ICSV14
Cairns • Australia
9-12 July, 2007
A SPECTRAL FINITE ELEMENT MODEL FOR THIN PLATE
VIBRATION
Mohammad Tawfik
Assistant Professor, Modeling and Simulation in Mechanics Department, German University in Cairo,
Cairo, Egypt,
Mohammad.Tawfik@guc.edu.eg
Abstract
The finite element analysis of plate vibration has become one of the classical problems that
received a lot of attention from the researchers through the past few decades. Different models
were developed including the classical polynomial elements, the hierarchical finite element
models, and most recently, the spectral finite element models. Developers of different models
introduced there models and performed many studies to prove that the models were suitable
and accurate with validations performed against analytical models, when available, and with
other numerical models. In this study, the first plate spectral finite element model is presented
with a generalized methodology for the derivation element matrices. Three different finite
element models will be compared in the study of free vibration characteristics of an isotropic
plate with different boundary conditions. The aim of the study is to point out the points of
weakness and strength of each model and to emphasise the ease and interchangeability of the
models. The models compared are a classical 3rd order/4-node element, a 7th order/ 16 node
element, and the proposed spectral finite element model.The models were created using a
symbolic manipulator, Mathematica® 4.1, in order to get the elements' matrices in closed
form to avoid the errors introduced by the numerical integration that is usually used in
creating the element matrices. Results are then compared with those obtained using numerical
integration performed by creating a similar code using MATLAB 6.1.
1. INTRODUCTION
The spectral finite element method is a development of the dynamic stiffness matrix
approach. The dynamic stiffness matrix approach is distinct from conventional finite element
in that it depends on the exact solution of the differential equations involved with the problem
[1]; that solution is used instead of the polynomial trial functions to result in a finite-elementlike model that could capture the exact dynamics of the structure.
The higher accuracy of the dynamic stiffness matrix method arises from the fact that the
higher the frequency of vibration, the shorter the wave length of the propagating wave; thus,
to accurately capture the dynamics of the structure, finer mesh is required for conventional
finite element models, which in turn means larger model and more computational effort. On
the other hand, the nature of the dynamic stiffness matrix is that the dynamic solution is
ICSV14 • 9-12 July 2007 • Cairns • Australia
embedded in the matrix model of the element [2], thus, less elements are required to present
the structure with the same or higher accuracy compared to conventional finite element
method.
Langley [3] suggested the application of dynamic stiffness matrix method for the study
of free and forced vibration of aircraft panels, and later [4], he coupled it with boundary
element method to study the acoustic radiations inside an aircraft fuselage. In both studies, the
accuracy of the model was illustrated.
The spectral analysis has been a method used for the approximate solution of different
types of structural vibrations problems [5]-[11]. Tawfik and Baz [12] resented, for the first
time, a spectral finite element model for a general plate. The model used numerical
integration for the evaluation of element matrices. They validated their results using classical
solutions as well as experimental results for a plate with bonded piezoelectric patches and
shunt electric circuits.
The previous survey showed that the spectral finite element model have proven higher
accuracy and efficiency in the modelling of the one-dimensional structures. It has also, proven
effective in the reduction of the finite element model for structures with viscoeleastic
components. On the other hand, the spectral finite element models for plates were only
developed for the Levy-type plates or the system matrices had to be evaluated using
numerical integration.
In this study, a spectral finite element model is going to be presented together with a 3rd
order/4-node and 7th order/16-node models for the analysis of the free vibration characteristics
of a plate with different boundary conditions. The main aim of this study is to emphasise the
simplicity of developing a generalized finite element model for different structure vibration
problems as well as presenting a solid base comparison of the different models.
2. PLATE FINITE ELEMENT MODEL
In this section, the different trial functions
used to model the plate are going to be
presented, then, the generalized finite element
model of a thin plate element, using the
classical plate theory, will be presented
independent of the trial functions. All the
elements presented are assumed rectangular
elements that are aligned with the global
coordinates.
2.1 Trial Functions
2.1.1 Conventional
element
3rd
order,
16-DOF,
Figure 1. Node numbering scheme for the 3rd
order and spectral finite element models.
