FE Analysis of Laminated Composite Plates Using A Higher Order Shear Deformation Theory With Assumed Strains
FE Analysis of Laminated Composite Plates Using A Higher Order Shear Deformation Theory With Assumed Strains
FE Analysis of Laminated Composite Plates Using A Higher Order Shear Deformation Theory With Assumed Strains
Key words
Laminate Composite Plate, Finite Element, Higher Order Shear
Deformation, Locking Phenomenon, Assumed Strain Method
1 INTRODUCTION
Laminated composite plates have been extensively used in many engineering disciplines such as civil
engineering, marine engineering and aerospace engineering due to its high strength to weight ratio
and excellent corrosion resistance. With the growing use of laminated composite material, it be-
comes very important to conduct numerical analysis and to use the resulting information in the
structural design process. This situation clearly has demanded the development of efficient and
accurate numerical analysis techniques which are necessarily required to predict the behaviors of
laminated plates.
In the early days, classical laminated plate theory (CLPT) has been mainly used with negligence
of the effect of transverse shear deformation. However, due to the increasing use of thick laminated
plate in construction, thick plate theories such as the first order shear deformation theory (FSDT)
and the higher order shear deformation theory (HSDT) are needed to take into account transverse
524 S. J. Lee et al / FE analysis of laminated composite plates using a higher order shear deformation theory with assumed strains
shear deformation through the thickness direction of the plates. In particular, The HSDTs do not
required shear correction factor and it can generally guarantee zero transverse shear stress values on
the top and bottom surfaces of the plate. Some important and early works on HSDT can be found
in the open literatures [1-5] where more realistic representation of transverse shear deformation were
generally tried to be provided. Later, Zhang and Yang [6] described some recent developments of
the FEs based on various laminated composite plate theories. Reddy [1] suggested a simple but very
useful HSDT for laminated composite plates. His version of HSDT is based on equivalent single
layer plate theory and it allows parabolic variation of transverse shear stress and also satisfies zero
shear stress boundary conditions at the top and bottom surfaces of the plate. Moreover, it does not
involve any unknown fields which do not have any physical meaning. Bose and Reddy [7, 8] ana-
lyzed laminated plates by using a unified third-order laminate plate theory that contains classical,
first-order and third-order theories and they presented analytical method using the Navier and Levy
equations and the FE method using the unified third order laminate plate theory. A review on the
various methods used in the estimation of transverse and inter-laminar stresses for laminated com-
posite plates and shell including both analytical and numerical methods was provided by Kant and
Swaminathan [9]. Kant and Manjunatha [10] provided the FE based on HSDT having twelve de-
grees of freedom per node. They presented three-dimensional stress and strain states to investigate
the flexure-membrane coupling behavior of unsymmetrical laminated plate. Akhars and Li [11] de-
veloped a spline finite strip method for static and free vibration analysis of composite plates using
Reddy’s HSDT. Pervez et al [12] developed a two dimensional serendipity FE based on a refined
HSDT having seven degrees of freedom per node to perform the linear static analysis of laminated
orthotropic composite plates. Latheswary et al [13] studied the behavior of laminated composite
plates under static loading by using a four-node nonconforming element based on HSDT. Goswami
[14] presented a simple C^0 FE formulation for nine-node FE with six degrees of freedom based on
HSDT.
From literature review, the previous FE developments using HSDT have mostly depended on
the standard strain definition. From this context, as shown in a recent work [15] laminated compo-
site plate FEs based on HSDT can produce the shear locking phenomenon. However, this problem
has not been paid much attention and uniform or selective reduced integration technique is just
adopted to rectify locking phenomenon although the assumed strain method becomes very popular
for the FE analysis of single-layered isotropic plate structures. So far, there have been a few re-
search works on the free vibration analysis of laminated plate using HSDT together with assumed
strain method [16, 17]. However, we found that there is no introduction of assumed strains in the
formulation of laminated composite plate element based on the HSDT for FE stress analysis, alt-
hough the stress evaluation at the layer and inter-layer of the laminated plate is very crucial.
Therefore, we here propose a new assumed strain laminated composite plate FE based on the
HSDT and provide a series of benchmark tests to prove its capability. More specifically, we provide
the description on the four-node lower order laminate composite plate FE with assumed strains and
its numerical results as benchmark test suites for an application of assumed strains into the lami-
nated composite structures.
The total domain ( Ω ) of laminated plate consists of the mid-surface and the thickness as shown
in Figure 1 and it can be defined as
⎧⎪ ⎡ h h ⎤ ⎫⎪
Ω = ⎪⎨( x1,x 2,x 3 ) ( x1,x 2 ) ∈ Ω0, x 3 ∈ ⎢ − , ⎥ ⎪⎬ (1)
⎪⎪⎩ ⎢⎣ 2 2 ⎥⎦ ⎪⎪⎭
⎛ ∂u ⎞
u1 ( x1,x 2,x 3 ) = u1 ( x1,x 2 ) + x 3θ2 + c1x 33 ⎜⎜ θ2 + 3 ⎟⎟⎟
⎜⎝ ∂x1 ⎟⎠
⎛ ∂u ⎞ (2)
u2 ( x1,x 2,x 3 ) = u2 ( x1,x 2 ) − x 3θ1 + c1x 33 ⎜⎜ −θ1 + 3 ⎟⎟⎟
⎜⎝ ∂x 2 ⎟⎠
u3 ( x1,x 2, 0 ) = u3 ( x1,x 2 ) = u3
The strains in the plate are defined by linear strain-displacement relationship as follows
1 ⎛ ∂u ∂u ⎞⎟ 1
εij = ⎜⎜⎜ j + i ⎟⎟⎟ = ( u j,i + ui,j ) (3)
2 ⎜⎝ ∂xi ∂x j ⎟⎠ 2
rameters c1 and c2 are −4 / 3h 2 and −4 / h 2 respectively and the individual strain terms are
−θ + u
⎪⎬,
⎪
{ γ( ) } = ⎪⎪⎨⎪
2
−θ + u
⎪⎬
⎪
.
