Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

Contents lists available at ScienceDirect

Journal of Wind Engineering

and Industrial Aerodynamics
journal homepage: www.elsevier.com/locate/jweia

Wind-induced responses and dynamics characteristics of an asymmetrical MARK

base-isolated building observed during typhoons
⁎
Dionysius M. Siringoringo , Yozo Fujino
Institute of Advanced Sciences, Yokohama National University, 79-1 Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan

A R T I C L E I N F O A BS T RAC T

Keywords: Responses of an asymmetrical base-isolated building during the passage of Typhoon Roke 2011, Typhoon
Base-isolated building Guchol 2012 and a bomb cyclone strong-wind event on April 3, 2012 were investigated in this paper. Wind-
Typhoon-induced full-scale measurement induced accelerations of the building were recorded by dense array permanent monitoring system and analyzed
Building vibration in detailed by spectra analysis, time-frequency wavelet analysis and system identification. Results of analysis
Typhoon response
revealed that wind-induced responses of the buildings were dominated by the first mode which is the
System identification
fundamental flexural mode coupled with torsion of the upper stories. Building responses demonstrated non-
linear characteristics of natural frequencies, damping ratio and mode-shapes. Natural frequencies of funda-
mental modes decreased with the increase of acceleration amplitude, with the maximum reduction of 8–10% of
natural frequencies observed during the peak wind speed. Damping estimates of the fundamental modes
increased with the increase of accelerations, while characteristics of mode-shapes were also found to be
dependent on the wind speed. Further analysis revealed that non-linear characteristics of building responses are
related to isolator deformations caused by wind-induced static force.

1. Introduction base-isolated building (Japan Society of Seismic Isolation JSSI, 2009).

Base isolation is a popular seismic mitigation technique for build-
ings and has been applied to more than 150 tall buildings with the
height of 60 m or more (Takenaka et al., 2004). As the number of base-
isolated tall buildings increases, consideration for wind-resistant de-
sign has become more significant than before. Unlike earthquake, wind
load has two different main characteristics, namely, long duration of
excitation and having static component (mean force). Because of
isolator characteristics, natural period of a base-isolated building is
longer than ordinary building resulting in larger wind-induced dis-
placement of the upper story.
There have been some studies on the behavior of base-isolated
buildings subjected to wind excitation. Most of them focus on modeling
building response to wind excitation (Kareem, 1997; Chen and
Ahmadi, 1992; Henderson and Novak, 1989), addressing the service-
ability aspect of wind-induced vibration (Liang et al., 2002) and
comparing wind-induced response and earthquake response of base-
isolated building (Vulcano, 1998). Several studies have experimentally
investigated behavior of isolator component especially creep behavior
that becomes an important consideration for relatively long duration of
wind excitation (Takenaka et al., 2004, 2009; Kochiyama et al., 2004).
Result of the study is included in the design guideline for wind effect of
Despite the efforts to investigate wind-response of base-isolated
building through modeling and simulation, there have been very few
studies that are based on real observation of full-scale base-isolated
building especially due to strong wind excitation (Yasui et al., 2002;
Sato et al., 2012). As comparison, there have been numerous studies on
the wind-response of tall conventional fixed base buildings under
strong wind events such typhoons (Campbell et al., 2005; Li et al.,
2000, 2003). Based on these studies, several conclusions are obtained
such as the amplitude-dependent characteristics of damping and
natural frequency (Li et al., 2000, 2003), and comparison between
predicted and measured wind-induced response of conventional fixed
base tall buildings (Jeary and Ellis, 1983). Recently, more base-isolated
buildings have been instrumented with permanent monitoring system
to observe their performance during large earthquake. The monitoring
system can also be used to capture the wind-induced responses during
strong wind or typhoon.
The objective of this study is to investigate wind-induced responses
and dynamic characteristics of a base-isolated building from full-scale
structural responses recorded during the passages of typhoons. At first
the paper outlines measurement program and describes the main
characteristics of typhoon-induced building responses in detailed using
temporal, spectral and time-frequency wavelet analysis. Further,

⁎
Corresponding author.
E-mail addresses: dion@ynu.ac.jp (D.M. Siringoringo), fujino@ynu.ac.jp (Y. Fujino).

