Professional Documents
Culture Documents
Dynamic stiffness elements and their applications for plates using rst order
shear deformation theory
M. Boscolo , J.R. Banerjee
School of Engineering and Mathematical Sciences, City University London, Northampton Square, London EC1V 0HB, UK
a r t i c l e
i n f o
Article history:
Received 7 June 2010
Accepted 9 November 2010
Available online 4 December 2010
Keywords:
Dynamic stiffness method
Hamiltons principle
Free vibration analysis
Plates
First order shear deformation theory
Symbolic computation
a b s t r a c t
Dynamic stiffness elements for plates are developed using rst order shear deformation theory to carry
out exact free vibration analysis of plate assemblies. The analysis has been facilitated by the application
of Hamiltonian mechanics and symbolic computation. The WittrickWilliams algorithm has been used as
the solution technique. Results have been extensively validated using published literature for both uniform and non-uniform plates. Some nite element results are also provided. The accuracy and computational efciency of the method are demonstrated. In the nal part of the investigation, signicant plate
parameters are varied and their subsequent effects on the free vibration characteristics are studied.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
Aircraft structures are generally modelled as assemblies of
thin-walled structural elements. In particular, the top and bottom
skins, torsion box, ribs and webs of the wing are idealised as plates.
Thus the free vibration analysis of such structures plays an important role in aircraft design. The analysis facilitates aeroelastic and
response analyses. The purpose of this paper is to develop the
dynamic stiffness method for an accurate and efcient free vibration analysis of plates and plate assemblies.
The usually adopted nite element method [1] (FEM) is a universal tool in structural analysis which can handle complex structures. With the advent of high speed computing, the tendency to
use FEM has increased enormously and many commercially available pre and post processing programs have broadened its appeal,
making it simple and straightforward to use. FEM is an approximate method, but it generally converges to the exact solution with
increasing number of elements. However, the accuracy of results
cannot be always guaranteed. This is particularly true in dynamic
analysis at high frequencies when the FEM may become unreliable.
Thus, there is, and there will always be a need to use analytical
methods based on classical theories, wherever possible, to validate
the FEM, provide further insights and importantly, restore condence in design. One such method is that of the dynamic stiffness
Corresponding author.
E-mail address: marco.boscolo.1@city.ac.uk (M. Boscolo).
0045-7949/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2010.11.005
method [28] (DSM) which gives exact results that are independent of the number of elements used in the analysis. For instance,
one single structural element can be used in the DSM to compute
any number of natural frequencies to any desired accuracy, which
of course, is impossible in the FEM. In DSM [28], once initial
assumptions about the displacement eld have been made, no
inaccuracy occurs in the analysis. However, for the fundamental
mode there is generally very little discrepancy in the frequencies
computed using FEM and DSM, but with increasing mode number,
signicant differences can arise in both response and stability
analyses.
The DSM at present has been developed mainly for onedimensional elements such as bars and beams [39]. This is
generally accomplished by using the exact closed form solution
of their governing differential equations of motion for harmonic
oscillation, and relating a state vector of loads to the corresponding
state vector of responses at the nodes. The relationship between
the two vectors establishes the frequency dependent dynamic
stiffness matrix of the element. There are well established
computer programs such as BUNVIS-RG [10] and PFVIBAT [11]
which demonstrate the accuracy and computational efciency of
the method. A strong point about DSM is that it has all the essential
features of FEM such as coordinate transformation, offset connections, assembly procedure, etc., and yet it retains the exactness of
results through the use of exact solution of the governing
differential equation. However, the solution techniques for FEM
and DSM are different. Unlike the conventional FEM which leads
to a linear eigenvalue problem, the DSM leads to a non-linear
396
2. Theory
2.1. Some basic preliminaries
Fig. 1 shows the notation for displacements and forces for a
thick plate in a rectangular Cartesian coordinate system. The displacement eld is described for both CPT and FSDT and the corresponding equations of motions with their natural boundary
conditions are briey summarised below. For brevity, only
Fig. 1. Coordinate system and notations for displacements and forces for a plate.
ux; y; z u0 x; y z
@w0
;
@x
v x; y; z v 0 x; y z
@w0
;
@y
wx; y; z w0 x; y:
The main focus of this work is on the out of plane vibratory motion
of the plate so that the displacements in the membrane mode
u0(x, y) and v0(x, y) are excluded. Clearly the only unknown in the
above expressions is the vertical displacement w0(x, y).
When deriving the DS matrix, there are some advantages of
using Hamiltons principle as opposed to Newtons second law.
The former provides not only a systematic deduction of the differential equations, but also links natural boundary conditions to the
displacement variables which facilitates the derivation of the DS
matrix greatly. The latter on the other hand is a direct approach
to derive the differential equations, but does not provide any additional information on the correspondence between forces and displacements which is essentially needed in DS formulation.
Although these advantages are not so apparent when CPT is used,
but for FSDT formulation, they are very useful (Section 2.2.3).
Application of Hamiltons principle to derive the governing differential equation of a plate undergoing free natural vibration
lends itself to the following equation and natural boundary conditions (details are not shown for brevity):
@ 4 w0
@ 4 w0
@ 4 w0
2 2 2
4
@x
@x @y
@y4
!
qh
@ 2 w0
0;
@t2
the BC are :
!
@ 3 w0
@ 3 w0
V x : D
2 m
dw0 ;
@x3
@x@y2
!
@ 2 w0
@ 2 w0
d/y ;
m
Mxx : D
@x2
@y2
397
2
8
2 0
@/y
@/x
@ w0
@ 2 w0
>
kGh
qh @@tw2 0
2 @x @y @y2
>
@x
>
>
>
2
>
2
>
> D @ /2y 1m @ /2y m1 @ 2 /x kGh @w0 /
>
y
>
2
2
@y@x
@x
@x
@y
>
<
2
h3 @ /y
q12
2 0
>
@t
>
>
> @2 /x 1m @2 /x m1 @2 /y
0
>
>
/x
D @y2 2 @x2 2 @y@x kGh @w
>
@y
>
>
>
>
:
h3 @ 2 /x
q12
0
@t 2
0
@w
Q x : kGh
/y dw0 ;
@x
@/y
@/
M xx : D
m x d/y ;
@x
@y
3
Gh @/y @/x
d/x :
M xy :
12 @y
@x
w0 x; y; t
1
X
W m xeixt sinam y;
m1
4
2
d Wm
d Wm
qhx2
2a2m
a4m
W m 0 m 1; 2; . . . ; 1:
4
2
dx
dx
D
8
Only the boundaries at x = 0 and x = b (Fig. 1(a)) are considered because a Levy
type solution of the differential equation is sought where two sides (y = 0 and y = L)
are assumed to be SS.
Using standard procedures, the solution of the above equation is obtained by solving the auxiliary equation from a trial solution [31]
398
r 1m
s
r
qh
a2m x
;
D
r 2m
s
r
qh
a2m x
:
D
W m x Am coshr 1m x Bm sinhr 1m x
C m coshr 2m x Dm sinhr 2m x:
10
r 1m
s
r
qh
;
a2m x
D
r 2m
s
r
qh
:
a2m x
D
11
Fig. 2. Boundary conditions for displacements and forces for a plate element.
W m x Am coshr 1m x Bm sinhr 1m x
C m cosr2m x Dm sinr2m x:
12
The procedure to obtain the DS matrix for the rst case is given
below. The DS matrix for the second case follows exactly the same
procedure but is not shown for brevity.
Now from the known displacement w0 (Eqs. (10) and (7)), the
rotation /y, the edge reactions or net shear force Vx, and bending
moment Mxx can be expressed in the following form using Eq. (3).
