for PZT6B-CoFe2O4 (Vf = 0.5).
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i
1 4
2 2 2 2 crack tip element(s). The use of higher order Gauss points for
qm h; arem ;aim
m ¼ pffiffiffi arem þ aim m þ are m sin2h are
m þ aim m 1 cos 2h
2 blending elements and triangularisation process at crack tip ele-
ments always increase this possibility. Fig. 42 shows one of the
In particular, if the roots of the characteristic Eq. (A.9) are purely an same case at the marked Gauss points for BaTiO3–CoFe2O4 with
imaginary then Vf = 0.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Authors have tabulated the roots of the characteristic Eq. (A.9)
1 4 2 h 2 i
qm h; arem ; aim
m ¼ pffiffiffi aimm þ 1 aim m cos 2h: ð34Þ with positive imaginary part for different volume fractions of a
2 MEE material and also for different MEE materials in Table 6. In
2 2 h 2 i
Now, if aim
m < 0:5 then sign of aim m þ 1 aim
m cos 2h depends particular, it is found that for BaTiO3–CoFe2O4, a volume fraction,
on cos 2h. Vf > 0.9996 and Vf < 0.0724 (up to four decimal places) give one
2
As weknow p and p, therefore the possi- of the roots as purely imaginary and of aim < 0:5. And if we re-
2 that
h h lies between
2
i m
bility of aim
m þ 1 aim m cos 2h to be negative cannot be ne- place the material constants of BaTiO3 with material constants of
glected. Hence, qm is imaginary. PZT-6B as considered in [43], a more critical case is obtained. In
There may be several cases possible where the position of the this case for all volume fractions one can get at least one of the
2
crack tip(s) is in such a way that it is perpendicular or nearly per- roots as purely imaginary with aim m < 0:5 as shown in Table 6.
pendicular to the Gauss point(s) of the blending element(s) or the The material constants for PZT-PIC 151 and PZT-6B are taken from
94 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
[44]. This implies that for all volume fractions, there is always a where
possibility of finding a Gauss point at blending element(s) or crack 8
tip element where modified radius is imaginary. Therefore, in this >
> re im wm ðh;are
m ;am Þ
im
im
> q
< m
h; am ; am cos 2
ifam > 0;
case the standard eight enrichment functions for crack tip cannot
g m ðhÞ ¼ ð37Þ
be taken guaranteed to represent the crack tip solution or to repre- >
> wm ðh;are
m ;am Þ
im
> qm h; are
: ; aim
sin if aim
m 6 0:
sent the field variables near the crack tip, which are actually real in m m 2
1 -1 -1
10 10 10
-2 -2
10 10
0 -3 -3
10 10 10
1 2 1 2 1 2
10 10 10 10 10 10
1/he 1/he 1/he
Fig. 44. Convergence of error in energy norm and relative error in IFs with respect to 1/he for BaTiO3–CoFe2O4 (Vf = 0).
1 -1 -1
10 10 10
-2 -2
10 10
0 -3 -3
10 10 10
1 2 1 2 1 2
10 10 10 10 10 10
1/he 1/he 1/he
Fig. 45. Convergence of error in energy norm and relative error in IFs with respect to 1/he for BaTiO3–CoFe2O4 (Vf = 1).
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 95
of error in total energy norm convergence and IFs is observed in ceramics. Therefore, a more suitable fracture toughness parameter
subsections 5.1 and 5.2. Hence, one could consider the eight basis is required which depends not only on h but also on the applied
functions based on the redefined modified radius as a generalized mechanical/electric/magnetic loadings. And for that more experi-
basis functions, which are applicable to all the MEE ceramics or their mental work on these ceramics is needed to be performed.
degenerated cases. Apart from the study on two-unequal-collinear cracks, an
interesting point regarding the application of standard eight basis
functions defined for these ceramics is also obtained. It is observed
9. Conclusions that the modified radius defined for these basis functions are not
applicable to all the MEE ceramics and for all the volume fractions.
From the last decade most of the work on two-collinear-cracks Therefore, authors alter the modified radius and redefined the
has been done on an infinite domain using analytic techniques. As basis functions which can handle the crack problems of all the
per the authors knowledge, this is the first attempt, to study the MEE ceramics and for all the volume fractions including their
quasi-static stationary and propagating two-unequal-collinear degenerated cases.
cracks in a finite specimen of 2-D MEE ceramic using X-FEM.
