Active Vibration Control of a Flexible Plate Structure

using Particle Swarm Optimization

S. Julai, Member, IEEE, M. O. Tokhi, Senior Member, IEEE, M. Mohamad and I. Abd. Latiff
Department of Automatic Control and Systems Engineering
The University of Sheffield
Mappin Street, Sheffield, S1 3JD, United Kingdom
[email protected]

Abstract—This paper presents investigations of modeling the There has been a great deal of interest in the use of PSO in
flexible plate structures using particle swarm optimization (PSO) system identification because of its profound intelligent
and active vibration control (AVC) of such structures. The background and that it could be used easily. Liu et al. have
optimization technique is utilized to obtain a dynamic model of a proposed the use of PSO in parameter identification with
flexible plate structure based on auto-regressive with exogenous nonlinear model structure of permanent magnet synchronous
(ARX) input structure. The structure is subjected to two different motors (PMSM) [10]. PSO has also been used to identify
disturbance signal types, namely random, and finite duration proton exchange membrane (PEM) fuel cell model parameters
step. The fitness function for the PSO is the mean-squared error through fitting the mathematical model to actual experimental
(MSE) between the measured and estimated outputs of the plate.
data, as reported in [11]. Kwok et al. have presented a new
The validation of the algorithm is presented in both time and
model for a magnetorheological damper to represent the
frequency domains. The developed PSO modeling approach is
used for AVC system design to suppress the vibration of the hysteretic relationship between the damping force and the
flexible plate. The performance of the controller is assessed in velocity where the model parameters were identified using
terms of level of attenuation achieved in the power spectral performance-enhanced techniques based on PSO [12].
density (PSD) of the observed signal. In the present work, a model-based AVC strategy for
suppressing vibration of a flexible plate at its first five
Keywords-Particle swarm optimization, system identification,
dominant modes is developed using PSO. A feedforward AVC
active vibration control, flexible plate structures
structure is considered for optimum reduction of the vibration.
I. INTRODUCTION
II. FLEXIBLE PLATE STRUCTURES
Unwanted vibrations can have many undesirable effects in
Flexible structures offer useful advantages in terms of fast
structures, such as damage to the structure or an adjoining
response, low energy consumption, reduced mass and low cost,
component. Over the past decade, significant advances have
in comparison to their rigid and bulky counterparts. Besides of
been made in active vibration control (AVC). The technology
their advantages, a major constraint to take into consideration is
has been studied and implemented in a variety of engineering
the system vibration arising from structural flexibility. Control
situations, as reported in [1-5]. AVC consists of artificially
of flexible systems has been a challenge for researchers and
generating canceling sources to interfere destructively with the
engineers [13-14].
unwanted source and thus result in a reduction in the level of
vibration (disturbances) at desired location [6]. AVC is realized Dynamic modeling and simulation of a flexible plate
by detecting and processing the vibration by a suitable structure using the finite difference (FD) method have been
electronic controller so that when superimposed on the reported in [15], where a flat, square plate with all edges
disturbances, cancellation occurs. Due to the broadband nature clamped has been considered. The classical dynamic equation
of these disturbances, it is required that the control mechanism of a thin rectangular plate is developed using a partial
realizes suitable frequency-dependent characteristics so that differential equation (PDE) formulation derived from [16, 17]
cancellation over a broad range of frequencies is achieved [1,
7]. ∂4w ∂4w ∂4w ρ ∂2w q
4
+2 2 2
+ 4
+ = (1)
Particle swarm optimization (PSO) shares many similarities ∂x ∂x ∂y ∂y D ∂t 2 D
with evolutionary computation techniques such as Genetic
Algorithm (GA). It is inspired by the ability of flocks of birds, where w is the lateral deflection in the z direction, ρ is the
schools of fish, and herds of animals to adapt to their
mass density per unit area, q = q ( x, y ) is the transverse external
environment, find rich sources of food by implementing an
information sharing approach which possesses the properties of force at point ( x, y ) and has dimensions of force per unit area,
easy implementation and fast convergence [8, 9]. PSO has been ∂ 2 w / ∂t 2 is the acceleration in the z direction,
widely applied in continuous and discrete optimization
problems and has received great attention in systems and D = [ Eh 3 / 12(1 − υ )] is the flextural rigidity with υ
control engineering, automatic recognition, radio systems, etc. representing the Poisson ratio, h the thickness of the plate, and

This work was supported in part by the Ministry of Higher Education,

Malaysia and the University of Malaya, Kuala Lumpur, Malaysia.