13
V xm x; y V xm x sinam y D Am r 31m 2 ma2m r 1m sinhr1m x
Bm r31m 2 ma2m r1m coshr 1m x
C m r 32m 2 ma2m r 2m sinhr2m x
14
Dm r32m 2 ma2m r2m coshr 2m x sinam y:
M xxm x; y Mxxm x sinam y D Am r21m ma2m coshr1m x
Bm r21m ma2m sinhr1m x C m r22m ma2m coshr2m x
Dm r 22m ma2m sinhr 2m x sinam y:
15
The boundary conditions for displacements (Fig. 2) are:
x 0;
W m W 1;
Uym Uy1 ;
x b;
Wm W2;
Uym Uy2 ;
16
i.e.
d AC;
19
where
Chi coshrim b;
Ci cosr im b;
S hi sinhr im b;
S i sinr im b
20
with i = 1, 2.
Subsequently, by applying the BCs for forces, i.e. substituting
Eq. (17) into Eqs. (14) and (15), the following matrix relationship
is obtained:
3 2
0
V1
6 M 7 6 L1
1
7 6
6
76
6
4 V 2 5 4 R1 S h1
L2
R1 Ch1
R2 S h2
3
Am
7
7
6
0
76 Bm 7
76
7
R2 Ch2 54 C m 5
L1 Ch1
L1 S h1
L2 Ch2
L2 S h2
M2
R1
R2
32
21
Dm
i.e.
F RC;
22
where
Ri Dr3im a2 rim 2 m;
Li Dr2im a2 m
23
with i = 1, 2.
Using Eqs. (19) and (22) the dynamic stiffness matrix K for the
plate element based on the CPT can be obtained by eliminating the
constant vector C to give:
F Kd;
24
where
similarly the BC for the forces (Fig. 2) are:
x 0;
V xm V 1 ;
x b;
V xm V 2 ;
K RA1 :
Mxxm M1 ;
17
Mxxm M2 :
By applying the BCs for displacements, i.e. substituting Eq. (16) into
Eqs. (10) and (13), the following matrix relationship is obtained:
3 2
W1
7 6
6
6 Uy1 7 6
7 6
6
6W 7 6
4 25 4
2
Uy2
32
r 1m
r2m
C h1
S h1
C h2
S h2
3
Am
76
7
76 Bm 7
76
7
76
7
54 C m 5
r 1m S h1
r 1m Ch1
r 2m S h2
r 2m Ch2
Dm
2
18
25
6
6
K 6
4
svv
sv m
fvv
smm
fv m
Sym
svv
fv m
fmm 7
7
7:
sv m 5
smm
26
Explicit expressions of the elements are derived by extensive algebraic manipulation using Mathematica [25]. These are given in the
Appendix for both sets of solutions (case 1 and case 2).
2.2.2. Effect of rotatory inertia
The effect of rotatory inertia can now be incorporated in the
analysis to derive the DS matrix for better accuracy. The authors
have managed to accomplish this without altering much of the
above expressions because neither the order, nor the nature of
the differential equation changes as a result of the rotatory inertia
effect. Thus it has been possible to simply redene the roots of the
auxiliary equation stemming from the governing differential equation. The effect of rotatory inertia essentially changes the coefcients of the differential operators of Eq. (8), but the rest of the
procedure in the derivation of DS matrix remains unchanged. The
rotatory inertia term associated with plate bending (exure) is given by Mindlin [24] and Reddy [24,26]:
4
Irot
Case 1.
v
!
u
u qx2 h3 2 qx2 h
qx h
t
2
;
am
P
D
24D
24D
2 3
28
r 2m
2
6
6
4
v
v
u
!2
u
u
3
2
u
qx h u
qx2 h3
qx2 h
t
2
t
;
am
D
24D
24D
v
v
u
!
u
u
u qx2 h3 2 qx2 h
2 h3
u
qx
t
2
t
:
am
D
24D
24D
hx2 q GhkD2 a2
GhkD
aGhk
GhkD
2
3
2
2 m1
D D a 2 Ghk x 12qh
D 12 m D
29
1
X
W m xeixt sinam y;
m1
1
X
32
m1
/x x; y; t
1
X
m1
8
2
dU
Ghk d dxW2m hx2 q a2 GhkW m aGhkUxm Ghk dxym 0;
>
>
>
>
>
< d2 U
3
2
dU
m
aD 12 m dxxm a2 D m1
Ghk x 12qh Uym 0;
D dx2ym Ghk dW
2
dx
>
>
>
>
>
: D 1m d2 Uxm aGhkW aD 1m dUym x2 qh3 a2 D Ghk U 0:
m
xm
2
2
dx
12
dx2
33
3 2
3 2 3
0
Wm
7
7 6 7
7 6
aD 12 m D
0 5;
U
4
4
5
y
m
5
1m 2
x2 qh3
0
U
xm
D
D a
Ghk
v
!2
u
qx2 h3 u
qx2 h3
qx2 h
2
;
am
6t
D
24D
24D
aGhk
30
D6 a1 D4 a2 D2 a3 DW 0;
v
v
u
!
u
u
u qx2 h3 2 qx2 h
2 h3
u
qx
;
ta2m
t
D
24D
24D
v
1
v
u 0
!
u
u
3
3 2
u
2
2
2
u B
qx h t qx h
qx hC
2
u
A:
t@am 24D
D
24D
34
12
r 2m
w0 x; y; t
Case 2.
r 1m
2.2.3. FSDT
Having established the methodology for DS formulation using
CPT, attention is now conned to the DS development using FSDT
which is a signicant step forward. The difcult arises due to the
complexity in the order of magnitude of the problem, exacerbated
by the increase in the number of variables.
As before, the solution of Eq. (5) is sought in the Levi form, i.e.
the two opposite sides of the plate are SS at y = 0 and y = L. Solutions complying with these conditions can be written as:
27
In essence, Eq. (27) should be included in Eq. (2) to account for the
effect of rotatory inertia. This will redene the roots of the auxiliary
Eqs. (9) and (11) as follows:
r 1m
Using the roots given by Eqs. (29) and (31) the DS matrix can
now be computed using the expressions of Eq. (26).
/y x; y; t
1
@ w
@ w
qh3
:
12
@x2 @t2 @y2 @t 2
399
35
where:
W W m or Uym
31
or Uxm :
36
Substituting a trial solution ek in Eq. (35) yields the following auxiliary equation:
k6 a1 k4 a2 k2 a3 0;
37
400
where:
2
a1
l3 a1 l2 a2 l a3 0;
39
1
3
l2
1
3
l3
1
3
44
for which the three roots (l1, l2, l3) are given by:
l1
r
p r
p!
3 p
q 3 p q
a1
;
2
2
r
r
p
p!
3 p
q
3 p
q
;
a1 b2
b1
2
2
r
r
p
p!
3 p
3 p
q
q
;
a1 b1
b2
2
2
40
where:
p
p
i 31
i 31
; b2
b1
2
2
3
p 2a1 9a1 a2 27a3 ; l a21 3a2 ;
q p2 4l
C 5m cosr3m x C 6m sinr 3m x:
h x2 qh x2 q 12Gk2 576DG2 hk
4
6 4 4
746496D G k m 1
3
3
gh k 12D2 x2 q Gh kx2 q1 m
2
2
6Dx2 qm2 1 24G2 hk
:
42
The discriminant is always positive, i.e. D > 0, in the present case
which means that the three roots are real, and consequently, there
are no more than four possible solutions, i.e. (i) all roots are positive,
(ii) one negative and two positive, (iii) two negative and one positive and (iv) all roots are negative. Descartess rule of signs [32]
was applied to predict the sign of the roots in advance, but no denitive conclusion on their signs could be drawn. Thus, it can be said
that there cannot be less than the four aforementioned cases. These
are as follows:
Case 1: l1, l2, l3 > 0.
