Journal of Sound and Vibration (1973) 27(3),325-344

A FINITE ELEMENT STUDY OF THE VIBRATION OF

TRAPEZOIDAL PLATES

RUTH M. ORRIS AND M. PETYT

Institute of Sound and Vibration Research,

University ofSouthampton, Southampton S09 5NH, England

(Received 19 September 1972)

Two high precision, conforming, plate bending elements, one a quadrilateral and the
other a triangle, are used to investigate the free vibration characteristics of triangular
and trapezoidal plates. An account of the derivation of the element mass and stiffness
matrices is included. These elements are used first to obtain the natural frequencies and
modeshapes of rectangular, trapezoidal and triangular, simplysupported plates with height
to base ratios of 1·5. The results of this analysis are compared with those found by other
authors. The natural frequencies and nodal patterns of both simply supported and
clamped trapezoidal plates with height to base ratios of 6 are then examined,by using the
elements described. The ratio of 6 has been chosen as typical of the dimensions of control
surface ribs. Four shapes of rib are considered, a rectangle, two symmetric trapezoids,
and an isosceles triangle.

I. INTRODUCTION
Aircraft substructures such as ailerons and tail planes usually consist of two spanwise skins
separated by several chordwise ribs. Random pressure fields, such as those caused by jet
noise or turbulent boundary layer fluctuations, can induce considerable responses over
limited regions of the aircraft structure. Therefore, when designing the aircraft to resist
acoustic fatigue, it is important to be able to predict the dynamic response ofthe su bstructures.
The initial step in predicting this response is to understand the vibration characteristics of
the beams and plates constituting the control surface. As a first stage in this work, the free
vibration characteristics of the rib of a typical substructure have been examined.
The chord length of the control surface is, in general, very much greater than the separation
of the two skins. Therefore, the aspect ratio of the ribs separating the two surfaces is usually
appreciably greater than unity. These ribs may be trapezoidal or triangular in shape. Fre-
quently sections of the ribs contain cut outs to reduce the weight, and cleats'are attached to
increase the stiffness. However, as an initial step in understanding their vibration character-
istics, it is reasonable to neglect these discontinuities, and represent the ribs by isotropic,
flat, trapezoidal or triangular plates.
Unfortunately, the literature on the vibration of simply supported and clamped trapezoidal
or triangular plates is meagre (see reference [1], for example). Recently, Chopra and Durvasula
[2, 3] have investigated the vibration characteristics of simply supported, symmetric and
unsymmetric, trapezoidal plates. They utilize the relationship that exists between the eigen-
values and eigenvectors of simply supported plates and fully fixed membranes. This reduces
the differential equation of the problem from fourth order to second order. The deflection
surface is expressed in terms of a Fourier sine series in transformed coordinates, and the
Galerkin method is used to solve the membrane equation. However, Chopra and Durvasula
325
326 RUTH M. ORRIS AND M. PETYT

restrict their investigations to plates with height to base ratios of 2·5 or less. In addition their
method is only applicable to plates with simply supported boundaries.
In the present analysis, the finite element method is used. This method is very versatile, as
once the element mass and stiffness matrices are set up, they can be used to investigate the
vibration characteristics of any plate, by defining the boundary conditions and a suitable
element mesh. Thus, results for either clamped or simply supported boundaries can be
obtained easily.
Two conforming plate bending elements are used in this work, a general quadrilateral and
a triangle. The plates analysed are assumed to be thin, so that effects due to shear deformations
and rotary inertia can be neglected.

t
/Ga:
! I \
! I \
I \

a
Figure J. Definition of the parameters used to describe the trapezoidal plate.

These elements are used first to investigate the vibration characteristics of simply supported
symmetric trapezoidal plates with height to base ratios of 1·5. The ratios of b to a (defined in
Figure 1) are chosen to be 1·0 (rectangular), 0·4 and 0·0 (triangular). These cases are among
those considered by Chopra and Durvasula [2]. Thus it his been possible to check the con-
vergence of the finite element analysis. Another aim of this preliminary investigation is to
determine a suitable mesh to use for analysing the trapezoidal and triangular plates. The
information gained from this analysis is then used to obtain mode shapes and natural
frequencies for trapezoidal plates with height to base ratios of 6 to 1. This ratio has been
selected as being typical of the dimensions of control surface ribs. Both simply supported and
clamped boundary condi tions are considered. Four values of the ratio b to a are investigated:
1·0 (rectangular), 0·8 and 0·2 (both trapezoids), and 0·0 (triangular).

2. ANALYSIS
2.1. CHOICE OF FINITE ELEMENTS
The two plate bending elements used in this project are based on the same assumptions as
the one by Fraeijs de Veubeke reported in references [4] and [5]. The computer routines
developed are for a general quadrilateral element and a triangular element. Both these elements
possess three degrees of freedom at their corner nodes, an out of plane displacement and two
rotations about the x and y axes of the global Cartesian system of coordinates. There are
additional nodes situated at the mid-points of the sides of the elements, and these nodes each
have a single degree of freedom, a rotation about the element side. At the interface between
elements the conditions of continuity of deflection and slope are satisfied.
 VIBRATION OF TRAPEZOIDAL PLATES 327
The quadrilateral element is obtained by a combination of the cubic deflection fields for
each of the four triangular regions defined by the edges of the quadrilateral and its diagonals.
These deflection fields are expressed in terms of the area coordinates of the triangular sub-
regions, instead of the oblique coordinate system defined by the diagonals of the quadrilateral
that is used by de Veubeke. The derivation of the displacement fields for the triangular sub-
regions, by using area coordinates, is identical to the analysis followed by Clough and
Felippa in deriving their refined quadrilateral element [6].

