You are on page 1of 20

STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. (2014)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/stc.1661

Seismic response analyses of an asymmetric base-isolated building


during the 2011 Great East Japan (Tohoku) Earthquake

Dionysius M. Siringoringo1,*,† and Yozo Fujino2


1
Department of Civil Engineering, University of Tokyo, Tokyo, Japan
2
Institute of Engineering Innovations, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan

SUMMARY
Analysis of strong motion recordings of a base-isolated building during the March 11, 2011, Great East
Japan (Tohoku) Earthquake is reported in this paper. The building, located in Tokyo Bay area, is an asymmetric
L-shaped structure consisting of seven-story and 14-story building with vertical opening. Vibration monitoring
system was installed on the building in 2010, and seismic responses were recorded including the strongest shaking
(peak ground acceleration 0.80–1.40 m/s2) experienced during the main shock of March 11, 2011, Great East
Japan Earthquake. The building survived the earthquake without structural damage. The study in this paper
includes response analysis, system identification, and seismic performance evaluation of the structure, especially
performance of base-isolation system. The study shows that despite considerable shift in effective frequency of
the building due to the increase in flexibility of isolation system during the main shock, large acceleration was
recorded on the superstructure with the peak floor accelerations of nearly 300 cm/s2. Two factors contribute to this
cause: one is the characteristics of the building where torsional modes dominate the seismic response of upper
stories and the other is resonance, where dominant frequencies of ground motions coincide with the natural
frequencies of torsional modes. Moreover, analysis shows that torsional modes were not significantly influenced
by performance of base isolation, so that even though the base isolation has functioned properly, the upper stories
still experienced large floor accelerations. The paper also discusses long-term observation of seismic responses
during aftershocks and various levels of earthquakes between 2010 and 2012. Copyright © 2014 John Wiley &
Sons, Ltd.

Received 17 October 2013; Revised 4 March 2014; Accepted 10 March 2014

KEY WORDS: building seismic monitoring; asymmetric base-isolated building; building torsional response; 2011
Great East Japan Earthquake; system identification

1. INTRODUCTION
Base isolation as a seismic mitigation technique for building and bridge has gained popularity in
Japan especially after the 1995 Hyogo-ken Earthquake. The favorable response during an
earthquake, ability to limit structural damage to a low and repairable level, and ability to maintain
functionality after a large earthquake are some advantages that have increased the acceptance of
base-isolation technology among engineers and building owners. Nowadays in Japan, base-
isolation system is widely used for hospital, school, university, and office buildings [1]. The
system usually consists of seismic isolators such as elastomeric or sliding type combined with
energy absorbing dampers. The isolator creates a structure with longer natural period, whereas
the damper provides higher damping to reduce the structural response. According to a recent
survey conducted in various locations in Japan, seismically isolated buildings generally showed

*Correspondence to: Dionysius M. Siringoringo, Department of Civil Engineering, University of Tokyo, Tokyo, Japan.

E-mail: dion@bridge.t.u-tokyo.ac.jp

Copyright © 2014 John Wiley & Sons, Ltd.


D. M. SIRINGORINGO AND Y. FUJINO

good performances during the 2011 Great East Japan (Tohoku) Earthquake, in that acceleration
responses of superstructures were less than the input ground motions as a result of isolation
effect [2,3].
There are two main concerns regarding the application of base-isolation system on an asymmetric
building, that is, asymmetricity of isolation system and asymmetricity of superstructure system. In both
cases, the asymmetricity could be caused by eccentricity in either superstructure or isolation system.
When eccentricity exists in a superstructure system, coupled lateral–torsional motions occur as they
do in the asymmetric nonisolated buildings. The effect of torsional motion can be significant when a
building has large eccentricity. Outer isolators will experience larger force than inner ones when
torsional motion is dominant. Large building rotation also contributes to significant corner deformation
such that outer columns will experience larger shear force as a result of combined shear and moment
from floor eccentricity.
The behavior of asymmetric base-isolated buildings and the effects of torsional motion on base
isolation performance have been extensively studied analytically, numerically, and experimentally
by many researchers. Lee [4], who defined the dynamic torque amplification factor as a ratio of
dynamic torque to the static torque at the center of stiffness, concluded that when eccentricity of
isolation systems is small (<0.2L, L is the longest plan dimension), displacements of the base
due to rotational motions remain small even if eccentricity of the structure is large. However, as
shown by Eisenberger [5], this conclusion is valid only for a certain type of ground motion. Pan
and Kelly [6] show analytically the importance of ratio between torsional and lateral frequency
on the rotational motion and corner deformation. Nagarajaiah [7] investigated the effects of
eccentricity of isolation, eccentricity of superstructure, and frequency ratio between superstructure
and base, and concluded that the main source of torsional motion in base-isolated buildings with
elastomeric bearings is the eccentricity of isolation system, in that increasing the eccentricity of
isolation system leads to the increase in torque amplification.
Experimental studies on a scaled model building in the shaking table tests have been performed to
investigate performance of asymmetrical base-isolated buildings. Employing models with various
eccentricities, Nakamura et al. [8] showed that the rotational motion of the structure can be minimized
by reducing the eccentricity of the isolation system. Hwang and Hsu [9], who tested asymmetric
superstructure with various eccentricities on the isolation system, concluded that increasing the
eccentricity will amplify the rotational response of the structure. Furthermore, corner displacement
was amplified by the rotational motion; with the contribution of the rotational motion to the maximum
corner, displacement can be as high as 30%.
Strong motion recordings of instrumented buildings have also been used to study the torsional
response of base-isolated buildings. The records from the 1985 Redlands earthquake (ML 4.8) on
the Foothill Community Law and Justice Center [10] revealed that the transverse motion of the
building is more dominant than the longitudinal motion. The rotational motion of the
superstructure due to spatial variability of ground motion and the extreme length of the structure
are considered as the main reason. Nagarajaiah and Xiahong [11] observed nominal torsional
response of the University of South California hospital building during the 1994 Northridge
earthquake. The building has an eccentricity ratio of 5%, and the ground motion had significant
energy in the higher mode.
The aforementioned studies, experiments, and strong motion observations provide valuable
insights on the behavior of asymmetrical base-isolated buildings and conditions that influence
the performance of isolation system. Recently, many base-isolation buildings, especially in an
earthquake-prone country such as Japan, are instrumented with seismic monitoring systems. The
monitoring systems provide seismic responses that are essential for seismic performance
evaluation. This paper describes a study on strong motion recordings from such building. The
study includes the following: (i) seismic response evaluation of base-isolated building during
the main shock of March 11, 2011, Great East Japan Earthquake using time-invariant and time-
variant system identification; (ii) investigation on the seismic response characteristics considering
the building asymmetricity; and (iii) evaluation of the long-term seismic response characteristics
of the building for over a year that includes the foreshock, main shock, aftershocks, and various
amplitude earthquakes in 2010–2012.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

