Computational Materials Science 116 (2016) 11–21

Contents lists available at ScienceDirect

Computational Materials Science

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

Computational simulations of thermal shock cracking by the virtual

crack closure technique in a functionally graded plate
V.N. Burlayenko a,⇑, H. Altenbach b, T. Sadowski c, S.D. Dimitrova a
a
Department of Applied Mathematics, National Technical University ‘KhPI’, 21 Frunze Str., 61002 Kharkov, Ukraine
b
Faculty of Engineering Mechanics, Otto-von-Guericke University of Magdeburg, 2 Universitätsplatz, 39106 Magdeburg, Germany
c
Department of Solid Mechanics, Lublin University of Technology, 40 Nadbystrzycka Str., 20-618 Lublin, Poland

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

Article history: Thermal cracking in a ceramic/metal functionally graded plate under thermal shock loading, when the
Received 30 June 2015 plate is cooled from high-temperature to ambient low one, is numerically analyzed with the commercial
Received in revised form 16 August 2015 finite element software ABAQUSTM. Continuous position-dependent functions were created for the
Accepted 17 August 2015
mechanical and thermal coefficients of FGM and, then, implemented as user-defined material properties
Available online 5 September 2015
via respective subroutines of the ABAQUS code. In order to find a temperature field and associated ther-
mal stresses in the FGM plate, a linear quasi-static thermoelastic problem for a plane strain state is
Keywords:
solved. The distributions of temperature and thermal induced stresses accounting for residual stresses
Functionally graded materials
Thermal shock
inside the plate are calculated under conditions of both the steady state and transient thermal processes.
Thermal fracture The solution of the transient heat conduction problem is used for crack propagation simulation using the
Finite element method virtual crack closure technique. The crack lengths developed on the ceramic surface during the thermal
Virtual crack closure technique shock for different graduation profiles are computed. It is shown that the crack lengths are influenced by
the material gradient profile of the functionally graded plate. Hence, the conclusions on the crack resis-
tance of FGM plates are drawn.
Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction tural components in transportation, energy, electronics and

It is well-known that traditional layered composite materials
and metals possessing high strength and toughness at room and
moderately elevated temperatures becomes incapable for employ-
ing under the ultra high temperatures. On the other hand ceramic
materials demonstrate superior properties in heat resistance. How-
ever, the use of ceramics in engineering applications is usually lim-
ited due to their low toughness. Combining gradually ceramic and
metal materials into one material system, so-called a functionally
graded material (FGM), allows one to get advantages of the both
materials as an alternative to structures with a bimaterial inter-
face. The ceramic constituents of FGMs withstand high tempera-
tures due to their excellent thermal resistance, while the metal
constituents provide better mechanical performance reducing the
possibility of catastrophic brittle fracture. FGMs originally devel-
oped as thermal barrier coatings for aerospace structures and
fusion reactors [1] have nowadays received a wide spread as struc-
⇑ Corresponding author.
E-mail addresses: burlayenko@kpi.kharkov.ua (V.N. Burlayenko), holm.
altenbach@ovgu.de (H. Altenbach), t.sadowski@pollub.pl (T. Sadowski),
s.dimitrova@kpi.kharkov.ua (S.D. Dimitrova).
biomedical engineering for the general use in high temperature
environments [2]. Hence, studies of thermal-induced stresses in
FGMs are extremely necessary. Moreover, the existence of micro-
cracks and other defects in FGMs due to features of fabrication pro-
cesses can essentially affect their material behavior. Therefore, the
fracture analysis of FGMs under thermal loading conditions are
important to ensure their durability in engineering applications.
A number of issues must be addressed in order to effectively
model thermal cracking in functionally graded materials. These
include profiles of material gradients and their magnitudes, speci-
men geometry, residual stress distributions, crack-wake and pro-
cess zone effects for advancing crack, etc. In this respect, various
analytical and computational approaches have been employed to
investigate crack-tip stress distribution and crack propagation in
FGMs [3]. The correspondence between crack-tip fields in homoge-
neous and non-homogeneous continua, which permits application
of standard analysis techniques in FGMs has been demonstrated in
[4,5]. This fact in conjunction with the simplest approximations of
graded material properties allowed researchers to get analytically
tractable solutions of many practical tasks reducing them to one-
dimensional models. For instance, a closed-form solution of ther-
mal stress intensity factor (TSIF) for a strip of metal/ceramic FGM