Authorized licensed use limited to: UNIVERSITI TEKNOLOGI MARA. Downloaded on November 14,2023 at 04:55:32 UTC from IEEE Xplore. Restrictions apply.
E the Young’s modulus. A simulation algorithm characterizing IV. PARTICLE SWARM OPTIMIZATION
the dynamic behaviour of the plate is developed through Particle swarm optimization is a stochastic, population-based
discretisation of the PDE in (1). The simulation is developed evolutionary algorithm first proposed by Kennedy and Eberhart
within the Matlab environment. This has been validated and [19]. Starting with a randomly initialized population, each
verified in [15]. particle in PSO flies through the problem space and remembers
the best position it has seen. The particles evaluate their
III. PARAMETRIC SYSTEM IDENTIFICATION positions relative a global fitness during each iteration, and use
Parameter estimation constitutes a procedure that makes it the memories of their best positions to adjust their own
possible to adjust a model with a specific structure. For this velocities and subsequent positions. In this way, the particles
purpose, it is necessary to determine the appropriate order and tend to fly towards better and better areas of the search space.
parameters for the model that best fits input–output data In a PSO algorithm, the position vector and the velocity
obtained during the experiment. Consider the system’s input vector of the i-th particle in the d-dimensional search space can
and output at time t represented by u (t ) and y (t ) , be represented as X i = ( x i1 , x i 2 ,..., x id ) and
respectively. As shown in Fig. 1, the system input, u (t ) , is Vi = (v i1 , v i 2 ,..., v id ) , respectively. The best position of each
given to both system to be identified and the model. The particle (which corresponds to the best fitness value obtained
difference between the outputs from the system and its model, by that particle at time t) is denoted as Pi ( p i1 , pi 2 ,..., pid ) , and
i.e. y (t ) and yˆ (t ) will then be fed into PSO optimization to
the fittest particle found so far at time t as
estimate the unknown parameters for the system model. Pg ( p g1 , p g 2 ,..., p gd ) . Then the new velocities and positions

System
of the particles for the next fitness evaluation are calculated
Measured
input, u(t) using the following equations:
output, y(t)
System
+ v id (t + 1) = ω × v id (t )
Estimated –
output, yˆ (t ) + c1 × rand(•) × ( p id − x id (t )) (4)
Model
+ c 2 × Rand (•) × ( p gd − x id (t ))
Identified
parameters PSO x id (t + 1) = x id (t ) + v id (t + 1) (5)
optimization
where ω is the inertia weight , c1 and c 2 are constants known
Figure 1. Block diagram of the PSO-SF optimization approach.
as time-varying acceleration coefficients, and rand(•) and
After the maximum iteration number has reached, the Rand (•) are two separately generated uniformly distributed
estimated parameters will be used to update the system model, random numbers in the range [0, 1].
and then the above process will be repeated. Auto-regressive
with exogenous (ARX) input structure is chosen to model the Ratnaweera et al. [20] introduced the concept of time-
system [18]. This is expressed as varying acceleration coefficients (TVAC) and in addition to
the time-varying inertia weight factor to effectively control the
n m global search and convergence to the global best solution. This
yˆ (t ) = ¦ a(i) ⋅ y(t − i) + ¦ b( j) ⋅ u(t − j)
i =1 j =0
(2) modification has been considered for avoiding premature
convergence in the early stages of the search and enhancing
convergence to the global optimum solution during the latter
where a(i) , b(j) are denominator and numerator polynomial stages of the search. The expression for ω , c1 , and c 2 in (4)
coefficients, n and m are number of coefficients in the for this TVAC concept are
denominator and numerator polynomials, y, u, and ŷ are
measured output, input, and estimated output, respectively. The (max _iter − iter )
ω = ω 2 + (ω1 − ω 2 ) ×
order of the transfer function depends on n. For optimization max _iter
formula, the mean-squared error (MSE) of the difference
iter
between the outputs from the system and its model is used as  c1 = c1i + (c1 f − c1i ) × (6)
defined by max _iter
iter
S
c 2 = c 2 i + (c 2 f − c 2 i ) ×
¦ ( y(i) − yˆ (i) )
1 2
min f error = max _iter
S i =1
where iter and max_iter are current iteration number and
subject to −1 ≤ a(i ) ≤ 1 and −1 ≤ b(i ) ≤ 1 (3) maximum number of allowable iterations, respectively.

