You are on page 1of 14

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2004; 33:723–736 (DOI: 10.1002/eqe.372)

Semi-active fuzzy control for seismic response reduction


using magnetorheological dampers

Kang-Min Choi1; ∗; † , Sang-Won Cho1 , Hyung-Jo Jung2 and In-Won Lee1


1 Department of Civil and Environmental Engineering; Korea Advanced Institute of Science and Technology;
373-1 Guseong-dong; Yuseong-gu; Daejeon 305-701; Korea
2 Department of Civil and Environmental Engineering; Sejong University; 98 Gunja-dong; Gwangjin-gu;

Seoul 143-747; Korea

SUMMARY
A semi-active fuzzy control strategy for seismic response reduction using a magnetorheological (MR)
damper is presented. When a control method based on fuzzy set theory for a structure with a MR
damper is used for vibration reduction of a structure, it has an inherent robustness, and easiness to treat
the uncertainties of input data from the ground motion and structural vibration sensors, and the ability
to handle the non-linear behavior of the structure because there is no longer the need for an exact
mathematical model of the structure. For a clipped-optimal control algorithm, the command voltage of
a MR damper is set at either zero or the maximum level. However, a semi-active fuzzy control system
has benet to produce the required voltage to be input to the damper so that a desirable damper force
can be produced and thus decrease the control force to reduce the structural response. Moreover, the
proposed control strategy is fail-safe in that the bounded-input, bounded-output stability of the controlled
structure is guaranteed. The results of the numerical simulations show that the proposed semi-active
control system consisting of a fuzzy controller and a MR damper can be benecial in reducing seismic
responses of structures. Copyright ? 2004 John Wiley & Sons, Ltd.

KEY WORDS: fuzzy control; semi-active control; MR damper; response reduction; vibration control

1. INTRODUCTION

In the eld of civil engineering, many control algorithms and devices have been proposed
over the last few decades for the purpose of protecting structures and their contents from the
damaging eects of environmental hazards such as earthquakes and wind loadings. Depending
on the control methods, vibration control in the structure can be divided into three categories;

∗ Correspondence to: Kang-Min Choi, Department of Civil and Environmental Engineering, Korea Advanced Institute
of Science and Technology, 373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Korea.
† E-mail:vision222@kaist.ac.kr

Contract=sponsor: National Research Laboratory (NRL) program, Ministry of Science and Technology, Korea;
contract=grant number: 2000-N-NL-01-C-251
Received 25 September 2002
Revised 29 August 2003
Copyright ? 2004 John Wiley & Sons, Ltd. Accepted 7 October 2003
724 K.-M. CHOI ET AL.

namely, passive control, active control and semi-active control or hybrid control (a combina-
tion of the previous types). The idea of passive control is to dissipate the vibratory energy in
the structural system. However, this kind of control system provides no extra assistance, so
it cannot adapt to varying loading conditions. Consequently its eect is limited. On the other
hand, the active control system reduces the structural response by using external energy sup-
plied by actuators to impart forces on the structures, generally depending on a sizeable power
supply. It is considerably more exible to reduce the structural responses for a wide variety
of loading conditions. However, there are some problems such as stability, cost eectiveness,
reliability, power requirements etc. Semi-active control has received a lot of attention recently
because it oers great adaptability without a large power requirement. Previous research has
demonstrated the eectiveness of semi-active systems in structural vibration control using vis-
cous uid dampers. The magnetorheological (MR) damper is one of the semi-active devices
that can provide reliable vibration control. Besides its minute power requirement, the MR
damper is reliable, fail safe, and is expected to be relatively inexpensive.
Classical control algorithms need an exact mathematical model for an actual structure and,
especially, an active control algorithm uses complicated mathematical equations to reduce
structure responses. Since buildings in civil engineering are getting much higher and bridges
are getting much longer than those structures are very complex multi-degree of freedom
systems, and it is very dicult to nd an exact mathematical model to describe the behavior
of the structures. Active vibration control of structural systems using fuzzy set theory has
been widely investigated in past years [1–7]. The fuzzy set theory was introduced by Zadeh
[1] in 1965. In 1974 Mamdani [2], by applying Zadeh’s theories of linguistic approach and
fuzzy inference, successfully used the ‘IF–THEN’ rule on the automatic operating control of
a steam generator. In civil engineering, the fuzzy set theory was applied by Brown and Yao
[3], Juang and Elton [4], Faravelli et al. [5], Teng et al. [6] and Wang and Lee [7]. The
most advantageous and controversial property is that the fuzzy controller does not rely on the
analysis and synthesis of the mathematical model of the process. Therefore, the uncertainties of
input data from the external loads and structural response sensors are treated in a much easier
way by the fuzzy controller than by classical control theory. Moreover, for a more realistic
implementation, non-linearities incorporated can be overcome easily, and by incorporating
human expertise into the fuzzy IF–THEN rules they can be constructed to control complex
structural systems.
Dyke et al. [8] proposed a semi-active clipped-optimal control algorithm to reduce the
structural response with a MR damper. In their approach, the command voltage of the MR
damper was adjusted by a linear optimal controller combined with a force feedback loop.
The command signal was set at either zero or the maximum level, depending on how the
damper’s force compared with the target optimal control force. However, it is important
that command signals are calculated based on the desirable control forces and input into an
actuator so that the actuator can produce forces as close as possible to the desired ones (i.e.
inverse dynamics). This means that, as time passes, the command signal is varied to reduce
eectively the structural response. To do so, owing to the damper’s non-linear characteristics,
it is dicult to obtain directly the command signal.
The objective of this paper is to propose a new semi-active control method for seismic
reduction of structures with a MR damper. To do this, a fuzzy set theory based structural
control algorithm is employed as a control algorithm and a MR damper is introduced as a
supplemental damping device. A fuzzy algorithm is used to calculate the command signal that

