pb-2 Rayleigh- Ritz method

for general plate analysis

K. M. Liew
Division of Apphed Mechamcs, School of Mechanical and Production Engineering,
Nanyang Technology University, Nanyang Avenue, Singapore

C. M. Wang
Department of Civil Engmeermg, Natwnal Untversltv of Singapore, Kent Ridge,
Singapore
(Recezved Aprd 1991, rewsed version accepted May 1991)

An alternative method to the widely used finite element rruethod for

elastic analysis of thin plates, the pb-2 Rayleigh-Ritz method, is
presented. The special feature of the pb-2 Rayleigh-Ritz method
lies in the definition of the Ritz functions which consist of the product
of a basic function and a two-dimensional polynomial function where
the number of terms may be increased until the desired accuracy is
achieved. The basic function is formed by the product of all the boundary equations; each of which Js raised to the power of either O, 1 or
2, corresponding to either free, simply supported or clamped edges,
respectively. Thus the basic function ensures the satisfaction of the
kinematic boundary conditions at the outset. The bending, buckling
and vibration analyses are presented in a unified form and the pb-2
Rayle~gh-Ritz method is applied to solve some plate examples to
illustrate ~ts simplicity and accuracy.
Keywords: bending, buckling, deflection, elastic, frequencies, pb-2
RayleJgh- Ritz method, plates, polynomials, vibration

Exact plate solutions are available only for certain

shapes, boundary and loading conditions (e g. in standard texts J-4). Apart from these hmlted classes of
plates, numerical techmques are necessary for analys~s.
The most widely used technique must be the fimte element method because of its versatihty in handhng any
arbitrarily shaped plate under all kinds of boundary and
loading conditions. In the fimte element method, there
are two versions: (1) the h-version and (2) thep-version.
The h-version FEM is the conventional approach where
the plate Is dwided into a finite number of elements of
maximum dmmeter, h The degrees of freedom at the
selected nodal points of the plate element are usually
approximated by polynomials of degree 1 - 3 . By refining and decreasing the mesh size, the plate solutions may
be obtained to the desired accuracy. Unlike the hversion, the p-versmn FEM uses a fixed number of
elements and increases the degree of polynomial, p to
perhaps 6 - 8 until the desired accuracy is achieved.
Based on the study by Babuska et al 5, the p-version
has some advantages such as" its rate of convergence
is not restricted by a fixed polynomial degree and may be
twice that of the h-version when the solutmn has a
singularity resulting from corner condmons and the
volume of the input data is much reduced due to the
0141 0296/93/010055
06
'~', 1 9 9 3 B u t t e r w o r t h - H e m e m a n n

employment of only a few elements However, the hversion when used with an optimal mesh design yields
a higher rate of convergence.
Although the h-version FEM Is a very useful tool, it
imposes some mconvemences on the analysts such as the
preparauon of an appropriate element mesh and the
requirement of a huge memory space for computational
purposes due to the large number of degrees of freedom
mvolved The p-version somewhat ~mproves on these two
respects by using fewer elements The method may stall
require a relatwely large number of dlscretizatlon nodes
to adequately approximate the curved boundary conditions. These inconveniences and approximations of the
boundary condmons via discretizatlon, however, do not
arise when using the more traditional Raylelgh-Ritz
method Being an approximate continuum method, it
ehminates the need for discretlzatlon, mesh generation
and thus large degrees of freedom by vlewmg the enUre
plate as a single super element with its deflection surface
approximated by statable Ritz functions_ The problem
now lies m finding that Ritz displacement function
which is general for any boundary condmons so that the
method may be as easdy automated and versatile as the
FEM. This problem has been a major setback for the
R a y l e i g h - R l t z method untd recent studies on the free

