Upload
km-liew
View
219
Download
0
Embed Size (px)
Citation preview
Nonlinear analysis of laminated composite plates usingthe mesh-free kp-Ritz method based on FSDT
K.M. Liew a,b, J. Wang a, M.J. Tan b, S. Rajendran b,*
a Nanyang Centre for Supercomputing and Visualisation, Nanyang Technological University, Nanyang Avenue,
Singapore 639798, Singaporeb School of Mechanical and Production Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798, Singapore
Received 4 September 2003; received in revised form 15 December 2003; accepted 23 March 2004
Abstract
A mesh-free kp-Ritz method of solution based on the kernel particle approximation for the field variables is pro-
posed for the large deflection flexural analysis of laminated composite plates. The first-order shear deformation theory
(FSDT) is used for modeling the flexure. The nonlinear solution algorithm is based on the total Lagrangian formulation
with Green’s strain measures and von Karman assumptions. The incremental form of nonlinear equations is obtained
by Taylor series expansion, and Newton’s method is used to solve these equations. Test problems involving square and
rectangular composite plates with SSSS and CCCC boundary conditions are solved to assess the efficacy of the pro-
posed method. The results are in excellent agreement with the series solution as well as the finite element solution
already reported in the literature.
2004 Published by Elsevier B.V.
Keywords: kp-Ritz method; Nonlinear bending; Laminated composite plates; FSDT; Newton methods; Arc length
1. Introduction
The nonlinear analysis of plates treats the stiffnesses and/or the loads as dependent upon the displace-
ments. Nonlinear problems can be classified as material nonlinear and geometric nonlinear problems [1].
The material nonlinearity is associated with the changes in material properties, e.g. elastoplastic problems.
The geometric nonlinearity is associated with significant changes in configuration (geometry). This paper
deals with only the geometric nonlinearity of composite laminated plates.Linear analyses may lead to inaccurate and inadmissible results under certain conditions and loadings.
In linear analysis of composite plates, the displacements and rotations are assumed to be small, and the
material property is regarded as linearly elastic. In the nonlinear analysis of laminated composite plates
*Corresponding author. Address: School of Mechanical and Production Engineering, Nanyang Technological University, Nanyang
Avenue, Singapore 639798, Singapore. Tel.: +65-6790-5865; fax: +65-6791-1859.
E-mail address: [email protected] (S. Rajendran).
0045-7825/$ - see front matter 2004 Published by Elsevier B.V.
doi:10.1016/j.cma.2004.03.013
Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
www.elsevier.com/locate/cma
based on a laminate theory with the von Karman strains, the geometry of the plates is assumed to beunchanged during the loading process. For problems with small strains and moderately large deflection and
rotations, these assumptions lead to very good results [2]. The first-order shear deformation plate theory is
used in this paper. In continuum mechanics, there are two incremental formulations to describe the
deformation and stress fields, i.e., Total Lagrangian (TL) formulation and Updated Lagrangian (UL)
formulation. In this paper, TL formulation is employed to describe the nonlinear behaviour of plates under
large deformations. Thus, all variables are referred to a fixed configuration, and changes in displacement
and stress fields are determined with respect to the fixed (reference) configuration.
Presently, the finite element method continues to dominate the numerical analyses of geometricalnonlinear problems. Many researchers have used it to solve large deformation problems of plates [3–7]. For
geometrically nonlinear problems, the incremental formulation is widely used. In the present study, the
discretization of composite laminated plates for nonlinear analysis is formulated in incremental form. The
formulations are derived by Ritz method with kernel particle discretization.
The numerical analysis of nonlinear problems leads to nonlinear algebraic equations. To solve these
equations, several methods have been developed. Among them, the Newton–Raphson method, modified
Newton–Raphson method and quasi-Newton method are widely used. Clarke and Hancock [7] have re-
viewed incremental-iterative solution techniques for geometrically nonlinear analyses. In this paper, themodified Newton–Raphson method is combined with the arc-length technique continuation to solve the
nonlinear equations.
In engineering analysis of nonlinear problems, the finite element method is still the most commonly used
method. Nevertheless, though relatively new, a class of particle methods or meshless methods has proven to
be a potentially attractive alternate numerical technique to solve the nonlinear problems especially for those
with large deformations [8–12]. The present work is an extension of our previous efforts to study the
laminated composite plates using the kernel particle Ritz method. The kernel particle Ritz (kp-Ritz)
method belongs to the broad class of meshless methods, and attempts a FSDT Ritz solution based onkernel particle approximations. In our previous works, the bending, free vibration and buckling analysis of
composite plates have been studied [13–15]. Free vibration analysis of laminated composite panels has also
been carried out [16–18]. In this paper, the nonlinear deformation of laminated composite plates is studied
using the kp-Ritz method.