The six roots are real r1, r1, r2, r2, r3, r3 where
p
p
p
r1 l1 ; r2 l2 and r3 l3 .
Thus the solutions can be written as:
45
43
46
401
C 1m c1 B2m ;
C 2m c1 B1m ;
C 3m c2 B4m ;
C 4m c2 B3m ;
C 5m c3 B6m ;
C 6m c3 B5m ;
47
52
1
3
Gh B1m a c1 r 1m coshr 1m x
12
B2m a c1 r 1m sinhr 1m x B3m a c2 r 2m coshr2m x
where:
di
53
3
Gk 12Ghk h qx2 6D1 m a2 r 2i
;
2
6r i 2G2 hk DGka2 1 m D1 mqx2 DGk1 mr2i
x0:
W m W 1;
xb:
Wm W2;
Uym Uy1 ;
Uym Uy2 ;
48
3
2
12Dr 2i 12Ghk 6Da2 m 1 h qx2 Gk a2 r 2i qx2 12G2 hk r2i
ci
2
6ari 2G2 hk D1 m qx2 Gk a2 r2i
49
with i = 1, 2, 3.
Thus Eq. (43) can be rewritten in terms of only one set of constants as:
Uxm Ux1 ;
Uxm Ux2 :
54
x0:
Qxm Q1 ;
xb:
Qxm Q2 ;
Mxxm Mxx1 ;
Mxxm Mxx2 ;
Mxym Mxy1 ;
2 0
1
7 6
6
6 Uy1 7 6 1
7 6
6
7 6
6
6 Ux1 7 6 0
76
6
7 6
6
6 W 2 7 6 d1 S h1
7 6
6
7 6
6
4 Uy2 5 4 Ch1
Ux2
c1 S h1
d1
d2
c1
c2
d1 Ch1
d2 S h2
d2 Ch2
d3 S h3
S h1
C h2
S h2
Ch3
c1 Ch1
c2 S h2
c2 Ch2
c3 S h 3
d3 32 B1 3
76 7
0 7 6 B2 7
76 7
76 7
c3 76 B3 7
76 7;
76 7
d3 Ch3 76 B4 7
76 7
6 7
S h3 7
5 4 B5 5
c3 Ch3
By substituting Eq. (50) into Eq. (6), the expressions for forces and
moments can be obtained as:
51
B6
56
55
Mxym Mxy2 :
i.e.
d AC;
57
7 6
6
6 7
6 Qx 7 6 L1 Ch L1 S h L2 Ch L2 S h L3 Ch L3 S h 7
B
2 7
1
1
2
2
3
3 76 4 7
6
76 7
6
7 6
6
7
7
6 Mxx2 7 4 R1 S h R1 Ch R2 S h R2 Ch R3 S h R3 Ch 56
B5 5
4
1
1
2
2
3
3
5
4
Mxy2
T 1 Ch1 T 1 S h1 T 2 Ch2 T 2 S h2 T 2 Ch2 T 2 S h2
B6
2
58
i.e.
F RC;
59
where
Li Ghkdi r i 1; Ri Daci m ri ; T i
1
3
Gh a ci ri
12
60
with i = 1, 2, 3.
From Eqs. (57) and (59) the dynamic stiffness matrix K for an
isotropic plate element based on FSDT is obtained as:
F Kd;
61
where
Fig. 3. Edge conditions of the plate element and sign conventions.
K RA1 :
62
402
Thus
2
6
6
6
6
K 6
6
6
6
4
sqq
sqm
sqt
fqq
fqm
smm
smt
stt
fqm
fqt
fmm
fmt
Sym
sqq
sqm
smm
fqt
7
7
7
7
7;
sqt 7
7
7
smt 5
fmt
ftt
63
stt
where the 12 independent terms sqq, sqm, sqt, smm, smt, stt, fqq, fqm, fqt,
fmm, fmt, ftt describe the relationships of shear, bending moment, and
twisting moment due to unitary displacements on the same (s)
nodal line, and on the far (f) nodal line. These are given in explicit
algebraic form in Appendix.
2.3. Application of the DSM
2.3.1. Assembly procedure and application of boundary conditions
The stiffness matrices given by Eqs. (26) and (63) are essential
requirements to compute exact natural frequencies of plates which
are simply supported on at least two opposite sides. For individual
plates, it is possible to determine the natural frequencies and mode
shapes by applying the CM dened in the introduction but clearly
that procedure cannot be easily extended to deal with plate assemblies as for the DSM. The assembly procedure in DSM is similar to
the nite element method and is schematically shown in Fig. 4.
Note that each strip of plate is connected through nodal lines instead of single points. Clearly the overall master stiffness matrix
will be banded as in the case of FEM.
The application of boundary conditions in DSM is also similar to
that of FEM. The penalty method is generally used to suppress any
particular degree of freedom. The method entails adding a large
stiffness to the appropriate term on the leading diagonal of the dynamic stiffness matrix.
The procedure for applying the boundary conditions is summarised as follows:
CPT.
Free (F): no penalty is applied.
Simply supported (SS): Wi is penalised.
Clamped (C): Wi and Uyi are penalised.
FSDT
Free (F): no penalty is applied.
Simply supported (SS): Wi and Uxi are penalised.
Clamped (C): W i ; Uyi , and Uxi are penalised.
where i the node to be constrained.
Because of the similarities, DS elements can be implemented in
FEM codes.
2.3.2. WittrickWilliams algorithm
Once the global dynamic stiffness matrix of a structure is
formed, the zeros of the determinant can be sought to determine
the natural frequencies. This procedure can be hugely cumbersome
because of the transcendental nature of the dynamic stiffness
j j0 sK ;
64
where j0 is the number of natural frequencies of single strip elements clamped on their opposite sides which are lower than the
trial frequency.
Assuming j0 is known, a suitable procedure can be devised, for
example the bi-section method, to bracket any natural frequency
between an upper and lower bound to any desired accuracy. This
allows quick and precise computations of the natural frequencies
without numerically searching for the zeros of the determinant
which may cause numerical difculties due to the transcendental
nature of stiffness element.
A drawback of the algorithm lays in computing the j0 values.
In this paper, j0 computation was avoided by using a sufciently
ne mesh to ensure that j0 = 0 at the frequency range of interest.
This was achieved by computing the rst CC natural frequency
of the largest strip by splitting it into narrower strips, applying
the DSM and subsequently compute the natural frequencies of
the global structure that are below the CC frequency computed
earlier.
2.3.3. Mode shape computation
The mode shapes are computed by using Eq. (61) and setting
the force vector to zero and one of the nodal displacements to an
arbitrary value. Then, the rest of the nodal displacements are determined in terms of the chosen one. For a ne mesh, the nodes will
be so close to each other that an accurate plot of the mode shapes
is possible from the nodal values of the displacements without the
need to process the results any further. Otherwise, for a coarse
mesh, once the nodal line values of the displacements have been
computed, the constants of the analytical solution (Eq. (50)) can
be computed by solving the system of algebraic Eq. (56) simultaneously. In this way, the analytical expressions of the mode shapes
can be obtained for any complex geometry and it can be plotted to
any desired accuracy.
