Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Article

Journal of Vibration and Control


1–14

Intelligent fuzzy logic with firefly ! The Author(s) 2015


Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
algorithm and particle swarm DOI: 10.1177/1077546315580693
jvc.sagepub.com
optimization for semi-active suspension
system using magneto-rheological
damper

Mat Hussin Ab Talib and Intan Zaurah Mat Darus

Abstract
This paper presents a new approach for intelligent fuzzy logic (IFL) controller tuning via firefly algorithm (FA) and particle
swarm optimization (PSO) for a semi-active (SA) suspension system using a magneto-rheological (MR) damper. The SA
suspension system’s mathematical model is established based on quarter vehicles. The MR damper is used to change a
conventional damper system to an intelligent damper. It contains a magnetic polarizable particle suspended in a liquid
form. The Bouc–Wen model of a MR damper is used to determine the required damping force based on force–
displacement and force–velocity characteristics. The performance of the IFL controller optimized by FA and PSO is
investigated for control of a MR damper system. The gain scaling of the IFL controller is optimized using FA and PSO
techniques in order to achieve the lowest mean square error (MSE) of the system response. The performance of the
proposed controllers is then compared with an uncontrolled system in terms of body displacement, body acceleration,
suspension deflection, and tire deflection. Two bump disturbance signals and sinusoidal signals are implemented into the
system. The simulation results demonstrate that the PSO-tuned IFL exhibits an improvement in ride comfort and has the
smallest MSE for acceleration analysis. In addition, the FA-tuned IFL has been proven better than IFL–PSO and uncon-
trolled systems for both road profile conditions in terms of displacement analysis.

Keywords
Firefly algorithm, fuzzy logic controller, magneto-rheological damper, particle swarm optimization, semi-active suspension
system

sustain any disturbance imposed upon the vehicle


1. Introduction body (Sharp and Hassan, 1986).
Vibration, load disturbance and road disturbance have Semi-active suspension is currently an attractive area
recently become important vehicle system issues. This is of research. This is because the system can be an active
because good comfort can be provided to drivers and device without requiring a large power source. In add-
passengers especially in terms of ride comfort and road ition, it can offer the reliability of passive devices
handling. Three types of suspension system namely pas- (Tsang et al., 2006). Unlike passive systems, the
sive, semi-active and active have their own abilities to
overcome disturbances with different techniques and Department of System Dynamics and Control, Universiti Teknologi
algorithms (Alexandru and Alexandru, 2011). Passive Malaysia, Johor Bahru, Malaysia
system is a conventional system consisting of a spring Received: 3 November 2014; accepted: 5 March 2015
and a fixed damper unit. This system performed poorly
in term of vehicle stability in recent findings as com- Corresponding author:
Intan Zaurah Mat Darus, Department of System Dynamics and Control,
pared with semi-active and active suspension system. Faculty of Mechanical Engineering, Universiti Teknologi Malaysia, Johor
The damper and fixed spring components of passive Bahru, Johor, Malaysia.
system do not have enough energy absorption to Email: intan@fkm.utm.my