2. Formulation for nonlinear flexure of plates
2.1. Total potential energy
The formulation for the large deflection analysis of laminated composite plates is derived here from Ritz
method based on the first-order shear deformation plate theory (FSDT). In geometrically nonlinear
analyses, the nonlinear terms are included in the strains.
In the first order plate shear deformation theory [2], the displacement field of a laminated composite
plate is expressed as
uðx; y; zÞ ¼ u0ðx; yÞ þ zhxðx; yÞ;vðx; y; zÞ ¼ v0ðx; yÞ þ zhyðx; yÞ;wðx; y; zÞ ¼ w0ðx; yÞ;
ð1Þ
where (u; v;w) are the displacements of a generic point (x; y; z) in the laminated plate, and (u0; v0;w0) are the
displacement projections on the mid-plane, and (hx; hy) denote the rotations of the transverse normal about
y- and x-axis, respectively. Considering Green’s expression as well as the lower order terms in a manner
4764 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
consistent with the usual von Karman assumptions, which concerns small strains and moderately large
rotations, the strain components at a generic point of the plate are expressed as
e ¼
exeyczxcyzcxy
8>>>>>><>>>>>>:
9>>>>>>=>>>>>>;¼
exey2ezx2eyz2exy
8>>>>>><>>>>>>:
9>>>>>>=>>>>>>;¼
ouox
þ zohxox
þ 1
2
owox
2
ovoy
þ zohyoy
þ 1
2
owoy
2
owox
þ hx
owoy
þ hy
ouoy
þ ovox
þ zohxoy
þ ohyox
þ ow
oxowoy
8>>>>>>>>>>>>>>>>><>>>>>>>>>>>>>>>>>:
9>>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>>;
; ð2Þ
which are re-expressed as linear and nonlinear components as
eL ¼eLxeLycLxy
8><>:9>=>; ¼
eLxeLy2eLxy
8><>:9>=>; ¼
ouoxovoy
ouoy
þ ovox
8>>>>>><>>>>>>:
9>>>>>>=>>>>>>;; ð3Þ
c ¼czxcyz
( )¼
2ezx2eyz
¼
owox
þ hx
owoy
þ hy
8>><>>:9>>=>>;; ð4Þ
j ¼jx
jy
jxy
8><>:9>=>; ¼
ohxoxohyoy
ohxoy
þ ohyox
8>>>>>><>>>>>>:
9>>>>>>=>>>>>>;ð5Þ
and
eNL ¼eNLx
eNLy
cNLxy
8><>:9>=>; ¼
eNLx
eNLy
2eNLxy
8><>:9>=>; ¼
1
2
owox
2
1
2
owoy
2
owox
owoy
8>>>>>>>><>>>>>>>>:
9>>>>>>>>=>>>>>>>>;: ð6Þ
With the usual assumptions of FSDT, the force and moment resultants of the plates are expressed asbNcMbQ8<:
9=; ¼A B 0
BT D 0
0 0 E
24 35 eL þ eNL
j
c
8<:9=; ð7Þ
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4765
in which, the membrane, membrane-bending, bending and transverse shear stiffness terms are repre-
sented by the sub-matrices, i.e., A, B, D and E, respectively. The strain energy of the laminated plate is
expressed as
U ¼ 1
2
Z ZX
eTL þ eTNL jT cT A B 0
BT D 0
0 0 E
24 35 eL þ eNL
j
c
8<:9=;dxdy
¼ 1
2
Z ZX
eTL þ eTNL jT A B
BT D
eL þ eNL
j
dxdy þ 1
2
Z ZXcTEcdxdy: ð8Þ
Considering a uniformly distributed transverse load q, the external work is given by
W ¼Z Z
Xwqdxdy: ð9Þ
Thus the total potential energy of the plate becomes
P ¼ U W
¼ 1
2
Z ZX
eTL þ eTNL jT A B
BT D
eL þ eNL
j
dxdy þ 1
2
Z ZXcTEcdxdy
Z ZXwqdxdy: ð10Þ
2.2. Ritz procedure for nonlinear equilibrium equations
By discretizing the problem domain X into NP scattered particles ðx1; x2; . . . ; xNPÞ, the displacement field
of a laminated plate using the kernel particle Ritz approach can be expressed as
½u; v;w; hx; hy T ¼XNP
I¼1
kðx; x xIÞ~/ðx xIÞ½uI ; vI ;wI ; hxI ; hyI T ¼XNP
I¼1
WIðx; yÞd I : ð11Þ
where NP is the total number of the particles which are used to discretize the domain, WIðx; yÞ are the shapefunctions, and d I ¼ ½uI ; vI ;wI ; hxI ; hyI T is a vector of unknown displacements of Ith particle. The shape
functions can be expressed by
WIðxÞ ¼ kðx; x xIÞ~/ðx xIÞ ¼ kðx; x xIÞ/ðx xIÞnðx xJ Þ
; ð12Þ
in which, kðx; x xIÞ is the coefficient function, /ðx xIÞ the kernel function and nðx xJ Þ the function to
