Original Article

Journal of Low Frequency Noise,

Vibration and Active Control
2018, Vol. 37(2) 216–234
Experimental investigation of semiactive ! The Author(s) 2018
DOI: 10.1177/0263092317711985
robust control for structures with journals.sagepub.com/home/lfn

magnetorheological dampers

Huseyin Aggumus and Saban Cetin

Abstract
This study investigates reducing the vibration of structures under different earthquakes using magnetorheological dam-
pers. To investigate the effect of the magnetorheological damper arrangement on reducing the vibration amplitude of
structures, experimental studies are conducted in which magnetorheological dampers are commanded by a robust
controller. So, the performance of the system is investigated using different combinations of two magnetorheological
dampers on a building and three different arrangements for dampers are considered for experimental study. Additionally,
an H1 controller is designed to determine the voltage transmitted to the magnetorheological dampers. The results show
that the magnetorheological damper commanded with the robust controller effectively reduces the vibration of a six-
story steel structure. Furthermore, the magnetorheological damper arrangement in which the one end is connected to
the ground reduces the vibration amplitudes.

Keywords
Structural control, magnetorheological damper, magnetorheological damper layout, H1 robust control, shaking table test

Introduction
The control of linear and nonlinear systems equipped with semiactive devices is interesting subject that can be
studied both experimentally and theoretically.1–10 In particular, semiactive systems have the reliability of passive
systems and consume less energy than active systems. Several systems are used to reduce undesired vibrations that
are emitted from buildings. Three types of systems have been developed and successfully studied in the literature.
These systems can be classified as passive, active, and semiactive systems. Active and semiactive systems can be
controlled electronically. Passive systems are simple, classical systems that are currently used. Their characteristic
features cannot be controlled with an external input. Their structure is simple, reliable, and low cost. When varying
conditions are considered as the damping force is held constant, their performance is limited. In active systems, the
necessary force of the system is achieved using active controllers (i.e. hydraulic cylinders). Such systems achieve
high performance over a wide frequency range. Regarding active systems, the use of equipment, such as sensors,
servo-valves, and high-energy demand leads to high system costs. However, semiactive systems can act as both
active and passive systems. Their performance is similar to active systems. Because semiactive systems can act as a
passive system in the case of problems in the control systems, they are more reliable than active systems.
Magnetorheological (MR) dampers are semiactive devices that can absorb the vibratory energy according to
the motion of the MR damper and input control voltage applied to them. Their advantages are mechanical
simplicity, high dynamic range, low power requirements, low cost, large force capacity, and inherit stability.

Department of Mechanical Engineering, Yildiz Technical University, Istanbul, Turkey

Corresponding author:
Huseyin Aggumus, Department of Mechanical Engineering Yildiz Technical University, Istanbul 34349, Turkey.
Email: haggumus@yildiz.edu.tr

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License (http://www.creativecommons.org/licenses/by/
4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and
Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).
Aggumus and Cetin 217

Because of these advantages, these devices are used in many application areas such as suspension systems,5,6,8,9
portal frame structure,11 and structural control.4,12–25
This study reports the control and various arrangement combinations of two MR dampers that are used to
reduce structural vibrations. A number of studies are conducted regarding the control of the MR damper and the
arrangement of multiple MR dampers. Dyke et al.12 employed MR dampers on the first floor of a three-story
structure and used a clipped-optimal control strategy with a momentum feedback base. Dyke et al.13 applied a
clipped optimal control algorithm, which was previously tested via simulations,12 by connecting the MR damper
to the test model between the ground floor and first floor. Jansen and Dyke14 implemented MR dampers in a
building model with six degrees of freedom (DOFs); the MR damper was laid out on the first two floors in a
parallel configuration. Yi et al.15 utilized a six-story structural model to test MR dampers located between the
ground floor and first floor and between the first floor and second floor. Two different semiactive control algo-
rithms are used, including Lyapunov-based control and clipped optimal control. Aldemir16 studied the perform-
ance of an optimal semicontrolled MR damper in a system with one DOF, which depended on the tuned mass
damper for reducing the vibration of the structure. Tusset et al.11 studied chaotic motions of a portal frame
structure under nonideal loading. Two control algorithms are used, including nonlinear feedforward control
and state feedback control. Cho et al.17 implemented an MR damper on the first two floors of a six-story building.
Model control was designed for a seismically stimulated six-story structure. Aldemir and Gavin18 implemented
passive, optimal, and pseudoskyhook control in a 1-DOF system and compared the responses of the system. Kim
et al.19 used a three-story structure model and fixed an MR damper on the first floor. A LMI-based semiactive
fuzzy logic controller was applied to command the MR damper. Lu et al.20 investigated the layouts of three MR
dampers in a six-story steel frame structural system and compared several semiactive control algorithms, including
decentralized sliding mode control, linear quadratic regulator (LQR) control, and passive-on and passive-off
control. The MR dampers were fixed to the structure in the four different cases. In the first case, a MR
damper was located on the first floor. For the second case, two MR dampers placed on the first and third
floors were used. In the third case, three MR dampers were located on the first, second, and third floors. For
the fourth case, three MR dampers were connected to the first, third, and fifth floors of a six-story steel frame
structure. Swartz and Lynch21 verified the LQR control method on a seismically stimulated six-story structure
model with ideal actuators using numeral simulations. Six MR dampers were fixed on all of the floors. Aldemir22
developed a causal suboptimal control, placing an MR damper on the first floor of the base-isolated structure with
2 DOFs. The results were compared with instantaneous optimally controlled and uncontrolled situations, and the
presented control algorithm was shown to be effective in reducing the effects of earthquakes on the structure.
Bitaraf et al.23 used two MR dampers, which were placed on the first and second floors of the structure, and
applied two control methods. The first was a direct adaptive control based on a simple adaptation technique. The
other was developed using genetic-based fuzzy control. Heo and Jeon24 used a three-story structural model. MR
dampers were placed on the first and second floors. The basis for the control algorithm was Lyapunov stability
theory. Rahbari et al.25 designed a LQR via an optimal control algorithm. Two structural models were used to
implement the controller as three- and 11-story structures. MR dampers were located on all of the floors of the
three-story structure and on the ninth, 10th, and 11th floors of the 11-story structure. Cetin et al.4 used a six-story
steel frame structural model, and an MR damper was implemented on the first floor. A nonlinear adaptive
controller based on the Lyapunov technique, which can balance parametric uncertainties, was used to command
the MR damper voltage.
The studies in the above literature review are considered from the view of their control methods, the MR
damper layouts, the MR damper numbers, and the number of floors. Considering the above literature review, the
recommended layout combination has not yet been examined, although the effects of the MR damper’s layout on
vibration control have been studied. Thus, one of the most important contributions of this study to the literature is
the recommended combination of the MR damper layout.
The present study investigates the arrangements of two MR dampers distributed on different floors based on
experimental studies. The primary contribution of this study is the investigation of three combinations of MR
damper layouts, two of which have not been studied by other researchers. A robust control method is used to
command the MR dampers. The MR damper is a semiactive control device that can be controlled by only the
voltage transmitted to the electromagnetic coils. An H1 robust controller is designed for the MR damper voltage,
which affects the damping force of the MR damper. The performance of the designed controller and MR damper
is experimentally tested using a shaking table by placing the various combinations on a six-story steel frame
structure. Three different combinations of the MR damper are tested in this study. The performance of the control
algorithm and the effect of the MR damper arrangements are evaluated based on the shaking table tests using
218 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