Ltd

Eng. Struct. 1993, Vol. 15, No 1 55

pb-2

Raylelgh-Ritz method for general plate analysts K M. Llew and C M Wang

vJbrallon of thin plate,, by' Lle,,v and Lain" Th% proposed the two-dimensional Ritz functions consisting of
the product of a basic function and orthogonal polynolmals whose degree may be increased untd the desired
accuracy is achlexed The basic funcuon is defined by
the product of equaUons of the speofied boundary shape
raised to the power of either 0, I or 2 corresponding to
free. simply supported or clamped edges, respectively
The basic function ensures automatic saUsfact~on of the
kmemauc boundary cond~uons at the outset w~thout
needing to use Lagrangmn multxphers (as m the
Lagrangmn multlpher method) These pb-2 Ritz functions (p, b, 2) denote polynomials, boundary expression
and d~menslon, respectwely) obvmte the tedious task of
choosing the form of the lnfimte series or tngonometlc
or algebraic functions to suit the condmons of support
along the edges As a result of thesepb-2 Ritz functions,
the R a y l e l g h - R i t z method can now be applied readily
to plates with any boundary condmons
This paper presents the pb-2 R a y l e l g h - R l t z method
for apphcations in bending, buckling and vibration
analyses of elastic thin plates By framing the energy
functional into a unified form, a general purpose software package has been developed to perform the three
kinds of analyses. To illustrate the slmphcity and
accuracy of the method, some plate examples are solved.

Nx

A*

} "/-

1(x,y)

Figure

Elastm thin plate

The work done, W~, by the transverse loads is given

by

Problem definition
Consider a flat, thin, lsotroplc and elastic plate of constant thickness, t with given support conditions as shown
in Figure 1. The plate may be subjected to transverse
and m-plane loads. The problems treated hereto a r e (a),
bending analysis of plates under transverse and m-plane
loads in which the deflection surface of the plate middle
plane is to be determined, from which the moments and
stresses may be obtained if required; (b), buckhng
analysis of plates under m-plane loadlngs in which the
critical load ~s to be determined; and (c), vibration
analysis of plates with/without m-plane loadlngs in
which the natural frequencies are to be determined. The
formulaUon and method of analyses for these problems
will be presented in a unified manner below

Rayleigh-Ritz method for general

analysis

pb-2

The strain energy, U, of the considered plate shown in

Figure I is given by ~

U=

o,

2-

I (\a.2

-2(1-

u)

\axOyJ J dA

(1)

where D = Et-~/[12(l - v2)] = flexural rigidity of the

plate, v = IS the Polsson ratio, w(x, y) IS the deflection
of the middle plane of the plate perpendicular to the xv
plane and A is the area of the plate. Note that the second
term in equaUon (1) vanishes for clamped plates and for
polygonal plates if one of the edge conditions is either
w = 0 or Ow/On = 0, where n is the direction normal to
the edge

56

p w dA* ,

P, w,

Eng. Struct. 1993, Vol. 15, No 1

(2)

i=l

while W2, for m-plane loads by

l t' [
W~-=2

(Ow~ 2 NfOW~ 2
N'\ox l

'~Oy]

aw O.~dA
+ 2 N , Ox 0 3 )

(3)

where p(x, y) is the transverse distributed load over the

area A* <_ A, P. is the concentrated vertical load. w, is
the vertical deflection under the point load P,; np is the
total number of point loads. N~, N,, N~, are the in-plane
normal loads in the x-direction, y-direcuon and in-plane
shearing loads, respectively
In the case of vibration analysis, the kinetic energy, T,
of the plate, after eliminating the time component from
the displacement function, is given by

T= 21 Otto'- 11'4w2 dA

Ox: Ov-~

Wt =

(4)

where p is the mass density per unit area of plate and ~0

IS the angular frequency of vibration
The unified expression of the energy functional, F, for
the various types of plate analyses can be written as

