Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 190 (2017) 54 – 61

Structural and Physical Aspects of Construction Engineering

Bending Analysis of FGM Plates under Thermal Load

Ladislav Satora*,Vladimir Sladeka, Jan Sladeka
a
Institute of Construction and Architecture, Slovak Academy of Sciences, Dúbravska cesta 9, 845 03 Bratislava, Slovakia

Abstract

In this paper, the bending of thin functionally graded plates under thermal load is considered within the classical theory of
thermoelasticity. The variable material properties of plate (such as the Young’s modulus, thermal expansion coefficient, etc.) are
allowed to be continuous functions of the position. The governing equations which are given by the 4th order partial differential
equations are decomposed into the 2nd order partial differential equations in order to overcome the inaccuracy of approximation
of high order derivatives of field variables. The strong form meshless formulations for solution of thin plate bending problem is
developed in combination with Moving Least Squares approximation scheme. The attention is paid to the numerical investigation
of the influence of several parameters of gradations of material coefficients on the deflection of the plate.
© 2017
© 2017TheTheAuthors.
Authors. Published
Published by Elsevier
by Elsevier Ltd. is an open access article under the CC BY-NC-ND license
Ltd. This
Peer-review under responsibility of the issue editors.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of SPACE 2016
Keywords: bending analysis; functionally graded materials; meshless methods; thermoelasticity

1. Introduction

Thermal stresses, especially at the interface between two different materials, often represent a significant factor in the
failure of laminated composite structures. Thus, there is a need to accurately predict thermal stresses in composite
structures [1]. Simultaneously, this is also a strong motivation to replace laminated plate structures by FGM ones if
possible. Therefore the study of behavior of FGM plates under thermal loadings has become attractive, what can be
confirmed by a huge amount of papers devoted to thermoelastic analyses with using various plate bending theories
for FGM plates under static and/or dynamic loadings [7-8]. In the classical thermoelasticity, the temperature field is
coupled with the elastic fields, in general, but the temperature change does not propagate in a wave form [2]. In this
paper, the bending of thin FGM plates under thermal load is considered within the classical theory of thermoelasticity.
The governing equations which are given by the 4th order partial differential equations (PDE) are decomposed into
the 2nd order PDEs in order to overcome the inaccuracy of approximation of high order derivatives of field variables.

* Corresponding author. Tel.: +421-25-930-9216

E-mail address: ladislav.sator@savba.sk

1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of SPACE 2016
doi:10.1016/j.proeng.2017.05.307
 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61 55

The strong form meshless formulations for solution of thin plate bending problem is developed in combination with
Moving Least Squares (MLS) approximation scheme [3]. The attention is paid to the numerical investigation of the
influence of several parameters of gradations of material coefficients on the deflection of the plate.

2. Governing equations for FGM plates in classical thermoelasticity

Let us consider FGM plates with unspecified in-plane gradation of Young’s modulus, thermal expansion and heat
conduction coefficients, specific heat capacity and mass density; and the power-law gradation in the transversal
direction as
p
§1 ·
E (x, z ) = E0 EH (x) EV ( z ) , EV ( z ) = 1 + ζ ¨ ± z ¸
©2 ¹
r
§1 ·
α (x, z ) = α 0α H (x)αV ( z ) , αV ( z ) = 1 + ξ ¨ ± z ¸
©2 ¹
s
§1 ·
k (x, z ) = k0 k H (x)kV ( z ) , kV ( z ) = 1 + ω ¨ ± z ¸ (1)
©2 ¹
q
§1 ·
c(x, z ) = c0 cH (x)cV ( z ) , cV ( z ) = 1 + χ ¨ ± z ¸
©2 ¹
g
§1 ·
ρ (x, z ) = ρ0 ρ H (x) ρV ( z ) , ρV ( z ) = 1 + ε ¨ ± z ¸
©2 ¹
where ζ , ξ , ω , χ , ε and p, r , s, q, g are the values of the level and exponent of the power-law gradation, respectively.