Two-collinear-unequal cracks in an infinite domain problem are Acknowledgement
simulated, analyzed and also validated. A square specimen having
width greater than or equal to five times of the length of a bigger The authors are grateful to Prof. R.D. Bhargava (Senior Professor
crack and the offset distances greater than or equal to 1.38 times and Head, (retd.), Indian Institute of Technology Bombay, Mumbai,
than the length of a bigger crack, may be considered as an infinite India) for the encouragement throughout the course of this work.
domain in the present set up. An excellent agreement of the behav-
iors is observed with an infinite domain under different parametric Appendix A
studies namely inter-crack space, crack length and the applied
mechanical/electrical/magnetic loads for quasi-static problem. Let (x, y) be the principal material and crack coordinates system
Total energy release rate strongly depends on the cracks dimen- and y be the poling direction as shown in Fig. 2.
sions and the distance between them. The effect of cracks on each With respect to the coordinate system (x, y) and under the
other reduces when the distance between the cracks is greater plane strain conditions (eyy = eyx = exy = Ey = Hy = 0), the constitutive
than or equal to the length of the bigger crack. It is concluded from relations for magnetoelectroelastic materials can be expressed as:
the investigation for TERR for asymmetric cracks orientation case 8 2 38
that TERR is either more (for the crack near to the edge of the spec- >
>
exx 9>
>
a11 a12 0 0 b21 0 d21 > rxx >
>
9
>
>
> >
> 6 7>
> >
>
imen) or approximately same at tips of cracks to the case of sym- >
> eyy > > 6 a12 a22 0 0 b22 0 d22 7>
> ryy >
>
>
> >
> 6 >
7> >
>
metric orientation for a fixed value of inter-crack space. The >
> >
> 6 7>
> >
>
>
> 2 e >
> 6 7>
> r >
>
concept of TERR cannot be taken as a fracture parameter as the >
>
<
xy > 6 0
>
=
0 a33 b13 0 d13 0 7>>
<
xy >
>
=
6 7
TERR is independent of the direction of the applied electrical and Ex ¼6
6 0 0 b13 d11 0 D11 7
0 7 Dx
>
> >
> 6 7>
> >
>
magnetic loads. However, MERR shows the shielding and unshield- >
> >
> >
> Dy >>
>
> E > 6 b21 b22 0 0 d22 0 D22 7
7> >
ing effect on the crack growth for applied negative and positive > y >
> > 6
> 6
>
>
7>
>
>
>
>
> >
> 6 >
> >
>
electrical/magnetic loads, respectively. The case of one edge and > H
> x >
>
>
> 4 0 0 d13 D11 0 f11 0 7 >
> x >
5> B >
>
another internal crack in a 2-D finite specimen of MEE specimen
>
: >
; >
: >
;
Hy d21 d22 0 0 D22 0 f22 By
is derived as a special case from two-unequal-collinear cracks
problem discussed. ðA:1Þ
Crack propagation of two-unequal-collinear edge cracks in MEE where coefficients aij, bij, dij, dij, Dij and fij are reduced material con-
material is successfully attempted here using X-FEM and modified stants defined in Rojas-Dı´az [30].
hoop stress intensity factor criterion with anisotropic fracture In the above Eq. (A.1), fexx eyy 2exy gT ; fEx Ey gT and fHx Hy gT de-
toughness parameter defined for polarized ceramics. The different note respectively the strain, electric field and magnetic field vec-
parametric studies conclude that propagation path for both the tors w.r.t. (x, y) system; frxx ryy rxy gT ; fDx Dy gT and fBx By gT
cracks are mesh insensitive, independent of domain radius but de- denote respectively the stress, electric displacement and magnetic
pend upon the crack growth increment. Due to asymmetry of the induction vectors w.r.t (x, y).