http://dx.doi.org/10.1016/j.commatsci.2015.08.038
0927-0256/Ó 2015 Elsevier B.V. All rights reserved.
12 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21
with exponentially varying thermo-mechanical properties with an
edge crack subjected to steady state thermal load was obtained in
[6], while the transient thermal problem for such type of graded
materials was presented in [7,8] for a cracked semi-infinite plate
and a plate with an edge crack, respectively. Authors in [9] have
used an analytical approach based on the layered model to study
edge cracked strips of ceramic/ceramic and ceramic/metal FGMs
with constant elastic moduli and thermal properties graded
according to the power law under transient thermal loading condi-
tions. Both the distributions of temperature and thermal-induced
stresses in vicinity of the crack have been found.
Different sophisticated techniques for solving more compli-
cated thermo mechanical fracture problems in FGMs have also
been found in the literature. In [10] the thermal shock problem
in a ceramic/metal FGM strip with thermal and thermoelastic con-
stants, varying as polynomial functions in the width direction was
considered and the influence of residual stresses in the FGM strip
on the TSIF and the length of cracks developing due to the thermal
shock were evaluated by comparing experimental data and theo-
retical results. Authors in [11] proposed a meshless method based
on the local boundary integral equation and the moving least
square approximation for computing the dynamic stress intensity
factors in continuously non-homogeneous FGM solids under a
transient dynamic load. A crack problem in functionally graded
semi-infinite plane with embedded periodic surface cracks sub-
jected to model I mechanical loads or thermal loading has been
solved by using Fourier integrals and Fourier series in [12]. An
approach based on integral equations has been used to study frac-
ture processes in the vicinity of interface cracks in functionally
graded/homogeneous bimaterials subjected to tensile loading and
a heat flux in [13,14]. This technique was also applied to investi-
gate the interaction of two edge cracks inclined arbitrary to the
interface of a functionally graded layer and homogeneous sub-
strate under cyclic heating–cooling thermal loading in [15]. A high
potential for solving fracture mechanics tasks of structural ele-
ments of FGMs has the direct approach that follows the original
idea of Cosserat, this technique was successfully applied for ana-
lyzing mechanical behavior of uncracked functionally graded
beams and plates in [16,17].
Despite these efforts, a clear description of fracture and crack
growth in graded materials has remained a challenging task yet
and requires powerful and precise numerical methods. The Finite
Element Method (FEM) is undoubtedly a favorite technique for
analyzing mechanical and thermal problems including those
involving FGMs. The main issue encountered in application of the
FEM to FGMs is concerned with modelling a material with contin-
uously varying properties. The simplest way involves the use of
homogeneous elements each with different properties, however,
it leads to a stepwise change in properties in the direction of the
material gradient. Such models have already been used by a num-
ber of researchers and reasonable results have been obtained, e.g.
[18] among many others. A more advanced way of including prop-
erty variation into a FE model is to utilize elements that themselves
contain a gradient in properties. Authors in [19] proposed a two-
dimensional (2-D) graded finite element with material properties
evaluated directly at the Gauss points. An alternative element
was developed in [20], a fully isoparametric element formulation
that interpolates material properties at each Gauss point from
the nodal values, using the same shape functions as the deforma-
tions was proposed. One can notice that comparisons of graded
and homogenous elements under various loading conditions with
analytical solutions in the literature showed that graded elements
give far greater accuracy for modelling FGMs [21].
Moreover, in the case of FGMs, appropriate modifications
accounting for material inhomogeneity must be made as compared
with the analysis of homogeneous materials for predictions of
stresses near the crack-tip and tracking the path of a crack growing
in the material, where mode mixity arises not only from geometri-
cal and loading configurations but also from the material nonho-
mogeneity. It has been proven in [22] that when the traditional
version of the J-integral is implemented into the FEM via the
domain integration method, it is path dependent in the case of
FGMs, and it can provide accurate results only if it is evaluated very
close to the crack tip using small elements. A modified path inde-
pendent ~J-integral, computed for the nonhomogeneous materials
was also proposed in [23] later. However, this quantity is cumber-
some to compute. Previous works on crack growth simulation in
FGMs includes a number of prediction models. For example, a local
remeshing technique to predict the crack path in the quasi-static
four-point bending fracture test has been used in [24]. The intrinsic
cohesive element formulation in the context of finite element
method can be found, e.g. in [25,26]. Recently, the novel extended
finite element method (XFEM) has been applied to model arbitrary
crack growth in FGMs under thermo-mechanical loading in [27,28]
among others.
In this paper, a finite element-based model is described, for a
graded ceramic/metal plate containing an edge crack initially per-
pendicular to the material gradient under thermal shock loading
conditions. Investigating this configuration provides insight into
the failure behavior of graded materials in thermal resistance
applications. Experiments have shown that for ceramic/metal
FGMs, cracks generally nucleate near the ceramic surface exposed
to the environment temperature and then grow towards the metal
side. To understand clearly this phenomenon by means of numer-
ical modelling, the paper proposes a finite element approach
within the ABAQUS code [29], where the two-dimensional graded
finite element with spatially varying thermal and mechanical prop-
erties is developed using appropriate user-defined subroutines.
The propagation of crack induced by thermal shock in conjunction
with residual stresses is simulated using the virtual crack closure
technique (VCCT). The VCCT originally developed for homogeneous
materials in [30] by using the Irwin’s assumption has proven its
efficiency for composite materials, e.g. [31,32]. This technique is
also very suitable for FGMs, since it is based on the local character-
istic of the crack tip parameters. The crack lengths depending on
the material profiles are discussed here in detail. This is of partic-
ular interest since works reported previously in the literature on
this subject are limited.
2. Material properties
In order to model temperature and strain/stress distributions in
FGMs, one needs an appropriate estimate for thermomechanical
properties of the graded composites. A large number of papers
on effective material properties of graded materials have been
published, and they can be classified largely into theoretical
homogenization techniques and experimental methods. The theo-
retical predictions are usually based on the framework of contin-
uum theory of nonhomogeneous media and often exploit other
simple rules of mixtures or variational methods or micro-
mechanical models [2].
In this paper, a plate of width w ¼ 10 mm and length
2l ¼ 40 mm made of a metal/ceramic functionally graded material
composed of particles of the titanium alloy Ti–6Al–4V and the zir-
conium dioxide ZrO2 ceramic is considered, Fig. 1. The mechanical
and thermophysical properties of the constituents of the graded
material at a reference temperature of 1000 K, taken from the
paper [18], are presented in Table 1.
The effective mechanical and thermophysical properties of the
Ti–6Al–4V/ZiO2 FGM are calculated from its constituent materials
and their volume fractions. It is assumed that the volume fraction
 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21 13

Fig. 2. Compositional profiles of the functionally graded material.