⎪⎪⎩ 1 3,2 ⎪⎪⎭ ⎪⎪⎩ 1 3,2 ⎪⎪⎭
In this study, each layer of laminate plate is assumed as orthotropic material and the normal
transverse stress (σ3' ) is assumed to be negligible. Therefore, the constitutive equation for the k th
layer with respect to the material coordinate system x1',x 2',x 3' ( ) can be written as
(k) (k)
{ σ' } = ⎡⎣C ⎤⎦ { ε' }
(k)
in which C ij are the components of rigidity matrix for the k th layer as follows
E1' E 2'
C11 = , C 22 = ,
' ' ' '
1 − ν12ν21 1 − ν12ν21
'
ν21E1' (7)
C12 = = C 21,
' '
1 − ν12ν21
' ' '
C 33 = G12 , C 44 = G13 , C 55 = G23
where E1' , E 2' are the Young’s modulus in x 1' , x 2' direction respectively, G12
'
, G13
'
, G23
'
are the
' '
shear modulus and ν12 , ν21 are the Poisson ratios.
where the transformation matrix ⎡⎣T ⎤⎦ between material coordinate system and global coordinate
system can be written as
⎡ ⎤
⎢ c2 s2 −2sc 0 0 ⎥
⎢ ⎥
⎢ s2 c2 2sc 0 0 ⎥
⎡T ⎤ = ⎢ ⎥ (9)
⎣ ⎦ ⎢ sc −sc c − s 2
2
0 0 ⎥
⎢ 0 0 0 c −s ⎥⎥
⎢
⎢⎣ 0 0 0 s c ⎥⎦
where c = cos ϑ and s = sin ϑ in which ϑ is the fiber’s angle as shown in Figure 1.
The stress-strain relationship in the global coordinate system can be written by using (6) and (8)
as follows
(k) (k)
= ⎡⎣T ⎤⎦ { σ ' } = ⎡⎣T ⎤⎦ ⎡⎣C ⎤⎦ { ε' }
(k) (k)
{σ }
(k) T (k)
= ⎡⎣T ⎤⎦ ⎡⎣C ⎤⎦ ⎡⎣T ⎤⎦ { ε } (10)
(k) (k)
= ⎡⎣Q ⎤⎦ { ε } .
The stress resultants are calculated by integration of the stresses through thickness direction of
laminated plate and five stress resultant terms such as { N }, { M }, { P }, { Q }, { R } can be
obtained as follows
⎧ ⎫
⎧
⎪
⎪ { N } ⎫⎪⎪⎪ ⎡
⎢ ⎡A⎤
⎣ ⎦
⎡ B ⎤ c ⎡ E ⎤ ⎤⎥ ⎪
⎣ ⎦ 1 ⎣ ⎦ ⎪
⎪
⎪ { ε( ) } 0 ⎪
⎪
⎪
⎪
⎪
⎪ ⎪ ⎢ ⎥⎪ ⎪
⎪
⎨
⎪
⎪
{ M } ⎪⎬⎪ =
⎪
⎢
⎢
⎡B ⎤
⎣ ⎦
⎡D ⎤ c ⎡F ⎤ ⎥ ⎪
⎣ ⎦ 1 ⎣ ⎦ ⎥⎨ ⎪ { ε( ) } 1 ⎪
⎬
⎪
⎪ { P } ⎪⎪⎪⎭ ⎢ ⎡E ⎤ ⎡F ⎤ c ⎡H ⎤ ⎥ ⎪ ⎪ ⎪
⎪
⎪
⎪
⎩ ⎢⎣ ⎣ ⎦ ⎣ ⎦ 1 ⎣ ⎦ ⎦⎪⎥ ⎪
⎪
⎪
⎩
{ ε( ) } 3 ⎪
⎪
⎪
⎪
⎭
(12)
⎧
⎧
⎪
⎪
⎪ {Q } ⎫
⎪
⎪
⎪
⎡
⎢ ⎡G ⎤
⎣ ⎦
⎪
c2 ⎡⎣ S ⎤⎦ ⎤⎥ ⎪
⎪
⎪
γ (
{ } ⎪⎫⎪⎪⎪
0)
⎨ ⎬= ⎢ ⎥⎨ ⎬
⎪
⎪
⎪
⎩
{R } ⎪
⎪
⎪
⎭
⎢
⎣
⎡S ⎤
⎣ ⎦ c2 ⎡⎣T ⎤⎦ ⎥ ⎪
⎦⎪⎪
⎪
⎩
γ( ){ } ⎪⎪⎪⎪⎭
2
Where the components of the rigidity matrices Aij , Bij , Dij , Eij , Fij , H ij , Gij , Sij , Tij can be
written as follows
k+1
nlayer x 3
( Aij ,Bij ,Dij ,Eij ,Fij ,H ij ) = ∑ ∫
k=1
⎢⎣ ij ⎥⎦ (
⎡Q k ⎤ 1,x , ( x )2 , ( x )3 , ( x )4 , ( x )6 dx ,
3 3 3 3 3 3 )
x 3k
k+1 (13)
nlayer x 3
(Gij ,Sij ,Tij ) = ∑ ∫
k=1
⎣⎢ ij ⎦⎥ (
⎡Q k ⎤ 1, ( x ) , ( x ) dx
3 3 3
2 4
)
x 3k
and the above equation can be explicitly rewritten in the following form:
nlayer
Aij = ∑ Qijk ( x 3k+1 − x 3k ) = Gij ,
k=1
∑ Qijk {( x 3k+1 ) − ( x 3k )
},
nlayer
1 2 2
Bij =
2 k=1
D = ∑ Q {( x ) − ( x ) } = S ,
nlayer
1 k k+1
3
k
3
ij ij 3 3 ij
3 k=1
(14)
E = ∑ Q {( x ) − ( x ) }
nlayer
1 k k+1
4
k
4
ij , ij 3 3
4 k=1
F = ∑ Q {( x ) − ( x ) } = T ,
nlayer
1 k k+1
5
k
5
ij ij 3 3 ij
5 k=1
H = ∑ Q {( x ) − ( x ) } .