2.2. DERIVATION OF DISPLACEMENT FUNCTIONS FOR THE SUBREGIONS BY USING AREA

COORDINA TES
The area coordinates, ~1' ~2' ~3 of a point P inside the triangular region may be defined as
the ratio

(I)

where Al is the area of the subtriangle subtended by the pointP, and A is the total area of the
triangle. The area coordinates are related by the condition

(2)
The relation between the global cartesian system and the triangular coordinates may be
expressed as

(3)

where Xl> Yl are the coordinates of the corners of the triangle. Relation (3) can be inverted
to give

(4)

where 2Aij = xIYj - XjYt> aj = Yj - Yk and b, = Xk - Xj'

Within the triangular subregion, 1, shown in Figure 2, the cubic deflection field may be
expressed in area coordinates as

11'(1) = L~f, ~~, ~~, ~r ~2' ~1 ~3' ~~ ~3' el e~, el ~~, e2 ~i, el ~2 (3J {CX~I)}, (5)
= LW~l) J{IX~1)}, (6)
where {IX)!)} is a 10 by I column vector of constants. These ten constants are evaluated in
terms of the deflections, WI' 11'2 and 11'0, and the rotations e.~l' By l , 8x 2 , 8y2 , Oxo and OyO, at nodes
1,2,0 and the single rotation Os at node 5, of the triangular subregion. This procedure yields
the relation
(7)

where {wl ll }' = LwoOx08yoWI8xI8YlW28:<28Y28sJ, and [C(l)] is a 10 by 10 matrix composed of

terms that are constants for the subregion considered. Equation (7) may easily be inverted to
yield the expression for the constants {<x\ll}:

(8)
328 RUTH M. ORRIS AND M. PETYT

Substituting equation (8) into (6) gives the deflection field within the subregion I as
w(l) = l W~ll J[C(llt 1{wlll} (9)
or
(9a)

The matrix [4>(1)] is identical to that given by Clough and Felippa [6] when appropriate
changes in notation are made.

C
1'6j:-';;
r

Figure 2. Quadrilateral plate-bending element.

2.3. ELEMENT STIFFNESS MATRIX

For an orthotropic plate, the strain energy in bending is given by T imoshenko and
Woinowsky-Krieger [7] as

(10)

where D y = E~ t 3/12, D; = E; (3/12, D 1 = E" (3/12 and D x y = Gt 3/12. The values of the four
constants, E~, E;, E" and G, depend on the elastic properties of the plate. For the particular
case of isotropy, E~ = E; = E/(1- y2), E" = vEj(l - y2) and G = E/2(1 + v),
Equation (10) may be written in the form

v=±JJ {1' }T[D]{I'}dxdy, (11)

Area
where
02W i}2w iFWJ

and
l
{y)T= - , - - , -
ax2 oxoy ay 2
 VIBRATION OF TRAPEZOIDAL PLATES 329
Let the curvature matrix for a triangular subelement, i, be given by
02 w(1) iF w(l) _ J.
02 w(1)
{'l'(I)}T= _
l _,_
ox2
_ ,_
oxoy oy2
(12)

By using expression (9) and differentiating, this may be written in the form

1
{yO)} = (2A(f)? [~I[WI] + ~2[W2] + ~3[W3]][C(l)tl{wll)}, (13)

where [W d, [W 2] and [W3] are each 3 by 10 matrices of constants.

Relation (13) may now be substituted into equation (11) to give the strain energy in bending
for subregion i as

V O ) =~ ff {y(f)}T[DHy(l)}dxdy. (14)
Areat

If the bending rigidity, D, remains constant over the area of the triangular subregion the
terms of the form

f f ~J~k[D]dxdy= [D] J J ~J~kdxdy

Area, Area ,
(15)

can be evaluated. When these integrations are carried out, equation (14) can be simplified to
yield
(16)
where
I 1
[K(n] = 24 . (2A(f»)3 . [C(l)]-T{[W I2]'[D][Wd -\- [W 23]'[D][W23] -\- [W31 ]'[D ][W 3 1 ]}[C(I)]- 1

and [Wlj] = [Wtl -\- [W J ] .

