You are on page 1of 22

ARTICLE IN PRESS

Journal of Wind Engineering


and Industrial Aerodynamics 93 (2005) 461–482
www.elsevier.com/locate/jweia

Dynamic characteristics and wind-induced


response of two high-rise residential buildings
during typhoons
S. Campbell, K.C.S. Kwok, P.A. Hitchcock
CLP Power Wind/Wave Tunnel Facility, Hong Kong University of Science and Technology, Hong Kong
S.A.R., P.R. China
Received 11 May 2004; received in revised form 28 December 2004; accepted 30 March 2005
Available online 28 June 2005

Abstract

Two high-rise residential buildings in Hong Kong, among the tallest in the world, were
equipped to monitor their wind-induced dynamic response. The effects of typhoon Imbudo
and typhoon Dujuan on the buildings are presented in detail. Upcrossing analyses performed
on the acceleration data indicate the wind-induced response of the buildings during typhoons
Imbudo and Dujuan follow a Gaussian distribution. Natural frequencies of vibration in two
orthogonal translational and torsional directions are estimated by empirical, numerical, and
experimental techniques. A comparison between these natural frequencies reveals that
experimental values are higher than empirically or numerically predicted values. This finding
suggests that tall, reinforced concrete (RC) buildings in Hong Kong are stiffer than similar
buildings in other countries.
r 2005 Published by Elsevier Ltd.

Keywords: Residential buildings; Typhoons; Full-scale measurement; Dynamic characteristics

Corresponding author. Tel.: 85223587151; fax: 85223581534.


E-mail address: kkwok@ust.hk (K.C.S. Kwok).

0167-6105/$ - see front matter r 2005 Published by Elsevier Ltd.


doi:10.1016/j.jweia.2005.03.005
ARTICLE IN PRESS
462 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

1. Introduction

The design and construction of tall buildings has changed fundamentally over
time. Traditionally, tall building structures used masonry cladding and often had
heavy internal partitions which added a significant amount of strength and stiffness
to the primary load-bearing components. This type of structural arrangement
influenced the dynamic response of the buildings through increased mass, stiffness
and damping, and also limited their attainable heights. As design and analysis
techniques and materials have developed, structural systems have evolved to allow
the construction of lighter, taller buildings that are potentially more responsive to
dynamic excitation.
The design of a tall building is often governed by serviceability rather than
strength requirements, which is related to the stiffness of the structural system.
Furthermore, the amount of damping is generally not a prescribed measure due
to the high degree of variability in its estimation. In fact, the total damping
of a structure is a consequence of factors encompassing building material, the
overall design of the structure and its interaction with the wind. Current designs
of tall buildings still remain focused on providing appropriate serviceability
and stiffness levels to accommodate the desires to build taller. As knowledge
in the field of engineering for wind-induced motion of structures continues to
increase, advances are made in preventing unwanted dynamic response and are
implemented into building designs. Concomitant to the development of these taller
buildings is the assessment of their full-scale performance necessary to validate the
adopted design.
In this study, two high-rise residential buildings in Hong Kong, of planform and
construction that are typical in the region, are the subjects of a full-scale
measurement program aimed at characterizing their wind-induced dynamic
response. Tall buildings in Hong Kong have the reputation of being stiffer than
buildings of similar height built in other countries, which would be reflected in the
dynamic motion of the subject buildings. This paper outlines the measurement
program and presents results obtained during typhoons Imbudo and Dujuan
in 2003.
The two high-rise residential buildings monitored and described in this paper,
designated Building C and Building E, are located within an urban environment with
significant exposure to wind from some directions. They belong to a group of five
buildings that are arranged in a line and are of similar plan form. Building C, located
in the middle of the group, is approximately 218 m tall. Building E, located at one
end of the group, is approximately 206 m tall. There is a larger gap between the
second and third buildings than between others causing building C to be slightly
more exposed than indicated by its position within the group.
Structurally, these buildings are reinforced concrete (RC) shear wall constructions
with a common podium. Three storys above the podium, lateral forces are resisted
by columns and a central lift core. These columns support a transfer plate which, in
turn, supports the tower superstructure. Above the transfer plate, lateral resistance is
shared through a combination of the central lift core and shear walls.
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 463