The 4-node element could be presented by the sketch in Figure 1. The transverse
displacement w(x,y) at any location x and y inside the plate element is expressed by
w( x, y ) = a1 + a2 x + a3 y + a4 xy + a5 x 2
+ a6 y 2 + a7 x 2 y + a8 xy 2 + a9 x 3 + a10 x 2 y
+ a11 xy 2 + a12 y 3 + a13 x 3 y + a14 x 2 y 2 + a15 xy 3 + a16 x 3 y 3
(1)
ICSV14 • 9-12 July 2007 • Cairns • Australia
where coefficients a1 through a16 are to be determined in terms of the nodal displacements
2.1.2 7th order, 64-DOF, element
To present the displacement distribution
in the 16-node element presented in Figure 2,
a 64-term polynomial may be used that
contains all the x and y-powers up to the 7th
order. The unknown coefficients of the
polynomial will extend from a1 through a64.
2.1.3 Spectral Element Trial Function
The proposed spectral finite element model is
based on the use of exponential functions as
trial functions instead of the usual
polynomials that are used with conventional
finite element models. The transverse
displacement w(x,y) at any location x and y
inside the plate element may be expressed by
w( x, y ) = a1e k x x e
ky y
+ a5 e − k x x e
+ a9 e ik x x e
+ a2 e k x x e
ky y
ky y
+ a13 e −ik x x e
Figure 2. Node numbering scheme for the 7th
order 16-node element.
−ky y
+ a6 e −k x x e
+ a10 e ik x x e
ky y
+ a3 e k x x e
−k y y
−k y y
+ a14 e −ik x x e
ik y y
+ a7 e − k x x e
+ a11e ik x x e
−k y y
+ a4 e k x x e
ik y y
ik y y
+ a15 e −ik x x e
− ik y y
+ a8 e −k x x e
+ a12 e ik x x e
ik y y
−ik y y
−ik y y
+ a16 e −ik x x e
(2)
−ik y y
where kx is the wave number in the x-direction and ky is the wave number in the y-direction;
and:
ω ρh
, and k x = kCos(θ ), k y = kSin(θ )
k =
D
2
1
4
where k is the wave number of a planar wave
propagating at an angle θ measured
anticlockwise from the positive x-axis
(Figure 3), ρ is the mass density, h plate
thickness, and D is the plate flexural rigidity
given by
Eh3
D=
12 1 − ν 2
Also, the coefficients a1 through a16 are to be
determined in terms of the nodal
displacements. The above proposed trial
function is a generalization of the beam trial
function proposed by Doyle [5].
(
)
Figure 3. Planar wave propagating in a plate.
ICSV14 • 9-12 July 2007 • Cairns • Australia
2.1.4 Unknown Coefficients in Terms of Nodal Displacement
In general, the displacement trial function could be written as,
w( x, y ) = H w {a}
(3)
where {a} = a1 L a16 for 3rd order and spectral element and {a} = a1 L a64 for the
7th order element.
For the elements under consideration, 4 degrees of freedom are associated with each
node; namely, w for the displacement, wx & wy for the slope in the x and y-directions
respectively, and wxy for the cross derivative of displacement. Thus, we may write,
w Hw
w H
x w, x
= H {a}
w y w, y
wxy H w, x , y
(4)
Where the subscript (,x & ,y) indicate the derivatives in the x and y-directions respectively.
Substituting the nodal coordinates into equation (4), we obtain the nodal bending
displacement vector {wb} in terms of {a} as follows,
{ wb } = [ Tb ]{ a }
(5)
where {wb} and [Tb] are the element degrees of freedom and the transformation matrix
respectively.
From equation (5), we obtain
{a} = [Tb ]−1 {wb }
Substituting equation (6) into equation (5) gives
w( x, y ) = [ H w ][ Tb ]
−1
{ wb } = [ N w ]{ wb }
(6)
(7)
where [Nw] is the shape function for bending given by
[ N w ] = [ H w ][ Tb ]−1
(8)
2.2 Strain Displacement Relations
Consider the lateral deflection; the classical plate strain-displacement relation for the lateral
deflections of thin plates can be written as follows
ICSV14 • 9-12 July 2007 • Cairns • Australia
∂2w
− 2
ε x ∂2x
{ε } = ε y = z − ∂ w2 = z{κ } = z[Cb ][Tb ]−1{wb } = z[Bb ]{wb }
γ ∂y2
xy
∂ w
− 2 ∂x∂y
(9)
Hw
, xx
[Cb ] = − H w, xyy
2 H
w, xy
(10)
Where {ε } is the strain vector, z is the vertical distance measure from the plate mid plane, and
the curvature vector {κ}, and
Thus, the strain-nodal displacement relation can be written as
{ε } = z{κ } = z[ Bb ]{ wb }
(11)
2.3 Element Matrices
Principal of virtual work states that
δΠ = δ (U − T ) = 0
(12)
where Π is the total energy of the system, U is the strain energy, T is the kinetic energy, and
δ(.) denotes the first variation.