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


2 Journal of Vibration and Control

variable damper of a semi-active system can actually be FA and PSO. Section 4 contains the analysis and dis-
effectively controlled in terms of damper stiffness based cussion and Section 5 provides the conclusion.
on the required values in a particular situation. An
intelligent magneto-rheological (MR) damper is used
2. Semi-active quarter car model
because of several advantages, such as fast time
response, low power requirement, high dynamic A semi-active system has a variable damper called a
range, and mechanical simplicity (Koo et al., 2004). MR damper system, which differs from a conventional
There are four major areas of MR damper technology passive system and a fully active system (Rashid et al.,
research, including mathematical and numerical 2011). Figure 1 can be represented as a quarter car
research (Yao et al., 2009; Ye et al., 2010), research model based on the mathematical equation of a semi-
on fluids (Yang et al., 2002), research on design and active suspension system. Using Newton’s second law,
development (Fujitani et al., 2003), and research on the mathematical equation can be described as follows:
MR damper control strategies (Caponetto et al., 2003;
Yao et al., 2012; Khiavi et al., 2013). However, ms x€ u þ Fd  kd ðxu  xs Þ ¼ 0 ð1Þ
from among those categories, MR damper control stra-
tegies will be investigated in this research. Currently, mu x€ u  Fd þks ðxu  xs Þ  kt ðxr  xu Þ ¼ 0 ð2Þ
many researchers have paid significant attention to
intelligent controllers such as neural networks (Guo where ms is sprung mass, mu is unsprung mass, xr is a
et al., 2004), PID neural networks (Liu et al., 2010), road profile, xu is unsprung mass displacement, xs is
fuzzy logic (Dong and Yu, 2013), and fuzzy PID sprung mass displacement, ks is spring stiffness, and
(Golob, 2001). Fd is damper force. The parameter model of the semi-
In this research, a mathematical model known as the active system is shown in Table 1. To implement the
Bouc–Wen model (Kwok et al., 2007) is used to deter- damper into a variable condition, the Bouc–Wen model
mine the actual damping force through the system. The is used to develop semi-active control for a MR
parameters of the Bouc–Wen model are very important damper. Basically, the characteristics of the MR
when determining the optimum value of the required
damping force for the system. The system identification
method can also be used to represent a MR damper
(Sarp Arsava et al., 2015). Xia (2003) used an optimal
neural network and system identification to predict the
MR damper model. The model was constructed using a
multi-layer perception optimal neural network and
system identification. This included a Gauss–Newton-
based Lavenberg–Marquardt training algorithm, an
optimal brain surgeon strategy, and an autoregressive
with exogenous variables (ARX) model. To control a
semi-active system, an intelligent fuzzy logic (IFL) con-
troller firefly algorithm (FA) and particle swarm opti-
mization (PSO) have been introduced in this research.
An IFL controller is one of the methods which have
been widely used by many researchers in different appli-
cations, such as a plate (Shirazi et al., 2011), an engine
turbine (Bazazzadeh et al., 2011), and a robot arm
(Mailah and Rahim, 2000). To implement IFL into
the system, the IFL gain factor must be properly Figure 1. Semi-active suspension model.
tuned. Thus, to optimize the best parameter values of
the gain factor, a FA and PSO optimization method
has been investigated in this research. The main object- Table 1. Parameter of suspension system.
ives of this paper are to develop an IFL controller Parameter Value
tuned by FA and PSO and to investigate the compara-
tive assessment of IFL–FA, IFL–PSO, and uncon- ms ¼ sprung mass 325 kg
trolled systems. This paper will be organized as mt ¼ unsprung mass 55 kg
follows. Section 2 presents semi-active quarter car ks ¼ spring stiffness 45.0 N/mm
model. In Section 3, the modeling and control suspen- kt ¼ tire stiffness 150.0 N/mm
sion system will be developed including fuzzy rules with

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


Ab Talib and Mat Darus 3

damper must be investigated in order to validate the Based on equation (9),  is a filter time constant and
performance of the damper to the system.  is a voltage input of the first filter. All model param-
Such characteristics using a proposed model were eters, including kD0, kD1, c0a, c0b, c1a, c1b, a , b , x0, ,
first introduced by Bouc (1967) and later modified by , n, and A, need to be defined in order to identify the
Wen (1976). This type of model has successfully proved MR damper system. The parameter values, which are
that hysteresis behavior of MR dampers has emerged as taken from previous work (Rashid et al., 2007), are
a potential technology to implement semi-active con- shown in Table 2. The parameters for the model were
trol in structures. The Bouc–Wen model is shown in chosen based on experimental system identification.
Figure 2. Due to this investigation, the predicted responses and
The damper force based on this model can be pre- the corresponding experimental data were compared
dicted using the following equations: and the parameters are taken when the behavior of
the damper is good in all regions, including a small
Fd ¼ cD1 y_ þ kD1 ðxD  x0 Þ ð3Þ