introduce the singularity to the kernel function which is associated with the particle xI . The kernel function
and the singularity function are defined as
/ðx xIÞ ¼1
ax/
x xIax
1
ay/
y yIay
ð13Þ
and
nðx xIÞ ¼ nðx xI ; y yIÞ ¼x xIax
2"
þ y yIay
2#p
ð14Þ
in which, ax and ay are the kernel support measurement along x-axis and y-axis, respectively. p is the
singularity order (p > 0). The correction function kðx; x xIÞ is written as
4766 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
kðx; x xIÞ ¼ aTðx xIÞbðxÞ; ð15Þ
where
aTðx xIÞ ¼ ½1; x xI ; y yI ; ðx xIÞ2; ðx xIÞðy yIÞ; ðy yIÞ2; ð16Þ
bTðxÞ ¼ ½b00ðxÞ; b10ðxÞ; b01ðxÞ; b20ðxÞ; b11ðxÞ; b02ðxÞ ð17Þ
and a is a vector of quadratic basis. The b(x) are functions of x which are yet to be determined. Thus, the
shape functions can be written as
WIðxÞ ¼ bTðxÞaðx xIÞ~/ðx xIÞ; ð18Þ
where
bðxÞ ¼ M1ðxÞað0Þ ð19Þ
and the moment matrix M is a function of x, given by
MðxÞ ¼XNP
I¼1
aðx xIÞaTðx xIÞ~/ðx xIÞ: ð20Þ
Therefore, the shape function can be expressed as
WIðxÞ ¼ aTð0ÞM1ðxÞaðx xIÞ~/ðx xIÞ: ð21Þ
In this paper, the cubic spline function
/ðzIÞ ¼
2
3 4z2I þ 4z3I for 06 zI 6 1
2
4
3 4zI þ 4z2I
4
3z3I for
1
26 zI 6 1
0 otherwise
8>>>><>>>>:
9>>>>=>>>>; ð22Þ
is chosen as the kernel function, where zI ¼ jx xI j=aI . The domain, in which /ðxJ xIÞ ¼ 1aI/ xJxI
aI
6¼ 0
for all particles xJ is called the support of /ðx xIÞ. The dilation parameter aI ¼ aðxIÞ is to control the size
of the support, and the parameter values are often different at various particles. In the numerical imple-
mentation, the dilation parameters are usually defined as aI ¼ sDI where s is a scaling factor, and DI is thelarger of the distances of the two neighbors of xI in one-dimensional case. For the 2-D problem, the dilation
parameters can similarly be defined in two directions.
With the assumptions of displacements as in Eq. (11), the strains of the composite laminate are expressed
using Eqs. (3)–(6) as
eL ¼XNP
I¼1
BLIdpI ; eNL ¼ 1
2
XNP
I¼1
Bndb; j ¼XNP
I¼1
BbIdbI ; c ¼XNP
I¼1
BqIdbI ; ð23Þ
where the displacements are re-expressed as two parts
dpI ¼ uI vI½ T; dbI ¼ wI hxI hyI½ T: ð24Þ
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4767
The strain matrices are given by
BLI ¼WI;x 0
0 WI ;y
WI;y WI ;x
24 35; BnI ¼
owox
0
0owoy
owoy
owox
26666664
37777775oWI
ox0 0
oWI
oy0 0
26643775; ð25Þ
BbI ¼0 WI;x 00 0 WI ;y
0 WI ;y WI ;x
24 35; BqI ¼WI ;x WI 0
WI;y 0 WI
: ð26Þ
The nonlinear equations of the plate can be derived by Ritz method. First, the strain energy is divided into
two parts UL and UN, corresponding to the linear strains and nonlinear strains, respectively. Expanding
Eq. (10), one obtains
U ¼ UL þ UN ¼ 1
2
Z ZX
eTLAeL
þ eTLBjþ jTBeL þ jTDjþ cTEcdxdy
þ 1
2
Z ZX
eTLAeNL
þ eTNLAeL þ eTNLBjþ jTBeNL þ jTDeNL þ eTNLAeNL
dxdy: ð27Þ
The Ritz minimization of the total potential energy yields the equations
oPodI
¼ oUL
odIþ oUN
odI oW
odI¼ 0; dI ¼ uI ; vI ;wI ; hxI ; hyI : ð28Þ
Using Eqs. (23) and (27),
oUL
odI¼ K0d: ð29Þ
where K0 is the linear stiffness matrix, a typical entry of which is defined as follows:
.TTT
TT
0 dxdyqJqIbJbILJbI
bJLILJLIIJ ∫∫Ω
+
=EBBDBBBBB
BBBABBK ð30Þ
From the nonlinear part of the strain energy
oUN
odI¼ Kn2d: ð31Þ
For the purpose of this derivation, the nonlinear strains in Eq. (6) are re-written as
eNL ¼ 1
2
owox
0
0owoy
owoy
owox
26666664
37777775owoxowoy
8>><>>:9>>=>>; ¼ 1
2cH bP: ð32Þ
4768 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
Then the following displacement-dependent stiffness matrix can be obtained as
.
