Original Article

Journal of Intelligent Material Systems

and Structures
2018, Vol. 29(5) 845–862
Micromechanics determination of Ó The Author(s) 2017
Reprints and permissions:
effective material coefficients of sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1045389X17721047

cement-based piezoelectric ceramic journals.sagepub.com/home/jim

composites

Albert Artha Saputra1, Vladimir Sladek2, Jan Sladek2

and Chongmin Song1

Abstract
Determination of effective composite properties for a real distribution of piezoelectric particles in cement matrix is con-
sidered using a representative volume element. The scaled boundary finite element method is proposed to analyse
cement-based piezoelectric ceramic composites. Quadtree hierarchical meshing structure is utilized to provide a mesh
that captures the material variation in the representative volume element effectively based on the colour of the piezo-
electric composite’s image. Analyses with carefully selected boundary conditions are performed to obtain the effective
material properties of various samples.

Keywords
Scaled boundary finite element method, representative volume element, piezoelectric ceramic, effective material proper-
ties, image-based analysis

Introduction catastrophic failure of structures, decrease maintenance

Monitoring of large civil structures has been widely
studied in recent years (Aizawa et al., 1998; Huston
et al., 1994; Pines and Lovell, 1998). High-performance
structures, such as high-rise buildings and long-span
bridges, as well as some buildings in which failure
would lead to disasters, such as nuclear waste contain-
ment structures, dams and bridge decks, are the most
probable candidates for the application of structure
monitoring. Monitoring is utilized to detect damage in
structures and analyse possibility of a collapse due to
natural disasters such as hurricanes, typhoons,
tornadoes and earthquakes. Currently, inspection
approaches are either passive or occur after a cata-
strophic structural failure. Embedded smart structure
technology (actuators and sensors) offers the ability to
assess damage and react to internal and external
changes in order to maintain their design functions.
Structures with the ability to sense and assess damage
are referred to as intelligent structures. The materials
that perform the sensing and reacting (actuating) tasks
are called smart materials. The wireless communication
technology permits this structural health monitoring
(SHM) to be performed remotely (Pines and Lovell,
1998). The SHM system is utilized to prevent
cost and guide construction. Usually, piezoelectric
materials play a significant role in the SHM since they
can be used as both actuators and sensors due to their
unique electromechanical coupling behaviour (Bhalla
and Kiong Soh, 2003; Giurgiutiu and Zagrai, 2002;
Lynch, 2007; Mascarenas et al., 2007; Song et al.,
2006). On the other hand, reinforced concrete made
from cement-based materials is the most popular mate-
rials used for larger civil engineering structures. Due to
a high impedance of pure piezoelectric materials, the
piezoelectric patches cannot be bonded on the surface
of the host concrete structure. To overcome such
incompatibility between concrete and piezoelectric
materials, one approach is to develop the cement-based
composites with piezoelectric properties. It is required
that the actuator could generate a strong power signal,
1
School of Civil and Environmental Engineering, The University of New
South Wales, Sydney, NSW, Australia
2
Institute of Construction and Architecture, Slovak Academy of Sciences,
Bratislava, Slovakia
Corresponding author:
Albert Artha Saputra, School of Civil and Environmental Engineering, The
University of New South Wales, Sydney, NSW 2052, Australia.
Email: a.saputra@unsw.edu.au
846
and the sensors have high sensitivity and specific fre-
quency response.
Normal mixing and spread techniques were applied
to produce the first 0–3 cement-based piezoelectric
smart composite, where the piezoelectric ceramic and
cement phases were of zero and three dimensions,
respectively (Li et al., 2002). The experimental results
showed that this new composite has a good compatibil-
ity with Portland cement concrete. If both lead zirco-
nate titanate (PZT) ceramic and cement phases are two
dimensions, they are referred as 2-2 composites. They
were fabricated and explored by Li et al. (2001). Such
composite exhibits a desirable direct piezoelectric effect
and inverse piezoelectric effect, so it can be used both
as sensor and actuator in intelligent structures. Since
the piezoelectric properties are determined by poling
conditions, the behaviour of cement-based piezoelectric
composites under various poling conditions was inves-
tigated (Chaipanich, 2007a, 2007b; Dong and Li, 2005;
Li et al., 2002, 2005; Shifeng et al., 2004; Wang et al.,
2012). The results revealed that the piezoelectric prop-
erties were enhanced with increasing volume fraction of
the PZT piezoelectric ceramic particles, poling voltage,
time and temperature. Recently, a novel cement–sand-
based piezoelectric smart composite has been developed
for conducting SHM for civil structures. A new compo-
site based on sand is also recently developed for the
first time by Zhao et al. (2016) to overcome the incom-
patibility between piezoelectric materials and reinforced
concrete containing cement and sand.
Works on the effect of particle size on the properties
of the PZT–cement-based composites are still very lim-
ited in the literature. Chaipanich (2007b) investigated
the effect of PZT particle size on the properties of PZT–
cement-based composites experimentally. The results
showed that the dielectric properties of the composites
increased marginally with PZT particle size. A notice-
able increase in d33 piezoelectric coefficient values was
also found when the particle size was increased. The
enhancement in the dielectric and piezoelectric proper-
ties was explained by lesser contacting surfaces between
the cement matrix and the PZT particles.
Other than experimentation, the effective material
properties for composites can also be obtained using
analytical approaches. Some examples include the use
of Rayleigh’s formalism (Kuo and Pan, 2011) or the
complex variable method along with the Mori-Tanaka
mean-field approximation (Wang and Pan, 2007) for
the determination of magnetoelectric effects of multi-
ferroic composite containing circular fibres. Another
approach involves the use of Green’s function and
interfacial operators for obtaining the effective proper-
ties of magnetoelectroelastic composite with coated
inclusions (Dinzart and Sabar, 2012).
Alternatively, in this article, the goal is to provide an
effective computational tool to evaluate the material
properties for the PZT–cement-based composites using
Journal of Intelligent Material Systems and Structures 29(5)
numerical analyses which are performed on the repre-
sentative volume element (RVE). The RVE contains
sufficient micro-structural information of the PZT par-
ticles in the cement matrix to be representative of any
similar volume taken from any location in the compo-
site. The effective material properties of the composite
can then be evaluated by considering special boundary
conditions for the RVE such as the one developed in
Sladek et al. (2017a). The solution of general boundary
value problems for coupled multi-field problems
requires advanced numerical methods due to the high
mathematical complexity. Such a multi-field problem is
described by a system of partial differential equations
because of the interactions among the electric and
mechanical fields involved in the constitutive equations.
The scaled boundary finite element method
(SBFEM) is adopted in this article. In the context of
this work, the SBFEM can be regarded as a technique
that generalizes the shape of finite elements from trian-
gles and quadrilaterals to arbitrary sided polygons.
This method is compatible and can be straightfor-
wardly incorporated into a finite element computer
program. It has been applied to analyse cracks in brittle
porous piezoelectric solids (Sladek et al., 2016). This
SBFEM has also been successfully applied to elasto-
statics (Deeks and Wolf, 2002; Song, 2004), elastody-
namic problems (Chiong et al., 2014) and fracture
mechanics (Bird et al., 2010; Li et al., 2013; Zhu et al.,
2013) to name a few. In the present article, SBFEM is
developed for two-dimensional (2D) boundary value
problem in a general piezoelectric solid under station-
ary boundary conditions. The SBFEM is applied here
on the micro-level (RVE) of the piezoelectric ceramic
particles in the Portland cement matrix to compute the
effective material properties using special boundary
conditions similar to the one provided in Sladek et al.
(2017a). The SBFEM has also been developed for
three-dimensional (3D) stress analysis (Saputra et al.,
2015, 2017).
In this article, the advantages of SBFEM with quad-
tree meshes for 2D image-based analysis (Saputra et al.,
2017) will be exploited for fast computation of effective
material properties. The quadtree structure can model
geometrical features of different scales efficiently and
automatically using a recursive approach. Each quad-
tree cell can be modelled as a scaled boundary polygo-
nal element which means there is no hanging node issue
commonly encountered when employing finite element
method (FEM) with quadtree meshes. Only the bound-
ary of the cells is discretized. Additionally, the element
solutions for cells of the same nodal pattern but having
different sizes are proportional. Hence, solutions
for the limited cell patterns generated can be precom-
puted and quickly extracted when required. This
increases the efficiency of the proposed approach for
the elastic analysis conducted in this article.
Saputra et al. 847

