Journal of Sound and Vibration 390 (2017) 2338

Contents lists available at ScienceDirect

Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsvi

Structural control and health monitoring of building

structures with unknown ground excitations: Experimental
investigation
Jia He a,b,n, You-Lin Xu a, Sheng Zhan a, Qin Huang c
a
Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hong Kong, China
b
College of Civil Engineering, Hunan University, Changsha, Hunan, China
c
Shenzhen Graduate School, Harbin Institute of Technology, Shenzhen, China

a r t i c l e i n f o abstract

Article history: When health monitoring system and vibration control system both are required for a
Received 1 March 2016 building structure, it will be beneficial and cost-effective to integrate these two systems
Received in revised form together for creating a smart building structure. Recently, on the basis of extended Kalman
23 November 2016
filter (EKF), a time-domain integrated approach was proposed for the identification of
Accepted 25 November 2016
structural parameters of the controlled buildings with unknown ground excitations. The
Handling Editor: L.G. Tham
Available online 1 December 2016 identified physical parameters and structural state vectors were then utilized to de-
termine the control force for vibration suppression. In this paper, the possibility of es-
Keywords: tablishing such a smart building structure with the function of simultaneous damage
Building structures
detection and vibration suppression was explored experimentally. A five-story shear
Ground excitations
building structure equipped with three magneto-rheological (MR) dampers was built.
Vibration control
Health monitoring Four additional columns were added to the building model, and several damage scenarios
Experimental investigation were then simulated by symmetrically cutting off these columns in certain stories. Two
MR dampers sets of earthquakes, i.e. Kobe earthquake and Northridge earthquake, were considered as
seismic input and assumed to be unknown during the tests. The structural parameters and
the unknown ground excitations were identified during the tests by using the proposed
identification method with the measured control forces. Based on the identified structural
parameters and system states, a switching control law was employed to adjust the current
applied to the MR dampers for the purpose of vibration attenuation. The experimental
results show that the presented approach is capable of satisfactorily identifying structural
damages and unknown excitations on one hand and significantly mitigating the structural
vibration on the other hand.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Research on vibration control and health monitoring of civil structures has been actively conducted in the past decades,
and many state-of-the-art reviews on these two areas are available [14]. A number of experiments have also been carried
out for investigating the efficiency of control systems or validating the accuracy of damage detection algorithms. In

n
Corresponding author at: College of Civil Engineering, Hunan University, Changsha, Hunan, China
E-mail address: hj19830507@126.com (J. He).

http://dx.doi.org/10.1016/j.jsv.2016.11.035
0022-460X/& 2016 Elsevier Ltd. All rights reserved.
24 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