 
1 þ 3ðjm  jc ÞV m
jðxÞ ¼ jc ; ð6Þ
3jc þ ðjm  jc ÞV c

cm qm V m þ cc qc V c
cðxÞ ¼ ; ð7Þ
qm V m þ qc V c
where E; m; q; a; j and c stand for Young’s modulus, Poisson’s ratio,
Fig. 1. Functionally graded plate. the mass density, the coefficient of linear thermal expansion, the
thermal conductivity and the specific heat, respectively. For sim-
of the metal phase in the direction of x-coordinate is approximated plicity, the material properties are assumed to be independent of
using a power function: temperature.

x p
Vm ¼ ð1Þ 3. Mathematical problem statement
w
The volume fraction of the ceramic phase is V c ¼ 1  V m . Here Cracking in the FGM plate, shown in Fig. 1, under thermal shock
and in what follows, the subscripts c and m of the material param- loading is discussed from the macroscopic standpoint, i.e. by eval-
eters denote the ceramic and metal phases, respectively. From (1) uating macroscopic stress components induced in the FGM plate
it follows that the FGM is ceramic rich when the parameter p > 1 due to both a temperature gradient and residual stresses. This
and metal rich when the parameter p < 1. Fig. 2 shows the volume problem is stated as a linear thermoelastic task in conjunction with
fraction of the metal phase for various values of p over the plate fracture mechanics of FGM structures, where the crack growth is
width w. driven by thermal stresses. Each the analysis, performed in the
The width variation of the thermo-mechanical parameters in paper is shortly presented below.
the FGM with negligible porosity are estimated by means of the
rule of mixtures of a two-phase material as follows, [18]: 3.1. Thermal analysis
( )
2=3
Ec þ ðEm  Ec ÞV m
EðxÞ ¼ Ec 2=3
; ð2Þ A generalized boundary value problem for thermal heat con-
Ec þ ðEm  Ec ÞðV m  V m Þ duction in a continuously non-homogeneous isotropic medium
occupying a domain X can be described by the following governing
mðxÞ ¼ mm V m þ mc V c ; ð3Þ equation [9]:

qðxÞ ¼ qm V m þ qc V c ; ð4Þ @Tðx; tÞ

qðxÞcðxÞ ¼ ½jij ðxÞT ;j ðx; tÞ;i þ qðx; tÞ ð8Þ
@t
am V m Em =ð1  mm Þ þ ac V c Ec =ð1  mc Þ
aðxÞ ¼ ; ð5Þ where Tðx; tÞ is a temperature field, qðx; tÞ is a density of body heat
V m Em =ð1  mm Þ þ V c Ec =ð1  mc Þ sources, jij ðxÞ is the thermal conductivity tensor, qðxÞ is the mass
density and cðxÞ is the specific heat, which are functions of spatial
position x ¼ ðx; y; zÞ in the Cartesian coordinate system. It is
Table 1 assumed that appropriate boundary conditions on a surface @ X of
Material properties of Ti–6Al–4V and ZiO2. the domain X and initial condition at a moment of time t ¼ 0 are
Property Constituents given as follows:
Ceramic ZiO2 Metal Ti–6Al–4V Tðx; tÞ ¼ Tðx; tÞ on @ XT
Young’s modulus E, MPa 117 66.2 ðx; tÞ
qðx; tÞ ¼ q on @ Xq
Poisson’s ratio m 0.333 0.32 ð9Þ
Mass density q, kg/m3 5600 4420 and
Coefficient of thermal Tðx; t ¼ 0Þ ¼ T 0 ðxÞ
expansion a, 106 1/K 7.11 10.3
Thermal conductivity j, W/m K 2.036 18.1 where Tðx; tÞ and q ðx; tÞ are a prescribed temperature and a heat
Specific heat c, J/kg K 615.6 808.3
pffiffiffiffiffi flux on the surfaces @ XT and @ Xq , respectively, such that
Fracture toughness K IC , MPa m 5 60
@ X ¼ @ XT [ @ Xq and T 0 ðxÞ is an initial temperature.
14 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21
Moreover, we assume that the heat flux in (8) obeys the law of
heat conduction in a non-homogeneous medium:
qðx; tÞ ¼ jij ðxÞT ;j ðx; tÞni ð10Þ
and the convection conditions between the medium and its sur-
roundings with the given heat transfer coefficient hðxÞ at an envi-
ronmental temperature T env ðxÞ:
qðx; tÞ ¼ hðxÞðTðx; tÞ  T env ðxÞÞ ð11Þ
on corresponding parts of the surface @ Xq with the unit outward
normal ni .
3.2. Stress analysis
In mechanical analysis we focus on FGMs undergoing elastic
deformations governed by the equations of non-homogeneous
elastic materials. Those basic equations include the equation of
motion without body forces:
@ 2 ui
rij;j ¼ qðxÞ ; ð12Þ
@t 2
where the inertial terms will be neglected in what follows, the
strain–displacement relations:
1 placements in the y-direction between released nodes at P, respec-
eij ¼ ðui;j þ uj;i Þ ð13Þ
2 tively. In the case of the problem at hand, the quantities in (16) will
and the Hooke’s law of an isotropic material with space-dependent
elasticity coefficients:
rij ¼ kðxÞeelkk dij þ 2lðxÞeelij ð14Þ
Here rij is the stress tensor, eij is the infinitesimal total strain
tensor, which is assumed to be decomposed into a linear elastic
part, eelij and strains caused by the thermal expansion,
ij ¼ aðxÞDTdij . Hence, the thermal analysis is involved into the
eth
mechanical one through the thermal strains. kðxÞ and lðxÞ are the
Lame’s constants and aðxÞ is the coefficient of thermal expansion,
which are pointwise functions of location x and DT is a relative
change in temperature at the point x.
Then, a boundary value problem of elasticity of FGMs consists of
the basic Eqs. (12)–(14) in the given domain X and the following
boundary conditions on its surface @ X:
rij nj ¼ ti on @ Xr
ð15Þ
i
ui ¼ u on @ Xu ;
where t i denotes the prescribed traction on the surface @ Xr with the
unit outward normal vector nj , u i stands for the prescribed displace-
ments on the surface @ Xu and @ Xr [ @ Xu ¼ @ X.
3.3. Fracture analysis
If other nonlinearities than a crack propagation can be neglected,
methods based on Linear Elastic Fracture Mechanics (LEFM) have
been proven to be effective in crack modelling. The virtual crack
closure technique is one of the most commonly applied approaches
in conjunction with the FEM for determining the components of
strain energy release rate along the crack front. Herewith, no
assumption of isotropy or homogeneity around the crack is neces-
sary. This approach allows to avoid both the usual extensive mesh
refinement and remeshing in the crack propagation analysis, but
it requires the existence of the pre-defined discontinuity.
An edge crack of length a located perpendicular to the material
gradient is assumed in the FGM plate as shown in Fig. 1. The crack
propagates when the TSIF at the crack tip exceeds the fracture
toughness of the FGM. The nature of the near crack tip displace-
ment field in FGMs is the same as for homogeneous materials
[4,5]. However, both stress intensity factors and fracture toughness
parameters are functions of material gradients. According to the
VCCT it is assumed that the strain energy released when a crack
is extended by a certain amount is identical to the energy required
to close the crack by the same amount [30]. It means that there is a
similarity between crack extension from node Q to node P along
one element and crack closure at node P as shown in Fig. 3a, where
the case of Mode I loading is only illustrated. Then, the energy to
close the crack and, thus, the energy to open the crack is defined
by the work done by the stresses rij (or equivalent nodal reaction
forces F i ) at the crack front (node Q) over the total appropriate
opening displacements Dui of the first node behind the crack front
(node P) produced by the virtual crack extension da in a single
finite element analysis [31]. It allows one to calculate, in turn,
the strain energy release rate (SERR) as a function of nodal forces
and displacements only. For instance, in the case of the opening
Mode, considered in Fig. 3a, GI can be computed as follows:
1 F Q DuP
 ¼ GI ; ð16Þ
2 DA
where DA ¼ dab is the area of the crack surface of the width b
formed due to the crack extension da, and F Q and DuP denote the
node reaction force in the y-direction at Q and the difference in dis-
be calculated from the solution of the thermoelasticity task men-
tioned above. Moreover, the mode I stress intensity factor can be
easily found by using the known relationship for the plane strain
state as follows:
 
