Professional Documents
Culture Documents
cylindrical shell
Ruisen Ming, Jie Pan, and Michael P. Norton
Department of Mechanical and Materials Engineering, The University of Western Australia,
Nedlands 6907, Australia
INTRODUCTION
tion components of different circumferential modes. Theoretically this method has no frequency limit and can be applied to the case where the vibration is the superposition of
many circumferential modes. It has been demonstrated that
in the far wave field at frequencies below the n3 cutoff
frequency, these two methods give comparable results.6
In this paper, a new method is proposed and its theoretical basis is outlined for the measurement of structural mobilities in a circular cylindrical shell. A point force excitation
is employed instead of a desirable circumferential modal
force excitation which is difficult to implement in practice.
Therefore, the response of the shell is the superposition of
different circumferential modes. This proposed method uses
an array of phase matched accelerometers to simultaneously
measure dynamical responses at several positions around the
cross section of interest in a circular cylindrical shell. Then
the method of least squares is used during the data processing to decompose the transfer function components of different circumferential modal accelerations to the input force. A
series of experiments was carried out under laboratory conditions to verify the feasibility of this proposed method.
I. STRUCTURAL MOBILITIES
A. Prediction
Consider a circular cylindrical shell of thickness h, radius a and length L. Let the shell be referenced to a cylindrical coordinate system (r, ,x) where x is taken in the axial
direction of the shell, measures the angle in the circumferential direction, and the r axis is directed outward along the
radial direction, as shown in Fig. 1. If only a radial force per
unit area, F, acts on the shell surface and if u, v and w
represent the displacement components of the shell middle
surface in the axial, tangential and radial directions, respectively, the equation of motion ReissnerNaghdiBerry
theory for an element of shell can be written as7,8
0001-4966/2000/107(3)/1374/9/$17.00
1374
n1
v n
V ns sin n n0
n1 s1
exp ik ns xi t ,
w n
n0
n0 s1
W ns cos n n0
exp ik ns xi t ,
2
w
1 2
1 2
1 2 v
0;
2
2
2
2
2
x
2a
2a x a x
cp t
1a
1 2
1 2
1 2
1 2 u
1 2
v
2a x
2 x 2 a 2 2 c 2p t 2
1
1 3
3
2
a2
a2 3 x2
u
1
1 3
3
2
2 2 3 2
a x
a
a
x
1b
w0;
v
1
a2 2
F
2 4
,
2 1 2
2 w
a
t
B
cp
1c
2 2 v
3w
43 3 w
;
a2 x
a2 x 2 x3
2a
v 2w
2w
2 2 ;
M x D 2
a
x
2b
u v
w
N x B
x a
2c
T x
B 1 1 u
v
w
12 2 4 2
,
2)
a
x
x
2
2d
u n n0
s1
U ns cos n n0
n0
3a
Mx
Nx
T x T x0
n0
J n G n ,G n A 0n X n ;
Mx
Nx
4a
T x T xL
n0
J n G n ,G n A Ln X n ,
4b
D
k ns a 2 n 2 n v ns ;
a2
G n 3,s
B
k a n v ns ;
a ns uns
G n 4,s
1 B
n uns 12 2 k ns a v ns
2 a
exp ik ns xi ti /2 ,
1375
Q x0 S x
3c
Q xl S x
3b
4n 2 k ns a
s1,2,3,4 ;
1375
that is, F n0
F n ,0,0,0 T . The force equilibrium at the
end of x0 gives a matrix equation, Q x0 F. The boundary condition at the other end (xL) will depend on practical considerations. For a free end condition, Q xl 0. For a
simply supported end condition, M x ,N x , w/ x ,u T xl
0, where M x and N x can be obtained from Eq. 4b. For a
clamped end condition, u, v ,w, w/ x xl 0. After X n is
determined, all the components of displacement at any position of the shell can be calculated and then the structural
mobilities can be evaluated from the ratios of the corresponding velocity components at the desirable position to the
acting external circumferential force.
B. Measurement
Y nv F
vF
v n n v n n H n
,
Fn i f 0
i
where H nv F is the transfer function of the tangential acceleration component of the nth circumferential mode to the input
point force. The above equations show that the measurement
of the structural mobility for the nth circumferential mode is
the measurement of the transfer function of the nth circumferential modal acceleration component to the input point
force. In order to determine the structural mobility, it is necessary to extract the corresponding transfer function component of the desirable circumferential mode from the measured transfer function signal.