earthquake data. The results are compared with situations in which MR damper is not connected and when the
MR damper is passive and semiactive.

Problem formulation
In this study, three different MR damper locations (arrangements) in a six-story steel frame structure are inves-
tigated, as shown in Figure 1. The three MR damper arrangements that are examined are listed below:

Case A: An MR damper is installed between the first floor and the ground (MR1).
Case B: An MR damper is installed between the second floor and the ground (MR2).
Case C: Two MR dampers are used. The first MR damper is installed between the first floor and the ground, and
the second MR damper is installed between the second floor and the ground (MR12).

The mathematical movement of the building models shown in Figure 1 can be given as

Ms x€ ðtÞ þ Cs x_ ðtÞ þ Ks ðtÞ ¼ HfðtÞ  Ms Lx€ g ðtÞ ð1Þ

where f(t) is the damping force of the MR damper and Ms, Cs, and Ks 2 R66 are the mass, damping, and stiffness
matrices, respectively. x€ ðtÞ, x_ ðtÞ and xðtÞ 2 R61 are the acceleration, velocity, and displacement vectors, respect-
ively. Unidirectional horizontal
 movement is considered
T in this model. The relative displacement vector for each of
the three models is x ¼ x1 x2 x3 x4 x5 x6 . If the MR damper is on the first floor, as shown in Figure 1,
the H vector, which indicates the placement of the control units in the system, is

 T
HMR1 ¼ 1 0 0 0 0 0 ð2Þ

If the MR damper is placed between the second floor and the ground

 T
HMR2 ¼ 0 1 0 0 0 0 ð3Þ

Figure 1. Three cases of the MR damper locations. (a) The MR damper is located between the first floor and the ground, (b) the MR
damper is located between the second floor and the ground, and (c) the MR damper is located on both the first and second floors.
MR: magnetorheological.
Aggumus and Cetin 219

If the MR damper is on both the first and second floors, HMR1 and HMR2 are separately
 
H ¼ HMR1 H MR2 ð4Þ
 T
The seismic input vector is L ¼ 1 1 1 1 1 1 . x€ g ðtÞ represents the earthquake ground acceleration.
Schematic models of the MR damper arrangements are shown in Figure 1.

Robust control design

The results of the previous section allow for the creation of a model that is used in the control design. In this
section, a robust controller is designed based on the reduced-order model (ROM) and then applied to the full-
order model (FOM). A mixed sensitivity problem is considered for the control design. The performance of the
designed controller is presented, and the semiactive controller implementation rule is defined.

