Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. 2013; 20:1021–1042


Published online 31 July 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/stc.1515

Semi-active direct control method for seismic alleviation of


structures using MR dampers

N. Mohajer Rahbari1, B. Farahmand Azar1, S. Talatahari2,*,† and H. Safari1


1
Department of Civil Engineering, University of Tabriz, Tabriz, Iran
2
Marand Faculty of Engineering, University of Tabriz, Tabriz, Iran

ABSTRACT
In the present study, a direct semi-active control method is introduced to mitigate the seismic responses of
structures equipped with magnetorheological (MR) dampers. Bouc–Wen model is utilized to investigate the
nonlinear behavior of the MR dampers, and an efficient controller is proposed based on this model features. In this
method, for conducting the MR damper control force close to the estimated optimal control force at any moment,
the optimal magnitude of current voltage that is applied to produce a magnetic field in the MR damper is calculated
through the use of linear quadratic regulator optimal control algorithm. This algorithm is applied to control seismic
vibrations of a three-story and an 11-story sample shear building that have been equipped with the MR damper
control system, and the performance of the controller is evaluated. Copyright © 2012 John Wiley & Sons, Ltd.

Received 13 February 2011; Revised 27 March 2012; Accepted 25 June 2012

KEY WORDS: structural control; semi-active control; direct control method; MR damper; Bouc–Wen model;
seismic response

1. INTRODUCTION
Preparing structures to deal with critical loads, including seismic excitations and wind force, is the
most important concern in the structural engineering, and although it is not taken into a deserved
consideration, structure would be faced with destructive phenomena and eventually it causes irrepara-
ble harm to the large numbers of people and structures. Current limitations on structural design, such as
low materials damping and vulnerability against unpredicted dynamic excitations because of the fixed
properties and incompatibility in loading conditions, made researchers to innovate new methods of
structural designing. Generally speaking, modern structural control systems can be categorized into
four major terms, namely passive, active, semi-active, and hybrid.
Semi-active control systems are a kind of fully-fledged passive control systems that are intelligent
and adaptable to variations of dynamic loads. On the other hand, in semi-actively controlled systems,
external energy is not injected into the system as active systems. Therefore, they do not require a large
power source, and the probability of instability in structure during the control time will also disappear.
Semi-active control systems have alternative and controllable dynamic properties that alter according
to the feedback and feedforward data at any moment. Thus, an optimal performance is achieved.
Reliability of the semi-active systems includes an added advantage, as these can operate as a passive
device in extreme cases [1].
Magnetorheological (MR) damper is a semi-active control device that provides a controllable
damping force by using MR liquid. MR liquid inside the damper is composed of magnetized tiny

*Correspondence to: S. Talatahari, Marand Faculty of Engineering, University of Tabriz, Tabriz, Iran.

E-mail: siamak.talat@gmail.com

Copyright © 2012 John Wiley & Sons, Ltd.


1022 N. MOHAJER RAHBARI ET AL.

particles that are scattered in a mineral liquid such as silicon oil. While this liquid is exposed to a magnetic
field, particles inside the liquid are polarized and chain form just in a few milliseconds. Hence, liquid will
be changed into semi-solid state and behave as a viscoplastic material. The reversible ability of converting
from a linear viscous fluid to a semi-solid form is the prominent property of MR liquids [2].
The conversion of MR liquid state through adjusting the magnitude of applied magnetic field
appears as a considerable change on damping force of MR damper. Therefore, for controlling purpose,
the damping force of a MR damper is set to an optimal value by reforming the magnitude of the applied
magnetic field according to a defined algorithm.

2. BOUC–WEN MODEL
In 1967, Bouc [3] proposed a uniform and adaptable mathematical model to describe nonlinear hysteresis
behavior of a single degree of freedom structure under the forced vibrations. Then, in 1976, Wen [4]
generalized this model for random vibrations and proposed an approximate solution procedure [5,6].
Afterwards, the Bouc–Wen model has been widely used as a mathematical expression of hysteresis and
nonlinear behavior of systems especially in civil and mechanical engineering. In this model, the nonlinear
force is related with the system deformation through a first-order nonlinear differential equation that has
several undefined parameters. By adjusting these parameters to proper values, the model’s response will
be coincided with the actual behavior of the nonlinear system.
For the first time, Spencer et al. [2] presented the force of MR dampers through applying a Bouc–Wen
model. The simple model of this system is illustrated in Figure 1. In this case, nonlinear force of MR
damper is calculated from Eq. (1), as fallows
 
F ¼ c0 x_ þ k0 x  x0 þaz (1)

where a is the Bouc–Wen model parameter related to the MR material yield stress, k0 and c0 are the stiff-
ness and damping coefficients, respectively, and z is the hysteretic deformation of model that is given as

z_ ¼ gjx_ jzjzjn1  bx_ jzjn þ Ax_ (2)


By adjusting model parameters for a prototype MR damper, it is possible to obtain force–displacement
cycles before and after the yield stress areas. The initial force caused by the existence of the accumulator in
damper is calculated by the initial deformation x0 of the spring k0.
Comparison between experimental results and the response of the Bouc–Wen model (Figure 2)
shows that in spite of possessing less accuracy to chase the force–velocity behavior of the MR damper
at low velocities [2], this model is able to generate the force–displacement curve of the damper
accurately [2,7]. This model has been used to implement a semi-active backstepping control method
in which system response has been showed to be in a great agreement with some experimental data
[7]. It has been also employed to study a neural network predictive control algorithm for calculating
the required voltage for MR dampers to provide the estimated optimal control force [8]. A shear mode
of MR dampers was studied in [9] to conduct a comparative study between four different semi-active
control strategies.

Figure 1. The simple Bouc–Wen model of a schematic magnetorheological damper [2].

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1023

Figure 2. Response curves predicted by the simple Bouc–Wen model in comparison with the experimental
results for the 3 kN, magnetorheological damper under sinusoidal excitation for the constant voltage, 1.5 V: (a)
force–displacement and (b) force–velocity [2].

For achieving optimal performance of control systems equipped with MR dampers, applied voltage
to the current driver must be varied according to the measured feedback at any moment to change the
damping force. Thus, for accounting this accordance, the coefficient a, stiffness k0, and damping
coefficient c0 in Eq. (1) are defined as a linear function of the efficient voltage as given in the following
equations [2,7–15]
aðuÞ ¼ aa þ ab u (3)
k0 ðuÞ ¼ k0a þ k0b u (4)

c0 ðuÞ ¼ c0a þ c0b u (5)


To accommodate the dynamics involved in the MR fluid reaching rheological equilibrium and in
driving the electromagnet in the MR damper, the following first-order filter is employed to calculate
efficient voltage, u [2,7–15]
u_ ¼ Zðu  vÞ (6)
where v is the applied voltage for the current generation.
Although the actual dynamics of electromagnetism of MR dampers is very complex because of the
eddy current effect and nonlinear magnetization of the MR material, comparison between numerically
estimated responses through utilizing given first-order filter (Eq. 6) with experimentally obtained data
from real MR dampers in [2,7,13,16] indicates that the first-order filter can acceptably account for the
complex behavior of MR dampers that is associated with the variable current input. Therefore, in this
paper, a first-order filter is utilized; however, to obtain more practical model, a more complex model
should be developed.