problem, crack propagation paths for bigger and smaller cracks Based on Lekhnitskii’s formalism, three complex potential func-
are also asymmetric to each other whereas paths are obtained per- tions v(x, y), #(x, y) and w(x, y) are introduced as:
fectly symmetric in case of two-equal-collinear edge cracks. The
difference in the crack propagation paths for different volume frac- @ 2 vðx; yÞ @ 2 vðx; yÞ @ 2 vðx; yÞ
rxx ¼ ; ryy ¼ ; rxy ¼ ;
tions of MEE ceramic is not significant here but still the behavior of @y2 @x2 @x@y
crack propagation paths agrees with Sih et al. [18,20] that an in- @#ðx; yÞ @#ðx; yÞ @wðx; yÞ @wðx; yÞ
Dx ¼ ; Dy ¼ ; Bx ¼ ; By ¼ :
crease in volume fraction decreases the crack growth rate. It is also @y @x @y @x
concluded that the increase in rxx/ryy ratio increases the crack ðA:2Þ
growth rate for a bigger edge crack. The same behavior is observed
for a smaller crack but up to a certain crack length, and it shows the The above defined relations in Eq. (A.2) in potential functions satisfy
reverse behavior after that. Electric and magnetic flux loadings al- the equilibrium equations
ways reduce the crack propagation growth rate. Although, the ef- @ rxx @ rxy @ rxy @ ryy @Dx @Dy @Bx @By
fect of magnetic loading is not significant but electrical loading þ ¼ 0; þ ¼ 0; þ ¼ 0; þ ¼ 0:
@x @y @x @y @x @y @x @y
has a significant effect. The above effect of loadings is also in agree-
ðA:3Þ
ment of the strain energy density function criterion for crack prop-
agation in MEE materials proposed by Sih et al. [18]. But the crack By substituting Eq. (A.2) into Eq. (A.1) and invoking the following
propagation paths obtained from the modified hoop stress inten- strain and electric field compatibility equations:
sity factor criterion with anisotropic fracture toughness parameter
@ 2 exx @ 2 eyy @ 2 exy @Ex @Ey @Hx @Hy
do not show the effect of electric/magnetic loading direction. This þ 2 ¼ 0; ¼ 0; ¼ 0: ðA:4Þ
@y 2 @x 2 @x@y @x @y @x @y
could not be validated with the experimental findings on these
96 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
The three coupled differential equations in v(x, y), #(x, y) and w(x, y) Substituting Eqs. (A.11) and (A.12) into Eq. (A.2) and introducing
can be obtained as: new complex potential functions
L4 vðx; yÞ L3 #ðx; yÞ M 3 wðx; yÞ ¼ 0; @ vk ðzk Þ @nk ðzk Þ
L3 vðx; yÞ L2 #ðx; yÞ þ M 2 wðx; yÞ ¼ 0; ðA:5Þ
nk ðzk Þ ¼ and 1k ðzk Þ ¼ for k ¼ 1; 2; 3; 4 ðA:14Þ
@zk @zk
M 3 vðx; yÞ M 2 #ðx; yÞ þ N2 wðx; yÞ ¼ 0: the stress, electric displacement and magnetic field components can
where be expressed in terms of the new complex potential functions as:
8 9 8 9
@2 @2 @2 @2 @2 @2
N 2 ¼ f22 2 þ f11 2 ; M2 ¼ D22 2 þ D11 2 ; L2 ¼ d22 2 þ d11 2 < rxx >
> = l2k >
4 <
P
>=
Dx P
4
k k lk
@x @y @x @y @x @y ryy ¼ 2Re 1 1k ðzk Þ; ¼ 2Re 1k ðzk Þ;
>
: >
; k¼1>
: >
; Dy k¼1 kk
@3 @3 @3 @3
r
xy lk
L3 ¼ ðb13 þ b21 Þ þ b22 3 ; M 3 ¼ ðd13 þ d21 Þ 2 þ d22 3 ; Bx P