consideration that the MR dampers are employed for vibration control and the EKF-based approach is used for damage
detection in this experimental study, some experiments focusing on the investigation of the above two aspects are briefly
introduced herein. For example, on the usage of clipped-optimal control strategy, a series of experiments was conducted by
Dyke et al. [5] to investigate the efficiency of MR dampers for seismic response reduction. With two MR dampers installed
on the first and second floor of a six-story building structure, the experimental studies were conducted to demonstrate the
capabilities of multiple MR devices for seismic control of civil structures [6]. By using a logic control algorithm to adjust the
properties of multiple MR dampers, an experimental investigation was carried out by Chen et al. [7] for seismic response
control of a complex building structure. For the damage detection on the basis of EKF, Zhou et al. [8] presented experimental
studies to validate the reliability of the adaptive EKF approach [9] for online identification of structural damage in a small-
scale three-story building structure. Moreover, the capability of various adaptive approaches, including adaptive least-
square estimation (ALSE), adaptive EKF, quadratic sum-squares error (QSSE), and the sequential non-linear least-square
estimation (SNLSE), in tracking structural damage were compared and demonstrated using experimental data [10]. Ex-
perimental studies were also performed by Yin et al. [11] for the verification of the capability of the adaptive EKF approach
for identifying and tracking damages in nonlinear building structures. By using the mutual information as a measure for
nonlinear signal cross correlation, a vibration-based health monitoring technique was recently developed by Trendafilova
et al. [12] and experimentally validated via a composite beam.
However, in most previous investigations, vibration control system and health monitoring system are treated separately
according to their individual primary objective pursued. Since sensors, data acquisition, transmission and processing devices
are required in both vibration control system and health monitoring system, integrating both systems together will be cost-
effective by sharing the same hardware devices and beneficial for creating a smart building structure where the function of
damage detection and vibration control can be simultaneously fulfilled. In recent years, increasing attentions have been paid
on the development of such smart civil structures. For example, with the aid of semi-active friction dampers, an integrated
system was developed for the purpose of vibration control and health monitoring on the basis of the variations of natural
frequencies and mode shapes [13,14]. Chen et al. [15] also proposed an integrated method in the time-domain for damage
detection and seismic response control of building structures equipped with semi-active friction dampers. More recently,
based on the adaptive structural control law, Lin et al. [16] presented a hybrid health monitoring system by combining the
global damage detection method with local damage identification method using infrared imaging technique. An integrated
semi-active control and health monitoring method was developed by Karami and Amini [17] to improve the control per-
formance on one hand, and to reduce the damage of frame structure caused by strong seismic excitations on the other hand.
Lei et al. [18] proposed an integrated algorithm for decentralized vibration control of tall shear building structures and for
identification of unknown earthquake-induced ground motion. On the basis of EKF and error tracking techniques, an on-line
integration technique was also presented by Lei et al. [19] for health monitoring and active control of the undamaged/
damaged structures. Amini et al. [20] proposed an integrated damage detection and semi-active control algorithm to locate
and characterize the damage in base-isolated structures and to suppress the excessive vibration of buildings caused by
seismic excitation. Based on recursive LSE, Xu et al. [21] proposed a real-time integrated procedure for accurately identifying
time-varying structural parameters and unknown excitations, as well as optimally mitigate excessive vibration of building
structure.
Although much progress has been made on the integration of vibration control system and health monitoring system,
the effectiveness of most aforementioned methodologies are examined through numerical examples. The experimental
investigation of integrated structural vibration control and health monitoring systems is limited. By using smart tuned mass
dampers, a time-frequency algorithm for system identification and response control of MDOF systems was developed and
experimentally investigated by Nagarajaiah [22]. Yang et al. [23] presented a hybrid real-time health monitoring and control
system for the building structures under seismic excitations and experimentally validated via a three-story aluminum frame
structure. Based on frequency response functions, a new method for system identification and damage detection of the
controlled buildings equipped with semi-active friction dampers was proposed and experimentally validated via a complex
building structure with a twelve-story main building and a three-story podium structure [24,25]. The authors [26] recently
proposed an integrated approach for simultaneously identifying structural damage and suppressing excessive vibration of
building structure, and verified the reliability and accuracy of the proposed approach through a numerical example. A
significant difference between the proposed approach and other existing EKF-UI (Extended Kalman Filter with Unknown
Input) based approaches is that the damage detection and vibration control are integrated and both are considered col-
lectively in the proposed approach while other EKF-UI approaches consider damage detection (system identification) only.
In the proposed approach, the control forces can be used not only for attenuating the structural vibration but also for aiding
the identification of structural parameters and excitations. Moreover, the projection matrix is presented and used to
transform the multiple regression equation into a simple regression equation. The closest solution of the unknown ex-
citations is obtained via the least-square estimation.
In this study, the possibility of establishing such a smart building structure was explored experimentally. First, the time-
domain integrated vibration control and health monitoring approach was briefly introduced. Then, a five-story shear-type
building structure equipped with three MR dampers was designed and introduced. Four additional columns were added to
the building model and several damage scenarios were then simulated by symmetrically cutting off these columns in certain
stories. The static tests and hammer tests were carried out to determine the dynamic properties of the building model when
the MR dampers were removed. Finally, the shaking table tests of the five-story building model equipped with MR dampers
 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 25

under Kobe earthquake and Northridge earthquake were carried out. The structural parameters and unknown ground
excitations were identified online. Based on the identified structural parameters and system states, a switching control law
was employed to adjust the current applied to the MR dampers for vibration mitigation. The experimental results show that
the proposed approach is capable of satisfactorily identifying structural damages and unknown ground motions on one
hand and significantly mitigating the structural vibration on the other hand. The performance of the integrated system for
vibration control and damage detection was investigated in detail.

2. The integrated vibration control and health monitoring approach

2.1. Identification of the structural parameters and unknown excitations

The equation of motion of a controlled building structure can be expressed as,

Mx (t ) + Cx (t ) + Kx (t ) = *f *(t ) + f (t ) (1)

where x (t ), x (t ), and x(t) are structural acceleration, velocity and displacement responses, respectively; M, C, and K are the
mass, damping and stiffness matrices, respectively; f(t) is the ground excitation in this study and is assumed to be unknown;
f *(t ) is the control force vector and it is assumed to be measurable in this study; and * are the influence matrices
associated with f(t) and f*(t), respectively.
By introducing as the unknown structural stiffness and damping coefficients and Z (t ) = x (t )T x (t )T T as the ex-
T

tended system state vector, the equation for the extended state vector of the structure can be derived from Eq. (1) as
x (t )

Z (t ) = M ( Cx (t ) Kx (t ) + *f *(t ) + f (t )) = g (Z (t ), f (t ), f *(t ), t )
1

0 (2)

In this study, the structural mass matrix M, the control force f *(t ) provided by MR dampers, and the structural accel-
eration x (t ) can be measured and assumed to be known. Therefore, the discretized observation equation at the time instant
tk kt with t being the time interval could be described as
yk = Kx k + Cx k fk + vk = h (Zk ) fk + vk (3)

in which yk = Mx k + *f *k is the measurement vector; vk is a measurement noise vector assumed to be a Gaussian white
noise vector with zero mean and a covariance matrix E [vk vTj ] = R k kj , where kj is the Kroneker delta.
^ and ^ be the estimates of Z and f , respectively. The priori estimation state ^
Let Zk|k fk | k k k Zk + 1 | k could be determined as,
(k + 1) t ^
^ ^ ^
Zk + 1 | k = Zk | k + kt g (Zk | k , fk | k, f *k , t ) dt
(4)

The priori estimation error covariance can be calculated as,

Pk + 1 | k = A1Pk | k A1T + A2R k A2T (5)

in which
0 0

A1 = I + t Uk | k + t M1(I ) Hk | k ; A2 = 1
t M (I )