Before writing of the governing equations in Kirchhoff- Love theory (KLT) of thin plates for thermoelastic boundary
value problems (BVP) we define the dimensionless coordinates and time variable as
xβ ∗ x ∗ ∗ t
xβ∗ := , x3 := 3 = h (x) z , h ( x ) = h0 h∗ ( x) , t := (2)
L h0 T

with L, h0 , T being the characteristic in-plane dimension, plate thickness and time interval. The dimensionless field
variable in the KLT (in-plane displacements, deflections, temperature) within classical thermoelasticity are introduced
as
u β ( x, t ) ∗ w(x, t ) ∗ ϑ ( x, t )
uβ∗ (x, t ) := , w (x, t ) := , ϑa (x, t ) := a , (3)
h0 h0 θ0
where we have omitted the superscript ∗ in coordinates and time variable and three fields ϑa ( x, t ) occur in the
truncated expansion of the temperature w.r.t. the z -coordinate: θ * ( x, z , t ) ≈ 1 + ϑ0* ( x, t ) + zϑ1* ( x, t ) + z 2ϑ2* ( x, t ) .

Combination of the variational formulation for quasi-static thermoelastic problems with assumptions of the KLT for
thin plates yields the strong form of the governing equations in view of the PDE
2 2 (θϑa ) 2 2
(θϑ ) ∗
¦ C a (x)ϑa (x, t ) + ¦ G, β (x)ϑa∗, β (x, t ) + ¦ G (θϑa ) (x)ϑa∗, ββ (x, t ) + ¦ D(θϑa ) (x)ϑa∗ (x, t ) = 0 (4a)
a =1 a =0 a =0 a =0

2
ªC (uu ) ( x)τ ∗(u ) ( x, t ) º + ªC (uw) ( x)τ ∗( w) ( x, t ) º + ¦ ªC (uϑa ) ( x)τ ∗(ϑa ) ( x, t ) º = 0 (4b)
¬ αβ ¼, β ¬ αβ ¼ , β a =0 ¬« αβ ¼», β
56 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61

2 ∗(ϑ )
ªC ( wu ) ( x)τ ∗(u ) ( x, t ) º + ªC ( ww) ( x)τ αβ
∗( w)
( x, t ) º + ¦ ªC ( wϑa ) (x)τ αβ a ( x, t ) º» = − q∗ ( x, t ) (4c)
¬ αβ ¼,αβ ¬ ¼,αβ a =0 «¬ ¼ ,αβ

plus two complementary algebraic equations should be composed from the possible thermal boundary conditions on
the bottom and top of the plate given by
Dirichlet type

1 1 θ (x, ±1 / 2, t )
ϑ0∗ (x, t ) ± ϑ1∗ (x, t ) + ϑ2∗ (x, t ) = −1 (5a)
2 4 θ0
or Neuman type
q (x, ±1 / 2, t )
± k0 k H (x)kV (±1 / 2) «ªϑ1∗ (x, t ) ± ϑ2∗ (x, t ) »º = . (5b)
¬ ¼ θ0
The most frequently applied boundary conditions for mechanical fields on the boundary edge of the plate are given
as follows:
(i) on clamped edge:

∂w∗
uα∗ ( x, t ) = 0 , w∗ (x, t ) =0, (x, t ) =0, (6a)
∂Ω ∂Ω ∂n
∂Ω

(ii) on simply supported edge:

∗ ∗( w)
nβ (x)Tαβ (x, t ) = 0 , w∗ (x, t ) = 0 , nα (x)nβ (x) M αβ (x, t ) = 0, (6b)
∂Ω ∂Ω ∂Ω

while for thermal fields on the boundary edge ∂Ω we name the following boundary conditions

(i) Dirichlet type: ϑ0∗ ( x, t ) = θ ( x, z = 0, t ) / θ 0 − 1 (7a)