nlayer
1 k k+1
7
k
7
ij ij 3 3
7 k=1
In this study, a four-node plate element having seven degrees of freedom per node is formulated
using the isoparametric formulation. Therefore the geometry and displacement fields of the pre-
sent FE can be defined in the following form:
4
xi = ∑Naxi
a
( i = 1, 3 ) ,
a=1
4
(15)
ui = ∑ N auia ( i = 1,7 )
a=1
where N a is the bilinear Lagrange shape function associated with node a , xi is the position vec-
tor of the plate and the nodal displacement vector uia = u a { } has seven components such as
⎧
⎪ ∂u a ∂u a ⎫
⎪
{ua } = ⎪⎨⎪u1a ,u2a ,u3a , θ1a , θ2a , ∂x3 , ∂x3 ⎪⎬⎪ (16)
⎪
⎩ 1 2 ⎪
⎭
where ∂u3a / ∂x1 , ∂u3a / ∂x 2 are two additional degrees of freedom related to the higher order
terms of (2) which do not appear in the FE based on the FSDT.
Using (15), the strains of (4) can be rewritten in the form of the strain-displacement relation ma-
trix as follows
4
∑ ( ⎡⎢⎣ B0a ⎤⎥⎦ + ⎡⎢⎣ B1a ⎤⎥⎦ + c1 ( x 3 ) ){u }
3
{ εp } = ⎡ Ba ⎤ a
⎢⎣ 2 ⎥⎦
a=1
4
(17)
{ γs } = ∑ ( ⎡ B a ⎤ + c ( x )2 ⎡ B a ⎤
⎣⎢ 3 ⎦⎥ 2 3 ⎣⎢ 4 ⎦⎥ ){u } a
a=1
⎡ ∂N a ⎤ ⎡ ∂N a ⎤
⎢ 00 0 0 0 0 ⎥⎥ ⎢ 0 0 0 0 0 0 ⎥⎥
⎢ ∂x1 ⎢ ∂x1
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎡ B a ⎤ = ⎢⎢ ∂N a ⎥ ⎢ ∂N a ⎥
⎢⎣ 0 ⎥⎦ ⎢ 0 0 0 0 0 0 ⎥ , ⎡⎢ B1a ⎤⎥ = ⎢ 0 0 0 − 0 0 0 ⎥,
∂x 2 ⎥ ⎣ ⎦ ⎢ ∂x 2 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ∂N a ∂N a ⎥ ⎢ ∂N a ∂N a ⎥
⎢ 0 0 0 0 0 ⎥⎥ ⎢ 0 0 0 − 0 0 ⎥⎥
⎢ ∂x 2 ∂x1 ⎢ ∂x1 ∂x 2
⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦
⎡ ∂N a ∂N a ⎤
⎢ 0 0 0 0 0 ⎥
⎢ ∂x1 ∂x1 ⎥
⎢ ⎥
⎢ ⎥
⎡ B a ⎤ = ⎢⎢ 0 0 0 − ∂N a 0 0
∂N a ⎥
⎥,
⎢⎣ 2 ⎥⎦ ⎢ ∂x 2 ∂x 2 ⎥
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 − ∂N a ∂N a ∂N a ∂N a ⎥
⎢ ∂x1 ∂x 2 ∂x 2 ∂x1 ⎥
⎢⎣ ⎥⎦
⎡ ⎤
⎢ 0 0 ∂N a 0 N 0 0 ⎥
⎢ ∂x a ⎥ ⎡ ⎤
⎡ B a ⎤ = ⎢⎢ 1 ⎥ ⎡ a⎤ ⎢ 0 0 0 0 Na Na 0 ⎥.