1
y_ ¼ ½z þ c0 x_ þ kD0 ðxD  yD Þ ð4Þ
c0 þ cD1 Table 2. Parameter values for MR damper RD-1005.

z_ ¼  xD
_  yDjzjzj
_ n1
_ jzjn þAðxD
_  yDÞ
ðxD _  yDÞ
_ Parameter Value

ð5Þ a 12.44 N/mm


b 38.43 N/V mm
where yD is internal displacement, xD is a damper dis- c0a 0.78 N s/mm
placement, x0 is the initial condition of damper deflec-
c0b 1.80 N s/V mm
tion, and z is the hysteretic restoring force. The voltage
c1a 14.64 N s/mm
applied depends on current driver from parameter
model as follows: c1b 34.62 N s/V mm
x0 0
 ¼ a þ b u ð6Þ kD0 37.81 N/mm
kD1 0.62 N/mm
c0 ¼ c0a þ c0b u ð7Þ A 2.68 m1
 0.647 m1
c1 ¼ c1a þ c1b u ð8Þ  0.647 m1
 90
where u represents output of the first order filter given n 10
as follows:

u_ ¼ ðu  Þ ð9Þ

Figure 2. Bouc–Wen model of a MR damper. Figure 3. Force–displacement characteristic.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


4 Journal of Vibration and Control

error between predicted and measured (experimental) input and output of MR damper are voltage and
force. damper force, respectively. The parameters obtained
Equations (3)–(9) are used to simulate the model previously can be proved to validate the model via
using a Matlab SIMULINK environment. The model force–displacement and force–velocity graphs, as
shown in Figures 3 and 4, respectively. It can be
observed that as the voltage increases, the correspond-
ing damping force increases as well.

3. Control design and modeling


3.1. Intelligent fuzzy logic controller
Semi-active vehicle suspension with a controller
system is a nonlinear structure with a parameter
system that is always changing as the vehicle’s ride
and handling changes (Wang et al., 2011). Thus, an
IFL controller is used to reduce unwanted vehicle
motion for disturbance rejection control. IFL consists
of four important sections, namely fuzzification, a
fuzzy inference engine, a fuzzy rule base, and defuzzi-
fication. Two inputs from the system, namely suspen-
sion displacement and velocity, were trained by the
fuzzy logic system to produce the targeted voltage as
Figure 4. Force–velocity characteristic. an output variable. In this controller algorithm, a scal-
ing factor is considered as an important parameter for
each variable input and output. GE, GV, and GU are
Table 3. Fuzzy rules for computing desired voltage output. the scaling factors for displacement input, velocity
input, and variable voltage output, respectively. FA
Velocity and PSO strategies are used to determine the best
Displacement N Z P value of the IFL scaling factors based on the mean
N M S L square error (MSE) of the system. A triangular and
Z L S L Gaussian membership function is used in the control-
ler design and three regions for input variable, namely
P L S M
negative (N), zero (Z), and positive (P), are used as

Figure 5. Membership function for displacement input.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


Ab Talib and Mat Darus 5

linguistic variables. For the output linguistic, three applications. The FA was basically developed by
output variables are set as small (S), medium (M), Yang (2007) based on the idealized behavior of the
and large (L). A fuzzy controller rule for computing flashing of fireflies in nature The main concept of firefly
the desired voltage output and membership functions is that each species of firefly produces its own pattern of
of input and output variables are shown in Table 3 flashes and the main purpose of their flashing is to
and Figures 5, 6, and 7, respectively. attract other fireflies as well as to send information
between fireflies (Fister et al., 2013). There are three
main idealized rules of firefly behavior:
3.2. Firefly algorithm
The firefly algorithm (FA) has become an important . Fireflies are unisex so that one firefly will be
method that has been widely applied in many attracted to other fireflies regardless of their sex.

Figure 6. Membership function for velocity input.