Identification of the system and model reduction

In most cases, controlling the lower modes effectively reduces the amplitude of the structural system. In particular,
the effects of the lower modes are greater than the higher modes in the vibrations of flexible systems, such as elastic
rotors, beams, and high-rise buildings.26 Therefore, controlling only the lower modes is important and sufficient to
mitigate vibration. Moreover, such flexible systems are modeled with many DOFs. The controller can be designed
based on these FOMs. Nevertheless, it is difficult to design and implement control using FOMs when the system
model is too large, especially in experimental studies and real applications. Therefore, using ROMs that capture
the dominant characteristics of the system represents a meaningful alternative for control design and
implementation.
In this study, a modal approach is chosen to reduce the order of the model. An H1 controller is designed based
on this ROM and is applied to the FOM. In this approach, the higher modes are truncated, and the lower modes
are considered in the model used to control the design. Neglected system dynamics occur in the reduction pro-
cedure. All of these dynamics induce system uncertainties. Considering again the structural system that was
modeled in the previous section, the FOM of the system in physical coordinates can be written in the state-
space model as follows
x_ f ¼ Af xf þ Bf u, yr ¼ Cf xf ð5Þ
where Af, Bf, Cf, and Df are the linear system matrices in the state-space form given, respectively, by
   
0 I 0  
Af ¼ , Bf ¼ , Cf ¼ Cy 0 , Df ¼ ½0 ð6Þ
M1
f Kf M1
f Cf M1
f Ff

Here, the Cy1 and Cy2 vectors represent the locations of the measurements and are defined as
   
Cy1 ¼ 1 0 0 0 0 0 , Cy2 ¼ 0 1 0 0 0 0 ð7Þ

The system matrix can be designed as

 
Af Bf
Pf ðsÞ ¼ ð8Þ
Cf 0

For the model order reduction, the system must be transformed from the physical space to the modal space.
Transformation from the physical space to the modal space can be performed using the following operator
 T
 ¼ 1 2 3 4 5 6 ð9Þ

Using this modal coordinate transformation vector, the equation in modal coordinates can be written as

€ þ Cf _ þ Kf  ¼ Hf f ð10Þ
220 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

where T Ms  ¼ I and x€ g ¼ 0. The matrices Cf , Kf , and Hf are defined, respectively, as

Cf ¼ T Cs  ¼ diag½c11 , c22 , c33 , c44 , c55 , c66 ,

 
Kf ¼ T Ks  ¼ diag !211 , !222 , !233 , !244 , !255 , !266 , ð!11 5 !22 5 !33 5 !44 5 !55 5 !66 Þ, ð11Þ
T T
Hf ¼  Hs ¼ ½H11 , H22 , H33 , H44 , H55 , H66 

When structural control is considered, controlling the first two modes provide good results for the earthquake
hazard mitigation of structural systems and reduces the amplitudes. Therefore, the ROM is formed using the first
two modes of the full-order system. Thus, the ROM is defined in the form of

€ þ Cr _ þ Kr  ¼ Hr f ð12Þ
 
where Cr ¼ diag½c11 , c22 , Kr ¼ diag !211 , !222 , and Hr ¼ ½H11 , H22 T . In the last stage, the ROM must be retrans-
formed to physical coordinates for the control design. Consequently, the ROM can be written in physical coord-
inates as follows
 
Ar Br
Pr ¼ ð13Þ
Cr Dr
where
   
0 I 0  
Ar ¼ , Br ¼ , Cr ¼ y 0 , y ¼ Cy 12 ð14Þ
Kr Cr Hr

Controller design
If the controlled variable converges to a value determined by the stability condition of the system, it is termed
regulator type. In this study, a regulator-type control system is considered. Thus, there is no need to follow the
reference input while controlling the system. In control theory, the regulator problem is the design of the controller
such that norm of the closed-loop transfer function is minimized and the closed-loop system is stable.
The parametric uncertainties are defined as differences between real values and the parameters of the mathem-
atical model.27 Systems models affected by parametric uncertainties reduce the control system performance.28 The
robustness of the control strategies is studied by many different applications considering the sensitivity of the
control system to parametric errors and measurement noise added to the system.11,26–32 Many dynamic destructive
effects, such as high-frequency dynamics, which cannot be modeled and which form in different parts of the
system, can be evaluated in a single block, such as ‘‘.’’ These uncertainties are called ‘‘nonstructural uncertain-
ties.’’ In linear time-invariant systems,  represents an unknown transfer function. Differences between real
systems and dynamic models are modeled as follows

t ðsÞ ¼ Pf ðsÞ  Pr ðsÞ ð15Þ

The primary point of H1 control design is stability of the feedback control system against unknown additive
uncertainties. Considering the block diagram of the control system shown in Figure 2, the two primary transfer
functions in this control system are S(s) and T(s). S(s) is defined as the sensitivity transfer function, and T(s) is the
complementary sensitivity transfer function. Generally, T(s) is the primary transfer function for the stability of a
closed-loop system. When T(s) and t(s) are considered stable, using a WT filter and provided that the upper limit
of t(s) satisfies
   
t ðj!Þ  WT ðj!Þ, 8! ð16Þ

a norm WT condition o from to z2, the feedback system can be stable

jjWT TðsÞjj1 5 1 ð17Þ

Aggumus and Cetin 221

Figure 2. Augmented system structure.

The second aim of the H1 controller is to improve the performance of the feedback control system. The
problem regarding improving the response performance is how to attenuate the influence of the disturbance w
on the output y of the plant. This issue is related to minimization while subjected to the condition of stability of the
closed-loop system. The H1 norm condition can be written as
jjSðsÞjj1 ¼ sup ½SðsÞ ð18Þ
where WS is the filter for the system output. A small control gain is used in a high-frequency region for robust
stability. Thus, specifying the WS(s) and WT(s) filters in the control system satisfies both robust stability and
response performance. This type of H1 controller is called a mixed sensitivity problem and is defined as
 
 WS S 
 
 WT T  5  ð19Þ
1

where  is a design parameter that is positive.