4
gk lk
@y2 @x @x @y @x @x
B
¼ 2Re
g
1k ðzk Þ: ðA:15Þ
y k¼1 k
@4 @4 @4
L4 ¼ a11 4 þ ð2a12 þ a33 Þ 2 2 þ a22 4 : Using the Eq. (A.15) into Eq. (A.1) and applying the gradient rela-
@y @y @x @x
tions, the general solution for displacement, electric and magnetic
ðA:6Þ potential can be written as:
Eliminating #(x, y) and w(x, y) in Eq. (6), the governing equations of 8 9 8 9 8 9
> ux > > Pk > > h y þ u >
>
> > > >
> >
> >
> >
>
plane piezoelectricity are reduced to the following eighth-order dif- < uy = P
4 < Q k =lk = < h x þ v =
ferential equation: ¼ 2Re nk ðzk Þ þ ðA:16Þ
h i >
>/> > k¼1>
> Sk > > >
> / > >
>
: > ; >
: >
; >
: >
;
L4 L2 N2 M 22 þ L3 ðL3 P2 2M 3 M 2 Þ þ L2 M23 vðx; yÞ ¼ 0: ðA:7Þ u Uk u
Eq. (A.7) can be written symbolically as (in Sosa [45]): where
F 1 F 2 F 3 F 4 F 5 F 6 F 7 F 8 vðx; yÞ ¼ 0 ðA:8Þ Pk ¼ a11 l2k þ a12 b21 kk d21 gk ;
where F k ¼ l @
@y
and lks(k = 1, 2, . . . , 6) are the roots of the
@
k @x Q k ¼ a12 l2k þ a22 b22 kk d22 gk ;
characteristic equation ðA:17Þ
Sk ¼ b13 lk þ d11 lk kk þ D11 lk gk ;
½a11 l4 þ ð2a12 þ a33 Þl2 þ a22 ½ðD11 l2 þ D22 Þ2 U k ¼ d13 lk þ D11 lk kk þ f11 lk gk :
ðd11 l2 þ d22 Þðf11 l2 þ f22 Þ ½ðb21 þ b13 Þl2 þ b22 2 ðf11 l2 þ f22 Þ
Here h⁄, u⁄, v⁄, /⁄ and u⁄ represent the rigid body motion, a refer-
½ðd21 þ d13 Þl2 þ d22 2 ðd11 l2 þ d22 Þ ence electric and magnetic potential and they are set to be zero.
þ 2½ðb21 þ b13 Þl2 þ b22 ½ðd21 þ d13 Þl2 þ d22 ðD11 l2 þ D22 Þ ¼ 0 On the basis of Griffith’s crack solution in classical elasticity, the
ðA:9Þ complex potential function 1k(zk) must have the following form:
Az
k k
In general, the roots of the Eq. (A.9) are complex with four conjugate 1k ðzk Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ Bk ; ðk ¼ 1; 2; 3; 4Þ ðA:18Þ
pairs, i.e. z2k a2
l1 ¼ a1 þ ib1 ; l2 ¼ a2 þ ib2 ; l3 ¼ a3 þ ib3 ; l4 ¼ a4 þ ib4 ; where Ak ¼ Are
im re im
k þ iAk ; Bk ¼ Bk þ iBk and a is the half crack length.
l5 ¼ l1 ; l6 ¼ l2 ; l7 ¼ l3 ; l8 ¼ l4 The above defined complex potential function satisfied the
ðA:10Þ boundary conditions at the crack flanks and the conditions of con-
pffiffiffiffiffiffiffi stant stresses and constant electric displacements at infinity. It also
where i ¼ 1; ak s and bk sðk ¼ 1; 2; 3; 4Þ are all real numbers. satisfies the equilibrium equations.
Eq. (A.8) can be solved by means of complex variables, and the Thus, one can know about the complete solution of the Griffith’s
general solution for the complex potential function v(x, y) as: crack problem using Eqs. (A.15)–(A.18). And it required only
16
P
4 arbitrary real constants, i.e. Are im re im
k ; Ak ; Bk ; Bk for k ¼ 1; 2; 3; 4 eval-
vðx; yÞ ¼ 2Re vk ðzk Þ ðA:11Þ uated from the boundary conditions at the crack surfaces and at
k¼1
infinity, and exclusion of the rigid body rotations.
where vk is an arbitrary function of complex variable zk = x + lky. Applying the boundary conditions at the crack surfaces.