2.3.1 The Potential Energy
The variation of the potential energy for a thin plate is given by
δU = ∫ {δε }T {σ }dV
(13)
V
where {σ} is the stress vector and V is the volume of the structure. Substituting from
equations (15) we get,
δU = ∫ z 2 {δwb }T [Bb ]T [Q ][Bb ]{wb }dV
V
= {δwb } [Tb ]
T
−T
h3
[Cb ][Q][Cb ]dA.[Tb ]−1 {wb }
∫
12 A
= {δwb } [k b ]{wb }
(14)
T
where z is the vertical distance measured from the plate mid-plane, []
. is the transpose of the
inverse, [kb] is the element bending stiffness matrix, and [Q] is the stress strain constitutive
elation. In this study, only isotropic plates are going to be used to illustrate the procedure and
−T
ICSV14 • 9-12 July 2007 • Cairns • Australia
the elements' differences, however, the same procedure is applicable to general orthographic
plates as well as composite plates.
2.3.2 The Kinetic Energy
The variation of the kinetic energy T of the plate element, ignoring the rotary inertia, is given
by,
δT = ∫ δwρ
V
∂2w
dV
∂t 2
(15)
where ρ is the density. The above equation can be rewritten in terms of nodal displacements
as follows
∂2w
&&b }dV
δT = ∫ δwρ 2 dV = ∫ {δwb }T [N w ]T ρ [N w ]{w
∂
t
V
V
= {δwb } [Tb ]
T
−T
∫ H
&&b }
= {δwb } [mb ]{w
A
w
T
ρ H w dA.{w&&b }
(16)
T
where [mb] is the element bending mass matrix. Finally, the element equation of free vibration
can be written as
[mb ]{w&&b } + [kb ]{wb } = {0}
(17)
Note that the matrices constituting the above equation are all dependent on the driving
frequency for the case of the spectral finite element model. Thus, the eigenvalue problem can
not be solved directly for the system natural frequencies; rather, an iterative method should be
used to obtain the eigenvalues.
2.4 Steps for Evaluating the Spectral Element Matrices
1. Determine the frequency at which the system is vibrating.
2. Evaluate the interpolation function, Hw, and its derivatives, H w, x , H w, y , and H w, xy ,at all
the element nodes
3. Evaluate the matrix Tb then invert it
4. Evaluate the integrals of the system matrices. Note that the matrix Tb is constant with
respect to the integrals which enable the simplification of the integral by delaying the
multiplication of the matrix until the end of the integral evaluation.
Note that the above steps are generic for any finite element model for the modelling or
rectangular plates except for that the frequency loop is must included for the spectral model.
Nevertheless, the integral mentioned in the fourth step could be evaluated using analytical or
numerical methods.
To get the system Eigenvalues for the proposed spectral finite element model, the problem
is solved using any, reasonable, initial frequency; then, the resulting eigenvalue should be
used in a second iteration. The model usually converges in less than 5 iterations except for
very low frequencies (less than 10Hz).
ICSV14 • 9-12 July 2007 • Cairns • Australia
3. NUMERICAL RESULTS
A symbolic manipulator, Mathematics® 4.1, was utilized for writing the finite element
procedure. The integrals needed for the element matrices where, thus, evaluated analytically
for the square plate elements used. Note that, each 7th order element introduces the same
number of nodes / DOF's that is introduced by the 3rd order and the spectral element; thus, in
all the results presented, the number of elements used for both models is devisable by 9.
3.1 Convergence
The convergence of the different models was tested against analytical results for a simply
supported plate as well as classical solutions presented for plates with clamped and free
boundary conditions. Table 1 through Table 3 presents the normalized frequency parameter
for a simply supported (SSSS) square plate; where the normalized frequency parameter is
given by the closed form solution given by
ω =ω
a2
π
ρh
2
D
= n2 + m2
(18)
where ω is the normalized frequency parameter, ω is the natural frequency of the plate, a is
the plate length, h is the plate thickness, and n and m are the mode number in the x and y
directions respectively.
The results presented in Table 1 through Table 3 show that all the proposed finite
element models are capable of converging fast for higher modes of vibration. It is also
noticeable that the error of the 3rd order and spectral models are similar for different modes
and different mesh sizes. Meanwhile, for the 7th order model, the error is higher in finer mesh
sizes!