Selection of frequency shaping filters

An important step in H1 control is to determine the frequency shape filters. Additive uncertainty is used to select
WT. A filter should cover the uncertainty to provide robust stability. Moreover, the objective of the control, which
is the control of the first two modes, should be considered. Furthermore, considering the spillover effect, the
controller should suppress the first two modes without exciting the truncated modes. In this manner, frequency
shape filters take the following form
2
s þ 2nm !nm s þ !2nm
WT ¼ kw , WS ¼ constant ð20Þ
s2 þ 2dm !dm s þ !2dm

where !nm is the frequency of the last controlled mode and !dm is the frequency of the first uncontrolled mode. In
particular, the WS filter is used to reduce the control system sensitivity included in the uncertainty effect and is
defined as a constant.

Controller application in semiactive system

The MR damper’s voltage should vary to ensure that the necessary power can be produced in the system. A
continuous state function is used to achieve these changes in voltage. The voltage of the MR damper is selected as
follows33,34

If Gðfc  fd Þsgnðfd Þ 4 Vmax , v ¼ Vmax

or Gðfc  fd Þsgnðfd Þ 4 Vmin , v ¼ Vmin ð21Þ
otherwise, v ¼ Gðfc  fd Þsgnðfd Þ
222 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

where Vmax is the maximum voltage in the MR damper, Vmin is the minimum voltage in the MR damper, fc is the
force necessary for the system and is determined by the controller, and fd is the force formed by the MR damper
and is measured by the system. Finally, G is the MR damper control gain.

Experimental performance analysis and results

Shaking table experiments are conducted in the Machine Theory, System Dynamics and Control Laboratory of
Yildiz Technical University to test the designed control algorithms. Experimental verification is completed using a
six-story steel frame structure, as shown in Figure 3. This type of steel frame structure model is generally used in
experimental studies in the literature. Each floor of the structural model is formed by connecting sheet metal using
screws. The weight of each floor is approximately 107.5 kg. Eight bars between floors, which represent columns,
are made of spring steel. Each bar is 3 mm thick and 50 mm wide. The distance between the floors is 250 mm, and
the floors and bars are connected using screws.
Sensors of different brands and types are used as the measurement equipment. A schematic view of the experi-
mental system (for MR12) is shown in Figure 4. A driving servo-motor system of the shaking table and the
application of the designed controllers in the system are conducted using a dSpace ACE Kit 1103. A 12-channel
NetdB signal analyzer is used to process the acceleration signals. In this experiment, the displacement and accel-
eration of each floor are measured. One Waycon-branded linear variable position sensor (LVDT), one Solartron
LVDT, one Burster LVDT, and three Opkon-branded resistive linear position sensors are used to measure the
displacement. Two Brüel&Kjaer accelerometers and four Dytran accelerometers are used to measure the acceler-
ation. Three types of RD-8041 MR dampers are installed in the building model, denoted as MR1, MR2, and MR12
and summarized in Table 1.
The force formed by the MR damper is measured using Brüel&Kjaer and Dytran force sensors. The processing
of the displacement and the force data is conducted using dSpace ACE Kit 1103. Two adjustable power generators
are used to supply the displacement sensors. One computer is used to data process the acceleration data, and
another computer is used to drive the shaking table and to control and process the data by communicating with
dSpace.

Figure 3. Experimental view of the structural model (MR12).

MR: magnetorheological.
Aggumus and Cetin 223

Figure 4. Schema of experimental system (MR12).

MR: magnetorheological.

Table 1. Abbreviations of MR dampers and placement types.

Abbreviation Type of connection (arrangement)

MR1 MR damper is placed between the first floor and the ground
MR2 MR damper is placed between the second floor and the ground
MR12 MR dampers on both the first and second floors
MR: magnetorheological.

Determining the model parameters of the experimental system

The parameters are determined using empirical expressions
 in structural dynamics. The mass value of system for
each floor is 107.5, and the mass matrix is Mf ¼ diag 107:5 107:5 107:5 107:5 107:5 107:5 . Considering
the connection with eight bars made of spring steel, the stiffness of the system for each floor is evaluated as
k1,2,3,4,5,6 ¼ 8*(12EI/l3) ¼ 145,152 N/m.4 According to the Rayleigh damping principle, if 0 ¼ 0.0265 and
0 ¼ 0.00011431, [Cs] ¼ 0[Ms] þ 0[Ks], the damping coefficient is C1,2,3,4,5,6 ¼ 16.59 N s/m, and the controller
gain G is 0.04 for MR1, 0.006 for MR2, and 0.009 for MR12. The maximum voltage is 2 Vmax, and the minimum
voltage Vmin is 0 V.

Application of the H1 controller

The H1 controller designed for a 6-DOF semiactive structural system, as described in ‘‘Robust control design’’
section, is applied to the structural models in Figure 1. A 0.5-scale Erzincan earthquake is considered as the seismic
excitation (see Figure 5).
Three different installation layouts (connection types) are studied. Abbreviations for these scenarios are shown
in Table 1. The output vector in equation (4) of the displacement feedback on the first floor for MR1 is
Cy1 ¼ ½ 1 0 0 0 0 0  and that for MR2 is Cy2 ¼ ½ 0 1 0 0 0 0 . Cy1 and Cy2 are applied separately
for MR12.
The last frequency values to be controlled and the first frequency values, which are not controlled, are used to
adjust the frequencies of the WT filter in equation (20). The last mode frequency to be controlled in the numerator
is considered as the next mode frequency, which is not controlled in the denominator. The aim is that control gain
is acquired in the frequency area to control the first two modes to sustain robust stability and such that it does not
simulate any other modes, which are not controlled. The coefficients of the controllers are kw ¼ 0.4, xnm ¼ 0.65,
xdm ¼ 0.15, !nm ¼ 43, and !dm ¼ 66. The H1 controller is applied to the experimental structural model using the
MR damper. The first two modes are the reduced forms of the 6-DOF systems for the H1 controller. The ROM
and FOM are illustrated in Figure 6, which shows that the first two modes of both models match. Therefore, the
224 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

Figure 5. East–West component of the Erzincan earthquake.