0 0 (6)
g (Zk , fk, f *k , kt ) h (Zk )
Uk | k = ; Hk | k = ; = In (T )1T
Zk ^ ^
Z = Z k | k, fk = fk | k
Zk ^
Zk = Zk | k (7)

With proper transformation, the posteriori estimation state can be obtained,

^ ^ ^
Zk + 1 | k + 1 = Zk + 1 | k + Gk + 1 yk + 1 h (Zk + 1 | k )
(8)

where Gk 1 is the EKF gain matrix at the time instant tk 1 (k 1)t, and can be given as
1
(
Gk + 1 = Pk + 1 | k HTk + 1 | k (Hk + 1 | k Pk + 1 | k HTk + 1 | k + R k + 1) ) (9)

The posteriori estimation error covariance can be found as

Pk + 1 | k + 1 = (I Gk + 1Hk + 1 | k ) Pk + 1 | k (I Gk + 1Hk + 1 | k )T + Gk + 1R k + 1GTk + 1 (10)

By assuming that the number of observed degrees of freedom is larger than the number of excitations, the unknown
26 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

Fig. 1. Flowchart of the presented algorithm for identifying structural parameters and unknown excitations.

excitations could be estimated basing on Eq. (3) and the estimated state vector,
^ ^
fk + 1 | k + 1 = (T )1T [ yk + 1 + h (Zk + 1 | k + 1)] (11)

The proposed algorithm can identify the unknown external excitations and structural parameters simultaneously, and
the flowchart of the proposed algorithm for the parameter and excitation identification is shown in Fig. 1.

2.2. Structural vibration control

In this study, structural vibration control is also taken into consideration for the building structures together with the
identification. An MR damper, which is a kind of semi-active control devices, is employed herein for vibration attenuation. A
variety of semi-active control strategies for the adjustment of the properties of MR dampers for the purpose of seismic
control have been proposed and investigated [2731]. The switching control algorithm is considered herein and can be
expressed as follows [31],

cd v + f
max
sgn (v) ( uv < 0)
d
ud =
min
cd v + f d sgn (v) ( uv 0) (12)

where ud is the MR damper force; cd is damping coefficient determined by the viscosity of MR fluid; and re- f dmax f dmin
spectively denote the maximum value and minimum value of the frictional forces which are related to the yielding shear
stress and can be achieved by adjusting the applied field; v is the relative velocity between the two adjacent floors with the
damper installed in this story; and u is the optimal control force determined by linear quadratic Gaussian (LQG). If the
optimal control force and the structural motion are in opposite direction, the maximum level of the current is applied to
obtain the maximum frictional force f dmax for the purpose of mitigating vibration as much as possible. Otherwise, the
commanded current is set to zero and the minimum frictional force f dmin is achieved.
Since the structural parameters and system states both have been identified in Section 2.1, they can be used herein for
determining the optimal control force u in the LQG algorithm to make sure control effectiveness.

u = R 1BT P {X} (13)

in which P is the solution of the following classical Riccati equation.

T
PBT R 1BP A P PA Q = 0 (14)

where Q is the positive semi-definite matrix associated with system states; R is the positive definite matrix associated with
the control forces; the vector X and the matrices A and B can be determined as follows,
 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 27

Fig. 2. Flowchart of the semi-active control algorithm for MR damper.

x id 0 I 0
X = ; A = ; B = 1
x id M1K
id M 1C
id M * (15)

where xid and x id are the estimated structural displacement and velocity, respectively; and Kid and Cid are the identified
stiffness and damping matrix, respectively;
The flowchart of the proposed control law is shown in Fig. 2.

2.3. The procedures for implementing the proposed integrated approach

Based on the equations presented in the previous two subsections, the integrated algorithm for real-time system
identification and vibration control of the building structure can be implemented step by step as follows:

Step 1: Calculate the observation equation basing on the measured acceleration responses and MR damper forces;
^
Step 2: Determine the priori estimation state Zk + 1 | k according to Eq. (4);
Step 3: Determine the priori estimation error covariance Pk + 1 | k using Eq. (5);
Step 4: Obtain the gain matrix Gk 1 according to Eq. (9);
Step 5: Obtain the posteriori estimation state Z ^
k + 1 | k + 1 according to Eq. (8);
Step 6: Update the posteriori estimation error covariance Pk + 1 | k + 1 by Eq. (10);
^
Step 7: Estimate the unknown excitations fk + 1 | k + 1 according to Eq. (11);
^
Step 8: Form the coefficient matrices according to Eq. (15) with the aid of the estimated state Zk + 1 | k + 1 determined in Step
5;
Step 9: Calculate P by solving Eq. (14), and determine the optimal control force according to Eq. (13);
Step 10: Determine the MR damper force according to Eq. (12).

It is clear that in the presented integrated approach, the measured control forces are employed for the identification of
structural parameters and unknown excitations on one hand, and the identified physical parameters and the estimated
structural responses are used in the control devices and control algorithm on the other hand. A schematic diagram for the
implementation of such integrated health monitoring and vibration control system for the building structure is plotted in
Fig. 3.