2. DESCRIPTION OF BASE-ISOLATED BUILDING AND SEISMIC ISOLATORS


The monitored structure is the SIT Building located in Tokyo Bay area. The building consists of two
parts: 14-story main building (M) and seven-story annex building (A) (Figure 1). The main building is
97.2 m long, 43.2 m wide, and 67.5 m high with 19 bays in longitudinal direction and six bays in
transverse direction. The annex building is 81 m long, 21.6 m wide, and 31.2 m high with 16 bays in
longitudinal direction and three bays in transverse direction. Both buildings are of braced steel frames
and connected at the corner by elevator shaft to form an L-shaped asymmetric structure. The 14-story
main building has vertical opening in the middle starting from the second floor to the seventh floor. The
opening divides the main building into the west section and east section, named hereafter as sections
MW and ME, respectively. Meanwhile, the seven-story annex building has some voids on floors to
provide access for escalator. Both buildings are connected by concrete slab at the basement level, on
top of which isolation systems are placed.
The isolation system consists of 59 natural rubber bearing (NRB) and 26 sliding bearing. There are
two types of NRB, namely, the 100 cm diameter with initial stiffness 11.9 kN/cm (43 units) and the
110 cm diameter with initial stiffness 12.9 kN/cm (16 units). The NRBs are made up of 26 layers of
rubber with 0.75 and 0.83 mm thickness for diameter 100 and 110 cm, respectively, and steel shims
of 0.45 cm. The design maximum deformation of NRB is set to 250% of the shear strain or about
49 cm. Meanwhile, the ultimate deformation is approximately 350% of the shear strain or about
69 cm. The sliding bearings’ surface is made of polytetrafluoroethylene with the initial stiffness
12.3, 17.5, and 22.9 kN/m for diameter 56.4, 71.4, and 94.4 cm, respectively. The thickness of
the sliding surface is 2 mm with the friction coefficient of 0.013. To limit the horizontal motion,
dampers are added to the isolation system. Two types of dampers are used, namely the lead
damper (28 units) and the U-shaped steel dampers (33 units). The layout of isolators and dampers
are given in Figure 1(b).
The eccentricity ratios of the building and isolator system are listed in Table I. The ratio is defined as
the eccentricity in the x- and y-directions with respect to building length in the corresponding direction
and presented separately according to building sections and floors. The table shows that the eccentricity

81m

16.2m
21.6m

Annex Building

Section
MW
21.6m

NRB (Natural Rubber Bearing)


SB (Sliding Bearing)
U-Shaped Damper
Lead Damper Vertical Void
(2 ~7 Fl)
97.2m
32.4m

Main
Building
Y Direction
32.4m

Section
ME
X Direction
(b)
43.8m

(a)
GPS Accelerometers on Upper floors

Network Time Network Network


Router
Protocol (NTP) Hub Hub
Server

Accelerometers GPS DF
on Upper floors

Data Server
Controller
Unit

(c) Accelerometers on Ground level


(d)

Figure 1. (a) SIT Building, (b) layout of isolation system, (c) sensor locations, and (d) architecture of the monitor-
ing system.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

Table I. Eccentricity ratio of the SIT Building.


Eccentricity ratio (%)
X-direction Y-direction
Annex building
1st Floor 0.72 1.56
2nd Floor 1.58 9.07
3rd Floor 0.64 3.41
4th Floor 0.47 7.86
5th Floor 0.49 3.69
6th Floor 1.94 5.76
7th Floor 1.91 5.50
Main building
1st Floor 3.00 0.53
2nd Floor 7.32 4.23
8th Floor 2.60 1.94
9th Floor 3.67 1.82
10th Floor 2.48 2.16
11th Floor 2.68 2.12
12th Floor 4.14 2.04
13th Floor 1.70 2.77
14th Floor 0.46 0.10
Main building section MW
3rd Floor 7.52 11.39
4th Floor 7.61 11.73
5th Floor 7.72 11.81
6th Floor 7.74 11.97
7th Floor 5.82 23.86
Main building section ME
3rd Floor 3.24 9.90
4th Floor 3.05 9.64
5th Floor 2.96 9.69
6th Floor 2.85 9.86
7th Floor 2.83 9.98
Isolators 3.41 8.83
Note: Eccentricity ratio = eccentricity/building length × 100%

in the y-direction is generally larger than in the x-direction for annex and main building, and the ratios
in some floors are larger than 10%, indicating significant asymmetricity of the building.

3. DESCRIPTION OF SEISMIC MONITORING SYSTEM


The building has a seismic monitoring system that consists of 21 triaxial accelerometers (18 accelerometers
on the structure and three accelerometers on the ground) and four triaxial displacement-meters measuring
relative displacement between basement and the first floor (Figure 1(c)). Vibration sensors are placed on
the basement (below the isolators) and the first, fourth, seventh, ninth, and fourteenth floors.
Acceleration responses are recorded by small servo-type accelerometers SQ-32 with resolution of
0.01 cm/s2 and measurement range of ±2000 cm/s2. The sensors are connected through local area
network, and the clock on each sensor is precisely synchronized with the Network Time Protocol server.
Furthermore, a GPS-connected controller unit is assigned to provide a global reference position and
synchronized time recording among the sensors. With this system, the delay time among sensors can be
reduced to the maximum of 4 ms [12]. The controller is connected to three accelerometers on the ground
level that act as trigger. Once these accelerometers record ground acceleration larger than the threshold
0.5 cm/s2, the controller will activate the monitoring system to record building responses. The responses
were sampled at 100 Hz and stored in a server for further analysis. It should be mentioned that horizontal
accelerations were oriented in the x- and y-directions according to the building orientation as opposed to

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

the commonly used north–south and east–west directions. The sensors nomenclature is organized as
follows: the first index denotes sensor type (A, accelerometer; D, displacement-meter), the second index
represents the building (M, main; A, annex building), the third index represents story level, and the last
index shows location of the sensor in the floor (i.e., S, south; E, east; W, west). The type of sensors and their
locations are shown in Figure 1(c).

4. DESCRIPTION OF RECORDED SEISMIC RESPONSES


At 02:46 PM Japan Standard Time (JST) on March 11, 2011, northeastern Japan was struck by the most
devastating earthquake in 130 years. The Great East Japan earthquake has the moment magnitude of
9.0, the largest ever recorded in Japan. The Japan Meteorological Agency (JMA) seismic intensity
5 (lower 5), out of the maximum scale of 7, was recorded on the building location. This seismic
intensity is equivalent to scale VII in the Modified Mercalli Intensity scale. Seismic responses
described in this study consist of the main shock event on March 11, 2011, at 02:47 PM, foreshock
event, and several aftershocks on the same days and few days afterward with the JMA seismic intensity
equal to or larger than 3. In the beginning, the discussion will focus on the building seismic responses
during the main shock event on March 11, 2011, at 02:47 PM JST. Afterward, responses due to several
small earthquakes in 2010, foreshock to the March 11, 2011, earthquake, aftershocks, and various
levels of earthquake between 2011 and 2012 are discussed to provide comparison of the response
characteristics. Table II provides detailed information on five largest earthquakes that give the largest
excitation recorded on the building in the period between 2010 and 2012.

5. CHARACTERISTICS OF RECORDED MAIN SHOCK GROUND MOTIONS


Figure 2 shows time histories of ground acceleration obtained from free-field AF0E sensor
during the main shock. The excitation lasted for about 10 min, in which the significant dura-
tion (i.e., 95% energy of the excitation) lasts for about 6 min since the arrival of the primary
wave. The maximum ground acceleration in the X, Y, and vertical directions were 0.16, 0.17,
and 0.08 g, respectively. The peak acceleration occurred at 143 and 130 s for the X and Y directions,
respectively. The ground accelerations have dominant frequency content between 0.7 and 1.2 Hz in both X
and Y directions.
In Figure 3, the response spectra of two ground accelerations, namely AM0W (basement level) and
AF0E (free-field sensor), are shown. The spectra are compared with the design response spectra
for level 2 earthquake with soil condition classes 2 and 3, as specified in the Japan’s Building
Code. As can be seen in the figure, the spectra of the main shock have large amplitude in the
period range of 0.7–1.2 s and at about 2.5 s. For the period range below 2.5 s, which is the period
range of the building, the spectra amplitudes of the present earthquake are still well below the
design spectra.