⎢⎣ 3 ⎥⎦ , B
⎥ ⎢⎣ 4 ⎥⎦ ⎢= ⎥
⎢ ∂N a ⎥ ⎢⎣ 0 0 0 −N a 0 0 Na ⎥⎦
⎢ 0 0 −N a 0 0 0 ⎥
⎢ ∂x 2 ⎥
⎣ ⎦
The total potential energy of the laminated plate can be written by using the stress resultants
and the corresponding strains as follows
Π=
1
2∫
dA
( {N } { ε( ) } + {M } { ε( ) } + c {P } { ε( ) } + {Q } { γ( ) } + c {R } { γ( ) })dA − P
T 0 T 1
1
T 3 T 0
2
T 2
(18)
1 T T
Π= {u } ( ⎡⎣ KN ⎤⎦ + ⎡⎣ KM ⎤⎦ + ⎡⎣ KP ⎤⎦ + ⎡⎣⎢ KQ ⎤⎦⎥ + ⎡⎣ KR ⎤⎦ ){u } − {u } { f } (19)
2
By minimization of the total potential energy, with respect to the nodal values u we obtain
⎡ K ⎤ {u } = { f } (20)
⎣ ⎦
where {u } is global vector of nodal displacements with typical nodal displacement sub-vector for
e=1
⎡ K ⎤ = ∧ ⎡ K (e) ⎤ (20)
⎣ ⎦ nel ⎢⎣ ⎦⎥
e=1
where ∧ [] is the finite element assembly operator and the element stiffness matrix is divided into
nel
five contributions
⎡ K (e ) ⎤ = ⎡ K (e ) ⎤ + ⎡ K (e ) ⎤ + ⎡ K (e ) ⎤ + ⎡ K (e ) ⎤ + ⎡ K (e ) ⎤
⎢⎣ ⎥⎦ ⎢⎣ N ⎥⎦ ⎢⎣ M ⎥⎦ ⎢⎣ P ⎥⎦ ⎢⎣ Q ⎥⎦ ⎢⎣ R ⎥⎦ (22)
in which ⎡⎢ K N(e) ⎤⎥ , ⎡⎢ K M
(e) ⎤ ⎡ (e) ⎤ ⎡ (e) ⎤
, K , K and ⎡⎢ K R(e) ⎤⎥ are
⎣ ⎦ ⎣ ⎦⎥ ⎣⎢ P ⎦⎥ ⎣⎢ Q ⎦⎥ ⎣ ⎦
⎡ K (e) ⎤ = ⎛ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎡ B a ⎤T ⎡ E ⎤ ⎡ Bb ⎤ ⎞⎟⎟dA ,
⎜⎜ B0
⎢⎣ N ⎥⎦ ∫ ⎢ ⎥ A ⎢ B ⎥ + ⎢ B
⎜⎝ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
0 1 ⎥ B ⎢ B0 ⎥ + c1 ⎢⎣ 2 ⎥⎦ ⎣ ⎦ ⎢⎣ 0 ⎥⎦ ⎟⎠
(e )
dA
⎡ K (e) ⎤ = ⎛ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎞⎟
⎜⎜ B0
⎢⎣ M ⎥⎦ ∫ ⎜⎝ ⎢⎣ ⎥⎦ ⎣ B ⎦ ⎢⎣ B1 ⎥⎦ + ⎢⎣ B1 ⎥⎦ ⎣ D ⎦ ⎢⎣ B1 ⎥⎦ + c1 ⎢⎣ B2 ⎥⎦ ⎣ F ⎦ ⎢⎣ B1 ⎥⎦ ⎟⎟⎠dA ,
(e )
dA
⎛ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ ⎤ ⎡ b ⎤ ⎞⎟
T
⎡ K (e) ⎤ = ⎜⎜c1 B0 2⎡ a⎤ ⎡
⎣⎢ P ⎦⎥ ∫ ⎜⎝ ⎣⎢ ⎦⎥ ⎣ E ⎦ ⎣⎢ B2 ⎦⎥ + c1 ⎣⎢ B1 ⎦⎥ ⎣ F ⎦ ⎣⎢ B2 ⎦⎥ + c1 ⎣⎢ B2 ⎦⎥ ⎣ H ⎦ ⎣⎢ B2 ⎦⎥ ⎟⎟⎠dA , (23)
(e )
dA
⎡ K (e) ⎤ = ⎛ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ T ⎞
⎜⎜ B3 G B3 + c2 ⎡ B4a ⎤ ⎡ S ⎤ ⎡ B3b ⎤ ⎟⎟dA ,
⎣⎢ Q ⎦⎥ ∫ ⎜⎝ ⎣⎢ ⎦⎥ ⎣ ⎦ ⎣⎢ ⎦⎥ ⎣⎢ ⎦⎥ ⎣ ⎦ ⎣⎢ ⎦⎥ ⎟⎠
(e )
dA
⎡ K (e) ⎤ = ⎛ ⎡ a ⎤T ⎡ ⎤ ⎡ b ⎤ 2 ⎡ a ⎤ ⎡ ⎤ ⎡ b ⎤⎞
T
⎟
⎜⎜c2 B3
⎣⎢ R ⎦⎥ ∫ ⎜⎝ ⎣⎢ ⎦⎥ ⎣ S ⎦ ⎣⎢ B4 ⎦⎥ + c2 ⎣⎢ B4 ⎦⎥ ⎣T ⎦ ⎣⎢ B4 ⎦⎥ ⎟⎟⎠dA .
(e )
dA
In this study, the assumed strain method is employed to alleviate the possible locking phenome-
non. Therefore, assumed strains are derived by using the interpolation functions based on La-
grangian polynomial and the strain values at the sampling points where the locking does not ex-
ist.
For natural assumed transverse shear strains γ( )( ) and γ( )( ) , the following sampling
0 A 0 A
13 23
points [18] are employed as shown in Figure 2:
Using (24), the assumed natural strains can be defined in the following form:
2 2
( 0 )( A ) = ( 0 )( A ) =
∑ ( ) ∑Qδ ( ξ1 ) γ23
δ δ
γ13 Pδ ξ2 γ13 , γ23 (25)
a=1 a=1
where δ denotes the position of the sampling point as shown in Figure 2 and the interpolation
functions P , Q are employed as follows
1 1
P1 = (1 + ξ2 ), P2 = (1 − ξ2 ),
2 2
(26)
1 1
Q1 = ( 1 + ξ1 ), Q2 = ( 1 − ξ1 ).