http://dx.doi.org/10.1016/j.jweia.2017.04.020
Received 22 July 2016; Received in revised form 26 February 2017; Accepted 26 April 2017
0167-6105/ © 2017 Elsevier Ltd. All rights reserved.
D.M. Siringoringo, Y. Fujino
Fig. 1. (a) Photo of SIT Building, (b) dimension, orientation and building's plan view with layout of isolators at basement level, (c) Layout and location of monitoring system (note: NRB:
natural rubber bearing, SB: sliding bearing, LD: lead damper and UD: U-shaped steel dampers).
dynamic characteristics of the building obtained from system identifi-
cations are described with respect to wind speed and direction. Finally,
the paper provides discussions on interpretation of the results and
compares the results with design condition and observations from
seismic records.
2. Description of base-isolated building and monitoring
system
The object structure is a mid-rise asymmetrical base-isolated
building named SIT Building located in Tokyo-Bay area. The building
consists of two parts: fourteen-story main building (M) and seven-story
annex building (A) (Fig. 1). Both buildings are of braced steel frames
and connected at the corner by elevator shaft forming an L-shaped
asymmetric structure. Dimension of the main building is 97.2 m long,
43.2 m wide and 67.5 m high with 19 bays in longitudinal direction and
6 bays in transverse direction. Annex building is 81 m long, 21.6 m
wide and 31.2 m high with 16 bays in longitudinal direction and 3 bays
in transverse direction. The main building has vertical opening in the
middle, starting from the second to the seventh floor, which divides the
building into the west section (MW) and east section (ME). Concrete
slab connects both buildings at the basement level and on top of the
slab the isolation systems are placed.
The main building's orientation and axis direction are depicted in
Fig. 1.(b). The building orientation is such that the x-axis deviates 42°
clockwise from the north. Due to its unique shape, the building has
significant eccentricity. The eccentricity ratio is defined as the distance
between center of stiffness and center of mass with respect to building
length in the corresponding direction. The ratio is presented separately
for each floor and each building section in x- and y-directions. The
eccentricity in y-direction is generally larger than in x-direction for
annex and main building; and the eccentricity ratios in some floors are
larger than 10%.
The building is base isolated with total of 146 units of isolator and
damper system consisting of 59 units of natural rubber bearing (NRB),
184
Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197
26 units of sliding bearing (SB), 28 units of lead damper (LD) and 33
units of U-shaped steel dampers (UD) (Fig. 1). Detail explanation of the
isolation system is given elsewhere (Siringoringo and Fujino, 2015a).
A permanent monitoring system consisting of 21 triaxial acceler-
ometers and four displacement-meters is deployed on the building
since 2010 as shown in Fig. 1(c). The sensors nomenclature is
organized as follow, the first index denotes sensor type (A: acceler-
ometer and D: displacement-meter), the second index represents the
building (M: main and 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 and W: West). Horizontal accelerations were
oriented in the x and y directions per building as opposed to the
commonly used North-South (NS) and East-West (EW) directions. The
x-axis coincides with the main building's weak axis, while the y-axis
coincides with the main building's strong axis. To avoid confusion, the
terms x-axis and y-axis according to Fig. 1. will be used throughout the
paper.
Eighteen accelerometers are placed on the building and three on the
ground outside the building. It should be mentioned, however, that
displacement-meters that measure relative displacement between base-
ment and the first floor are activated by an underground trigger that
functions only when ground motion with acceleration larger than 5 cm/
s2 occurs as in the case of earthquake. These sensors were not triggered
by typhoon-induced vibration; therefore, in this paper we present
analysis only from acceleration responses of upper stories. The accel-
erometers are the servo type with resolution of 0.01 cm/s2 and
measurement range ± 2000 cm/s2. Sensor network is connected to a
GPS controller unit to provide global reference position and to
synchronize time recording with 4 ms maximum delay time among
sensors (Siringoringo and Fujino, 2015a). The monitoring system
records building accelerations at 100 Hz sampling frequency and stores
the data in a server for further analysis.
Wind records were provided by Edogawa weather station, the
closest weather station located approximately 2 km southeast of the
building, since onsite wind measurement system was unavailable on
D.M. Siringoringo, Y. Fujino
Fig. 2. Ten-minute averaged and instantaneous peak wind speed and average wind direction for the case of (a) Typhoon Roke, September 5, 2011 (24 h), (b) Bomb-cyclone, April 3,
2002 (24 h), (c) Typhoon Guchol, June 19–20, 2012 (48 h). Note: arrows in the building inset denote the dominant wind direction with respect to building's orientation.
the building at the time of the typhoons and storm event. Considering
topography of the area surrounding the building and the proximity of
Edogawa weather station, the collected wind information is deemed
reasonable to reflect overall wind characteristics affecting the building.
The wind records consist of 10-min average wind speed and direction,
as well as maximum instantaneous wind speed within ten-minute
duration. The wind data was recorded at the standard height of 10.2 m
from sea level (Japan Meteorological Agency (JMA) (2016)).
3. Description of typhoons and strong wind event
In this study, the building responses were recorded during two
typhoons and one low pressure storm event called bomb cyclone. The first
typhoon is the typhoon Roke 2011 or the 15th typhoon of 2011 in Japanese
typhoon counting system that passed through the building location on
September 5th, 2011. The typhoon recorded the lowest instantaneous
mean sea-level pressure and the maximum recorded wind speed from all
meteorological stations are 940 hPa and 43.7 m/s, respectively. At
Edogawa weather station, the maximum recorded instantaneous wind
speed was 41 m/s (Fig. 2.a). The dominant wind direction during the
largest wind speed of typhoon Roke was between southwest and southeast
direction which was nearly parallel to the building's x-axis.
The second typhoon is Typhoon Guchol 2012 or the 4th Typhoon of
2012 that passed through area near the building on June 19, 2012
afternoon until June 20 evening. The lowest instantaneous mean sea-
level pressure and maximum recorded speed of the typhoon from all
meteorological stations are 930 hPa and 51 m/s, respectively. Typhoon
Guchol recorded the maximum wind speed 32.6 m/s at Edogawa
weather station, with dominant winds direction were from south and
southeast that was almost perpendicular to building's y-axis. Fig. 2(c)
illustrates the 10-min mean and maximum instantaneous wind speed
for typhoon Guchol.
The third event is a wind storm caused by an extra-tropical cyclone
also known as bomb cyclone. The storm occurred on April 3, 2012 with
the recorded maximum instantaneous wind speed of 51.1 m/s at
Tobishima Sakata, Yamagata Prefecture. The highest measured wind
speed at the Edogawa weather station was 33.4 m/s, with the dominant
wind direction during the storm was between southwest to south
185
Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197
direction. Variations of 10-min mean and maximum instantaneous
wind speed during the storm are shown in Fig. 2(b).
4. Characteristics of building's dynamic responses
4.1. Response amplitude with respect to wind speed
Acceleration responses recorded at the east corner of the 14th floor
of main building (sensor AM14E) are shown in Fig. 3 for typhoon Roke,
Guchol and bomb cyclone. The records were obtained during 10-min
duration of the largest wind speed of the respective events. For the
typhoon Roke 2011 the peak acceleration recorded by sensor AM14E
was 4.74 cm/s2 at 18:29 Japan Standard Time (JST) of Sept 5, 2011
caused by the maximum instantaneous wind speed of 41 m/s. The peak
acceleration due to bomb-cyclone was about 2.31 cm/s2 occurring at
the instantaneous time of the maximum wind speed 33.4 m/s. In the
case of typhoon Guchol, the typhoon gained strength rather slowly but
lasted for two days. The increase in building acceleration was noted at
19:00 JST in the first day June 19, 2012 and reached the maximum
around midnight with the peak building acceleration of 3.73 cm/s2 due
to maximum wind speed of 32.6 m/s.
Variations of 10-min root-mean square (RMS) and peak building
accelerations during the 10-min for the entire day of typhoon are presented
in Fig. 4. The largest accelerations for the three events were recorded by
sensor AM14E in x-direction, and the amplitude varied between 4.5 and
6 cm/s2. In y-direction, the peak accelerations were smaller between 2 and
4 cm/s2. By comparing the daily variation of wind speed as shown in Fig. 2
and the building acceleration, one can observe that as expected, building
accelerations increase with increasing wind speed. It should be mentioned,
however that there is a sudden increase in building acceleration around
21:15 JST of typhoon Roke. During this time instance, the peak and RMS
of accelerations are very different with the ones that precede and follow
after, whereas the wind speed at this time instant is smaller than before.
This spike in acceleration was not caused by wind response but due to a
small earthquake Mw 4.3 about 200 km to the northeast side of the
building.
Fig. 5 illustrates the relationships between the peak (Pacc) and RMS
values (σacc) of 10-min x and y-direction acceleration response with respect
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

5
Acc (cm/s ) 2
Peak Acceleration : 4.74 cm/s
2

-5
(a) 18:20 18:21 18:22 18:23 18:24 18:25 18:26 18:27 18:28 18:29 18:30

5
Acc (cm/s )
2

-5
18:20 18:21 18:22 18:23 18:24 18:25 18:26 18:27 18:28 18:29 18:30
(b)

2
Acc (cm/s )
2

-2 2
Peak Acceleration : 2.31cm/s
(c) 19:40 19:41 19:42 19:43 19:44 19:45 19:46 19:47 19:48 19:49 19:50

2
Acc (cm/s )
2

-2
19:40 19:41 19:42 19:43 19:44 19:45 19:46 19:47 19:48 19:49 19:50
(d)

5
Acc (cm/s )
2

0
2
Peak Acceleration : 3.73cm/s
-5
00:00 00:01 00:02 00:03 00:04 00:05 00:06 00:07 00:08 00:09 00:10
(e)

5
Acc (cm/s )
2

-5
00:00 00:01 00:02 00:03 00:04 00:05 00:06 00:07 00:08 00:09 00:10
(f) Time (Hour)

Fig. 3. Acceleration responses recorded on the top of main building (AM14E) for: (a) typhoon Roke X-direction, (b) typhoon Roke Y-direction, (c) bomb-cyclone X-direction, (d) bomb-
cyclone Y-direction, (e) typhoon Guchol X-direction, (f) typhoon Guchol Y-direction.

to 10-min mean wind speed for all three events. It is evident from this
figure that the increases in acceleration responses were associated with the
increase in the approaching mean wind speed. The regression curves of
acceleration responses presented in these figures are the fitted power
regression curve with 95% confidence bounds and expressed as:
σacc = a1 U b1
Pacc = a2 U b2
where the values of a1, a2, b1 and b2 as the results of curve fitting are given
in Table 1.
The figures with the constant of regression values of a1, a2, b1 and b2
listed in Table 1 demonstrate that accelerations in x-direction that is the
weak-axis of the main building increase more significantly than accelera-
tions in y-direction. Note that the peak accelerations on the main building
in x and y directions are due to south to southwest winds, which is the
dominant wind directions for the three events. Looking at the wind
(1)
directions and building accelerations as shown by the values of coefficient
of regression during the peak of typhoons, it can be concluded that
(2)
typhoon-induced responses of the main building were mainly the along-
wind responses.

Fig. 4. Ten-minute averaged and instantaneous peak of floor acceleration of the main building (14th Floor) for the case of (a) Typhoon Roke, September 5, 2011, (b) Bomb-cyclone,
April 3, 2002, (c) Typhoon Guchol, June 19–20, 2012.

186
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

Fig. 5. Relationship between 10-min mean wind velocity and 10-min mean and peak building accelerations recorded at the 14th floor of main building in X direction (AM13E-X) for: (a)
Typhoon Roke, (b) Bomb-cyclone, (c) Typhoon Guchol; and in Y direction (AM13E-Y) for (d) Typhoon Roke, (e) Bomb-cyclone, (f) Typhoon Guchol.