where S represents the number of input/output samples, y (i ) is Idris and Tokhi, [21], have proposed the use of spread factor
for guaranteed fast convergence as an improvement to the
the measured output and yˆ (i ) is the calculated model output. TVAC concept. They have established that by varying the
parameter, i.e. continuously modify the value of inertia weight,

Authorized licensed use limited to: UNIVERSITI TEKNOLOGI MARA. Downloaded on November 14,2023 at 04:55:32 UTC from IEEE Xplore. Restrictions apply.
it will produce superior results compared to the fixed one. For
fast convergence purposes, the particles have to know their re
location and relative distance from each other in exploring the Primary
search space. In doing so, the proposed spread factor measures Controller source Detector
the distribution of particles in the search space as well as the rf
d rg
precision and accuracy of the particles with respect to the Observed
global optimum. Precision (or spread of particles in the search Observer signal
space) refers to the maximum distance between particles in the Secondary
source
rh
best and worst positions while accuracy (or deviation of the
particles) refers to the distance of average particle position
from the global best particle. The value of spread factor varies Figure 2. AVC feedforward structure.
from the maximum range of the search space down to the
desired convergence precision and can be calculated as
G
SF = 0.5( spread + deviation) (7) YD
UM UC +
UD E + M C L H +
Y
This factor is used to modify the inertia weight as +
YC
F
ω = exp(−iter/ ( SF × max _iter ))
c1 = 2 × (1 − iter/ max _iter ) (8) Figure 3. Block diagram for AVC.

c2 = 2 The objective in Fig. 2 and 3 is to force the observed signal

Y to zero. U M and Y are detected signal and observed signal,
where c1 is linearly reduced to zero from its initial value of 2,
respectively, which are treated as system input and output to
and c 2 is maintained at 2 to ensure all particles are pulled design the controller. In regards to the state of the secondary
towards global minimum. source, the system behaviour can be characterized by two
The PSO algorithm used for optimization in this work is subsystems: when the secondary source is off, Q0 and when
mainly based on (4), (5), (7) and (8). This is denoted as PSO- the secondary source is on, Q1 :
SF. For comparison purposes, optimization using (6) has also
been carried out.
Y Y
Q0 = , Q1 = (9)
V. ACTIVE VIBRATION CONTROL UM U C =0
UM U C ≠0

A schematic diagram of the feedforward AVC structure

considered is shown in Fig. 2. An unwanted (primary) point The required optimal controller design rule given in terms of
source is considered acting on the flexible plate. The primary the transfer characteristics Q0 and Q1 can be obtained as
signal is detected by a detector located at a distance re relative
to the primary source. This detected signal is processed by a −1
controller of suitable transfer characteristics and fed to a ª Q º
C = «1 − 1 » (10)
cancelling (secondary) point source located at a distance d ¬ Q0 ¼
relative to the primary source and a distance r f relative to the
detector. The generated secondary signal is then superimposed The implementation of AVC can be achieved with the
on the primary signal so as to achieve vibration reduction at following steps:
and in the vicinity of an observation point located at distances
rg and rh relative to the primary and secondary sources, 1. Obtain transfer functions Q0 and Q1 using the
measurement of the input and the output signals when the
respectively [22].
secondary source is off, U C = 0 , and secondary source is
A block diagram of the AVC structure is shown in Fig. 3. In on, U C ≠ 0 , respectively.
this diagram, E, F, G and H are the transfer functions of the
paths through the distances re , r f , rg and rh , respectively. 2. Use (10) to obtain the corresponding transfer function of
The transfer characteristics of the detector, controller and the controller.
secondary source are represented as M, C and L, respectively. 3. Implement the controller.
U D and U C are the primary and secondary signals at the
source locations whereas Y D and YC are the corresponding VI. SIMULATION RESULTS
signals at the observation points, respectively. U M is the An aluminium type plate with specifications given in Table 1
detected signal and Y is the observed signal. was simulated. The plate was divided into 20 × 20 sections and
simulation performed with a sampling time of 0.0016 sec. Tests