∂Ω

∗
(ii) Neuman type: −k0 k H (x)kV (0)nβ (x)ϑ0, β (x, t ) = q (x, 0, t ) / θ0 (7b)
∂Ω
The deformation fields are defined as

∗(u )
ταβ (x, t ) =
H ∗
2
( )
uα , β (x, t ) + u∗β ,α (x, t ) +νδαβ uγ∗,γ (x, t ) (8a)

∗( w)
ταβ (x, t ) := Hw,∗αβ (x, t ) +νδαβ ∇2 w∗ (x, t ) (8b)

∗(ϑ )
ταβ a (x, t ) := (1 +ν )ϑa∗ (x, t )δαβ , (8c)

while coefficients C (<) , G (<) , D(<) are defined in Appendix A.

The gradients of deformation fields include higher order derivatives of primary field variables, which lead to
increasing inaccuracy of the solution. To overcome this problem, we decompose original 4th order governing PDEs
into coupled set of 2nd order PDEs by introducing new field variables as

m∗ (x, t ) := ∇ 2 w∗ (x, t ) , sα∗ (x, t ) := ∇ 2 uα∗ (x, t ) (9)

Then, we receive
 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61 57

∗(u ) H ∗ H + 2ν ∗ ∗(u ) ∗
ταβ ,β (x, t ) = sα (x, t ) + uβ , βα (x, t ) , ταβ ,αβ (x, t ) = ( H +ν ) sβ ,β (x, t ) (10a)
2 2

∗( w) ∗ ∗( w) 2 ∗
ταβ ,β (x, t ) = ( H +ν )m,α (x, t ) , ταβ ,αβ (x, t ) = ( H +ν )∇ m (x, t ) (10b)

∗(ϑ )
(
ταβ ,aβ (x, t ) := (1 +ν ) ϑa∗ (x, t ) − t1ϑa∗ (x, t ) ),α , ∗(ϑ )
( )
ταβ ,aαβ (x, t ) := (1 +ν ) ∇2ϑa∗ (x, t ) − t1∇2ϑa∗ (x, t ) . (10c)

Now, 3 equations given by (9), plus 4 equations given by (4), plus 2 complementary algebraic equations composed
from the possible thermal boundary conditions on the bottom and top of the plate given by Eq. (5) yield the complete
set of governing equations for the primary field variables

{w∗ (x, t ), m∗ (x, t ), uβ∗ (x, t ), sβ∗ (x, t ), ϑ0∗ (x, t ),ϑ1∗ (x, t ),ϑ2∗ (x, t )} .
3. Moving Least Squares (MLS) approximation technique

In this paper, we have used the strong formulation with employing the meshless approximation for primary field
variables by Central Approximation Node (CAN) concept of Moving Least Square (MLS) approximation technique
[4-5].
Without going into details [6], the approximation of field variable u (x) around the central approximation node xq
can be expressed by
Nq
u (x) ≈ ¦ uˆ a φ (q,a) (x) , a = n( q , a ) , (11)
a =1
where a is the global number of the a -th node from the influence domain of x q , N q is the number of nodal points
in the influence domain, and φ ( q ,a ) ( x) is the shape function associated with the node n( q, a ) . The CAN can be
selected as the nearest node to the field point x .
The derivatives of the field variable u ( x) can be approximated by derivatives of approximated fields (D0-approach)
[4], i.e.