The stiffness matrices for the four subelements, defined by equation (16), are combined
into a 19 by 19 matrix for the complete quadrilateral, K T , defined by
4
{WT}'[KT]{WT} = 2: {w~O}'[K(I)] {w~I)}. (17)
1

where {WT}' = LWoexoe yOWI eXl ej'l W2 e x 2 ()Y2W3 e x3 e y3 W4(),,4eY4()S e 6e7 eBJ and {W~l)} is defined
by equation (7).
Finally, the additional three degrees of freedom at the internal node, 0, introduced in
deriving the displacement fields in the triangular subregions, are eliminated by establishing
slope compatibility between these subregions. This is done by enforcing continuity of normal
slope at points midway along the sides of adjacent su bregions: that is, at nodes 9, 10 and 11,
shown on Figure 2. For example, the normal slope at node 9, between triangles 1 and 2, is
given in terms of the coordinates of triangle 1 by
e~l, = Lb~I)J{W~lJ}. (18)

In terms of the local coordinates of triangle 2, the normal slope at node 9 is given by
e~2) = Lb~2J J{W\2J}. (19)
For internal compatibility between regions 1 and 2, the normal slopes at node 9 must be
equal: that is,
(20)
330 RUTH M. ORRIS AND M. PETYT

Equations (18), (19) and (20) may be combined to give

[B(9l]{WT} = O. (21)

Similar expressions can be derived by considering slope compatibility at nodes 10 and 11

between triangles 2 and 3, and 3 and 4, respectively. The three equations obtained can be
combined in the matrix expression
(22)
where [BJT = [B(9lB(lOlB(lll] and {WT} is defined with equation (17).
The terms in equation (22) corresponding to the internal node, 0, can be partitioned off to
yield

[Bo ,: B]
I n
{W---
o}
w = 0 (23)
n

{Wo} = _[BO]- 1[Bn]{wn}, (24)

= [L]{wn } , (24a)

where [LJ is a 3 by 16 matrix of terms that are constants for a particular element. The 19 by 19
element stiffness matrix, [KT], defined by equation (17), can then be reduced to the final 16
by 16 element stiffness matrix, [Kn], by the transformation
[Knl = [N]/[KT1[N] (25)
where [NT] = [L:I], and [I] is a 16 by 16 identity matrix.

2.4. ELEMENT MASS MATRIX

The consistent mass matrix may be derived by considering the kinetic energy of the plate
in bending,
T=tpt f J (w)2dA. (26)
AJ'ea

Here, the dot denotes differentiation with respect to time. For harmonic vibration of angular
frequency W, this becomes

T= tpro t
2
f Jw 2
dA. (27)
Area

Within the subregion i the kinetic energy, TW, is then given by

T(l) = !PW 2 t f J W(l)1 dA. (28)

At'./

Upon using relation (9) this becomes

t-» = -tw2{wlll}T[M(ll]{w~/)}, (29)
where
[M(l)] = pt[C<Il]-T J f {W~YLW~J dxdy[C(I)]-l. (29a)
Area,
 VIBRATION OF TRAPEZOIDAL PLATES 331
The integrals over the area of the subregion can be evaluated to yield

(30)

where [m] is a matrix of numerical constants.

The procedure for combining the mass matrices for the four subregions, and eliminating
the additional degrees of freedom introduced at the internal node, is identical to that used to
obtain the element stiffness matrix.
The analysis used to derive the mass and stiffness matrices for the triangular plate-bending
element, shown in Figure 3, is similar to that used in deriving the element matrices for the

c:
T~=-1r
y

Figure 3. Triangular plate bending elemen t.

quadrilateral. However, in the case of the triangle there are three triangular subregions, and
these are defined by the centroid of the element.

2.5. ASSEMBLY OF THE ELEMENTS AND EIGENVALUE PROBLEM

Once the mass and stiffness matrices for the individual elements have been evaluated these
matrices can be assembled to form, respectively, the mass and stiffness matrices for the
complete structure. At this stage the boundary conditions of the problem are applied by
constraining some degrees of freedom at external nodes to be zero.
During the assembly procedure some care is necessary to ensure continuity in direction of
the slope normal to the edge between adjacent elements. When quadrilateral elements are
joined this continuity is obtained by reversing the direction of the normal slopes at the mid-
side nodes for two adjacent sides of each element. The nodes chosen in this work are 5 and 6,
in sides I and 2. Provided that the element nodes are numbered in the same direction as the
element numbering in the plate (that is, in either a clockwise, or an anticlockwise direction),
the direction of the normal slope between elements is then always consistent. For the meshes
shown in Figures 6(b), 7(a), 7(b) and 9(d), triangular elements are combined with quadri-
lateral elements. In these examples it is necessary to reverse the sign of the slope normal
to the edge of the triangle at nodes 4 and 5, in sides I and 2 of the element, to obtain
inter-element continuity of normal slope.
With three degrees of freedom at a nodal point, the boundary conditions applied along a
simply supported edge require that the out of plane displacements, and the slopes tangential
332 RUTH M. ORRIS AND M. PETYT

to the boundary, be zero. Thus, for the simply supported trapezoidal plate, shown in Figure 4,
wand By are zero along edges AD and Be. The degrees of freedom at the corner nodes of the
elements are referred to the global system of axes, x, y, z, Therefore, to apply the condition of
zero tangential slope along edges AB and CD, it is necessary to rotate the reference axes for
the element corner nodes situated along these edges from the global x, y, z system, to the
x', y', z system, through the angle 0:: Here, 0: is the semi-vertical angle of the triangle obtained

D A

c B

Figure 4-. Trapezoidal plate with global (x,y) axes and local (x',y') axes-indicated.