2
1
2
12
10
TTTTT
T
2 dxdy
nJnInJbIbJnILJnI
nJLI
IJn ∫∫Ω
++=
ABBBBBBBBABB
ABBK ð33Þ
From the differential of the external work, the force vector is obtained as
F I ¼oWodI
¼ dxdy: ð34Þ
where qI ¼ ½0; 0; q; 0; 0T, and the shape function matrix is defined as
WI ¼
WI 0 0 0 0
0 WI 0 0 0
0 0 WI 0 00 0 0 WI 0
0 0 0 0 WI
266664377775: ð35Þ
Thus the nonlinear equilibrium equations may be obtained as
K sðdÞd ¼ F; ð36Þ
where the secant stiffness matrix is defined as
K sðdÞ ¼ K0 þ Kn2ðdÞ: ð37Þ
3. Solution of the nonlinear equations
3.1. Incremental form of nonlinear equations
For solving the nonlinear system, incremental-iterative Newton-type solution methods should be used.
To get the incremental form, Eq. (36) is re-expressed as
gðdÞ ¼ K sd F ¼ 0: ð38Þ
The applied load is assumed to be proportional to a fixed load F0 as
F ¼ kF0: ð39ÞThus the nonlinear equilibrium equations become functions of displacements and the load scaling factor k.Hence, Eq. (38) may be re-written as
gðd; kÞ ¼ K sðdÞd kF0 ¼ 0: ð40ÞFor an equilibrium configuration near the old configuration,
gðd þ Dd; kþ DkÞ ¼ 0: ð41ÞApplying the Taylor series expansion to the above equation
gðd þ Dd; kþ DkÞ ¼ gðd; kÞ þ K tDd DkF0 ¼ 0; ð42Þ
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4769
where Dd is the displacement increment as the load changes from kF0 to ðkþ DkÞF0, and the following
substitution is implicit:
oUodI
¼ K sd; ð43Þ
o2UodIodJ
¼ K t: ð44Þ
The tangent stiffness matrix can be obtained as
K t ¼ K0 þ Kn þ KG; ð45Þin which K0 is the linear stiffness matrix as in Eq. (31), KG the geometrical stiffness matrix and Kn the
displacement-dependant stiffness matrix. The last two matrices can be expressed as
dxdynJnInJbIbJnILJnI
nJLInIJ ∫∫ ∫∫Ω
++
=ABBBBBBBBABB
ABBK
TTTT
T0ð46Þ
and
ð47Þ
Finally the incremental form of the nonlinear equilibrium equations is expressed by using the tangent
stiffness matrix and the displacement increments as
K tDd ¼ DkF0 gðd; kÞ: ð48ÞThe incremental formulae for iteration are written as
Ddm ¼ ½K tðdmÞ1½DkmF0 gðdm; kmÞ¼ ½K tðdmÞ1½DkmF0 K sðdmÞdm þ kmF0;
dmþ1 ¼ dm þ Ddm;
8<: ð49Þ
where the subscript m refers to the load step number.
As Dk is a variable to be solved for each incremental step, an additional constraint equation is needed.
The arc-length continuation used here provides this constraint. In the arc-length continuation method, for
each Dk step, a corresponding external load increment is applied, and subsequently iterations are necessary
to reach a new equilibrium using Eq. (48). Let the superscript n denote the nth iteration cycle and subscript
m represents the mth load step, the equilibrium iterations start from n ¼ 2, and n ¼ 1 corresponds to theexternal load increment of every load step m. Then the generalized equation of the incremental-iterative
equation is
½K tmDdnm ¼ DknmF0 gðdn1
m ; kn1m Þ ¼ DknmF0 ½K sðdn1
m Þdm kn1m F0: ð50Þ
The displacement increment Dd is composed of two parts, one is from the external load increment and the
other is from the residual forces, i.e.