3. SYSTEM DESCRIPTION
The standard description of the state space equation is used to formulate and solve modern control
problems. The equation of motion for a controlled structural system under seismic excitation is
represented as
MS €x þ CS x_ þ KS x ¼ Γf  MS Λ€x g (7)
that can be rewritten as follows

€x ¼ M1 _  M1
S CS x
1
S KS x þ MS Γf  Λ€
xg (8)
thus, the following phrase can be deduced
    n o    
x_ 0  I x 0 0
¼  þ f €x g (9)
€x 1  1
MS KS MS CS _
x 1
MS Γ Λ
To apply a linear control theory to mitigate the seismic response of structures, the second-order differ-
ential equation, Eq. (7), is redefined as a first-order differential equation in terms of the state variable as

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1024 N. MOHAJER RAHBARI ET AL.

nxo
S¼ (10)
x_
Therefore, the state of a controlled structural system subjected to a seismic excitation can be given as
the following equation

S_ ¼ AS þ Bf þ E€x g (11)
where
 
x_
S_ ¼ (12)
€x 2n1
"  
0nn  Inn
A¼  (13)
M1 1
S KS MS CS 2n2n

0nr
B¼ (14)
M1
S Γ 2nr
 
0n1
E¼ (15)
Λ 2n1
here, f is the vector of the dampers’ forces. The rank of matrix Г, dampers’ position, is n  r, in which n
and r denote the number of stories and dampers, respectively.

4. DIRECT CONTROLLER DESIGN


In the early semi-active control algorithms used for MR dampers, the controllers usually work in a
bang–bang (on–off) manner, and the damper force usually approaches to a desired control force in each
time step only by cutting the current off or applying maximum voltage [9,14]. So, in these methods,
computing the optimal voltage is not considered. In 2008, Nguyen et al. [17] developed a direct
controller design for buildings with one MR damper installed in the first story based on Kwok’s model that
employs a hyperbolic tangent function to produce the nonlinear behavior of the MR damper [18,19].
Herein, the well-established Bouc–Wen model is used to develop a direct controller design that is useful
for any arrangement of MR dampers. In this method, after computing the optimal control force, the
optimal input voltage to the current driver is calculated via simulating a linear quadratic regulator
(LQR) controller to conduct the MR damper force close to the optimal control force at each time step.
Through the use of the Bouc–Wen model for the description of the MR damper force, the equation
of motion for a building with r dampers installed in consecutive stories can be given as
M2S €x þ CS x_ þ KS x ¼ 3
ðc0a1 þ c0b1 u1 Þx_ i þ ðk0a1 þ k0b1 u1 Þxi þ ðaa1 þ ab1 u1 Þz1
6 ðc0a2 þ c0b2 u2 Þx_ iþ1 þ ðk0a2 þ k0b2 u2 Þxiþ1 þ ðaa2 þ ab2 u2 Þz2 7
6 7
6 ðc0a3 þ c0b3 u3 Þx_ iþ2 þ ðk0a3 þ k0b3 u3 Þxiþ2 þ ðaa3 þ ab3 u3 Þz3 7
6 7
Γ66:
7  MS Λ€x g
7
(16)
6: 7
6 7
4: 5
_
ðc0ar þ c0br ur Þx iþðr-1Þ þ ðk0ar þ k0br ur Þxiþðr-1Þ þ ðaar þ abr ur Þzr
where i is the story number related to the first installed damper. x, x_ , and z represent relative displace-
ment, relative velocity of story, and hysteretic displacement in the Bouc–Wen model, respectively,
which can be rewritten as
xj ¼ xjþ1  xj
; j ¼ i e i þ ðr  1Þ (17)
x_ j ¼ x_ jþ1  x_ j
 
z_ k ¼ gx_ iþðk-1Þ zk jzk jn1  bx_ iþðk-1Þ jziþk jn þ Ax_ iþðk-1Þ ; k ¼ 1 e r (18)
By using these formulae, the aforementioned equation of motion will be written as

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1025

2 3 2 3
ðc0b1 x_ i þ k0b1 xi þ ab1 z1 Þu1 aa1 z1
6 ðc0b2 x_ iþ1 þ k0b2 xiþ1 þ ab2 z2 Þu2 7 6 aa2 z2 7
6 7 6 7
6 ðc0b3 x_ iþ2 þ k0b3 xiþ2 þ ab3 z3 Þu3 7 6 aa3 z3 7
6 7 6 7
 x_ þ Kx
MS €x þ C  ¼ Γ6 : 7 þ Γ6 : 7  MS Λ€x g (19)
6 7 6 7
6: 7 6: 7
6 7 6 7
4: 5 4: 5
ðc0br x_ iþðr-1 Þ þ k0br xiþðr-1Þ þ abr zr Þur aar zr
where matrices C and K
 are damping and stiffness matrices of the structure, respectively, which are the
modified forms of Cs and Ks in Eq. (16), as
 ðj; jÞ ¼ Cs ðj; jÞ þ c0a
C

C ðj; j-1Þ ¼ Cs ðj; j-1Þ  c0a
 ðj-1; jÞ ¼ Cs ðj-1; jÞ  c0a ; j ¼ i e i þ ðr  1Þ (20)
C
 ðj-1; j-1Þ ¼ Cs ðj-1; j-1Þ þ c0a
C

 ðj; jÞ ¼ Ks ðj; jÞ þ k0a


K
 ðj; j-1Þ ¼ Ks ðj; j-1Þ  k0a
K
 ðj-1; jÞ ¼ Ks ðj-1; jÞ  k0a ; j ¼ i e i þ ðr  1 Þ (21)
K
 ðj-1; j-1Þ ¼ Ks ðj-1; j-1Þ þ k0a
K
Regarding Eq. (19), we will have

S_ ¼ Ad S þ Bd f d þ Ed (22)
where
  
0  I
Ad ¼  (23)
M1 
S K MS C
1 

 
0
Bd ¼ (24)
M1
S Γ
2 32 3
f1 u1
6 f2 76 u2 7
6 76 7
6 f3 76 u3 7
6 76 7
fd ¼ 6
6 : 76 : 7
76 7 (25)
6 : 76 : 7
6 76 7
4 : 54 : 5
fr ur

f k ¼ c0bk x_ iþðk-1Þ þ k0bk xiþðk-1Þ þ abk zk ; k ¼ 1e r (26)


32
aa1 z1
6 aa2 z2 7
6 7
   6 aa3 z3 7
0 0 6 6:
7
7
Ed ¼  €x g þ 1 6 7 (27)
Λ MS Γ 6 : 7
6 7
4: 5
aar zr
Assuming a purely active control system, an LQR controller is designed that minimizes the
following cost function to find the optimal input control force

Z1

1
J¼ ðCSÞT Q ðCSÞ þ f T R f dt (28)
2
0

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1026 N. MOHAJER RAHBARI ET AL.

where Q and R are referred to as weighting matrices, whose magnitudes are assigned according to the
relative importance attached to the state variables and the control forces in the minimization procedure
[20,21]. Thus, the optimal control force vector at each time step can be given as

f opt ¼ R1 BT PS (29)


in which P is the Riccuti matrix and S is the state feedback of the system at each time step.
Capturing the feedback data at each time step, the hysteretic signals zk in Eq. (26) could be calculated
by means of the Bouc–Wen model (Eq. 2). Hence, by taking fd  fopt through inspiration from LQR
method and ignoring the term Ed in Eq. (22) (seismic excitation and hysteretic variables given in Eq.
27), the efficient voltage vector of dampers can be obtained by solving the following system of equations
2 32 3
f1 u1
6 f 76 u2 7
6 2 76 7
6 f3 76 u3 7
6 76 7
f opt ¼ 6
6 : 76 : 7
76 7 (30)
6 : 76 : 7
6 76 7
4 : 54 : 5
fr ur
It should be noticed that hysteretic variables zk are used, in an inverse direction, directly to obtain the
final control voltages by means of Eq. (30) that is almost a common method to other control algorithms
that include Bouc–Wen models to obtain the control voltages [22].
On the other hand, by solving the differential Eq. (6) for each time step, Eq. (31) is obtained through
the following form