by extending the sloping sides of the trapezoid. The required transformation for each of the
nodes concerned is given by

(31)

or
(32)

The transformed element stiffness matrix, [Kt,,], is then given by

(33)
where

(34)

and the position of A in the matrix is determined by the number, i, of the element node
constrained.
The boundary conditions applied along a clamped edge are that the deflection, the
tangential slope, and the rotation about the edge are all zero. Thus, boundary nodes possessing
three degrees of freedom have all three freedoms constrained. In addition, the single, rota-
tional, degree of freedom contributed by nodes mid-way along element sides, is also zero on
a clamped boundary.
 VIBRATION OF TRAPEZOIDAL PLATES 333
If {wn is the vector of nodal displacements for the complete, constrained, system, these
are related to the nodal displacements of the individual elements by the relation
{w:} = [UO]{wn. (35)

The matrices [vcl are defined by the element node numbering, and the constraints applied to
the system. The overall mass matrix, [M], and the overall stiffness matrix, [K], obtained by
assembling the mass and stiffness matrices for the individual elements and applying the
constraints of the problem, are then given, respectively, by
[M] = 2 [Uo]T[M~][U0],
e

[K] = 2 [UoF[K~][U°]· (36)

e

In the absence of external forces on the system, the requirement that the total energy of the
system, L = T - V, is a minimum yields the characteristic equation
([K] - Q)2[M]){wf) = O. (37)
The solution of equation (37) for the natural frequencies, ro, and the corresponding nodal
displacements of the complete system, {wf}, is a standard eigenvalue problem.
The size of the eigenvalue problem can be reduced by taking the symmetry of the trapezoid
into account. Thus it is only necessary to analyse one-half of the plate. For each type of
boundary condition, clamped or simply supported, there are two possible modes of vibration
of the trapezoid plate, symmetric or antisyrnrnetric with respect to the axis aa', shown in
Figure 4. When the plate vibrates in a symmetric mode, the normal slope along the edge aa'
is zero. The constraints for this case are 8y = 0 at element corner nodes along this boundary,
and 0" = 0 at mid-side nodes. When the plate is vibrating in an antisymmetric mode, there is
a nodal line along the axis aa', This boundary is then effectively simply supported, and the
constraints that apply are those for a simply supported edge.

3. RESULTS
3.1. TRAPEZOIDAL PLATE WITH A HEIGHT TO BASE RATIO OF 1·5
Various element configurations have been considered when analysing the simply supported
trapezoidal plate with a height to base ratio of 1·5. The different divisions used are shown in
Figures 5, 6 and 7. The non-dimensionalized frequency parameters obtained for the first
six modes, when using these meshes, are given in Table I. Results are presented for the three
cases bla = 1'0, 0·4 and 0·0. For comparison, the exact theoretical result for the rectangular

Figure 5. Element mesh used to describe one half of the simply supported rectangular plate, h/a - 1'5.
334 RUTH M. ORRIS AND M. PETYT

(0) (b)

Figure 6. The two different element meshes used to describe one half of the simply supported trapezoidal
plate, hla = 1-5, bla ~ 0'4. (a) Mesh of 12quadrilateral elements; (b) mesh of 12quadrilateral and 2 triangular
elements .

(0 ) (b)

Figure 7. The two different element meshes used to describe one half of the simply supported triangular
plate, hla = 1'5, b ]« = 0'0. (a) Mesh of 10 quadrilateral and 2 triang ular elements; (b) mesh of 9 quadrilateral
and 3 triangular elements .

TABLIl I
Comparison between the frequencies A= m(d 2 Irr 2)(ptID)+ ofsimply
supportedplates with hla = 1'5

Mode 2 3 4 5 6

(a) bla = 1·0

12 square elements 3'2505 6·253 10'029 11'279 13·033 18·070
(Exact) theory] 3·25 6·25 10·0 11'25 no 18-0
(b) bla = 0'4
12 quadrilateral elements 5'3927 9·438 14·744 15·964 21·911 23·250
12 quadrilateral and 2
triangular elements 5·4616 9·463 14'753 16·146 21·968 23'267
Chopra and Durvasulaf 5·3896 9'424 11·685 15-911 21·700 23'146
(c) bla = 0'0
10 quadrilateral and 2
triangular elements 7·4685 14·927 20·026 24'440 32·862 36·611
9 quadrilateral and 3
triangular elements 7·7755 15·161 20·418 25-400 33-295 37'517
Chopra and Durvasulaj 7·4630 14'877 19'935 24'183 32-523 35'403

t Reference [1].
:t Reference [2].
 VIBRATION OF TRAPEZOIDAL PLATES 335
plate, and the results obtained by Chopra and Durvasula [2]for the trapezoidal and triangular
plates are also included in Table 1.
An exact theoretical solution for the natural frequencies and normal modes of the simply
supported trapezoidal plate is available only for the rectangular case, when bla = 1·0. For
this case the quadrilateral element is used as a square, and the mesh, consisting of 12 elements,
is shown in Figure 5. The values obtained for the frequency parameters of the first six modes
are shown in Table I. From these results it is apparent that the analysis with only 12 elements
yields values for the first six frequencies that differ by less than O· 5 %from those given by the
exact theory. The nodal patterns obtained by this analysis coincide with those predicted
theoretically.

a/o Mode
( 2 :3 4 5 6

10 D EJ rn § E8 ~
3·25 6'25 10·03 (('28 18·07

0·4

5·39 9044 14·74 15·96 21·91 23'25

0·0

7'47 14·93 20'03 24'44 32'86 36'61

Figure B. Mode shapes and frequency parameters, A., obtained for the first six modes of the simply supported
plates with hla ... 1'5.