Table II. Five earthquakes with the largest excitation recorded on the SIT Building between 2010 and 2012.
Max recorded ground Max recorded
acceleration (cm/s2) acceleration (cm/s2)
Earthquake JMA seismic
(date-month-year), JST Mw intensity AM0W-Y AM0W-X AA7S-Y AM11E-X
11-03-2011, 02:47 PM (main shock) 9.0 5 110.53 80.46 297.42 204.12
11-03-2011, 03:16 PM (after shock 1) 7.7 4 53.03 38.24 239.99 87.99
29-05-2012, 01:36 AM 5.2 4 29.58 28.20 32.83 29.48
24-11-2012, 05:59 PM 4.8 4 27.09 18.49 44.58 15.98
07-12-2012, 05:19 PM 7.3 4 29.20 26.37 49.79 38.78

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

(a) (b)
Figure 2. (a) Recorded ground acceleration time histories (AF0E: free-field) during the main shock; (b) Fourier
spectra of ground acceleration.

6. CHARACTERISTICS OF RECORDED BUILDING RESPONSES


The characteristics of accelerations recorded on the top floor of the annex and main buildings are
shown in Figure 4. The responses have similar characteristics with the accelerations on the lower floors
with one dominant peak. Similar to the accelerations on the lower and ground floors, the peak
accelerations occur at about 143 and 130 s for the X and Y directions, respectively, indicating no
significant time lag between the response of the ground floor and the upper floors. Table III lists the
peak accelerations recorded by all sensors during the main shock. Strong amplification was recorded
on the annex building in the Y direction. The peak acceleration on the south section of seventh floor
(AA7S) is nearly 300 cm/s2 (Figure 5) or about three times of the basement acceleration. Large accel-
eration was also recorded on the fifth floor (AA7S), with peak acceleration of 265 cm/s2. This large ac-
celeration can be considered exceptional considering that the peak acceleration on the basement level
of the southeast section of the annex building is only 96 cm/s2. The transfer functions of accelerations
of the annex building in the Y direction (Figure 6) reveal that the responses were dominated by fre-
quency peak at 0.86 Hz.

1.4
DS-L2-S2
DS-L2-S3
1.2
Spectra Acceleration(g)

AM0W-Y
AM0W-X
AF0E-Y
1 AF0E-X

0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
Period(sec)

Figure 3. Acceleration response spectra at surface level for 5% damping and comparison with design spectra at
surface level (note: DS, design spectra; L2, level 2 earthquake; S2/S3, site class 2/3 for Tokyo area).

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

Figure 4. Recorded horizontal accelerations at the top floor of the building: (a) seventh floor of the annex building,
(b) 14th floor of the main building section NW, and (c) 14th floor of the main building section NE. Fourier spectra
of horizontal acceleration on the (d) seventh floor of the annex building, (e) 14th floor of the main building section
NW, and (f) 14th floor of the main building section NE.

Table III. Summary of the SIT Building peak responses during the main shock of March 11, 2011, Great East
Japan Earthquake (units: acceleration in cm/s2; displacement in cm).
Peak response Peak response
Main Annex
building X-direction Y-direction Vertical building X-direction Y-direction vertical
AM0W 80.46 110.53 86.88 AA4S 69.61 265.15 78.94
AM4W 66.75 84.09 107.07 AA7S 85.31 297.43 99.32
AM7W 69.67 81.36 122.62 AA0S 80.72 96.03 76.54
AM9W 87.93 80.92 149.88 AA0W 111.53 108.82 87.35
AM14W 236.52 141.41 178.92 AA1S 92.53 89.41 84.38
AM4E 129.07 92.23 72.10 DA0S 5.60 4.82 0.29
AM7E 112.36 82.36 90.27 DA0W 7.88 5.06 0.24
AM9E 116.29 74.91 103.74
AM14E 204.12 137.38 117.43
AM1W 85.35 118.04 49.67
AM1E 120.66 94.68 46.09
AM0E 82.39 110.62 88.83
DM0W 8.04 4.78 0.47
DM0E 9.90 5.12 0.57

Response amplification of the upper stories was also observed on the main building in the X
direction. The peak accelerations on the fourteenth floor of the east and west sections of the main
building were about 204 and 236 cm/s2, respectively, or about 2.5 times of the ground acceleration.
Figure 7 shows the transfer functions of accelerations of the main building in the X direction, and it
can be seen that large acceleration responses were mainly due to frequency component at 0.58 Hz,
whereas responses in the Y direction were dominated by frequency component at 0.44 Hz.
The significant acceleration amplification of the upper stories is rather unexpected for a base-
isolated building. A common assumption in the design of seismically isolated building is that lateral
deformation is concentrated on isolator level and the upper structure behaves rigidly without significant
dynamic amplification. In Japan, although not explicitly prescribed in the code, on the basis of
experiments on medical facilities and electronic appliances, the common design practice is to limit

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

14 14 14
13 13 13
12 12 12
11 11 11
10 10 10
9 9 9

Floor
8 8 8
AA7S
7 7 AA7S 7
6 6 6
5 5 5
AA4S AA4S 4
4 4
3 3 3
2 2 2
1 1 1
0 0 0
50 100 150 200 250 300 50 100 150 200 250 300 50 100 150 200 250 300
Max Acc X-Dir (cm/s2) Max Acc Y-Dir (cm/s2) Max Acc Vertical (cm/s2)

Figure 5. Recorded maximum acceleration amplitude for each floor in each direction (note: 0 = basement level).

15
Transfer Ratio

0.86Hz AA4S
AA7S
10

0
0.5 1 1.5 2 2.5
Frequency (Hz)

Figure 6. Transfer functions of recorded accelerations on the annex building in the Y direction.

8
Transfer Ratio
Transfer Ratio

8 AM9W-Y AM9W-X
0.44Hz 0.58Hz
AM14W-Y AM14W-X
6 6
4 4
2 2
0 0
0.5 1 1.5 2 2.5 0.5 1 1.5 2 2.5
(a) Frequency (Hz) (c) Frequency (Hz)
0.58Hz
Transfer Ratio

8 AM9E-X
Transfer Ratio

8
0.44Hz AM9E-Y AM14E-X
6 AM14E-Y 6

4 4
2 2
0 0
0.5 1 1.5 2 2.5 0.5 1 1.5 2 2.5
(b) Frequency (Hz) (d) Frequency (Hz)

Figure 7. Transfer functions of recorded accelerations on the main building in the (a,b) Y and (c,d) X directions.

the horizontal floor acceleration to maximum 300 cm/s2 [13]. Avoiding large floor acceleration is
necessary to protect nonstructural components, highly sensitive equipment, and computer facilities
inside the building. In the following sections, we shall discuss the reason behind significant dynamic
amplification of the upper stories.

7. SYSTEM IDENTIFICATION TECHNIQUES


The seismic behavior of the building is examined thoroughly by evaluating vibration modes pertinent
to the seismic responses using system identification techniques. The general approach of earthquake-
induced system identification is to use the input–output relationship to recreate structural models that

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

are capable of reproducing the actual responses, for example, via output-error minimization of parametric
time-invariant model [14], MIMO state-space with system realization model [15–17], or time-variant
adaptive least square technique and autoregressive exogenous (ARX) model [18,19]. In this study, two
types of nonparametric system identification procedures were employed, namely the time-invariant
system using moving window and the time-variant system using recursive least square (RLS).