Figure 7. Membership function for voltage output.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


6 Journal of Vibration and Control

. The attractiveness is proportional to the brightness, being the randomization parameter, "i is a vector of
and they both decrease as their distance increases. random numbers drawn from Gaussian distribution
The less bright firefly will move towards the brighter or uniform distribution. For most implementations,
firefly. If there is no particularly bright firefly, it will 0 can be taken as 1 and a e [0, 1]. Based on FA
move randomly. rules, the steps of the FA can be summarized as the
. The brightness of a firefly is determined by the value pseudo code as in Hashmi et al. (2013) and the param-
of the objective function (Pal et al., 2012). eter of fireflies used in this research are shown in Table 4.

In the simplest way, light intensity I(r) varies with a


distance r, i.e.
3.3. Particle swarm optimization
Particle swarm optimization (PSO) was introduced by
I ¼ I0 er ð10Þ Eberhart and Kennedy (1995) inspired by the behavior
of bird flocking and fish flock movement behavior.
where I0 is the original light intensity and  is the light With this inspiration, swarm represents a number of
absorption coefficient. The attractiveness of firefly algo- potential solutions to the problem and the individual
rithm is proportional to the light intensity and the
equation of attractiveness  can be defined as:

m Table 5. Parameters of PSO algorithm.


ðrÞ ¼ 0 er , ðm  1Þ ð11Þ
Parameter name Value
where r is the distance between each two fireflies and 0 No. of iterations 40
is their attractiveness at r ¼ 0. The light absorption Inertia weight, w 1
coefficient  is actually determined by the variation Correction factor, c1 & c2 2
of attractiveness when the number of firefly commu- Swarm size 30
nities is increased. The distance between a couple of
fireflies i and j at xi and xj, respectively, is the
Cartesian distance:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u d Start
uX
rij ¼ jjxi  xj jj ¼ t ðxi , k  xj , kÞ2 , ð12Þ
k¼1
Initialization of algorithmic parameter
The movement of a firefly i when attracted to and swarm position
another more attractive (brighter) firefly j can be
defined as:
Calculate fitness of each particle
rij2
xi ¼ xi þ 0 e ðxj  xi Þ þ "i ð13Þ
Update pbest and gbest
where the second term is due to the attraction and the
third term can be defined as a randomization with a
Update particle velocity and position

Table 4. Parameters of firefly algorithm. no


Termination based
Parameter name Value on stopping criteria

Firefly size 30
Max generation, k 40
Alpha,  0.25 yes
Beta,  0.8
Gamma,  1 End
Lower boundary 0
Upper boundary 8
Figure 8. Steps in the PSO algorithm.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


Ab Talib and Mat Darus 7

in PSO algorithm is called a particle which means each particle respectively. With a given fitness function, the
individual in search space can be adjusted dynamically particle position can be found with the best evaluation.
based on the movement of position and velocity. To evaluate the best particle position, two ‘‘best’’
Position and velocity of particle represents the candi- values, namely pbest and gbest, are updated. Thus,
date solution to the problem and flying direction of the based on these two ‘‘best’’ values, the position of a

Figure 9. Block diagram of IFL with FA and PSO tuning.

Figure 10. Bumps road profile.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


8 Journal of Vibration and Control

Figure 11. Sinusoidal road profile.