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
SEMI-ACTIVE FUZZY CONTROL 725

varied as time passed, into the MR damper based on structural responses. Since a MR damper
is an energy-dissipation device that cannot add mechanical energy to the structural system, the
proposed control strategy is fail-safe, in that it guarantees the bounded-input, bounded-output
stability of the controlled structure. To demonstrate the eectiveness of the proposed method,
a three-story building structure is presented.

2. SEMI-ACTIVE CONTROL STRATEGY USING A MR DAMPER

2.1. Introduction to fuzzy control


Figure 1 shows an algorithm of fuzzy control inference. It consists of three basic parts; fuzzi-
cation where continuous input variables are transformed into linguistic variables, fuzzy rule
inference that handles rule inference consisting of fuzzy IF–THEN rules, and defuzzica-
tion that ensures exact and physically interpretable values for control variables. The design
of fuzzy control may include: the denition of input and output variables; the selection of
the data manipulation method; the membership function design; and the rule base design.
Using fuzzy rules and membership functions, fuzzy control converts linguistic variables into
numerical values required in most applications.
The design of the fuzzy controller selects the response quantities to be used as inputs to
the fuzzy controller and the distribution and type of membership functions to be used for the
selected input variables. Moreover, we must consider what control functions are needed, and
then dene them as output variables.
Then, the fuzzy inference rule is completely based on the selected input variables. Usually,
we use the form ‘If (a set of conditions to be satised) Then (a set of consequences to be
inferred)’ to describe the expert knowledge. For example, the multiple-input multiple-output
IF–THEN rules of the fuzzy control are shown in the form:

Rj : If x1 is A1j and xp is Apj then y1 is B1j and ym is Bmj (1)

where Rj denotes the j-th rule of the fuzzy inference rule, j = 1; 2; : : : ; q, x1 ; x2 ; : : : ; xp are the
inputs of the fuzzy controller, Aij is the linguistic value with respect to xi of rule j; y1 ; y2 ; : : : ; ym
are the outputs of the fuzzy controller and Bij is a fuzzy singleton function dened by experts.

Fuzzification

Input variables Fuzzy inference Output variables

Defuzzification

Figure 1. The algorithm of fuzzy control inference.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
726 K.-M. CHOI ET AL.

Figure 2. Simple mechanical model of the MR damper (Spencer et al. [9]).

Then, the inference conclusion obtained via fuzzication is defuzzied into a crisp output.
This paper adopts the center-of-gravity (COG) method among the defuzzication methods.
The COG method is dened as follows:
N  
j p
B
l=1 l 
i=1 Ai j (xi )

yj =    (2)
N p
l=1 i=1 Aj (xi )
i

where A is the membership function of Aji .