The present analyses give information concerning governing equations for stationary problems which are
how the homogenized elastic properties of the material expressed by the equilibrium of forces and the Maxwell
vary in terms of the volume fraction of piezoelectric equations are reduced to the Gauss law (Parton and
PZT ceramic particles. The influence of the ceramic par- Kudryavtsev, 1988)
ticles’s distribution and shape will also be investigated.
sij, j ðxÞ = 0 ð7Þ

Governing equations for piezoelectric Dj, j ðxÞ = 0 ð8Þ

materials where the body force vector and the volume density of
The coupling of the mechanical and electrical fields in free charges are assumed to be vanishing.
piezoelectric solids is expressed via constitutive equations
expressed as follows (Parton and Kudryavtsev, 1988)
Image-based quadtree meshing algorithm
sij ðx, t Þ = cijkl ekl ðx, t Þ  eijk Ek ðx, t Þ ð1Þ This study involves the use of images which show the
Dj ðx, t Þ = ejkl ekl ðx, t Þ + hjk Ek ðx, t Þ ð2Þ distribution of the material in the composite. In order
to capture the material distribution accurately and effi-
where the elastic stresses sij and electric displacements ciently, a hierarchical quadtree structure is utilized. For
Di are fields conjugated to the elastic strains eij and elec- more details on the quadtree algorithm, readers are
tric field vector Ei , respectively. The material coeffi- referred to Saputra et al. (2017).
cients cijkl , hjk and eijk are the elastic coefficients, The quadtree recursive meshing approach starts with
dielectric permittivities and piezoelectric coefficients, a square cell which covers the entire image. At each
respectively. The secondary fields feij , Ei g are given by meshing step, the colour of the pixels contained in a cell
gradients of the primary fields (the displacement vector is tested against a criterion of homogeneity. If the range
ui and the electric potential f) as of colours in the cell exceeds a user-defined threshold,
the cell is divided into 4 identical square cells, otherwise
1 
the division process is terminated as it is considered as
eij ðx, t Þ = ui, j ðx, t Þ + uj, i ðx, t Þ ð3Þ
2 homogeneous. As an example, consider a sample image
Ei ðx, t Þ =  f, i ðx, t Þ ð4Þ in Figure 1(a) where grey represents the piezoelectric
ceramic particles and black represents the mortar. The
Note that the scalar potentials are sufficient in case quadtree decomposition of this image is shown in
of quasi-static approximation, when the external load- Figure 1(b) which shows blue cells representing the
ing is remarkably slow as compared with the character- piezoelectric particles and purple cells representing the
istic frequency of electric processes. mortar. Typically, smaller cells are found in between
If one is interested in transversely isotropic piezo- the material interfaces as depicted in Figure 1(b). This
electric material with poling direction along the positive meshing procedure is automatic and can efficiently cap-
x3-axis, the x1 x2 plane is isotropic. Therefore, either ture features with different scales.
the x1 x3 or x2 x3 plane can be employed for the study Each quadtree cell is modelled as a scaled boundary
of plane electromechanical phenomena. Considering polygonal element which will be discussed in the next
the x1 x3 plane, the plane strain conditions require section. Since elastic analysis is considered, the element
e22 = e12 = e23 = E2 = 0. Then, equations (1) and (2) solutions for cells with the same nodal configurations
are reduced to the matrix form as follows (Parton and but having different sizes are proportional. There are
Kudryavtsev, 1988) only 16 cell patterns in each material that must be com-
2 3 2 32 3 2 3 puted for the balanced quadtree mesh utilized here. The
s11 c11 c13 0 e11 0 e31  
4 s33 5 = 4 c13 E element solutions for the rest of the cells can be simply
c33 0 54 e33 5  4 0 e33 5 1
E3 scaled from these 16 unique cells; hence, these unique
s13 0 0 c44 2e13 e15 0
cells are commonly referred to as master cells.
ð5Þ
2 3
    e   
D1 0 0 e15 4 11 5 h 0 E1 Summary of SBFEM
= e33  11
D3 e31 e33 0 0 h33 E3 The heterogeneities on the micro-structural length scale
2e13
ð6Þ can be accounted in macro-structural analyses via
homogenization when the effective (overall) material
For analyses of the influence of micro-structural coefficients are obtained from solutions of appropriate
parameters on the effective global material behaviour, boundary value problems on mesoscale level in a RVE.
it is sufficient to consider stationary boundary condi- Note that both the macroscale and mesoscale analyses
tions. Then, all fields are independent of time and the are performed within the theory of continua in contrast
848 Journal of Intelligent Material Systems and Structures 29(5)

(a) (b)
Figure 1. Quadtree decomposition for a piezoelectric ceramic composite image: (a) a sample image and (b) quadtree mesh for the
sample image.

to the multiscale modelling when different physical

models as well as different computational methods are
employed on different length scales.
In the present article, the SBFEM is developed for
electroelastic materials. Making use the standard Voigt
notation, equations (5) and (6) can be written in a com-
pact form as
8 9 2 38 9
>
> s11 >
> c11 c13 e31 > > e11 >
>
>
> > 6 > >
< s33 >
= 6 c13 c33 e33 7 >
7< e33 =
>
 = s13 = 6
s 6 c44 e15 7 2e13 = Ce
7>
>
> >
> 4 5> >
>
>
> D1 >> e15 h11 > E1 >
> >
: ; : ;
D3 e31 e33 h33 E3
ð9Þ
Figure 2. A scaled boundary polygonal element.
where s  denotes the generalized stress vector contain-
ing the elastic stress components sij and the electric dis-
placements Di . The e denotes the generalized strains where L is the linear differential operator and u denotes
containing the elastic strains eij and the electric field the generalized displacement which consists of the dis-
vector Ei . The symmetric C matrix contains the mate- placement components ui and the electric potential f.
rial coefficients. The governing equations in the domain without consid-
Using equations (3) and (4), the generalized strains e ering body forces and electric charges are expressed
can be expressed as using equations (9) and (10) as
2 ∂ 3
LT s
 =0 ð11Þ
6 ∂x1 7
6 7
6 ∂ 7
6 7
6 ∂x3 78 9
6 7< u1 =
6 ∂ ∂ 7 Scaled boundary transformation of geometry
e = 6
6 ∂x3
7 u3 = L
7: ; u ð10Þ
6 ∂x1 7 f In the SBFEM, only the boundary of a polygonal ele-
6 ∂ 7
6 7 ment is discretized with 1D (line) elements Se , as shown
6 ∂x1 7
6 7 in Figure 2. A scaling centre O is selected at a point
4 ∂ 5 from which the whole boundary is directly visible.
∂x3 Based on this scaling centre, local scaled boundary
Saputra et al. 849