The finite element method described here, and the Galerkin method as used by Chopra
and Durvasula, are both equivalent to the Ritz method. Either method of analysis yields
frequencies which are upper bounds. Thus from Table I it is apparent that the results obtained
by Chopra and Durvasula for the trapezoidal and triangular plates are slightly more accurate
than those obtained from this finite element analysis.
The values found for the non-dimensionalized frequency parameters of the trapezoidal
and triangular plates when the element configurations labelled (a) in Figures 6 and 7 are
used, are generally lower than those values obtained when the meshes labelled (b) are used.
This is apparent from the results presented in Table 1. The former arrangements of elements,
(a), give values for the frequencies of the first six modes of the trapezoidal plate (bja = 0-4),
and the first five modes of the triangular plate (b]« = 0,0) that differ by less than I %from the
frequencies presented by Chopra and Durvasula [2]. The second element configuration used
for analysing the trapezoidal plate (Figure 6(b)) yields slightly higher frequency values than
the first configuration (Figure 6(a)), but the results are still within 1 %of those obtained by
336 RUTH M. ORRIS AND M. PETYT

Chopra and Durvasula for this case . However, it must be noted that 14 elements are used in
this mesh (Figure 6(b)), compared with only 12 in the previous configuration (Figure 6(a)).
For the triangular plate, both meshes (F igur es 7(a) and 7(b) ) consis t of 12elements. The values
found for the frequencies of the first five modes, when using 12 elements in the arrangement
labelled (b), differ by about 2 % from those found by Chopra and Durvasula. Thus, for the
triangular plate, when 12 elements are used in the analysis, the values obtained for the
frequencies can vary by at least 1 %when the particular arrangement chosen for the elements
is altered.
The nodal patterns found for the first six modes of the trapezoidal plate, with bla = 0'4,
and for the triangular plate, with bla = 0,0 , are shown in Figure 8. They are similar to the
nodal figures presented by Chopra and Durvasula [2].
The nodal patterns shown in Figure 8 account for the discrepancy observed between the
frequencies found for the triangular plate when the 12 elements used are arranged in the two
different configurations shown as Figures 7(a) and 7(b). When the nodal patterns are com-
pared with these configurations, it is apparent that the pattern defined by the element
boundaries for arrangement (a) in Figure 7 is closer to the nodal pattern of the plate than
that for arrangement (b) . This explanation is also applicable to the results obtained with the
two different meshes used in analysing the trapezoidal plate (Figures 6(a) and 6(b)).

3.2. TRAPEZOIDAL PLATES WITH A HEIGHT TO BASE RATIO OF 6

The preceding analysis sho ws that a mes h of 12 quadrilateral and tr iangular elem ents can
be used to obtain satisfactory values for the natural frequencies and mode sh apes of a simply
supported trapezoidal plate with a height to base ratio of 1·5. However, the aim of this study
is to investiga te the natural frequencies and normal modes of a trapezoidal plate with an

TABLE2
Comparison between the frequencies A. = w(d 2 /n 2)(pt{ D)!' ofsimply supported and
clamped rectangular plates with hJa = 6

Mode 2 3 4 5

(a) Symmetric modes of the simply supported plate

12 rectangular elements 37·005 40·032 45·220 52-843 63·296
Exacttheoryt 37 40 45 52 61
(b) Antisymmetric modes of the simply supported plate
12 rectangular elements 145-354 148,290 153'565 161·736 173'512
Exact theory'[ 145 148 153 160 169
(c) Symmetric modes of the clamped plate
12 rectangular elements 82-409 84·673 89·001 96·289 109'175
24 rectangular elements 82·295 84·385 88·176 94·116 102'810
Upper and lower values
obtained 82-30 84·71 88·80 94·90 103'3
from Leissat 82-20 83-98 87'14 91·77 98'03

(d) Antisymmetric modes of the clamped plate

12 rectangular elements 227·456 230'336 235-883 245'305 260'822
24 rectangular elements 226·195 228·857 233'599 240·903 251·518
Upper and lower values
obtained 226'0 228·8 233·5 240·4
from Leissat 225·7 228·0 231-6 236·9

t Reference [lj.
 vmRATION OF TRAPEZOIDAL PLATES 337
aspect ratio of 6,0, and to consider both simply supported and clamped boundary conditions.
For this latter case, a reasonable deduction to make is that a good initial estimate of the
natural frequencies and nodal patterns can be obtained by using the arrangement of elements
found most satisfactory in the preceding analysis. This deduction follows from the statement
of Chopra and Durvasula [2], that the nodal patterns for trapezoidal plates with height to
base ratios of 2·5 are similar to those for the plates with this ratio equal to 1·5.
An exact theoretical solution is available for the rectangular plate with simply supported
boundaries (see reference [1], for instance). The values given by the theory for the frequency
parameters of the plate with an aspect ratio of 6 are shown in Table 2. These can be compared
with the frequencies obtained by using 12 rectangular elements, in the arrangement of Figure
9(a), also shown in Table 2. From this comparison it is apparent that the analysis using 12