2.2. Phenomenological model of the MR damper


Spencer et al. [9] proposed a phenomenological model to portray the behavior of a prototype
MR damper. The simple mechanical idealization of the MR damper depicted in Figure 2 is
based on a Bouc–Wen hysteresis model and is governed by the following seven simultaneous
equations:
f = c1 ẏ + k1 (x − x0 ) (3)
1
ẏ = {z + c0 ẋ + k0 (x − y)} (4)
(c0 + c1 )
ż = −|ẋ − ẏ|z |z |n−1 − (ẋ − ẏ)|z |n + A(ẋ − ẏ) (5)
 = a + b u (6)
c1 = c1a + c1b u (7)
c0 = c0a + c0b u (8)
u̇ = −(u − v) (9)

Here, the accumulator stiness is represented by k1 , the viscous damping observed at large
and low velocities is represented by c0 and c1 , respectively; k0 is present to control the
stiness at large velocities; and x0 is the initial displacement of spring k1 associated with
the nominal damper force due to the accumulator; ;  and A are hysteresis parameters for the

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
SEMI-ACTIVE FUZZY CONTROL 727

Table I. Parameters for the prototype MR damper model.


Parameter Value

c0a (Ns=cm) 21.0


c0b (Ns=cm) 3.50
k0 (Ns=cm) 46.9
c1a (Ns=cm) 283
c1b (Ns=cm V) 2.95
k1 (N=cm) 5.00
 a (N=cm) 14.3
b (N=cm V) 140
(cm−2 ) 695
(cm−2 ) 363
A 363
n 2
(s−1 ) 190

..
xg
v fMR y
MR
Structure
Damper
.
x, x

fMR
Clipped Optimal
algorithm algorithm
fd
Control Law

Figure 3. Control diagram for the semi-active clipped-optimal control system (Dyke et al. [8]).

yield element;  is the evolutionary coecient. A total of 14 model parameters are obtained
to characterize the prototype MR damper using experimental data and a constrained non-linear
optimization algorithm. The resulting parameters are given in Table I.

2.3. Semi-active clipped-optimal control system design


Dyke et al. [8] have proposed a semi-active clipped-optimal control strategy based on ac-
celeration feedback for the MR damper. In their approach, a linear optimal controller was
designed to adjust the command voltage of the MR damper. The command signal was set at
either zero or the maximum level, depending on how the damper’s force compared with the
desired optimal control force. The control diagram for the semi-active clipped-optimal control
system is shown in Figure 3.
The algorithm for selecting the command signal is graphically represented in Figure 4 and
can be concisely stated as

fd ; fd · ẋdev ¡0
fMR = (10)
0; otherwise

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
728 K.-M. CHOI ET AL.

fd

v = Vmax
v=0

v=0
f

v=0 v=0

v = Vmax

Figure 4. Graphical representation of the algorithm (Dyke et al. [8]).

..
xg
v .
MR fMR x, x
Structure
Damper
.
x, x

Fuzzy Controller

Figure 5. Control diagram for the semi-active fuzzy control using a MR damper.

where fMR is the control force of the MR damper, and ẋdev is the velocity across the damper.
This equation means that the control forces that cannot be achieved by a MR damper are
clipped in the secondary bang–bang-type controller. Therefore, the command signal is set
at either zero or the maximum level, depending on how the damper’s force compared with
the target optimal control force. However, it is important that command signals are calcu-
lated based on the desired control forces and inputs put into an actuator so that the actuator
can produce forces as close as possible to the desired ones. This means that as time passes
the command signal is varied to eectively reduce the structural response. To do so, ow-
ing to the damper’s non-linear characteristics, it is dicult to obtain directly a command
signal.

2.4. Semi-active fuzzy control system using a MR damper


The strategy of the semi-active fuzzy control algorithm for seismic protection using
a MR damper is comprised of only one controller compared to the semi-active clipped-
optimal controller which has two controllers. The one semi-active fuzzy controller produces
directly the desired command voltage using fuzzy rule inference as the desired force de-
manded is varied so that the actuator can produce forces as close as possible to the de-
sired forces. The control diagram for the semi-active fuzzy control system is shown in
Figure 5.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
SEMI-ACTIVE FUZZY CONTROL 729

MR damper

..
xg

Figure 6. Three-story building structure employing a MR damper (Dyke et al. [8]).