E
K 2I ¼ GI ð17Þ
1  m2 tip
Here, ‘tip’ means that values of E and m should be taken at the
crack tip.
Further Fig. 3b illustrates an example of using the VCCT to
model the Mode I crack growth in the case of two-dimensional
four-noded elements. Nodes 2 and 5 will start to release when
the following crack propagation criterion is met at the current cal-
culation step:
1 F v 2;5 v 1;6
 ¼ GI P GIC ; ð18Þ
2 bd
where GIC is the Mode I critical SERR related to the critical stress
intensity factor (or fracture toughness), also in the Mode I, by the
equation similar to (17), F v 2;5 is the vertical reaction force between
nodes 2 and 5, v 1;6 is the vertical (opening) displacement between
released nodes 1 and 6, b is the width and d is the length of the ele-
ment at the crack front.
Similar arguments can be used and the equations can be
obtained for the Mode II and Mode III components of the SERR in
the case of two- or three-dimensional cracks, e.g. [33]. However,
in the mixed-mode case, the node at the crack tip will separate, sim-
ulating the crack growth, if the equivalent strain energy release rate
at the crack tip exceeds the equivalent critical strain energy release
rate for the material: Geq P GeqC , where, for instance, the total strain
energy release rate GT can be used as Geq . To model failure driven by
several fracture mode types at the same time, the power law model
will be used in the succeeding calculations as follows:
 k1  k2  k3
GI GII GIII Geq
þ þ ¼ ð19Þ
GIC GIIC GIIIC GeqC
The model should be provided by the mixed-mode test data
such as components of the critical SERRs GIC ; GIIC and GIIIC and the
fitting coefficients k1 ; k2 and k3 for the material studied.
 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21
Fig. 3. VCCT approach for pure Mode I: (a) crack open and closure, (b) crack growth criterion.
4. Finite element modelling
In the thermal cracking problem described in the previous sec-
tion, the plate is presented by 2-D model with the crack plane
which is a plane of symmetry as shown in Fig. 1. The finite element
model is created using the general purpose finite element analysis
software ABAQUSTM, [29]. The continuous variation of the thermo-
mechanical properties in the graded plate is incorporated into
the finite element model by specifying each of properties (2)–(7)
at the Gauss points of a finite element via appropriate subroutines
of ABAQUS. The user-defined subroutines UMAT and UMATH have
been used to implement gradients of the mechanical properties
such as the Young’s modulus, Poisson’s ratio and the coefficient
of thermal expansion, and the graded thermal properties like con-
ductivity and specific heat, respectively. While the user subroutine
USDFLD defines the spatial variation of the mass density.
The simulation procedure of the FGM plate is conventionally
divided into thermal analysis, stress analysis and failure analysis
considered in Section 3. In the thermal analysis, the plate is ini-
tially assumed to be stress-free at a processing temperature of
T 0 ¼ 1000 K. Then, it is left in an environment with temperature
of T env ¼ 300 K with free convection at all surfaces, as a result
residual thermal stresses appear. Next the thermal shock as a cycle
of heating and cooling is modelled, Fig. 4, i.e. simulating the appli-
cation of FGM layer to a thermal barrier coating [1]. In this regard
the analysis is divided in two steps. Under heating the ceramic sur-
face of the plate x ¼ 0 is exposed to a high temperature
T H ¼ 1700 K, while metal x ¼ w is kept at a low temperature
T L ¼ T env ¼ 300 K. When the temperature field within the plate
Fig. 4. A scheme of thermal shock conditions.
15
reaches the steady state, the ceramic surface is cooled. There, at
the ceramic surface, forced convection takes place with a convec-
tion coefficient of h ¼ 2000 W=m2 K, all other surfaces are assumed
to be thermally insulated. This step provides the temperature evo-
lution of each element of the model during the transient cooling
treatment.
A sequentially coupled thermal-stress analysis is conducted
with ABAQUS/Standard to model dropping of the temperature
from T 0 to T env and heating the ceramic surface at T H . First, the
steady-state heat transfer analysis is carried out. Then, the known
steady-temperature distributions are used as an input for comput-
ing the thermal stress field, accounting for residual stresses, in the
static analysis. Once the initial distributions of temperature and
associated thermal stresses throughout the domain occupied by
the FGM is known, the transient solution of macromechanical
problems as the stress analysis and fracture can be undertaken. A
fully coupled thermal-stress analysis available in ABAQUS/Stan-
dard is employed to evaluate the distributions of transient temper-
ature and stresses arising from fast cooling of the ceramic surface
up to T L . It should be mentioned that the inertia effects are not
taken into account and coupling of the thermal and structural
problems are assumed to be through the calculated temperature
distribution only. The computed transient stress fields known at
each point of the model as a function of temperature and material
gradient are used as driving forces for the crack growth in the
graded plate. The VCCT implemented into ABAQUS/Standard is
applied to estimate the status of the crack front and to advance
the crack if the crack propagation criterion (19) is met at any point
of the model.
The 2-D plate model is discretized with quadratic plane strain
elements (CPE8T) available in ABAQUS for thermal and displace-
ment analyses. The mesh created for the temperature computa-
tions is also the one used for the mechanical analysis. There are a
total of 2047 two dimensional finite elements with 14730 numbers