(a) (b) (e) Cd)

Figure 9. Meshes of 12 elements used to describe the clamped and simply supported plates with hla » 6·0.
(a) Rectangular plates; (b) trapezoidal plate, bla - 0·8; (c) trapezoidal plate, b]a = 0'2; (d) triangular plate.

elements gives values for the frequencies of the first three symmetric, and the first three anti-
symmetric modes, that are within I % of the theoretical results. The nodal lines found from
the analysis with a mesh of 12rectangular elements coincide with those predicted theoretically.
Theoretical values of the frequency parameters of clamped rectangular plates with various
aspect ratios are given by Leissa [1]. Upper and lower bounds for the frequency parameters
of the clamped plate with an aspect ratio of 6 may be obtained from this source. These bound-
ing values are shown in Table 2. When a mesh of only 12 elements is used, the frequencies
found by the analysis are approximately 1 %above the upper bound value.With 24 rectangular
elements arranged as shown in Figure IO(a), slightly lower frequency values are obtained.
For the symmetric modes these frequencies are slightly below the upper bound value. For
the antisymmetric modes, frequencies within 0·2 % above the upper bound, and within 2 %
of the lower bounds shown are found.
Initially, meshes of 12 elements have been used to describe the trapezoidal plates, with bla =
0·8 and 0'2, and the triangular plate, with bja = 0·0. The element configurations used are
shown in Figure 9(b), (c) and (d). The results obtained from these analyses, for both the simply
supported and clamped boundary conditions, are given in Table 3. For the triangular plate
with clamped boundaries, however, a mesh of 12 elements does not yield sufficient nodal
degrees of freedom, after the constraints have been applied, to define the nodal patterns of the
338 RUTH M. ORRIS AND M. PETYT

3rd, 4th and 5th symmetric and antisymmetric modes uniquely. Therefore, additional analyses
have been carried out, in which meshes of 24 elements were used. Two different arrangements
of these elements have been considered. In one the elements are arranged in threes across the
semi-breadth of the plate, to give 8 elements along the plate length, and in the other the
elements are arranged in pairs to give 12 elements parallel to the longer edge of the plate.
These element configurations are shown in Figure 10. The results obtained for the different
plates, when using these meshes, are given in Table 3. The particular mesh consisting of 24
elements arranged in threes has been used only for analysing the plate with clamped
boundaries. When the boundaries are simply supported some nodal displacements are
retained along the plate edge after the constraints have been applied. For these conditions
only two elements are needed to define the variation in nodal pattern across the plate breadth.
The results presented in Table 2 for the simply supported trapezoidal plate with bla = 0'2,
and the simply supported triangular plate, show that doubling the number of elements used

(b) (e) (d)

(i) (Ii) (i) (ii)

Figure 10. Meshes of 24 elements used to check the convergence of the results for the plates with h/a"" 6·0.
(a) Clamped rectangular plate; (b) clamped trapezoidal plate, bla ~ 0'8; (c) trapezoidal plate, bla » 0·2;
(i) clamped, (ii) clamped and simply supported; (d) triangular plate, bla ~ 0'0, (i) clamped, (ii) clamped and
simply supported.

in the analysis has caused a decrease of only 1 %in the frequencies found for the first symmetric
and antisymmetric modes. The reductions obtained in the frequencies of the higher modes
are greater. For the fifth antisymmetric mode of the triangular plate a reduction of nearly
40 % occurs in the frequency when the number of elements used in the analysis is doubled,
although for the third anti symmetric mode this reduction is only about 20 %. The results for
the trapezoidal plate, and the symmetric modes of the triangular plate, show smaller percent-
age decreases, of between I %and 20 %, when the number of elements is increased by a factor
of two.
When the number of elements used in the analysis is doubled the frequencies found for the
clamped trapezoidal plate with bja = 0·8 decrease by about I % to 3 %, depending on the
order of the mode. However, for the clamped trapezoidal plate with bla = 0,2, and for the
clamped triangular plate, the convergence of the results when the number of elements used
is increased is less satisfactory. The frequencies of the first symmetric and antisymmetric
modes decrease by 5 % to 10 %when the number of elements used is doubled, depending on
the element configuration used. The lowest frequency values are those found when the 24
 TABLE 3
Comparison between the frequencies A = ro(d 2In2)(pt ID)i: ofsimply supported and clamped trapezoidal plates with hIa = 6·0
No. of bla= 0·8 bla=0'2 bla =0,0
Mode elements Symmetric Antisymmetric Symmetric Antisymmetric Symmetric Antisyrnrnetric