This type of controller has two merits. One is to have all the advantages of the fuzzy
controller, and the other is to give directly the desired command voltage. This means that
it is possible to eectively control the dynamic responses of a real civil structure caused by
earthquakes and strong winds.

3. NUMERICAL EXAMPLES

To evaluate the performance of the proposed semi-active fuzzy control system using a MR
damper, a three-story shear building structure is studied. Simulated results of the proposed
control system are compared to those of an uncontrolled system, two passive systems (i.e.
passive-o and passive-on; for passive-o, the command voltage to the MR damper is held
at 0 V and for passive-on, the command voltage to the MR damper is held at the maximum
voltage level (2:25 V)); a semi-active clipped-optimal control system; and a semi-active fuzzy
control system.

3.1. Three-story shear building structure


A model of a three-story building congured with a single MR damper is considered and
shown in Figure 6. The MR damper is rigidly connected between the ground and the rst
oor of the structure. The equations of motion of the structure are given by

M x + C ẋ + K x = f − M  xg (11)

where f is the measured control force, and x = [x1 x2 x3 ]T is a vector of the displacements
of the three oors of the structure relative to the ground. The system matrices are listed in
Table II. This system is a simple model of the scaled, three-story, test structure, described
in Reference [8], which has been used in previous active control studies at the Structural
Dynamics and Control=Earthquake Engineering Laboratory (SDC=EEL) at the University of
Notre Dame. Because the MR damper is attached between the rst oor and the ground, its

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
730 K.-M. CHOI ET AL.

Table II. Relevant parameters of the three-story building.


 
98:3 0 0
 
Mass matrix, M M= 0 98:3 0  kg
0 0 98:3
 
175 −50 0
  N sec
Damping matrix, C C =  −50 100 −50 
m
0 −50 50
 
12:0 −6:84 0
  N
Stiness matrix, K K =  −6:84 13:7 −6:84  105
m
0 −6:84 6:84

Location matrix of control forces,   = [ −1 0 0 ]T

Location matrix of external loads,   = [ 1 1 1 ]T

displacement is equal to the displacement of the rst oor of the structure relative to the
ground.
The structural measurements used for calculating the desired control force include the ab-
solute accelerations of the three oors of the structures, and the displacement and the velocity
of the MR damper (i.e. y = [xa1 xa2 xa3 x1 ẋ1 ]T ). Thus, Equation (11) can be written as

ż = Az + Bf + E xg (12)

y = C z + Df + v (13)

where
0 I 0 0
A= ; B= ; E=
−M −1 K −M −1 C M −1
 
   
−M −1 K −M −1 C M −1 
   
C = 1 0 0 0 0 0 ; D=  0 
000 100 0

3.2. Semi-active clipped-optimal control system design


For a rst (primary) optimal control design, the absolute acceleration of the ground is taken
to be stationary white noise, and an innite horizon performance index is chosen
  
1
J = lim E {(C z) Q(CZ ) + rfc2 } dt (14)
→∞  0

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
SEMI-ACTIVE FUZZY CONTROL 731

A wide variety of controllers were evaluated by Dyke et al. [8]. The best results were
obtained using r = 10−17 and choosing all of the elements of the weighting matrix Q to be
zero, except for Q33 = 1. Furthermore, the measurement noise is assumed to be identically
distributed, statistically independent Gaussian white noise processes, and Sxg xg =Svi vi =  = 50.

3.3. Semi-active fuzzy control system design


The design of the semi-active fuzzy controller selects the response quantities to be used as
input to the fuzzy controller and the distribution and type of membership functions to be used
for the selected input variables.
In this example, the controller is designed using two input variables (i.e. one is the rst-
oor velocity and the other is the third-oor velocity), each one having ve membership
functions, and one output variable (i.e. command voltage) with three membership functions.
The membership functions chosen for the input and output variables are triangular shaped, as
illustrated in Figure 7. The denitions of the fuzzy variables of input membership function
are as follows: NL = Negative Large, NS = Negative Small, ZE = Zero, PS = Positive Small
and PL = Positive Large. A reasonable range of input values must be selected for the input
membership functions since, if the range is too large or too small, the outermost membership
functions will rarely or essentially be used, respectively, and thus limit the variability of the
control system. The denitions of the fuzzy variables of the output membership function are
as follows: ZE = Zero, M = Medium and L = Large. The voltage is the fuzzy control system
output for the structural control system.
The fuzzy inference rule is completely based on the structural rst-oor velocity and the
third-oor velocity. In this example, the basic concept of the inference rule is that, if the

NL NS ZE PS PL

-X 0 X
(a)

ZE M L

(b) 0 Y

Figure 7. Input and output membership functions: (a) input membership


function; and (b) output membership function.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
732 K.-M. CHOI ET AL.