From Eqs. (A.5) and (A.11), the general solution for the complex
potential function #(x, y) and w(x, y) can be expressed as: 1. rxy = 0 for (jxj < a, y = 0), Eqs. (A.15)–(A.18) results into
2 0 13
P
4 @ vk ðzk Þ P4 @ v ðzk Þ
#ðx; yÞ ¼ 2Re kk ; wðx; yÞ ¼ 2Re gk k ðA:12Þ 6P 4
B Ak ðx þ lk yÞ C7
k¼1 @zk k¼1 @zk 2Re4 lk @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi þ Bk A5 ¼ 0
2
k¼1 ðx þ lk yÞ a 2
where
2x h
re re im re re im
l3 ðlk Þ m3 ðlk Þ ) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi lim 1 A1 þ l 1 A1 þ lim 2 A2 þ l 2 A2
kk ¼ ;g ¼ ; 2 2
l1 ðlk Þ k l1 ðlk Þ a x i
re re im re re im
2 þ lim3 A3 þ l3 A3 þ lim4 A4 þ l4 A4
l1 ðlk Þ ¼ d22 þ d11 l2k f22 þ f11 l2k D22 þ D11 l2k ; m3 ðlk Þ h
re im im re im im
l3 ðlk Þ ¼ ðd13 þ d21 Þl2k þ d22 D22 þ D11 l2k 2 lre 1 B1 l1 B1 þ lre 2 B2 l2 B2
i
ðb13 þ b21 Þl2k þ b22 f22 þ f11 l2k þ lre re im im
þ lre re im im
3 B3 l3 B3 4 B4 l4 B4 ¼0
¼ ðb13 þ b21 Þl2k þ b22 D22 þ D11 l2k
re re im re re im
) lim 1 A1 þ l1 A1 þ lim2 A2 þ l2 A2
ðd13 þ d21 Þl2k þ d22 f22 þ f11 l2k :
re re im re re im
ðA:13Þ þ lim3 A3 þ l 3 A3 þ lim 4 A4 þ l 4 A4 ¼0 ðA:19Þ
R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98 97
and 1. ryy ¼ r1
yy ; jx þ iyj ! 1
3 n o
P re im
lre1 Bre1 lim im
1 B1 þ lre re im im
2 B2 l2 B2 þ lre re im im
3 B3 l3 B3 ) 2Re Ak þ Bre im
k þ i Ak þ Bk ¼ r1
yy
k¼1
lre4 Bre4 lim im re
þ 4 B4 ¼0 ðA:20Þ ) 2 Are re re 1
1 þ A2 þ A3 þ A4 ¼ ryy ðA:27Þ
2. ryy = 0 for (jxj < a, y = 0), Eqs. (A.15)–(A.18) results into
2 0 13
2. rxy ¼ 0; jx þ iyj ! 1 ) lre1 Are1 þ Bre1
6P4
B Ak ðx þ lk yÞ C7 re re
2Re4 @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi þ Bk A5 ¼ 0 re re
lre re
2 A2 þ B2 l3 A3 þ B3
k¼1 ðx þ lk yÞ2 a2
re
2x lre re im
4 A4 þ B4 þ l1 Aim
1 þ B1
im
þ lim Aim im
2 þ B2
) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Aim1 þ Aim2 þ Aim 3 þ Aim
4
2
a2 x2
þ 2 Bre re re
1 þ B2 þ B3 þ B4
re
þ lim
3 Aim im
3 þ B3 þ lim
4 Aim im
4 þ B4 ¼0 ðA:28Þ
¼0 ) Aim
1 þ Aim
2 þ Aim
3 þ Aim
4 ¼0 ðA:21Þ
n 2 o
and 3.