7.1. Time-invariant system identification


The first system identification, a time-invariant MIMO system, is based on the system realization using
information matrix (SRIM) [20]. The SRIM algorithm utilizes correlation functions between input and
output data to realize a state-space model and to estimate the modal parameters. The identification
procedure starts by estimating the observability matrix Op from the so-called information matrix that
is composed by the correlation functions of input and output data. To determine the observability
matrix Op, one can start by obtaining the matrices of input–output correlation data:
^ xx OT
Rhh ¼Op R (1)
p

^ xx are defined as follows:


where the quantities Rhh and R

Rhh ¼Ryy Ryz R1 T


zz Ryz (2a)

^ xx ¼Rxx Rxz R1 RT :


R (2b)
zz xz

The quantity Rhh is determined from the input (z) autocorrelation matrix Rzz, the input–output
correlation matrix Ryz, and the output (y) autocorrelation matrix Ryy, and it exists only if the input
autocorrelation matrix Rzz is a nonsingular matrix. To obtain the solution for matrix Op,
factorization of Equation (1) is required. In this factorization, the observability matrix is divided into
three matrices using the singular value decomposition as follows:
^ xx OT ½:; 1 : ðp  1Þm ¼ H2N Σ2 VT :
Rhh ½:; 1 : ðp  1Þm ¼ Op R (3)
p 2N 2N

Finally, following the identity in Equation (3), the observability matrix can be obtained as follows:

^ xx OT ½:; 1 : ðp  1Þm ¼ Σ2 VT :
Op ¼ H2N and R (4)
p 2N 2N

Given the observability matrix, the system matrix A can be estimated as follows:
A ¼ Op ð1 : ðp  1Þm; :ÞOp ðm þ 1 : pm; :Þ (5)

where the asterisk (*) denotes the pseudo inverse matrix. The integer p should be chosen such that
matrix Op(m + 1 : pm, :) of dimension (p  1)m × 2N has rank larger than or equal to 2N; hence,
p ⩾ 2N/m + 1.
The modal parameters of the structural system can be estimated by solving the eigenvalues problem
of matrix A as follows:
^ Λ
AΦ¼ e Φ:
^ (6)

Matrices Λe and Φ ^ denote the eigenvalues and eigenvectors of matrix A, respectively. The
eigenvalues and eigenvectors can be real or complex, where in the latter case, they appear in complex
conjugate pairs. The eigenvalues eλi are actually expressed in z-domain and, therefore, can be related to
the modal characteristics of dynamics system using the following transformation:
 
λi ¼ ln eλi =Δt: (7)

After transformation, the natural frequency (ωi) and modal damping ratio (ξ i) can now be estimated
as follows:

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ωi ¼ Reðλi Þ2 þ Imðλi Þ2 ; ξ i ¼ Reðλi Þ=ωi : (8)

Mode shapes matrix in coordinate system is obtained by transforming the eigenvectors in z-domain
into coordinate domain using the output transformation matrix R,

^
Φ¼RΦ: (9)

Detailed information on the system identification algorithm and example of application is given
in [20,21].
To implement the system identification effectively, a set of input–output data was selected.
Responses from triaxial accelerometers located at the basement, the same level as the isolators (i.e.,
AA0S, AA0W, AM0S, and AM0W), were selected as inputs. Meanwhile, accelerations from 13
sensors on the upper stories (i.e., AA1S, AM1W until AM14E) were selected as outputs. Note that
the input and output data are derived from multiaxial accelerations, so that the identified mode shapes
have three-dimensional shape. It should be mentioned that in the time-invariant assumption, the modal
parameters remain constant during a specific time window in which the input–output data were
analyzed. However, considering that the response of base-isolated building may enter nonlinear region
during large excitation, this assumption may not be satisfied throughout the whole responses.
Therefore, a piecewise linear analysis was conducted using moving time window. A shorter time
window was selected for analysis, during which the modal parameters were assumed to remain
constant. For this reason, the total time history responses were divided into several time windows
consisting of 50 s of input and output data. Each time window generates a set of modal parameters,
from which the change of parameters with respect to time can be evaluated. By applying this
procedure, the 600 s main shock response was divided into 12 time windows of input–output data.

7.2. Time-variant system identification


The second system identification is the RLS method using the ARX model. In this technique, the
modal parameters are considered as time-variant quantities. Consider the ARX model given by the
following equation:
AðqÞyðk Þ ¼ BðqÞzðkÞ þ wðkÞ (10)

where y(k) and z(k) are the output and input sequences, respectively, and w(k) denotes the white noise
signal. A(q) and B(q) represent the polynomial functions that include the autoregressive and moving
average coefficients {ai}, {bi} and are defined as follows:

AðqÞ ¼ 1 þ a1 q1 þ ⋯ þ ana qna (11a)

BðqÞ ¼ b1 q1 þ ⋯ þ bnb qnb : (11b)

The quantity q j describes a backward shift operator q jy(k) = y(k  j), whereas na and nb are the
orders of the system output and input, respectively.
The autoregressive and moving average coefficients {ai}, {bi} are evaluated using input and output
sequences z(k) and y(k) recorded on the building. The input is the acceleration response located at the
basement, whereas the outputs are the accelerations on the upper floors. The model parameters are
assumed to be estimated by the following equation:

θðN Þ ¼ RðN Þ1 ΨðN Þ; (12)


where

θðN Þ ¼ ½ a1 ; a2 ; ⋯ana ; b1 ; b2 ; ⋯bnb T (13)

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

1X N
R ðN Þ ¼ λNk φðk ÞφT ðk Þ (14)
N k¼1

1X N
Ψ ðN Þ ¼ λNk yðkÞφT ðkÞ: (15)
N k¼1
The quantity λ denotes the coefficient of forgetting factor, whereas φ(k) is given by the following:

φðk Þ ¼ ½ yðk  1Þ; ⋯; yðk  naÞ; zðk  1Þ; ⋯; zðk  nbÞT : (16)

In the ARX model, the natural frequencies ωj and damping ratios ξ j are evaluated as modulus rj and
argument pj of the poles of polynomial functions by employing the following equations [22,23]:
 
ln 1=rj
ωj ¼ and (17a)
2πξ j Δt
 
ln 1=r j
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ξj ¼     2ffi : (17b)
2
pj þ ln 1=r j
In this study, the RLS–ARX is implemented as a SISO system. The response from the basement
(AM0W) was selected as the input, whereas the accelerations from the upper stories (i.e. AA7S,
AM14W, and AM14E) were used as outputs. Accordingly, three separate SISO systems were
employed, namely AM0W–AA7S system for the annex building, AM0W–AM14W system for the
main building west section, and AM0W–AM14E system for the main building east section. In the
calculation, 5 s of response delay was selected, and λ = 0.99 was chosen as the forgetting factor.