Table 1 Normalized frequency parameter convergence for an SSSS plate compared to exact results for
spectral finite element model
ω
Exact
#
1
2
3
4
5
6
7
8
9
10
n
1
1
2
2
1
3
2
3
1
4
M
1
2
1
2
3
1
3
2
4
1
2
5
5
8
10
10
13
13
17
17
3x3
Elements
ω
2.07
5.18
5.18
8.43
10.73
11.03
14.44
14.44
18.92
18.92
Error
3.7%
3.7%
3.7%
5.4%
7.3%
10.3%
11.1%
11.1%
11.3%
11.3%
6x6
Elements
ω
2.04
5.07
5.07
8.19
10.09
10.14
13.31
13.31
17.22
17.22
Error
2.0%
1.4%
1.4%
2.4%
0.9%
1.4%
2.4%
2.4%
1.3%
1.3%
9x9
Elements
ω
2.03
5.05
5.05
8.12
10.05
10.07
13.18
13.18
17.09
17.09
Error
1.4%
0.9%
0.9%
1.5%
0.5%
0.7%
1.4%
1.4%
0.5%
0.5%
12x12
Elements
ω
2.02
5.04
5.04
8.09
10.04
10.05
13.13
13.13
17.05
17.05
Error
1.1%
0.7%
0.7%
1.2%
0.4%
0.5%
1.0%
1.0%
0.3%
0.3%
15x15
Elements
ω
2.02
5.03
5.03
8.07
10.03
10.04
13.11
13.11
17.04
17.04
Error
0.9%
0.6%
0.6%
0.9%
0.3%
0.4%
0.8%
0.8%
0.2%
0.2%
Table 2 Normalized frequency parameter convergence for an SSSS plate compared to exact results for
the 3rd order element.
#
n
m
Exact
3x3
Elements
6x6
Elements
9x9
Elements
12x12
Elements
15x15
Elements
ICSV14 • 9-12 July 2007 • Cairns • Australia
ω
1
2
3
4
5
6
7
8
9
10
1
1
2
2
1
3
2
3
1
4
1
2
1
2
3
1
3
2
4
1
ω
2.07
5.21
5.21
8.50
11.20
11.58
15.26
15.26
19.82
19.82
2
5
5
8
10
10
13
13
17
17
Error
3.7%
4.2%
4.2%
6.3%
12.0%
15.8%
17.4%
17.4%
16.6%
16.6%
ω
2.04
5.07
5.07
8.20
10.10
10.15
13.33
13.33
17.29
17.29
ω
Error
2.0%
1.5%
1.5%
2.5%
1.0%
1.5%
2.5%
2.5%
1.7%
1.7%
2.03
5.05
5.05
8.12
10.05
10.07
13.19
13.19
17.10
17.10
Error
1.4%
1.0%
1.0%
1.5%
0.5%
0.7%
1.4%
1.4%
0.6%
0.6%
ω
2.02
5.04
5.04
8.09
10.04
10.05
13.13
13.13
17.06
17.06
Error
1.1%
0.7%
0.7%
1.2%
0.4%
0.5%
1.0%
1.0%
0.3%
0.3%
ω
2.02
5.03
5.03
8.07
10.03
10.04
13.11
13.11
17.04
17.04
Error
0.9%
0.6%
0.6%
0.9%
0.3%
0.4%
0.8%
0.8%
0.2%
0.2%
Table 3 Normalized frequency parameter convergence for an SSSS plate compared to exact results for
the 7th order element.
1x1
Elements
ω
ω
Exact
#
1
2
3
4
5
6
7
8
9
10
n
1
1
2
2
1
3
2
3
1
4
m
1
2
1
2
3
1
3
2
4
1
2
5
5
8
10
10
13
13
17
17
2.08
5.38
5.38
8.99
10.29
11.02
14.95
14.95
17.96
17.96
2x2
Elements
Error
4.0%
7.6%
7.6%
12.3%
2.9%
10.2%
15.0%
15.0%
5.6%
5.6%
ω
2.16
5.19
5.19
8.18
10.07
10.68
13.58
13.58
17.44
17.44
3x3
Elements
ω
Error
8.0%
3.8%
3.8%
2.2%
0.7%
6.8%
4.5%
4.5%
2.6%
2.6%
2.11
5.18
5.18
8.46
10.12
10.17
13.42
13.42
17.24
17.24
Error
5.5%
3.7%
3.7%
5.8%
1.2%
1.7%
3.2%
3.2%
1.4%
1.4%
4x4
Elements
ω
2.09
5.14
5.14
8.35
10.14
10.18
13.48
13.48
17.12
17.12
Error
4.3%
2.8%
2.8%
4.3%
1.4%
1.8%
3.7%
3.7%
0.7%
0.7%
5x5
Elements
ω
2.07
5.11
5.11
8.28
10.12
10.14
13.39
13.39
17.14
17.14
Error
3.6%
2.3%
2.3%
3.5%
1.2%
1.4%
3.0%
3.0%
0.8%
0.8%
Table 4 through Table 6 present the frequency parameter results obtained for a clamped
square plate (CCCC) with Poisson’s ratio of 0.3 compared to classical results presented in the
book by Leissa [13]. The frequency parameter ω is given by the equation
*
ω = π 2ω
*
(19)
It can be easily noticed from the results presented that very reasonable results were
obtained for this case with only 36 elements in the cases of 3rd order and spectral elements and
only 4 elements for the case of 7th order element.