Ddnm ¼ ½ðK tÞm
1 DknmF0
½K sðdn1
m Þdm kn1m F0
¼ Dknm½d f m þ ½DdRnm; ð51Þ
where
½d f m ¼ ½ðK tÞm1F0 ð52Þ
4770 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
and
½DdRnm ¼ ½ðK tÞm1gn1m ¼ ½ðK tÞm
1½K sðdn1m Þdn1
m kn1m F0: ð53Þ
The increment of the load level parameter Dknm is obtained by solving the arc-length constraint equation
(Eq. (58)). The total displacements and the updated load level parameter are obtained as
dnm ¼ dn1
m þ Ddnm; knm ¼ kn1
m þ Dknm: ð54Þ
3.2. Iterative strategy of the arc-length technique
In the modified Riks method proposed by Crisfield [19], the additional arc-length constraint equation is
written as
½ðDdaÞnmTðDdaÞnm ¼ l2m; ð55Þ
where lm is the generalized ‘arc-length’ of the tangent at converged state of the last load step in the load–
deflection space, and ðDdaÞnm are the accumulated displacements in the mth load step until the nth iteration,
i.e.
ðDdaÞnm ¼ dnm dm1: ð56Þ
Eq. (55) can be re-written as
½ðDdaÞn1
m þ Dknmðd fÞm þ ðDdRÞnmT½ðDdaÞn1
m þ Dknmðd fÞm þ ðDdRÞnm l2m ¼ 0; ð57Þ
and simplified into a quadratic equation,
a1ðDknmÞ2 þ a2Dk
nm þ a3 ¼ 0; ð58Þ
where the coefficients are defined as
a1 ¼ ðd fÞTmðd fÞm;a2 ¼ 2½ðDdaÞn1
m þ ðDdRÞnmTðd fÞm;
a3 ¼ ½ðDdaÞn1
m þ ðDdRÞnmT½ðDdaÞn1
m þ ðDdRÞnm l2m:
ð59Þ
The constraint equation may have two real roots, and the correct choice is the one makes the angle between
the incremental displacement vector of last iteration and that of current iteration positive. The angles aremeasured by
a1 ¼ ½ðDdaÞn1
m þ ðDknmÞ1ðd fÞm þ ðDdRÞnmTðDdaÞn1
m ;
a2 ¼ ½ðDdaÞn1
m þ ðDknmÞ2ðd fÞm þ ðDdRÞnmTðDdaÞn1
m :ð60Þ
The correct root is the one that makes a positive. If both are positive, the one that makes the a smaller is
selected.
3.3. Strategy for load incrementation
When the arc-length method is incorporated into the modified Newton method to solve the nonlinear
equation, the load incrementation is implicitly controlled. If the (m 1)th step load km1F0 and thecorresponding displacements dm1 have been known, then the arc-length in the mth load step is defined
as
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4771
l2m ¼ ðDk1mÞ2ðd fÞTmðd fÞ; ð61Þ
where Dk1m is the first load increment in the mth load step. Then the arc-length used in the current increment
can be obtained from the previous increment as
lm ¼ lm1
JdesJm1
0:5–1
; ð62Þ
where Jdes is the desired number of the iterations (3–5), and Jm1 is the number of iterations in the previous
(m 1)th load step. The new load increment parameter is obtained as
Dk1m ¼ lmffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðd fÞTmðd fÞm
q : ð63Þ
To begin with, the first load increment Dk11, it is usually chosen to be 20–40% of the anticipated maximum
load.It should be noted that the load incremental parameter Dk1m may have different signs and the correct sign
is decided by Crisfield’s method [19]. When the determinant of the tangent stiffness matrix does not change
sign, the current sign of Dk1m follows that of the previous increment. Otherwise, Dk1m changes its sign.
4. Numerical examples
In this section, a few examples of composite laminated plates with large enough loads to cause significantgeometric nonlinearity are numerically investigated. Simply supported and clamped boundary conditions
are considered. The results are compared with those calculated by other methods. Exact or close-form
solutions of nonlinear analysis of plates are not available yet. Hence in this paper, the nonlinear deflections
of plates are compared with experimental results or numerical results of other methods such as the finite
element method.
In the following examples, regular particle distribution is set as 17 · 17. The Gauss integration scheme is
set as 2 · 2. Since the singularity has been introduced to the kernel function to construct the shape func-
tions, the boundary conditions can be treated as in FEM techniques. As no coordinate transformation isneeded, the singular kernel method is more suitable for nonlinear analysis than the other methods in
particle approximations. The coding procedures may be seen in Appendix A.
4.1. Isotropic plate
The nonlinear flexure of a square isotropic plate is carried out first. The plate is subjected to a uniformly
distributed transverse load of intensity q, and it is simply supported on all the four edges (SSSS). The
geometry and the load are shown as in Fig. 1. The geometry parameters and nondimensional materialproperties are set as
a ¼ b ¼ 10; h ¼ 0:2; E ¼ 3 106; v ¼ 0:316: ð64Þ
As the thickness-to-edge ratio is quite small (h=b ¼ 0:02), so the present results could be compared with
Chia’s results [20] for the thin plate.