u ¼ v þ u0  vÞeZt (31)
where u0 is the damper efficient voltage at the beginning of each time step and v is the applied voltage
to produce the current step. So, the input voltage will be equal to
u  u0 eZt
v¼ (32)
1  eZt
Eventually, by choosing a small time interval, Δt, in which the system state can be assumed
constant, the optimal voltage applied to each damper in each time step is given as
8
>
<0 vk ⩽0
uk  u0k eZΔt
vk ¼ 0⩽vk ⩽Vmax ; k ¼ 1e r (33)
>
: 1  eZΔt
Vmax Vmax ⩽vk
It is obvious that the direct control algorithm is applicable for any arbitrary arrangement of dampers,
and it can also be used with any other alternative controllers to solve optimal control problem. It should
be noticed that the presented algorithm will have its most suitable performance when the magnitude of
dampers’ forces could satisfy the estimated optimal control forces (fopt).
Some experimental studies [2,7,13] explicitly acknowledged that the Bouc–Wen model is one of the
best phenomenological models proposed for description of hysteretic behavior of MR dampers;
however, its major drawback that makes the Bouc–Wen model more complicated to be used than its
counterparts is to solve the nonlinear first-order differential equation (Eq. 2) that produces the
hysteretic behavior. The most of proposed mechanical models for MR dampers are composed of
dashpot, spring, or mass elements in parallel with a nonlinear hysteretic element, although the main
difference between them is only on the interpretation of the hysteretic element. Hence, it appears that
the proposed semi-active direct controller can be easily generalized for some other models such as
Bingham model [2], Kowk model [18], and nonsymmetrical Bouc–Wen model [23].
Furthermore, experimental works on parameter identification of MR dampers showed that the
parameters a, k0, and c0 are not always related to input current through linear equations as given in Eqs
(3)–(5), but quadratic or cubic relations are better identified for large-scale MR dampers [16,23–25].
However, the nonlinear description of these parameters is also easily applicable for direct controller to find
the optimal input voltage. The change is merely the conversion of Eq. 30 from a linear equation to a

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1027

quadratic or a cubic equation. To sum up, the optimal control force vector (fopt) is calculated assuming a
purely active system at first; efficient voltage u for dampers is then gained through solving a quadratic or
a cubic system of equations, and finally, u is transferred to Eq. 33 to obtain the optimal input voltage v.
Although simple Bouc–Wen model defined by Eqs (1) and (2) can yield convincing accuracy,
experiments have shown that, in particular, the first parameter (g) in Eq. (2), in some cases, may have
a nonlinear dependence on the applied current [22,24–26]. Considering a model in which the para-
meters in differential Eq. (2) are related to voltage u, it should be noted that the hysteretic variable
zk in Eq. (26) becomes completely immeasurable. To accommodate this version of Bouc–Wen model
in the proposed direct controller, a simple efficient method proposed in [22] could be easily engaged to
avoid the need of using the real value of the hysteretic signal zk and input an upper or lower bound of it
instead gained from the sign of damper’s velocity according to feedback data. In this case, it is much
better to import the normalized Bouc–Wen model [6,22,24–26] into direct controller that is free of
parameter redundancy demonstrated for simple Bouc–Wen model [27].

5. NUMERICAL EXAMPLES
5.1. Three-story building
A three-story flexible shear building is examined in this section to evaluate the performance of the pro-
posed method. Structural properties of this building are given in Table I.
Rayleigh’s damping matrix in Table I can be calculated from the following relationship
CS ¼ a0 MS þ a1 KS (34)
where
2o1 o2
a0 ¼ x (35)
o1 þ o2

2
a1 ¼ x (36)
o1 þ o2
Two MR dampers with the maximum capacity of 1000 kN are installed in the first story to control
the seismic responses of the structure. Table II provides the optimal values of the Bouc–Wen model

Table I. Structural properties of the three-story building in the first example.


Mass matrix (MS)
2 3
345:6 0 0
34
3  10 0 345:6 0 5ðkgÞ
0 0 345:6

Stiffness matrix (KS)


2 3
2:4 -1:2 0  
3:8  10 -1:2 2:4 -1:2 5 m
84 N

0 -1:2 1:2

Rayleigh’s damping matrix (CS)


2 3
6:5666 -2:5701 0  
10 -2:5701 6:5666 -2:5701 5 N:sec
54
m
0 -2:5701 3:9964

Periods of the first three modes (T)


0.67, 0.24, 0.17 (s)
Damping ratios of the first two modes (x)
1%

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1028 N. MOHAJER RAHBARI ET AL.

Table II. Optimal values of the parameters of the Bouc–Wen model for a 1000 kN magnetorheological damper [10].
Parameter Unit Value
aa kN/m 26
ab kN/m/V 29.1
c0a kN.s/m 105.4
c0b kN.s/m/V 131.6
x0 m 0
g m2 141
b m2 141
A — 2075
n — 2
Z s1 100

parameters for this damper that has been determined through solving a constrained nonlinear optimiza-
tion problem [10] to best fit the experimental data for a full-scale 20 ton MR fluid damper [16]. To ob-
tain the data of a 100 ton damper (i.e., 1000 kN), the experimental data of the 20 ton damper have been
linearly scaled up five times in the damper force and 2.5 times in the stroke of the device [10]. Max-
imum input voltage for this damper is 10 V.
For the LQR controller design, a cost function is chosen that weights the absolute acceleration of the
top floor. Thus, matrix C is obtained through the following equation
h 
C ¼ M1 M1 CS
S KS S nn
(37)

Numerous controllers have been tested, and the best result was obtained by defining weight matrices
Q and R as follows
2 3
0 0 0
Q ¼ 40 0 05 ; R ¼ 1012 Irr
0 0 1

Simulation was conducted in the MATLAB platform. To consider the robustness of the newly de-
veloped semi-active controller and record-to-record variability of ground motions, three earthquake
records with different frequency contents were chosen to excite the structure. Characteristics of the se-
lected earthquakes are gathered in Table III, and the accelerograms are shown in Figures 3–5.
Two passive-performance modes are also investigated as shown next to indicate the fail-safety prop-
erty of the MR damper control system and the efficiency of the semi-active direct controller:
1. Passive-off; in this case, the applied voltage is kept equal to zero (V = 0) during the control time.
2. Passive-on; in this case, the applied voltage is kept at its maximum value (Vmax = 10 V) during
the control time [7,9,11–15].
Moreover, the clipped optimal semi-active controller is also investigated to show the effectiveness
of the proposed modulating control law. The clipped optimal controller has been proved to be one of
the most promising and applicable bang–bang (on–off) controllers that can more effectively mitigate

Table III. Properties of three selected earthquakes records.


Closest distance to Site condition
Earthquake Date Station fault rupture (km) (NEHRP) Duration (s) PGA (g)
Imperial Valley 1940 El Centro 16.9 C 53.73 0.348
Northridge 1994 LA, Hollywood Stor 25.5 C 39.98 0.358
Cape Mendocino 1992 Rio Dell Overpass 18.5 B 35.98 0.385
NEHRP: National Earthquake Hazards Reduction Program; PGA: Peak Ground Acceleration.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1029

Figure 3. Ground acceleration for the El Centro earthquake.

Figure 4. Ground acceleration for the Northridge earthquake.