Table 6. Parameter tuning results. A block diagram of the semi-active system with IFL
control based on the FA tuning algorithm and PSO
Parameter name Values
algorithm is shown in Figure 9.
Firefly algorithm
GE 1.4595
4. Simulation results and analysis
GV 0.3394
GU 6.7199 As shown in Figures 10 and 11, the input from two
Particle swarm optimization different disturbances, namely bumps input and sinus-
GE 0.8297 oidal input, were introduced to excite the system. The
GV 0.4749 amplitudes for two bump disturbances are 0.03 m and
GU 3.9865 0.01 m, respectively, whereas for the sinusoidal wave,
the frequency is 0.5 Hz with maximum amplitude
0.0125 m. A simulation study for tuning the IFL scaling
factor was performed using the Matlab SIMULINK
particle can be adjusted by changing its velocity dynam- environment. The parameter tuning results for the pro-
ically toward the global optimum. The basic formula posed controller using FA and PSO algorithms are
for updated velocity and updated position for each par- shown in Table 6.
ticle can be calculated, as shown in equations (14) and
(15), respectively:
4.1. Bumps road profile
  Based on Figures 12 and 13, it is clearly observed that
vi kþ1 ¼ w vi k þc1 ðrand1 Þðpbesti  xi Þ
ð14Þ IFL with FA and PSO techniques manages to reduce
þ c2 ðrand2 Þðgbesti  xi Þ the amplitude oscillation better than an uncontrolled
system. It can be seen that the semi-active suspension
system with IFL–FA and IFL–PSO controllers could
xi kþ1 ¼ xi k þvi kþ1 ð15Þ improve the vehicle’s ride performance up to 85.4%.
For comparison between IFL–PSO and IFL–FA con-
where w is an inertia weight parameter, c1, c2 are weight trolled MR damper, the IFL–FA semi-active suspen-
factors, vik is the velocity of particle i in the kth iter- sion system made a slight improvement in terms of
ation, xik is the position of particle i in the kth iteration, body vertical displacement, whereas for body acceler-
and, rand1, rand2 are random numbers between 0 and 1. ation, the IFL–PSO semi-active suspension system
The PSO algorithm parameters and the main procedure made a better improvement than IFL–FA. A vehicle’s
are shown in Table 5 and Figure 8, respectively. body acceleration actually determines the ride comfort

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


Ab Talib and Mat Darus 9

of passengers and drivers. Thus, the IFL–PSO control- Figure 14 and 15 show the analysis of suspension
ler has drawn significant attention compared to an deflection and tire deflection, respectively. Based on
IFL–FA controller. The PSO technique has a main tire deflection analysis, it is clearly proved that the ver-
advantage for finding the best parameter values of an tical force from vehicle body has been transmitted to
IFL controller. For example, the swarm particle is not the tire system in order to ensure that body displace-
easily trapped into local optima. Thus, it can be moved ment and acceleration are always minimized. In terms
in any direction and is able to achieve the best position of overall performance, it can be mentioned that both
based on a given fitness function. However, both of the IFL–FA and IFL–PSO have their own capability to
proposed controllers are able to come out with robust improve the performance of a semi-active system
controllers due to vehicle body improvement (body ver- using a MR damper. Without controller response, the
tical displacement and body acceleration). comfort and stability of the passengers and driver

Figure 12. Body displacement response.

Figure 13. Body acceleration response.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


10 Journal of Vibration and Control

might not achieve the targeted level due to the proposed improvement in terms of body vertical displacement
situation. and body vertical acceleration, respectively. As men-
tioned previously, this is due to the fact that the par-
ameter results during tuning optimization are affected
4.2. Sinusoidal road profile
by the benefits of both techniques. For instance, PSO
A sinusoidal road profile is another disturbance input and FA techniques have their own advantages, which
imposed on the system. The results of vertical body are good in terms of global search performance and
displacement and body acceleration are shown in very effective when solving high nonlinear problems,
Figures 16 and 17, respectively. It is clearly observed respectively. However, the capability of the PSO tech-
that the proposed controllers constantly managed to nique did make an advantage for body vehicle due to
control the disturbance road profile. For comparison the better performance of body acceleration analysis as
between IFL–PSO and IFL–FA controlled MR dam- compared with the FA technique. Analysis of suspen-
pers, IFL–FA and IFL–PSO did make a slight sion deflection and tire deflection response are shown in

Figure 14. Suspension deflection response.

Figure 15. Tire deflection response.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


Ab Talib and Mat Darus 11

Figure 16. Body displacement response.

Figure 17. Body acceleration response.

Figures 18 and 19, respectively. It can be observed that An overview of the results about MSE and percent-
both of the proposed controllers are able to control the age improvement in terms of body displacement and
body deflection as compared with an uncontrolled body acceleration for the proposed controller strategy
system. are stated in Tables 7 and 8, respectively.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


12 Journal of Vibration and Control

Figure 18. Suspension deflection response.