 
coordinates are introduced namely a radial coordinate 1 x3 ðhÞ, h x3 ðhÞ
½JðhÞ1 = ð18Þ
j and a circumferential coordinate h. The radial coor- jJðhÞj x1 ðhÞ, h x1 ðhÞ
dinate j takes the value of 0 at the scaling centre and 1
at the boundary S. By scaling the boundary S along j with the determinant of the Jacobian given as
with a scaling factor 0  j  1, the whole polygonal ele-
jJðhÞj = x1 ðhÞx3 ðhÞ, h  x3 ðhÞx1 ðhÞ, h ð19Þ
ment is covered.
Each line element Se is interpolated based on the The linear differential operator L in equation (10) is
shape functions N(h) defined for 1  h  1. Any point transformed to the scaled boundary coordinates using
on the boundary (x1 , x3 ) can be described as equation (17) as
x1 ðhÞ = NðhÞx1 ð12aÞ ∂ 1 ∂
L = b 1 ð hÞ + b 2 ð hÞ ð20Þ
x3 ðhÞ = NðhÞx3 ð12bÞ ∂j j ∂h

where x1 and x3 are the vectors containing the nodal with

coordinates for each line element Se . Any point within 2 3
the polygonal element (^x1 , ^x3 ) can be expressed using x3 ðhÞ, h 0 0
6 0 x1 ðhÞ, h 0 7
equation (12) as 1 6 6 x1 ðhÞ, h
7
7
b 1 ð hÞ = x3 ðhÞ, h 0
jJðhÞj 6
4 0 0 x3 ðhÞ, h 5
7
^x1 ðj, hÞ = jx1 ðhÞ = jNðhÞx1 ð13aÞ
0 0 x1 ðhÞ, h
^x3 ðj, hÞ = jx3 ðhÞ = jNðhÞx3 ð13bÞ
ð21aÞ
Subsequently, the differential operator is trans- 2 3
x3 ðhÞ 0 0
formed from the Cartesian coordinate to the scaled 6 0 x 1 ð hÞ 0 7
boundary coordinates using the following relationship 1 6 6 x1 ðhÞ
7
b 2 ð hÞ = x3 ðhÞ 0 7 ð21bÞ
8 9 8 9 jJðhÞj 6
4 0
7
∂ > ∂ > 0 x3 ðhÞ 5
>
> > >
> >
< = < = 0 0 x1 ðhÞ
∂j ^ ∂x1
= Jðj, hÞ ð14Þ
>
> ∂ > > ∂ >
: > ; >
: >
;
∂h ∂x3 Generalized displacements, strains and stresses in
with the Jacobian matrix defined as radial direction
  Generalized displacement functions u(j) are intro-
^ðj; hÞ ¼ ^x1 ðj; hÞ; j ^x3 ðj; hÞ; j
J duced along a radial line (line passing the scaling cen-
^x ðj; hÞ; h ^x3 ðj; hÞ; h tre and a node of a line element). The generalized
 1  ð15Þ
x 1 ð hÞ x3 ðhÞ displacement at any point (j, h) within each sector Ve
¼
jx1 ðhÞ; h jx3 ðhÞ; h (an area enclosed by scaling a line element Se along
the radial direction) as shown in Figure 2 is interpo-
The Jacobian matrix in equation (15) can be written lated by
in terms of the Jacobian matrix on the boundary J(h),
that is uðj, hÞ = Nu ðhÞuðjÞ = ½N1 ðhÞI, . . . , Nn ðhÞIuðjÞ ð22Þ
    
^ðj, hÞ = 1 0 x 1 ð hÞ x 3 ð hÞ 1 0 where Ni is the shape function of the ith node, I is a
J = JðhÞ
0 j x1 ðhÞ, h x3 ðhÞ, h 0 j 3 3 3 identity matrix and n is the total number of nodes
ð16Þ on the line element.
By substituting equations (20) and (22) into equa-
By rearranging equation (14) and subbing equation tion (10), the following generalized strain expression is
(16) into it, the following relationship can be obtained obtained
8 9 8 9
> ∂ > > ∂ > 1
>
< >
= h < >
i1 > = eðj, hÞ = B1 ðhÞuðjÞ, j + B2 ðhÞuðjÞ ð23Þ
∂x1 ^ðj, hÞ ∂j j
= J
>
> ∂ >> >
> ∂ >
: ; : > ; with
∂x3 ∂h8 9
2 3> ∂ > ð17Þ
1 0 < => > B1 ðhÞ = b1 ðhÞNu ðhÞ ð24aÞ
= ½JðhÞ1 4 1 5 ∂j
0 > ∂ > B2 ðhÞ = b2 ðhÞNu ðhÞ, h ð24bÞ
j >: > ;
∂h
The generalized stresses can be obtained by substi-
where ½J(h)1 is obtained from equation (16) as tuting equation (23) into equation (9) to yield
850 Journal of Intelligent Material Systems and Structures 29(5)

 
1 A block diagonal Schur decomposition is utilized to
 ðj, hÞ = C B1 ðhÞ
s uðjÞ, j + B2 ðhÞ
uð j Þ ð25Þ
j decompose Z in equation (30) to obtain

ZC = CS ð31Þ
Scaled boundary finite element equation in
where the matrix S is the real Schur form partitioned
generalized displacement
into 2N  1 block diagonal matrices which can be
Based on the weighted residual method (Song and expressed as
Wolf, 1997) or virtual work principle (Deeks and Wolf,   
2002), the scaled boundary finite element equation in 0 I
S = diag S1 , . . . , SN1 , , SN + 2 , . . . , S2N
generalized displacement can be obtained as 0 0
  ð32Þ
uðjÞ, jj + E0  E1 + ET1 j
E0 j2  uðjÞ, j  E2 
uð j Þ = 0
ð26Þ where I is a 3 3 3 identity matrix which is associated
with the three modes of translational rigid body
where the coefficient matrices E0 , E1 and E2 are defined motions for an in-plane problem. The matrix C in
as equation (31) is the transformation matrix with inde-
pendent vectors partitioned into 2 3 2N blocks of
ð
+1
matrices which can also be grouped into 2 3 2 equal
E0 = BT1 ðhÞCB1 ðhÞjJðhÞjdh ð27aÞ blocks of larger matrices such that
1  
Cu1    CuN CuN +1    Cu2N
ð C= q
+1
 C1u    CqN CqN +1    Cq2N
BT2 ðhÞCB1 ðhÞjJðhÞjdh ð33Þ
E1 = ð27bÞ Cn Cup
=
1 Cqn Cqp
ð
+1
The general solutions for the generalized displace-
E2 = BT2 ðhÞCB2 ðhÞjJðhÞjdh ð27cÞ ments u(j) and generalized internal nodal forces q(j) in
1 equation (29) can be expressed as