Chia used series method to solve the nonlinear bending of isotropic plates with all simply supported
edges (SSSS) or clamped edges (CCCC). The deflection–load curve of the SSSS and CCCC isotropic plate
4772 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
solved by the present method is shown in Figs. 2 and 3, respectively. The nondimensional deflection andloads in Fig. 2 are defined as the follows:
w ¼ wc
h; q ¼ qb4
Eh4; ð65Þ
where wc is the center deflection. Similarly, with respect to Fig. 3
w ¼ wc
h; q ¼ qb4
Dh; ð66Þ
where D ¼ Eh3=12ð1 m2Þ is the flexural rigidity of the plate.
The results are compared with the series solutions of Chia [20] which is based on thin plate theory.
Though a slight difference exists between the solutions, the agreement is still quite good.
0 8 16 24 32 40 48 56 64 72 80 880.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
w
q
Chia (1980) [20] Present
Fig. 2. The load–deflection curves of an SSSS square isotropic plate.
a
y
o b
x
qh
Fig. 1. An SSSS plate with uniformly distributed transverse loads.
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4773
4.2. Laminated composite plates
Symmetric and antisymmetric cross-ply composite laminated plates of square and rectangular geometry
are considered with fully clamped or simply supported edges. Uniformly distributed lateral pressures are
applied to the plates.First, a square four-layered symmetric cross-ply laminate (0/90/90/0) is considered. The geometry
and material properties are set as follows:
a ¼ b ¼ 12 in:; h ¼ 0:138 in:; E1 ¼ 1:8282 106 psi; E2 ¼ 1:8315 106 psi;
G12 ¼ G23 ¼ G13 ¼ 3:125 105 psi; m12 ¼ m23 ¼ m13 ¼ 0:23949: ð67Þ
The plate is clamped at the edges (CCCC), and thus the boundary conditions are set as
½u; v;w; hx; hy TI ¼ 0; ð68Þ
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.20.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
w
q
FEM (Ganapathi, 1996) [22] Experimental
(Zaghloul et al., 1975) [21] Present
Fig. 4. The load–deflection curves of a CCCC square symmetric cross-ply composite laminated plate ð0=90=90=0Þ.
0 25 50 75 100 125 150 175 200 225 250 275 300 3250.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
w
q
Chia (1980) [20] Present
Fig. 3. The load–deflection curves of a CCCC square isotropic plate.
4774 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
where the subscript I refers to a typical node on the edges.
The plot of center deflection versus lateral load intensity is shown in Fig. 4. The present solution is
compared with available results obtained by experiments [21] and the finite element method [22]. The
present result is in close agreement with the finite element result, although it has a considerable deviation
from the experimental curve. This discrepancy is due to the assumption of an ideal clamped boundary
condition for the computation, which are difficult to ensure in experiments.
A rectangular three-layer symmetric cross-ply plate (0/90/0) is also studied. The nondimensional
material properties are as follows:
E1 ¼ 25E2; G12 ¼ G13 ¼ 0:5E2; G23 ¼ 0:2E2; m12 ¼ 0:25: ð69Þ
0 40 80 120 160 200 240 280 320 360 4000.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
w
q
PresentSavithri et al. (1993) [23]
Fig. 5. The load–deflection curve of an SSSS square symmetric cross-ply composite laminated plate ½ð0=90=0Þ; h=b ¼ 0:1; a=b ¼ 1.
0 40 80 120 160 200 240 280 320 360 400 4400
25
50
75
100
125
150
175
200
225
250
275
σ x
q
Fig. 6. The load–stress curve of an SSSS square symmetric cross-ply composite laminated plate ½ð0=90=0Þ; h=b ¼ 0:1; a=b ¼ 1.
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4775
The plate is simply supported along all edges (SSSS), and is subjected to the uniformly distributed trans-
verse load. The geometric parameters h=b and a=b are varied from 0.01 to 0.1 and 0.5 to 2.0, respectively.
The results are compared with the series solutions of Savithri et al. [23]. For comparison, the nondimen-
sional load parameter, the center deflection and the extreme fibre stress are computed as:
q ¼ qb4
E2h4; w ¼ wc
h; rx ¼
rxb4
E2h4; ð70Þ
where q is the uniform lateral pressure, wc is the center transverse displacement of the plate and rx is the
extreme fibre stress. The load–displacement plots are shown in Figs. 5–9, which suggest that the results are
in close agreement with the series solutions.