8. RESULTS OF SYSTEM IDENTIFICATION AND COMPARISON WITH ANALYTICAL


MODEL
The MIMO time-invariant SRIM system identification results in three dominant modes: the first
translation mode identified at 0.44–0.55 Hz, the first torsional modes identified at 0.58–0.68 Hz, and
the second torsional mode at 0.85–0.95 Hz. The identified frequencies differ slightly for each time
window as listed in Table IV. And, as illustrated in Figure 8, the first mode has characteristic of large
modal displacement at the isolator layer and small modal displacement of the upper stories resembling

Table IV. Comparison between identified and estimated frequencies of the SIT Building.
Translational Torsional mode Torsional mode
Estimated by FEM mode (main building) (annex building)
Assumed initial stiffness of isolator (Hz) 0.50 0.64 1.05
Assumed large deformation of isolators (Hz) 0.16–0.18 — —

Identified from main shock ω1 (Hz) ξ1 (%) ω2 (Hz) ξ2 (%) ω3 (Hz) ξ3 (%)
Frame 1 (t = 1–50s) 0.61 2.83 0.68 3.79 0.96 3.20
Frame 2 (t = 51–100 s) 0.49 8.87 0.68 5.81 0.97 3.51
Frame 3 (t = 101–150 s) 0.50 11.32 0.64 5.90 0.84 8.70
Frame 4 (t = 151–200 s) 0.44 14.60 0.58 7.22 0.85 13.47
Frame 5 (t = 201–250 s) 0.45 11.62 0.58 11.60 0.90 3.91
Frame 6 (t = 251–300 s) 0.48 8.10 0.63 5.04 0.82 3.93
Frame 7 (t = 301–350 s) 0.49 5.69 0.63 4.99 0.94 3.30
Frame 8 (t = 351–400 s) 0.50 4.19 0.63 5.52 0.94 3.19
Frame 9 (t = 401–450 s) 0.50 3.26 0.65 4.21 0.96 4.64
Frame 10 (t = 451–500 s) 0.51 2.82 0.67 4.39 0.98 3.70
Frame 11 (t = 501–550 s) 0.51 3.28 0.67 3.96 0.97 4.73
Frame 12 (t = 551–600 s) 0.52 3.56 0.67 3.70 0.99 4.49

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

Figure 8. First three modes identified from system identification of the main shock (note: results are obtained from
frame 4, and the blue lines are from sensors on the east side of the main building).

the rigid body motion of the superstructure. This is a typical characteristic of the fundamental base-
isolation mode. The second is the torsional mode with large modal displacement concentrated on the
top corner of the main building. In this mode, the modal displacements of the superstructure of the
annex building are smaller. On the other hand, the third mode is the torsional mode with large modal
displacements concentrated on the corner of the annex building and small modal displacements on the
upper floors of the main building. In both torsional modes, the modal displacements on isolation level
are significantly smaller than they are in the first translation mode.
In design, two scenarios of isolator behavior and their influence on modal parameters were consid-
ered. One is the case of small and medium earthquake in which the initial stiffness of isolators are
considered, and the other one is during large earthquake, where the isolators enter the secondary
stiffness. For the small and moderate earthquake, the isolators are expected to remain on the initial
stiffness, and the estimated natural frequencies are as shown in Table IV. In the design for large
earthquake (maximum input larger than 500 cm/s2), substantial deformation of isolation up to 40 cm
is assumed, and as a result, the first translation mode dominates the building vibration. In such condi-
tion, the first natural frequency of the building could reach as low as 0.16 Hz, and the structure moves
as rigid body where the main deformation concentrates on isolator layer. When large deformation oc-
curs on the isolators during large earthquake, it is expected that the torsional modes of the upper struc-
ture become insignificant and translation mode due to the flexibility of isolation dominates vibration of
the building. Comparing the identified natural frequencies with the estimates from two scenarios, one
can see that the identified frequencies are closer to the response estimates for small and moderate
earthquakes. This goes to say that the present condition represents a case of intermediate level of
ground motion (i.e., maximum input between 100 and 500 cm/s2), where the effect of torsional motion
is still dominant, and this dominant torsional motion results in large peak horizontal acceleration on the
upper structure.

8.1. Variation of natural frequencies and damping ratios during the main shock
A variation of natural frequencies and damping ratios over time during the main shock can be observed
from the results of RLS–ARX and SRIM as described in Figure 9. Note that the results from the time-

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

Figure 9. Time variation of natural frequency and damping ratio of the first three modes identified from the main
shock by RLS–ARX model and SRIM algorithm.

invariant identification agree well with the results of the time-variant identification for the three modes
indicating that time variation of the modal parameters can be captured reasonably by time-invariant
identification when moving time window analysis is applied. The general trend shows that strong
shaking causes a significant reduction in the natural frequency. The natural frequency of the first mode
decreases from about 0.6 Hz in the beginning of the response to about 0.44 Hz during the largest
excitation (t = 150–250 s). A similar reduction in the natural frequencies was also evident from
the second and third modes, although to a lesser extent. In the SRIM identification, reductions
of the natural frequencies were identified during the largest excitation ( frames 3 and 4). The
second and third modes also show the reduction from 0.68 and 0.95 Hz in the beginning of
the response to 0.58 and 0.85 Hz, respectively, during the largest excitation. The frequencies
increase slightly near the end of excitation when excitation amplitude becomes smaller but do
not recover to their conditions prior to the arrival of a seismic wave. The results suggest that
effective linear vibration properties of the isolated structure are strongly dependent on the inten-
sity of the excitation.
While reducing the natural frequency, the strong shaking increases the effective damping ratio.
The increases in damping ratio were observed in the first three modes. The result of RLS–ARX
shows that the damping ratio of the first mode increases from about 5% in the beginning of the
response to about 20% during the largest excitation. Similarly, the damping ratio of the second
and third modes also increases from about 3% and 4% during the beginning of the responses to
about 10% and 15%, respectively, during the largest excitation. Note, however, that the damping
estimates are generally decreasing nearly to their original values at the end of the responses.
The increase in damping was also evident from the results of the time-invariant SRIM
identification, where significant damping for the first mode about 15% was identified from the time
window (t = 150–200 s), which corresponds to the largest excitation. Similar results can be observed
from other modes (Table IV). The increase in damping is related to the performance of isolators and
dampers that become fully engaged during the largest excitation, as shown by the largest isolator
deformation occurring during t = 130–170 s. Large damping in the present earthquake is in
agreement with common practice in Japan, that is, to design a number of dampers so that the
equivalent damping is about 15–20% [13].

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

8.2. Variation of mode shapes during main shock


As previously mentioned, the SRIM identification is a MIMO system, whereas RLS–ARX is a SISO
system, so only the SRIM provides mode shapes information. Figure 8 shows the first three mode
shapes estimated by SRIM from the main shock (time window t = 150–200 s). As shown in the figure,
only the first mode is significantly affected by flexibility of isolators in that the relative modal
displacement between the basement and the first story is large. The other two modes are more affected
by the upper stories modal displacement in that the second mode at 0.58 Hz is dominated by the
torsional response of the main building upper story, whereas the third mode at 0.85 Hz is dominated
by the torsional response of the corner of the annex building.
Figure 10 shows the first mode shapes for the main shock generated from SRIM analysis during
several time frames. One can observe a significant isolator flexibility in the Y direction during the
relatively strong shaking (i.e., frames 3 and 4) and the time-varying characteristics of the mode
shape. Results are shown for frame 2 (t = 51–100 s) when the initial shear waves have arrived and
accelerations are steadily increasing, frame 3 (t = 101–150 s) and frame 4 (t = 151–200 s) when
accelerations are steadily increasing and reaching the maximum amplitude, and frame 6 (t = 251–300 s)
after the principal body waves have passed the site. Results for frames 3 and 4 clearly indicate
increasing relative isolator flexibility as the amplitude of shaking increases. Near the end of the
response, that is, when the amplitude of shaking has decayed, the mode shape resembles that of
the beginning of the response. These results indicate that the modal displacement ratio between
the basement and the first floor is strongly dependent on the amplitude of shaking as a result of
nonlinear isolator response.