Although the numerical values of the natural frequencies are
not inuenced by the mesh size since each element is analytically
exact, sometimes it is advisable to use a ner mesh if the plots of
the mode shapes are required. The reason for a ner mesh is essentially twofold. First, the nodal values can be used to plot the modes
and the calculation of the constants can be avoided since it
403
Table 1
p
^ xh q=G) for an isotropic SSSSSSSS square plate: b/L = 1, b/h = 10, m = 0.3, k = 5/6. The errors incurred by the CPT and CPT with rotatory
Validation of natural frequencies (x
inertia when compared with the FSDT are reported in brackets. Classical method results (CM) by Reddy and Phan [21].
mn
FSDT
CM
FSDT
DSM
CPT + Irot
CM
CPT
CM
CPT (error %)
DSM
1
1
2
1
2
1
3
2
3
1
2
4
3
0.0930
0.2219
0.3406
0.4149
0.5206
0.6520
0.6834
0.7446
0.8896
0.9174
0.9984
1.0764
1.1268
0.09303
0.22193
0.34056
0.41494
0.52056
0.65201
0.68341
0.74465
0.88965
0.91740
0.99844
1.07642
1.12685
0.0955
0.2360
0.3732
0.4629
0.5951
0.7668
0.8090
0.8926
1.0965
1.1365
1.2549
1.3716
1.4457
0.09553
0.23599
0.37319
0.46293
0.59506
0.76684
0.80904
0.89256
1.09650
1.13649
1.25494
1.37117
1.44747
0.0963
0.2408
0.3853
0.4816
0.6261
0.8187
0.8669
0.9632
1.2040
1.2521
1.3966
1.5411
1.6374
0.09632
0.24079
0.38527
0.48159
0.62606
0.81870
0.86686
0.96318
1.20397
1.25213
1.39660
1.54108
1.63740
1
2
2
3
3
4
3
4
4
5
5
4
5
(2.69)
(6.34)
(9.58)
(11.57)
(14.31)
(17.61)
(18.38)
(19.86)
(23.25)
(23.88)
(25.69)
(27.38)
(28.45)
(3.54)
(8.50)
(13.13)
(16.06)
(20.27)
(25.57)
(26.84)
(29.35)
(35.33)
(36.49)
(39.88)
(43.17)
(45.31)
Table 2
p
p
^ xh q=G) for an isotropic SSSSSSSS rectangular plate: b=L 2; b=h 10; m 0:3; k 5=6. The errors incurred by the CPT and CPT with
Validation of natural frequencies (x
rotatory inertia when compared with the FSDT are reported in brackets. Classical method results (CM) by Reddy and Phan [21].
mn
FSDT
CM
FSDT
DSM
CPT + Irot
CM
CPT
CM
CPT (error %)
DSM
1
2
1
3
2
3
4
1
2
4
3
5
5
0.07036
0.13729
0.20123
0.24235
0.26250
0.35948
0.37818
0.39666
0.45089
0.48608
0.53754
0.53754
0.63609
0.07036
0.13729
0.20123
0.24236
0.26251
0.35948
0.37818
0.39667
0.45090
0.48609
0.53754
0.53754
0.63610
0.07180
0.14273
0.21281
0.25908
0.28207
0.39575
0.41822
0.44062
0.50729
0.55133
0.61680
0.61680
0.74563
0.07180
0.14273
0.21281
0.25908
0.28208
0.39575
0.41823
0.44062
0.50730
0.55134
0.61680
0.61680
0.74563
0.07224
0.14448
0.21671
0.26487
0.28895
0.40935
0.43343
0.45751
0.52974
0.57790
0.65014
0.65014
0.79462
0.07224
0.14448
0.21671
0.26487
0.28895
0.40935
0.43343
0.45751
0.52975
0.57791
0.65014
0.65014
0.79462
1
1
2
1
2
2
1
3
3
2
3
1
2
(2.04)
(3.96)
(5.76)
(6.90)
(7.46)
(10.09)
(10.59)
(11.08)
(12.51)
(13.42)
(14.74)
(14.74)
(17.22)
(2.67)
(5.23)
(7.70)
(9.29)
(10.07)
(13.87)
(14.61)
(15.34)
(17.49)
(18.89)
(20.95)
(20.95)
(24.92)
404
Table 3
p
^ xL2 qh=D) for isotropic square plates with various boundary conditions: b/L = 1, b/h = 10, m = 0.3, k = 5/6. The errors incurred by the CPT and CPT with
Natural frequencies (x
rotatory inertia when compared with the FSDT are reported in brackets. CPT frequencies are validated against results published by Leissa [33].
SSSSSSSS
SSFSSSS
SSCSSSS
mn
FSDT
CPT (error %)
mn
FSDT
CPT (error %)
mn
FSDT
CPT (error %)
1
1
2
2
1
3
2
3
4
1
3
4
19.0650
45.4827
45.4827
69.7944
85.0380
85.0380
106.683
106.683
133.621
133.621
140.057
152.608
19.7392 (3.54)
49.3480 (8.50)
49.3480 (8.50)
78.9568 (13.13)
98.6960 (16.06)
98.6960 (16.06)
128.304 (20.27)
128.304 (20.27)
167.783 (25.57)
167.783 (25.57)
177.652 (26.84)
197.392 (29.35)
1
1
2
2
1
3
2
3
1
3
2
4
11.3731
26.1545
38.2861
53.2462
55.6234
78.3664
81.0756
92.2072
97.3551
117.908
119.844
127.461
11.6845 (2.74)
27.7563 (6.12)
41.1967 (7.60)
59.0655 (10.93)
61.8606 (11.21)
90.2941 (15.22)
94.4837 (16.54)
108.918 (18.12)
115.685 (18.83)
145.638 (23.52)
148.509 (23.92)
159.080 (24.81)
1
2
1
2
3
1
3
2
4
1
3
4
22.3886
47.1039
52.1496
74.1051
85.8759
93.2268
109.259
112.741
134.084
141.936
144.117
154.145
23.6463 (5.62)
51.6743 (9.70)
58.6464 (12.46)
86.1345 (16.23)
100.269 (16.76)
113.228 (21.45)
133.791 (22.45)
140.845 (24.93)
168.958 (26.01)
187.436 (32.06)
188.113 (30.53)
201.724 (30.87)
26.6683
49.1129
59.2102
78.8130
86.8440
101.371
112.058
118.922
134.595
148.316
149.990
155.788
28.9508 (8.56)
54.7431 (11.46)
69.3270 (17.09)
94.5853 (20.01)
102.216 (17.70)
129.095 (27.35)
140.204 (25.12)
154.775 (30.15)
170.346 (26.56)
199.810 (34.72)
206.697 (37.81)
208.391 (33.77)
1
2
1
2
3
1
3
2
1
4
3
2
SSFSSF
1
1
1
2
2
2
1
3
3
2
3
1
1
2
3
1
2
3
4
1
2
4
3
5
1
2
1
2
3
1
3
2
4
3
4
1
SSFSSC
9.4406
15.3893
33.8599
36.3570
42.7927
62.1467
66.1965
76.6330
82.4088
92.6128
101.029
110.024
9.6314
16.1348
36.7256
38.9450
46.7381
70.7401
75.2834
87.9867
96.0405
111.0254
122.0400
133.7046
(2.02)
(4.84)
(8.46)
(7.12)
(9.22)
(13.83)
(13.73)
(14.82)
(16.54)
(19.88)
(20.80)
(21.52)
1
1
2
2
1
3
2
3
1
3
2
4
1
2
1
2
3
1
3
2
4
3
4
1
1
1
2
2
1
3
2
3
1
4
3
2
SSCSSC
12.2492
30.4086
38.6343
55.8017
62.7260
78.5265
85.8935
93.6825
105.613
120.987
126.029
127.540
12.6874 (3.58)
33.0651 (8.74)
41.7019 (7.94)
63.0148 (12.93)
72.3976 (15.42)
90.6114 (15.39)
103.161 (20.10)
111.896 (19.44)
131.428 (24.44)
152.773 (26.27)
159.302 (26.40)
162.371 (27.31)
1
2
1
2
3
1
3
2
4
3
1
4
1
1
2
2
1
3
2
3
1
3
4
2
Fig. 5. Mode shapes (mode m = 2, n = 2) for square plates with different boundary conditions.