0 40 80 120 160 200 240 280 320 360 4000.0
0.1
0.2
0.3
0.4
0.5
0.6
w
q
Present Savithri et al. (1993) [23]
Fig. 8. The load–deflection curve of an SSSS rectangular symmetric cross-ply composite laminated plate ½ð0=90=0Þ; h=b ¼0:1; a=b ¼ 0:5.
0 40 80 120 160 200 240 280 320 360 4000.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
w
q
PresentSavithri et al. (1993) [23]
Fig. 7. The load–deflection curve of an SSSS rectangular symmetric cross-ply composite laminated plate ½ð0=90=0Þ; h=b ¼0:1; a=b ¼ 2.
4776 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
5. Conclusions
In this paper, a kernel particle Ritz solution method has been proposed for the geometrically nonlinear
analysis of plate flexure, which is based on the total Lagrangian formulation and first-order shear defor-
mation plate theory. The nonlinear equations are solved by using the modified Newton’s method and arclength continuation. Typical problems involving isotropic as well as laminated composite plates have been
solved considering different aspect ratios and boundary conditions. The numerical results are in excellent
agreement with those obtained by finite element and series solution methods reported by previous
researchers. Thus the proposed kernel particle Ritz method promises to be potential alternative to the finite
element method in solving the large deformation plate flexure problems.
Appendix A. Coding procedure
For numerical implementation, the code includes three main steps, namely pre-processing, solving and
post-processing procedures. In the code developed for the present research, the nonlinear equations may be
solved by the modified Newton–Raphson method alone or combining it with the arc-length technique. The
algorithm is described in the following flow chart:
(1) Pre-processing step:
• Input the material properties and compute the stiffness coefficients A, B, D and E;• Input the fixed external loading, scaling factor s, and the geometry (length a, width b and thickness
h);• Set up particle coordinates;
• Determine the domain of influence of every particle;
• Read variables related with boundary conditions;
• Set up Gauss quadrature cells, Gauss points, weights, Jacobian;
• Compute the equivalent fixed external force vector F0 in Eq. (39);
0 40 80 120 160 200 240 280 320 360 4000.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
w
q
a/b= 2a/b= 1a/b=0.5
Fig. 9. Comparisons of the load–deflection curves of SSSS rectangular symmetric cross-ply composite laminated plates with different
aspect ratios ½ð0=90=0Þ; h=b ¼ 0:1.
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4777
• Input the parameters for increment/iteration control;
For a new computation, set initial displacements, initial internal forces, initial load-level factor,
load factor increment, desired iteration times, d0, k0, Dk11, R0 and Jdes;
For a re-solution, read the displacements, internal forces, external forces and etc. of last successful
load step. d0, k0 and R0 from the previous saved file.
(2) Solution step:
• Start the load incremental loop m Save data of the (m 1)th load step; Compute the tangent stiffness K tðdm1Þ as in Eq. (45), check the sign of the determinant of K t to
determine if the limit point or singular point is passed;
Add boundary conditions and solve the tangent displacement ðd fÞm due to the fixed external force
F0: ðd fÞm ¼ ½ðK tÞm1F0;
Update the load level factor;
If m ¼ 1, compute first arc-length l21 ¼ ðDk11Þ2ðd fÞT1 ðd fÞ1;
If mP 2, compute arc length lm ¼ lm1
ffiffiffiffiffiffiffiJdesJm1
q;
Compute the initial load increment Dk1m ¼ lmffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðd f ÞTmðdf Þm
p in the mth load step;
Update the total displacements and load level: d1m ¼ dm1 þ Dk1mðd fÞm, k
1m ¼ km1 þ Dk1m;
Save k1m; Start equilibrium iteration loop using the modified N–R method with arc-length continuation
method. In the nth iteration
– Calculate residual forces gn1m ¼ K sðdn1
m Þ kn1m F0;
– Save residual forces;
– Compute displacement increments: ðDdaÞn1
m ¼ dn1m dm1;
– Enforce boundary conditions, and compute ðDdRÞnm ¼ ½K t1
m gn1m ;
– Solve a1ðDknmÞ2 þ a2Dk
nm þ a3 ¼ 0 to obtain Dknm, and select correct root or stop the iteration if the
maximum load has been reached or there are no real roots;
– Update the load increment;
– Save increment of the load level and update the displacement and load level;
– Obtain the displacement increments: Ddnm ¼ Dknmðd fÞm þ ðDdRÞnm;
– Update total displacements: dnm ¼ dn1
m þ Ddnm;
– Update total load level: knm ¼ kn1m þ Dknm;
– Check convergence by using the norm of the residual forces: kgnmk6 bkFn
mk ¼ bk6 knmF0k;– If the norm is less than the tolerance limit, stop iteration and save the iteration number.
End of the iteration loop.
• End of the load increment loop.
(3) Post-processing step:
• Obtain the displacements and strain matrices in each load step, then calculate the stresses of the plate
at any point in the plate;
• Draw the deformation mode and stress distributions, and etc.