2 2
4
{ γs } = ∑ ( ⎡⎢⎣ B3a( A ) ⎤⎥⎦ + c2 ( x 3 ) ){u }
2
⎡ Ba ⎤ a
(27)
⎢⎣ 4 ⎥⎦
a=1
where B3a(A) is the assumed natural strain-displacement relationship matrix and explicitly can be
written as
1 a Ξ 1 1 a
Ξ1 =
4
( )
ξ1 1 + ξ2a ξ2 , Ξ2 = 1 = 1 + ξ2a ξ2 , Ξ3 =
a
ξ1 4
( ξ + ξ2 ,
4 2
) ( )
1 a Ξ 1 1 a
Ξ4 =
4
( )
ξ2 1 + ξ1a ξ1 ,Ξ5 = 1 = 1 + ξ1a ξ1 , Ξ6 =
a
ξ2 4
(ξ + ξ1 ,
4 1
) ( ) (29)
1 1 1 1
αi =
4
(
−x1i + x 2i + x 3i − x 4i , β ) i =
4
( )
x i − x 2i + x 3i − x 4i , γ i = (−x1i − x 2i + x 3i + x 4i )
4
where x ai , ξia are the coordinates of nodal point a in global coordinate system and natural coor-
dinate system respectively.
In previous section, the assumed strains are defined in natural coordinate system and so trans-
verse shear rigidity matrix of (13) can be also defined in the natural coordinate system as follows
∂ξi ∂ξj
Gij = G ,
∂x α ∂x β αβ
(30)
∂ξi ∂ξj
SijQ = S αj , SijS = S .
∂x α ∂x β iβ
⎛ ⎡ ( A ) ⎤T ⎞
⎡ K (e) ⎤ = ⎜⎜ B ⎡G ⎤ ⎡ B ( A ) ⎤ + c ⎡ B ⎤T ⎡ SQ ⎤ ⎡⎢ B ( A ) ⎤⎥ ⎟⎟dA
⎣⎢ Q ⎦⎥ ∫ ⎜⎝ ⎢⎣ 3 ⎥⎦ ⎣ ⎦ ⎢⎣ 3 ⎥⎦ 2 ⎣ 4 ⎦ ⎣⎢ ⎥⎦ ⎣ 3 ⎦ ⎟⎟⎠ ,
dA
(31)
⎛ ⎡ ( A ) ⎤T ⎞
⎡ K (e) ⎤ = ⎜⎜c B ⎡ SS ⎤ ⎡ B ⎤ + c 2 ⎡ B ⎤T ⎡T ⎤ ⎡ B ⎤ ⎟⎟dA.
⎢⎣ R ⎥⎦ ∫ ⎢ ⎥
⎝⎜ 2 ⎣ 3 ⎦ ⎢⎣ ⎥⎦ ⎣ 4 ⎦ 2 ⎣ 4 ⎦ ⎣ ⎦ ⎣ 4 ⎦⎟ ⎟⎠
dA
In this study, (31) will be used instead of the terms ⎡⎢ KQ(e) ⎤⎥ and ⎡ K (e) ⎤ in (23). Note that the
⎢⎣ R ⎥⎦
⎣ ⎦
standard FE (HAD4) is derived by using the original terms in (23).
In order to investigate the accuracy and reliability of the newly developed laminated plate ele-
ment HSA4, a series of numerical test for symmetric and unsymmetric laminated plates are con-
sidered. The square plates with the simply supported boundary conditions are used in the test.
More specifically, two sets of boundary conditions, SS 1 and SS 2 are employed in numerical tests
as follows
∂u3
SS1 : u1 = 0, u3 = 0, θ2 = 0, = 0 at x 2 = 0, x 2 = a ,
∂x1
∂u3
u2 = 0, u3 = 0, θ1 = 0, = 0 at x1 = 0, x1 = a ,
∂x 2
(32)
∂u
SS2 : u2 = 0, u3 = 0, θ2 = 0, 3 = 0 at x 2 = 0, x 2 = a ,
∂x1
∂u3
u1 = 0, u3 = 0, θ1 = 0, = 0 at x1 = 0, x1 = a
∂x 2
In this study, the laminated plates are also assumed to be subjected to sinusoidal load or uni-
form load as follows
πx1 πx
L1 = q ( x1,x 2 ) = q 0 sin sin 2
a a (34)
L2 = q ( x1,x 2 ) = 1 ( uniform load ) .
All the numerical results provided in Tables and Figures are described as non-dimensionalized
values by using the following form
⎛ E h 3 ⎞⎟ ⎛ h 2 ⎞⎟ ⎛ 2 ⎞
w = u3 ⎜⎜⎜ 100 × 2 ⎟⎟, σ1 = σ1 ⎜⎜⎜ ⎟⎟, σ2 = σ2 ⎜⎜ h ⎟⎟⎟,
⎜⎝ qa 4 ⎟⎟⎠ ⎜⎝ qa 2 ⎟⎠ ⎜⎜ qa 2 ⎟
⎝ ⎠
(35)
⎛ h ⎞⎟
2 ⎛ ⎞ ⎛ ⎞
τ12 = τ12 ⎜⎜⎜ ⎟⎟, τ13 = τ13 ⎜⎜ h ⎟⎟⎟, τ 23 = τ 23 ⎜⎜ h ⎟⎟⎟.
⎜⎝ qa ⎟⎠
2 ⎜
⎝ qa ⎟⎠
2 ⎜
⎝ qa ⎟⎠
2
⎛a a ⎞ ⎛a a h ⎞ ⎛a a h ⎞
u3 : ⎜⎜ , , 0 ⎟⎟⎟, σ1 : ⎜⎜ , ,± ⎟⎟⎟, σ2 : ⎜⎜ , ,± ⎟⎟⎟,
⎜⎝ 2 2 ⎟⎠ ⎜⎝ 2 2 2 ⎟⎠ ⎜⎝ 2 2 2 ⎟⎠
⎛ (36)
h⎞ ⎛ a ⎞ ⎛a ⎞
τ12 : ⎜⎜ 0, 0,± ⎟⎟⎟, τ13 : ⎜⎜ 0, , 0 ⎟⎟⎟, τ 23 : ⎜⎜ , 0, 0 ⎟⎟⎟
⎜⎝ ⎟
2⎠ ⎜
⎝ 2 ⎠ ⎟ ⎜
⎝2 ⎟⎠
where a is the width of the plate and h is the thickness of the plate.