405
SS-F-SS-F
SS-SS-SS-SS
SS-C-SS-C
140
120
120
100
100
Error, %
Error, %
SS-F-SS-F
SS-SS-SS-SS
SS-C-SS-C
140
80
60
80
60
40
40
20
20
0
10
100
10
b/h
140
120
SS-F-SS-F
SS-SS-SS-SS
SS-C-SS-C
140
120
100
Error, %
100
Error, %
100
b/h
80
60
80
60
40
40
20
20
10
100
10
b/h
100
b/h
(c) 7 th Mode
Fig. 6. Convergence of CPT natural frequencies towards FSDT against thickness ratio for different boundary conditions.
Table 4
p
^ xL2 qh=D) for an isotropic SSSSSSSS square plate: b/L = 1, b/h = 10, m = 0.3, k = 5/6. The exact value obtained by the DS elements is compared to the
Natural frequencies (x
values obtained by FEM with different mesh sizes.
CPT
FSDT
FEM
10 10
100 100
DSM
FEM
11
10 10
DSM
100 100
11
^
x
Error %
^
x
Error %
^
x
^
x
Error %
^
x
Error %
^
x
0.09511
0.23676
0.23676
0.37027
0.47503
0.47503
0.59761
0.59761
0.80420
0.81289
1.3
1.7
1.7
3.9
1.4
1.4
4.5
4.5
1.8
0.7
0.09630
0.24074
0.24074
0.38505
0.48146
0.48146
0.62556
0.62556
0.81848
0.81848
0.0
0.0
0.0
0.1
0.0
0.0
0.1
0.1
0.0
0.0
0.09632
0.24079
0.24079
0.38527
0.48159
0.48159
0.62606
0.62606
0.81870
0.81870
0.09237
0.22020
0.22020
0.32916
0.41248
0.41248
0.49571
0.49571
0.62596
0.64391
0.7
0.8
0.8
3.3
0.6
0.6
4.8
4.8
4.0
1.2
0.09370
0.22537
0.22537
0.34781
0.42520
0.42520
0.53512
0.53512
0.67281
0.67281
0.7
1.5
1.5
2.1
2.5
2.5
2.8
2.8
3.2
3.2
0.09303
0.22193
0.22193
0.34056
0.41494
0.41494
0.52056
0.52056
0.65201
0.65201
406
Table 5
Natural frequencies (x) for isotropic stepped plates (SSSSSSSS) with various
stringer thicknesses ratios. E = 71,200 MPa, m = 0.3,q = 2700 kg/m3, b/L = 1, b1 = 0.3b,
b2 = 0.1b, b3 = 0.2b, b/t1 = 1000. CM: classical method by Chopra [34] and corrected by
Guo et al. [35], DFSM: dynamic nite strip method [35], DSM (CPT): dynamic stiffness
method with classical plate theory developed in this paper.
t2/t1 = 1.2
t2/t1 = 1.5
mn
CM
DFSM
DSM (CPT)
mn
CM
DFSM
DSM (CPT)
1
1
2
2
1
3
2
3
1
4
3
2
31.74
79.30
80.60
127.3
156.4
162.5
207.1
207.8
/
/
/
/
31.74
79.29
80.60
127.3
156.3
162.4
207.0
207.7
/
/
/
/
31.738
79.298
80.606
127.309
156.362
162.490
207.137
207.833
273.604
277.029
288.246
319.387
11
12
21
22
13
31
23
32
14
41
33
24
33.17
81.69
87.64
134.2
161.5
179.0
221.1
222.0
/
/
/
/
33.17
81.69
87.67
134.2
161.5
179.0
221.0
221.8
/
/
/
/
33.171
81.695
87.678
134.230
161.495
179.096
221.104
221.959
284.582
305.75
309.201
331.024
1
2
1
2
3
1
3
2
4
1
3
4
Table 6
Natural frequencies (x) for an isotropic thick stepped plate computed by DSM based on different plate theories: CPT, CPT with rotatory inertia, and FSDT. E = 71200 MPa, m = 0.3,
q = 2700 kg/m3, b/L = 1, b1 = 0.3b, b2 = 0.1b, b3 = 0.2b, b/t1 = 10, t2/t1 = 1.5, k = 5/6.
SSSSSSSS
CPT
SSCSSC
CPT + Irot
FSDT
CPT
CPT + Irot
FSDT
mn
mn
mn
mn
mn
mn
1
1
2
2
1
3
2
3
1
4
3
2
4
3
1
4
3317.10
8169.43
8767.77
13422.98
16149.55
17909.59
22110.42
22195.91
28458.24
30574.96
30920.08
33102.40
34340.33
41239.72
42361.48
42513.24
1
1
2
2
1
3
3
2
1
4
3
2
4
3
1
4
3283.27
7995.98
8545.86
12937.92
15919.35
17040.96
20909.12
21449.62
26007.54
28202.95
29279.49
29846.95
31447.46
36342.73
38910.39
39056.25
1
1
2
2
1
3
2
3
1
4
3
2
4
3
4
1
3154.61
7339.71
7854.54
11433.49
13481.87
14698.52
17404.82
17630.01
21566.83
22796.97
22976.23
24377.75
25168.74
28997.32
29710.50
29778.97
1
2
1
2
3
1
3
2
4
3
1
4
2
4
3
5
4715.32
9661.77
11366.75
15895.44
18604.82
20908.31
24267.81
26520.18
31254.01
34988.73
35222.01
36343.84
39574.41
46150.98
47114.01
47342.38
1
2
1
2
3
1
3
2
4
1
4
3
2
3
5
4
4663.71
9405.74
11033.20
15196.79
17674.04
20750.18
22677.83
25931.32
28770.11
32352.55
33018.56
33403.57
35825.95
41637.54
42015.25
42722.08
1
2
1
2
3
1
3
2
4
1
3
4
2
3
4
5
4247.36
8395.51
9416.91
12775.79
15004.08
15939.68
18475.41
19298.84
22995.49
24131.01
24323.97
25733.40
26500.53
30559.19
30622.63
31740.18
1
2
1
2
3
1
3
2
4
1
3
4
2
4
5
3
1
2
1
2
3
1
2
3
4
1
3
4
2
4
5
3
1
2
1
2
3
1
3
2
4
1
3
4
2
4
3
5
1
1
2
2
1
3
2
3
1
3
4
2
4
3
4
1
1
1
2
2
1
3
2
3
1
4
2
3
4
4
1
3
Fig. 8. Natural modes for the stepped plates. Modes computed by DSM and FSDT.
1
1
2
2
1
3
2
3
1
4
3
2
4
4
3
1
407
Table 7
Natural frequencies (x) for an isotropic thick stepped plates computed by FEM (NASTRAN) compared with the exact values computed by the DSM. E = 71200 MPa, m = 0.3,
q = 2700 kg/m3, b/L = 1, b1 = 0.3b, b2 = 0.1b, b3 = 0.2b, t2/t1 = 1.5.
CPT, b/t1 = 1000, (SSSSSSSS)
FEM
10 10
100 100
DSM
FEM
51
10 10
DSM
100 100
51
^
x
Error %
^
x
Error %
^
x
^
x
Error %
^
x
Error %
^
x
32.80
80.35
86.56
129.54
157.40
177.53
210.09
213.26
285.33
287.65
304.52
324.37
332.01
1.12
1.65
1.27
3.49
2.54
0.87
4.98
3.92
0.26
5.92
1.51
2.01
3.32
33.17
81.67
87.66
134.15
161.42
179.06
220.89
221.79
284.52
305.68
308.75
330.74
343.09
0.02
0.03
0.02
0.06
0.04
0.02
0.10
0.08
0.02
0.02
0.15
0.09
0.09
33.17
81.69
87.68
134.23
161.50
179.10
221.10
221.96
284.58
305.75
309.20
331.02
343.40
4217.26
8348.32
9358.28
12385.63
14973.60
15936.29
17666.76
18483.53
22292.85
22700.80
23536.93
24231.91
24819.87
0.71
0.56
0.62
3.05
0.20
0.02
4.38
4.22
3.06
5.93
3.24
5.83
6.34
4287.44
8561.59
9572.40
13084.94
15445.34
16636.62
19053.29
20115.94
23801.80
24765.00
25318.64
26644.37
27265.66
0.94
1.98
1.65
2.42
2.94
4.37
3.13
4.23
3.51
2.63
4.09
3.54
2.89
4247.36
8395.51
9416.91
12775.79
15004.08
15939.68
18475.41
19298.84
22995.49
24131.01
24323.97
25733.40
26500.53
investigated to determine any required number of natural frequencies and mode shapes to any desired accuracy.