9. RELATIONSHIP BETWEEN ISOLATOR DEFORMATION, MODE SHAPES, AND GROUND


EXCITATION
The influences of excitation intensity on the isolator’s relative displacement for the first three dominant
modes during the main shock are shown in Figure 11. In this figure, abscissa quantifies the ground

14 14

9 9
Floor

7 7

4 4

1 1

0 0

0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1


Modal Displacement Modal Displacement
Ratio (Y Direction) Ratio (X Direction)

Figure 10. Time variation of the first mode during the main shock.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

1 1 1
(a) (b) (c)
0.8 0.8 0.8

0.6 0.6 0.6

RIMD

RIMD
RIMD
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
0 50 100 150 0 20 40 60 80 100 0 100 200 300
Sa (cm/s2) Sa (cm/s2) Sa (cm/s2)

Figure 11. Influence of the shaking intensity on the isolator’s relative displacement during the main shock for the
(a) first mode, (b) second mode, and (c) third mode.

motion amplitude in terms of response spectra of acceleration AM0W for 5% damping. The response
spectra of horizontal ground acceleration were calculated at the frequencies that correspond to the first
three modes (i.e., 0.45, 0.58, and 0.85 Hz). In such a way, we can quantify the frequency components
of the ground motion that are relevant to the natural frequencies of the structure. It is noted, however,
that the response spectra amplitudes have linear relationship with the root mean square (RMS) of
ground motion in that large accelerations (in terms of RMS) result in large acceleration spectra,
although the linear trend of each mode is not necessarily the same.
To measure the influence of shaking intensity on isolator displacement, an index called relative
isolator modal displacement (RIMD) is utilized. For each mode shape, the RIMD measures the
relative modal displacement at isolator level with respect to the modal displacement at the roof
level. The mode shapes used here are derived from the time-invariant system identification of
the main shock using moving window scheme as explained earlier. Note that when the isolator
displacement is large, the RIMD value will become closer to unity, and for an ideal rigid
superstructure, the RIMD value is equal to one. The RIMD index also confirms the influence
of isolator displacement on the global modal displacement for each mode. As can be seen in
the figure, only the first mode shows a clear trend in which the RIMD values increase as the
spectra acceleration increase for that mode increases. The other modes, on the other hand, do
not share the same tendency. These results indicate that only the first translation mode is
significantly affected by the isolators’ deformation.

10. ISOLATOR’S BEHAVIOR DURING LARGE EARTHQUAKE


Figure 12(a) shows the orbit displacement of the base slab above the isolation layer relative to the slab
below the isolation layer. The maximum relative displacement about 10 cm was recorded in the X
direction by sensor DM0E located near the east corner of the main building. The maximum
displacement in the X direction is about twice than the displacement in the y direction. The design

400
cumulative
disp(cm)

300
Y-Relative Displacement(cm)

10
8
(a) 200
100
6 (b)
4 0
0 50 100 150 200 250 300 350 400 450
2
Time(s)
0
400
-2
cumulative
disp(cm)

-4 300
-6 200
-8 100
-10 (c)
0
-10 -5 0 5 10 0 50 100 150 200 250 300 350 400 450
X-Relative Displacement(cm) Time(s)

Figure 12. (a) Orbit motion of isolator displacements shows relative motion between the basement and the
first floor recorded by displacement meter at DM0E; cumulative isolator displacement measured at (b) DM0E-X
direction and (c) DM0E-Y direction.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

maximum deformation of NRB is set to about 250% of the shear strain or equal to 49 cm, which means
that the maximum recorded displacement corresponds to 20.4% of the design maximum shear strain.
Figure 12(b,c) shows the cumulative story displacement of the isolated layer recorded by sensor
DM0E. The maximum cumulative displacement is 415 cm, which is considerably large considering
that the maximum recorded shear strain is only 20.4% of maximum design. The long duration of the
earthquake is thought to be the reason behind this large cumulative story displacement.
To evaluate the time variation of the isolator during the main shock, a force–deformation
relationship of the isolator system is approximated. The force is estimated from the total acceleration
immediately above the isolator by filtering out the frequency components larger than 1 Hz that is the
frequency range that corresponds to the frequency contents of the first three dominant modes. The
filtered total acceleration is assumed to be proportional to the total shear force imparted to the isolators,
whereas the relative story displacement recorded by the displacement sensor provides the deformation
of the isolator system. By employing the two quantities, one can plot an ‘equivalent hysteresis loops’ of
the isolator system from which the time variation of the isolator effective stiffness can be inferred.
An example of variations of hysteresis characteristics is shown in Figure 13 for several time
windows. The straight line on each figure connects the maximum and minimum loads to show the
approximate effective stiffness of the isolator layer. As illustrated in the figure, the effective stiffness
is relatively high before the passage of the shear wave (t = 100–130 s) and becomes smaller with the
arrival of the largest amplitude of shear wave that causes the increase in the relative displacement of
the isolator layer. The effective stiffness remains small even after the passage of the shear wave after
200 s, indicating that the system has been softened from the initial condition. The results are consistent
with the time variation of the natural frequencies estimated by the system identification. Moreover, as
shown in the figure, the largest isolator deformation occurs during t = 130–170 s, which is the time
when the hysteresis loop reaches the maximum area signifying the increase of energy dissipation. This
condition is consistent with the fact during this period damping ratio also reaches the maximum value
as shown by the results of system identifications.

11. LONG-TERM VARIATION OF MODAL PARAMETERS


Since being installed in September 2010, more than 140 seismic events have been recorded by the
monitoring system until December 2012. Most of the earthquakes were of small and moderate scale
(JMA seismic intensity of 3 or less), and only five events including the March 11, 2011, main shock
have seismic intensity equal or larger than 4. With relatively large earthquake database, we can eval-
uate the behavior of the building and observe the trend in the modal parameters with respect to the level
of seismic excitation and their long-term variations. For this purpose, the linear time-invariant system

100 100 100 100


t:100-130s t:170-200s t:200-230s
Acc(cm/s2)

50
Acc(cm/s2)

50 50
Acc(cm/s2)

50

0 0 0 0

-50 -50 -50 -50

t:100-170s
-100 -100 -100 -100
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
X-Relative disp(cm) X-Relative disp(cm) X-Relative disp(cm) X-Relative disp(cm)
100 100 100 100
t:230-260s t:260-300s t:300-400s t:400-500s
Acc(cm/s2)

50 50
Acc(cm/s2)

50
Acc(cm/s2)

50

0 0 0 0

-50 -50 -50 -50

-100 -100 -100 -100


-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
X-Relative disp(cm) X-Relative disp(cm) X-Relative disp(cm) X-Relative disp(cm)