Authorized licensed use limited to: UNIVERSITI TEKNOLOGI MARA. Downloaded on November 14,2023 at 04:55:32 UTC from IEEE Xplore. Restrictions apply.
were carried out with different disturbance signal types, namely − 1.441z 3 + 0.346 z 2 + 0.571z − 0.444
random, and finite duration step from t = 0.3 to t = 0.5 sec. Q0( random) = (11)
z 4 + 0.344 z 3 − 0.254 z 2 + 0.011z − 0.068
TABLE I. PARAMETERS OF THE FLEXIBLE PLATE
− 1.222 z 3 + 0.287 z 2 − 0.467 z + 0.424
Q1( random) = (12)
Parameter Value z 4 − 0.118 z 3 + 0.419 z 2 − 0.222 z − 0.057
Length, L 1.0 m
Width, W 1.0 m − 0.759 z 3 − 0.449 z 2 + 0.758 z + 0.203
Thickness, T 3.2004 × 10
−3
m
Q0(step) = (13)
z 4 + 0.617 z 3 + 0.032 z 2 + 0.174 z − 0.025
−11 2
Moment of inertia, I 5.1924 × 10 kg.m
Mass density per area, ρ 3
2.71 × 10 kg/m
2
− 0.547 z 3 − 0.599 z 2 − 0.321z + 0.546
Q1(step) = (14)
Young’s Modulus, E 10
7.11 × 10 N/m
2
z 4 − 0.074 z 3 + 0.589 z 2 − 0.066 z − 0.364
Simulation time, t 4 seconds
Poisson’s ratio, υ 0.3 The system performance with the random disturbance is
shown in Fig. 8. It is found that the spectral attenuations
achieved were 32.56 dB, 4.31 dB, 0.94 dB, 3.92 dB and 0.91
The disturbance source, detection point, secondary source, dB, at the first five modes respectively (as shown in Fig. 8 (b)).
and observation point were located at (7, 7), (8, 8), (10, 9) and For the case when the simulated plate was disturbed by finite
(11, 12). Here, the system is characterized with the detected duration step disturbance, Fig. 9 (b) shows the spectral
signal as input and observed signal as output. An order 4 was attenuation of 37.39 dB, 16.72 dB, 6.46 dB and 3.21 dB at the
formed adequate to model the system for both disturbances. first five modes respectively. Notice that PSO-AVC has
The best results for PSO-AVC optimization were obtained with successfully attenuated system vibration at the first dominant
100 individuals and maximum iteration of 400. mode for all disturbances. This attenuation affects the system
Figures 4 – 7 show results of the measured and estimated response by reducing its magnitude which can be seen in Figs.
output thus obtained with PSO subjected to different 8 (a) and 9 (a). System response in 3 dimensional plot before
disturbance signal types. As shown in Figs. 4 (a), 5 (a), 6 (a) and after cancellation can be viewed in Figs. 8 (c)-(d) and 9
and 7 (a), the PSO-SF achieved the best MSE levels of (c)-(d).
5.702 × 10 −16 , 5.805 × 10 −16 , 7.53 × 10 −16 1.273 × 10 −15 in the Table 2 summarizes the spectral attenuation obtained
74th, 74th, 172nd, and 280th generations, respectively, for through the above-mentioned strategy. The results show that
modeling using random and finite duration step signals, for significant vibration reduction has been achieved with the
both Q0 and Q1 . On the other hand, PSO-TVAC achieved PSO-AVC at the first dominant mode of the flexible plate,
higher MSE level than PSO-SF, followed by minor attenuation at the other modes.
−15 −15 −15
i.e. 4.984 × 10 , 4.891 × 10 , 1.522 × 10 and
TABLE II. SPECTRAL ATTENUATION ACHIEVED FOR RESONANCE MODES
1.688 × 10 −14 at the maximum generation of 1000. This
showed that PSO-TVAC needed longer time to achieve the Type of Attenuation mode (dB)
same level of MSE obtained with PSO-SF according to the disturbance
1 2 3 4 5
objective function assigned.
Random 32.56 4.31 0.94 3.92 0.91
Figs. 4 (b), 5 (b), 6 (b), and 7 (b) show the measured and Finite step 37.39 16.72 6.46 0.28 3.21
predicted responses of the system in the time domain. It is
noted that the two match one another closely. This is further
demonstrated by frequency domain plots of the measured and VII. CONCLUSION
estimated outputs, as shown in Figs. 4 (c), 5 (c), 6 (c), and 7 This work has investigated parametric modeling and active
(c), which indicate that the model has successfully captured the vibration control of flexible plate structure using PSO. The
system dynamics with the best model order of 4 for each of the vibration modes of the flexible plate structure have been
disturbance signals. This was also confirmed by correlation detected successfully, and this has lead to a good controller
tests using 95% confidence bands defined as r < 1.96 N , design. Auto-regressive with exogenous (ARX) input structure
where N is the data length and r is the correlation function, to has been used to identify the parameters, and the estimated
indicate if the estimated correlations are significant or not. The output was validated using correlation tests. PSO-SF has been
correlation functions for all cases were found to be within that shown to outperform PSO-TVAC in minimizing the prediction
band, indicating an adequate fitted model. The pole-zero error, resulting in a good level of accuracy of the estimated
diagrams (Figs. 4 (d), 5 (d), 6 (d), and 7 (d)) show that all poles model. The performances of the control system with different
and zeros lie inside the unit circle, indicating that the models types of disturbances have also been assessed. Based on the
are stable with the transfer functions results achieved, the PSO-AVC method has performed
successfully in suppressing the vibration of the simulated
flexible plate.