The panel studied by Chopra had a thickness ratio of b/t1 = 1000
and, as a consequence, there is no reason to use FSDT as opposed to
CPT. The same panel with a thickness ratio b/t1 = 10 and t2/t1 = 1.5
has been studied with two different sets of boundary conditions
SSSSSSSS and SSCSSC, respectively. The rst 16 natural frequencies are shown in Table 6. It can be seen again that the effect
of shear deformation is not negligible for this thickness ratio and
the presence of the two intermediate thick strips shown in Fig. 7
has an even more pronounced effect. In fact the error on the fundamental natural frequency for the simply supported panel due to
using CPT is around 5% while it is only 3% for a constant thickness
plate (Table 3). Clearly, the error on the rst frequency worsens at
11% for the clamped BCs. Two mode shapes (m = 2, n = 1 and
m = 4, n = 1) for the two boundary conditions are plotted in Fig. 8
for demonstration purposes.
Also for this case, results obtained by using the FEM (NASTRAN)
have been compared with the exact results computed by using the
DSM. The results are summarised in Table 7.
A coarse mesh (10 10) and a ner mesh (100 100) were
used as before. CPT based nite element results converge to the exact solution. On the other hand, when FSDT elements are used, the
FE solution converges to a higher value than the exact one obtained
by the DSM. Accuracy of the FEM decreases for higher natural frequencies, as in the previous case. Dynamic stiffness elements, not
only provide the exact value for any number of natural frequencies
but also they offer signicant savings in computer time.
4. Concluding remarks
An exact dynamic stiffness method for plate elements with two
opposite sides simply supported has been developed using classical plate theory with and without the inclusion of rotatory inertia
effect. The theory has then been signicantly enhanced by applying
the rst order shear deformation theory. Explicit expressions for
the terms of the dynamic stiffness matrix for all cases have been
derived by extensive use of symbolic computation. The dynamic
stiffness elements are sufciently general to be assembled to
investigate the free vibration behaviour of complex structures following a procedure similar to that used in the nite element method. Once the dynamic stiffness matrix for the overall structure is
formulated the eigenvalue problem is solved by using a modied
version of the Wittrick and Williams algorithm.
408
sqq
Ghk
Ch3 c1 d3 r2 c2 d2 r 3 c3 d2 2r3 c2 d3
D
d1 r 1 c1 r 2 c2 c3 d2 c2 r 1 c1 r3 c3 d3
S h1 S h2 d3 r 2 c21 r 3 c2 c3 d2 r 3 c21 c22 d3
d1 r2 c1 r 1 c2 c3 d2 r1 c22 r 3 c1 c3 d3
Ch2 d1 r 1 c1 r 2 c2 c3 d2 c2 r 1 c1 r 3 c3 d3
c1 d2 2r 2 c3 d2 r2 c2 r 3 c3 d3
S h1 S h3 d1 r 2 c1 c2 r 1 c23 d2 c2 r3 c1 r 1 c3 d3
d2 r 2 c21 c23 d2 r3 c21 r 2 c2 c3 d3
Ch1 2r1 c2 c3 d21 r 1 c1 c3 d1 d2 r 2 c2 c3 d1 d2 r 1 c1 c2 d1 d3
CLT
Case 1
svv r 2m R1 r 1m R2 r1m Ch2 S h1 r 2m Ch1 S h2 =D;
65
2
2
sv m r1m R2 Ch1 Ch1 Ch2 S h1 R1 S h1 S h2 ;
66
=D;
r2m R2 S h1 S h2 R1 Ch1 Ch2 Ch2 S 2h2
smm L2 L1 r 2m Ch2 S h1 r 1m Ch1 S h2 =D;
fvv r 2m R1 r 1m R2 r 2m S h2 r 1m S h1 =D;
fv m Ch2 Ch1 r 2m R1 r 1m R2 =D;
fmm L1 L2 r 2m S h1 r 1m S h2 =D;
67
68
69
70
71
D r21m S h1 S h2 r 22m S h1 S h2
2
r1m r 2m Ch1 Ch2 S 2h1 S 2h2 :
r 3 c2 c3 d1 d3 r 2 c1 c2 d2 d3 r3 c1 c3 d2 d3
S h2 S h3 r 1 c22 c23 d21 c1 r 3 c2 r2 c3 d2 d3
d1 r 1 c1 c2 d2 r 2 c23 d2 r 3 c22 d3 r1 c1 c3 d3
Ch2 Ch3 2r 1 c2 c3 d21 d1 r 1 c1 r 2 c2 c3 d2
c2 r 1 c1 r3 c3 d3 c1 2r 2 c3 d22 r 2 c2 r3 c3 d2 d3
2r 3 c2 d23
;
83
sqm
S h2 S h3 c2 d1 c1 d2 c1 d3
Ch2 Ch3 c3 d1 c1 d3
1 r 2 d2 c3 d1 c1 d3 1 Ch1 Ch2
72
Case 2
svv r 2m R1 r 1m R2 r2m Ch1 S 2 r 1m C2 S h1 =D;
sv m r 2m R1 C22 C2 Ch1 S 22 R2 S 2 S h1 ;
r1m R1 S 2 S h1 R2 C2 Ch1 Ch1 S 2h1 =D;
smm L1 L2 r 1m Ch1 S 2 r2m C2 S h1 =D;
fvv r 2m R1 r 1m R2 r 2m S 2 r 1m S h1 =D;
fv m C2 Ch1 r 2m R1 r 1m R2 =D;
fmm L1 L2 r 1m S 2 r 2m S h1 =D;
75
76
77
78
R1 D r 31m a2 r1m 2 m ; R2 D r 32m a2 r 2m 2 m ;
2
a m; L2 D
D r 21m S 2 S h1
r 22m S 2 S h1
r 22m
Ch2 Ch3 c2 d1 c1 d2 S h2 S h3 c3 d1 c1 d3
1 r 1 d1 c3 d2 c2 d3 1 Ch1 Ch2
S h3 c3 d1 c1 d3 S h2 c2 d1 c1 d2 Ch1 Ch3 c2 d1
84
c1 d2 S h1 S h3 c3 d2 c2 d3 ;
74
L1 Dr 21m
S h3 c3 d2 c2 d3 S h1 c2 d1 c1 d2
73
79
80
a m :
81
2
2
2
r 1m r 2m C2 Ch1 S 2 S h1 :
82
FSDT
Explicit expressions for the coefcients of the DS matrix are
given only for case 1 for brevity. The coefcients for the other 3
cases can be obtained by following the procedure reported in
Section 2.2.3. Given the complexity of these coefcient, the
use of a symbolic computation program such as Mathematica
[25] is essential. In order to avoid numerical instabilities and
overows, the expressions need to be simplied and carefully
constructed. Full simplication is achieved by using the explicit
expressions of the terms appearing in the matrix A and R (Eqs.