(a) Simply supported plates

1 12 42·963 161'57 55·907 192'24 59·889 201·65
24: 12 x 2 55'584 190·73 59'323 199'40
2 12 48·928 176-07 76·147 242-30 84·997 268·11
24: 12 x 2 75·024 232·95 83-811 251-02
3 12 54·854 188·73 97·134 31H1 116·71 364·78
24: 12 x 2 94·669 273·30 109·05 301·00
4 12 62·454 200·73 131-18 377·75 166·80 447·83
24: 12 x 2 115·30 313·12 136·06 351·50
5 12 72·600 214·44 161·08 467·27 211·59 579·02
24: 12 x 2 137·27 354·17 165·74 412·37
(b) Clamped plates
1 12 93·682 253·22 121·20 314·30 130·82 335·90
24: 8 x 3 93·036 250·45 117·45 305-83 125'51 324'53
24: 12 x 2 114·37 293·63 121·35 307·13
2 12 102·71 273·21 154·16 391·61 ]73·00 439·19
24: 8 x 3 101'55 268·98 149'18 370·34 166'24 404·09
24: 12 x 2 141·42 346·86 154·97 372·79
3 12 111-09 291·36 190'55 476·35 230·16 560·67
24: 8 x 3 109'26 285'78 178·83 428·88 202'18 492·17
24: 12 x 2 168·37 398·18 ]89·60 437·90
4 12 119-61 308·47 244'55 583·26 302·19 732·33
24: 8 x 3 116·76 302-04 210'24 524'52 260·68 627·62
24: 12 x 2 196·46 450·43 226·91 505·82
5 12 128·66 326·39 292·05 707·22 402·49 960·48
24: 8 x 3 125-31 317·40 265·28 604·26 324·21 737·33
24: 12 x 2 226·48 504·70 266·36 578·90

w
w
ID
340 RUTH M. ORRIS AND M. PETYT

elements are arranged in two rows of 12 (Figure IO(d)(ii)). Further reductions could have
been obtained in the frequencies found for the clamped triangular plate if more elements had
been used in the analysis. The results presented here indicate that it would be preferable to use
these extra elements to improve the definition of the variation in nodal pattern along the axis of
/)/0 Mode
I 2 3 4 5

1·0

37·01 40·03 45·22 52'84 63·30

0·8

42·96 48·93 54·85 62-45 72·60

0·2

55·58 75'02 94·67 115·3 137·3

0'0

59'32 83·81 109·1 136·1 165·7

Figure 11. Nadal patterns and values of the frequency parameter, A, for the symmetric modes of the simply
supported plate, hla - 6'0.

symmetry ofthe plate. However, when the number of elements used in the analysis is increased,
the size of the eigenvalue problem increases, and difficulties arise due to increased computer
time and storage requirements. Because of these problems, and the slow convergence of the
results for this long triangular plate, the value of additional analysis with more elements is
doubtful.
The nodal patterns obtained for the different plates from these analyses are shown in
Figures 11,12,13 and 14. One effect of the large value chosen for the height to base ratio of the
 VIBRATION OF TRAPEZOIDAL PLATES 341
trapezoid is to cause a considerable separation in the frequencies of the first symmetric mode
of the plate, and the first antisymmetric mode. The effect occurs for both simply supported
and clamped boundary conditions. The separation in the frequencies of the two types of modes
is greatest for the rectangular plate. For this plate the first five symmetric modes all have

b/o Mode
I 2 3 4 5

1'0

145-4 148<3 153·6 161-7 173·5

0-8

176'1 188'7 200·7 214-4

0'2

190'7 232·9 273-3 354'2

0-0

19904 251-0 301-0 351·5 412-4

Figure 12. Nodal patterns and values of the frequency parameter, it, for the antisymmetric modes of the
simply supported plate, hla ~ 6'0.

appreciably lower frequencies than that of the first antisymmetric mode. From Figures 11,
12, 13 and 14, it is apparent that as the ratio bla decreases, the separation between the fre-
quencies of the first symmetric and the first antisymmetric modes also decreases. However,
it is apparent from the results presented that for all the plates examined the frequency of the
fifth symmetric mode is lower than that of the first antisymmetric mode.
A second effect apparent from the results presented in Table 3 and Figures 11, 12, 13 and 14
is that the frequencies of each mode increase with decreasing b{a ratio. This effect becomes
342 RUTH M. ORRIS AND M. PETYT

more noticeable for the higher mode numbers. The frequencies of the first symmetric and
antisyrnmetric modes of the triangular plate, for both clamped and simply supported
boundaries, are roughly one and a half times the corresponding frequencies for the rect-
angular plate. For the fifth symmetric and antisymrnetric modes the frequencies of the tri-
angular plate are nearly three times those of the rectangular plate.

a/a Mode
I 2 3 4

1·0

82'29 84·38 88·18 94'12 102'8

0·8

93·04 101'5 109·3 116'8 125·3

0·2

114·4 141·4 166-4 196·5

0'0

121'3 155·0 189·6 226·9 266-4

Figure 13. Nodal pa tterns and values of the frequency parameter, 1, for the symmetric modes of the damped
plate, hla = 6·0.