F = U-/31W, -/32W2 - /33T

(5)

in which the scalar indicators /3 take on

/3~ = 1, /32 =/33 = 0

for bending analysis of

plates under transverse loads

pb-2 Rayleigh-Ritz method for general plate analysis: K. M. Dew and C. M. Wang

~ I I I i I I i I F--L
I

A
-0.5

B
+0.5

0
I

v=0.3

=g
1

0.0005

- x

05/

b/2

ay

I-

0/2

%L

./2

0.0010

0.5
C3

0.0015

0.0020

b
0.04
.03

.02

P.01

"-~

-0.01

-0.02

c/o = 0.25~

-0.03

cla = 0 . 5 0 - - - - - /

o.j o.o,

-O.Oq

c/a = 1 . 0 ~
-0.06

-0,07

Figure 2

(a), Trapeszo~dal plate w=th clamped and s~mply supported edges and subject to uniformly d=stnbuted load, (b), deflect=on profiles
along line AB, (c), variation of moment, M X, along AB, (d), variation of moment, My, along AB

~l = B2 = 1, B3 = 0

B2= 1, B] = B 3 = O
B3= 1, B, = B 2 = 0

for bending analysis of

plates under combined
transverse and in-plane
loads

The transverse deflection surface may be parametenzed

by
m

w(x, y)= ~ c,4~,(x,y)

for buckhng analysis of

plates

(7)

I=1

for free vibration analysis of

plates

3 2 = 3 , = 1 , 3t=O for free vibration ahalysis of

plates under the influence
in-plane loads
(6)

where c, is the unknown coefficient to be varied and the

Ritz function, ~,, may be taken as the product of the
boundary function, 4h, which serves as the basic function and polynomial function, f , I.e.

~,,(x, y) = f(x, y)0,(x, y)

Eng. Struct. 1993, Vol. 15, No 1

(8)

57

pb-2 Rayletgh-Rttz method for general plate analysts K M. Ltew and C M Wang
where X = ptco=a:be/D. X, = o/a-'t~ D and the elements 111 the matrices are

in whmh
II,

[,(x, y)]~'

0j(x,y) = H

(9)

I=[

+ R)f < I

where n~ is the number of supporting edges, P, is the

boundary equation of the/th supporting edge (Ftgure 1 )
and f~,, depending on the support edge condition, lakes

16)

+ 2(1 - V,~b; t~,,

on

If i th edge is free
if i th edge is simply supported
if ith edge is clamped

f~,=l
fl, = 2

M, = _.,R~)

17)

(10)

N. = N, _hZR
ram_., + Iv,"a 2nomlrc,f+ N . ab(ROlm
and the polynomial function f may be generated as
follows

+ R
1')
--q

18)

R{;U"= t'4 I3"+q0'(~'~1)][3'+'4~;(~i~)

3~pO~ q
0~/3 ~

f(x, y) = x~y ' cos: 71"0- r 2 - 1 )

d,4

19)

2
_ ~ 7r(t--r

+ x'y' sin-

2-

1)

' ,,{*

(11)

O' = t

tip

pO,(~, 7I)M* + ~

p,O,(~,, ~1,)

(20)

where
r = f~/~t-

S =

E[

where d,4 = d~drl. For bending (deflection) analysis,

/33 -- 0, the simultaneous equations in equation (15) are
solved for [c] and back substituted into equation (7) for
the deflection surface In the case of buckling and vthration analyses, /3, = 0, the eigenvalues are obtained by
solving the set of homogeneous equations (7)

(12)

11
-

7r(l-

1
r2 1

r 2-

1)

COS 2

2
t-

, ~-(t-r
sin-

1)

(12)

[ ] denotes the maximum integer function, for example

Ix/2] = 1. Note that 4h ensures that the Ritz functions
satisfy the kinematic boundary conditions. The number
of polynomial terms can be taken to any m number,
rather than the conventional procedure of generating the
polynomial terms in a spectral set In previous studies by
Liew and Lain 6 7, the polynomials are orthogonalized
The orthogonabhzaUon process is omitted in the current
work, because it makes the pb-2 Rayleigh-Ritz method
more cumbersome with perhaps a minimal advantage in
obtaining a standard eigenvalue problem
For convenience, the coordinates will be normahzed

2.