Table III. Fuzzy inference rule.


NL NS ZE PS PL

NL L L M L M
NS L M ZE M ZE
ZE M ZE M ZE M
PS ZE M ZE M L
PL M L M L L

rst-oor velocity and the third-oor velocity are very large, then the output variable (voltage)
is large. The fuzzy inference rule and control surface are shown in Table III.

3.4. Control results


The test structure is controlled under normal (100%), high (150%) and low (50%) excitation
levels of the El Centro earthquake. Because the system under consideration is a scaled model,
the earthquake must be reproduced at ve times the recorded rate. The MR damper attached
three-story building is the prototype model that Dyke et al. [8] used. Figures 8–10 show the
structural responses of uncontrolled and semi-active fuzzy control systems such as the displace-
ment and the acceleration during the normal (100%), high (150%) and low (50%) El Centro
earthquakes. As seen from the gures, compared to the cases without a MR damper, the third-
oor displacement and the third-oor acceleration responses are reduced signicantly. Figure
11 shows the command voltage sent to the MR damper due to the normal El Centro earth-
quake. Therefore, it is veried that the proposed semi-active fuzzy controller could signicantly
reduce the building structural response due to the wide range of the El Centro earthquake.
The peak responses for the controlled structure are compared to those of an uncontrolled
system, two passive systems (i.e. passive-o and passive-on), a semi-active clipped-optimal
control system, and a semi-active fuzzy control system as shown in Tables IV to VI.
As seen from Table I, the maximum ratios of the peak responses in the passive-o system,
passive-on system, and semi-active fuzzy control system are 0.46, 0.32 and 0.29 for the
displacement, 0.51, 0.55 and 0.50 for the interstory displacement, and 0.51, 0.55 and 0.50 for
the acceleration, respectively. The performance of the system employing the semi-active fuzzy
control system surpasses those of both passive systems. Notice that these performance gains
are achieved by the semi-active fuzzy controller while requiring smaller control forces than
are required in the passive-on case. This means that the semi-active fuzzy control system uses
the forces more eciently to control structural vibration. The maximum ratios of the peak
responses in the semi-active clipped-optimal control system and semi-active fuzzy control
system are 0.23 and 0.29 for the displacement, 0.50 and 0.50 for the interstory displacement,
and 0.82 and 0.50 for the acceleration, respectively. The overall performance of the semi-
active fuzzy control system is much superior to that of the uncontrolled system although
the performance of the semi-active clipped-optimal control system is slightly better for the
maximum ratio of the peak displacement and signicantly worse for the peak acceleration
than that of the semi-active fuzzy controller, respectively. Notice that the peak control force
of the semi-active fuzzy control system is relatively small compared to that of the semi-active
clipped-optimal control system. It is demonstrated that the semi-active fuzzy control system
is very eective in reducing the structural responses due to the normal El Centro earthquake.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
SEMI-ACTIVE FUZZY CONTROL 733

1
Uncontrolled
Controlled

Displacement (cm)
0.5

-0.5

-1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(a) Time (sec)
1500
Uncontrolled
Acceleration (cm/s2)

1000 Controlled
500
0
-500
-1000
-1500
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(b) Time (sec)

Figure 8. Time histories of responses in the third oor under the normal (100%) El Centro
earthquake: (a) displacement; and (b) acceleration.

1.5
Uncontrolled
1 Controlled
Displacement (cm)

0.5
0
-0.5
-1
-1.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(a) Time (sec)

2000 Uncontrolled
Acceleration (cm/s2)

Controlled
1000

-1000

-2000
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(b) Time (sec)

Figure 9. Time histories of responses in the third oor under the high (150%) El Centro
earthquake: (a) displacement; and (b) acceleration.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
734 K.-M. CHOI ET AL.