Figure 13. Variation of hysteresis characteristics of isolator during the main shock recorded from displacement
meter DM0E in the X direction along the time of the main shock.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

identification (SRIM) was utilized. The identification was performed separately for each input–output
data set, and the results were evaluated with respect to the RMS of the input amplitude, which is the
acceleration from sensor AM0W. In the implementation of the SRIM algorithm, the moving time
window of 50 s was employed in order to capture the variation of the modal parameters in a single
seismic event.
Figure 14 shows the sequential long-term variation of the first mode. Note that we focus the
discussion on the first mode, because it is the mode that is significantly affected by the isolator
behavior. As shown in the figure, for one seismic event, the natural frequency reaches the minimum
value in the beginning of the record that is during the largest excitation and increases gradually toward
the end. A similar behavior is repeated in the next earthquake events. These repetitions result in the
sharp turns in alternating directions on the natural frequency as time progresses. The minimum
frequency of the first mode for the entire monitoring record is 0.44 Hz, observed during the main shock
of the March 11, 2011, earthquake. Prior to that event, the lowest observed frequency was slightly
higher 0.55 Hz, recorded during the foreshock event on February 9, 2011. Note that the natural
frequency of the first mode was higher before the foreshock event with the average of 0.6 Hz. The main
shock of the March 11, 2011, earthquake was followed by several aftershocks with moderate level of
excitations, and in the course of 2 months after the main shock, the building experienced frequent
earthquakes with JMA seismic intensity 3 or 4. One can note from the figure that it takes about
2 months (from March 2011 to May 2011) for the natural frequency to return to its pre-main shock
value (i.e., the lowest frequency recorded during the foreshock event).
Figure 15(a,b) shows the trend of the natural frequency of the first mode plotted with respect to the
excitation amplitude. As expected, the frequency decreases as the input increases indicating the
increase in flexibility as a result of the isolation system. The decrease in the natural frequency signifies
Natural Frequency (Hz)

0.6

0.5
(a)
0.4
9/10 2/11 03/11 4/11 5/11 6/11 7/11 8/11 9/11 10/11 12/11
Time (month/year)
Natural Frequency (Hz)

0.6

0.5

(b)
0.4
1/12 2/12 3/12 4/12 5/12,6/12 7/12 8/12 9/12,10/12 11/12 12/12
Time (month/year)
Damping Ratio (%)

15

10

0
9/10 2/11 03/11 4/11 5/11 6/11 7/11 8/11 9/11 10/11 12/11
(c) Time (month/year)
15
Damping Ratio (%)

10

0
1/12 2/12 3/12 4/12 5/12,6/12 7/12 8/12 9/12,10/12 11/12 12/12
(d) Time (month/year)

Figure 14. Long-term time sequential variation of the first mode: (a) natural frequency from September 2010 to
December 2011, (b) natural frequency from January 2012 to December 2012, (c) damping ratio from September
2010 to December 2011, and (d) damping ratio from January 2012 to December 2012. Note the circle in (a)
showing the variation of frequency in one earthquake event.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

0.8 15

Damping Ratio(%)
Before March 11,2011 Before March 11,2011

Frequency (Hz)
Mainshock Mainshock
0.7 Aftershocks(Mar 2011) 10 Aftershocks(Mar 2011)

0.6
5
0.5

0.4 0 (c)
(a) 10-2 10-1 100 101 10-2 10-1 100 101
RMS of Input Acc (cm/s2) RMS of Input Acc (cm/s2)
0.8 15

Damping Ratio(%)
Apr-May 2011
Frequency (Hz)

Apr-May 2011
Jun-Dec 2011 Jun-Dec 2011
0.7 Jan-Dec 2012
Jan-Dec 2012 10

0.6
5
0.5

0.4 0
(b) 10-2 10-1 100 101 (d) 10-2 10-1 100 101
2)
RMS of Input Acc (cm/s RMS of Input Acc (cm/s2)

Figure 15. Long-term variation of natural frequency and damping ratio of the first mode with respect to input am-
plitude. (a) Natural frequency for earthquakes before March 11, 2011, during the main shock, and aftershocks, (b)
natural frequency for earthquakes in 2011 and 2012, (c) damping ratio for earthquakes before March 11, 2011,
during the main shock, and aftershocks, and (d) damping ratio for earthquakes in 2011 and 2012 (note: aftershocks
in (a) and (c) are all earthquakes that come from Tohoku area (plate boundary area) that occur between March 11
and March 31, 2011).

the increase in flexibility of the structure as a direct consequence of base-isolation function during
earthquakes. An interesting feature observed in these figures is that for the same level of excitation,
the natural frequencies of the first mode during aftershocks were lower than those observed before
the main shock.
The prolonged effect of reduction in the natural frequency is thought to be caused by the stiffness
reduction of the isolator system as results of the property known as the Mullins’ effect [24], which
is a temporary and recoverable reduction in modulus due to cyclic straining history on the rubber
property of isolators. The reduction in effective modulus with cycling of elastomeric isolator system
depends on the material properties of elastomer and the strain history. Considering the high amplitude
and long duration of excitation experienced during the main shock of March 11, 2011, earthquake, and
many large aftershocks that follow, it is understandable that recovery of effective modulus takes longer
time since all events occurring within a short time span affect the strain history of the isolator system.
Damping values obtained from system identification analyses suggest a wide range between 1%
and 15%. From Figure 14(c,d), it can be seen that damping reaches the maximum value in the
beginning of the record, that is, during the largest excitation, and decreases gradually toward the
end of the record, as opposed to the trend in the natural frequency. The largest modal damping
of about 15% was observed during the main shock. The dependence of damping to excitation
intensity can be observed clearly from Figure 15(c,d), where the damping ratio increases as the
input acceleration increases. The increase in damping is attributed to the base-isolation system that
becomes fully engaged during large excitation. Moreover, energy dissipation mechanism in base
isolation also includes vibration absorption devices such as u-shaped steel dampers and lead
damper, which increase the overall structural damping when the isolation system begins to function
completely during large earthquakes.

12. CONCLUSIONS
Strong motion recordings of an asymmetric base-isolated building located in Tokyo Bay area during
the March 11, 2011, Great East Japan (Tohoku) Earthquake and various levels of earthquakes that
follow are investigated in this paper. On the basis of response analysis and system identification, the
following conclusions are drawn:

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
ASYMMETRIC BASE-ISOLATED BUILDING IN 2011 GREAT EAST JAPAN EARTHQUAKE