of degrees of freedom in the model. The crack is modelled by an
actual small gap between the finite elements located along the
crack surfaces. In the fracture analysis, to simulate the stress singu-
larity in the vicinity of the crack tip and to extract the TSIF, the
degenerated quarter-point quadratic elements are used and these
singular crack-tip elements are surrounded by the standard
eight-node quadratic elements for the rest of the two dimensional
model as shown in Fig. 11 for the sake of demonstration. The radius
of these elements is taken as 0.05a to compute the stress intensity
factors within the desired level of accuracy. Then, to evaluate TSIFs
the J-integral analysis is employed with the ABAQUS software. The
contour integral in terms of the energy release rate and, hence, the
stress intensity factor is calculated by utilizing the domain integral
16 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21
method, [29]. It is important to notice that, in essence, this integral
quantity is path dependent for FGMs. Therefore, only one ring of
elements completely surrounding the crack tip from one crack sur-
face to the opposite crack surface is used for calculating contour
integral in the considered cracked FGM plate.
Including a contact definition into the FE model is of importance
to prevent interpenetration of crack surfaces under compressive
stresses. In this regard, the traction free crack surfaces are sub-
jected to frictionless contact conditions. The interpenetration con-
ditions are realized by using the general contact algorithm with the
’hard’ pressure-overclosure contact model within the small-sliding
tracking algorithm in ABAQUS/Standard [29]. Those contact condi-
tions are maintained behind the crack tip in the crack propagation
analysis also. Another contact model, suitable for the crack propa-
gation analysis within the ABAQUS code can be found in [34,35].
5. Model validation
For verification of the proposed finite element model imple-
menting the user-defined subroutines, we consider the problem
presented in [6], which is one of the few analytic fracture solutions
available for a FGM plate, Fig. 1. Thereby, these results are used as a
basis of comparison in the current study. At the beginning, the
thermal stress problem of the FGM plate in absence of the crack
is considered. The authors in [6] have assumed that the Young’s
modulus, the thermal conductivity and the specific heat are vary-
ing exponentially along the coordinate x (Fig. 1) only as the
following:
EðxÞ ¼ E1 ebx ; bw ¼ lnðE2 =E1 Þ ¼ 0:37498
xx
aðxÞ ¼ a1 e ; xw ¼ lnða2 =a1 Þ ¼ 0:51283 ð20Þ
jðxÞ ¼ j1 egx ; gw ¼ lnðj2 =j1 Þ ¼ 2:5014
where E1 ; a1 ; j1 and E2 ; a2 , j2 are material parameters of the FGM at
x ¼ 0 and x ¼ w, respectively. The Poisson’s ratio is a constant value
of 0.25. Then, if the surface temperatures T 1 ¼ Tðx ¼ 0Þ and
T 2 ¼ Tðx ¼ wÞ are prescribed, the steady state temperature distribu-
tion can easily be calculated analytically by the expression:
Z ,Z ! mic surface is cooled, its surface temperature drops dramatically
x w
dx dx
TðxÞ ¼ T 1 þ ðT 2  T 1 Þ ; ð21Þ
0 jðxÞ 0 jðxÞ
and, then, the distribution of non-zero stress ryy is defined by the
formula:
increasing the metal volume fraction with higher thermal conduc-
EðxÞ
ryy ðxÞ ¼ Ax þ B  aðxÞðTðxÞ  T 0 Þ ; ð22Þ
1m
where the unknown values A and B can be obtained from the
boundary conditions of the plate. For unconstrained boundaries
the following conditions are used:
Z w Z w
ryy ðxÞdx ¼ 0; ryy ðxÞxdx ¼ 0 ð23Þ
0 0
Without a loss of generality we can assume that the tempera-
ture at the stress free state T 0 is 10 °C, and the surface temperature
T 2 is held at T 0 , while the surface temperature T 1 is varied such
that the ratio T 1 =T 0 takes values 5, 10 and 20. The temperature
and thermal stress distributions computed analytically by using
(21)–(23) and numerically by the FE model developed in ABAQUS
are compared in Fig. 5a and b, respectively. In the graph the lines
illustrate the analytical solutions, while the markers show the
finite element simulations. One can see a perfect agreement
between the both results. Fig. 6 shows comparisons of the temper-
ature and thermal stress distributions in the FGM plate in the case,
when the surface temperature T 2 is assumed lower than T 0 and is
kept at 0:5T 0 , and the surface temperature T 1 is defined by the ratio
T 1 =T 0 taking values: 0.5, 0.2, 0.1 and 0.05. This graph also clearly
presents that the finite element solutions are in very good accor-
dance with the analytical ones.
To verify the performance of the finite element model for the
fracture analysis, TSIFs of the two-dimensional model with an
embedded edge crack of relative length a=w ¼ 0:2 under tempera-
ture and stress state shown in Fig. 6 are compared. In [6] the TSIFs
have been calculated by using the integral singular equations,
while the singular finite elements around the crack tip in the FE
model are employed in the FE simulations. The agreement between
those results observed in Fig. 7 is quite good.
6. Numerical results and discussion
Next the finite element model, developed and tested in the pre-
vious sections, is used to gain a deeper insight into the thermal
cracking behavior of functionally graded ceramic–metal plates.
The thermomechanical properties defined by the expressions
(1)–(7) with respective material constants listed in Table 1 are
assigned to the finite element model. The conditions of the previ-
ously described thermal shock loading of the FGM plate with an
pre-existing edge crack a in the ceramic-rich surface is therefore
considered. The sequentially and fully coupled thermal-
displacement analyses are carried out. The results of simulations
of thermal cracking in the FGM plate with various graduation pro-