Nq
u,ijk ...l ( x) ≈ ¦ uˆ a φ,(ijkq,a...)l (x) . (12)
a =1

4. Numerical experiments

In presented numerical investigations of thin elastic plates under stationary thermal loads within the KLT, we consider
a square plate L × L with clamped all the edges, L = 1 , and various gradations of material properties: (i) in-plane
gradation; (ii) transversal gradation. Poisson's ratio is assumed to be constant ν = 0.3 , while various values of the
power-law gradation exponent and levels of gradation are considered. The stationary thermal loading is applied with
prescribed zero temperature ( θ * = 0 ) on bottom surface and a finite temperature ( θ * = 20 ) on top surface, while the
heat flux is vanishing ( q * = 0 ) on the plate edges

In all numerical computations, we have used a uniform distribution of nodal points (36x36 nodes) with δ being the
distance between two neighbour nodes. The other parameters in the MLS-approximation have been taken as: radius
58 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61

of the interpolation domain ρ a = 3.001δ , shape function parameter c a = δ , and cubic polynomial basis. In stationary
problems, the thermal task can be solved separately from the mechanical one. First of all we shall present the
distribution of temperatures and heat fluxes through the plate thickness (Figure 1).

Fig. 1. Transversal distribution of: (a) temperature; (b) heat flux.

Figure 1 illustrates the influence of the levels of gradation of various material properties ( ξ for transversal gradation
of thermal expansion coefficient, ζ for transversal gradation of Young's modulus, ω for transversal gradation heat
conduction coefficient) on the distribution of temperatures and heat fluxes across the plate thickness in case of
considered BVP. It can be seen from Figure 1 that the distribution of the temperature field is linear as long as the heat
conduction coefficient is homogeneous ( ω = 0 ). Rather strong influence of the parameter ω on the distribution of
the temperature and heat flux is observed.

Fig. 2. In-plane distribution of (a) deflections and (b) in-plane displacements of FGM plates with shown levels ( κ 0 , κ1 , κ 2 ) of in-plane
power-law gradation of material coefficients ( E , α, k ), respectively.

In what follows, we concise to presentation of numerical results for deflections and in-plane deformations of the square
thin elastic FGM plates under stationary thermal loads with assuming power-law gradations of various material
coefficients.
 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61 59

The significant influence of gradation parameters on deflections and in-plane deformations can be observed in thin
elastic FGM plates under stationary thermal loads (Figures 2). The levels of the in-plane power-law gradation of the
thermal expansion coefficient play the crucial role for the mechanical responses of plates to stationary thermal load
conditions presented above. It can be seen that the deflection response arises only if κ 0 ≠ 0 ∧ κ1 ≠ 0 , while the in-
plane displacements are developed if κ 0 ≠ 0 or κ1 ≠ 0 or κ 0 ≠ 0 ∧ κ1 ≠ 0 . The employed values for the exponents of
gradation b0 = b1 = b2 = 1 correspond to linear gradation.

5. Conclusions

Incorporating the interaction between the elastic and thermal fields within the theory of classical thermoelasticity, we
derived the mathematical model for 2D problems describing quasi-static behaviour of bending modes, in-plane
deformations and heat conduction problems in the mid-plane of the plate. In all the derivations, we supposed
continuous variation of material coefficients (FGM plates) and variable plate thickness. The transversal gradation of
the Young's modulus brings a new physical phenomenon in bending analysis, namely the coupling between the
bending and in-plane deformation modes.
The derived governing equations are given by a system of the coupled 4th order PDE with variable coefficients. In any
numerical method, the accuracy of approximations of derivatives of field variables is decreasing with increasing the
order of derivatives. For this purpose, we decomposed the formulation with high order derivatives into the formulation
for coupled fields with derivatives not higher than 2nd order. The price which should be paid for such decrease of the
order of derivatives is introduction of new field variables.
We proposed and developed the strong formulation for solution of complex multifield problems represented by
systems of the PDE with variable coefficients. The strong formulation offers the method which is mathematically as
simple as and physically correct because of preserving the physical nature of the problems. For spatial approximations
of field variables, we employed Moving Least Squares approximation technique. The numerical experiments
document the creation of finite responses in deflections, in-plane displacement and thermal fields in FGM plates under
thermal loads as well as indicate the significant influence of gradation parameters on the distributions of the
mechanical and thermal fields.