The nodal patterns for both the clamped and the simply supported, large aspect ratio,
trapezoidal plates, resemble those obtained for the first five modes of the trapezoidal plate
with a height to base ratio of 1·5. For the plates with simply supported boundaries, one change
in the nodal pattern caused by a decrease in the ratio bla is a shift of the nodal lines towards
the base of the trapezoid (Figure 11). A second effect that is apparent is the curving of the
nodal lines parallel to the shorter edges of the plate. The amount of curvature is greatest for
 VIBRATION OF TRAPEZOIDAL PLATES 343
the triangular plate. However, for this large aspect ratio plate, the curving of the nodal
lines is a small effect.
The effect of clamped boundaries on the nodal pattern of the rectangular plate is to cause
the nodal lines parallel to the shorter edges to move inwards slightly from the nodal positions
for the simply supported plate. A comparable effect occurs for the trapezoidal plates. The

hlo Mode
1 2 3 4 5

)'0

226·2 228·9 233'6 240·9 251·5

0'8

250-4 269·0 285·8 302·0 317·4

0'2

293·6 346·9 398·2 450·4 504·7

0·0

307'1 372·8 437·9 505·8 578'9

Figure 14. Nodal patterns and values of the frequency parameter, 1, for the antisymrnetric modes of the
clamped plate, hla - 6'0.

nodal line nearest the base of the trapezoid plate occurs slightly further from the base.when
the boundaries are clamped than when they are simply supported. When there are two or
more nodal lines across the plate, the nodal line nearest the top of the trapezoid also occurs
nearer the base when the boundaries are clamped. Thus, the separation between the crosswise
nodal lines of the clamped plates is less than that between the corresponding nodal lines for
the plate with the boundaries simply supported.
344 RUTH M. ORRIS AND M. PETYT

4. CONCLUSIONS
Two conforming plate bending elements, a general quadrilateral and a triangle, have been
used to investigate the vibration characteristics of rectangular, trapezoidal and triangular
plates with height to base ratios of 6·0. Two different trapezoids have been considered, one
approaching a rectangular shape, with bla = 0,8, and the other tending to a triangular shape,
with bla = 0·2. The effects of both simply supported, and fully fixed boundaries, have been
considered.
The generally high precision performance of the two elements described is confirmed by the
results presented. However, it has been observed that the results converge more rapidly as
the number of elements employed in the analysis increase if the quadrilateral and triangular
elements used have moderate aspect ratios of not more than about 10 for the triangle or 20
for the quadrilateral. Thus, the results obtained for the trapezoidal plates with hla = 6·0 and
bla = 0·8 exhibit better convergence characteristics, as the element mesh is made finer, than
those for the triangular plate, or the trapezoidal plate with bla = 0·2.
When the ratio bla is decreased, an increase occurs in the frequency parameters of the
plates. The frequencies of the first symmetric and antisymmetric modes of both the clamped
and simply supported triangular plates, with tiia = 6'0, are nearly twice those of the corres-
ponding rectangular plate. For all the ratios of bla considered, when the height to base ratio
of the plate is 6,0, the first antisymmetric mode of the plate has a higher frequency than the
fifth symmetric mode.
The number of nodal lines present in the vibration pattern of any of the first five symmetric
or antisymmetric modes of the triangular or trapezoidal plates is equal to the number of
lines present in that particular mode for the simply supported rectangular plate. However,
as the ratio bja decreases, from 1·0 for the rectangle, the separation between nodal lines
crossing the axes of symmetry decreases, and the lines begin to curve slightly. In addition,
these nodal lines across the breadth of the plate move nearer to the base of the trapezoid,
or triangle, as the ratio bla decreases.
The nodal patterns for the clamped plates are similar to those of the simply supported
plates. However, for clamped boundaries the separation of nodal lines across the plate is
slightly less than that found for simply supported boundaries. The nodal line nearest the
base of the trapezoid also occurs slightly farther from the base for the plate with clamped
boundaries than for the plate with boundaries that are simply supported.

REFERENCES

1. A. W. LEISSA 1969 National Aeronautics and Space Administration NASA SP-160. Vibration of
plates.
2. I. CHOPRA AND S. DURVASULA 1971 Journal of Sound and Vibration 19, 379-392. Vibration of
simply-supported trapezoidal plates. Part I. Symmetric trapezoids.
3. I. CHOPRA and S. DURVASULA 1972 Journal of Sound and Vibration 20, 125-134. Vibration of
simply-supported trapezoidal plates. Part II. Unsymmetric trapezoids.
4. B. FRAEIJS DE VEUBEKE 1966 Proceedings of the First Conference on Matrix Methods in Structural
Mechanics, AFFDL-TR-66-80. Bending and stretching of plates-special methods for upper
and lower bounds.
5. B. FRAEIJS DE VEUBEKE 1968 International Journal ofSo/ids andStructures 4,95-108. A conforming
finite element for plate bending.
6. R. W. CLOUGH and C. A. FELIPPA 1968 Proceedings ofthe Second Conference on Matrix Methods
in Structural Mechanics, AFFDL-TR-68-150. A refined quadrilateral element for analysis of
plate-bending,
7. S. TIMOSHENKO and S. WOINOWSKy-KRIEGER 1959 Theory of Plates and Shells. New York:
McGraw-Hill Book Co., Inc.