Authorized licensed use limited to: UNIVERSITI TEKNOLOGI MARA. Downloaded on November 14,2023 at 04:55:32 UTC from IEEE Xplore. Restrictions apply.
 -14 -6 -13 -5
x 10 x 10 x 10 x 10
1.5 2 2 3
PSO-SF optimization PSO-SF estimated PSO-SF optimization PSO-SF estimated
PSO-TVAC optimization 1.5 Measured output PSO-TVAC optmization Measured output
2
Best MSE PSO-SF = 5.702e-16 1.5 Best MSE PSO-SF = 7.531e-16
1
1

Deflection (m)

Deflection (m)
-15
0.5 x 10 1
MSE

MSE
1
0 6
0
0.5 4
-0.5 2
0.5
-1
-1 100 200 300

0 -1.5 0 -2
0 50 100 150 200 250 300 0 1 2 3 4 0 50 100 150 200 250 300 0 1 2 3 4
No. of iteration Time (sec) No. of iteration Time (sec)

(a) MSE vs. no. of iteration (b) Measured and estimated outputs (a) MSE vs. no. of iteration (b) Measured and estimated outputs

-2
10 1 1
PSO-SF estimated PSO-SF estimated
-2
Measured output 10 Measured output
-3
10 0.5 0.5

Imaginary axis
Imaginary axis
Magnitude (dB)

Magnitude (dB)
-4 -3
10 0 10 0

-5
10 -0.5 -0.5
-4
10

-6
10 -1 -1
0 50 100 150 -1 -0.5 0 0.5 1 0 50 100 150 -1.5 -1 -0.5 0 0.5 1
Frequency (rad/sec) Real axis Frequency (rad/sec) Real axis

(c) Spectral density of the outputs (d) Pole-zero diagram (c) Spectral density of the outputs (d) Pole-zero diagram

Figure 4. PSO prediction in characterizing Q0 using random disturbance. Figure 6. PSO prediction in characterizing Q0 using finite step disturbance.