Table 4 Frequency parameter convergence for CCCC plate compared to classical results [13] for the
spectral element
3x3
Elements
#
1
2
3
4
5
Classical
35.11
72.93
72.93
107.5
131.7
ω
*
36.14
74.92
74.92
111.69
156.19
Error
2.9%
2.7%
2.7%
3.9%
18.6%
6x6
Elements
ω
*
36.00
73.52
73.52
108.48
132.46
Error
2.5%
0.8%
0.8%
0.9%
0.6%
9x9
Elements
ω
*
35.99
73.42
73.42
108.27
131.76
Error
2.5%
0.7%
0.7%
0.7%
0.1%
12x12
Elements
ω
*
35.99
73.40
73.40
108.24
131.64
Error
2.5%
0.6%
0.6%
0.7%
0.0%
15x15
Elements
ω
*
35.99
73.40
73.40
108.22
131.60
Error
2.5%
0.6%
0.6%
0.7%
0.0%
ICSV14 • 9-12 July 2007 • Cairns • Australia
6
7
8
9
10
131.7
164.4
164.4
210.3
210.3
156.59
189.27
189.27
300.24
300.24
18.9%
15.2%
15.2%
42.7%
42.7%
133.06
166.00
166.00
214.01
214.01
1.1%
1.0%
1.0%
1.7%
1.7%
132.38
165.21
165.21
211.31
211.31
0.6%
0.5%
0.5%
0.5%
0.5%
132.26
165.07
165.07
210.78
210.78
0.5%
0.4%
0.4%
0.2%
0.2%
132.22
165.02
165.02
210.59
210.59
0.4%
0.4%
0.4%
0.1%
0.1%
Table 5 Frequency parameter convergence for CCCC plate compared to classical results [13] for the
3rd order element
3x3
Elements
#
1
2
3
4
5
6
7
8
9
10
Classical
35.11
72.93
72.93
107.5
131.7
131.7
164.4
164.4
210.3
210.3
ω
*
36.20
75.01
75.01
111.82
156.53
156.93
189.61
189.61
300.61
300.61
Error
3.1%
2.9%
2.9%
4.0%
18.9%
19.2%
15.4%
15.4%
42.9%
42.9%
6x6
Elements
ω
*
Error
2.5%
0.8%
0.8%
0.9%
0.6%
1.1%
1.0%
1.0%
1.8%
1.8%
36.00
73.53
73.53
108.49
132.47
133.07
166.02
166.02
214.03
214.03
9x9
Elements
ω
12x12
Elements
ω
*
35.99
73.42
73.42
108.28
131.77
132.39
165.22
165.22
211.31
211.31
Error
2.5%
0.7%
0.7%
0.7%
0.1%
0.6%
0.5%
0.5%
0.5%
0.5%
15x15
Elements
*
Error
2.5%
0.6%
0.6%
0.7%
0.0%
0.5%
0.4%
0.4%
0.2%
0.2%
35.99
73.40
73.40
108.24
131.64
132.26
165.07
165.07
210.78
210.78
ω
*
Error
2.5%
0.6%
0.6%
0.7%
0.0%
0.4%
0.4%
0.4%
0.1%
0.1%
35.99
73.40
73.40
108.23
131.61
132.23
165.03
165.03
210.63
210.63
Table 6 Frequency parameter convergence for CCCC plate compared to classical results [13] for the
7th order element
#
1
2
3
4
5
6
7
8
9
10
Classical
35.11
72.93
72.93
107.5
131.7
131.7
164.4
164.4
210.3
210.3
ω
3x3
Elements
*
35.99
73.42
73.42
108.26
137.29
138.07
168.81
168.82
230.45
231.15
Error
2.5%
0.7%
0.7%
0.7%
4.3%
4.9%
2.7%
2.7%
9.6%
9.9%
6x6
Elements
ω
*
35.99
73.39
73.39
108.22
131.59
132.22
165.01
165.01
210.58
210.58
Error
2.5%
0.6%
0.6%
0.6%
0.0%
0.4%
0.4%
0.4%
0.1%
0.1%
ω
9x9
Elements
*
35.99
73.39
73.39
108.22
131.58
132.21
165.00
165.00
210.52
210.52
Error
2.5%
0.6%
0.6%
0.6%
-0.1%
0.4%
0.4%
0.4%
0.1%
0.1%
12x12
Elements
ω
*
35.99
73.39
73.39
108.22
131.58
132.21
165.00
165.00
210.52
210.52
Error
2.5%
0.6%
0.6%
0.6%
-0.1%
0.4%
0.4%
0.4%
0.1%
0.1%
15x15
Elements
ω
*
35.99
73.39
73.39
108.22
131.58
132.21
165.00
165.00
210.52
210.52
Error
2.5%
0.6%
0.6%
0.6%
-0.1%
0.4%
0.4%
0.4%
0.1%
0.1%
Table 7 through Table 9 present the frequency parameter results for a free square plate. It
should be noticed that 0% relative error was obtained for the case of 144 spectral elements for
almost all the modes. It should also be noticed that the 9th and 10th modes had negative
relative error which is not seen as a mistake in the finite element model since the results are
already compared to classical approximate solution of the problem. Similar results were