The coefficient matrices are obtained for each line X

N1
element individually. They are later assembled for each uðjÞ = Cui jSi ci + CuN cN
polygonal element similar to the assembly of global i¼1

stiffness matrix in the FEM.   X

2N

The generalized internal nodal forces associated with + CuN lnj+CuN +1 cN +1 + Cui jSi ci ð34aÞ
i=N +2
the generalized displacements are denoted as q(j). On a
surface with a constant j, the generalized internal nodal X
N 1 X
2N
qðjÞ = Cq1 jSi ci + CqN +1 cN +1 + Cq1 jSi ci
forces equal to i=1 i=N +2
ð34bÞ
 uðjÞ, j + ET1 
qðjÞ = E0 j uð j Þ ð28Þ
where ci are the integration constants. Note that CqN = 0
as demonstrated in Song (2004). For the bounded
Solution procedure domain, the condition of finiteness of the strain energy at
In order to solve the scaled boundary finite element the scaling centre leads to cN +1 = 0 as ln j tends to nega-
equation, a matrix function solution procedure is per- tive infinity and ci = 0 for N + 2  i  2N as the jSi
formed as explained in Song (2004). Equations (26) and tends to infinity for Si with positive real parts. Hence, the
(28) are first transformed into a system of first-order general solutions reduces to
ordinary differential equations yielding
X
N 1
    uðjÞ = Cui jSi ci + CuN cN = Cun jSn cn ð35aÞ
uðjÞ uð j Þ

j = Z ð29Þ i=1
qðjÞ ,j

q ðj Þ
X
N 1

with Z being the Hamiltonian coefficient matrix defined qðjÞ = Cqi jSi ci = Cqn jSn cn ð35bÞ
i=1
as
  where Sn = diag(S1 , . . . , SN 1 , 0) and cn is the vector
E1
0 E1
T
E1
0 containing all of the integration constants that are
Z= ð30Þ
E2 + E1 E1 T
0 E1 E1 E1
0 related to the bounded domain, that is, ci for 1  i  N .
Saputra et al. 851

The integration constants are determined after

obtaining the displacement solution at the boundary

u(j = 1) based on the rearrangement of equation (35a)
as follows

1
cn = Cun uð j = 1Þ ð36Þ

The generalized nodal forces on the boundary q(j = 1)

can be expressed using equations (35b) and (36) as

1
qðj = 1Þ = Cqn Cun 
 uð j = 1Þ ð37Þ

Since 
q(j = 1) must be equal to the applied general-
ized nodal forces on the boundary of a polygonal ele-
ment, an expression of the stiffness matrix K for each
element is obtained directly from equation (37) as
Figure 3. Boundary conditions appropriate for evaluation of

1
K = Cqn Cun ð38Þ ceff eff eff
11 , c13 and e31 .

After obtaining the stiffness matrix for each quad-
tree cell (polygonal element) using equation (38), the
rest of the solution is similar to the FEM. This includes
the assembly for the global stiffness matrix for the
whole domain which is obtained through the assembly
of each quadtree cell’s stiffness matrix, the generalized
nodal forces computation and the solution of the global
linear equations. Hence, an existing code and user inter-
displacements. Since the quadtree cells are typically
small enough around the boundary of the material
phases where the stress change is higher, the expression
for the average stresses and electric displacements
shown in equation (39) can be simplified by only sam-
pling the values at the scaling centres of the quadtree
cells. These can be written as follows
face for the FEM or extended finite element method 
1 X N
(XFEM) can be easily modified or coupled to include sij = s k Ak ð40aÞ
ab k = 1 ijðscÞ
the use of scaled boundary polygonal element such as
the one performed in Natarajan and Song (2013). 
1 X N
Dj = D k Ak ð40bÞ
ab k = 1 jðscÞ
Computation of effective material
properties where Ak is the area of cell k, skijðscÞ and DkjðscÞ are the
stress and electric displacement at the scaling centre of
According to the constitutive relationships as shown in cell k, respectively.
equation (9), we can calculate the effective material coef-
ficients of piezoelectric ceramic particles in Portland
cement matrix. These effective material properties are Boundary conditions set 1
computed using the known average values of the sec- If the boundary conditions are selected as shown in
ondary fields and average values of conjugated fields Figure 3, the average values of the secondary fields are
obtained from numerical solutions of a few properly given as
selected boundary value problems on a RVE sample.
Let us consider a rectangular RVE sample O = he11 i = e11 = constant ð41aÞ
f8x = (^x1 , ^x3 ); ^x1 2 ½0, a, ^x3 2 ½0, bg. The average stress
he33 i = 0 ð41bÞ
and electric displacement are given by
ð he13 i = 0 ð41cÞ

1
sij = sij dO ð39aÞ hE i i = 0 ð41dÞ
ab
O
ð Then, we can get the following effective material

1 coefficients
Dj = Dj dO ð39bÞ
ab
O hs11 i
ceff
11 = ð42aÞ
where the integrands are obtained from the solution of e11
considered boundary value problems. hs33 i
In this article, we shall use another approach for the ceff
13 = ð42bÞ
e11
evaluation of average values of stresses and electric
852 Journal of Intelligent Material Systems and Structures 29(5)

Figure 4. Boundary conditions appropriate for evaluation of Figure 5. Boundary conditions appropriate for evaluation of
ceff eff eff
33 , c13 and e33 . ceff eff
44 and e15 .

hD3 i Boundary conditions set 3

eeff
31 = ð42cÞ
e11
If the boundary conditions are selected as shown in
where the average values of the conjugated fields hs11 i, Figure 5, the average values of the secondary fields are
hs33 i and hD3 i are given by equation (40), with the val- given as
ues at the scaling centres being obtained from the solu-
he11 i = 0 ð45aÞ
tion of the considered boundary value problem.
he33 i = 0 ð45bÞ
he13 i = e13 = constant ð45cÞ
Boundary conditions set 2
h Ei i = 0 ð45dÞ
If the boundary conditions are selected as shown in
Figure 4, the average values of the secondary fields are Now, we can calculate the following effective mate-
given as rial coefficients