Table 1
Parameters of regression curved fitted in Eqs. (1) and (2). artificially offset to show that peaks of all spectra located in the same
frequency. The figures reveal clearly the dominant influence of low
Typhoon Sensor Root mean squared of Peak acceleration (Pacc) frequency peak at 0.58 Hz for the east corner of the main and annex
acceleration (σacc)
building in both x and y directions. Whereas for the west corner of
a1 b1 R2 a2 b2 R2 main building that is the middle corner of the L- shape building
frequency peak at 0.74 Hz dominates the spectra in x direction. Also
Roke AM14E-X 3.70E 2.875 0.8893 6.61E−05 3.259 0.9035 note that the same pattern of spectra peaks appear on different floors
−05
located on the same side indicating continuity of the responses among
AM14E-Y 1.89E 2.145 0.815 2.81E−05 3.344 0.8605
−04
the floors. As expected, the height of spectra peaks that indicate
Bomb AM14E-X 5.09E 2.047 0.8178 8.93E−04 2.472 0.7892 vibration level are different according to the floor in that the higher
Cyclo- −04 floor has higher level of vibration and thus higher peak of spectra. The
ne AM14E-Y 2.38E 1.292 0.7773 4.37E−03 1.717 0.6745 fact that spectra peak at 0.58 Hz and 0.78 Hz appear in both x and y
−03
directions indicate the occurrence of torsional vibration modes. We
Guchol AM14E-X 2.30E 2.352 0.7671 3.59E−04 2.783 0.7342
−04 shall see this in more detail when discussing the mode-shapes from
AM14E-Y 1.68E 1.452 0.6394 1.60E−03 2.079 0.6474 system identification.
−03 Fig. 7 illustrates the spectra of building accelerations for 10-min
period at 10:30 JST earlier that day where RMS of wind speed was
between 2.3 m/s and the peak was 4.1 m/s, as a comparison of spectral
4.2. Spectral characteristics of building accelerations
characteristics during moderate or ambient wind. The figures exhibit
similar pattern with that of strong wind accelerations. The spectra are
Most of the records consist of small to moderate accelerations
dominated by low frequencies peaks while contribution of higher
except for several hours when the typhoon peaks passed the building
resonant frequencies at 2 Hz, 3 Hz and 3.5 Hz are more apparent this
location. For typhoon Roke the large accelerations period, which is
time as shown by the relatively higher peaks for those respective
when peak accelerations on top of the building are larger than 0.5 cm/
frequencies. It is interesting to note that frequencies associated with
s2, lasted for about 6 h from 16:00 to 22:00 JST. For bomb-cyclone, the
the peaks for the ambient wind are slightly higher than that the
large acceleration period was about 8 h from 15:00 to 23:00 JST. And
corresponding peaks during the strong wind. For example, frequency of
for the typhoon Guchol the duration of large accelerations period was
the first peak changed to 0.63 Hz from 0.58 Hz, while the second peak
about 9 h, starting from June 19, 2012 at 19:00 JST until 04:00 JST
shifted to 0.79 Hz from 0.7 Hz.
the next day.
The same trend of frequency shift was observed from building
Spectral characteristics of building accelerations are evaluated for
accelerations due to the bomb-cyclone and typhoon Guchol as listed in
moderate and strong wind excitations to show the difference in
Table 2. In both cases, the acceleration spectra during the highest wind
characteristics. Fig. 6 shows the spectra amplitude of accelerations of
speed are characterized by lower resonant frequency peaks at the range
the main and annex building at 18:30 JST for the 10-min period of the
of 0.58–0.6 Hz for the first peak and 0.73–077 Hz for the second peak.
strongest wind of typhoon Roke (i.e. wind speed RMS=30.1 m/s and
Meanwhile for lower wind speed ( < 3.4 m/s) the peaks of acceleration
peak=41 m/s). Note that in the figures the x-axis of each spectrum was
spectra are at the higher frequency range of 0.64–0.65 Hz for the first

187
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

AM14E-X AM09E-X AM07E-X AM04E-X AM14E-Y AM09E-Y AM07E-Y AM04E-Y

15000 10000
Fourier Amplitude
0.58Hz 0.58Hz
0.74Hz
10000
5000
5000

(a) 0 (b) 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
AM14W-X AM09W-X AM07W-X AM04W-X AM14W-Y AM09W-Y AM07W-Y AM04W-Y
10000 10000
Fourier Amplitude

0.58Hz
0.74Hz

5000 5000

(c) 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (d) 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

6000 4000
Fourier Amplitude

0.58Hz AA7S-X AA4S-X 0.58Hz AA7S-Y AA4S-Y

0.74Hz 3000
4000
2000
2000
1000

0 0
(e) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (f) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (Hz) Frequency (Hz)

Fig. 6. Fourier spectral characteristics of building accelerations during the strongest wind period of typhoon Roke (18:30 JST Sept 5,2011, rms=30.1 m/s and peak=41 m/s) for: (a) east
corner of main building in X-direction, (b) east corner of main building in Y-direction, (c) west corner of main building in X-direction, (d) west corner of main building in Y-direction, (e)
south corner of annex building in X-direction, and (f) south corner of annex building in Y-direction.

peak and 0.78–0.79 Hz for the second peak. The results indicate 4.3. Time-frequency characteristics of building accelerations
reduction of resonant frequencies of the building during strong wind
excitation. We shall see this in more detailed when discussing the The Fourier spectra analysis provides a basic description on the
variation of resonant frequencies with respect to wind speed from energy distribution of acceleration responses in frequency domain for
system identification. stationary signal. In the case of typhoon responses, however, it is not
uncommon that they become non-stationary (see for instance (Xu and

AM14E AM09E AM07E AM04E AM14E AM09E AM07E AM04E

1000 800 0.63Hz
0.63Hz
Fourier Amplitude

0.79Hz 600

500 400

200

(a) 00 (b) 0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

AM14W AM09W AM07W AM04W AM14W AM09W AM07W AM04W

1000 0.79Hz 800
Fourier Amplitude

0.63Hz
600

500 400

200

0 0
(c) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (d) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

400 300
0.63Hz
Fourier Amplitude

AM07E 0.63Hz AM07E

300 0.79Hz AM04E AM04E
200
200
100
100

0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
(e) (f)
Frequency (Hz) Frequency (Hz)

Fig. 7. Fourier spectral characteristics of building accelerations during the low wind period of typhoon Roke (10:30 JST Sept 5,2011, rms=2.3 m/s and peak=4.1 m/s) for: (a) east
corner of main building in X-direction, (b) east corner of main building in Y-direction, (c) west corner of main building in X-direction, (d) west corner of main building in Y-direction, (e)
south corner of annex building in X-direction, and (f) south corner of annex building in Y-direction.