obtained for the case of the 3rd order element model while only 4 elements were sufficient for
the 7th order model.
Table 7 Frequency parameter convergence for FFFF plate compared to classical results [13] for the
spectral element.
#
Classical
3x3
6x6
9x9
12x12
15x15
ICSV14 • 9-12 July 2007 • Cairns • Australia
Elements
ω
1
2
3
4
5
6
7
8
9
10
13.47
19.6
24.27
34.8
34.8
61.09
61.09
63.69
69.5
77.59
*
28.37
31.13
34.09
40.91
40.91
64.47
64.47
65.72
71.29
78.88
Error
110.6%
58.9%
40.5%
17.5%
17.5%
5.5%
5.5%
3.2%
2.6%
1.7%
Elements
ω
*
14.81
20.51
24.98
35.21
35.21
61.26
61.26
63.80
69.37
77.25
Error
10.0%
4.7%
2.9%
1.2%
1.2%
0.3%
0.3%
0.2%
-0.2%
-0.4%
Elements
ω
*
13.75
19.78
24.41
34.88
34.88
61.12
61.12
63.71
69.28
77.19
Error
2.0%
0.9%
0.6%
0.2%
0.2%
0.0%
0.0%
0.0%
-0.3%
-0.5%
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.10
61.10
63.69
69.27
77.18
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
Elements
ω
*
Error
19.60
24.27
34.80
34.80
61.10
61.10
63.69
69.27
77.17
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
Table 8 Frequency parameter convergence for FFFF plate compared to classical results[13] for the 3rd
order element.
3x3
Elements
#
1
2
3
4
5
6
7
8
9
10
Classical
13.47
19.6
24.27
34.8
34.8
61.09
61.09
63.69
69.5
77.59
ω
*
13.49
19.64
24.36
35.01
35.01
61.45
61.45
64.55
69.58
77.63
Error
0.2%
0.2%
0.4%
0.6%
0.6%
0.6%
0.6%
1.3%
0.1%
0.1%
6x6
Elements
ω
*
13.47
19.60
24.28
34.82
34.82
61.20
61.20
63.77
69.39
77.35
Error
0.0%
0.0%
0.0%
0.1%
0.1%
0.2%
0.2%
0.1%
-0.2%
-0.3%
9x9
Elements
ω
*
13.47
19.60
24.27
34.81
34.81
61.12
61.12
63.71
69.29
77.21
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
12x12
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.10
61.10
63.69
69.27
77.19
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
15x15
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.10
61.10
63.69
69.27
77.18
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
Table 9 Frequency parameter convergence for FFFF plate compared to classical results [13] for the 7th
order element
1x1
Elements
#
1
2
3
4
5
6
7
8
9
10
3.2
Classical
13.47
19.6
24.27
34.8
34.8
61.09
61.09
63.69
69.5
77.59
ω
*
13.47
19.60
24.27
34.80
34.80
61.11
61.11
63.69
69.28
77.20
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
2x2
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.09
61.09
63.69
69.27
77.17
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
3x3
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.09
61.09
63.69
69.27
77.17
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
4x4
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.09
61.09
63.69
69.27
77.17
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
5x5
Elements
ω
*
13.47
19.60
24.27
34.80
34.80
61.09
61.09
63.69
69.27
77.17
Wave Direction
An important parameter of the spectral finite element model derived in section 2 is the wave
direction angle θ. That parameter seems to affect all the equations derived for the spectral
Error
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
-0.3%
-0.5%
ICSV14 • 9-12 July 2007 • Cairns • Australia
finite element model. In all the previous results, θ was taken as 45o. It seemed natural that the
effect of this parameter should be tested.