Ghk
c2 d1 c1 d2 1 r 3 d3 1 Ch1 Ch3
D
S h2 c3 d2 c2 d3 S h1 c3 d1
sqt
Ghk
Ch3 c1 d2 d3 r2 d2 r3 d3 r 1 d21 c3 d2 c2 d3
D
d1 r 2 c3 d22 2r 3 c3 d2 d3 r 3 c2 d23
S h1 S h2 c3 d21 r 2 d2 r 3 d3 r 3 d2 d3 c3 d2 c2 d3
d1 r 1 c3 d22 r 2 c1 r 1 c2 d2 d3 r 3 c1 d23
Ch2 c1 d2 d3 r 2 d2 r 3 d3 r1 d21 c3 d2 c2 d3
d1 r 2 c3 d22 2r 2 c2 d2 d3 r 3 c2 d23
S h1 S h3 r 2 d2 d3 c3 d2 c2 d3 d21 r 2 c2 d2 r 3 c2 d3
d1 r 2 c1 d22 r 3 c1 r 1 c3 d2 d3 r 1 c2 d23
Ch1 r 1 c3 d21 d2 r 2 c3 d1 d22 r 1 c2 d21 d3 2r 1 c1 d1 d2 d3
r 2 c1 d22 d3 r 3 c2 d1 d23 r 3 c1 d2 d23
S h2 S h3 c1 d2 d3 r 3 d2 r 2 d3 r 1 d21 c2 d2 c3 d3
d1 r 1 c1 d22 r3 c2 d2 d3 r 2 c3 d2 d3 r 1 c1 d23
Ch2 Ch3 c1 d2 d3 r 2 d2 r 3 d3 r 1 d21 c3 d2 c2 d3
d1 r 2 c3 d22 2r 1 c1 r2 c2 r 3 c3 d2 d3 r3 c2 d23
;
85
409
D
r 3 c2 d1 r2 c3 d1 r 3 c1 d2 r 1 c3 d2 r 2 c1 d3 r 1 c2 d3 Ch3 S h1 S h2 c2 d1 c1 d2 S h3 Ch2 S h1 c3 d1 c1 d3
D
Ch1 S h2 c3 d2 c2 d3 ;
smm
D
r 3 amc3 1 Ch1 Ch3 S h2 d2 c2 d1 c1 d2 1 Ch1 Ch2 S h3 2c3 d1 d2 c2 d1 d3 c1 d2 d3
D
S h1 Ch2 Ch3 d1 c2 d1 c1 d2 d1 c2 d1 c1 d2 S h2 S h3 c3 d21 c3 d22 c1 d1 d3 c2 d2 d3
r 1 amc1 S h2 d2 Ch1 Ch3 d2
1 Ch1 Ch2 S h3 d3 c3 d2 c2 d3 S h1 c3 d1 d2 c2 d1 d3 2c1 d2 d3 Ch2 Ch3 c3 d1 d2 c2 d1 d3 2c1 d2 d3
S h2 S h3 c2 d1 d2 c1 d22 c3 d1 d3 c1 d23 r 2 amc2 1 Ch1 Ch2 S h3 d3 c3 d1 c1 d3
1 Ch1 Ch3 S h2 c3 d1 d2 2c2 d1 d3 c1 d2 d3 S h1 Ch2 Ch3 d1 c3 d1 c1 d3 d1 c3 d1 c1 d3 S h2 S h3 c2 d21
c1 d1 d2 c3 d2 d3 c2 d23
;
86
smt
87
Gh
Ch2 c1 r 2 c2 r3 c3 d2 d3 d1 r 1 c1 r 2 c2 c3 d2 c2 r 1 c1 2r2 c2 r 3 c3 d3 S h1 S h3 r2 c2 r 3 c3 d1 c2 d1 c1 d2
12D
r 1 c1 r 2 c2 c3 d2 d3 c2 r1 c1 r 2 c2 d23 Ch3 c1 r 2 c2 r 3 c3 d2 d3 d1 c3 r 1 c1 r 2 c2 2r 3 c3 d2 r 1 c1 c2 r 3 c2 c3 d3
S h1 S h2 c3 r 1 c1 r 3 c3 d22 c2 r 1 c1 r3 c3 d2 d3 r 2 c2 r3 c3 d1 c3 d1 c1 d3 Ch1 r 1 c1 c3 d1 d2 r2 c2 c3 d1 d2
r 1 c1 c2 d1 d3 r 3 c2 c3 d1 d3 2r 1 c21 d2 d3 r 2 c1 c2 d2 d3 r 3 c1 c3 d2 d3 S h2 S h3 c2 r1 c1 r 3 c3 d1 d2 c1 r 1 c1 r 3 c3 d22
stt
88
Ghk
r 1 S h2 S h3 c22 d21 2r 1 c2 c3 d21 r1 S h2 S h3 c23 d21 r 2 S h1 S h3 c1 c2 d1 d2 r1 S h2 S h3 c1 c2 d1 d2 r 1 c1 c3 d1 d2 r 2 S h1 S h2 c1 c3 d1 d2
D
fqq
c2 r 1 c1 r 3 c3 d3 Ch2 c1 d3 r2 c2 d2 r 3 c3 d2 2r3 c2 d3 d1 r 1 c1 r2 c2 c3 d2 c2 r1 c1 r 3 c3 d3 ;
fqm
Ghk
Ch1 c3 d2 c2 d3 S h2 c2 d1 c1 d2 r 1 d1 r 3 d3 S h3 r 1 d1 r 2 d2 c3 d1 c1 d3 Ch2 c3 d1 c1 d3 S h1 c2 d1
D
c1 d2 r 2 d2 r 3 d3 S h3 r 1 d1 r 2 d2 c3 d2 c2 d3 Ch3 c2 d1 c1 d2 S h1 r 2 d2 r 3 d3 c3 d1 c1 d3 S h2 r 1 d1
r 3 d3 c3 d2 c2 d3 g;
fqt
89
90
Ghk
r 2 S h1 S h3 c2 d21 d2 r 1 S h2 S h3 c2 d21 d2 r 1 c3 d21 d2 r 2 S h1 S h2 c3 d21 d2 r2 S h1 S h3 c1 d1 d22 r 1 S h2 S h3 c1 d1 d22 r 2 c3 d1 d22
D
r1 S h1 S h2 c3 d1 d22 r 1 c2 d21 d3 r 3 S h1 S h3 c2 d21 d3 r3 S h1 S h2 c3 d21 d3 r 1 S h2 S h3 c3 d21 d3 2r1 c1 d1 d2 d3 r 2 S h1 S h2 c1 d1 d2 d3
r3 S h1 S h3 c1 d1 d2 d3 2r 2 c2 d1 d2 d3 r1 S h1 S h2 c2 d1 d2 d3 r 3 S h2 S h3 c2 d1 d2 d3 2r3 c3 d1 d2 d3 r 1 S h1 S h3 c3 d1 d2 d3
r2 S h2 S h3 c3 d1 d2 d3 r 2 c1 d22 d3 r 3 S h2 S h3 c1 d22 d3 r3 S h1 S h2 c3 d22 d3 r 2 S h1 S h3 c3 d22 d3 r 3 S h1 S h2 c1 d1 d23 r1 S h2 S h3 c1 d1 d23
r3 c2 d1 d23 r 1 S h1 S h3 c2 d1 d23 r 3 c1 d2 d23 r 2 S h2 S h3 c1 d2 d23 r 3 S h1 S h2 c2 d2 d23 r 2 S h1 S h3 c2 d2 d23 Ch2 Ch3 c1 d2 d3 r 2 d2 r 3 d3
r1 d21 c3 d2 c2 d3 d1 r2 c3 d22 2r 1 c1 d2 d3 r3 c2 d23 Ch1 Ch3 c1 d2 d3 r 2 d2 r 3 d3 r 1 d21 c3 d2 c2 d3
d1 r 2 c3 d22 2r 2 c2 d2 d3 r 3 c2 d23
Ch2 c1 d2 d3 r 2 d2 r 3 d3 r 1 d21 c3 d2 c2 d3 d1 r2 c3 d22 2r 3 c3 d2 d3 r 3 c2 d23
;
D
r 3 c2 d1 r 2 c3 d1 r 3 c1 d2 r1 c3 d2 r 2 c1 d3 r 1 c2 d3 S h2 S h3 c3 d2 c2 d3 S h1 S h2 c2 d1 c1 d2 S h3 c3 d1 c1 d3 ;
D
D
r 3 c2 d1 r 2 c3 d1 r 3 c1 d2 r 1 c3 d2 r2 c1 d3 r 1 c2 d3 Ch3 S h1 d1 S h2 d2 Ch2 S h1 d1 S h3 d3 Ch1 S h2 d2 S h3 d3 ;
D
91
fmm
92
fmt
93
ftt
3
Gh
r 2 S h1 S h3 c22 d21 r2 S h1 S h2 c2 c3 d21 r 3 S h1 S h3 c2 c3 d21 r 3 S h1 S h2 c23 d21 r 2 S h1 S h3 c1 c2 d1 d2 r 1 S h2 S h3 c1 c2 d1 d2 r1 c1 c3 d1 d2
12D
r3 S h1 S h3 c1 c3 d1 d2 r 2 c2 c3 d1 d2 r 3 S h2 S h3 c2 c3 d1 d2 2r 3 c23 d1 d2 r 1 S h2 S h3 c21 d22 r1 S h1 S h2 c1 c3 d22 r 3 S h2 S h3 c1 c3 d22
r3 c3 d3 Ch3 c1 r 2 c2 r3 c3 d2 d3 d1 r 1 c1 r 2 c2 c3 d2 c2 r 1 c1 2r 2 c2 r 3 c3 d3 ;
94
410
where
D 2c3 d2 c2 d3 1 Ch1 Ch3 S h2 c2 d1 c1 d2 1 Ch1 Ch2 S h3 c3 d1 c1 d3 S h1 2c2 d1 c1 d2 c3 d1
c1 d3 2Ch2 Ch3 c2 d1 c1 d2 c3 d1 c1 d3 S h2 S h3 c22 c23 d21 c21 c23 d22 2c1 c3 d1 d3 c21 c22 d23 2c2 d2 c1 d1 c3 d3
95
References
[1] Zienkiewicz OC, Taylor, RL. The nite element method. The basis. 5th ed. vol. 1.
Butterworth-Heinemann, 2000.
[2] Banerjee JR. Dynamic stiffness formulation for structural elements: a general