188
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

Table 2 peaks for 5-h scalogram from 17:00 to 22:00 JST are lower than 0.6 Hz
Peaks resonant frequencies on spectra of building accelerations identified during (Fig. 8(b)). The results agree well with the spectral characteristics as
typhoons at peak wind speed and low wind speed (ambient).
previously described by Figs. 6 and 7.
Typhoon roke Bomb-cyclone Typhoon guchol To have a better description on evolution of frequency content, the
wavelet instantaneous spectra analysis is performed by slicing the
Ambient Peak Ambient Peak Ambient Peak scalogram at a specific time and expanding the view in frequency
9:50a 18:30 3:30 19:30 5:50 00:20 domain. The resulting plot shows variation of wavelet coefficients
[0.7 m/s] [41 m/s] [3.4 m/s] [32 m/s] [1.0 m/s] [34 m/s]
across the range of frequencies. Examples of instantaneous wavelet
1st peak 0.65 Hz 0.580 Hz 0.64 Hz 0.599 Hz 0.65 Hz 0.589 Hz spectra for several time instants are shown in Fig. 8(c). Again from
2nd peak 0.79 Hz 0.740 Hz 0.78 Hz 0.761 Hz 0.79 Hz 0.760 Hz these figures, one can observe clearly a general decreasing trend in the
peak frequency as wind speed increases. During moderate wind speed
Note:
a as shown in time instant between 14:30 and 16:30 JST, the peak
Time is in Japan Standard Time (JST).
frequency is about 0.62 Hz and 0.613 Hz respectively. When the wind
speed as well as building acceleration reached the peak around 18:00–
Chen, 2004; Chen and Xu, 2004)), which means characteristics of
19:00 JST, the peak frequency dropped to about 0.56 Hz. When wind
energy distribution and frequency vary with respect to time. Under this
speed and building acceleration decreased near the end of the day, the
condition, it is necessary to use signal processing tools that can produce
peak frequency recovered back to 0.62 Hz. The same trend was
a time-frequency distribution of the responses. For this purpose, the
observed from the bomb-cyclone and typhoon Guchol data (Fig. 9).
Morlet-type Wavelet transform is utilized. The energy distribution in
For the bomb-cyclone, the instantaneous spectra of acceleration show a
the Morlet-type Wavelet transform is represented by plotting the
drop in frequency from about 0.62 Hz in the beginning of the event to
squared modulus of wavelet coefficients C(a,t) called the scalogram
0.55 Hz during the peak wind speed at 21:30 JST. The peak frequency
plot (Cohen, 1995).
was later recovered when the strong wind period end around midnight.
An example of scalogram of acceleration from sensor AM14E-X for
Similarly, for the typhoon Guchol, the peak frequency began to drop
the typhoon Roke is illustrated in Fig. 8. The energy distribution in
when the typhoon gained its strength around 23:00 JST and reached
time and frequency domain can be seen clearly by the contrast in
the lowest frequency at 0.58 Hz around midnight. The peak frequency
colors. It is evident from the figures that low order frequencies that
was later recovered to 0.62 Hz after the high wind speed period was
correspond to fundamental modes (i.e the first and second modes)
over.
within the range of 0.58–0.8 Hz dominate the energy distribution
throughout the entire acceleration responses. Higher order modes at
about 2 Hz and 3.5 Hz were also observed, but their contributions are 5. System identification methodology and results
rather weak. The figures show that for small to moderate wind speed,
as shown by the 5-h scalogram from 12:00 to 17:00 JST, the dominant In order to extract modal properties of the building from recorded
peaks are slightly higher than 0.6 Hz. Meanwhile, the corresponding dynamic responses, two system identification methodologies were

Fig. 8. Morlet Wavelet Scalogram of acceleration responses recorded by sensor AM14E in X direction, showing (a) 5-h of time-frequency variation during small wind excitation, (b) 5-h
of time-frequency variation during the peak wind speed, (c) time evolution of wavelet instantenous spectra.

189
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

0.015 0.08 0.2 0.08

t=20:20 0.8 t=21:30s
t=16:30 t=18:00
2

0.06 0.15 t=23:10

Wavelet Coeff

0.06
0.01 0.6
f=0.618Hz 0.04 0.1 0.04
f=0.613Hz 0.4 f=0.625Hz
f=0.571Hz f=0.55Hz
0.005
0.02 0.05 0.2 0.02

0 0 0 0 0
(a) 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Freq(Hz) Freq(Hz) Freq(Hz) Freq(Hz) Freq(Hz)
0.04 0.4 4 0.5 0.05
t=23:10 t=01:30s t=03:30
t=21:30 t=00:30
Wavelet Coeff 2

0.03 0.3 0.4 0.04

3
0.3 0.03
0.02 0.2 2
f=0.64Hz f=0.62Hz 0.2 0.02 f=0.621Hz
f=0.58Hz f=0.59Hz
0.01 0.1 1 0.1 0.01
0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0
(b) 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Freq(Hz) Freq(Hz) Freq(Hz)
Freq(Hz) Freq(Hz)

Fig. 9. Time evolution of Morlet Wavelet instantenous spectra of acceleration responses recorded by sensor AM14E in X direction, for (a) bomb-cyclone and (b) typhoon Guchol.

utilized, namely, the multi-output time-invariant system identification throughout the entire 10-min records. Therefore, a piece-wise linear
using information matrix (SRIM (Juang, 1997)) and the covariance- analysis was conducted using shorter moving time window, during
driven stochastic subspace identification (SSI (Peter van Overschee and which the modal parameters were assumed to remain constant. To this
Bart De Moor, 1996)). The SRIM method is based on the realization of end, the total 10-min acceleration responses were divided into ten time
dynamical system from cross-correlation functions of multi-channel windows consisting of 1-min data each from which a set of modal
accelerations. The identification process starts by constructing the parameters is generated.
multi-output correlation matrix defined as a correlation matrix be- Table 3 list the representative values of frequencies and damping
tween acceleration responses from multiple channels on the building to ratios obtained from both system identifications. Two cases are selected
a specific acceleration response, so called the reference channel. In this for each typhoon, namely the frequencies and damping during peak
case the acceleration from channel AM0W (x and y directions) were wind excitation and smaller wind which can be considered as the case
selected as the reference. Further details on description of the SRIM for ambient excitation. Frequencies of both vibration modes identified
method and implementation on the multi-channel output-only system from SRIM and SSI are very similar indicating the accuracy of
are presented elsewhere (Siringoringo and Fujino, 2008). Meanwhile identification, and their values are also very close to those obtained
the SSI method is based on covariance matrix computed from from spectra of accelerations. One can observe from the table that the
covariance of all channels on the upper-stories in x and y directions. identified frequencies of the first and second modes drop during the
Details on description of the SSI method on the multi-channel output highest wind speed. The results that are very similar with the peak
scheme can be found elsewhere (Peter van Overschee and Bart De resonances observed from frequency spectra of accelerations. The table
Moor, 1996). also show that although not very significant, the damping ratios
Both system identifications are time domain and based on linear obtained during high wind speed are generally larger than that
system assumption, which means modal parameters are considered to estimated from low wind responses. The tendency is consistent for
remain constant within the specific time-window of analyzed data. both modes from two system identifications.
However, considering the possibility of non-linear responses of the The table provides information on frequencies and damping
building during large excitation, this assumption may not be satisfied estimated from earthquakes, that is the biggest earthquake March 11,

Table 3
Examples of Identified natural frequencies and damping ratio for lower wind speed (ambient) and at the peak wind speed during typhoons and comparison with frequencies estimated
from FEM and earthquakes.

Typhoon roke Bomb-cyclone Typhoon Guchol

Ambient Peak wind speed Ambient Peak Wind speed Ambient Peak wind speed

9:50 [0.7 m/s]a 18:30 [41 m/s] 3:30 [3.4 m/s] 19:10 [32 m/s] 5:50 [1 m/s] 24:20 [34 m/s]

Frequency (Hz)**

1st Mode 0.620/0.630 0.556 /0.586 0.643/0.631 0.566/0.571 0.640/0.645 0.581/0.585

2nd Mode 0.782/0.784 0.740/0.749 0.784/0.773 0.755/0.761 0.784/0.793 0.760/0.756

Damping ratio (%)

1st Mode 1.54/2.18 3.07/3.06 1.80/2.26 3.06/2.54 1.51 /1.67 2.92/2.35
2nd Mode 1.94/1.10 2.77/2.62 1.67/2.31 3.44/2.84 1.16 /1.22 2.18/2.54

Finite Element Model Freq (Hz) 2011 March 11 Earthquake

Assumed Initial Stiffness of Isolator Assumed Large isolator deformation Main shock Aftershock(Eq. (4))
Freq (Hz) Damp.Ratio (%) Freq (Hz) Damp.Ratio (%)
1st Mode 0.50 0.16–0.18 0.440 14.60 0.530 9.780
2nd Mode 0.64 – 0.580 7.22 0.700 3.830

Note:
a
Frequency and damping ratio were identified from system identifications and presented as SRIM/SSI.