It should be noted that Reddy’s displacement function definition of (2) can reserve zero shear
stress boundary condition at the top and bottom of the laminated plate but this condition cannot
be reserved anymore when assumed strain method is adopted with the HSDT. Therefore, the
following modified strain definition is used when we calculate transverse shear stress value to-
gether with (10):
4
{ γs } = ∑ ( ⎡⎢⎣ B3a( A ) ⎤⎥⎦ + c2 ( x 3 ) ⎡ B a( A ) ⎤
){u }, where c
2 a
⎢⎣ 3 ⎥⎦ 2 = 4 / h 2. (37)
a=1
With (37), we can enforce the values of γs to zero at top ( x 3 = h / 2 ) and bottom ( x 3 = −h / 2 )
of laminated plates and eventually the zero shear stress values can be achieved at the top and
bottom of the plates.
(a) Convergence rate test: A four-layer symmetric cross-ply ( 0° / 90° / 90° / 0° ) laminated com-
posite plate is analyzed to check the convergence rate of the present FE (HSA4) and its perfor-
mance for the variation of thickness values. In this test, material property, boundary condition
and load case such as M1, SS1, L1 are used. Five cases with 4 × 4 , 8 × 8 , 12 ×12 , 16 ×16 and
32 × 32 FE meshes are employed in the test. Two aspect ratios a / h = 10 , 100 are used in FE
analysis to show the applicability of the present plate element to both thin and thick plates. The
center deflection and stresses are calculated for both thick and thin plates and summarized in
Tables 1 and 2. The present solutions are compared with the exact elasticity solution [19] and a
FE solution [14]. The convergence test result is plotted in Figure 3.
Figure 3 The convergence test results of the present FE (HSA4) for a / h = 10 and 100.
The present plate element shows a good convergence rate. In particular, we can see that the
present FE solution has a good agreement with the elasticity solution [19] and achieves a certain
level of convergence with the case of 16 ×16 FE mesh for both thin and thick plates. We there-
fore decided to use the 16 ×16 FE mesh with 256 four-node elements for numerical test through-
out this study. Note that FE reference solution [14] was produced by a nine-node plate element so
that the reference solution with n × n FE mesh is compared to the present solution with the
2n × 2n FE mesh.
Table 2 Convergence test of a simply supported ( SS1 ) symmetric cross-ply ( 0° / 90° / 90° / 0° ) laminated plate under sinusoidal
transverse load ( a / h = 100 ).
(b) Locking test: The same plate is also used to investigate the possible appearance of locking
phenomenon again. For this purpose, we use four different aspect ratios such as
a / h = 4, 10, 20, 100 with the 16 ×16 FE mesh. Numerical results such as central deflection and
Table 3 The non-dimensionalized deflection and stresses of a simply supported ( SS1 ) symmetric cross-ply
( 0° / 90° / 90° / 0° ) laminated plate under sinusoidal transverse load.
Figure 4 Variation of the in-plane stress ( σ1 ) through the thickness direction ( z / h ) of a simply supported ( SS1 ) cross-ply
( 0° / 90° / 90° / 0° ) laminated composite plate under sinusoidal transverse load ( a / h = 4 and 100)
Figure 5 Variation of the in-plane stress ( σ2 ) through the thickness direction ( z / h ) of a simply supported ( SS 1 ) cross-ply
( 0° / 90° / 90° / 0° ) laminated composite plate under sinusoidal transverse load ( a / h = 4 and 100)
Figure 6 Variation of the transverse shear stress ( τ 13 ) through the thickness direction ( z / h ) of a simply supported ( SS 1 )
cross-ply ( 0° / 90° / 90° / 0° ) laminated composite plate under sinusoidal transverse load ( a / h = 4 and 100)
Figure 7 Variation of the transverse shear stress ( τ 23 ) through the thickness direction ( z / h ) of a simply supported ( SS1 )
cross-ply ( 0° / 90° / 90° / 0° ) laminated composite plate under sinusoidal transverse load ( a / h = 4 and 100)
This example is used to investigate the performance of the present FE HSA4 on the analysis of
unsymmetric composite plate. A two-layer unsymmetric cross-ply ( 0 / 90º ) laminated plate is
analyzed in the test. The material property, boundary condition and load case such as
M1, SS1, L1 are used. The central deflection and stresses are calculated for four aspect ratios
a / h = 4, 10, 20, 100 and summarized in Table 4.
Table 4 The non-dimensionalized deflection and stresses of a simply supported ( SS1 ) unsymmetric cross-ply ( 0 / 90º ) laminated
plate under sinusoidal transverse load.