rxx ¼ 0; jx þ iyj ! 1 ) lre1 2 lim
1 Are re
1 þ B1
Bre
1 þ Bre
2 þ Bre
3 þ Bre
4 ¼0 ðA:22Þ n 2 o
3. Dy = 0 for (jxj < a, y = 0), Eqs. (A.15)–(A.18) results into
þ lre2 2 lim
2 Are re
2 þ B2
2 0 13 n 2 o
þ lre3 2 lim Are re
3 þ B3
B Ak ðx þ lk yÞ
3
6P 4
C7
2Re4 kk @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi þ B k A5 ¼ 0 n
2 2 o h
k¼1 ðx þ lk yÞ a 2
þ lre4 2 lim
4 Are re re im
4 þ B4 2 l1 l1 Aim im
1 þ B1
2x h
im re im im re im
) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi k1 Are 1 þ k 1 A1 þ k2 Are 2 þ k 2 A2 þ lre im
Aim im
þ lre im
Aim im
2 2 2 l2 2 þ B2 3 l3 3 þ B3
a x i
im re im im re im i
þ k3 Are 3 þ k 3 A3 þ k4 Are 4 þ k 4 A4 þlre im
4 l4 Aim im
4 þ B4 ¼0 ðA:29Þ
h
re im im re im im
2 k1 Bre 1 k1 B1 þ k2 Bre 2 k2 B2
i
re
þ k3 Bre
im im re
þ k4 Bre
im im re re re re
3 k3 B3 4 k4 B4 ¼0 re re
4. Dy ¼ D1
y ; jx þ iyj ! 1 ) 2 k1 A1 þ B1 þ k2 A2 þ B2
im re re im
) k1 A1 þ k1 A1 þ k2 A2 þ k2 A2
im re re im re re re
þk3 Are re
3 þ B3 þ k4 A4 þ B4
re
h
im re im im re im
þ k3 Are 3 þ k3 A3 þ k4 Are 4 þ k 4 A4 ¼0 ðA:23Þ im
þ 2 k1 Aim im im
þ k2 Aim im im
þ k3 Aim im
1 þ B1 2 þ B2 3 þ B3
i
and im
þk4 Aim 4 þ B4
im
¼ D1 y ðA:30Þ
re im im re im im re im im
k1 Bre
1 k1 B1 þ k2 Bre 2 k2 B2 þ k3 Bre
3 k3 B3
h
re im im
þ k4 Bre 4 k4 B4 ¼0 ðA:24Þ 5. Dx ¼ 0; jx þ iyj ! 1 ) lre1 kre1 lim im
Are re
1 k1 1 þ B1
6P 4
B Ak ðx þ lk yÞ C7
2Re4 gk @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ Bk A5 ¼ 0 þ lre3 kre3 lim im
3 k3 Are re
3 þ B3
k¼1 ðx þ lk yÞ2 a2
i
2x h re im im
þ lre
4 k4 l4 k4 Are re
4 þ B4
) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi gim 1 Are
1 þ gre
1 Aim
1 þ gim 2 Are
2 þ gre
2 Aim
2
2 2
a x i h
þ gim re
gre3 Aim þ gim re re im lre1 kim im re
1 þ l1 k 1 Aim im
1 þ B1
3 A 3 þ 3 4 A4 þ g4 A4
h
re im im re im im
2 gre 1 B1 g1 B1 þ gre 2 B2 g2 B2 þ lre2 kim im re
2 þ l2 k 2 Aim im
2 þ B2
i
þ gre re im im
þ gre re im im
3 B3 g3 B3 4 B4 g4 B4 ¼0
þ lre3 kim im re
3 þ l3 k 3 Aim im
3 þ B3
im re re im im re re im re re im
) g1 A1 þ g1 A1 þ g2 A2 þ g2 A2 þ gim 3 A3 þ g3 A3 i
þ lre
im im re
Aim im
re re im 4 k4 þ l4 k4 4 þ B4 ¼0 ðA:31Þ
þ gim 4 A4 þ g4 A4 ¼0 ðA:25Þ
and
re re re
re re
6. By ¼ B1y ; jx þ iyj ! 1 ) 2 g1 A1 þ B1 þ g2 A2 þ B2
re
re re
gre1 Bre1 gim im
1 B1 þ gre re im im
2 B2 g2 B2 þ gre re im im
3 B3 g3 B3 þgre re re
3 A3 þ B3 þ g4 A4 þ B4
re
h
re im im
þ gre
4 B4 g4 B4 ¼0 ðA:26Þ þ 2 gim1 Aim
1 þ B1
im
þ gim2 Aim
2 þ B2
im
þ gim3 Aim3 þ B3
im
i
Applying the boundary conditions at infinity and using Eqs. (A.15)– þgim Aim im
¼ B1
4 4 þ B4 y ðA:32Þ
(A.18), we get
98 R.R. Bhargava, K. Sharma / Computational Materials Science 60 (2012) 75–98
re
7. Bx ¼ 0; jx þ iyj ! 1 ) lre re im im
A1 þ Bre [3] Z.-G. Zhou, L.-Z. Wu, B. Wang, Eur. J. Mech. A-Solid. 24 (2005) 253–262.