files (Fig. 2) are presented in Figs. 8–14 below.
In Fig. 8a and b the distributions of transient temperature, nor-
malized with respect to the environmental temperature
T env ¼ 300 K, over the width of the functionally graded plate with
volume fraction parameter 0.5 and 2, respectively, are shown.
The non-dimensional time quantity s ¼ t=t total , where ttotal is the
time of total analysis, is used here. At s ¼ 0 we assume that the
cooling starts in the FGM plate, and before this the plate was sub-
jected to both a temperature gradient, arising from the tempera-
ture difference between the temperatures T 0 and T env , and a
steady state heating. Thereby, a steady state temperature distribu-
tion is illustrated in Fig. 8 at the moment of time s ¼ 0. As the cera-
with time up to T L and the highest temperature value moves from
the surface into the domain of the first quarter of the plate width.
Herewith, one can see that the temperature distributions in the
functionally graded plate with bigger parameter p ¼ 2 is slightly
higher than those with smaller exponent p ¼ 0:5, because of
tivity in the latter graduation profile. For the sake of clearness, con-
tour plots of the distributions of steady state and final transient
temperatures in the FGM plate with p ¼ 0:5 are presented in Fig. 9.
The distributions of transient stress ryy associated with the
respective transient temperature in Fig. 8 are illustrated in
Fig. 10a and b, accordingly. One can see that at the moment of time
s ¼ 0, when the steady state temperature distribution is reached in
the heating phase, the compressive stress concentration near the
crack tip occurs. Herewith, the FGM plate with bigger metal vol-
ume fraction p ¼ 0:5 has a bigger thermally induced steady com-
pressive stress. In the first few milliseconds after the cooling
starts, the thermal stress near the crack tip becomes rapidly tensile
and its value is significantly bigger than the residual compressive
stress developed under heating. The tensile stress sharply
decreases with distance from the crack toward the metal surface.
Also the largest tensile stress near the crack tip quickly decreases
with time of cooling. These results are in compliance with those
presented in [5]. From Fig. 10 it is obviously that the tensile stress
concentration is bigger in the FGM with smaller metal volume frac-
tion p ¼ 2. This is due to the higher tensile strength of ceramic. One
can notice that the temperature and stress distributions associated
 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21
Fig. 5. Comparison of steady state solutions for various T 1 =T 0 with T2 = T0: (a) temperature distribution normalized by T 0 , (b) stress distribution normalized by
r0 ¼ E1 a1 T 0 =ð1  mÞ.
Fig. 6. Comparison of steady state solutions for various T 1 =T 0 with T2 = 0.5T0: (a) temperature distribution normalized by T 0 , (b) stress distribution normalized by
r0 ¼ E1 a1 T 0 =ð1  mÞ.
Fig. 7. Comparison of the normalized TSIF for the cracked FGM strip under stress
state shown in Fig. 6.
with the other metal volume fraction parameters are characterized
by results lying in the middle between those two considered cases
of p. For the demonstration’s sake, contour plots of the distribution
of final transient von Mises stress and the distributions of steady
state and final transient stress ryy at the domain surrounding the
crack tip in the FGM plate with p ¼ 0:5 are presented in Fig. 11.
Analyzing the last two graphs one can see that the rapid change
of the sign of stress and its high concentration at the ceramic sur-
face under cooling are preconditions for initiating and even propa-
gating an edge crack in the FGM plate. The FE studies in [18] has
showed that the simulation of crack growth in a FGM is a challeng-
17
ing task. Due to graded material properties the mode mixity takes
place always, as a result the crack unstably grows along a curve
path that is numerically difficult to be calculated and predicted.
To avoid those complications in the current simulations we assume
that the critical strain energy released rate does not depend on the
volume fraction of the composite components. Its value is defined
by the critical strain energy of the ceramic material GcIC because the
crack is predominantly located in the ceramic-rich zone of the
plate. This assumption will provide predictions that can be inter-
preted as a lower bound of the thermal cracking resistance of the
FGM plate. Moreover, since the reliable sources for temperature
dependence of the studied graded material is not found, we restrict
our simulations with an assumption on temperature independence
of the critical strain energy released rate over the domain of the
FGM plate. Finally, the power law fracture criterion in ABAQUS
(19) was utilized with input arguments such that
GIC ¼ GIIC ¼ GIIIC ¼ 0:19 N=mm2 and k1 ¼ k2 ¼ k3 ¼ 1. The thermal
crack growth in the FGM plate with a pre-existing crack of relative
length a=w of 0.2 and 0.1 is shown in Fig. 12a and b, respectively.
The obtained relative crack lengths versus the non-dimensional
time of cooling are presented in dependence of the composition
profile of FGM. Comparing the results for the lengths of crack
extension, one can conclude that the composition profile with
higher metal volume fraction has better resistance to the thermal
crack growth. It is true for the both cracks with considered initial
lengths. These results are in compliance with those obtained using
the XFEM in [28] and show the clear effect of the compositional
profile on the crack resistance in FGMs.
Moreover, it follows from Fig. 12 that the shorter initial crack
has a longer final crack path. This is due to the higher TSIF induced
18 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21