1.!

a
I

-O

D
H
4~

5=]

L
O

x
~=-;
a

7=-

(13)

where a 1s the maximum diameter of the plate m the xdirection and b the maximum diameter in the y-directton
(see Figure 1).
Applying the Raylelgh-Ritz method

OF
--=0,
Oc,

t = 1, 2,

,m

Substituting equations (1)-(13)

yields

0=60
0.I

O3
r"

8=45

(14)
Into equation

( [ K ] - ),,~fl~ [ M ] - X,/3 2 [ N ] ) [ c ]

(14)

- / 3 , [Q} = 0
(15)

58

0=90
8=75

C
13

Eng. Struct. 1993, Vol. 15, No 1

Ol
0.5

b
Figure 3

I
1.0

I
1 5

I
2 0

.l
2.5

Side r a t m , a/b

(a), S k e w plate w i t h free and simply s u p p o r t e d edges

and sublect to u n i f o r m c o m p r e s s i o n , (b), buckhng factors for
various side ratios and s k e w angles

Rayleigh-Ritz method for genera/plate analysis: K. M. L#ew and C. M. Wang

pb-2

oil

~lll

__1~

_.J

//

\\

,=~,,
I
!

I
I

\\

"-_:---c-'-J

"~'1I?1 ((~
11111!
iIl.II
/ sr;ll
JJ'#l i~%~,_...
-/if
_~.. / #

r-m- -rr~
~1~6

2R

~2 ~

I0

./(ir] =0

~]~ 17.]0

gq
SO

ol(2r)~O 15

xl "io.zz
b

~= = zl .z~

,,,!.ii

,,,,,, ,;~,

;-z~ zl .z6

lilllrrl = 1.00

Figure 4

(a), Truncated circular plate with clamped circular per=meter and free straight edges, (b), first three v~bratmn modes and
natural frequencies for var0ous plate dimensions

Eng. Struct.

1993,

Vol.

15, No 1

59

pb 2 Rayletgh

Ritz method for general plate analyms K M Llew and C. M Wang

Numerical examples

Bemhng anah,st ~ oJ tlapejmhd plate under tottfl~rmly

dt,~tJvbttled load
Consider a trapezoidal plate in Figure 2a under a
uniformly distributed load, q Two opposite sides of the
plate are simply supported while the other two are
clamped The deflections and moments are to be deterlnlned The basra funct,on for the plate is

~ = [ x _ ~ _ O 2 5 a 2]

e[

~e

_ (!b - c)<2_,- + a)
\

4a

+ 2c) 2]

(21)

Subsntutlng the basic function into equatmn (8) and

applying the pb-2 R a y l e i g h - R i t z method, results were
obtained for v = 0 3, b/a = 1.0 and c/a = 0.25, 0 50
and 1.00 The deflection profiles along the line AB of
Figure 2a are presented in Figure 2b and the moments,
M, = -D(O2w/Ox 2 + vO2w/Oy'-) and M, = -D(O~-w/
3v 2 + uO2w/Ox2) are given In Figures 2c and 2d, respectwely The maximum deflection and moment values for
c/a = 1.0 are in total agreement with those given in
Reference 1 (pp 187)

Buckhng anah'st,s oJ skew plate under uniform

compressmn
Consider a skew plate
uniform In-plane forces,
and the other two sides
load is to be determined

as shown
two edges
are free.
The basic

In Ftgure 3a under
are simply supported
The elastic buckling
funcnon for the plate

IS

~, = [3' - x

(22)

tan 0] [3' - (a - a)tan 0]

of the buckhng intensity factor k = ofiet

sln40/(rc2D) (where /~ = b sec 0) with respect to side
raUos are presented in Figure 3b for various skew angles
Results

of the plate, 0 = 90 , 75, 60 , 45 and u = 0 3. The

results for the special case of 0 = 0 are confirmed by
independent numerical solunons tabulated in Reference
8