At this point, it should be noticed that all the simply supported symmetricantisymmetric modes (e.g. mode (1,4) and (4,1)) are the ones that give exactly the same
values for the normalized frequency parameters (see Table 1); in contrast with all other
symmetric-symmetric and anisymmetric-antisymmetric modes (e.g. mode (1,3), (3,1) and
(2,4), (4,2)).
Table 10 presents the values obtained for the normalized frequency parameter for a
square simply supported plate with different values of the direction angle θ. The number of
elements was taken to be 81 elements. It can be noticed that as the angle is deviated from the
symmetric 45o, the previous observation does not hold true anymore which can be understood
easily to be a result of the broken symmetry. Nevertheless, the normalized frequency
parameter results deviate slightly as the angle is changed from 5o to 85o. It should also be
noticed that the spectral finite element model proposed here fails as the angle becomes 0o or
90o.
Figure 4 presents the variation of the
normalized frequency parameter of modes
(1,3) and (3,1) for the plate with the change
in the wave angle from 5o to 85o. A
distinction can not be made between both
curves since, for a square plate, both should
be equal theoretically. On the other hand, it
is noticeable that the difference between
both corves is less than 0.3%. Figure 5
presents contour shading for the mode shape
of the plate for both modes when the wave
Figure 4. Variation of the normalized frequency
propagation angle is 5o (note that 5o and 85o
parameter
with the wave direction angle for modes
are similar cases). Note that there is no
(1,3)
and
(3,1)
of a simply supported square plate.
significant difference between Figure 5 and
Normalized frequency parameter (rad.)
10.09
10.08
10.08
10.07
10.07
10.06
10.06
10.05
10.05
0
10
20
30
40
50
60
70
80
90
Wave Direction (deg.)
the plot in Figure 6 which presents the plots for the modes shapes at wave propagation angle
of 45o. Nevertheless, the mode shapes shown match the theoretical expectations.
Table 10 Change of the resulting normalized frequency parameter with the wave direction angle θ
#
1
2
3
4
5
6
7
8
9
10
N
1
1
2
2
1
3
2
3
1
4
m
1
2
1
2
3
1
3
2
4
1
Exact
2
5
5
8
10
10
13
13
17
17
θ=5o
θ=15o θ=25o θ=35o θ=45o θ=55o θ=65o θ=75o θ=85o
2.03
5.05
5.05
8.13
10.05
10.08
13.19
13.21
17.07
17.16
2.03
5.05
5.05
8.13
10.05
10.08
13.19
13.20
17.06
17.14
2.03
5.05
5.05
8.12
10.05
10.07
13.18
13.19
17.07
17.12
2.03
5.05
5.05
8.12
10.05
10.07
13.18
13.19
17.07
17.10
2.03
5.05
5.05
8.12
10.05
10.07
13.18
13.18
17.09
17.09
2.03
5.05
5.05
8.12
10.05
10.07
13.18
13.19
17.07
17.10
2.03
5.05
5.05
8.12
10.05
10.07
13.18
13.19
17.07
17.12
2.03
5.05
5.05
8.13
10.05
10.08
13.19
13.20
17.06
17.14
2.03
5.05
5.05
8.13
10.05
10.08
13.19
13.21
17.07
17.16
ICSV14 • 9-12 July 2007 • Cairns • Australia
(a)
(b)
Figure 5. Mode shapes of modes (a) (3,1) and (b) (1,3) for a simply supported square plate.
Wave propagation angle 5o.
(a)
(b)
Figure 6. Mode shapes of modes (a) (3,1) and (b) (1,3) for a simply supported square plate.
Wave propagation angle 45o.
13.21
Normalized frequency parameter (rad.)
Figure 7 presents the variation of the
normalized frequency parameter of modes
(1,4) and (4,1) as obtained by the spectral
finite element model. Note the intersection of
both curves at propagation angle of 45o.
Figure 8 presents the contour shading plots of
the mode shape at wave propagation angle of
5o. Note the distinct shape of the modes (1,4)
and (4,1) that appear in those plots. On the
other hand, Figure 9 presents the plots for the
modes (1,4) and (4,1) with propagation angle
of 45o, those plots match the results
mentioned by Leissa [13].