Fig. 8. Transient temperature distribution under cooling in the FGM plate with volume fraction parameter: (a) p ¼ 0:5, (b) p ¼ 2.

Fig. 9. Contour plots of temperature distribution: (a) steady state distribution, (b) final transient distribution.

Fig. 10. Transient stress ryy distribution under cooling in the FGM plate with volume fraction parameter: (a) p ¼ 0:5, (b) p ¼ 2.
 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21 19

Fig. 11. Contour plots of the stress distribution: (a) final transient von Mises stress, (b) steady state stress ryy at the crack tip, (c) final transient stress ryy at the crack tip.

Fig. 12. Thermal crack growth under thermal shock loading for different composition profiles in the FGM plate with pre-existing crack: (a) a=w ¼ 0:2, (b) a=w ¼ 0:1.

around the crack. The variations of normalized total energy release
rate with respect to the non-dimensional time of analysis for the
both cracked FGM plates are shown in Fig. 13. The bigger values
of the energy release rate in the plate with shorter initial crack is
clearly seen. The normalized critical energy release rate increases
with time and is the highest in the cooling phase when the crack
grows. Herewith, the smaller is the metal volume fraction in the
FGM profile, the greater critical energy release rate is, i.e. the
longer crack length in the FGM is.
To visualize the mechanics of thermal crack growth under cool-
ing, the series of contour plots of the transient thermal stress ryy
distributions at different non-dimensional moment of time is pre-
sented in Fig. 14. A temperature gradient during heating leads to a
compressive stress due to the constraint to the thermal expansion
caused by spatially varying thermo-mechanical properties of FGM
as shown in Fig. 14a. Subsequent cooling of the plate causes the
ceramic surface to experience a tensile stress as seen in Fig. 14b.
Further one can see in Fig. 14c that the tensile stress increases
through the plate thickness that gives rise to a bending moment
opening the crack. As soon as the TSIF induced by this stress
exceeds the critical stress intensity factor value for the given
FGM, the crack starts to grow as illustrated in Fig. 14d. Finally,
the crack is arrested by the compressive stress appearing in
domains of the plate behind the crack tip, Fig. 14e.
20 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21

Fig. 13. Normalized critical energy release rate for different composition profiles in the FGM plate with pre-existing crack: (a) a=w ¼ 0:2, (b) a=w ¼ 0:1.

Fig. 14. Contour plots of transient thermal stress ryy at different moments of non-dimensional time: (a) s ¼ 0, (b) s ¼ 0:01, (c) s ¼ 0:1, (d) s ¼ 0:5, (e) s ¼ 1.

7. Conclusions stress distributions. The variation of the material’s volume fraction

The thermal cracking has been simulated for the functionally
graded plate with a variety of graduation profiles using the finite
element method within the ABAQUS code. The all graded
thermo-mechanical properties of the FGM are programmed at
the element level and are implemented into ABAQUS as user-
defined subroutines. The thermal stresses under the both steady
state and transient-temperature field are treated as a quasi-static
thermoelastic two-dimensional problem for a plane strain state.
In the fracture analysis the singular elements surrounding crack
tip are used to model stress singularity and to extract TSIFs in
the cracked FGM plate. A fully coupled thermal-stress analysis in
conjunction with the VCCT is used to study the thermal crack
growth in dependence of the graduation profile of FGM.
The obtained results showed that the developed FE model
enables to make accurate and reliable predictions of FGMs’
mechanical behavior. The comparisons between the known analyt-
ical solutions and finite element data gave a very good agreement
in the thermal, stress and fracture analyses. Moreover, the calcula-
tions clearly demonstrated the great influence of spatially varying
material properties on the transient temperature and thermal
is also the most important factor in designing the functionally
graded plates with high crack resistance. The higher is content of
metal in the graded ceramic/metal composite, the higher ability
of the FGM to resist a thermal crack. Fracture toughness also varies
with the volume fraction parameter, but further investigations are
required to clarify the influence of this specific factor.
Acknowledgements
The first author acknowledgements the German Science Foun-
dation DFG (AL 341/46-1) for supporting his research in Otto-
von-Guericke University Magdeburg. The third author would like
to mention that this work was also financially supported by Min-
istry of Science and Higher Education of Poland within the statu-
tory research Grant No. S/20/2015.
References
[1] M. Koizumi, Functionally Gradient Mater. 34 (1993) 3–10.
[2] Y. Miyamoto, W.A. Kaysser, B.H. Rabin, A. Kawasaki, R.G. Ford, Functionally
Graded Materials Design, Processing and Applications, Springer Science
+Business Media, New York, 1999.
 V.N. Burlayenko et al. / Computational Materials Science 116 (2016) 11–21 21