3. Experimental validation of the proposed integrated approach

3.1. Experimental setup

To experimentally investigate the performance of the proposed integrated approach for damage detection and vibration
control of a building structure, a five-story building model was designed and built in the Dynamic Structural Laboratory of
The Hong Kong Polytechnic University, as shown in Fig. 4. The building model consisted of five rigid plates of 850 mm 
500 mm  16 mm and four equal sized rectangular columns of the cross section of 50 mm  6 mm. The total height of this
building model was 1750 mm, with the height of each story being identical. The cross section of the column was arranged in
28 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

Fig. 3. A schematic diagram of the implementation of the integrated system.

Fig. 4. Configuration of the five-story building model: (a) Elevation view (unit: mm); (b) Plane view (unit: mm); (c) Photograph of the model.

such a way that the stiffness of the building structure in the x-direction was much smaller than that in the y-direction. Each
steel floor plate was highly rigid in the horizontal direction compared with the columns, which led to a shear type de-
formation. Four additional columns with the cross section of 10 mm  6 mm were added in the building model to simulate
damage when they were symmetrically cut off in some stories. The columns and plates were properly welded together to
form rigid joints. All of the columns were eventually welded to a thick steel plate (see Fig. 4) that was in turn bolted firmly to
a shaking table. A series of silicon oil dampers were designed and installed between each of the adjacent floor plates to
increase the structural damping. As can be seen from Fig. 4, the lumped mass for each floor was composed of the masses of
 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 29

Fig. 5. Calibration of Wonder Box device: (a) Calibration setup of Wonder Box; (b) Output voltages versus input voltages.

the plate, columns, oil damper, MR damper, and the auxiliary for connecting the dampers to the plate. With the mea-
surement of each component, the mass matrix for the building model can be determined as Mdiag[67.955; 61.918; 56.837;
63.079; 59.922] (unit: kg) in which diag[] denotes a diagonal matrix. The Rayleigh damping assumption was used to
construct the damping matrix in this study.
The building model was fixed on the shaking table of 3 m  3 m, which was built by MTS Corporation. During the tests,
the structural acceleration response of each floor was measured by a piezoelectric accelerometer with a sampling frequency
of 1000 Hz. The ground acceleration induced by the shaking table was also measured by the accelerometer and used for the
comparison with the identified one. Moreover, the force transducers were placed in series with the MR dampers to measure
the control forces which were then used for the parameter and excitation identification. The displacement response of each
floor was also obtained but solely utilized for the assessment of the control performance.
Three MR dampers (RD-1097-01) manufactured by the Lord Corporation, USA, were respectively installed on the 1st, 2nd,
and 5th floor. To achieve the best performance of the MR damper, the Rheonetic Wonder Box device controller kit (RD-3002-
03) designed by the Lord Corporation was employed along with the MR damper. The output current supplied to the MR
damper by the Rheonetic Wonder Box device depended on the command voltage (the input voltage) from the computer.
Through the calibration of the Wonder Box device (see Fig. 5), it was found that the output current of such device would
reach its saturation value when the input voltages were larger than a certain value (about 1.8 V). Therefore, the maximum
and minimum input voltages were set to be 1.8 V and 0 V, respectively. A Simulink model was built, and a dSPACE real-time
simulator and control system was employed to process and analyze the aforementioned measurements and provide the
desired control signals according to Eq. (12).

3.2. Damage scenarios

The static tests of the building model as shown in Fig. 6 were first conducted for determining the structural stiffness in

Fig. 6. Static tests of the building model: (a) Without additional columns; (b) With four additional columns.
30 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

Table 1
The measured stiffness of the building model.

k1 (N/m) k2 (N/m) k3 (N/m) k4 (N/m) k5 (N/m)

5 5 5 5
Without additional columns 2.181  10 2.085  10 2.107  10 2.166  10 2.213  105
With two additional columns 2.418  105 2.323  105 2.363  105 2.412  105 2.483  105
With four additional columns 2.657  105 2.576  105 2.619  105 2.639  105 2.713  105

the cases without and with additional columns. As the number of the calibrated mass block increased, the increased de-
formation of the building model in the x-direction was recorded by dial gauges and used for the determination of the
structural stiffness. The measured results are given in Table 1.
Since the structural damage usually results in the reduction of structural stiffness, the damage is simulated in this study
by symmetrically cutting off the additional columns in certain stories as shown in Fig. 7. One advantage of such damage
pattern is that the mass remains unchanged while the structural stiffness is reduced. In this experimental study, four
damage scenarios were considered. In Case 1, two additional columns in the symmetrical position of the first floor were cut
off; in Case 2, the remaining two additional columns (i.e. all of the four additional columns) in the first floor were cut off; in
Case 3, two additional columns in the third floor were cut off; in Case 4: four additional columns in the third floor all were
cut off.
For each of the aforementioned four damage scenarios, hammer tests were carried out to capture the properties of the
building model when the MR dampers were removed as shown in Fig. 8. The impact force was applied on the top floor of the
building model and the acceleration response of each floor was measured. The modal parameters of the building, including
its natural frequencies, damping ratios, and modal shapes, were identified basing on the analysis of the corresponding
frequency response functions. Since the mass of an MR damper was 0.525 kg only, the loss of the total mass of each floor
was relatively small as compared with the original building model. Moreover, the stiffness provided by the MR damper was
quite limited. From these points of view, the variations of the natural frequencies caused by the removal of the MR dampers
would be small. The first two natural frequencies and damping ratios of the building model in the four damage cases are
shown in the Table 2 as examples.