he11 i = 0 ð43aÞ hs13 i

ceff
44 = ð46aÞ
2e13
he33 i = e33 = constant ð43bÞ
hD1 i
he13 i = 0 ð43cÞ eeff
15 = ð46bÞ
2e13
h Ei i = 0 ð43dÞ where the average values of the conjugated fields hs13 i
and hD1 i are given by equation (40), with the values at
Then, we can get the following effective material the scaling centres being obtained from the solution of
coefficients the considered boundary value problem.
hs33 i After solving the above considered boundary value
ceff
33 = ð44aÞ
e33 problems on the RVE sample with multiple material
phases, all the effective material coefficients can be com-
hs11 i puted. Some of them can be calculated duplicitously.
ceff
13 = ð44bÞ
e33
hD3 i Numerical examples
eeff
33 = ð44cÞ
e33
The material coefficients for PZT-5H are given as
where the average values of the conjugated fields
hs11 i, hs33 i and hD3 i are given by equation (40), with c11 = 12:6 3 1010 N m2 , c13 = 5:3 3 1010 N m2
the values at the scaling centres being obtained from c33 = 11:7 3 1010 N m2 , c44 = 3:53 3 1010 N m2
the solution of the considered boundary value
problem. e15 = 17 C m2 , e31 =  6:5 C m2
Saputra et al.
Figure 6. Geometry for the square PZT material embedded in
the middle of a square cement matrix.
D piezoelectric phases (Sladek et al., 2012). In the follow-
A B
C pixel. This results in a total of 3496 polygonal ele-
Figure 7. The quadtree mesh for the square PZT material
embedded in the middle of a square cement matrix.
e33 = 23:3 C m2 , h11 = 15:1 3 109 CðV mÞ1
h33 = 13 3 109 CðV mÞ1
Portland cement matrix with isotropic material
properties
E = 1:4 3 1010 N m2 , n = 0:2
h11 = 15:1 3 106 CðV mÞ1 , h33 = 13 3 106 CðV mÞ1
Note that the electric permittivity in the Portland
cement matrix is significantly larger than the value in
853
-10
10
6
FEM
SBFEM
4
-2
-4
-6
0 0.2 0.4 0.6 0.8 1
Figure 8. Displacement u1 results (m) along line AB for the
square PZT material embedded in the middle of a square
cement matrix.
the piezoelectric PZT ceramic as the gradient of the
electric potential inside the cement matrix is signifi-
cantly smaller than in the piezoelectric phase. A similar
approach is utilized for layered piezomagnetic and
ing sections, all the superscript ‘eff’ indicating effective
material property computed based on the boundary
value problems will be omitted for conciseness.
Verification example
In order to verify the accuracy of the present method, a
simple example is conducted by considering a square
PZT-5H material embedded in the middle of a square
Portland cement matrix, as shown in Figure 6.
The dimension of the problem considered is a/
L = 150/256. An image with resolution of 256 pixels
by 256 pixels is considered. Each pixel represents
0.1 mm by 0.1 mm. The maximum cell size used is
8 3 8 pixels and the minimum cell size is set to be 1
ments with 4089 nodes. The quadtree mesh is shown
in Figure 7.
In order to verify the result, a finite element analysis
(FEA) is performed in the commercial software
ANSYS using linear element (PLANE13). A uniform
fine mesh with each element having the size of 0.1 mm
by 0.1 mm is utilized. There are a total of 65,536 ele-
ments with 66,049 nodes. One can see a significantly
higher number of elements in the FEA than in the
quadtree mesh used in the SBFEM.
The analysis is conducted for boundary conditions
set 2 with e33 = 1 3 106 . The results from both the
present method and the FEA for u1 along line AB
(shown in Figure 7) are plotted in Figure 8. In addition,
the u3 and f results from both methods along line CD
(shown in Figure 7) are plotted in Figures 9 and 10,
854 Journal of Intelligent Material Systems and Structures 29(5)

10-8 0.1
3 FEM
FEM SBFEM
SBFEM
2.5 0.05

2
0
1.5

1 -0.05

0.5
-0.1
0 0.2 0.4 0.6 0.8 1
0
0 0.2 0.4 0.6 0.8 1
Figure 10. Electric potential f results (V) along line CD for
Figure 9. Displacement u3 results (m) along line CD for the the square PZT material embedded in the middle of a square
square PZT material embedded in the middle of a square cement matrix.
cement matrix.

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  SBFEM 2
u1 X N
ui  uFE
respectively. All the displacement and electric potential APD = 100t Ai H ðuFE i Þ i
ð47Þ
At i = 1 uFE
i
results from both methods are in excellent agreement.
Additionally, the results for s11 , s33 and D3 from P
both methods are plotted for comparison in Figures 11 where At = Ni= 1 Ai . In equation (47), Ai is the weight
to 13, respectively. It can also be seen from these fig- related to a node i. N = Nt  NB with Nt representing
ures that the use of quadtree mesh captures the varia- all the nodes in the scaled boundary quadtree mesh and
tion in stresses (s11 , s33 ) and electric displacements NB representing the nodes on the boundary of the
(D3 ) accurately and efficiently as the mesh is refined cement matrix (including both the large and small
automatically around the boundary of the smaller square boundaries). H(z) is the Heaviside unit step func-
square containing the PZT-5H material where the var- tion. uSBFEM
i is the result at node i obtained from the
iation is normally the largest. SBFEM and uFE i is the corresponding result obtained
The average percentage difference (APD) between from the FE analysis. Only results which lie between
the FE results and the SBFE results is computed for the the 2nd percentiles to the 98th percentiles of the non-
various generalized displacements and stresses accord- zero results are included to give meaningful compari-
ing to the following equation son. The APD is provided in Table 1.

Figure 11. Elastic stress s11 results (Pa) for the square PZT material embedded in the middle of a square cement matrix: (a)
SBFEM and (b) FEM.
Saputra et al. 855

Figure 12. Elastic stress s33 results (Pa) for the square PZT material embedded in the middle of a square cement matrix: (a)
SBFEM and (b) FEM.

Figure 13. Electric displacement D3 results (C m2 ) for the square PZT material embedded in the middle of a square cement
matrix: (a) SBFEM and (b) FEM.

It is observed that all of the results in Table 1 are in Table 1. Average percentage difference (APD) between FEM
good agreement (within 2% difference), therefore veri- and SBFEM for the square PZT material embedded in the middle
fying the accuracy and efficiency of the present method. of a square cement matrix.

Result APD

Effect of RVE size on the effective material u1 0.49

properties u3 0.02
f 0.34
It is imperative that the sample size of the cement-based s11 0.86
piezoelectric ceramic composites must be sufficiently s33 0.22
D3 1.90
large or can be regarded as a RVE in order to effec-
tively represent the micro-structural heterogeneities in APD: average percentage difference.
the composite (Gitman et al., 2007).
A range of square RVE sizes are investigated. The with the image size given in brackets are 1.39 mm
side lengths of the square RVEs included in the study (300 3 300 pixels), 1.62 mm (350 3 350 pixels),
856 Journal of Intelligent Material Systems and Structures 29(5)

Figure 14. Five different images of RVE for cement-based piezoelectric ceramic composite with a side length of 1.39 mm.

material properties. The volume fractions of all the

Table 2. Range of volume fraction for the different sample
sizes. samples are approximately 30% with ceramic particle
sizes ranging from 92 to 230 mm. In the images, black
Sample size (mm) Minimum (%) Maximum (%) represents the cement matrix and grey represents the
PZT-5H ceramic particles.
1.39 29.6 30.2 Five different random images are analysed for each
1.62 29.9 30.5
1.85 29.6 30.5 sample size. Similar investigations were conducted in
2.08 29.9 30.5 the literature (Kari et al., 2007). The five images used
2.31 29.8 30.5 for the 1.39 mm sample are shown in Figure 14 as an
example.
The variation of volume fraction of the samples is
1.85 mm (400 3 400 pixels), 2.08 mm (450 3 450
summarized in Table 2.
pixels) and 2.31 mm (500 3 500 pixels) are investi-
The variation of the effective material properties
gated in order to determine the appropriate RVE size
namely c33 , c13 and e33 are presented for the various
which will yield minimal variation in the effective
sample sizes in Tables 3 to 5, respectively.