190
D.M. Siringoringo, Y. Fujino
2011 and Aftershock (Eq. (4)) (Siringoringo and Fujino, 2015a, 2015b)
for comparison. During the largest earthquake, the frequency dropped
significantly to 0.44 Hz and 0.58 Hz for the first and second mode,
respectively. The significant frequency drop is consequence of func-
tioning of isolation system and this is accompanied by increase in
damping. In the case of aftershock, the frequency is also smaller
compared to structure frequencies during the ambient excitation, but
not as significant as the main shock. The natural frequencies estimated
during aftershock are comparable with the values obtained from
typhoon during the peak response. Comparison between wind-induced
and earthquake-induced frequencies of the building will be discussed in
more detail in later part of this paper.
5.1. Variation of identified natural frequencies with respect to
building responses
In general, both system identifications generate three dominant
modes below 1.00 Hz. Comparisons with the previous identification
from seismic response and finite element model reveal that the modes
can be classified as the first translational mode of main building at
0.55–0.64 Hz, torsional mode of main building at 0.74–0.80 Hz and
torsional mode of annex building at 0.85–0.95 Hz. In the following
section, we shall focus the analysis on the first two modes, since they
are the modes that govern dynamic responses of the building during
typhoons and bomb-cyclone as shown in acceleration spectra and time-
frequency wavelet analysis.
Fig. 10 illustrates the time variation of natural frequencies of the
first two modes generated by both system identifications. Each data
point in the figure corresponds to frequency estimates from 1-min
acceleration records. Note that both system identifications give very
similar results during the observation time indicating high reliability of
frequency estimates. In general, the frequency estimates are relatively
constant throughout the observed period except during the time when
high wind speed occurs. During the period of largest wind speed,
gradual frequency reductions appear as noted in the time period
between 17:00 and 20:00 JST for typhoon Roke, 18:00 and 21:00
JST for bomb-cyclone; 23:00 and 2:00 JST for typhoon Guchol. The
average frequency reductions are about 10% and 8% for the first and
second mode, respectively. The general trends of natural frequencies
resulted from the two system identifications are shown in Fig. 11 with
respect to RMS of accelerations on the top of the building. In general,
Fig. 10. Time variations of the first and second mode natural frequencies identified from system identification SRIM during (a) Typhoon Roke, September 5, 2011, (b) Bomb-cyclone,
April 3, 2002, (c) Typhoon Guchol, June 19–20, 2012. Time variations of the first and second mode natural frequencies identified from system identification SSI during (d) Typhoon
Roke, September 5, 2011, (e) Bomb-cyclone, April 3, 2002, (f) Typhoon Guchol, June 19–20, 2012.
191
Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197
the results from SSI system identifications are slightly higher than that
of SRIM for the same interval of wind responses, but overall values are
very similar. The general trend is that the natural frequency estimates
decrease with the increase of RMS of building accelerations.
5.2. Variation of identified damping ratios with respect to building
responses
System identifications that are based on the realization of dynami-
cal system such as the SRIM and SSI methods are known to give
relatively accurate estimates of natural frequency and mode-shapes.
Estimation of accurate damping ratio is more difficult and highly
subjected to factorization and selection of system order in addition to
condition of the signal. Variations of damping estimates are generally
larger, especially for short duration of the records or when the system is
subjected to narrowband nonstationary excitation (Siringoringo and
Fujino, 2008).
The damping ratios of the first two modes were identified using 1-
min acceleration records by implementing the SRIM and SSI identifi-
cations. Generally, the damping ratios were identified between 0.5–5%.
The largely scatter results, however, create difficulty in drawing a trend
or meaningful conclusion directly from 1-min acceleration records with
respect to the response amplitude. To give a better view of the results,
the average and standard deviation of damping ratios during 10-min
data are computed and the results are shown in the Fig. 12 as the dot
and bar, respectively. The results still reveal quite large variation; but
despite the variation there is a general tendency that damping ratios
increase at the period of largest wind speed. The gradual increase of
damping ratio is noted in the time period between 17:00 and 20:00 JST
for typhoon Roke, 18:00 and 21:00 JST for bomb-cyclone; 23:00 and
2:00 JST for typhoon Guchol. The similar tendency is observed from
both system identification methods as shown in the two columns of
figure.
To have comparison of damping estimates that are not obtained
from system identification based on realization technique, a separate
damping identification procedure is conducted. Modal damping ratio is
estimated by computing the autocorrelation of acceleration at each
sensor node and applying a band-pass filter to the autospectrum to
isolate the mode of interest. Afterwards, the single-mode autospectrum
is transformed to time domain using an inverse Fourier transform to
obtain the free-vibration response. Modal damping ratio is extracted by
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

Fig. 11. Variation of natural frequency of the (a) first mode and (b) second mode identified from 1-min wind-induced accelerations of the building with respect to the RMS of top-floor
acceleration.

Fig. 12. Time variations of damping ratios for the first and second mode identified from system identification SRIM during (a) Typhoon Roke, September 5, 2011, (b) Bomb-cyclone,
April 3, 2002, (c) Typhoon Guchol, June 19–20, 2012. Time variations of damping ratios for the first and second mode identified from system identification SSI during (d) Typhoon
Roke, September 5, 2011, (e) Bomb-cyclone, April 3, 2002, (f) Typhoon Guchol, June 19–20, 2012.

the logarithmic decrement method from the free-vibration response. In
principle, the autocorrelation of the response of a single-degree-of-
freedom system can be considered proportional to its free vibration
response (Brincker et al., 2001; Magalhães et al., 2010).
In the case of acceleration responses of the building, there are two
modes that significantly dominate the response as revealed by spectra
figures, that is the first mode from about 0.55–0.62 Hz that can be
extracted from sensors in x-direction of east corner of the main
building (AME) and the second mode 0.7 Hz that can be extracted
from sensors in x-direction of west corner of the main building (AMW).
Since both modes are well separated and very dominant in its
respective direction especially during large wind excitation, assumption
of single-degree-of-freedom vibration can be applied to the autocorre-
lation of the acceleration responses. Generally, it is noted that the
accuracy of damping estimates obtained from free decay response
depend on the number of cycles used when extracting the free decay
response. This suggests that a longer data is essential to give better
estimates. For this reason, the damping estimation is conducted using
10-min instead of 1-min acceleration data.
Results of damping ratio estimates are plotted in Fig. 13(b) with
respect to the root-mean-squared of accelerations of top floor. For
comparison, the average damping ratios for 10-min data obtained from