Figure 5. Ground acceleration for the Cape Mendocino earthquake.

structural responses rather than its other well-established rivals such as decentralized bang–bang con-
trol and control based on lyapunov stability theory [9,14]. In this algorithm, to conduct the MR damper
control force close to the desired optimal control force, applied voltage to the current driver is set as the
following description at any time step:
• If the damper force is equal to the optimal control force (fMR = fopt), then applied voltage is not
changed.
• If the absolute of MR damper force is less than the absolute of calculated optimal control force
and both of them have the same sign, applied voltage should be increased to its maximum value.
• Otherwise, the input voltage is set to zero.
The clipped optimal method can be summarized in the following equation:
 
v ¼ Vmax H f opt  f MR f MR (38)
in which Vmax denotes the maximum voltage that is associated with the saturation of magnetic field in
the MR damper and H{.} is the Heaviside function [7,9–15].

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1030 N. MOHAJER RAHBARI ET AL.

Maximum structural responses for the cases of uncontrolled, passive-on, passive-off, clipped opti-
mal and the new direct control are presented in Tables IV-VI for three prescribed earthquake records
of El Centro, Northridge, and Cape Mendocino. According to the obtained results, the direct controller
can efficiently mitigate the structural responses. For the El Centro earthquake, although this control
system burdens a slight increase to the maximum absolute acceleration of the first story, it decreases
this response for the upper stories by approximately 28%. To conduct an accurate comparative study,
average response reduction of different control strategies is summarized in Table VII. It can be seen
that the direct controller diminishes the peak displacements, drifts, and absolute accelerations of stories
with an average of 32.43%, 33.01%, and 26.16%, respectively, in comparison with uncontrolled

Table IV. Maximum structural responses value due to the El Centro earthquake for the first example.
Story 1 2 3 1 2 3 1 2 3 1
Absolute Control
Control method Displacement (cm) Drift (cm) acceleration (g) force (kN)
Uncontrolled 6.316 11.058 14.024 6.316 5.119 3.126 0.748 1.055 1.400 —
Pass-off 5.685 9.991 12.614 5.685 4.442 2.660 0.748 0.951 1.190 259.7
Reduction (%) 9.98 9.65 10.05 9.98 13.23 14.90 0.04 9.92 15.00 —
Pass-on 4.366 8.080 10.135 4.366 3.842 2.472 0.937 0.968 1.103 2000.0
Reduction (%) 30.87 26.93 27.73 30.87 24.95 20.92 25.25 8.23 21.20 —
C-O 3.944 7.629 9.706 3.944 3.708 2.233 0.722 0.766 1.001 2000.0
Reduction (%) 37.56 31.01 30.79 37.56 27.56 28.58 3.54 27.37 28.51 —
Direct-C 3.897 7.463 9.578 3.897 3.633 2.238 0.770 0.757 1.002 2000.0
Reduction (%) 38.29 32.51 31.70 38.29 29.02 28.40 2.90 28.28 28.40 —

Table V. Maximum structural responses value due to the Northridge earthquake for the first example.
Story 1 2 3 1 2 3 1 2 3 1
Absolute Control
Control method Displacement (cm) Drift (cm) acceleration (g) force (kN)
Uncontrolled 5.180 9.255 11.483 5.180 4.079 2.334 0.684 0.855 1.047 —
Pass-off 4.737 8.483 10.568 4.737 3.780 2.092 0.601 0.781 0.935 233.0
Reduction (%) 8.56 8.34 7.98 8.56 7.33 10.37 12.20 8.71 10.73 —
Pass-on 3.363 5.971 7.651 3.363 2.782 2.094 0.652 0.718 0.935 2000.0
Reduction (%) 35.08 35.49 33.37 35.08 31.81 10.27 4.72 16.02 10.68 —
C-O 3.337 6.070 7.596 3.337 2.732 1.804 0.522 0.703 0.805 2000.0
Reduction (%) 35.58 34.42 33.85 35.58 33.02 22.71 23.64 17.81 23.07 —
Direct-C 3.316 6.022 7.522 3.316 2.706 1.786 0.516 0.679 0.801 2000.0
Reduction (%) 36.00 34.94 34.50 36.00 33.67 23.47 24.53 20.63 23.49 —

Table VI. Maximum structural responses value due to the Cape Mendocino earthquake for the first example.
Story 1 2 3 1 2 3 1 2 3 1
Absolute Control
Control method Displacement (cm) Drift (cm) acceleration (g) force (kN)
Uncontrolled 6.958 11.743 14.828 6.958 5.585 4.227 1.185 1.235 1.892 —
Pass-off 6.687 11.282 14.193 6.687 5.335 3.913 1.075 1.160 1.751 312.8
Reduction (%) 3.89 3.92 4.29 3.89 4.47 7.42 9.22 6.03 7.46 —
Pass-on 4.998 8.946 11.037 4.998 4.097 2.364 0.815 0.874 1.056 2000.0
Reduction (%) 28.16 23.82 25.57 28.16 26.65 44.08 31.16 29.26 44.20 —
C-O 4.889 8.773 10.806 4.889 3.883 2.237 0.758 0.912 1.000 2000.0
Reduction (%) 29.73 25.29 27.13 29.73 30.46 47.08 36.02 26.16 47.14 —
Direct-C 4.841 8.701 10.741 4.841 3.860 2.242 0.726 0.899 1.002 2000.0
Reduction (%) 30.42 25.91 27.57 30.42 30.89 46.95 38.71 27.23 47.05 —

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1031

Table VII. The average response reductions for the first example.
Status Earthquake Displacement Drift Absolute acceleration Total average
Pass-off El Centro 9.89 12.70 8.29 10.29
Northridge 8.29 8.75 10.55 9.20
Cape Mendocino 4.03 5.26 7.57 5.62
Average 7.40 8.90 8.80 8.37
Pass-on El Centro 28.51 25.58 1.40 18.50
Northridge 34.65 25.72 10.47 23.61
Cape Mendocino 25.85 32.96 34.88 31.23
Average 29.67 28.09 15.58 24.45
C-O El Centro 33.12 31.23 19.80 28.05
Northridge 34.62 30.44 21.51 28.86
Cape Mendocino 27.38 35.76 36.44 33.19
Average 31.71 32.48 25.92 30.03
Direct-C El Centro 34.17 31.90 17.93 28.00
Northridge 35.15 31.04 22.88 29.69
Cape Mendocino 27.96 36.09 37.66 33.90
Average 32.43 33.01 26.16 30.53

structure. The direct control method with overall response reduction of 30.53% moderately excels in
reducing all structural responses because of all three earthquakes compared with the bang–bang
clipped optimal controller. It is also recognizable that the direct semi-active controller considerably
performs superior to those passive modes namely passive-on and passive-off.
Seismic responses of the top story, for both uncontrolled and direct semi-active control method, are
illustrated in Figures 6–8. It is observed that, in addition to mitigation of peak responses, the direct
controller could also reduce seismic responses during the control time.
Figure 9 illustrates the comparison of the input control force for passive-on and direct control
method. Despite the better performance of the direct controller system in reducing the structural
responses, the input control force in this method is less than the passive-on mode. This means that
applying a larger control force is not always an effective way to reach a suitable performance of a
control system. It is also obvious in Tables IV-VI that the maximum damper force for both passive-on
and direct method is equal to 2000 kN identically during the control time that stands for well-adaptability
of the semi-active direct approach that is able to employ the whole capacity of dampers to mitigate the
responses of buildings during seismic excitations.
Figure 10 shows the optimal voltage to the current driver calculated by the direct semi-active control
method. As mentioned earlier, in the most of previously proposed semi-active algorithms for MR
damper control systems, the damper force is controlled in a bang–bang manner, but the presented direct
algorithm calculates the optimal magnitude of voltage applied to the current driver directly at any
moment, and it can take any value between zero and Vmax.

Figure 6. Comparison of the top story displacement in the cases of uncontrolled and direct controller system due to
the El Centro earthquake for the first example.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1032 N. MOHAJER RAHBARI ET AL.