1 g1 l1 g1 1 [4] Z.-G. Zhou, P.-W. Zhang, L.-Z. Wu, Philos. Mag. 90 (2010) 1743–1769.
re re
þ l2 g2 lim 2 g2
im
Are
2 þ B2
re [5] B.-L. Wang, H.Y. -Zhang, J.-C. Han, Arch. Appl. Mech. 77 (2007) 541–558.
re re re [6] B.M. Singh, J. Rokne, R.S. Dhaliwal, Eur. J. Mech. A-Solid. 28 (2009) 599–
þ l3 g3 lim 3 g3
im
A3 þ Bre 3 609.
re re [7] X.-C. Zhong, Appl. Math. Model. 35 (2011) 2930–2944.
þ l4 g4 lim 4 g4
im
Are
4 þ B4
re
[8] R. Rojas-Diaz, F.G. -Sanchez, A. Saez, Ch. Zhang, Int. J. Fract. 157 (2009) 119–
h 130.
im im re
lre 1 k1 þ l1 k1 Aim
1 þ B1
im
[9] Y.-D. Li, K.Y. Lee, J.-W. Pan, Collinear unequal crack series in
magnetoelectroelastic materials: anti-plane case, J. Appl. Math. Mech.
im im re
þ lre 2 k 2 þ l2 k 2 Aim2 þ B2
im
doi:http://dx.doi.org/10.1002/zamm.201000053/pdf.
[10] Y.-D. Li, K.Y. Lee, Eng. Fract. Mech. 77 (2010) 2772–2790.
im im re
þ lre 3 k 3 þ l3 k 3 Aim3 þ B3
im [11] F.G. -Sanchez, R. Rojas-Diaz, A. Saez, Ch. Zhang, Theor. Appl. Fract. Mech. 47
(2007) 192–204.
im i
[12] B.-L. Wang, Y.-W. Mai, Comput. Meth. Appl. Mech. Engg. 196 (2007) 2044–
þ lre im im re
4 g4 þ l4 g4 A4 þ Bim 4 ¼0 ðA:33Þ 2054.
[13] Y. Li, W. Feng, Z. Xu, Comput. Meth. Appl. Mech. Engg. 198 (2009) 2347–2359.
[14] S.B. Park, C.T. Sun, Int. J. Fract. 70 (1995) 203–216.
[15] S. Kumar, R.N. Singh, Acta Mater. 44 (1996) 173–200.
8. Now, to remove the rigid rotation, the following condition is [16] S. Kumar, R.N. Singh, Acta Mater. 45 (1997) 859–868.
imposed as stated in Sosa [45]: [17] X.-L. Xu, R.K.N.D. Rajapakse, Acta Mater. 48 (2000) 1865–1882.
[18] Z.F. Song, G.C. Sih, Theor. Appl. Fract. Mech. 39 (2003) 189–207.
[19] G.C. Sih, R. Jones, Z.F. Song, Theor. Appl. Frac. Mech. 40 (2003) 161–186.
uy;x ux;y ¼ 0; jx þ iyj ! 1 [20] G.C. Sih, H.Y. Yu, Compos. Struct. 68 (2005) 1–11.
) X 1 Are þ Bre þ X 2 Are þ Bre þ X 3 Are þ Bre [21] N. Moës, J. Dolbow, T. Belytschko, Int. J. Numer. Meth. Eng. 46 (1999) 131–
1 re 1 re 2re 2re 3re 3re 150.
þ X 4 A4 þ B4 Y 1 A1 þ B1 Y 2 A2 þ B2 [22] T. Belytschko, T. Black, Int. J. Numer. Meth. Eng. 45 (1999) 601–620.
re
Y 3 Are re re
3 þ B3 Y 4 A4 þ B4 ¼ 0 ðA:34Þ [23] G. Ventura, E. Budyn, T. Belytschko, Int. J. Numer. Meth. Eng. 58 (2003) 1571–
1592.
[24] R. Huang, N. Sukumar, J.-H. Prévost, Int. J. Solids Struct. 40 (2003) 7539–
where 7552.
2 3 [25] E. Be’chet, H. Minnebo, N. Moës, B. Bugardt, Int. J. Numer. Meth. Eng. 64 (2005)
Q re re im im
k lk þ Q k lk
1033–1056.
X k ¼ 4 2 2 P k lk Pk lk 5 and Y k
re re im im
[26] P. Laborde, J. Pommier, Y. Renard, M. Salaün, Int. J. Numer. Meth. Eng. 64
lrek þ lim k (2005) 354–381.