3.3. Implementation of the proposed identification and control algorithms

After the MR dampers were installed to the building model, the integrated system was established and the shaking table
tests could be conducted for the aforementioned four damage cases. The dSPACE real-time simulator and control system was
employed to realize the integration of damage detection and vibration control of the building model as schematically shown
in Fig. 9. It can be seen that the dSPACE real-time simulator and control system consists of three processing blocks:
(1) DS2003 (A/D) block that is used for receiving analog signals from sensors and transforming them to digital signals;
(2) DS2102 (D/A) block that is used for transforming digital signals to analog signals and then sending them to the control
devices; and (3) the algorithms written using the MATLAB/Simulink block program, which are used to identify the structural
parameters and unknown excitations, as well as to create the corresponding control forces.
The presented integrated vibration control and health monitoring approach was first programmed by the MATLAB/
Simulink and then converted to the executive codes in the main program ControlDesk of dSPACE via real time interface
(RTI). Some details on how to build the Simulink model on the basis of the associated block programs are shown in Fig. 10. It
should be pointed out that each block marked by the bold line stands for the corresponding subsystem as shown in Figs. 10
(b, c) and it is built according to the equations given in Sections 2.1 and 2.2.
After the implementation of the proposed identification and control algorithms, the shaking table tests were conducted.
Two sets of earthquakes, i.e. Kobe earthquake and Northridge earthquake, were considered in this experimental

Fig. 7. Damage pattern in the experiment.

 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 31

Fig. 8. The hammer tests of the building model without MR dampers.

Table 2
The properties of the model determined by the hammer test.

Natural frequencies Damping ratios

f1 (Hz) f2 (Hz) 1 (%) 2 (%)

Case 1 2.907 8.533 0.86 1.03

Case 2 2.845 8.408 0.89 1.08
Case 3 2.813 8.346 1.01 1.09
Case 4 2.782 8.283 1.09 1.15

Fig. 9. Schematic diagram of the integrated system.

investigation. The structural parameter and earthquake identification results and the control performance under these two
earthquakes are given in the following subsections.

3.4. Damage detection and vibration control under Kobe earthquake

The scaled Kobe earthquake with a peak acceleration of 0.84 m/s2 was employed as the seismic input. The initial values
for the identification of the structural stiffness and Rayleigh damping coefficients were assigned as 1.2  105 N/m, 0.01 s-1,
and 1  10-4 s, respectively. The building model is static before the test, and thus the initial values for the ground excitation
and structural responses were all assumed to be zeros. Since the seismic input was used and assumed to be unknown, the
32 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

Fig. 10. MATLAB/Simulink block diagrams for the implementation of integrated system: (a) Block diagram of identification and control integration;
(b) Block diagram for identification in Sub1; (c) Block diagram for control in Sub2.

excitation influence matrix in Eq. (1) was equal to -M  [1 1 1 1 1]T. As mentioned before, the structural acceleration
responses and the MR damper forces all were measured and thus the observation equation can be calculated according to
Eq. (3) for the real-time parameter and excitation identification. The estimated structural parameters and the system states
were then used for the vibration control basing on the procedures described in Section 2.3. The aforementioned four da-
mage cases were considered, and the identified structural stiffness under Kobe earthquake were given in Table 3. By
comparing the values of the identified stiffness with those obtained from the static tests, it can be found from Table 3 that
the proposed approach is capable of identifying the structural stiffness satisfactorily. Taking Case 4 as an example, the
identified stiffness during the earthquake is presented in Fig. 11 as solid line whereas that determined by the static test is
shown as dashed line. It can be seen that the identified results by the proposed integrated approach can be stably converged
to the measured ones.
By using the proposed approach, the Rayleigh damping coefficients can also be estimated. However, the values of
Rayleigh damping coefficients are usually small and not easy for the assessment. Thus, the first two damping ratios, which
are determined basing on the identified Rayleigh damping coefficients and the first two natural frequencies as shown in
 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 33

Table 3
The identified structural stiffness (under Kobe earthquake).

k1 k2 k3 k4 k5

5 5 5 5
Case 1 Identified (N/m) 2.389  10 2.601  10 2.533  10 2.604  10 2.742  105
Relative error (%) 1.20 0.97 3.28 1.32 1.06

Case 2 Identified (N/m) 2.147  105 2.619  105 2.535  105 2.608  105 2.684  105
Relative error (%) 1.51 1.67 3.21 1.17 1.07

Case 3 Identified (N/m) 2.111  105 2.638  105 2.388  105 2.662  105 2.765  105
Relative error (%) 3.16 2.41 1.06 0.87 1.92

Case 4 Identified (N/m) 2.086  105 2.675  105 2.142  105 2.663  105 2.783  105
Relative error (%) 4.31 3.84 1.64 0.91 2.58

Fig. 11. The identified structural stiffness under Kobe earthquake.