References
[1] R.D. Cook, D.S. Malkus, M.E. Plesha, Concepts and applications of finite element analysis, 3rd ed., John Wiley and Sons, New
York, 1989.
[2] O.O. Ochoa, J.N. Reddy, Finite element analysis of composite laminates, Kluwer Academic Publishers, Dordrecht, Boston, 1992.
[3] P.G. Bergan, Solution algorithms for nonlinear structural problems, Comput. Struct. 12 (1980) 497–509.
4778 K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779
[4] M.E. Fares, Non-linear bending of composite laminated plates using a refined first-order theory, Compos. Struct. 46 (1999) 257–
266.
[5] E. Carrera, M. Villani, Large deflections and stability FEM analysis of shear deformable compressed anisotropic flat panels,
Compos. Struct. 29 (1994) 433–444.
[6] Q.H. Cheng, T.S. Lok, Z.C. Xie, Geometrically non-linear analysis including shear deformation of composite laminates, Thin-
Walled Struct. 35 (1999) 41–59.
[7] M.J. Clarke, G.J. Hancock, A study of increment-iterative strategies for non-linear analyses, Int. J. Numer. Methods Engrg. 29
(1990) 1365–1391.
[8] J.S. Chen, C. Pan, C.T. Wu, W.K. Liu, Reproducing kernel particle methods for large deformation analysis of non-linear
structures, Comput. Methods Appl. Mech. Engrg. 139 (1996) 195–227.
[9] S. Li, W. Hao, W.K. Liu, Meshfree simulations of shear banding in large deformation, Int. J. Numer. Methods Engrg. 37 (2000)
7185–7206.
[10] K.M. Liew, T.Y. Ng, Y.C. Wu, Meshless method for large deformation analysis––a reproducing kernel particle method, Engrg.
Struct. 24 (2002) 543–551.
[11] K.M. Liew, Y.C. Wu, G.P. Zou, T.Y. Ng, Elasto-plasticity revisited: numerical analysis via reproducing kernel particle method
and parametric quadratic programming, Int. J. Numer. Methods Engrg. 55 (6) (2002) 669–683.
[12] K.M. Liew, H.Y. Wu, T.Y. Ng, Meshless method for modeling of human proximal femur: treatment of nonconvex boundaries
and stress analysis, Comput. Mech. 28 (5) (2002) 390–400.
[13] J. Wang, K.M. Liew, M.J. Tan, Meshless analyses of shear deformable plates, in: Proceedings of the Eighth Annual International
Conference on Composite Engineering, Tenerife, Spain, 5–11 August, 2001, pp. 965–966.
[14] J. Wang, K.M. Liew, M.J. Tan, S. Rajendran, Analysis of rectangular laminated composite plates via FSDT meshless method, Int.
J. Mech. Sci. 44 (2002) 1275–1293.
[15] K.M. Liew, J. Wang, T.Y. Ng, M.J. Tan, Free vibration and buckling analyses of shear-deformable plates based on FSDT
meshfree method, J. Sound Vib., in press.
[16] X. Zhao, K.M. Liew, T.Y. Ng, Vibration analysis of laminated composite cylindrical panels via a mesh-free approach, Int. J. Solid
Struct. 40 (2003) 161–180.
[17] K.M. Liew, X. Zhao, T.Y. Ng, Element-free kp-Ritz method for vibration of laminated rotating cylindrical panels, Int. J. Struct.
Stab. Dyn. 2 (4) (2002) 523–558.
[18] X. Zhao, T.Y. Ng, K.M. Liew, Free vibration of laminated two-side simply-supported cylindrical panels via the kp-Ritz method,
Int. J. Mech. Sci. 46 (2004) 123–142.
[19] M.A. Crisfield, Non-linear finite element analysis of solids and structures, John Wiley and Sons, Chichester, UK, 1991.
[20] C.Y. Chia, Nonlinear analysis of plates, McGraw-Hill, New York, 1980.
[21] S.A. Zaghloul, J.B. Kennedy, Nonlinear behaviour of symmetrically laminated plates, J. Appl. Mech., Trans. ASME 42 (1) (1975)
234–236.
[22] M. Ganapathi, O. Polit, M. Touratier, A C0 eight node membrane-shear-bending element for geometrically nonlinear (static and
dynamic) analysis of laminates, Int. J. Numer. Methods Engrg. 39 (20) (1996) 3453–3474.
[23] S. Savithri, T.K. Varadan, Large deflection analysis of laminated composite plates, Int. J. Non-linear Mech. 28 (1) (1993) 1–12.
K.M. Liew et al. / Comput. Methods Appl. Mech. Engrg. 193 (2004) 4763–4779 4779