Figure 6. FOM and ROM: (a) MR1 and (b) MR2.

FOM: full-order model; MR: magnetorheological. ROM: reduced-order model.

first two modes of the ROM present the same characteristics as the first two modes of the FOM. The H1 control is
designed using MATLAB. The frequency response of the H1 control is shown in Figure 7. To suppress the low-
frequency region, the controller has the highest gain around the range of the first two modes. Accordingly, to avoid
the spillover effect, the lowest gain is provided around the truncated modes, in which the control effect is not
expected. Finally, the open- and closed-loop responses of both the FOM and ROM are shown in Figure 8. The
figure shows that the controllers achieve good performance and reduce the frequency range of the controlled
modes without exciting the truncated modes. The MR1 and MR2 controllers are used together for the MR12
controller.
Here, the amplitude values indicate the maximum value of the vibration mode of the system for the passive and
controller cases.
Aggumus and Cetin 225

Figure 7. Controller frequency response: (a) MR1 and (b) MR2.

MR: magnetorheological.

Figure 8. Open- and closed-loop responses of the system: (a) MR1 and (b) MR2.
MR: magnetorheological.
226 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

Figure 9. Displacements of all floors for the combinations of the MR damper layout: (a) MR1, (b) MR2, and (c) MR12.
MR: magnetorheological.
Aggumus and Cetin 227

Figure 9. Continued.

Displacement responses
The time responses for the situations of all combinations of the MR damper’s layout (MR1, MR2, and MR12)
are shown in Figures 9 to 12. The displacements of the MR1, MR2, and MR12 situations are compared in
Figures 9 and 11. The acceleration responses are illustrated in Figures 10 and 12. The passive (MR damper
disconnected), passive with MR damper (MR damper connected), and H1 controlled situations are compared
to determine the performance of the controller. The cases in which the MR damper is only on the first floor
(MR1), on the second floor (MR2), and on both the first and second floors (MR12) are examined. The per-
formance of the controller is compared with the cases in which the MR damper is not connected and the MR
damper is connected but uncontrolled. ‘‘MR1’’ denotes that the MR damper is connected to the first floor and
that there is no voltage. The case in which the MR damper is connected and controlled case is denoted as ‘‘MR1
(H1).’’ The ‘‘MR2 (H1)’’ and ‘‘MR12 (H1)’’ represent for the situations in which the MR damper is on the
second floor and there is no voltage and in which the MR damper is on second floor and controlled, respect-
ively. ‘‘MR12’’ and ‘‘MR12 (H1)’’ are typified in the same manner. The displacement responses of all floors are
shown in Figure 9. Comparisons of the passive, uncontrolled, and controlled cases for all combinations of the
MR damper’s layout are depicted in Figure 9(a) to (c), respectively. The connection of the MR damper reduces
the vibration of each floor. The amplitudes of the displacement in both situations corresponding to the passive
MR and the MR with the controller are lower than the displacement amplitude in the situation corresponding
to the passive system without the MR damper. Furthermore, implementation of the robust controller improves
the vibration reduction performance of the MR damper. The MR damper with the robust controllers is better
than the MR-uncontrolled cases. The acceleration responses of each floor for all combinations of the MR
damper layout are shown in Figure 10. The connection of the MR damper and the application of the robust
controller improve the acceleration responses of each floor. As shown in Figures 11 and 12, the controlled cases
for all situations are compared. The maximum displacement responses are compared in Figure 11. The best
performance is achieved by MR12 (H1), in which an MR damper is on both the first and second floors and is
commanded by the H1 controller. Figure 12 compares the maximum acceleration responses. MR12 (H1) again
exhibits the best performance. Generally, implementation of the controller reduces the maximum absolute values
of the accelerations. However, some of the maximum absolute values for the controlled cases are greater than
the MR damper-passive cases.
228 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

Figure 10. Accelerations of all floors for the combinations of the MR damper layout: (a) MR1, (b) MR2, and (c) MR12.
MR: magnetorheological.
Aggumus and Cetin 229

Figure 10. Continued.

Figure 11. Comparison of the maximum displacements of all floors for the controlled cases (MR1(H1), MR2(H1), and MR12(H1)).
MR: magnetorheological.