Table 2, is employed herein for ease of comparison. The identified results are given in Table 4. It can be seen that, as
compared with the results obtained from the hammer tests (see Table 2), the identified results by means of the proposed
approach are still acceptable.
Besides the identification of structural parameters, the unknown ground motion can also be simultaneously identified by
the proposed approach. Also, taking Case 4 as an example, the time history of the identified seismic excitation is given in
Fig. 12 as dashed curve, whereas the solid curve is the corresponding measured one obtained directly from the accel-
erometer. For clarity of comparison, only the segment from 4 s to 8 s is shown in Fig. 12. It can be found from Fig. 12 that the
identified ground excitation is close to the measured one. The good agreement between the identified and measured ex-
citations also implies that the estimated structural damping ratios shown in Table 4 are acceptable.
Furthermore, based on the proposed integrated approach, the structural vibration can be significantly reduced with the
aid of the MR dampers. Two sets of shaking table tests, saying the building model with and without MR dampers, were
conducted in the lab. The aforementioned four damage scenarios were considered. The switching control law introduced in

Table 4
The identified damping ratios (under Kobe earthquake).

Case 1 Case 2 Case 3 Case 4

1 (%) 0.67 0.86 0.82 1.06

2 (%) 0.28 0.45 0.88 0.89
34 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

1
Identified

Kobe ground excitation (m/s2 )

Measured
r
0.5

-0.5

-1
4 5 6 7 8
time (s)
Fig. 12. The comparison of the identified Kobe ground excitation.

Section 2.2 was employed for adjusting the properties of the MR dampers. The maximum acceleration and displacement
responses of the building model under the identical Kobe earthquake are given in Table 5 for comparison. Here, si and ai (i
1, 2, , 5) are the peak displacement and acceleration of the i-th floor obtained from the sensors, respectively. It is observed
from Table 5 that the structural vibration is significantly reduced. Moreover, the comparison of time histories of the
measured displacement and acceleration of the top floor are shown in Fig. 13. Here, only the building model in Case 4 is
plotted as an example. Similar results can be obtained from the other cases.

3.5. Damage detection and vibration control under Northridge earthquake

To investigate the robustness of the proposed integrated approach, the shaking table tests of the building model under
the scaled Northridge earthquake with a peak acceleration of 2.28 m/s2 were conducted. The aforementioned four damage
scenarios were considered. Based on the proposed integrated approach, the structural stiffness can be identified and shown
in Table 6. It can be found from Table 6 that the maximum relative error for all the cases is only 4.37% indicating the
structural stiffness can be identified accurately. Similarly, taking Case 4 as an example, the identified stiffness during the
earthquake is plotted in Fig. 14 as solid line whereas that determined by the static test is shown as dashed line. It can be
found that the identified results by means of the proposed integrated approach can be stably converged to the measured
ones. The Rayleigh damping coefficients can be simultaneously identified as well, and the first two damping ratios de-
termined on the basis of the identified damping coefficient and the first two natural frequencies are given in Table 7 for ease
of comparison. It can be seen from Table 7 that the identification results of the structural damping is still acceptable to some
extent.
Furthermore, the unknown ground motion can also be identified by the proposed approach. Here, taking Case 4 as an
example, the time history of the identified ground excitation is shown in Fig. 15 as dashed curve, whereas the solid curve is
the corresponding measured one. For clarity of comparison, only the segment from 4 s to 8 s is plotted in Fig. 15. It is clear
that the identified ground excitation is close to the measured one.
The comparison of the structural responses in terms of the peak acceleration and displacement with and without control

Table 5
Peak responses of the building model under Kobe earthquake.

si (mm) ai (m/s2)

s1 s2 s3 s4 s5 a1 a2 a3 a4 a5

Case 1 Without control 5.83 9.22 12.58 14.74 14.65 1.95 3.09 4.12 4.66 4.80
With control 3.45 4.19 5.17 5.69 5.79 1.11 1.37 1.30 1.45 1.61

Case 2 Without control 6.22 9.40 12.45 14.31 15.14 1.97 2.93 3.67 4.17 4.35
With control 3.53 4.24 5.19 5.70 5.77 1.12 1.35 1.31 1.46 1.67

Case 3 Without control 5.58 8.36 11.24 12.81 13.47 1.82 2.59 3.22 3.72 3.95
With control 3.41 3.86 4.81 5.39 5.43 1.31 1.42 1.33 1.73 1.70

Case 4 Without control 5.19 7.86 10.88 12.38 13.04 1.59 2.28 2.97 3.53 4.03
With control 3.36 3.81 4.78 5.33 5.36 1.15 1.29 1.21 1.56 1.53
 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 35

Fig. 13. The comparison of the time histories of structural responses with and without control (Kobe earthquake).

Table 6
The identified structural stiffness (under Northridge earthquake).

k1 k2 k3 k4 k5

Case 1 Identified (N/m) 2.368  105 2.641  105 2.537  105 2.605  105 2.686  105
Relative error (%) 2.07 2.52 3.13 1.29 1.01

Case 2 Identified (N/m) 2.143  105 2.616  105 2.537  105 2.612  105 2.668  105
Relative error (%) 1.83 1.55 3.13 1.02 1.66

Case 3 Identified (N/m) 2.095  105 2.626  105 2.392  105 2.599  105 2.756  105
Relative error (%) 3.89 1.94 1.23 1.52 1.58

Case 4 Identified (N/m) 2.121  105 2.634  105 2.199  105 2.598  105 2.751  105
Relative error (%) 2.71 2.25 4.37 1.55 1.41

Fig. 14. The identified structural stiffness under Northridge earthquake.

is shown in Table 8. Four damage scenarios are considered and the absolute acceleration and displacement measurements
obtained from the sensors are employed for comparison. It can be found from Table 8 that the structural vibration can be
attenuated substantially. Moreover, Fig. 16 gives the comparison of time histories of the measured displacement and
36 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

Table 7
The identified damping ratios (under Northridge earthquake).