approach. Comput Struct 1997;63(1):1013.
[3] Banerjee JR. Coupled bending-torsional dynamic stiffness matrix for beam
elements. Int J Numer Methods Eng 1989;28:128398.
[4] Banerjee JR. Free vibration analysis of a twisted beam using the dynamic
stiffness method. Int J Solids Struct 2001;38(3839):670322.
[5] Banerjee JR. Free vibration of sandwich beams using the dynamic stiffness
method. Comput Struct 2003;81(1819):191522.
[6] Banerjee JR. Development of an exact dynamic stiffness matrix for free
vibration analysis of a twisted Timoshenko beam. J Sound Vib 2004;270(1
2):379401.
[7] Banerjee JR, Su H, Jackson DR. Free vibration of rotating tapered beams using
the dynamic stiffness method. J Sound Vib 2006;298(4-5):103454.
[8] Banerjee JR, Cheung CW, Morishima R, Perera M, Njuguna J. Free vibration of a
three-layered sandwich beam using the dynamic stiffness method and
experiment. Int J Solids Struct 2007;44(2223):754363.
[9] Williams FW, Wittrick WH. An automatic computational procedure for
calculating natural frequencies of skeletal structures. Int J Mech Sci
1970;12(9):78191.
[10] Anderson MS, Williams FW. BUNVIS-RG: exact frame buckling and vibration
program, with repetitive geometry and substructuring. J Spacecraft Rockets
1987;24:35361.
[11] Akkeson B. A computer program for plane frame vibration analysis by an exact
method. Int J Numer Methods Eng 1976;10:122131.
[12] Wittrick WH, Williams FW. A general algorithm for computing natural
frequencies of elastic structures. Quarterly J Mech Appl Math
1970;24(3):26384.
[13] Williams FW, Howson WP. Compact computation of natural frequencies and
buckling loads for plane frames. Int J Numer Methods Eng 1977;11:106781.
[14] Wittrick WH. A unied approach to initial buckling of stiffened panels in
compression. Int J Numer Methods Eng 1968;11:106781.
[15] Wittrick WH. General sinusoidal stiffness matrices for buckling and vibration
analyses of thin at-walled structures. Int J Mech Sci 1968;10:94966.
[16] Williams FW, Wittrick WH. Computational procedures for a matrix analysis of
the stability and vibration of thin at-walled structures in compression. Int J
Mech Sci 1969;11(12):97998.
[17] Wittrick WH, Williams FW. Buckling and vibration of anisotropic or isotropic
plate assemblies under combined loadings. Int J Mech Sci 1974;16(4):20939.
[18] Williams FW, Kennedy D, Butler R, Anderson MS. VICONOPT: program for
exact vibration and buckling analysis or design of prismatic plate assemblies.
AIAA J 1991;29:19278.
[19] Anderson MS, Kennedy D. Inclusion of transverse shear deformation in exact
buckling and vibration of composite plate assemblies. Tech. Rep. NASA/CR1993-4510, Langley, Hampton; 1993.
[20] Anderson MS, Kennedy D. Transverse shear deformation in exact buckling and
vibration of composite plate assemblies. AIAA J 1993;31(10):19635.
[21] Reddy JN, Phan ND. Stability and vibration of isotropic, orthotropic and
laminated plates according to a higher-order shear deformation theory. J
Sound Vib 1985;98(2):15770.
[22] Carrera E, Boscolo M. Classical and mixed nite elements for static and
dynamic analysis of piezoelectric plates. Int J Numer Methods Eng
2007;70:113581.
[23] Leung AYT, Zhou WE. Dynamic stiffness analysis of laminated composite
plates. Thin-Walled Strcut 1996;25(2):10933.
[24] Mindlin RD. Inuence of rotatory inertia and shear on exural vibrations of
isotropic, elastic plates. J Appl Mech 1951;18:10316.
[25] Wolfram S. The mathematica book. 4th ed. Cambridge University Press; 1999.
[26] Reddy JN. Theory and analysis of elastic plates. CRC Press; 2006.
[27] Reddy JN, Miravete A. Practical analysis of composite laminates. CRC Press;
2000.
[28] Timoshenko S, Woinowsky-Krieger S. Theory of plates and shells. McGraw-Hill
Book Company; 1959.
[29] Lowe PG. Basics principles of plate theory. Surrey University Press; 1982.
[30] Reissner E. On the theory of bending of elastic plates. J Math Phys
1944;23(4):18491.
[31] Ince EL. Ordinary differential equations. Dover Publications; 1920.
[32] Ferrar WL. Higher algebra. Oxford: Clarendon Press; 1948.
[33] Leissa AW. Vibration of plates. Tech. Rep. NASA SP-160, National Aeronautics
and Space Administration, Washington; 1969.
[34] Chopra I. Vibration of stepped thickness plates. Int J Mech Sci 1974;16:33744.
[35] Guo SJ, Keane Aj, Moshere-Torbati M. Vibration analysis of stepped thickness
plates. J Sound Vib 1997;204(4):64557.
[36] Cheung YK. Finite strip method in structural analysis. Pergamon Press; 1976.