Table 3. Variation of c33 for the different sample sizes including mean value m and standard deviation s.

Sample size (mm) Minimum (109 N m2 ) Maximum (109 N m2 ) m (109 N m2 ) s (109 N m2 )

1.39 23.7 24.4 24.1 0.2

1.62 23.9 24.5 24.2 0.2
1.85 23.5 24.2 23.9 0.3
2.08 23.6 24.3 24.0 0.3
2.31 23.6 24.1 23.9 0.2

Table 4. Variation of c13 for the different sample sizes including mean value m and standard deviation s.

Sample size (mm) Minimum (109 N m2 ) Maximum (109 N m2 ) m (109 N m2 ) s (109 N m2 )

1.39 6.5 6.6 6.5 0.1

1.62 6.5 6.6 6.6 0.1
1.85 6.5 6.7 6.6 0.1
2.08 6.5 6.6 6.6 0.1
2.31 6.5 6.6 6.6 0.0

Table 5. Variation of e33 for the different sample sizes including mean value m and standard deviation s

Sample size (mm) Minimum (C m2 ) Maximum (C m2 ) m (C m2 ) s (C m2 )

1.39 3.2 3.4 3.3 0.1

1.62 3.2 3.4 3.3 0.1
1.85 3.1 3.3 3.2 0.1
2.08 3.1 3.3 3.3 0.1
2.31 3.2 3.3 3.3 0.0
Saputra et al. 857

Figure 15. Images of cement-based piezoelectric ceramic composite with a volume fraction of (a) 10.7%, (b) 20.8%,
(c) 30.2%, (d) 41.0% and (e) 51.0%.

Figure 16. Quadtree meshes for images of cement-based piezoelectric ceramic composite with a volume fraction of
(a) 10.7%, (b) 20.8%, (c) 30.2%, (d) 41.0% and (e) 51.0%.

The relative maximum percentage difference from
the mean value of all the effective material properties
shows small fluctuations (within 3%) which do not
seem to get smaller as the RVE size increases.
Therefore, it is concluded that the smallest image size
investigated is already larger than the minimum RVE.
Effective material properties for varying volume
fractions
A series of sample cement-based piezoelectric ceramic
composites (each with a resolution of 330 3 330 pixels
where the image represents a sample size of 1.52 mm
by 1.52 mm) are investigated. Various volume fractions
of the ceramic composites are considered namely
10.7%, 20.8%, 30.2%, 41.0% and 51.0%. In all the
samples, the approximate diameters of the ceramic par-
ticles vary from 92 to 230 mm. These images are shown
in Figure 15. Similar to the previous section, the black
colour represents the cement matrix and the grey col-
our represents the PZT-5H ceramic particles. The cor-
responding quadtree meshes are given in Figure 16.
The computed effective material coefficients for
various volume fractions considered are tabulated in
Table 6. The effective elastic material properties for
varying volume fractions are also presented in Figures
17 and 18. Additionally, the effective piezoelectric
material properties for varying volume fractions are
shown in Figures 19 and 20.
One can observe that the effective elastic material
properties grow with increasing volume fraction of
piezoelectric ceramic particles. Similarly, the values of

Table 6. Effective material properties computed based on the images in Figure 15.

Image a b c d e

Elements 7203 12,045 16,770 20,502 23,373

Nodes 9212 15,334 21,220 25,842 29,286
% Volume fraction 10.7 20.8 30.2 41.0 51.0
c11 (109 N m2 ) 18.1 21.1 24.6 29.8 35.7
c13 (109 N m2 ) 4.7 5.6 6.7 8.4 10.5
c33 (109 N m2 ) 18.0 20.9 24.2 29.4 35.9
c44 (109 N m2 ) 6.7 7.7 8.9 10.6 12.5
e15 (C m2 ) 1.0 1.9 2.8 4.2 5.7
e31 (C m2 ) 20.5 21.1 21.7 22.9 23.9
e33 (C m2 ) 1.1 2.2 3.3 4.8 6.5
858 Journal of Intelligent Material Systems and Structures 29(5)

% %
Figure 17. Variation of c11 and c33 with the volume fraction of Figure 20. Variation of e31 with the volume fraction of ceramic
ceramic composite. composite.

% %
Figure 18. Variation of c13 and c44 with the volume fraction of Figure 21. Variation of d33 with the volume fraction of ceramic
ceramic composite. composite from the present method and the experimental
results from Xing et al. (2008).

the effective piezoelectric coefficients. Figure 21 pre-

sents a comparison of effective piezoelectric strain fac-
tor obtained by experimental measurement for three
different volume fractions of the ceramic composite
and the result from the present computational method.
The results are in good agreement.

Effect of circular particle size on the effective

% mic particles on the effective material properties is con-
Figure 19. Variation of e15 and e33 with the volume fraction of
ceramic composite.
effective piezoelectric coefficients also increase as the
volume fraction grows larger.
In the literature, one can find only experimental
results for the piezoelectric strain factor (Xing et al.,
2008). It is easy to get piezoelectric strain factor from
material properties
A study on the effect of different sizes of circular cera-
ducted here. Two cases are considered with the first
case having only one large circular PZT-5H material at
the centre and another one having four equal-sized cir-
cular PZT-5H material with equal spacing at the centre
of a square cement matrix. The volume fraction of the
PZT-5H is chosen to be approximately 30%.
Note that the geometry of the RVE with four inclu-
sions can be obtained as a composition of four RVEs
with one inclusion each and subsequent re-scaling by
Saputra et al.
Figure 22. Finite element mesh for (a) sample with one large
inclusion and (b) sample with four small inclusions.
the factor 1=4. Thus, it can be said that the RVE with
four inclusions is composed of four identical cells with
each cell involving one inclusion. Furthermore, from
the applied boundary conditions and geometrical sym-
metry, one can identify the boundary conditions on
each cell and show that these boundary conditions are
identical with boundary conditions applied to the RVE
with one inclusion. The proof is similar to that given
for the RVE consisting of elastic skeleton with four cir-
cular voids (Sladek et al., 2017b).
Since the considered boundary value problems are
solved within the linear theory, the solution in each cell
of the RVE with four inclusions is expected to be iden-
tical with the solution in the RVE with one inclusion,
where the diameter of inclusion is two times larger than
in the RVE with four inclusions in order to keep the
same volume fraction of inclusions in both of the con-
sidered RVEs.
Analysis using FEM. The size of the square RVE is set to
be 1.5 mm by 1.5 mm. For the case with one large cir-
cular inclusion, the radius of the circle in radius of the
circle is 0.464 mm. While for the case with four circular
inclusion, each circle has a radius of 0.232 mm. Hence,
the total area of circular PZT-5H is approximately
30% of the RVE’s area for both samples. Two different
meshes in ANSYS (coarse and fine) are produced for
each sample to check the convergence of the results.
Only quadrilateral elements are used in the meshes with
the element type (PLANE13) which has linear shape
functions. The fine mesh for the case with one large
inclusion are shown in Figure 22(a) and the fine mesh
for the case with four smaller inclusions are shown in
Figure 22(b).
Analysis using SBFEM. Similarly, the size of the square
RVE is set to be 1.5 mm by 1.5 mm. For the case with
one large circular inclusion, an image with a resolution
of 256 3 256 pixels is utilized. On the other hand, for
the case with four circular inclusion, a larger image with
859
Figure 23. Quadtree meshes for (a) sample with one large
inclusion and (b) sample with four small inclusions.
resolution of 512 3 512 pixels is used to capture the
detail of the circular geometry.
In order to make the area of PZT-5H material to be
30% of the RVE’s area, the radii of the circles in are set
to be equal to the side lengths of 79 pixels. The maxi-
mum size in the quadtree mesh is set to be 8 3 8 pixels
and the minimum size is set to be 1 pixel. The meshes
are shown in Figure 23 for both cases.
Result comparison. The results from both FEM and
SBFEM for the one large circular particle are presented
in Table 7, while the results for the problem with four
small circular particles are summarized in Table 8.
After analysing both Tables 7 and 8, it was found
that the results between the coarse and fine mesh are
very close (less than 0.4% difference) which shows that
the FE result has converged. The results between the
FE analysis and SBFE analysis also show very good
agreement (less than 1.5% difference). The effective
material properties do not change significantly in the
two cases investigated. This observation agrees with the
above mentioned expectation for the uniform distribu-
tion of inclusions. Therefore, it can be concluded that
the size of the particle does not affect the effective mate-
rial properties. Similar conclusion was also obtained by
Kari et al. (2007) which mentions that the effective
material properties are insensitive to the particle size in
linear elastic case unless debonding or damage in the
material is considered.
Effect of random particle size and shape on the
effective material properties
The effect of the random ceramic particle size and
shape on the effective material properties is investigated
here. Three images with different average particle sizes
are considered where each image represents a sample
size of 1.52 mm by 1.52 mm. The minimum and maxi-
mum particles sizes in each image are tabulated in
Table 9. Note that since the smallest particle size
requires higher resolution to capture the shape of the
860 Journal of Intelligent Material Systems and Structures 29(5)