Figure 19. Tire deflection response.

system. Based on the simulation results, the IFL–FA


system shows a better improvement than the IFL–PSO
5. Conclusion
and uncontrolled systems for body displacement ana-
The performance of an IFL controller with FA and lysis with 75.9% and 9.53% for the bumps road profile
PSO tuning for has been considered for a semi-active and sinusoidal road profile, respectively. However, for

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


Ab Talib and Mat Darus 13

Table 7. MSE and improvement for body displacement. semiactive suspension. IEEE Transactions on Control
Systems Technology 11(6): 786–798.
MSE % Reduction Dong X and Yu M (2013) Genetic algorithm based fuzzy
Bumps logic control for a magneto-rheological suspension.
Uncontrolled 3.424  104 Journal of Vibration and Control 20(9): 1343–1355.
Fister I, Yang X-S and Brest J (2013) A comprehensive review
IFL–PSO 8.9495  105 73.8
of firefly algorithms. Swarm and Evolutionary Computation
IFL–FA 8.2324  105 75.9 13: 34–46.
Sinusoidal Fujitani H, Tomurac T, Hiwatashid T, et al. (2003)
Uncontrolled 9.565  105 Development of 400 kN magnetorheological damper for
IFL–PSO 9.1904  105 3.91 a real base-isolated building. Smart Structures and
IFL–FA 8.6528  105 9.53 Materials 5052: 265–276.
Golob M (2001) Decomposed fuzzy proportional–integral–
derivative controllers. Applied Soft Computing 1(3):
201–214.
Table 8. MSE and improvement for body acceleration. Guo DL, Hu HY and Yi JQ (2004) Neural network control
for a semi-active vehicle suspension with a magnetorheo-
MSE % Reduction logical damper. Journal of Vibration and Control 10:
Bumps 461–471.
Uncontrolled 4.2733 Hashmi A, Goel N, Goel S, et al. (2013) Firefly Algorithm for
unconstrained optimization. IOSR Journal of Computer
IFL–PSO 0.6239 85.4
Engineering (IOSR-JCE) 11(1): 75–78.
IFL–FA 0.8821 79.3 Kennedy J and Eberhart R (1995) Particle swarm optimiza-
Sinusoidal tion. Proceedings of ICNN’95 – International Conference
Uncontrolled 0.0189 on Neural Networks. vol. 4, November/December, Perth,
IFL–PSO 0.0139 20.8 Asutralia, pp. 1942–1948.
IFL–FA 0.0149 20.4 Khiavi AM, Mirzaei M and Hajimohammadi S (2013) A new
optimal control law for the semi-active suspension system
considering the nonlinear magneto-rheological damper
body acceleration analysis, the IFL–PSO system has model. Journal of Vibration and Control 20(14):
been proven better than the IFL–FA and uncontrolled 2221–2233.
systems with 85.4% and 20.8% reduction for bumps Kwok NM, Ha QP, Nguyen MT, et al. (2007) Bouc Wen
road profile and sinusoidal road profile, respectively. Model parameter identification for a MR damper fluid
damper using computationally efficient GA. ISA
Transactions 46: 167–179.
Acknowledgements Koo J-H, Goncalves FD and Ahmadian M (2004)
The authors would like to express their gratitude to Minister Investigation of the response time of magnetorheological
of Education Malaysia (MOE) and Universiti Teknologi fluid dampers. Smart Structures and Materials 5386:
Malaysia (UTM) for funding and providing facilities to con- 63–71.
duct this research. Liu W, Shi W, Liu D, et al. (2010) Experimental modeling of
magneto-rheological damper and PID neural network
Funding controller design. 2010 Sixth International Conference on
Natural Computation. vol. 4, 10–20 August, Yantai,
This work was supported by FRGS (grant number 4F395)
Shandong, pp. 1674–1678.
and UTM Research University (grant number 05H71).
Mailah M and Rahim NIa (2000) Intelligent active force con-
trol of a robot arm using fuzzy logic. 2000 TENCON
References Proceedings, Intelligent Systems and Technologies for the
Alexandru C and Alexandru P (2011) A comparative analysis New Millennium. vol. 2, 24–27 September, Kuala Lumpur,
between the vehicles’ passive and active suspensions. Malaysia, pp. 291–296.
International Journal of Mechanics 4(5): 371–378. Pal SK, Rai C and Singh AP (2012) Comparative study of
Bazazzadeh M, Badihi H and Shahriari A (2011) Gas turbine firefly algorithm and particle swarm optimization for
engine control design using fuzzy logic and neural net- noisy non-linear optimization problems. International
works. International Journal of Aerospace Engineering Journal of Intelligent Systems and Applications 4(10):
2011: 1–12. 50–57.
Bouc R (1967) Forced vibration of mechanical systems with Rashid MM, Rahim NA, Member S, et al. (2007)
hysteresis. Proceedings of the Fourth Conference on Development of a semi-active car suspension control
Nonlinear Oscillation. Prague, Czechoslovakia, p. 315. system using magnetor-heological damper. International
Caponetto R, Diamante O, Fargione G, et al. (2003) A soft Journal of Mechanical and Material Engineering
computing approach to fuzzy sky-hook control of (IJMME) 2(2): 93–108.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015