192
D.M. Siringoringo, Y. Fujino
Fig. 13. Variation of damping ratio with respect to the RMS of top-floor acceleration. Identified from system identification methods using 10-min wind-induced accelerations (a) first
mode and (b) second mode. Estimated by autocorrelation of top-floor acceleration (c) first mode and (d) second mode. Comparisons with damping ratios estimated from small
earthquakes: (e) first mode and (f) second mode (Siringoringo and Fujino, 2008).
system identifications are also shown in the figure. It is also observed
that damping ratios obtained from autocorrelation functions are
generally larger than that obtained from system identifications for
the same time interval. Generally, the damping ratios obtained from
autocorrelation functions are between 1.5–5% where the first mode's
damping is usually larger than the second mode for the same time
interval. It is evident from the figure that in general, damping ratios
obtained from system identifications and autocorrelation functions
increase as the building accelerations increase. The increasing trend is
more obvious for damping ratios obtained from autocorrelation func-
tions than from system identification.
The increase of damping is known to be associated with the sources
of damping in the building and their contributions to the total damping
when building response increases (Tamura and Suganuma, 1996). In
the case of a base-isolated building, there are three possible sources of
damping during the wind-induced vibration, namely; the structural
damping from upper stories, damping from isolators and dissipative
energy devices at the basement, and aerodynamic damping due to
building vibration.
The higher damping estimates at higher building acceleration can
be attributed to the contribution of damping from isolators and
dissipative energy devices, while contribution of the aerodynamics
damping may not be so significant considering the wind velocity and
level of acceleration recorded on the top floor, as well as the total height
of building of 60.4 m. Contributions of damping from isolators and
dissipative energy devices are expected since they are activated due to
wind-induced deformation of isolators. Structural damping can be
approximated from structural responses during ambient vibration
where the effects of damping from isolators and aerodynamic damping
are considered small. In such case, the structural dampings are
approximated to be 2% and 1.5% for the first and second mode,
respectively.
Fig. 13(c) compares the trend of damping increase with respect to
increase in building's response for small earthquakes. Two points are
noted from the figures: (1) the trend generally agrees well with the
estimated from wind-induced response for the same interval of
building's accelerations, (2) the estimates of structural damping from
small earthquakes are found within good agreement with the ones
obtained from wind-induced responses.
5.3. Variation of identified mode-shapes with respect to building
responses
Typical shapes of the first two modes resulted from system
193
Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197
identification are shown in Fig. 14. It should be mentioned that since
both SRIM and SSI system identification results in the identical mode-
shapes for the same frequency, only one representative mode-shape is
shown in the figure for each frequency. In the figure, modal displace-
ments of each floor are normalized to the maximum absolute value of
the three-dimensional modal displacement. The figures were selected
to highlight characteristics of corresponding mode-shapes identified
from three cases, namely, the peak of typhoon Roke (18:30 JST), the
small ambient wind condition (11:00 JST) and large earthquake, which
in this case is the main shock of 2011 Great East Japan earthquake.
Note that in the next section, comparisons of mode-shapes from
moderate earthquakes will also be explained in detailed.
As illustrated in the Fig. 14, the first mode has characteristic of
large modal displacement in y-direction especially at the upper stories
of main building and smaller participation of motion in x-direction.
The mode exhibits a typical characteristic of flexural mode with small
participation of torsional motion on the upper stories that is due to
asymmetrical characteristics of the building. The first mode identified
from wind-excitation differs from the fundamental base-isolation mode
identified from large earthquake March 11, 2011. The mode identified
from the large earthquake exhibits a sliding mode characteristic in that
modal displacements are concentrated at the isolation layer. The large
modal displacement at the isolation layer is commonly observed when
the isolation has fully functioned as in the case of large seismic
excitation (Siringoringo and Fujino, 2015a, 2015b).
The main difference between the first mode generated by small and
large wind is on the amplitude of modal displacement in x and y
direction. In the case of small or ambient wind, modal displacement in
y direction is smaller than that of the strong wind. As a result, degree of
coupling between x and y direction is larger on the mode from ambient
wind than that of the stronger wind. From previous results of earth-
quake-induced mode (Siringoringo and Fujino, 2015b), we know that
large modal displacement in y-direction at the lower story is associated
with condition of isolator. Similarly, we can relate the difference in
modal displacement during ambient and strong wind with the condi-
tion of isolator deformation. Therefore, the difference in mode shape
amplitude is thought to be attributed to larger deformation of isolators
under strong wind excitation. As will be discussed later, this deforma-
tion can be related to the drag force acting on the building whose
magnitude is proportional to the wind speed.
As a torsional mode of main building, the second mode shows large
modal displacements concentrated on top corner of the main building
with opposite phase between the east and west corner. Meanwhile,
modal displacements of the superstructure of annex building are
D.M. Siringoringo, Y. Fujino
Fig. 14. Estimated mode-shape of the building during the peak acceleration of wind-induced response due to typhoon Roke for (a) the first mode 0.57 Hz and (b) second mode 0.75 Hz.
(Note: bold line=strong wind, dashed line=small wind, dotted lines=large earthquake (March 11, 2011)).
relatively small. Unlike the first mode, the second mode identified from
wind-induced vibration has similar characteristics with the corre-
sponding mode identified from large earthquake. This is understand-
able since the torsional modes of the building are not significantly
affected by condition of isolators as revealed in the previously research
(Siringoringo and Fujino, 2015a,b).
6. Relationship between estimated frequency and isolator
deformation
The results of spectra analysis, time-frequency wavelet scalogram,
instantaneous wavelet spectra and system identification have all
established that the fundamental frequencies of the building decrease
with the increase in wind speed. The frequencies, however, recover
soon after the wind speeds are back to normal. Since the building does
not suffer any structural damage and the upper structure is expected to
remain elastic throughout the strong wind excitation, the decrease in
frequencies is thought to be attributed to the stiffness decrease of the
seismic isolation layer due to isolators deformation.
Previous studies on the building (Siringoringo and Fujino,2015a,
2015b) revealed that characteristics of fundamental frequencies and
associated modes depend on the isolator condition. Since the isolation
system is designed primarily for seismic loading, two scenarios of
isolator behavior and the effect to modal parameters were considered.
One is the case of small and medium earthquake in which the initial
stiffness of isolators is considered, 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 3. In the design for large earthquake (maximum input
larger than 500 cm/s2), substantial deformation of isolation 40 cm is
assumed, as a result the first translation mode dominates the building
vibration. In such condition, the first natural frequency of the building
could reach as low as 0.16 Hz and the structure moves as a rigid body
where the main deformation concentrates on isolator layer. In design,
194
Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197
when large deformation occurs on the isolators during large earth-
quake, contribution of the torsional modes of upper-structure is
expected to be insignificant and due to the flexibility of isolation,
translation mode is expected to dominate vibration of the building.
Frequency reduction and change in fundamental mode shape
during strong wind excitation could be explained by employing similar
analogy to earthquake loading especially the effect of shearing force on
isolator deformation. Directions of the oncoming winds during the
largest wind speed period of the three events are on the south-
southeast or south-southwest (see Fig. 2), indicating the dominant
along-wind excitation with respect to main building's x-axis. In such
condition, the along-wind exerts drag force on the main building. The
total shear force (V) acting on the isolator layer is the summation of
shear forces on all floors given by:
n
V MWL = ∑ Fdi
i =1 (3)
1
Fdi is the average shear force on each floor defined as Fdi = 2 ρAi Cd Ui 2 ,
where ρ, Ai, Cd and Ui are the air density, frontal area, drag coefficient
and horizontal mean wind velocity of the ith floor, respectively. Since
the summation of shear forces is computed from mean wind speed, the
approach is termed as the mean wind load (MWL) method.
For comparison, the total shear forces on the isolators are also
computed using the Equivalent Static Wind Load (ESWL). By employ-
ing this approach, the total shear force (V) acting on the isolator layer is
defined as the summation of equivalent static wind loads on all floors:
n
V ESWL = ∑ WDi
i =1 (4)
The ESWL at height i is given as WDi = qi CD GD A, where q i =ρUi2 /2,
CD is the aerodynamic factor, A the frontal area and GD is the gust
loading factor. The ESWL method used in this comparison utilizes the
gust loading factor based on base bending moment (BBM) as described
in AIJ-RLB-2004 (AIJ-RLB, 2004). The BBM-based gust loading factor
is defined as:
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