Table 7. Effective material properties for sample with one large circular particle.

Analysis FEM (coarse mesh) FEM (fine mesh) SBFEM

Elements 4224 16,896 2704

Nodes 4273 16,993 3269
% Volume fraction 30.0 30.0 29.9
c11 (109 N m2 ) 23.9 24.0 24.0
c13 (109 N m2 ) 6.2 6.2 6.2
c33 (109 N m2 ) 23.7 23.7 23.7
c44 (109 N m2 ) 8.6 8.6 8.6
e15 (C m2 ) 2.5 2.4 2.5
e31 (C m2 ) 21.4 21.4 21.4
e33 (C m2 ) 3.3 3.3 3.3

FEM: finite element method; SBFEM: scaled boundary finite element method.

Table 8. Effective material properties for sample with four small circular particles.

Analysis FEM (coarse mesh) FEM (fine mesh) SBFEM

Elements 16,896 67,584 10,816

Nodes 16,993 67,777 12,945
% Volume fraction 30.0 30.0 29.9
c11 (109 N m2 ) 23.9 23.9 23.9
c13 (109 N m2 ) 6.1 6.1 6.1
c33 (109 N m2 ) 23.7 23.7 23.7
c44 (109 N m2 ) 8.3 8.3 8.3
e15 (C m2 ) 2.2 2.2 2.2
e31 (C m2 ) 21.4 21.4 21.5
e33 (C m2 ) 3.3 3.3 3.3

FEM: finite element method; SBFEM: scaled boundary finite element method.

Figure 24. Images with a volume fraction of approximately
30% with (a) small particles and (b) large particles.
particles accurately, a bigger image is utilized in the
analysis (three times the resolution of the other images
in each dimension). All the images have a volume frac-
tion of approximately 30%.
The images containing the small and large particles
are shown in Figure 24 with the corresponding quad-
tree meshes shown in Figure 25. Please note that
image with medium particles indicated in Table 9 is
the same as Figure 15(c) with its mesh being shown
on Figure 16(c).
The effective material properties for the various
images with different particles sizes investigated are
Figure 25. Quadtree meshes for images with a volume fraction
of approximately 30% with (a) small particles and (b) large
particles.
shown in Table 10. Note that the effective material
properties for image with medium particle are obtained
from the ones corresponding to Image c in Table 6.
It can be seen that the piezoelectric coefficient does
not really change with the particle size as shown by the
results in Table 10. Chaipanich (2007a) has investigated
experimentally the effect of PZT particle size on dielec-
tric and piezoelectric properties of PZT–cement compo-
sites who found that there is an enhancement in
piezoelectric properties as the particle size increases.
This was contributed to lesser contacting surfaces
Saputra et al. 861

Table 9. Information on the range of particle sizes, image size and volume fraction for RVEs with different average particle sizes.

Particles Min. size (mm) Max. size (mm) Size (pixels) Volume fraction

Small 18 55 990 3 990 30.0

Medium 92 230 330 3 330 30.2
Large 369 553 330 3 330 30.2

Table 10. Effective material properties computed based on the of the present method for three-dimensional analyses is
RVEs with different average particle sizes. also possible by employing the approach developed in
Saputra et al. (2017).
Particles Small Medium Large

Elements 183,975 16,770 6660 Acknowledgements

Nodes 230,880 21,220 8508
c11 (109 N m2 ) 24.1 24.6 24.6 The authors would like to thank Dr Zhang Zihua for the
c13 (109 N m2 ) 6.7 6.7 6.7 piezoelectric ceramic composite images.
c33 (109 N m2 ) 24.0 24.2 24.3
c44 (109 N m2 ) 8.7 8.9 9.1
e15 (C m2 ) 2.8 2.8 2.8 Declaration of conflicting interests
e31 (C m2 ) 21.6 21.7 21.5 The author(s) declared no potential conflicts of interest with
e33 (C m2 ) 3.2 3.3 3.3 respect to the research, authorship and/or publication of this
article.

between the cement matrix and the PZT particles.

However, in this article, it is considered that ideal
bonding is present between the cement matrix and the
PZT particles; therefore, the size and shape of ceramic
particles with similar volume fraction have no signifi-
cant effect on the composite’s effective material proper-
ties. This observation also agrees with the previous
investigation with circular ceramic particles.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: The authors acknowledge the support by the Slovak
Science and Technology Assistance Agency registered under
number APVV-14-216 as well as VEGA 2/0046/16 and the
support by the Australian Research Council under grant num-
ber DP160104628.