Figure 12. Comparison of the maximum accelerations of all floors for the controlled cases (MR1(H1), MR2(H1), and MR12(H1)).
MR: magnetorheological.
230 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

Assessment of the structural vibration performance

To evaluate the test results in detail, the performance indices formed by Ohtori et al.35 are used as follows4
8 9 8  9
jdi ðtÞj
<max hi >
> = <maxx€ i ðtÞ=
t,i t,i
J1 ¼ max , J2 ¼ max
>
: 
max >
; : x€ max
i ;
8 9 8  9 ð22Þ
<maxx€ i ðtÞ=
kd k
<max hii =
t,i t,i
J3 ¼ max , J4 ¼ max  
: kmax k ; : x€ max
i
 ;

max ¼ max j
d i ðtÞj
hi : The maximum interstory drift ratio of the uncontrolled structure.
hi: distance between floors,
di: displacement between floors,
x€ max
i ðtÞ: absolute acceleration without the controller.
In the performance indices, the passive cases with the controller and MR damper are proportional to the passive
case without the MR damper. The performance index values obtained by the experimental data are shown in
Table 2. All of the performance index values are less than 1. Thus, the system recovered as a result of the MR
damper’s control and connection to system. The amplitudes of the system in both situations (MR passive, MR
with controller) are less than the amplitude in the passive situation without the MR damper. The first performance
index compares the maximum values of the amplitudes between floors. The maximum value of the vibration
amplitude, as a result of the connection of the MR damper to the system according to this performance standard,
is shown in Table 2. The J1 performance standard is less than 1 for the MR1, MR2, and MR12 cases.
Thus, the system amplitude decreases for a passive MR damper in the system. When the passive cases with the
MR damper are compared, the best performance is obtained for the MR12 case in which the MR damper is placed
on both the first and second floors. When the passive situations and situations with the controller are compared,
situations with the controller decrease the vibration amplitudes more effectively than in the passive cases.
Similarly, when the three connection types are compared, the best performance is acquired for the MR12 situation,
as shown in Table 2. The second performance index is based on the maximum acceleration value. The J2 per-
formance index is less than 1 for the MR1, MR2, and MR12 situations. Thus, for the passive MR damper system,
the acceleration of the system also decreases. When we compare the passive situations, the best performance is
achieved for the MR12 situation in which MR damper is placed on both the first and second floors. When the
situations with the controller and passive situations are compared, MR12 has the best performance. Although
situations with the controller compress the displacement of vibrations, the acceleration response worsens in situ-
ations with the controller. This occurs because the controlled voltage while the MR damper is functioning occa-
sionally locks the MR damper, and thus, the acceleration values of the floor on which the MR damper is placed
increase. The third performance index is based on the maximum value of the displacement norm. Again, the best
performance index is achieved by MR12 (H1). The fourth performance is based on the maximum value of the
acceleration norm. The performances of MR12 (H1) and MR2 (H1) recovered. The best performance is achieved
by MR12 (H1). As this performance index involves all acceleration values, the fact that the values of MR2 (H1)
and MR12 (H1) are better than the passive values indicates that the controllers decrease the total acceleration
values even though they fail regarding the maximum acceleration values.

Table 2. Assessment of the controller according to the performance indices.

Abbreviation J1 J2 J3 J4

MR1 0.7516 0.7964 0.6405 0.8517

MR1 (H1) 0.6615 0.8896 0.3766 0.9514
MR2 0.5815 0.6672 0.3277 0.6649
MR2 (H1) 0.4264 0.8125 0.2557 0.5393
MR12 0.5320 0.5083 0.2855 0.5436
MR12(H1) 0.3488 0.6018 0.2250 0.4538
MR: magnetorheological.
Aggumus and Cetin 231

Figure 13. Maximum displacement values of the floors.

Figure 14. RMS displacement values of the floors.

RMS: Root mean square.

Table 3. Absolute maximum acceleration values for all floors.

Abbreviation First floor Second floor Third floor Fourth floor Fifth floor Sixth floor

Passive 0.6363 0.8193 0.7545 0.6887 0.6978 0.9118

MR1 0.3356 0.3627 0.6012 0.5719 0.5438 0.7262
MR1 (H1) 0.3597 0.3506 0.5266 0.5993 0.6283 0.8111
MR2 0.3304 0.3235 0.5950 0.5204 0.6084 0.5668
MR2 (H1) 0.3114 0.3059 0.4669 0.4281 0.4958 0.7408
MR12 0.3367 0.3191 0.4106 0.3805 0.4252 0.4635
MR12 (H1) 0.3339 0.3175 0.3646 0.3775 0.4179 0.5487
MR: magnetorheological.

Table 4. RMS values for all floors.

Abbreviation First floor Second floor Third floor Fourth floor Fifth floor Sixth floor

Passive 0.0090 0.0185 0.0252 0.0299 0.0328 0.0341

MR1 0.0054 0.0113 0.0153 0.0183 0.0200 0.0208
MR1 (H1) 0.0034 0.0066 0.0087 0.0104 0.0113 0.0118
MR2 0.0028 0.0058 0.0079 0.0095 0.0104 0.0108
MR2 (H1) 0.0021 0.0044 0.0058 0.0072 0.0079 0.0083
MR12 0.0023 0.0048 0.0067 0.0081 0.0089 0.0093
MR12 (H1) 0.0019 0.0038 0.0050 0.0062 0.0070 0.0074
MR: magnetorheological; RMS: Root mean square.
232 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