Figure 7. Comparison of the top story drift in the cases of uncontrolled and direct controller system due to the El
Centro earthquake for the first example.

Figure 8. Comparison of the top story absolute acceleration in the cases of uncontrolled and direct controller
system due to the El Centro earthquake for the first example.

Figure 9. Comparison of the input control force in the cases of passive-on and direct controller system due to the
El Centro earthquake for the first example.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1033

Figure 10. Optimal applied voltage to magnetorheological dampers in the direct control.

5.2. Eleven-story building


Herein, an 11-story shear frame building located in the city of Rasht in the north of Iran is chosen as the
second test structure [15,28]. The structure represents a typical medium-size multistory building with
the natural period of 0.956 s. The structural properties for this building are listed in Table VIII.
Damping ratio for the first two modes of this building is taken 5%, and Rayleigh’s damping matrix
can be then calculated from Eq. (34).
Three 1000 kN MR dampers (Table II) are used to control the seismic responses of the building.
From Table VIII, it comes to this point that the mass and stiffness of all stories are roughly in the same
range, and drift will be an important type of response to control for upper stories in terms of interstory
shear and evaluating them to remain within the linear range; hence, dampers are placed in the top three
stories that have been shown to perform in the optimal way [15]. As the previously similar studies and
experimental verifications [7–9,11–15], dampers are installed in such a way that they are resisting
stories drift as schematically shown in Figure 11. The proposed semi-active direct controller and the
clipped optimal algorithm along with two prescribed passive scenarios are designed to control system
response. The weight matrices (i.e., Q and R) used in Eq. (28) are given in Table IX to weight the
absolute acceleration of all stories. It is obvious that a larger weighting value is assigned for top story
acceleration by taking Q(11,11) = 10.
Tables X–XII present the maximum structural responses of the uncontrolled and controlled structure
with four aforementioned control algorithms because of the three prescribed earthquake excitations,

Table VIII. Mass and stiffness data of building stories for the second example [25].
Story Mass (kg)  105 Stiffness (N/m)  108
1 2.15 4.68
2 2.01 4.76
3 2.01 4.68
4 2 4.5
5 2.01 4.5
6 2.01 4.5
7 2.01 4.5
8 2.03 4.37
9 2.03 4.37
10 2.03 4.37
11 1.76 3.12

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1034 N. MOHAJER RAHBARI ET AL.

Figure 11. Schematic diagram of the magnetorheological damper control system for the second example.

Table IX. Q and R weight matrices.


Q R
2 3
1 0 0 0 0 0 0 0 0 0 0
60 1 0 0 0 0 0 0 0 0 0 7
6 7
60 0 1 0 0 0 0 0 0 0 0 7
6 7
60 0 0 1 0 0 0 0 0 0 0 7
6 7
60 0 0 0 1 0 0 0 0 0 0 7 1012  I3  3
6 7
60 0 0 0 0 1 0 0 0 0 0 7
6 7
60 0 0 0 0 0 1 0 0 0 0 7
6 7
60 0 0 0 0 0 0 1 0 0 0 7
6 7
60 0 0 0 0 0 0 0 1 0 0 7
6 7
40 0 0 0 0 0 0 0 0 1 0 5
0 0 0 0 0 0 0 0 0 0 10

and Table XIII demonstrates the average of structural response reductions. In these tables, it is revealed
that the clipped optimal controller works a little better than the direct method in the case of displace-
ment reduction due to all excitations other than Northridge earthquake. However, the direct modulating
control algorithm outperforms the clipped optimal controller in decreasing stories’ drifts and performs
significantly greater than the clipped optimal in contracting the acceleration of stories by in excess of
two times on average. Comparing the responses of the structure separately for the selected earthquake
records, according to Table XIII, shows that the performance of the direct method is insignificantly bet-
ter than the clipped optimal controller for the El Centro earthquake, whereas the proposed direct
method is absolutely more effective in comparison with clipped optimal with relative improvement
of (18.76–16.44)  100/18.76 = 12.36% and (19.76–17.41)  100/19.76 = 11.8% for Northridge and
Cape Mendocino earthquakes, respectively. It should be noted that, although the controller’s priority
is to mitigate the responses of the top stories, it can be shown that the performance of the proposed

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1035

Table X. Maximum structural response values due to the El Centro earthquake for the second example.
Story 1 2 3 4 5 6 7 8 9 10 11
Displacement (cm)
Uncontrolled 2.24 4.32 6.27 8.12 9.73 11.12 12.49 13.73 14.77 15.49 15.98
Pass-off 2.14 4.12 5.98 7.72 9.24 10.52 11.77 12.97 13.95 14.63 15.07
Reduction (%) 4.3 4.5 4.6 4.9 5.1 5.4 5.8 5.5 5.5 5.6 5.7
Pass-on 1.56 2.98 4.30 5.53 6.59 7.64 8.62 9.46 10.10 10.53 10.64
Reduction (%) 30.3 30.9 31.4 31.9 32.3 31.3 31.0 31.1 31.6 32.0 33.4
C-O 1.37 2.64 3.86 5.06 6.20 7.21 8.10 8.85 9.40 9.75 9.84
Reduction (%) 38.8 38.8 38.5 37.6 36.3 35.1 35.1 35.5 36.4 37.1 38.4
Direct-C 1.39 2.69 3.94 5.16 6.31 7.34 8.27 9.03 9.57 9.86 9.97
Reduction (%) 37.7 37.6 37.3 36.5 35.2 33.9 33.8 34.2 35.2 36.4 37.6
Drift (cm)
Uncontrolled 2.24 2.08 1.96 1.89 1.77 1.69 1.54 1.37 1.08 0.74 0.51
Pass-off 2.14 1.98 1.86 1.78 1.68 1.60 1.46 1.30 1.03 0.70 0.47
Reduction (%) 4.3 4.6 5.1 6.0 5.3 5.1 5.0 4.9 4.9 5.7 8.0
Pass-on 1.56 1.42 1.32 1.32 1.24 1.13 1.00 0.88 0.68 0.46 0.13
Reduction (%) 30.3 31.5 32.5 30.6 30.0 33.0 34.9 35.9 37.0 38.6 74.8
C-O 1.37 1.28 1.25 1.25 1.17 1.07 0.90 0.80 0.66 0.41 0.19
Reduction (%) 38.8 38.4 36.2 34.0 33.9 36.5 41.6 41.5 39.1 45.0 62.6
Direct-C 1.39 1.30 1.27 1.26 1.19 1.09 0.94 0.80 0.63 0.38 0.15
Reduction (%) 37.7 37.4 35.0 33.3 32.9 35.4 39.2 41.8 41.9 49.1 69.6
Absolute acceleration (g)
Uncontrolled 0.27 0.39 0.47 0.53 0.55 0.57 0.57 0.64 0.77 0.86 0.93
Pass-off 0.28 0.38 0.47 0.52 0.54 0.54 0.54 0.61 0.74 0.84 0.89
Reduction (%) 1.5 0.8 0.6 1.2 2.2 4.8 4.7 4.5 3.5 3.1 3.9
Pass-on 0.30 0.31 0.34 0.37 0.37 0.37 0.45 0.51 0.55 0.59 0.62
Reduction (%) 8.7 20.7 26.8 30.2 32.3 35.6 21.4 20.3 29.1 31.8 32.7
C-O 0.31 0.35 0.32 0.35 0.36 0.42 0.53 0.57 0.75 0.86 0.77
Reduction (%) 13.8 9.3 32.1 33.6 34.4 26.4 6.5 11.0 2.7 0.2 17.1
Direct-C 0.31 0.36 0.33 0.34 0.36 0.42 0.49 0.55 0.67 0.72 0.71
Reduction (%) 13.7 6.3 29.9 35.2 34.7 25.6 14.3 14.1 13.1 15.9 23.0
Damper force (kN)
Pass-off — — — — — — — — 81 78 75
Pass-on — — — — — — — — 748 586 665
C-O — — — — — — — — 939 918 794
Direct-C — — — — — — — — 961 896 751