Since the accelerometer is attached on the shell outer
surface during the measurement, its output signal is not equal
to the acceleration component of the shell middle surface
unless the accelerometer main axis is lying along the radial
direction. The relationships between the measured transfer
function signals and the transfer function of the acceleration
components on the shell middle surface to the input point
force are given by7
uF
Hm
H uF d
vF
Hm
n0
F n cos n 0 ,
n nw
n n H wF
w
n
,
Fn i f 0
i
H wF
n
H wF
;
x
d
d H wF
1 H v F
;
a
a
8
wF
Hm
H wF ,
uF
Hm
n0
F f 0 x0 a a 0
H uF
n d
H wF
n
cos n n0
x
n0
vF
Hm
n1
uF
cos n n0 ;
H mn
9a
d v F nd wF
H
sin n n0
H
a n
a n
n0
vF
H mn
sin n n0 ;
9b
H wF
n cos n n0 .
9c
wF
Hm
n0
uF
Note that the modal transfer function components, H mn
,
wF
vF
H mn and H n , are complex numbers and they are characterized by an amplitude and phase or real and imaginary parts.
When the structural wave field is dominated by several circumferential modes, the measured transfer functions will be
approximately equal to the sum of those dominating transfer
1376
1377
e w
i1
wF
H m,i
H wF
n cos n n0 ,
10
1377
should be negligible. For the case of free-damped end conditions, the shell was supported at a position 0.5 m away
from its free end by a piano wire with a very soft foam pad
and its another end was buried 0.5 m in a dry sand filled
box which intended to provide an effectively absorptive termination for all circumferential modes and wave types. For
both cases, the free end of interest was driven in the radial
direction by a mechanical shaker fed with a pseudo-random
noise signal.
To measure the driving force, an impedance head was
mounted at the driving location. The shaker and the impedance head were connected by a steel rod of 30 mm in length
and 1 mm in diameter to avoid possible axial and tangential
force excitations. Since the excitation was stable, the responses at different positions around the cross section of interest were measured by one B&K 4375 accelerometer at
different times. Therefore, the measured data should contain
no phase matching error. For the measurement of axial and
tangential acceleration components, an aluminum cube of 10
mm side dimensions was used to construct a rotational accelerometer. The masses of the accelerometer and the cube
are 2.4 g and 2.7 g, respectively. The predicted normalized
mass loading errors of measured accelerations are negligible
1378
v x
w
rx
xr
um
vm ,
wm
11
where x is the tangential component of the transverse sensitivity of the accelerometer whose main axis is directed
along the axial direction; (u m , v m ,w m ) are the measured acceleration signals.
1379
1379
n0,1,2,3,4,5 are considered in the date processing for estimating the n1,2 circumferential modal mobility components from the same measurement data. It is shown that all
the results agree well, especially at frequencies close to the
resonance frequencies. The results shown in the following
figures are obtained by considering 4 n1,2,3,4 circumferential modes only from data measured on 12 positions.
Twelve measurement positions were uniformly distributed
the radial angle between the neighboring positions was 30
around the free end cross section. For the calculation of the
axial transfer function component, the first order derivative
was estimated from the data measured around the
of H wF
n
cross-section of the free end and that of 15 mm away from
the free end. Every measured transfer function signal both
real and imaginary parts was recorded in a dual-channel
real-time frequency analyzer.
B. Measurement of structural mobilities
For the cylindrical shell under test, the mobility component of the breathing n0 mode is quite small compared
with those of n1,2 circumferential modes at most frequencies, as shown in Figs. 2 and 3. The measured n0 mobility
components were much higher than the predicted ones at
1380
1380
and predicted real and imaginary parts of the input and cross
radial force mobilities of the finite circular cylindrical shell
under the free-damped end conditions for n1 and 2, respectively. Figures 8c and 10c show the negative real parts of
the cross mobility Y nv F . It can be seen that the measured
mobility curves under the free-damped end conditions is
smoother than those under the freefree end conditions
shown in Figs. 47, even at the troughs, except for the frequencies very close to the cutoff frequencies where a big
jump or fluctuation is observed. The reason is that the
damped end gave a very effectively absorptive termination.
Again, at frequencies above the n3 cutoff frequency, the
mobility curves are not correct because of the same reason
for Figs. 47. The measured mobility curves fluctuate while
the predicted ones do not. The frequency averages of the
measured real parts of the mobilities are slightly smaller than
their predicted ones while those of the measured imaginary
parts of the mobilities are a little larger than their predicted
ones. This is because the predictions were made based on the
semi-infinite circular cylindrical shell model but this assumed model does not accurately represent the actual cylindrical shell system. The damped end absorbed most but not
Ming et al.: Mobilities of shells
1381
all of the incident energy. From these figures it may be concluded that for cylindrical shells with free-damped end conditions this proposed method gives accurate results.
ACKNOWLEDGMENT
III. CONCLUSIONS
1382
1382