C′g
GD = 1 + gD 1 + φD2 RD
Cg
where, gD is the peak factor, C′g and Cg are the fluctuating and mean
coefficient for along-wind overturning moment, ϕD is the mode-shape
dependent correction factor for along-wind mode and RD the resonance
factor.
Since the displacement-meters at isolators were not activated
during the typhoons, we cannot evaluate directly the measured
deformation of isolators due to static wind force. However, we can
estimate the deformation by computing the total shear force and use
the designed value of lateral stiffness of isolator. To this end, average
shear force or the equivalent static wind load on each floor caused is
computed by taking wind angle of attack into consideration. The mean
wind velocity at each story of the building is estimated using the
variation of wind with height formula as follows:
⎛ z ⎞α
Ui = Uz 0i ⎜ i ⎟ cos θi
⎝ z0 ⎠
where Uz 0i is the mean of standard wind measured at height z0=10.2 m
(Japan Meteorological Agency (JMA), 2016), and zi is the respective
height of each floor, θi is the mean of relative angle between wind mean
direction and the angle parallel to main building's x-axis. The mean of
wind angle of attack is assumed constant along the building height.
Furthermore, the static deformation of isolators associated with the
total shear force are estimated assuming the designed lateral stiffness
of isolators Kiso =727,560 kN/m, as:
VMWL, ESWL
δisolator =
Kisolator
For the MWL approach, the total shear force is computed by
estimating the frontal area (Ai) of each story based on design drawing
and taking the value of drag coefficient Cd as 1.08 as suggested for the
type of building geometry by AIJ-RLB-2004. Meanwhile, for the ESWL
method, calculation of the gust loading factor GD is conducted using the
procedure described in AIJ-RLB-2004. Parameters for gD, Cg, Cg’ and
RD are all selected using terrain category III and their associated values
as defined in the AIJ-RLB-2004 code (AIJ-RLB, 2004). The mode-
shape dependent correction factor ϕD is computed by using the first
mode identified from system identification for each typhoon event.
Note that since the analysis is aimed at finding the maximum wind load
and maximum isolator displacement at each typhoon event, the
frequency and damping ratios associated with the first mode is used
to calculate the gust load factor since it is the mode that most
significantly affected by isolator deformation (Siringoringo and
Fujino, 2015a).
Table 4 lists the results of estimates of total shear force and isolator
displacement during the period of peak wind speed for the three events
computed by MWL and ESWL methods. The results show that the total
shear force and the estimated isolator displacement resulting from
(5) ESWL method are consistently larger than that of MWL. This can be
explained as the consequence of including the gust loading factor,
whose values are between 2 and 3, when calculating the static load in
the ESWL method. The table also includes the maximum isolator
displacements observed from six earthquakes that give the first natural
frequency at the same range with the natural frequency from typhoon
events. It is evident from the table that the maximum isolator
displacements estimated from MWL method agree well with the values
observed from earthquakes, while those estimated from ESWL are
several times larger.
As suggested in design, the maximum deformation of isolator
significantly affects the natural frequency of the fundamental mode of
the building. Previous observation during various levels of earthquake
(Siringoringo and Fujino, 2015b) has revealed a relationship between
the maximum recorded isolator deformation and the identified natural
frequency of the first mode as shown by a trend line in Fig. 15. The
figure also shows that the maximum static deformations of isolator
(6)
estimated from wind-induced by MWL and ESWL methods during the
largest wind speed of the three events. The values are plotted with
respect to the identified natural frequency of the first mode.
Note that while the total maximum of isolator displacement due to
wind-induced vibration is a summation of static and dynamic deforma-
tion, the static component is expected to dominate the total maximum
displacement, so it is reasonable to compare the maximum recorded
isolator deformation from earthquakes with the maximum static
deformation of isolator estimated from wind-induced responses.
The trend line shown in Fig. 15 is obtained from several moderate
earthquakes in 2011 and a large one that is the main shock of March
(7) 11, 2011 earthquake where the first frequency dropped to 0.448 Hz
and the maximum recorded isolator displacement is about 5.50 cm.
The more moderate earthquakes cause smaller isolator displacements,
generally between 0.15 and 0.30 cm. The results shown in Fig. 15
reveal that wind-induced maximum isolator deformations estimated
from MWL method are generally in the good agreement with the trend
line established previously from earthquake records. More specifically,
the results of maximum static deformations of isolator estimated from
wind-induced vibration during the largest wind speed of the three
events are comparable with the results from moderate earthquakes
shown in Table 4. The results from ESWL, however, are slightly off the
trend line. As mentioned previously, the total shear force and the
estimated isolator displacement resulting from ESWL method are
consistently larger than that of MWL as the consequence of including
the gust loading factor, whose values are between 2 and 3; while
natural frequencies remain the same. As the result, the maximum static
deformations of isolator estimated by ESWL method are slightly off the
trend line.
The results suggest that drag force acting on the building due to
mean wind part induces initial static deformation on the isolators. The

Table 4
Estimated maximum total shear force acting on the isolators resulted from static wind force computed by MWL and ESWL methods, recorded moderate earthquakes and the
corresponding isolator maximum displacements.

Estimated Values Typhoon roke Bomb-cyclone Typhoon guchol

18:30 JST [28.5 m/s] 21:10 JST [22.7 m/s] 00:20 JST [26 m/s]

MWL ESWL MWL ESWL MWL ESWL

Total Shear Force (V) [tonf] 1521.32 4417.58 1363.55 3862.07 1593.86 4679.61
Max Isolator Disp. (cm) 0.209 0.607 0.187 0.531 0.219 0.643
Frequency of 1st mode 0.565 0.565 0.57 0.57 0.576 0.576

Eq. (1) Eq. (2) Eq. (3) Eq. (4) Eq. (5) Eq. (6)
16-4-2011 15-7-2011 25-7-2011 (1) 25-7-2011(2) 21-4-2011 31-7-2011
Max Isolator Disp. (cm) 0.276 0.282 0.176 0.229 0.269 0.2234
Frequency of 1st mode 0.536 0.545 0.572 0.548 0.543 0.550

195
D.M. Siringoringo, Y. Fujino Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197

0.65

0.6

EQ
0.55
EQ-Trend Line
Roke(MWL)
0.5 Roke(ESWL)
BombCy(MWL)
BombCy(ESWL)
0.45 Guchol(MWL)
Guchol(ESWL)

0.4
0.1 0.2 0.3 0.4 0.5 0.6 0.8 1.0 2.0 3.0 4.0 5.0 6

Fig. 15. Comparison between identified frequency of the first mode and isolator maximum displacement obtained from earthquake measurements (Siringoringo and Fujino, 2015b) and
estimated from wind-induced responses (MWL) and (ESWL) method.

Main Bld Mid Column Main Bld Mid Column Annex Bld X-Axis
Main Bld Outer Column Main Bld Outer Column Annex Bld Y-Axis
14 14 8
EQ1
7 EQ1
12 12
EQ2
6 EQ2
10 10 EQ3
5 EQ3
8 8 EQ4
Floor
Floor

Floor
4 EQ4
6 6
3
4 4
2
2 2 1

0 0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
(a) Modal Disp Ratio X-Axis (b) Modal Disp Ratio Y-Axis (c) Modal Disp Ratio

14 14 8 Roke
Roke
12 12 7
B.Cyc
B.Cyc
10 10 6
Guchol
5 Guchol
8 8
Floor

Floor

Floor

4
6 6
3
4 4
2
2 2
1
0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1 0
-1 -0.5 0 0.5 1
(d) Modal Disp Ratio X-Axis (e) Modal Disp Ratio Y-Axis (f) Modal Disp Ratio

Fig. 16. Comparison of mode shapes of the first mode identified from moderate earthquakes: (a) main building x-axis, (b) main building y-axis and (c) annex building; with the
corresponding mode shapes identified from typhoons and bomb-cyclone: (d) main building x-axis, (e) main building y-axis and (f) annex building (Note: detail on selected earthquakes
and typhoons events are as explained in Table 3).

initial isolator deformation due to static wind force become the original
point of isolator displacement caused by building vibration associated
with the fluctuating wind part. Due to this fact, frequency of wind-
induced building vibration is not equal to the frequency at the initial
stiffness of the isolators but already affected by their initial static
deformations. In the case when wind velocity is high for a relatively
long period of time and wind acting on the same direction as in the case
of present analysis, the incremental static deformation of isolators
could accumulate and resulted in significant initial displacement.
In addition to natural frequency, mode shapes of the first mode
obtained from the moderate earthquakes and during the largest wind of
the three events are evaluated as illustrated in Fig. 16. One can observe
clearly that the mode shapes obtained from the wind responses analysis
are in the good agreement with the corresponding mode-shapes
obtained previously from earthquake records. The results imply that
not only the natural frequencies and estimated maximum isolator
displacement but also the identified mode-shapes of obtained from the
wind-induced responses analysis are consistent; suggesting that ap-
proximation of the maximum deformation of isolator by computing the
total shear force is satisfactory.
7. Conclusions
Analyses of wind-induced responses and dynamics characteristics
of an asymmetrical base-isolated building during the passage of
Typhoon Roke 2011, Typhoon Guchol 2012 and a Bomb Cyclone