13.21
13.20
13.20
13.19
13.19
13.18
13.18
0
10
20
30
40
50
60
70
80
90
Wave Direction (deg.)
Figure 7. Variation of the normalized frequency
parameter with the wave direction angle for
modes (1,4) and (4,1) of a simply supported
square plate.
ICSV14 • 9-12 July 2007 • Cairns • Australia
(b)
(a)
Figure 8. Mode shapes of modes (a) (1,4) and (b) (4,1) for a simply supported square plate.
Wave propagation angle 5o.
(a)
(b)
Figure 9. Mode shapes of modes (a) (4,1) and (b) (1,4) for a simply supported square plate.
Wave propagation angle 45o.
3.3
Analytical vs. Numerical Integration
A similar code was written using MATLAB® 6.1 script. Numerical integration was used for
the evaluation of the element matrices. The numerical integration used Gauss-Legendre
quadrature method for numerical integration [14]. Different number of quadrature points were
used. In all the above mentioned models, the results obtained using numerical integration was
not of significant difference that those obtained by analytical integration from as little as 5
integration points in each direction.
4. CONCLUDING REMARKS
The main objective of this study was to show the possibility of preparing a generic finite
element model for the solution of the plate vibration problem. The study used three different
models to emphasize the possibility of the generalization
ICSV14 • 9-12 July 2007 • Cairns • Australia
In this study, a new spectral finite element model is proposed, namely, using
exponential functions as trial functions instead of the conventional polynomial trial functions
which have proven superior in the one-dimensional structures. The model integrals and
routines for the calculations of the normalized frequency parameter were performed
symbolically using Mathematics® version 4.1.
The effect of the estimated wave direction has shown to be of minor effect on the
resulting frequency parameter. Thus, the use of 45o angle is a reasonable choice for most
problems except for those with indicated excitation direction. Also, the effect of the direction
on mode shape has shown to be of minor.
The results obtained from the different models emphasised that the use of sophisticated
models is not of great significance on the results obtained for square plate vibration problem.
REFERENCES
[1] AYT Leung and SP Zeng, “Analytical formulation of dynamic stiffness,” Journal of
Sound and Vibration 177(4), 555-564 (1994).
[2] JR Banerjee, “Dynamic stiffness formulation for structural elements: A general
approach,” Computer and Structures 63(1), 101-103 (1997).
[3] RS Langley, “Application of dynamic stiffness method to the free and forced vibrations
of aircraft panels,” Journal of Sound and Vibration 135(2), 319-331 (1989).
[4] RS Langley, “A dynamic stiffness/boundary element method for the prediction of interior
noise levels,” Journal of Sound and Vibration 163(2), 207-230 (1993).
[5] JF Doyle, “Wave propagation in structures: spectral analysis using fast discrete fourier
transforms,” Mechanical Engineering Series, 2nd ed., Springer-Verlag, 1997.
[6] R Ruotolo, “A spectral element for laminated composite beams: Theory and application
to pyroshock analysis,” Journal of Sound and Vibration 270, 149-169 (2004).
[7] U Lee, “Vibration analysis of one-dimensional structures using the spectral transfer
matrix method,” Engineering Structures 22(6), 681-690 (2000).
[8] M Krawczuk, “Application of spectral beam Finite element with a crack and iterative
search technique for damage detection,” Finite Element in analysis and Design 38, 537548 (2002).
[9] U Lee and J Kim, “Spectral element modelling for the beams treated with active
constrained layer damping,” International Journal of Solids and Structures 38(32), 56795702 (2001).
[10] G Wang and NM Wereley, “Spectral finite element analysis of sandwich beams with
passive constrained layer damping,” 40th AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference and Exhibit, St. Louis, MO, Apr. 12-15,
1999, Collection of Technical Papers. Vol. 4 (A99-24601 05-39), Reston, VA, American
Institute of Aeronautics and Astronautics, 2681-2694 (1999).
[11] F Birgersson, NS Ferguson and S Finnveden, S., “Application of the spectral finite
element method to turbulent boundary layer induced vibration of plates,” Journal of
Sound and Vibration 259(4), 873-891 (2003).
[12] M Tawfik and A Baz, "Experimental and spectral finite element study of plates with
shunted piezoelectric patches," International Journal of Acoustics and Vibration 9(2),
87-97 (2004).
[13] A Leissa, Vibration of plates, 2nd edition, Acoustical Society of America, 1993.
[14] D Zwillinger, Standard mathematical tables and formulae, CRC press, 30th ed., 1996.