Case 1 Case 2 Case 3 Case 4

1 (%) 0.68 0.65 0.9 0.75

2 (%) 0.67 0.54 0.95 0.85

Northbridge ground excitation (m/s )

2
Identifi
fied
Measured
r
2

-1

-2
4 5 6 7 8
time (s)
Fig. 15. The comparison of the identified Northridge ground excitation.

Table 8
Peak responses of the building model under Northridge earthquake.

si (mm) ai (m/s2)

s1 s2 s3 s4 s5 a1 a2 a3 a4 a5

Case 1 Without control 5.89 7.85 10.27 12.35 13.64 3.12 3.45 3.44 4.52 5.74
With control 4.66 6.33 8.31 9.64 9.98 1.89 1.95 2.73 3.09 3.41

Case 2 Without control 5.51 7.48 9.84 11.73 12.86 2.85 2.86 3.16 4.18 4.88
With control 4.76 6.36 8.30 9.59 9.92 1.67 1.93 2.62 2.76 3.03

Case 3 Without control 5.38 7.30 9.81 11.66 12.74 2.72 2.61 3.13 3.95 4.71
With control 4.27 5.68 7.81 9.28 9.56 1.71 1.73 2.29 2.27 2.45

Case 4 Without control 5.36 7.18 9.91 11.72 12.78 2.74 2.60 3.02 3.76 4.56
With control 4.12 5.37 7.56 8.94 9.20 1.77 1.69 2.13 2.11 2.26

Fig. 16. The comparison of the time histories of structural responses with and without control (Northridge earthquake).
 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338 37

acceleration of the top floor. Here, only the building model in Case 4 is plotted as an example. Similar results can be obtained
from the other cases.

4. Conclusions

In this paper, the possibility of the establishment of a smart building structure with the function of simultaneously
identifying structural damage and mitigating structural vibration was explored experimentally. A five-story shear building
structure was designed and used for the validation of the presented approach. Three MR dampers respectively located on
the 1st, 2nd, and 5th floor were employed for vibration attenuation. Four damage scenarios were considered by symmetrically
cutting off the additional columns in certain stories. Two sets of earthquakes, i.e. Kobe earthquake and Northridge earth-
quake, were used as seismic input and assumed to be unknown during the tests. The structural parameters and the un-
known ground excitations were identified with acceptable accuracy. The maximum relative error of the identified stiffness
was 4.31% under Kobe earthquake and 4.37% under Northridge earthquake. It should be noted in this study the acceleration
responses and control forces were all assumed to be measured for the identification. Based on the identified structural
parameters and system states, a switching control law was employed to adjust the current applied to the MR dampers for
the purpose of vibration attenuation. As compared with the building model without control, the structural responses were
substantially reduced with control in terms of peak displacement and acceleration. Notably, only the time-invariant para-
meters are considered in this study. If the structural damage process and mechanism shall be considered, the proposed
integrated approach shall be extended to simultaneously consider the identification of time-varying structural parameters
and unknown excitations as well as the suppression of excessive structural vibration.

Acknowledgements

The authors gratefully acknowledge the financial support from the Research Grants Council of Hong Kong through a GRF grant
(PolyU 5319/10E) and The Hong Kong Polytechnic University through an FCE Postdoctoral Fellowship Scheme (Grant no. ZVAU). The
authors would also like to express their sincere thanks to Dr. Z.H. Chen of Fuzhou University, China, for his help during the course of
this work.