-14 -6 -12 -5
x 10 x 10 x 10 x 10
1.4 1
PSO-SF optimization PSO-SF estimated PSO-SF optimization 3 PSO-SF estimated
1.2 2
PSO-TVAC optimization Measured output PSO-TVAC optimization Measured output
0.8
1 Best MSE PSO-SF = 5.805e-16 Best MSE PSO-SF = 1.273e-15 2
Deflection (m)

Deflection (m)
1 -14
0.8 0.6 x 10 1
MSE

MSE

3
0.6 0 0.4 2 0
0.4 1
0.2 -1
0.2 -1 0 100 200 300
-2
0 0
0 50 100 150 200 250 300 0 1 2 3 4 0 50 100 150 200 250 300 0 1 2 3 4
No. of iteration Time (sec) No. of iteration Time (sec)

(a) MSE vs. no. of iteration (b) Measured and estimated outputs (a) MSE vs. no. of iteration (b) Measured and estimated outputs

1 1
-3 PSO-SF estimated PSO-SF estimated
10 Measured output 10
-2
Measured output
0.5 0.5
Imaginary axis
Imaginary axis
Magnitude (dB)

Magnitude (dB)

-4
10
-3
0 10 0

-5
10
-0.5 -0.5
-4
10
-6
10 -1 -1
0 50 100 150 -1 -0.5 0 0.5 1 0 50 100 150 -1 -0.5 0 0.5 1
Frequency (rad/sec) Real axis Frequency (rad/sec) Real axis

(c) Spectral density of the outputs (d) Pole-zero diagram (c) Spectral density of the outputs (d) Pole-zero diagram

Figure 5. PSO prediction in characterizing Q1 using random disturbance. Figure 7. PSO prediction in characterizing Q1 using finite step disturbance.

Authorized licensed use limited to: UNIVERSITI TEKNOLOGI MARA. Downloaded on November 14,2023 at 04:55:32 UTC from IEEE Xplore. Restrictions apply.
 -6
2
x 10
-50 [3] M. H. Fernandes, I. Garitaonandia, J. Albizuri, J. M. Hernández, and D.
1.5
Uncontrolled
-60
Uncontrolled Barrenetxea, “Simulation of an active vibration control system in a
PSO-SF controlled PSO-SF controlled
-70 centerless grinding machine using a reduced updated FE model,”
1
International Journal of Machine Tools and Manufacture, vol. 49, 2009,

Magnitude (dB)
Deflection (m)

-80
0.5
-90
pp. 239-245.
0
-100 [4] A. Rama Mohan Rao and K. Sivasubramanian, “Optimal placement of
-0.5
-110 actuators for active vibration control of seismic excited tall buildings
-1 -120 using a multiple start guided neighbourhood search (MSGNS)
-1.5 -130 algorithm,” Journal of Sound and Vibration, vol. 311, 2008, pp. 133-
0 1 2 3 4 0 50 100 150
Time (sec) Frequency (rad/sec) 159.
[5] Y. H. Guan, T. C. Lim, and W. Steve Shepard, “Experimental study on
(a) System response in time domain (b) System response in frequency domain
active vibration control of a gearbox system,” Journal of Sound and
Vibration, vol. 282, 2005, pp. 713-733.
-7 -7
x 10 x 10 [6] M. O. Tokhi and M. A. Hossain, “Self-tuning active vibration control of
15 15
flexible beam structures,” Research Report no. 531, 1994, Department of
Automatic Control and Systems Engineering, The University of
Deflection (m)
Deflection (m)

10 10

5 5 Sheffield, Sheffield, UK.

0 0 [7] M. O. Tokhi and R. R. Leitch, “Design and implementation of self-
-5 -5 tuning active noise control systems,” IEE Proceedings-D, vol. 138, issue
20 20
20 20 5, 1991, pp. 421–430.
10 15 10 15
5
10
5
10 [8] J. Kennedy, and R. C. Eberhart, “Particle swarm optimization”,
y (j) y (j)
x (i) x (i) Proceedings of the IEEE Conference on Neural Networks IV,
Piscataway, NJ, 1995, pp. 1942-1948.
(c) System response before cancellation (d) System response after cancellation
[9] R. C. Eberhart, and Y. Shi, “Tracking and optimizing dynamic systems
Figure 8. Performance of PSO-SF AVC using random disturbance. with particle swarms”, Proceedings of the Congress on Evolutionary
Computation, Seoul, South Korea, 2001, pp. 94-100.
[10] L. Liu, L. Wenxin, and C. David A. “Particle swarm optimization-based
parameter identification applied to permanent magnet synchronous
-5
motors,” Engineering Applications of Artificial Intelligence, vol. 21 (7),
x 10
3 -20 2008, pp. 1092-1100.
Uncontrolled Uncontrolled
2
PSO-SF controlled -40 PSO-SF controlled [11] M. Ye, X. Wang, and Y. Xu, “Parameter identification for proton
exchange membrane fuel cell model using particle swarm optimization,”
Magnitude (dB)