Acknowledgement

The financial support of the Slovak Research and Development Agency under the contract No. APVV-14-0440 is
greatly acknowledged.

References
[1] Vel.S.S. and B. R.C, Generalized plane strain thermoelastic deformation of laminated anisotropic thick plates, Int. J. Solids Struct, 38
(2001), pp. 1395-1414
[2] G. A.E. and L. K. ., Thermoelasticity, J. Elast., 2 (1972), pp. 1-7
[3] P. Lancaster and K. Salkauskas, Surfaces generated by moving least squares method, Math. Comput., 37 (1981), pp. 141-158
[4] V. Sladek, J. Sladek and C. Zhang, Computation of stresses in non-homogeneous elastic solids by local integral equation method: a
comparative study, Computational Mechanics, 41 (2008), pp. 827-845
[5] V. Sladek, J. Sladek and C. Zhang, Local integral equation formulation for axially symmetric problems involving elastic FGM, Engineering
Analysis with Boundary Elements, 32 (2008), pp. 1012-1024
[6] V. Sladek, J. Sladek and L. Sator, Physical decomposition of thin plate bending problems and their solution by mesh-free methods,
Engineering Analysis with Boundary Elements, 37 (2013), pp. 348-365
[7] L. Sator, V. Sladek, J. Sladek and D. Young, Elastodynamics of FGM plates by mesh-free method, Composite Structures, 140 (2016), pp.
309-322
[8] L. Sator, V. Sladek and J. Sladek, Coupling effects in elastic analysis of FGM composite plates by mesh-free methods, Composite
Structures, 115 (2014), p. 100-110

Appendix A
In this Appendix, we present the explicit expressions for all coefficients included in the governing equations given
by Eq. (4a).
60 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61

2
(θϑa ) § L· k H ( x) ª
C (x) := ¨ ¸ kV (1/ 2) + (−1)a kV (−1/ 2) º , a = 1, 2 (A.1)
¬ ¼
( )
2
© h0 ¹ h∗ (x)

G, β
(θϑa )
(
(x) := d(0)a + ωd( s )a k H , β (x) , ) a = 0,1, 2

(
G (θϑa ) (x) := d(0) a + ω d( s ) a k H (x) , a = 0,1, 2 )
L2 ρ0 c0
D (θϑa ) (x) :=
k0 T
( )
d(0) a + ε d( g ) a + χ d( q ) a + εχ d( g + q ) a ρ H (x)cH (x) , a = 0,1, 2

1 −ν L
C (uu ) (x) := 12
H h0
(
d(0)0 + ζ d( p )0 D1H (x) ) (A.2)

1 −ν ª 4 º
C (uw) (x) := 12 ζ «(c1 − 1)d( p)1 − 3 c1c2 d ( p )3 » D2 H (x)
H ¬ ¼
2
1 −ν § L ·
C (uϑa ) (x) := −12 (
¨ ¸ α 0θ 0 d(0) a + ξ d( r ) a + ζ d( p ) a + ξζ d( p + r ) a B1H (x)
H © h0 ¹
)
1 −ν L ª 4 º L
C ( wu ) (x) := 12 ζ (1 − c1 )d( p )1 + c1c2 d( p )3 » D2 H (x) = − C (uw) (x)
H h0 «¬ 3 ¼ h0

1 −ν 16 ª 16 º½
C ( ww) ( x) := 12 {( c1 − 1) d(0)2 − 9 c1c2 d(0)6 + ζ «(c1 − 1)d ( p )2 − 9 c1c2 d ( p )6 » ¾ D3H ( x)
H ¬ ¼¿
2
1 −ν § L· ­ 4
C ( wϑa ) (x) := 12 ¨ ¸ α 0θ 0 ®(c1 − 1)d(0) a +1 − c1c2 d(0) a +3 +
H © 0¹
h ¯ 3