2 im re im
3 [27] A. Yazid, N. Abdelkader, H. Abdelmadjid, Appl. Math. Model. 33 (2009) 4269–
Q k lk Q k lk
re 4282.
¼ 4 2 2 P k lk þ P k lk 5 ; k
re im im re [28] T. Belytschko, R. Gracie, G. Ventura, Model. Simul. Mater. Sc. 17 (2009) 043001.
lrek þ lim [29] R. Rojas-Dı´az, N. Sukumar, A. Saez, F.G. Sanchez, Crack analysis in
k magnetoelectroelastic media using the extended finite element method, in:
¼ 1; 2; 3; 4: ðA:35Þ International Conference on Extended Finite Element Methods – Recent
Developments and Applications, 2009.
[30] R. Rojas-Dı´az, N.Sukumar, A. Sáez, F. Garcı́a-Sánchez, Fracture in
Now, using Eqs. (A.19)–(A.35),
we can evaluate all the 16 arbi- magnetoelectroelastic materials using the extended finite element method,
trary real constants, i.e. Are im re im
k ; Ak ; Bk ; Bk for k ¼ 1; 2; 3; 4 and Int. J. Numer. Meth. Eng. (2011) doi:http://dx.doi.org/10.1002/nme.3219.
hence the field variables near the tip of the crack. [31] E. Be’chet, M. Scherzer, M. Kuna, Int. J. Numer. Meth. Eng. 77 (2009) 1535–
Further, applying the classical definition of intensity factor and 1565.
[32] M.A. Gregory, C.T. Herakovich, J. Compos. Mater 20 (1986) 67–85.
Eqs. (A.19)–(A.35), the IFs are given as [33] M. Doblare, F. Espiga, M. Alcantud, Eng. Fract Mech 37 (1990) 953–967.
pffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi [34] L. Janski, P. Steinhorst, M. Kuna, IUTAM (2009) 163–173.
KI ¼ lim 2pr ryy ðx ¼ a þ r; y ¼ 0Þ ¼ r1yy pa ðA:36Þ [35] R. Rojas-Dı´az, M. Denda, F. Garcı́a-Sánchez, A. Sáez, Eur. J. Mech. A-Solid. 31
r!0
pffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi (2012) 152–162.
KIV ¼ lim 2prDy ðx ¼ a þ r; y ¼ 0Þ ¼ D1 y pa ðA:37Þ [36] M. Denda, Key Eng. Mater. 383 (2008) 67–84.
r!0 [37] B.N. Rao, M. Kuna, Int. J. Fract. 153 (2008) 15–37.
pffiffiffiffiffiffiffiffiffi p ffiffiffiffiffiffi
KV ¼ lim 2prBy ðx ¼ a þ r; y ¼ 0Þ ¼ B1 y pa ðA:38Þ [38] S. Osher, J.A. Sethian, J. Comput. Phys. 79 (1988) 12–49.
r!0 [39] M. Stolarska, D.L. Chopp, N. Moës, T. Belytschko, Int. J. Numer. Meth. Eng. 51
(2001) 943–960.
This implies that IFs for Griffith’s crack problem are indepen- [40] W.-Y. Tian, U. Gabeert, Eur. J. Mech. A-Solid. 23 (2004) 599–614.
dent of the material constants. [41] G.G. Pisarenko, V.M. Chusko, S.P. Kovalev, J. Am. Ceram. Soc. 68 (1985) 259–
265.
[42] J. R’ethor’e, S. Roux, F. Hild, Comptes Rendus Mecanique 338 (2010) 121–
References 126.
[43] C.-C. Ma, J.-M. Lee, Int. J. Solids Struct. 46 (2009) 4208–4220.
[1] C.-F. Gao, H. Kessler, H. Balke, Int. J. Eng. Sci. 41 (2003) 983–994. [44] Z.-C. Ou, Y.-H. Chen, Int. J. Fract. 130 (2004) 427–454.
[2] Z.-G. Zhou, B. Wang, Y.-G. Sun, Int. J. Eng. Sci. 42 (2004) 1155–1167. [45] H. Sosa, Int. J. Solids Struct. 28 (1991) 491–505.