0.5
Uncontrolled
Controlled

Displacement (cm)
0.25

-0.25

-0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(a) Time (sec)

750
Uncontrolled
Controlled
Acceleration (cm/s2)

500
250
0
-250
-500
-750
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(b) Time (sec)

Figure 10. Time histories of responses in the third oor under the low (50%) El Centro
earthquake: (a) displacement; and (b) acceleration.

1.5
Voltage (V)

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec)

Figure 11. Applied voltage in the semi-active fuzzy case due to the normal (100%) El Centro earthquake.

At high excitation, the performance results of the semi-active fuzzy controller are still
better than those of the passive controllers. The maximum ratios of the peak responses in
the passive-o system, the passive-on system, and the semi-active fuzzy control system are
0.51, 0.32 and 0.29 for the displacement, 0.54, 0.51 and 0.42 for the interstory displacement,
and 0.54, 0.51 and 0.42 for the acceleration, respectively. Also, notice that these performance
gains are achieved by the semi-active fuzzy controller while requiring smaller control forces
than are required in the passive-on case.
In the low excitation simulation, the ratios of the third-oor displacement of the passive-
o and passive-on system are 25.0% larger and 25.0% smaller than those of the semi-active

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
SEMI-ACTIVE FUZZY CONTROL 735

Table IV. Peak responses under the normal (100%) El Centro earthquake (ratio).
Control Semi-active Semi-active
strategy Uncontrolled Passive-o Passive-on clipped-optimal fuzzy

0:549 (1:00) 0:211 (0:33) 0:080 (0:15) 0:114 (0:21) 0:101 (0:18)
xi
0:836 (1:00) 0:357 (0:38) 0:197 (0:24) 0:185 (0:22) 0:184 (0:22)
(cm)
0:973 (1:00) 0:452 (0:46) 0:307 (0:32) 0:219 (0:23) 0:282 (0:29)
0:549 (1:00) 0:211 (0:38) 0:080 (0:15) 0:114 (0:21) 0:101 (0:18)
di
0:317 (1:00) 0:153 (0:48) 0:158 (0:50) 0:90 (0:28) 0:137 (0:43)
(cm)
0:202 (1:00) 0:104 (0:51) 0:111 (0:55) 0:101 (0:50) 0:101 (0:50)

xai 879 (1:00) 421 (0:48) 281 (0:32) 721 (0:82) 400 (0:46)
1071 (1:00) 488 (0:46) 499 (0:47) 746 (0:70) 438 (0:41)
(cm=s2 ) 1404 (1:00) 721 (0:51) 772 (0:55) 706 (0:50) 704 (0:50)
f(N) – 259 981 953 843

Table V. Peak responses under the high (150%) El Centro earthquake (ratio).
Control Semi-active Semi-active
strategy Uncontrolled Passive-o Passive-on clipped-optimal fuzzy

0:823 (1:00) 0:349 (0:42) 0:137 (0:17) 0:178 (0:22) 0:164 (0:20)
xi
1:255 (1:00) 0:587 (0:47) 0:313 (0:25) 0:294 (0:23) 0:300 (0:24)
(cm)
1:460 (1:00) 0:751 (0:51) 0:465 (0:32) 0:371 (0:25) 0:425 (0:29)
0:823 (1:00) 0:349 (0:42) 0:137 (0:17) 0:178 (0:22) 0:164 (0:20)
di
0:475 (1:00) 0:250 (0:53) 0:221 (0:47) 0:126 (0:27) 0:185 (0:39)
(cm)
0:303 (1:00) 0:165 (0:54) 0:156 (0:51) 0:142 (0:47) 0:127 (0:42)

xai 1318 (1:00) 615 (0:47) 509 (0:39) 959 (0:73) 547 (0:42)
1606 (1:00) 755 (0:47) 616 (0:38) 1054 (0:66) 552 (0:34)
(cm=s2 ) 2106 (1:00) 1146 (0:54) 1083 (0:51) 985 (0:48) 885 (0:42)
f(N) – 363 1180 1163 969

Table VI. Peak responses under the low (50%) El Centro earthquake (ratio).
Control Semi-active Semi-active
strategy Uncontrolled Passive-o Passive-on clipped-optimal fuzzy