direct algorithm is not only superior for mitigation of drift for upper stories but also its functionality in
alleviation of drifts and mainly absolute acceleration overshadows the clipped optimal method.
Moreover, despite the direct controller mitigates the acceleration of stories with an average of
7.90%, that is, less than its corresponding value for the passive-on mode with an average of 14.53%,
this algorithm impressively attenuates the displacement and drift of stories with average of 28.30%
and 34.05%, respectively, that completely overcomes passive-on case with the corresponding values
of 22.44% and 27.44%.
Besides this, the direct controller with overall response reduction of 23.42% outperforms its rival. It
should be highlighted that the values of ‘total average’ given in Tables VII and XIII are proposed to
tackle about superiority of different methods that present the average of mean values for the various
types of response. Although they seem physically meaningless, however, they could greatly simplify
the comparing process over the vast amount of obtained results. As a result, drawing a useful analogy
between different methods according to Table XIII, it comes to this point that the direct controller
relatively improves the general condition of the structure by approximately (23.42–21.47)  100/
21.47 = 9.1% and (23.42–21.53)  100/21.52 = 8.8% compared with the passive-on and clipped
optimal controller, respectively.
To assess the potential effectiveness of the semi-active MR damper direct control system, Figure 12
represents the performance comparison of an optimum passive tuned mass damper (TMD) system
designed by employing genetic algorithm (GA) for this model [28] with the MR damper control system

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1036 N. MOHAJER RAHBARI ET AL.

Table XI. Maximum structural response values due to the Northridge earthquake for the second example.
Story 1 2 3 4 5 6 7 8 9 10 11
Displacement (cm)
Uncontrolled 1.79 3.55 5.30 7.06 8.72 10.20 11.45 12.48 13.24 13.75 14.09
Pass-off 1.75 3.47 5.18 6.88 8.50 9.94 11.16 12.16 12.89 13.36 13.65
Reduction (%) 2.4 2.3 2.2 2.7 2.6 2.5 2.5 2.5 2.6 2.9 3.2
Pass-on 1.50 2.93 4.27 5.55 6.63 7.74 8.69 9.48 10.08 10.47 10.57
Reduction (%) 16.5 17.5 19.5 21.4 24.0 24.1 24.1 24.0 23.9 23.9 25.0
C-O 1.57 3.00 4.30 5.43 6.34 7.07 7.67 8.17 8.52 8.82 8.92
Reduction (%) 12.1 15.6 18.8 23.2 27.3 30.6 33.0 34.5 35.6 35.9 36.7
Direct-C 1.54 2.94 4.22 5.33 6.23 6.95 7.53 8.01 8.36 8.60 8.70
Reduction (%) 14.0 17.2 20.3 24.6 28.6 31.8 34.2 35.8 36.9 37.5 38.2
Drift (cm)
Uncontrolled 1.79 1.76 1.77 1.78 1.67 1.50 1.28 1.08 0.85 0.60 0.42
Pass-off 1.75 1.72 1.73 1.74 1.62 1.46 1.25 1.05 0.83 0.59 0.40
Reduction (%) 2.4 2.2 2.5 2.3 2.6 2.5 2.6 3.0 1.6 2.5 5.8
Pass-on 1.50 1.44 1.40 1.36 1.27 1.15 1.00 0.87 0.69 0.47 0.13
Reduction (%) 16.5 18.2 21.1 23.6 23.6 23.4 22.3 19.3 18.8 22.5 68.0
C-O 1.57 1.46 1.34 1.24 1.07 0.94 0.79 0.71 0.54 0.41 0.19
Reduction (%) 12.1 17.4 24.4 30.2 35.7 37.2 38.2 34.4 36.1 32.5 55.4
Direct-C 1.54 1.43 1.31 1.22 1.04 0.93 0.78 0.70 0.56 0.36 0.16
Reduction (%) 14.0 18.9 25.8 31.2 37.8 38.2 38.7 34.8 33.2 40.7 61.5
Absolute acceleration (g)
Uncontrolled 0.32 0.36 0.46 0.50 0.52 0.56 0.60 0.60 0.60 0.65 0.75
Pass-off 0.33 0.38 0.48 0.52 0.54 0.58 0.61 0.60 0.59 0.65 0.75
Reduction (%) 3.2 5.8 4.7 3.1 3.7 3.2 0.9 0.9 1.3 0.6 0.5
Pass-on 0.35 0.40 0.46 0.48 0.50 0.52 0.52 0.52 0.54 0.59 0.63
Reduction (%) 10.0 9.8 0.4 4.0 3.1 6.6 13.1 14.0 10.2 8.6 15.8
C-O 0.39 0.51 0.54 0.53 0.54 0.52 0.51 0.53 0.81 0.84 0.71
Reduction (%) 22.5 40.7 18.0 5.1 3.5 6.7 15.7 12.5 35.0 29.5 5.1
Direct-C 0.39 0.50 0.54 0.52 0.55 0.52 0.51 0.50 0.73 0.73 0.70
Reduction (%) 22.8 37.3 18.0 2.7 5.3 6.4 14.9 17.1 20.8 13.4 7.2
Damper force (kN)
Pass-off — — — — — — — — 79 76 74
Pass-on — — — — — — — — 736 598 688
C-O — — — — — — — — 960 881 707
Direct-C — — — — — — — — 971 875 709

in the cases of semi-active direct control, semi-active clipped optimal control, and passive-on operation
due to the El Centro earthquake. Figure 13 shows a schematic TMD system for this example. To find
the optimal parameters of TMD, GA-based TMD (GATMD) controller takes the integral of the time
multiplied absolute value of the error (ITAE) based on relative displacement of all stories in the build-
ing as a performance index of the optimization criterion to be solved using GA without specifying
which mode should be controlled [28].
It is plainly observable in Figure 12 that passive MR dampers do not perform as well as GATMD
system in displacement reduction because of their particular given arrangement, whereas a different
arrangement of MR dampers might have greatly enhanced the structural response, even in comparison
with the GATMD system. On the contrary, the fail-safe semi-active MR damper control systems with
both aforementioned semi-active control algorithms could reach the efficiency of such a GATMD
system in terms of mitigation of stories displacement by only use of three dampers installed in the
top stories of the building. Although the GATMD results in more reduction for displacement of bottom
stories, semi-active MR damper systems operate better for top stories and perform similar to the
GATMD system for middles. It should be mentioned that the clipped optimal controller diminishes
the displacement of stories a little bit (by almost 1.12%) more than direct controller under the El Centro
earthquake (Table XIII); however, the direct algorithm is considerably better at alleviating drift and
acceleration of stories as shown in Tables X through XIII.
Figures 14–16 illustrate the displacement, drift, and absolute acceleration of the top story in a
comparable manner for the uncontrolled structure and the building equipped with MR dampers. As