a /h Theory w σ1 σ2 τ12
Present(HSA4) 1.6836 -0.7847 0.7847 -0.0540
Elasticity[19] 1.7287 -0.7723 0.8036 -0.0586
HSDT *[9] 1.6800 -0.7510 0.7720 -0.0557
5
HSDT **[9] 1.7037 -0.7662 0.7662 -0.0572
Reddy[1] 1.6760 -0.8385 0.8385 -0.0558
Latheswary et al [13] - - - -
Present(HSA4) 1.2168 -0.7228 0.7228 -0.0525
Elasticity[19] 1.2318 -0.7317 0.7353 -0.0540
HSDT *[9] 1.2192 -0.7269 0.7273 -0.0533
10
HSDT **[9] 1.2274 -0.7286 0.7286 -0.0539
Reddy[1] 1.2161 -0.7468 0.7468 -0.0533
Latheswary et al [13] 1.2161 0.7517 0.7517 0.0532
Present(HAS4) 1.0995 -0.7136 0.7136 -0.0521
Elasticity[19] 1.1060 -0.7200 0.7206 -0.0529
HSDT *[9] 1.1025 -0.7189 0.7186 -0.0527
20
HSDT **[9] 1.1078 -0.7185 0.7185 -0.0530
Reddy[1] 1.1018 -0.7235 0.7235 -0.0527
Latheswary et al [13] - - - -
Present(HSA4) 1.0619 -0.7087 0.7087 -0.0520
Elasticity[19] 1.0742 -0.7219 0.7219 -0.0529
HSDT*[9] 1.0651 -0.7161 0.7161 -0.0525
100
HSDT**[9] 1.0695 -0.7152 0.7152 -0.0527
Reddy[1] 1.0651 -0.7161 0.7161 -0.0525
Latheswary et al [13] 1.0647 0.7206 0.7206 0.0524
The numerical results are compared with the exact elasticity solution [19], analytical solutions [2,
8] and the FE solution [20]. In this example, we found that the present solutions have good
agreements with the reference solutions regardless of aspect ratios. In specific, stress values are
reasonably accurate for all aspect ratios a / h = 4, 10, 20, 100 .
(a) Uniform load ( L2 ): The unsymmetric angle-ply ( ⎡⎣ 45° / −45° ⎤⎦ ) laminated plate are analyzed
n
with four aspect ratios a / h = 4, 10, 20, 100 with three different numbers of layers such as n=2,
4, 8. Material properties, boundary condition and load case are used as M 2, SS2, L2 respectively.
The numerical results are non-dimensionalized and summarized in Table 5. From numerical re-
sults, both the present HSA4 and the standard HSD4 have good agreements with reference solu-
tion [9] for the aspect ratio a / h = 4 . However, for the aspect ratio a / h = 100 the standard FE
HSD4 shows a great discrepancy with the reference solution although it exhibit reasonably good
performance up to a / h = 20 . However, the present FE HSA4 does not show any shear locking
phenomenon for thin angle-ply unsymmetric laminated plate. In this section, we also investigate
the effect of different of layer number on the central deflection with the assumption that plates
keep the same total thickness value. From numerical result, when we double the number of layers
from n=2 to n=4, the center deflection is reduced up to 60% in case of the aspect ratio
a / h = 100 and 25 % for aspect ratio a / h = 4 respectively. In other words, thick laminated
plate is very sensitive to variation of the number of layers than thin laminated plate. We also
found that the increase of the number of layers in laminated plate generally tends to reduce the
deflection of laminated plate when the plate has the same thickness value.
Table 5 The non-dimensionalized deflection of a simply supported ( SS1 ) unsymmetric angle-ply ( ⎡⎣ 45° / -45° ⎤⎦ ) laminated plate
4
(b) Sinusoidal load ( L1 ): In this example, the analysis of unsymmetric angle-ply ( ⎡⎣ 45° / −45° ⎤⎦ )
4
laminated plate is carried out with two aspect ratios a / h = 10, 100 and sinusoidal load. This
example provides the results of more detailed investigation on the performance of the present FE
HSA4 for unsymmetric angle-ply plate. Material properties, boundary condition and load case are
used as M1,SS2,L1 respectively. The non-dimensionalized deflection and stresses are summarized
in Table 6. From numerical results, the HSA4 has excellent agreements with the reference solu-
tions [10, 13]. However, the standard FE HSD4 has the error of 35% compared to the reference
solutions for aspect ratio a / h = 100 . Figures 8 and 9 show the variation of non-dimensionalized
in-plane stress σ1 and transverse shear stress τ13 through the thickness direction for the aspect
ratios a / h = 10 and 100 . In particular, the present element HSA4 can produce a good parabolic
shape than that of HSD4 for the transverse shear stress τ13 as shown in Figures 8 and 9. The
present FE shows also a good performance to predict the maximum stress value of in-plane stress
σ1 and transverse shear stress τ13 .
Table 6 The non-dimensionalized deflection of a simply supported ( SS2 ) unsymmetric angle-ply ( ⎡⎣ 45° / -45° ⎤⎦ ) laminated plate
4
under sinusoidal transverse load.
Kant and Pandya [20] 0.4193 0.1633 0.1633 0.1601 0.2347 0.2347
Kant and Pandya [20] 0.2469 0.1462 0.1462 0.1430 0.2344 0.2344
Figure 8 Variation of the in-plane stress ( σ1 ) through the thickness ( z / h ) of a simply supported 𝑺𝑺𝟐 unsymmetric angle-ply
( ⎡⎣ 45° / -45° ⎤⎦ ) laminated composite plate under sinusoidal transverse load ( a / h = 10 and 100 )
4
Figure 9 Variation of transverse shear stress ( τ 13 ) through the thickness ( z / h ) of a simply supported 𝑺𝑺𝟐 unsymmetric angle-
ply ( ⎡⎣ 45° / -45° ⎤⎦ ) laminated plate under sinusoidal transverse load ( a / h = 10 and 100 )
4
The analysis of a five-layer unsymmetric and unbalanced sandwich plate with isotropic core is
carried out and its numerical results are presented. Material properties, boundary condition and
load case are used as M 3,SS1,L1 respectively. The thickness of core is equal to total thickness of
four face sheets which have all the same thickness. The non-dimensionalized deflection and in-
plane stresses are summarized in Table 7. From numerical results, both HSD4 and HSA4 have a
good agreements with the reference solutions up to the aspect ratio a / h = 10 . For the aspect
ratio a / h = 100 , the present element HSA4 have approximately the 1-8 % of differences with
reference solutions [1, 21] but the standard HSD4 have almost the 60 % of difference with the
reference solutions because of locking phenomenon.