14 Journal of Vibration and Control

Rashid MM, Rahim NA, Member S, et al. (2011) Analysis Wen YK (1976) Method for random vibration of hysteretic
and experimental study of magnetorheological-based systems. Journal of Engineering Mechanics (American
damper for semiactive suspension system using fuzzy Society of Civil Engineers) 102(2): 249–263.
hybrids. IEEE Transactions on Industry Applications Xia P-Q (2003) An inverse model of MR damper using opti-
47(2): 1051–1059. mal neural network and system identification. Journal of
Sarp Arsava K, Nam Y and Kim Y (2015) Nonlinear system Sound and Vibration 266(5): 1009–1023.
identification of smart reinforced concrete structures Yang G, Spencer BF, Carlson JD, et al. (2002) Large-scale
under impact loads. Journal of Vibration and Control. MR fluid dampers: modeling and dynamic performance
First published on 14 January 2015. DOI: considerations. Engineering Structures 24(3): 309–323.
1077546314563966. Yang X (2007) Firefly algorithms for multimodal optimiza-
Sharp RS and Hassan SA (1986) The relative performance tion. Lecture Notes in Computer Sciences 5792: 169–178.
capabilities of passive, active and semi-active car suspen- Yao J-L, Shi W-K, Zheng J-Q, et al. (2012) Development of a
sion systems. Proceedings of the Institution of Mechanical sliding mode controller for semi-active vehicle suspen-
Engineers, Part D: Journal of Automobile Engineering sions. Journal of Vibration and Control 19(8): 1152–1160.
200(3): 219–228. Yao K, Zhao X and Hou Z (2009) Damping characteristics
Shirazi AHN, Owji HR and Rafeeyan M (2011) Active vibra- modeling and dimulation of MR damper. 2009
tion control of an FGM rectangular plate using fuzzy logic International Conference on Information Management,
controllers. Procedia Engineering 14: 3019–3026. Innovation Management and Industrial Engineering. vol.
Tsang HH, Su RKL and Chandler AM (2006) Simplified 2, 26–27 December, Xián, China, pp. 492–495.
inverse dynamics models for MR fluid dampers. Ye MY, Jiang H, Xu YS, et al. (2010) Bouc–Wen hysteresis
Engineering Structures 28(3): 327–341. model parameter identification by means of hybrid intel-
Wang Y, Utsunomiya K and Bortoff AS (2011) Nonlinear ligent technique. Advanced Materials Research 108–111:
control design for a semi-active vibration reduction 1397–1402.
system. Control Conference (CCC), 2011 30th Chinese.
22–24 July, Yantai, China, pp. 5833–5837.

Downloaded from jvc.sagepub.com at East Tennessee State University on July 2, 2015

You might also like