References

[1] G. Housner, L. Bergman, T. Caughey, A. Chassiakos, R. Claus, S.F. Masri, R. Skelton, T.T. Soong, B.F. Spencer Jr, Y. Yao, Structural control: past, present, and
future, J. Eng. Mech. 123 (9) (1997) 897971.
[2] B.F. Spencer Jr., S. Nagarajaiah, State of the art of structural control, J. Struct. Eng. 129 (7) (2003) 845856.
[3] W. Fan, P.Z. Qiao, Vibration-based damage identification methods: a review and comparative study, Struct. Health Monit. 10 (1) (2011) 83111.
[4] Y.L. Xu, Y. Xia, Structural Health Monitoring of Long-span Suspension Bridges, Spon Press, London, 2012.
[5] S.J. Dyke, B.F. Spencer Jr., M.K. Sain, J.D. Carlson, An experimental study of MR dampers for seismic protection, Smart Mater. Struct. 7 (1998) 693703.
[6] F. Yi, S.J. Dyke, J. Caicedo, J. Carlson, Experimental verification of multi-input seismic control strategies for smart dampers, J. Eng. Mech. 127 (11) (2001)
11521164.
[7] J. Chen, Y.L. Xu, W.L. Qu, Z.L. Wu, Seismic response control of a complex structure using multiple MR dampers: experimental investigation, Earthq. Eng.
Eng. Vib. 3 (2) (2004) 181193.
[8] L. Zhou, S. Wu, J. Yang, Experimental study of an adaptive extended Kalman filter for structural damage identification, J. Infrastruct. Syst. (2008) 4251,
http://dx.doi.org/10.1061/(ASCE)1076-0342(2008)14:1(42).
[9] J.N. Yang, S. Lin, H.W. Huang, L. Zhou, An adaptive extended Kalman filter for structural damage identification, Struct. Control Health Monit. 13 (2006)
849867.
[10] H.W.Huang, J.N.Yang, and L.Zhou, Comparison of various structural damage tracking techniques based on experimental data, in: Proceedings of SPIE
6932, Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems, 2008, 6, 9, 3225, http://dx.doi.org/10.1117/12.774621,
2008.
[11] Q. Yin, L. Zhou, T. Mu, J.N. Yang, Experimental study on damage detection of base-isolated structure using an adaptive extended Kalman filter, J. Theor.
Appl. Mech. 51 (4) (2013) 10131026.
[12] I. Trendafilova, R. Palazzetti, A. Zucchelli, Damage assessment based on general signal correlation: application for delamination diagnosis in composite
structures, Eur. J. Mech. A/Solids 49 (2015) 197204.
[13] Y.L. Xu, B. Chen, Integrated vibration control and health monitoring of building structures using semi-active friction dampers: part I methodology, Eng.
Struct. 30 (7) (2008) 17891801.
[14] B. Chen, Y.L. Xu, Integrated vibration control and health monitoring of building structures using semi-active friction dampers: part II numerical
investigation, Eng. Struct. 30 (3) (2008) 573587.
[15] B. Chen, Y.L. Xu, X. Zhao, Integrated vibration control and health monitoring of building structures: a time-domain approach, Smart Struct. Syst. 6 (7)
(2008) 811833.
[16] C.H. Lin, N. Sebastijanovic, H.T.Y. Yang, Q. He, X. Han, Adaptive structural control using global vibration sensing and model updating based on local
infrared imaging, Struct. Control Health Monit. 19 (6) (2012) 609626.
[17] K. Karami, F. Amini, Decreasing the damage in smart structures using integrated online DDA/ISMP and semi-active control, Smart Mater. Struct. 21
(2012) 105017.
[18] Y. Lei, D.T. Wu, S.Z. Lin, Integration of decentralized structural control and the identification of unknown inputs for tall shear building models under
unknown earthquake excitation, Eng. Struct. 52 (2013) 306316.
[19] Y. Lei, H. Zhou, L.J. Liu, An on-line integration technique for structural damage detection and active optimal vibration control, Int. J. Struct. Stab. Dyn. 14
(5) (2014) 1440003.
[20] F. Amini, S.A. Mohajeri, M. Javanbakht, Semi-active control of isolated and damaged structures using online damage detection, Smart Mater. Struct. 24
38 J. He et al. / Journal of Sound and Vibration 390 (2017) 2338

(2015) 105002.
[21] Y.L. Xu, Q. Huang, Y. Xia, H.J. Liu, Integration of health monitoring and vibration control for smart building structures with time-varying structural
parameters and unknown excitations, Smart Struct. Syst. 15 (3) (2015) 807830.
[22] S. Nagarajaiah, Adaptive passive, semiactive, smart tuned mass dampers: identification and control using empirical mode decomposition, Hilbert
transform, and short-term Fourier transform, Struct. Control Health Monit. 16 (7-8) (2009) 800841.
[23] H.T.Y. Yang, J.Z. Shan, C.J. Randall, P.K. Hansma, W.X. Shi, Integration of health monitoring and control of building structures during earthquakes, J. Eng.
Mech. 140 (5) (2014) 04014013.
[24] Q. Huang, Y.L. Xu, J.C. Li, Z.Q. Su, H.J. Liu, Structural damage detection of controlled building structures using frequency response functions, J. Sound Vib.
331 (15) (2012) 34763492.
[25] Y.L. Xu, Q. Huang, S. Zhan, Z.Q. Su, H.J. Liu, FRF-based structural damage detection of controlled buildings with podium structures: experimental
investigation, J. Sound Vib. 333 (13) (2014) 27622775.
[26] J. He, Q. Huang, Y.L. Xu, Synthesis of vibration control and health monitoring of building structures under unknown excitation, Smart Mater. Struct. 23
(2014) 105025.
[27] S.J. Dyke, B.F. Spencer Jr., M.K. Sain, J.D. Carlson, Modeling and control of magnetorheological dampers for seismic response reduction, Smart Mater.
Struct. 5 (1996) 565575.
[28] S.J.Dyke, and B.F.SpencerJr., A comparison of semi-active control strategies for the MR damper, in: Proceedings of Intelligent Information Systems,
IEEE, Grand Bahama Island, 1997, pp. 580584.
[29] L. Jansen, S.J. Dyke, Semiactive control strategies for MR Dampers: comparative study, J. Eng. Mech. 126 (8) (2000) 795803.
[30] Y.L. Xu, W.L. Qu, J.M. Ko, Seismic response control of frame structures using magnetorheological/electrorheological dampers, Earthq. Eng. Struct. Dyn.
29 (2000) 557575.
[31] J.P. Ou, Structural Vibration Control: Active, Semi-active, and Intelligent Control, Science Press, Beijing, 2003.