0:274 (1:00) 0:091 (0:33) 0:040 (0:15) 0:053 (0:19) 0:051 (0:19)
xi
0:418 (1:00) 0:157 (0:38) 0:087 (0:21) 0:087 (0:21) 0:105 (0:25)
(cm)
0:487 (1:00) 0:194 (0:40) 0:119 (0:24) 0:119 (0:24) 0:155 (0:32)
0:274 (1:00) 0:091 (0:33) 0:040 (0:15) 0:053 (0:19) 0:051 (0:19)
di
0:157 (1:00) 0:066 (0:42) 0:079 (0:50) 0:053 (0:34) 0:072 (0:46)
(cm)
0:194 (1:00) 0:038 (0:20) 0:056 (0:29) 0:051 (0:26) 0:051 (0:26)

xai 439 (1:00) 195 (0:44) 141 (0:32) 447 (1:02) 188 (0:43)
535 (1:00) 220 (0:41) 251 (0:47) 354 (0:66) 192 (0:36)
(cm=s2 ) 702 (1:00) 264 (0:38) 387 (0:55) 356 (0:51) 355 (0:51)
f(N) – 159 492 421 393

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736
736 K.-M. CHOI ET AL.

fuzzy control system, respectively. However, the ratios of the maximum oor acceleration of
the passive-o and passive-on system are 13.7% smaller and 7.8% larger than those of the
semi-active fuzzy control system, respectively; and the ratio of the third-oor displacement and
maximum interstory displacement of the semi-active clipped-optimal control system are 25.0%
and 26.1% smaller than those of the semi-active fuzzy control system, respectively. However,
the ratio of the maximum oor acceleration of the semi-active clipped-optimal control system
is larger than that of the uncontrolled system and 92.5% larger than that of the semi-active
fuzzy control system. At low excitation, it is needed to select the proper controller to suit the
designer’s purpose.
It is demonstrated that the semi-active fuzzy control system is very eective in reducing
the building structural responses due to the wide range of earthquake loading conditions.

4. CONCLUSIONS

A semi-active fuzzy control method using a MR damper is presented for seismic response
reduction. Only one presented controller produces directly the desired command voltage using
fuzzy rule inference as the desired force demanded is varied so that the actuator can produce
forces as close as possible to the desired forces. This type of controller has all the advantages
of the fuzzy control algorithm in previous studies. The eectiveness of the MR damper in
reducing the structural responses for a wide range of loading conditions has been demon-
strated via two types of numerical simulations. In addition, the semi-active control system
has many attractive features, such as the bounded-input, bounded-output stability and small
energy requirements. The results of this investigation, therefore, indicate that the semi-active
fuzzy control strategy using a MR damper could be used for control of seismically excited
structures.

ACKNOWLEDGEMENTS
This research was supported by the National Research Laboratory (NRL) program (Grant No. 2000-
N-NL-01-C-251) from the Ministry of Science and Technology in Korea. This nancial support is
gratefully acknowledged.

REFERENCES
1. Zadeh LA. Fuzzy set. Information and Control 1965; 8:338–353.
2. Mamdani EH. Application of fuzzy algorithms for control of simple dynamic plants. Proceedings of the IEE
1974; 121:1585–1588.
3. Brown CB, Yao JTP. Fuzzy sets and structural engineering. Journal of Structural Engineering (ASCE) 1983;
109:1211–1225.
4. Juang C, Elton DJ. Fuzzy logic for estimation of earthquake intensity based on building damage records. Civil
Engineering Systems 1986; 3:187–191.
5. Battaini M, Casciati F, Faravelli L. Fuzzy control of structural vibration. An active mass system driven by a
fuzzy controller. Earthquake Engineering and Structural Dynamics 1998; 27:1267–1276.
6. Teng TL, Peng CP, Chuang C. A study on the application of fuzzy theory to structural active control. Computer
Methods in Applied Mechanics and Engineering 2000; 189:439– 448.
7. Wang AP, Lee CD. Fuzzy sliding mode control for a building structure based on genetic algorithms. Earthquake
Engineering and Structural Dynamics 2002; 31:881–895.
8. Dyke SJ, Spencer BF, Sain MK, Carlson JD. Modeling and control of magnetorheological dampers for seismic
response reduction. Smart Materials and Structures 1996; 5:565–575.
9. Spencer BF, Dyke SJ, Sain MK, Carlson JD. Phenomenological model for magnetoreological dampers. Journal
of Engineering Mechanics (ASCE) 1997; 123:230 –238.

Copyright ? 2004 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2004; 33:723–736