Table 7 The non-dimensionalized deflection and stresses of a simply supported 𝑺𝑺𝟏 unsymmetric cross-ply
( 0° / 90° / core / 0° / 90° ) sandwich plate under sinusoidal transverse load.
a /h Theory w σ1 σ2 τ12
Present(HSA4) 12.9278 -1.4673 1.4673 0.1970
HSD4 13.2060 -1.4773 1.4773 0.1992
4 HSDT*[9] 14.1627 -1.6445 1.4931 0.2031
HSDT**[9] 14.3440 -1.5328 1.5328 0.2196
Reddy[1] 8.7941 -0.9937 0.9937 0.1291
Present(HSA4) 3.0264 -0.7504 0.7504 0.0874
HSD4 3.0260 -0.7310 0.7310 0.0855
10 HSDT*[9] 3.3032 -0.8140 0.7606 0.0946
HSDT**[9] 3.3197 -0.7771 0.7771 0.0951
Reddy[1] 2.3075 -0.6815 0.6815 0.0787
Present(HSA4) 1.0105 -0.6029 0.6029 0.0649
HSD4 0.4306 -0.2558 0.2558 0.0275
100 HSDT*[9] 1.0697 -0.6231 0.6226 0.0691
HSDT**[9] 1.0763 -0.6216 0.6216 0.0696
Reddy[1] 1.0595 -0.6214 0.6214 0.0690
5 CONCLUSIONS
A four-node laminated composite plate element having seven degrees freedom per node is newly
developed by using the HSDT and assumed strains to perform the FE stress analysis. The accu-
racy and reliability of new laminated composite plate element is thoroughly tested by using five
numerical tests for both symmetric and unsymmetric situations. From numerical results, the pre-
sent FE does not produce any locking phenomenon for the cases with very small aspect ratio of
the plate and it can be applicable to most types of laminated composite plate structures. Further
investigation into the performance of the present FE for the case subjected to concentrated or line
loads is underway and the vibration, stability and transient response of laminated plates are also
of prime importance as future investigations.
References
[1] Reddy JN. A simple Higher order Theory for laminated composite plates. ASME J Appl Mech 1984;51:745-
752.
[2] Bert CW. A critical evaluation of new plate theories applied to laminated composites. Compos Struct
1984;2:329-347.
[3] Lo KH, Christensen RM, Wu EM. A high-order theory of plate deformation-part 2: laminated plates. Appl
Mech 1977;44:669-676.
[4] Reddy JN. A review of the literature on finite element analysis of progressive failure in laminated composite
plates. Shock Vib Digest 1985;17(4):3-8.
[5] Reddy JN. Mechanics of laminated composite plate. CRC Press, 1997.
[6] Zhang YX, Yang CH. Recent developments in finite element analysis for laminated composite plates. Com-
pos Struct 2009;88:147-157.
[7] Bose P, Reddy JN. Analysis of composite plates using various plate theories. Part 1: Formulation and ana-
lytical solutions. Struct Engng Mech 1998;6(6):583-612.
[8] Bose P, Reddy JN. Analysis of composite plates using various plate theories. Part 2: Finite element model
and numerical results. Struct Engng Mech 1998;6(7):727-746.
[9] Kant T, Swaminathan K. Estimation of transverse/interlaminar stresses in laminated composites-a selective
review and survey of current developments. Compos Struct 2000;49:65-75.
0
[10] Kant T, Manjunatha BS. An unsummetric FRClaminate C finite element model with 12 degrees of freedom
per node. Eng Comput 1988;5:300-308.
[11] Akhras G, Li W. Static and free vibration analysis of composite plates using spline finite strips with higher
order shear deformation. Compos Part B: Eng 2005;36:496-503.
[12] Pervez T, Seibi AC, Al-Jahwari FKS. Analysis of thick orthotropic laminated composite plates based on
higher order shear deformation theory. Compos Struct 2005;71:414-422.
[13] Latheswary S, Valasrajan KV, Rao YVKS. Behavior of laminated composite plates using higher order shear
deformation theory. IE(I) J-AS 2004;85:10-17.
0
[14] Goswami S. A C plate bending element with refined shear deformation theory for composite structures.
Compos Struct 2006;72:375-382.
[15] Lee SJ, Kim HR. Finite element analysis of symmetric and unsymmetric laminated plates based on higher
order shear deformation theory. In: Proceedings of Autumn Congress of Architectural Institute of Korea,
Cheongju, 26-27 Oct 2007.p. 249-252.
[16] Nayak AK, Moy SSJ, Shenoi RA. Free vibration analysis of composite sandwich plates based on Reddy’s
higher –order theory. Compos Part B: Eng 2002;33:505-519.
[17] Nayak AK, Shenoi RA. Assumed strain finite elements for buckling and vibration analysis of initially
stressed damped composite sandwich plates. Sandwi Struct and Mater 2005;7:307-334.
[18] Lee SJ. Free-vibration analysis of plates by using a four node finite element formulated with assumed natu-
ral transverse shear strain. J Sound Vibration 2004;278:657-684.
[19] Pango NJ. Exact solutions for rectangular bidirectional composites and sandwich plates. J Compos Mater
1970;4:20-34.
[20] Kant T, Pandya BN. A simple finite element formulation of a higher order theory of unsymmetrically lami-
nated composite plates. Compos Struct 1988;9:215-246.
[21] Kant T, Swaminathan K. Analytical solutions for the static analysis of laminated composite and sandwich
plates based on higher order refined theory. Compos Struct 2002;56:329-344.