1. The recorded input excitations in the 2011 Great East Japan (Tohoku) Earthquake are still below
the seismic design ground motions. The building experienced intense shaking for about 10 min
with peak ground acceleration 0.80–1.40 m/s2 but survived the earthquake without structural
damage.
2. Seismic response of the building is dominated by three vibration modes: (i) the first translational
mode characterized by large modal displacement at the isolator layer and rigid body movement
of the upper stories; (ii) torsional mode with dominant modal displacement at the corner of the
main building; and (iii) torsional mode with dominant modal displacement at the corner of the
annex building. During the main shock, the first mode shows large lateral modal displacement
at the isolator layer indicating that base-isolation system has functioned. This result is supported
by recorded relative displacement between isolation and the first story that reached the
maximum amplitude of 10 cm.
3. The building experienced frequency reduction and increase of damping during the main shock.
Frequency reduction indicates the increase in flexibility due to base-isolation system, as
demonstrated by reduction of the first mode from about 0.6 Hz at the beginning of the response
to about 0.44 Hz during the largest excitation. Meanwhile, the increase in damping is a
consequence of shear deformation in the isolation system and functioning of the damper system
during the large excitation. By employing time-variant and time-invariant system identifications
in an intra-event or using various levels of earthquakes, analysis shows that frequency reduction
and increase in damping are strongly related to the excitation intensity.
4. Despite the functioning of base-isolation system, strong response amplifications on the upper
stories were recorded during the main shock. Peak acceleration on the top of the annex building
is nearly 300 cm/s2, or about three times of the basement acceleration. Similarly, a strong
amplification was also observed on the main building, with peak acceleration on the 14th floor
of the west corner of about 236 cm/s2, which is 2.5 times of the ground acceleration. Results of
frequency analysis and system identification of a superstructure reveal that unlike a typical base-
isolated building where the first sliding mode dominates the response, the torsional modes
dominate the horizontal accelerations of the superstructure, and they cause amplification of up-
per stories. Analysis also shows that the torsional modes were not significantly influenced by
performance of the isolation system. Moreover, natural frequencies of the torsional modes coin-
cided with the frequencies of ground motions that have the largest energy. Combinations of
these factors are thought to be the reason behind strong amplification of the superstructure.
5. Significant response amplification could be critical to nonstructural elements or sensitive
equipment on the building and should be carefully considered in the design. At the current state,
the torsional effect is not very harmful to the building, and no nonstructural components were
damaged during the earthquake. However, further study on the possibility of larger and more
significant consequence in case of a larger earthquake is needed. And in an extreme case,
countermeasures such as additional damping on the upper floor could be considered to suppress
excessive vibration.
6. Using over 140 recorded earthquakes between 2010 and 2012, we observe that for the same
level of excitation, the natural frequencies of the first mode remain low for about 2 months after
the main shock. The prolonged effect of reduction in the natural frequency is thought to be
caused by the stiffness reduction of the isolator system as results of the temporary and
recoverable reduction in modulus due to cyclic straining history on a rubber. The high level
and long duration of excitation experienced during the main shock of the March 11, 2011,
earthquake and occurrence of many moderate level aftershocks that follow closely after the main
shock are considered to influence the strain history of the isolator system and prolong its
stiffness reduction.

ACKNOWLEDGEMENTS

This research is financially supported by Strategic Research Program of Japan Science and Technology Agency
(JST) in the area of advanced integrated sensing technology (CREST) under the title ‘Risk Monitoring and
Disaster Management of Urban Infrastructure’ (principle investigator: Yozo Fujino). The authors gratefully

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc
D. M. SIRINGORINGO AND Y. FUJINO

acknowledge this support. The authors also gratefully acknowledge the valuable discussions on monitoring system
with Dr. Katsuaki Konno of Shibaura Institute of Technology and Dr. Takafumi Nishikawa of Nagasaki
University. The assistance from NS Consultant Ltd. on the initial design documents and drawing and from
Mr. Hisashi Yamasaki, former graduate student on data analysis, is also highly appreciated. The opinions, findings,
and conclusions expressed in this material are those of the authors and do not necessarily reflect those of the
abovementioned institutions.

REFERENCES

1. Higashino M, Okamoto S. Response Control and Seismic Isolation of Buildings – CIB Series. Taylor & Francis Group:
New York, NY, USA, 2006.
2. Japan Society of Seismic Isolation. Report of response-controlled buildings by the investigation committee, 2012.
3. Kani N, Ogino N, Kitamura Y, Kitamura H, Fukazawa Y. Performance of response-controlled building during the huge 2011
earthquake. Proc. 13th World Conf. on Seismic Isolation, Energy Dissipation and Active Vibration Control of Structures,
Sendai, Japan, 2013.
4. Lee DM. Base isolation for torsion reduction in asymmetric structures under earthquake loading. Earthquake Engineering &
Structural Dynamics 1980; 8(4):349–359.
5. Eisenberger M, Rutenberg A. Seismic base isolation of asymmetric shear buildings. Engineering Structures 1986;
8(1):2–8.
6. Pan TC, Kelly JM. Seismic response of torsionally coupled base isolated structures. Earthquake Engineering & Structural
Dynamics 1983; 11(6):749–770.
7. Nagarajaiah S, Reinhorn AM, Constantinou MC. Torsion in base-isolated structures with elastomeric isolation systems.
Journal of Structural Engineering, ASCE 1993; 119(10):2932–2951.
8. Nakamura T, Suzuki T, Okada H, Takeda T. Study on base-isolation for torsional response in asymmetric structures under
earthquake motion. Proc. 9th World Conf. on Earthquake Engineering, Tokyo, Japan, 1988; 675–680.
9. Hwang JS, Hsu TY. Experimental study of isolated building under triaxial ground excitations. Journal of Structural
Engineering, ASCE 2000; 126(8):879–886.
10. Papageorgiou AS, Lin BC. Study of the earthquake response of the base-isolated Law and Justice Center in Rancho
Cucamonga. Earthquake Engineering & Structural Dynamics 1989; 18(8):1189–1200.
11. Nagarajaiah S, Xiaohong S. Response of base-isolated USC hospital building in Northridge earthquake. Journal of
Structural Engineering, ASCE 2000; 126(10):1177–1186.
12. Konno K, Nishikawa T, Fujino Y. Estimation of Lateral and Rotational Vibrations of a Base-Isolated Building during the
2011 Great East Japan Earthquake. Journal of Japan Association for Earthquake Engineering 2013; 13(3):314–329 (in
Japanese).
13. Pan P, Zamfirescu D, Nakashima M, Nakayasu N, Kashiwa H. Base-isolation design practice in Japan: introduction to the
post-Kobe approach. Journal of Earthquake Engineering 2005; 9(01):147–171.
14. Chaudhary MTA, Fujino Y. System identification of bridges using recorded seismic data and its application in structural
health monitoring. Structural Control and Health Monitoring 2008; 15(7):1021–1035.
15. Arici Y, Mosalam KM. Modal identification of bridge systems using state-space methods. Structural Control and Health
Monitoring 2005; 12(3):381–404.
16. Siringoringo DM, Fujino Y. Observed dynamic performance of the Yokohama-Bay Bridge from system identification using
seismic records. Structural Control and Health Monitoring 2006; 13(1):226–244.
17. Hong AL, Betti R, Lin CC. Identification of dynamic models of a building structure using multiple earthquake records.
Structural Control and Health Monitoring 2009; 16(2):178–199.
18. Chu SY, Lo SC. Application of the on-line recursive least-squares method to perform structural damage assessment.
Structural Control and Health Monitoring 2011; 18(3):241–264.
19. Minami Y, Yoshitomi S, Takewaki I. System identification of super high-rise buildings using limited vibration data during
the 2011 Tohoku (Japan) earthquake. Structural Control and Health Monitoring 2013; 20(11):1317–1338.
20. Juang JN. System realization using information matrix. Journal of Guidance, Control, and Dynamics 1997; 20(3):492–500.
21. Siringoringo DM, Fujino Y. System identification applied to long-span cable-supported bridges using seismic records.
Earthquake Engineering & Structural Dynamics 2008; 37:361–386.
22. Safak E. Adaptive modeling, identification, and control of dynamic structural systems. I: theory. Journal of Engineering
Mechanics ASCE 1989; 115(11):2386–2405.
23. Safak E. Adaptive modeling, identification, and control of dynamic structural systems. II: applications. Journal of Engineering
Mechanics ASCE 1989; 115(11):2406–2426.
24. Clark PW, Aiken ID, Kelly JM. Experimental Studies of the Ultimate Behavior of Seismically-Isolated Structures, UCB/
EERC-97/18. Earthquake Engineering Research Center, University of California: Berkeley, CA, USA, 1997.

Copyright © 2014 John Wiley & Sons, Ltd. Struct. Control Health Monit. (2014)
DOI: 10.1002/stc

You might also like