The maximum displacement values of each floor are shown in Figure 13. The passive case, which corresponds to
when the MR damper is not connected to the structure, performs worse than the other cases. The displacement
values of each floor for the passive case are greater than both the uncontrolled and controlled cases. The connec-
tion of the MR damper to the structure reduces the displacement of all floors. In addition, all of the H1-controlled
cases perform well compared to cases when the MR damper is connected but not controlled. For example, the
maximum displacement values of each floor in the MR1 (H1) case are smaller than the MR1 uncontrolled case.
Similarly, the application of the robust controller reduces the maximum displacement values in both the MR2 and
MR12 cases. The RMS values of each floor are shown in Figure 14. The RMS values of the passive case are larger
than those of the other cases. Again, connecting the MR damper reduces the RMS values. In addition, the
MR1(H1), MR2(H1), and MR12(H1) cases perform well in terms of reducing the vibrations on each floor
compared to the uncontrolled cases. Thus, the H1 controller exhibits good performance in reducing the displace-
ment of all floors, and it exhibits the best performance for the MR12 situation. The maximum displacement values
shown in Figure 13 and the RMS values shown in Figure 14 verify the above summarized performance analysis.
The absolute maximum values of the accelerations are given in Table 3. The absolute maximum values of the
accelerations for all combinations of the MR dampers are lower than those in the passive case. The RMS values
are shown in Table 4. All of the absolute maximum values of the accelerations for all combinations of the MR
dampers are lower than for the passive case. In general, implementation of the controller reduces the maximum
absolute values of the accelerations. However, some of the maximum absolute values for the controlled cases are
greater than the MR damper-passive cases.

Conclusions
In this study, a H1 robust controller is designed to command MR damper voltage by placing an MR damper on
different floors to reduce building vibrations during earthquakes. The designed controller is experimentally tested
in the laboratory using a shaking table and a six-story structural model. The performances of the controller and
MR damper are investigated. Comparisons are made among cases when the MR damper is not connected, passive,
and controlled with H1. Additionally, we study how the building reacts when the MR damper is placed in various
combinations. The evaluations are based on the displacement and acceleration responses and the performance
indices. Moreover, the maximum displacement values and RMS values of each story are considered. When the
passive damper situations are considered, the structural system amplitudes decreased compared to the case without
the MR damper for all three combinations. The arrangement of the MR damper in which one end is connected to
the ground is effective in reducing the vibration amplitudes. In addition, the MR damper connected between the
second floor and the ground is more effective than the damper connected between the first floor and the ground.
However, the best combination of MR dampers is the connection between both the first floor and the ground and
between the second floor and the ground. Furthermore, the designed controller clearly improves the system
performance. The best performance is acquired for the MR12 (H1) situation, in which an MR damper is located
on both the first and second floors and the dampers are commanded by the H1 controller. In one MR damper
case, the MR damper arrangement which is connected between ground and second floor reduces the vibration
amplitudes more than the MR damper arrangement which is connected between ground and first floor. In future
studies, the case for which the MR damper is connected to the system between the first and second floors will be
examined and compared to the previous cases.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this
article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:
This study was supported by The Scientific and Technological research council of Turkey (TUBITAK) under grant no. 213M664.

References
1. Atabay E and Ozkol I. Application of a magnetorheological damper modeled using the current-dependent Bouc–Wen
model for shimmy suppression in a torsional nose landing gear with and without freeplay. J Vib Control 2014; 20:
1622–1644.
Aggumus and Cetin 233