196
D.M. Siringoringo, Y. Fujino
April 2012 have been presented in this paper. Based on detail analyses
of the measured data, conclusions of the study are summarized as
follows:
1. Wind-induced accelerations of the building increased with the
increase of approaching wind speed. The relationship between wind
speed and building acceleration can be described by power equa-
tions. Observation from measured wind direction revealed that the
most significant building responses during the typhoon were the
along-wind responses.
2. Responses of the building were dominated by the first fundamental
flexural mode slightly coupled with torsion. Due to building's
asymmetric configuration, influence of the second mode, a torsional
mode, on the overall responses of the floors on the west corner
(inner corner) of the L-shape building is quite significant.
Contributions of other higher flexural and torsional modes on the
total responses were found to be insignificant.
3. Wind-induced responses of the building revealed non-linear char-
acteristics. Natural frequencies of the fundamental modes (1st and
2nd modes) decreased with the increase of amplitude of building
acceleration. During the peak wind speed of typhoons and bomb
cyclone, the drops of 8–10% of natural frequencies were observed.
4. Damping estimates of the fundamental modes increased with the
increase of building accelerations. The increase in damping is
thought to be contributed by isolators and dissipative energy devices
that were activated due to wind-induced initial deformation.
5. Shapes of the fundamental modes identified from wind-induced
vibration had different pattern compared to that of the large earth-
quake, but show similar pattern with the ones from moderate
earthquakes. Due to asymmetrical characteristics of the building,
small participation of torsional motion on the upper stories, which
was not observed during large earthquakes, was observed on the first
mode during the wind-induced vibration. Characteristics of the
mode-shapes were also found to be strongly influenced by the wind
speed. Larger modal displacements of the main building's lower
stories, which associated with isolator displacement, were observed
during the periods of largest wind speed.
6. The analyses show that reduction in natural frequencies and change
in mode shapes during strong wind are related to deformation of
isolators due to static wind force. The wind-induced maximum
deformations of isolators were estimated by computing the total
shear force on the isolators using the mean wind load (MWL) and
the equivalent static wind load (ESWL) approaches. The results were
compared with the deformations measured during moderate earth-
quakes and better agreements were obtained for the results of the
mean wind load (MWL) than the equivalent static wind load (ESWL)
approach.
References
AIJ-RLB, 2004. Recommendations on Loads for Building. Architectural Institute of
Japan (AIJ), (in Japanese).
Brincker, Rune, C. Ventura, Palle Andersen, 2001. Damping estimation by frequency
domain decomposition. IMAC XIX, Kissimmee, USA, 9(2001), p. 72.
197
Journal of Wind Engineering & Industrial Aerodynamics 167 (2017) 183–197
Campbell, S., Kwok, K.C.S., Hitchcock, P.A., 2005. Dynamic characteristics and wind-
induced response of two high-rise residential buildings during typhoons. J. Wind
Eng. Ind. Aerodyn. 93 (6), 461–482.
Chen, J., Xu, Y.L., 2004. On modelling of typhoon‐induced non‐stationary wind speed for
tall buildings. Struct. Des. Tall Spec. Build. 13 (2), 145–163.
Chen, Y., Ahmadi, G., 1992. Wind effects on base-isolated structures. J. Eng. Mech. ASCE
118 (8), 1708–1727.
Cohen, L., 1995. Time-Frequency Analysis 299. Prentice Hall PTR, Upper Saddle River,
New Jersey.
Henderson, P., Novak, M., 1989. Response of base‐isolated buildings to wind loading.
Earthq. Eng. Struct. Dyn. 18 (8), 1201–1217.
Japan Meteorological Agency (JMA), 2016. Regional Meteorological Observatory List of
the Automated Meteorological Data Acquisition System (AMeDAS).
Japan Society of Seismic Isolation (JSSI), 2009. Guideline of Wind-resistant Design for
Base-isolated Buildings (in Japanese).
Jeary, A.P., Ellis, B.R., 1983. On predicting the response of tall buildings to wind
excitation. J. Wind Eng. Ind. Aerodyn. 13 (1–3), 173–182.
Juang, J.N., 1997. System realization using information matrix. J. Guid. Control Dyn. 20
(3), 492–500.
Kareem, A., 1997. Modelling of base-isolated buildings with passive dampers under
winds. J. Wind Eng. Ind. Aerodyn. 72, 323–333.
Kochiyama O., Ikenaga M., Nakamura T., Kaneko S., Takenaka Y., Yoshikawa K., Wada
A., 2004. Study on LRB performance of high-rise base-isolated building during
strong wind. In: Proceedings 13th World Conference on Earthquake Engineering,
Aug. 2004.
Li, Q.S., Yang, K., Wong, C.K., Jeary, A.P., 2003. The effect of amplitude-dependent
damping on wind-induced vibrations of a super tall building. J. Wind Eng. Ind.
Aerodyn. 91 (9), 1175–1198.
Li, Q.S., Fang, J.Q., Jeary, A.P., Wong, C.K., Liu, D.K., 2000. Evaluation of wind effects
on a supertall building based on full‐scale measurements. Earthq. Eng. Struct. Dyn.
29 (12), 1845–1862.
Liang, B., Shishu, X., Jiaxiang, T., 2002. Wind effects on habitability of base-isolated
buildings. J. Wind Eng. Ind. Aerodyn. 90 (12), 1951–1958.
Magalhães, F., Cunha, Á., Caetano, E., Brincker, R., 2010. Damping estimation using free
decays and ambient vibration tests. Mech. Syst. Signal Process. 24 (5), 1274–1290.
Sato D., Suzuki, H., Tamura, T., Fugo,Y., Nakamura, O., Kasai K., Kitamura H., 2012.
Evaluation of Wind-Induced Responses of Isolated Highrise Building Based on
Monitoring Records. In: Proceedings 20th Natl. Symp. on Wind Engineering, Tokyo,
pp. 365–370 (in Japanese).
Siringoringo, D.M., Fujino, Y., 2008. System identification applied to long‐span cable‐
supported bridges using seismic records. Earthq. Eng. Struct. Dyn. 37 (3), 361–386.
Siringoringo, D.M., Fujino, Y., 2015a. Seismic response analyses of an asymmetric base‐
isolated building during the 2011 Great East Japan (Tohoku) Earthquake. Struct.
Control Health Monit. 22 (1), 71–90.
Siringoringo, D.M., Fujino, Y., 2015b. Long‐term seismic monitoring of base‐isolated
building with emphasis on serviceability assessment. Earthq. Eng. Struct. Dyn. 44
(4), 637–655.
Takenaka Y., Yoshie K. Kitamura H., Ohkuma T., 2009. Study on wind-resistant design
method of seismically base-isolated buildings. In: Proceedings JSSI 15th
Anniversary International Symposium on Seismic Response Controlled Buildings for
Sustainable Society, Tokyo, September 16–18.
Takenaka Y., Suzuki M., Yoshikawa K., Nakamura T., Kouchiyama O., Ikenaga M., 2004.
Simplified prediction method of wind-induced response of base-isolated tall building
using LRB. In: Proceedings 18th Natl. Symp. on Wind Engineering, Tokyo, pp. 365–
370 (in Japanese).
Tamura, Y., Suganuma, S.Y., 1996. Evaluation of amplitude-dependent damping and
natural frequency of buildings during strong winds. J. Wind Eng. Ind. Aerodyn. 59
(2), 115–130.
van Overschee, Peter, De Moor, Bart, 1996. Subspace Identification for Linear Systems:
Theory, Implementation, Applications. Kluwer Academic Publishers, P.O. Box 17,
3300 Dordrecht, The Netherlands.
Vulcano, A., 1998. Comparative study of the earthquake and wind dynamic responses of
base-isolated buildings. J. Wind Eng. Ind. Aerodyn. 74, 751–764.
Xu, Y.L., Chen, J., 2004. Characterizing nonstationary wind speed using empirical mode
decomposition. J. Struct. Eng. 130 (6), 912–920.
Yasui H., Ohkuma T., Koga S., Shimomura S., 2002. Full-scale measurement of wind-
induced response of a base-isolated midrise building. In: Proceedings 17th Natl.
Symp. on Wind Engineering, Tokyo, pp. 445–450 (in Japanese).