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1037

Table XII. Maximum structural responses values due to the Cape Mendocino earthquake for the second example.
Story 1 2 3 4 5 6 7 8 9 10 11
Displacement (cm)
Uncontrolled 1.90 3.70 5.50 7.36 9.14 10.78 12.24 13.53 14.56 15.28 15.78
Pass-off 1.82 3.57 5.34 7.16 8.88 10.47 11.86 13.10 14.09 14.77 15.24
Reduction (%) 4.0 3.5 2.9 2.8 2.8 2.9 3.1 3.2 3.3 3.3 3.4
Pass-on 1.71 3.34 4.95 6.53 7.94 9.16 10.36 11.44 12.26 12.83 13.02
Reduction (%) 9.9 9.7 10.0 11.4 13.2 15.1 15.4 15.5 15.8 16.0 17.5
C-O 1.48 2.94 4.39 5.85 7.28 8.62 9.79 10.79 11.57 12.03 12.23
Reduction (%) 22.1 20.6 20.2 20.6 20.3 20.0 20.1 20.2 20.5 21.3 22.5
Direct-C 1.50 2.97 4.43 5.90 7.36 8.73 9.93 10.94 11.67 12.09 12.27
Reduction (%) 21.0 19.7 19.5 19.8 19.5 19.0 18.9 19.2 19.9 20.8 22.2
Drift (cm)
Uncontrolled 1.90 1.82 1.84 1.89 1.81 1.66 1.50 1.37 1.11 0.78 0.54
Pass-off 1.82 1.77 1.79 1.83 1.75 1.59 1.48 1.34 1.09 0.76 0.51
Reduction (%) 4.0 2.9 2.4 2.9 3.5 3.9 1.5 2.1 2.1 2.6 5.0
Pass-on 1.71 1.65 1.62 1.58 1.53 1.43 1.29 1.13 0.88 0.59 0.19
Reduction (%) 9.9 9.6 12.2 16.2 15.3 13.5 14.3 17.8 20.9 24.2 65.3
C-O 1.48 1.46 1.45 1.51 1.47 1.34 1.18 1.06 0.78 0.56 0.22
Reduction (%) 22.1 19.9 20.9 19.9 18.6 18.9 21.8 23.0 30.4 28.3 58.9
Direct-C 1.50 1.47 1.47 1.53 1.48 1.37 1.20 1.03 0.77 0.46 0.19
Reduction (%) 21.0 19.2 20.3 19.2 18.2 17.4 20.3 24.9 30.8 40.5 63.9
Absolute acceleration (g)
Uncontrolled 0.42 0.50 0.57 0.57 0.59 0.56 0.62 0.66 0.74 0.87 0.97
Pass-off 0.43 0.52 0.57 0.57 0.57 0.54 0.61 0.64 0.74 0.86 0.96
Reduction (%) 3.0 3.0 0.4 1.5 2.7 3.6 2.4 2.2 0.4 0.9 0.6
Pass-on 0.41 0.43 0.44 0.44 0.45 0.50 0.57 0.62 0.69 0.74 0.79
Reduction (%) 2.4 14.6 23.1 24.1 24.1 10.0 8.9 5.0 6.7 14.2 18.9
C-O 0.43 0.45 0.46 0.43 0.46 0.49 0.63 0.69 0.80 1.03 0.87
Reduction (%) 3.6 11.0 20.6 25.4 21.1 12.3 1.7 5.0 8.2 18.7 10.4
Direct-C 0.43 0.45 0.44 0.43 0.44 0.49 0.59 0.64 0.68 0.76 0.80
Reduction (%) 2.9 10.7 22.3 25.1 24.7 12.1 4.7 1.9 7.7 12.6 17.8
Damper force (kN)
Pass-off — — — — — — — — 81 78 75
Pass-on — — — — — — — — 771 591 798
C-O — — — — — — — — 1000 885 840
Direct-C — — — — — — — — 1000 935 818

Table XIII. The average response reductions for the second example.
Status Earthquake Displacement Drift Absolute acceleration Total Average
Pass-off El Centro 5.17 5.36 2.52 4.35
Northridge 2.58 2.74 2.04 1.09
Cape Mendocino 3.20 2.99 0.78 2.32
Average 3.65 3.70 0.42 2.59
Pass-on El Centro 31.58 37.20 24.73 31.17
Northridge 22.17 25.20 5.03 17.47
Cape Mendocino 13.58 19.93 13.81 15.77
Average 22.44 27.44 14.52 21.47
C-O El Centro 37.06 40.69 14.51 30.75
Northridge 27.57 32.15 10.40 16.44
Cape Mendocino 20.76 25.69 5.77 17.41
Average 28.46 32.84 3.29 21.53
Direct-C El Centro 35.94 41.21 18.03 31.73
Northridge 29.00 34.06 6.77 18.76
Cape Mendocino 19.96 26.88 12.44 19.76
Average 28.30 34.05 7.90 23.42

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1038 N. MOHAJER RAHBARI ET AL.

it can be seen, this control system is capable of mitigating all seismic responses of building during
earthquake excitations. Figures 17–19 display the optimal applied voltages to the current driver calcu-
lated by the direct controller.

Figure 12. Displacement reduction of the magnetorheological (MR) damper control system in the cases of direct
controller and passive-on performance in comparison with genetic algorithm-based tuned mass damper (GATMD)
due to the El Centro earthquake for the second example.

Figure 13. Schematic diagram of the tuned mass damper (TMD) system for the second example.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1039

Figure 14. Comparison of the 11th story displacement in the cases of uncontrolled and direct controller due to the
El Centro earthquake.

Figure 15. Comparison of the 11th story drift in the cases of uncontrolled and direct controller due to the El Centro
earthquake.

Figure 16. Comparison of the 11th story acceleration in the cases of uncontrolled and direct controller due to the
El Centro earthquake.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1040 N. MOHAJER RAHBARI ET AL.

Figure 17. Optimal applied voltage to magnetorheological dampers in story 9.

Figure 18. Optimal applied voltage to magnetorheological dampers in story 10.

Figure 19. Optimal applied voltage to magnetorheological dampers in story 11.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
SEMI-ACTIVE DIRECT CONTROL METHOD 1041

6. DISCUSSION AND CONCLUSION


A direct semi-active controller based on the Bouc–Wen model has been implemented in the present
study to control the seismic vibrations of frame buildings equipped with MR dampers. To lead the
MR damper force close to the optimal control force during the control time, this method introduces
a direct procedure for calculation of optimal voltage applied to the current driver with getting inspira-
tion from the LQR control algorithm.
Two sample shear buildings with semi-active MR damper control systems were evaluated and the
obtained results indicate that the proposed direct control algorithm can calculate the optimal input volt-
age in accordance with the received feedback data at any time step and accordingly changes the control
force produced by the MR damper at once. Furthermore, this semi-active controller improves the
dynamic characteristics of the MR damper continuously during the control period.
As the obtained results for the second case study indicated, the passive-on control is unexpectedly
more effective on reducing the absolute acceleration for the second case study than the semi-active
controllers. This common drawback with semi-active controllers has been previously seen in several
other studies [7,11,12,14] that employed the clipped optimal method along with H2/LQG algorithm,
and the reason is because of inaccuracies involved in LQR algorithm affecting the system mainly by
ignoring the seismic excitation term for both semi-active controllers to find the true optimal control
forces (vector of fopt) and maybe by ignoring the hysteretic variables in Eq. (27) to find the optimal
control voltage in direct algorithm. The arrangement and number of MR dampers might be another
culprit. However, it seems the direct control algorithm can be used with the other algorithms solving
optimal control problem to solve this problem. Moreover, despite this drawback, both semi-active
controllers perform much better than passive-on mode in contracting the displacement and interstory
drift. Therefore, meeting the two criteria of three means that semi-active methods are superior to
passive-on. In other words, semi-active methods especially the proposed direct modulating control
algorithm generally show more effective performance of seismic mitigation in comparison with a
passive damper operating at the maximum power allowed.
It was also illustrated that the proposed direct controller more or less outperforms its bang–bang
counterpart namely clipped optimal controller somehow, and this superiority is especially manifested
in terms of mitigating absolute accelerations of building stories.
The direct controller was shown to successfully reduce the peak responses of both buildings and signif-
icantly alleviates structural vibrations during earthquake excitations. Comparison between the GATMD
control system and the semi-active MR damper direct control system for the 11-story model demonstrated
that the direct control system can reach the efficiency of GATMD controller in terms of reducing displace-
ment of stories by employing a few number of MR dampers in the building in addition to highly mitigation
of other structural responses. However, MR dampers will not cause instability in the structure and operate
as a safe passive device after probable failure in the power system. It was shown that the proposed direct
controller surpasses its clipped optimal counterpart and the so-called passive-on performance.
It is noteworthy to mention that the performance of the resulting controlled system and the requirements
of the control device are highly dependent on the control algorithm employed, and algorithms that explicitly
incorporate actuator dynamics and control-structure interaction into the design process may offer additional
performance gains [9,14]; therefore, as a suggestion for future works, the proposed direct controller can be
further implemented by an appropriate dynamic model of a real current driver that accounts for operating
principles. Furthermore, utilizing an accurate control algorithm that could account for uncertainties and
input excitations would considerably promote the proposed direct control algorithm because of its high
association with accurate calculation of optimal control force. As the last suggestion, finding a way to
reduce the acceleration response of semi-active controllers (which is the potential risk of structure
associated with the nonstructural component damage) can be considered as a future work.