Conclusion
The SBFEM is applied to solve properly chosen bound-
ary value problems for a RVE to obtain effective mate-
rial coefficients in cement-based piezoelectric ceramic
composite. It was shown that the SBFE analysis with
quadtree mesh can analyse images of cement-based
piezoelectric ceramic composite efficiently. The accu-
racy of the SBFEM is verified by comparing the result
with the FEM. The present numerical method also
allows the determination of the appropriate RVE size
to be performed easily. In the study, it was found that
the effective material properties of the composite vary
mainly with the volume fraction of ceramic particles.
Good agreement with experimental results is also
obtained. The size, shape and distribution of inclusions
within the RVE affect the macro-structural effective
values of material coefficients less significantly and only
if such irregularities lead to micro-structural boundary
value problems which are not re-scaling invariant in the
context of linear theory. The method presented here
can be extended to perform computation of effective
material properties of composites involving magnetoe-
lectrostatic materials (Kuo and Pan (2011), Wang and
Pan (2007) and Dinzart and Sabar (2012). An extension
References
Aizawa S, Kakizawa T and Higasino M (1998) Case studies
of smart materials for civil structures. Smart Materials and
Structures 7(5): 617–626.
Bhalla S and Kiong Soh C (2003) Structural impedance based
damage diagnosis by piezo-transducers. Earthquake Engi-
neering & Structural Dynamics 32(12): 1897–1916.
Bird G, Trevelyan J and Augarde CE (2010) A coupled BEM/
scaled boundary FEM formulation for accurate computa-
tions in linear elastic fracture mechanics. Engineering Anal-
ysis with Boundary Elements 34(6): 599–610.
Chaipanich A (2007a) Dielectric and piezoelectric properties
of PZT–cement composites. Current Applied Physics 7(5):
537–539.
Chaipanich A (2007b) Effect of PZT particle size on dielectric
and piezoelectric properties of PZT–cement composites.
Current Applied Physics 7(5): 574–577.
Chiong I, Ooi ET, Song C, et al. (2014) Computation of
dynamic stress intensity factors in cracked functionally
graded materials using scaled boundary polygons. Engi-
neering Fracture Mechanics 131: 210–231.
Deeks AJ and Wolf JP (2002) A virtual work derivation of
the scaled boundary finite-element method for elasto-
statics. Computational Mechanics 28(6): 489–504.
Dinzart F and Sabar H (2012) Magnetoelectric effect in
coated fibrous magnetic–piezoelectric composites. Journal
862
of Intelligent Material Systems and Structures 23(11):
1249–1261.
Dong B and Li Z (2005) Cement-based piezoelectric ceramic
smart composites. Composites Science and Technology
65(9): 1363–1371.
Gitman I, Askes H and Sluys L (2007) Representative vol-
ume: existence and size determination. Engineering Frac-
ture Mechanics 74(16): 2518–2534.
Giurgiutiu V and Zagrai AN (2002) Embedded self-sensing
piezoelectric active sensors for on-line structural identifica-
tion. Journal of Vibration and Acoustics: Transactions of
the ASME 124(1): 116–125.
Huston DR, Fuhr PL, Ambrose TP, et al. (1994) Intelligent
civil structures-activities in vermont. Smart Materials and
Structures 3(2): 129–139.
Kari S, Berger H, Rodriguez-Ramos R, et al. (2007) Compu-
tational evaluation of effective material properties of com-
posites reinforced by randomly distributed spherical
particles. Composite Structures 77(2): 223–231.
Kuo HY and Pan E (2011) Effective magnetoelectric effect in
multicoated circular fibrous multiferroic composites. Jour-
nal of Applied Physics 109(10): 104901.
Li C, Man H, Song C, et al. (2013) Fracture analysis of piezo-
electric materials using the scaled boundary finite element
method. Engineering Fracture Mechanics 97: 52–71.
Li Z, Dong B and Zhang D (2005) Influence of polarization
on properties of 0–3 cement-based PZT composites.
Cement & Concrete Composites 27(1): 27–32.
Li Z, Zhang D and Wu K (2001) Cement matrix 2-2 piezoelec-
tric composite – part 1. sensory effect. Materials and struc-
tures 34(8): 506–512.
Li Z, Zhang D and Wu K (2002) Cement-based 0-3 piezoelec-
tric composites. Journal of the American Ceramic Society
85(2): 305–313.
Lynch JP (2007) An overview of wireless structural health
monitoring for civil structures. Philosophical Transactions
of the Royal Society A: Mathematical, Physical and Engi-
neering Sciences 365(1851): 345–372.
Mascarenas DL, Todd MD, Park G, et al. (2007) Develop-
ment of an impedance-based wireless sensor node for
structural health monitoring. Smart Materials and Struc-
tures 16(6): 2137–2145.
Natarajan S and Song C (2013) Representation of singular
fields without asymptotic enrichment in the extended finite
element method. International Journal for Numerical Meth-
ods in Engineering 96(13): 813–841.
Parton VZ and Kudryavtsev BA (1988) Electromagnetoelasti-
city: Piezoelectrics and Electrically Conductive Solids. New
York: Taylor & Francis.
Pines DJ and Lovell PA (1998) Conceptual framework of a
remote wireless health monitoring system for large civil
structures. Smart Materials and Structures 7(5): 627–636.
Saputra AA, Birk C and Song C (2015) Computation of
three-dimensional fracture parameters at interface cracks
Journal of Intelligent Material Systems and Structures 29(5)
and notches by the scaled boundary finite element method.
Engineering Fracture Mechanics 148: 213–242.
Saputra AA, Talebi H, Tran D, et al. (2017) Automatic
image-based stress analysis by the scaled boundary finite
element method. International Journal for Numerical
Methods in Engineering 109(5): 697–738.
Shifeng H, Jun C, Futian L, et al. (2004) Poling process and
piezoelectric properties of lead zirconate titanate/sulphoa-
luminate cement composites. Journal of Materials Science
39(23): 6975–6979.
Sladek J, Sladek V and Pan E (2017a) Effective properties of
coated fiber composites with piezoelectric and piezomag-
netic phases. Journal of Intelligent Material Systems and
Structures 28(1): 97–107.
Sladek J, Sladek V, Krahulec S, et al. (2012) MLPG analysis
of layered composites with piezoelectric and piezomagnetic
phases. Computers Materials & Continua 29(1): 75–101.
Sladek J, Sladek V, Krahulec S, et al. (2016) Crack analyses in
porous piezoelectric brittle materials by the SBFEM. Engi-
neering Fracture Mechanics 160: 78–94.
Sladek V, Musil B, Sladek J, et al. (2017b) Microstructural
evaluation of effective elasticity coefficients in materials
with micro-voids. Procedia Engineering 190: 170–177.
Song C (2004) A matrix function solution for the scaled
boundary finite-element equation in statics. Computer
Methods in Applied Mechanics and Engineering 193(23):
2325–2356.
Song C and Wolf JP (1997) The scaled boundary finite-
element method – alias consistent infinitesimal finite-
element cell method – for elastodynamics. Computer
Methods in Applied Mechanics and Engineering 147(3–4):
329–355.
Song G, Sethi V and Li HN (2006) Vibration control of civil
structures using piezoceramic smart materials: a review.
Engineering Structures 28(11): 1513–1524.
Wang F, Wang H, Song Y, et al. (2012) High piezoelectricity
0–3 cement-based piezoelectric composites. Materials Let-
ters 76: 208–210.
Wang X and Pan E (2007) Magnetoelectric effects in multifer-
roic fibrous composite with imperfect interface. Physical
Review B 76(21): 214107.
Xing F, Dong B and Li Z (2008) Dielectric, piezoelectric, and
elastic properties of cement-based piezoelectric ceramic
composites. Journal of the American Ceramic Society 91(9):
2886–2891.
Zhao P, Kim S and Hinderliter B (2016) Investigation of
cement–sand-based piezoelectric composites. Journal of
Intelligent Material Systems and Structures 27(12):
1666–1672.
Zhu C, Lin G and Li J (2013) Modelling cohesive crack
growth in concrete beams using scaled boundary finite ele-
ment method based on super-element remeshing tech-
nique. Computers & Structures 121: 76–86.