Vibration anah'sts of truncated circular plate

Consider a truncated circular plate as shown In Ftgure
4a. The circular perimeter is clamped while the straight
edges are free Its natural frequencies are to be determined The basic function for the plate is
~1 :

(23)

( X2 -~- ~)2 -- r 2 ) 2

where r is the radms of the mrcular plate. Applying the

pb-2 R a y l e i g h - R i t z m e t h o d , the vlbratmn modes and

natural frequencies, X = X~r 4/(ab)2, for various values
of a/(2r) = 0.50, 0 75, 1.00 are obtained. Figure 4b
shows

60

the

first

three

vibration

modes

Eng. Struct. 1993, Vol. 15, No 1

and

the

corresponding frequencies [or these plate dllnen,,ions

For the special case ol a/(2r) = 1 00, corresponding to
a clamped circular plate, the results agree with the exact
solutions of X~ = 1022. X~ = 2 1 26 and ;~,~:-21 26
(see Reference 9)

Conclusions
The paper presents a unified functional and its associated
set of simultaneous equations for bending, buckling and
vibration analyses of elastic thin plates. The method of
solution is via the recently developed pb-2 R a y l e l g h Ritz method The advantages of the method are that
firstly, it does not require mesh generation as in FEM,
and secondly, ~t uses a relatively smaller memory space
than the FEM (For example, in the buckling problem
above, only 40 terms of polynomial are used which
means that the matrix size in the elgenvalue problem is
only 40 40 In the FEM, the minimum number of
degree of freedoms for a plate element is 9 if a triangular
plate element with three degrees of freedom per node
was adopted For the FEM to have the comparable
matrix size to that of pb-2 method, it would mean that
it can only use at most 5 triangular plate elements It xs
doubtful that such a coarse mesh could furnish an
accurate critical load)
Thirdly, it does not reqmre dlscrenzation of curved
boundaries since it uses the exact boundary expressions;
this preserves the actual shape of the plate and thus provldes more accurate solutions
Fourthly, it improves on the Lagranglan mulUpller
method in terms of computaUonal effort because by
implicitly embedding the boundary equations into the
Ritz functions, it keeps the matrix size small while the
latter increases the m a m x size by the number of
Lagranglan multipliers used
Finally, it should be noted that the pb-2 R a y l e l g h Ritz method can be applied to a w~de class of twodimensional engineering problems (e.g , prestressed flat
membrane problems, torsion problems, waveguide problems, etc ) which requires the minimization of a functional with respect to a scalar field subject to various
restraints at the boundaries_

References
1 Timoshenko, S P and Womowsky-Krteger S Theory ojplate~ and
shells, McGraw-Hill, New York, 1959
2 T~moshenko, S P and Gere, J M Theor~ of elaatu 3tabthtv,
McGraw-Hall. New York. 1961
3 Letssa, A W Vibration of plates, NASA Sp-160. 1969
4 Handboo~ oJ Structural Stabdtt3 (Ed Column Research Committee ot
Japan). Corona Pubhshmg Company, L t d . Tokyo, 1971
5 Babuska. I . Szabo, B A and Katz, I N The p-version ot the finite
element method', S I A M J Numet Anal, 1981 18, (3), 5 1 5 - 5 4 5
6 Llew. K M and Lam, K Y 'Apphcatlon of two-dimensional
orthogonal plate functmn to flexural vibration of skew plates'. J Sound
Vtbr 1990. 139, 12) 2 4 1 - 2 5 2
7 LJew, K M and Lam. K Y A 'Raylmgh-R~tz approach to
transverse vlbranon of ~sotroplc and amsotrop~, trapezoidal plates
using orthogonal plate functmns'. Int J Soh& Stru<t 1901.27, (21
189-203
8 Llbove, C Elastic stability, Chapter 44 of Handbook of engmeermg
mechamcs (Ed Flugge) McGraw-Hall, USA, 1962
9 Carrmgton, H 'The frequencies of v~bratmn of fiat c~rcular plates
fixed at the circumference'. Phd Mag 1925. 50. 1261-1264