REFERENCES

1. Nagarajaiah S. Adaptive Passive, Semiactive, Smart Tuned Mass Dampers: Identification and Control Using Empirical
Mode Decomposition, Hilbert Transform, and Short-Term Fourier Transform. Structural Control and Health Monitoring
2009; 16(7–8):800–841.
2. Spencer BF Jr, Dyke SJ, Sain MK, Carlson JD. Phenomenological Model of a Magnetorheological Damper. Journal of
Engineering Mechanics (ASCE) 1997; 123(3):230–238.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc
1042 N. MOHAJER RAHBARI ET AL.

3. Bouc R. Forced vibrations of mechanical systems with hysteresis. Proceeding of the 4th Conference on Nonlinear
Oscillations, Prague, Czechoslovakia, 1967.
4. Wen YK. Method for random vibration of hysteretic systems. Journal of the Engineering Mechanics Division 1976;
102(2):249–263.
5. Foliente GC. Hysteresis Modelling of Wood Joints and Structural Systems. Journal of Structural Engineering (ASCE) 1995;
121(6):1013–1022.
6. Ikhouane F, Rodellar J. Systems with Hysteresis: Analysis, Identification and Control Using the Bouc-Wen Model. John
Wiley & Sons Ltd: West Sussex, 2007.
7. Zapateiro M, Karimi HR, Luo N, Phillips BM, Spencer BF Jr. Semiactive Backstepping Control for Vibration Reduction in a
Structure with Magnetorheological Damper Subject to Seismic Motions. Journal of Intelligent Material Systems and
Structures 2009; 20(17):2037–2053.
8. Karamodin A, Kazemi H. Semi-Active Control of Structures Using Neuro-Predictive Algorithm for MR Dampers.
Structural Control and Health Monitoring 2010; 17(3):237–253.
9. Jansen LM, Dyke SJ. Semi-Active Control Strategies for MR Dampers: A Comparative Study. Journal of Engineering
Mechanics (ASCE) 2000; 126(8):795–803.
10. Jung HJ, Spencer BF Jr, Lee IW. Control of Seismically Excited Cable-Stayed Bridge Employing Magnetorheological Fluid
Dampers. Journal of Structural Engineering (ASCE) 2003; 129(7):873–883.
11. Dyke SJ, Spencer Jr BF. Seismic Response Control Using Multiple MR Dampers. Proceedings of the 2nd International
Workshop on Structural Control, Hong Kong University of Science and Technology Research Centre, Hong Kong, 1996;
163–173.
12. Dyke SJ, Spencer BF Jr, Sain MK, Carlson JD. Modelling and Control of Magnetorheological Dampers for Seismic
Response Reduction. Smart Materials and Structures 1996; 5(5):565–575.
13. Yi F, Dyke SJ, Caicedo JM, Carlson JD. Experimental Verification of Multiinput Seismic Control Strategies for Smart
Dampers. Journal of Engineering Mechanics (ASCE) 2001; 127(11):1152–1164.
14. Dyke SJ, Spencer Jr BF. A Comparison of Semi-Active Control Strategies for the MR Damper. Proceedings of the 1997
IASTED International Conference on Intelligent Information Systems (IIS ’97), Grand Bahama Island, BAHAMAS, 1997.
15. Farahmand Azar B, Mohajer Rahbari N, Talatahari S. Seismic Mitigation of Tall Buildings Using Magneto-Rheological
Dampers. Asian Journal of Civil Engineering 2011; 12(5):637–649.
16. Yang G, Spencer BF Jr, Carlson JD, Sain MK. Large-scale MR Fluid Dampers: Modelling and Dynamic Performance
Considerations. Engineering Structures 2002; 24(3):309–323.
17. Nguyen MT, Dalvand H, Yu YH, Ha QP. Seismic Responses of Civil Structures under Magnetorheological-Device Direct
Control. 25th International Symposium on Automation and Robotics in Construction (ISARC), Institute of Internet and
Intelligent Technologies, Vilnius Gediminas Technical University, Lithuania, 2008; 106–112.
18. Kwok NM, Ha QP, Nguyen TH, Li J, Samali B. A Novel Hysteretic Model for Magnetorheological Fluid Dampers and
Parameter Identification Using Particle Swarm Optimization. Sensors and Actuators 2006; 132(2):441–451.
19. Ha QP, Kwok NM, Nguyen MT, Li J, Samali B. Mitigation of Seismic Responses on Building Structures Using MR
Dampers with Lyapunov-Based Control. Structural Control and Health Monitoring 2008. doi:10.1002/stc.218.
20. Cheng FY, Jiang H, Lou K. Smart Structures: Innovative Systems for Seismic Response Control. Taylor & Francis Group:
Boca Raton, 2008.
21. Soong TT. Active Structural Control: Theory and Practice. John Wiley & Sons Inc.: New York, 1990.
22. Bahar A, Pozo F, Acho L, Rodellar J, Barbat A. Hierarchical Semi-Active Control of Base-Isolated Structures Using a New
Inverse Model of Magnetorheological Dampers. Computers and Structures 2010; 88(7–8):483–496.
23. Kwok NM, Ha QP, Nguyen MT, Li J, Samali B. Bouc-Wen Model Parameter Identification for a MR Fluid Damper Using
Computationally Efficient GA. ISA Transactions 2007; 46(2):167–179.
24. Rodriguez A, Iwata N, Ikhouane F, Rodellar J. Model Identification of a Large-Scale Magnetorheological Fluid Damper.
Smart Materials and Structures 2009; 18(1):1–12.
25. Aguirre N, Ikhouane F, Rodellar J, Christenson R. Parametric Identification of the Dahl Model for Large Scale MR Damper.
Structural Control and Health Monitoring 2011. doi:10.1002/stc.434.
26. Bahar A, Pozo F, Acho L, Rodellar J, Barbat A. Parameter Identification of Large-Scale Magnetorheological Dampers in a
Benchmark Building. Computers and Structures 2010; 88(3–4):198–206.
27. Ma F, Zhang H, Bockstedte A, Foliente GC, Paevere P. Parameter Analysis of the Differential Model of Hysteresis. Journal
of Applied Mechanics 2004; 71(3):342–349.
28. Shayeghi A, Eimani Kalasar H, Shayeghi H. Seismic Control of Tall Building Using a New Optimum Controller Based on
GA. International Journal of Applied Science, Engineering and Technology 2009; 5(2):85–92.

Copyright © 2012 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2013; 20:1021–1042
DOI: 10.1002/stc

You might also like