ª 4 º ª 4 º
+ξ «(c1 − 1)d( r ) a +1 − c1c2 d( r ) a +3 » + ζ «(c1 − 1)d( p) a +1 − 3 c1c2 d( p) a +3 » +
¬ 3 ¼ ¬ ¼

ª 4 º½
+ξζ «(c1 − 1) d ( p + r ) a +1 − c1c2 d ( p + r ) a +3 » ¾ B2 H (x) ,
¬ 3 ¼¿

( ) ( )
j j
where D jH ( x) := h∗ ( x) EH ( x) , B jH ( x) := h∗ (x) EH ( x)α H (x) .

In evaluation of these coefficients, we have used the closed form integrals given in Appendix B.
 Ladislav Sator et al. / Procedia Engineering 190 (2017) 54 – 61 61

Appendix B
In this Appendix, we present the analytical expressions for integrals employed in evaluation of coefficients
summarized in Appendix A.
The auxiliary integrals used in evaluation of coefficients (A.1):
1/2 1/2 ª §1 · º
s
³ z kV ( z )dz = ³ z «1 + ω ¨ ± z ¸ » dz = d (0) a + ω d ( s ) a , a = 0,1, 2
a a
(B.1)
−1/2 −1/2 «¬ ©2 ¹ »¼

1/2 1/2 ª §1 ·
g
§1 ·
q
§1 ·
g +q º
³ z ρV ( z )cV ( z ) dz = ³ z «1 + ε ¨ ± z ¸ + χ ¨ ± z ¸ + εχ ¨ ± z ¸
a a
» dz =
−1/2 −1/2 «¬ ©2 ¹ ©2 ¹ ©2 ¹ »¼

= d(0)a + ε d( g )a + χ d(q )a + εχ d( g + q )a

in which
1/2 1 + (−1)a
d(0) a := ³ z a dz = (B.2)
−1/2 (a + 1)2a +1
p a a § a ·§ 1 · k
1/2 §1 · 1§ 1· 1
d( p ) a := ³ z a ¨ ± z ¸ dz = (±1) a ³ ¨ y − ¸ y p dy = (±1)a ¦ ¨ ¸ ¨ − ¸ .
−1/2 ©2 ¹ 0© 2¹ k =0 © k ¹ © 2 ¹ p + 1 + a − k

The auxiliary integrals used in evaluation of coefficients (A.2):

1/2 1/2 ª §1 ·
pº
³ zEV ( z )dz = ³ z «1 + ζ ¨ ± z¸ » dz = ζ d ( p )1 (B.3)
−1/2 −1/2 «¬ © 2 ¹ »¼

1/2
2
1/2
2
ª §1 · º
p
³ z EV ( z )dz = ³ z «1 + ζ ¨ ± z ¸ » dz = d(0)2 + ζ d ( p )2
−1/2 −1/2 «¬ ©2 ¹ »¼

1/2
n +1
1/2
n +1
ª §1 · º
p
³ z EV ( z )dz = ³ z «1 + ζ ¨ ± z ¸ » dz = d (0) n + ζ d ( p ) n
−1/2 −1/2 «¬ ©2 ¹ »¼

1/2
n +1
1/2
n +1
ª §1 ·
pº
³ z EV ( z )dz = ³ z «1 + ζ ¨ ± z¸ » dz = d (0) n +1 + ζ d ( p ) n +1
−1/2 −1/2 «¬ ©2 ¹ »¼

1/2
n
1/2
n
ª §1 · ºª
p
§1 · º
r
³ z EV ( z )αV ( z )dz = ³ z «1 + ζ ¨ ± z ¸ » «1 + ξ ¨ ± z ¸ » dz = d (0) n + ξ d ( r ) n + ζ d ( p ) n + ξζ d ( p + r ) n
−1/2 −1/2 «¬ ©2 ¹ »¼ «¬ ©2 ¹ »¼