2. Castão KAL, Goes LCS and Balthazar JM. A note on the attenuation of the Sommerfeld effect of a non-ideal system
taking into account a MR damper and the complete model of a DC motor. J Vib Control 2011; 17: 1112–1118.
3. Cornejo C and Alvarez-Icaza L. Passivity based control of under-actuated mechanical systems with nonlinear dynamic
friction. J Vib Control 2012; 18: 1025–1042.
4. Cetin S, Zergeroglu E, Sivrioglu S, et al. A new semiactive nonlinear adaptive controller for structures using MR damper:
design and experimental validation. Nonlinear Dyn 2011; 66: 731–743.
5. Paksoy M, Guclu R and Cetin S. Semiactive self-tuning fuzzy logic control of full vehicle model with MR damper. Adv
Mech Eng 2014; 6: 816813.
6. Song X and Ahmadian M. Characterization of semi-active control system dynamics with magneto-rheological suspensions.
J Vib Control 2010; 16: 1439–1463.
7. Tusset AM, Balthazar JM, Chavarette FR, et al. On energy transfer phenomena, in a nonlinear ideal and non-ideal
essential vibrating systems, coupled to a (MR) magneto-rheological damper. Nonlinear Dyn 2012; 69: 1859–1880.
8. Yildiz AS, Sivrioğlu S, Zergeroğlu E, et al. Nonlinear adaptive control of semi-active MR damper suspension with
uncertainties in model parameters. Nonlinear Dyn 2015; 79: 2753–2766.
9. Malekshahi A, Mirzaei M and Aghasizade S. Non-linear predictive control of multi-input multi-output vehicle suspension
system. J Low Freq Noise Vib Active Control 2015; 34: 87–106.
10. Xia Z, Wang X, Hou J, et al. Non-linear dynamic analysis of double-layer semi-active vibration isolation systems using
revised Bingham model. J Low Freq Noise Vib Active Control 2016; 35: 17–24.
11. Tusset AM, Piccirillo V, Balthazar JM, et al. On suppression of chaotic motions of a portal frame structure under non-
ideal loading using a magneto-rheological damper. J Theor Appl Mech 2015; 53: 653–664.
12. Dyke SJ, Spencer BF Jr, Sain MK, et al. Modeling and control of magnetorheological dampers for seismic response
reduction. Smart Mater Struct 1996; 5: 565–575.
13. Dyke SJ, Spencer BF Jr, Sain MK, et al. An experimental study of MR dampers for seismic protection. Smart Mater Struct
1998; 7: 693–703.
14. Jansen LM and Dyke SJ. Semiactive control strategies for MR dampers: comparative study. J Eng Mech 2000; 126:
795–803.
15. Yi F, Dyke SJ, Caicedo JM, et al. Experimental verification of multi input seismic control strategies for smart dampers.
J Eng Mech 2001; 127: 1152–1164.
16. Aldemir U. Optimal control of structures with semiactive-tune mass dampers. J Sound Vib 2003; 266: 847–874.
17. Cho SW, Kim BW, Jung HJ, et al. Implementation of modal control for seismically excited structures using magnetor-
heological dampers. J Eng Mech 2005; 131: 177–184.
18. Aldemir U and Gavin HP. Optimal semiactive control of structures with isolated base. Int Appl Mech 2006; 42:
235–240.
19. Kim Y, Langari R and Hurlebaus S. Semiactive nonlinear control of a building with a magnetorheological damper system.
Mech Syst Signal Process 2009; 23: 300–315.
20. Lu KC, Loh CH, Yang JN, et al. Decentralized sliding mode control of a building using MR dampers. Smart Mater Struct
2008; 17: 055006.
21. Swartz RA and Lynch JP. Strategic network utilization in a wireless structural control system for seismically excited
structures. J Struct Eng 2009; 135: 597–608.
22. Aldemir U. Causal semiactive control of seismic response. J Sound Vib 2009; 322: 665–673.
23. Bitaraf M, Ozbulut OE, Hurlebaus S, et al. Application of semi-active control strategies for seismic protection of buildings
with MR dampers. Eng Struct 2010; 32: 3040–3047.
24. Heo G and Jeon J. Application of unified Lyapunov control algorithm and mathematical demonstration for structure
control. KSCE J Civil Eng 2011; 15: 479–486.
25. Rahbari NM, Azar BF, Talatahari S, et al. Semi-active direct control method for seismic alleviation of structures using MR
dampers. Struct Control Health Monit 2013; 20: 1021–1042.
26. Sivrioglu S, Tanaka N and Yuksek I. Acoustic power suppression of a panel structure using H1 output feedback control. J
Sound Vib 2002; 249: 885–897.
27. Peruzzi NJ, Chavarette FR, Balthazar JM, et al. The dynamic behavior of a parametrically excited time-periodic MEMS
taking into account parametric errors. J Vib Control 2016; 22: 4101–4110.
28. Nozaki R, Balthazar JM, Tusset AM, et al. Nonlinear control system applied to atomic force microscope including
parametric errors. J Control Automat Electr Syst 2013; 24: 223–231.
29. Gudarzi M. -Synthesis controller design for seismic alleviation of structures with parametric uncertainties. J Low Freq
Noise Vib Active Control 2015; 34: 491–511.
30. Balthazar JM, Bassinello DG, Tusset AM, et al. Nonlinear control in an electromechanical transducer with chaotic
behaviour. Meccanica 2014; 49: 1859–1867.
31. Tusset AM, Balthazar JM, Bassinello DG, et al. Statements on chaos control designs, including a fractional order dynam-
ical system, applied to a ‘‘MEMS’’ comb-drive actuator. Nonlinear Dyn 2012; 69: 1837–1857.
32. Cetin S, Sivrioglu S, Zergeroglu E, et al. Semi-active H robust control of six degree of freedom structural system using MR
damper. Turkish J Electr Eng Comput Sci 2011; 19: 797–805.
234 Journal of Low Frequency Noise, Vibration and Active Control 37(2)

33. El-Kafafy M, El-Demerdash SM and Rabeih AAM. Automotive ride comfort control using MR fluid damper. Engineering
2012; 4: 179–187.
34. Lam AHF and Lio WH. Semi-active control of automotive suspension systems with magnetorheological dampers. Int J
Veh Des 2002; 33: 50–75.
35. Ohtori Y, Christenson RE, Spencer BF Jr, et al. Benchmark control problem for seismically excited nonlinear buildings. J
Eng Mech 2004; 130: 366–385.

Appendix
Notation

Cs stiffness matrix of the structural system

Cy1, Cy2 the locations of the measurements of the system
di displacement between floors
E Young’s modulus
fC the force is measured by the system
fd the force necessary for the system
f(t) damping force of the MR damper
hi distance between floors
H the placement of the control units
I cross-sectional moment of inertia
Jn the performance indices of the system
Ks damping matrix of the structural system
l the length of the column
L the seismic input vector
MR1 MR damper is placed between the first floor and the ground
MR12 MR dampers on both the first and second floors
MR2 MR damper is placed between the second floor and the ground
Ms mass matrix of the structural system
Pf (s) the system matrix
Pr (s) the reduced-order model of the system
S(s) the sensitivity transfer function
T(s) the complementary sensitivity transfer function
Vmax the maximum voltage in the MR damper
Vmin the minimum voltage in the MR damper
WT, WM filters
x€ ðtÞ acceleration vector
xðtÞ displacement vector
x€ g ðtÞ the earthquake ground acceleration
x_ ðtÞ velocity vector
x€ max
i (t) absolute acceleration without the controller
0 , 0 Rayleigh damping coefficients
 design parameter
t ðtÞ the uncertainty
 the modal space of the system
 the maximum singular value of S(s)
!dm the frequency of the first uncontrolled mode
!nm the frequency of the last controlled mode