[3] T. Sadowski, Non-symmetric thermal shock in ceramic matrix composite [18] T. Fujimoto, N. Noda, J. Am. Ceram. Soc. 84 (7) (2001) 1480–1486.
(CMC) materials, in: R. de Borst, T. Sadowski (Eds.), Lecture Notes on [19] M.H. Santare, J. Lambros, J. Appl. Mech. 67 (4) (2000) 819–822.
Composite Materials – Current Topics and Achievements, Solid Mechanics [20] J.H. Kim, G.H. Paulino, J. Appl. Mech. 69 (2002) 502–514.
and its Applications, vol. 154, Springer Science+Business Media B.V., 2009, pp. [21] W.G. Buttlar, G.H. Paulino, S.H. Song, J. Eng. Mech. 132 (3) (2006) 240–248.
99–148. [22] P. Gu, M. Dao, R.J. Asaro, J. Appl. Mech. 66 (1999) 101–108.
[4] J.W. Eischen, Int. J. Fract. 34 (3) (1987) 3–22. [23] G. Anlas, N.H. Santare, J. Lambros, Int. J. Fract. 104 (2000) 131–143.
[5] Z.H. Jin, N. Noda, J. Appl. Mech. 61 (1994) 738–740. [24] M.T. Tilbrook, R.J. Moon, M. Hoffman, Eng. Fract. Mech. 72 (2005) 2444–2467.
[6] F. Erdogan, B.H. Wu, J. Therm. Stresses 19 (3) (1996) 237–265. [25] Z.H. Jin, G.H. Paulino, R.H. Dodds, Eng. Fract. Mech. 70 (14) (2003) 1885–1912.
[7] Z.H. Jin, N. Noda, Int. J. Solids Struct. 31 (2) (1994) 203–218. [26] S. Rangaraj, K. Kokini, J. Eng. Mater. Technol. 126 (2004) 103–115.
[8] N. Noda, J. Therm. Stresses 20 (3–4) (1997) 373–387. [27] C. Comi, S. Mariani, Comput. Methods Appl. Mech. Eng. 196 (2007) 4013–4026.
[9] Z.H. Jin, G.H. Paulino, Int. J. Fract. 107 (2001) 73–98. [28] I.V. Ivanov, T. Sadowski, D. Pietras, Eur. Phys. J. Special Top. 222 (7) (2013)
[10] T. Sadowski, A. Neubrand, Thermal shock crack propagation in functionally 1587–1595.
graded strip, in: A.M. Brandt, V. Li, I.H. Marshal (Eds.), Brittle Matrix [29] ABAQUS 6.11 User’s Manuals, Dassault Systèmes Simulia Corp., Providence, RI,
Composites, vol. 7, Zturek RSI and Woodhead Publ., Warsaw-Cambridge, USA, 2011.
2003, pp. 81–90. [30] E.F. Rybicki, M.F. Kanninen, Eng. Fract. Mech. 9 (1977) 931–938.
[11] J. Sladek, V. Sladek, Ch. Zhang, Comput. Mater. Sci. 32 (2005) 532–543. [31] R. Krueger, P.J. Minguet, Compos. Struct. 81 (2007) 41–59.
[12] B. Yildirim, Ö. Kutlu, S. Kadioğlu, Int. J. Solids Struct. 48 (2011) 3020–3031. [32] V.N. Burlayenko, T. Sadowski, Budownictwo i Architektura 2 (1) (2008) 95–
[13] V. Petrova, S. Schmauder, Comput. Mater. Sci. 52 (1) (2012) 171–177. 109.
[14] V. Petrova, S. Schmauder, Comput. Mater. Sci. 64 (2012) 229–233. [33] R. Krueger, Appl. Mech. Rev. 57 (2) (2004) 109–143.
[15] V. Petrova, T. Sadowski, Meccanica 49 (2014) 2603–2615. [34] V.N. Burlayenko, T. Sadowski, Modeling the dynamic debonding growth of
[16] H. Altenbach, V.A. Eremeyev, Mechanics of viscoelastic plates made of FGMs, sandwich plate, in: Yu.V. Mikhlin, M.V. Perepelkin (Eds.), Proceedings of the
in: J. Murin, V. Kompis, V. Kutis (Eds.), Computational Modelling and 4th International Conference ‘Nonlinear Dynamics’, June, 19–22, 2013,
Advanced Simulations, Computational Methods in Applied Sciences, vol. 24, Sevastopol, Ukraine, Tochka Publ., Kharkiv, 2013, pp. 225–230.
Springer, Heidelberg, 2011, pp. 33–48. [35] V.N. Burlayenko, T. Sadowski, J. Appl. Nonlinear Dynam. 3 (4) (2014) 369–379.
[17] M. Birsan, H. Altenbach, T. Sadowski, V.A. Eremeyev, D. Pietras, Compos. Part B:
Eng. 43 (2012) 1315–1328.