-60
Deflection (m)

1 International Journal of Hydrogen Energy, vol. 34 (2), 2009, pp. 981-

-80 989.
0
-100 [12] N. M. Kwok, Q. P. Ha, T. H. Nguyen, J. Li, and B. Samali, ”A novel
-1
-120
hysteretic model for magnetorheological fluid dampers and parameter
identification using particle swarm optimization,” Sensors and Actuators
-2
0 1 2 3 4
-140
0 50 100 150 A: Physical, vol. 132 (2), 2006, pp. 441-451.
Time (sec) Frequency (rad/sec)
[13] I. Z. Mat Darus and M. O. Tokhi, “Soft computing-based active
(a) System response in time domain (b) System response in frequency domain vibration control of a flexible structure,” Engineering Applications of
Artificial Intelligence, vol 18, 2005, pp. 93-114.
x 10
-5
x 10
-5 [14] M. H. Shaheed and M. O. Tokhi, “Dynamic modelling of a single-link
2 2 flexible manipulator: parametric and non-parametric approaches,”
Robotica, vol. 20, 2002, pp. 93-109.
Deflection (m)

Deflection (m)

1.5 1.5

1 1 [15] I. Z. Mat Darus and M. O. Tokhi, “Modelling of a flexible plate structure

0.5 0.5
using finite difference method”, Proceedings of the 2nd World
Engineering Conference, Malaysia, July 2002, pp. 483-487.
0 0
20
20
20
20
[16] S. Timoshenko and S. Woinowsky-Krieger, “Theory of plates and
10
10
15 10
10
15 shells”, McGraw-Hill Book Company, New York, 1959.
5 5
y (j) x (i) y (j) x (i) [17] A. W. Leissa, “Vibration of plates”, Scientific and Technical
Information Division Office of NASA, NASA SP-160, Washington, D.C,
(c) System response before cancellation (d) System response after cancellation 1969.
[18] L. Ljung and T. Glad, Modelling of Dynamic Systems. New Jersey:
Figure 9. Performance of PSO-SF AVC using finite duration step
Prentice-Hall, 1994.
disturbance.
[19] J. Kennedy and R. C. Eberhart, “Particle swarm optimization,”
Proceedings IEEE Conference on Neural Networks IV, Piscataway, NJ,
1995, pp. 1942-1948.
REFERENCES
[20] A. Ratnaweera, S. Halgamuge and H. Watson, “Self-organizing
[1] A. Rashid and C. Mihai Nicolescu, “Active vibration control in hierarchical particle swarm optimizer with time varying acceleration
palletised workholding system for milling,” International Journal of coefficients,” IEEE Transactions on Evolutionary Computation, vol. 8,
Machine Tools and Manufacture, vol. 46, 2006, pp. 1626-1636. issue 3, 2004, pp. 240-255.
[2] M. Moshrefi-Torbati, A. J. Keane, S. J. Elliott, M. J. Brennan, D. K. [21] I. Abd Latif and M. O. Tokhi, “Fast convergence strategy for particle
Anthony, and E. Rogers, “Active vibration control (AVC) of a satellite swarm optimization using spreading factor,” To appear.
boom structure using optimally positioned stacked piezoelectric
[22] M. O. Tokhi and R. R. Leitch, “Active noise control,” Clarendon Press,
actuators,” Journal of Sound and Vibration, vol. 292, 2006, pp. 203-220.
Oxford, 1992.

Authorized licensed use limited to: UNIVERSITI TEKNOLOGI MARA. Downloaded on November 14,2023 at 04:55:32 UTC from IEEE Xplore. Restrictions apply.