2. Monitoring program setup

Dynamic building accelerations were measured using two orthogonally mounted


accelerometers. These devices were placed at the roof level above the highest
occupied floor of each building inside the core of the structure. The input axes of the
accelerometers, denoted X and Y, were aligned with the core walls, as seen in Fig. 1.
Channels X and Y are aligned approximately 1121 and 221 from North, respectively.
A signal conditioner was used to convert the raw current from each accelerometer
to a measurable voltage signal. Data records of 10 min duration were sampled at
20 Hz with a 16-bit A/D converter and stored on a dedicated PC after signal
amplification and the application of a 5 Hz low-pass filter.
Onsite restrictions prevented the installation of an anemometer on either building
to measure the winds associated with typhoon Imbudo and typhoon Dujuan. Wind
records were obtained from an anemometer located on Stonecutters Island, located
to the west of the Kowloon peninsula approximately 3 km from the subject
buildings, which generally reflects the overall wind characteristics affecting buildings
C and E. Stonecutters Island is a small island marking the entrance to Hong Kong’s
shipping container terminals. The propeller anemometer is supported on a mast 50 m
above ground with near-field topography consisting mostly of water to the
southwest, stacked shipping containers to the north, and a 65 m tall hill about
0.5 km to the east.

3. Typhoon events

3.1. Typhoon Imbudo

The Hong Kong Observatory (HKO) responded to typhoon Imbudo by issuing


standby signal #1 at 20:20 on 22 July, 2003. The eye of typhoon Imbudo came within
a distance of 280 km at its closest to Hong Kong at approximately 05:00, July 24.
The highest signals raised during typhoon Imbudo were #8NE raised at 22:40, 23
July and #8SE at 05:15, 24 July. The lowest instantaneous mean sea-level pressure of
997.5 hPa was recorded at the HKO headquarters at 04:13 and 04:14, 24 July. All
signals were lowered at 12:40, 24 July with the passage of typhoon Imbudo [1]. The
path of the typhoon’s eye can be observed in Fig. 2.

Building C Building E
NORTH Y Y
22º

X X

Fig. 1. Building plans and accelerometer orientation.


ARTICLE IN PRESS
464 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

Fig. 2. Track of typhoon Imbudo (after HKO).

30
50m
55m
25
Wind Speed (m/s)

20

15

10

0
7/23/03 7/23/03 7/23/03 7/23/03 7/24/03 7/24/03 7/24/03 7/24/03 7/25/03
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00
Date/Time

Fig. 3. 10-min wind speed record taken at Stonecutters Island during typhoon Imbudo.

Anemometer records from Stonecutters Island recorded during typhoon Imbudo


were used to calculate 10-min mean wind speeds, which are presented in Fig. 3.
The maximum 10-min mean wind speed was 17.0 m/s measured at 20:09, 23 July
and the maximum 3-s gust wind speed was 27.3 m/s measured at a time of 15:21,
23 July.
The 10-min mean wind directions for typhoon Imbudo were also calculated and
are included in Fig. 4. The dominant wind direction of typhoon Imbudo tended to be
between 501 and 1201 from 23 to 25 July. This is consistent with the eye of typhoon
Imbudo passing to the south of Hong Kong.
While the anemometer at Stonecutters Island better represents the wind field near
Buildings C and E, it has not been operating for a sufficiently long period to use its
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 465

360
50m
300

240
Heading (˚)
180

120

60

0
7/23/03 7/23/03 7/23/03 7/23/03 7/24/03 7/24/03 7/24/03 7/24/03 7/25/03
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00
Date/Time

Fig. 4. 10-min wind direction record taken at Stonecutters Island during typhoon Imbudo.

data alone for a detailed statistical analysis of wind speed and return period.
Therefore, wind data records were also procured for purposes of estimating the
return period of typhoon Imbudo from an anemometer based at Waglan Island.
Waglan Island is located to the southeast of Hong Kong Island, with rela-
tively uninterrupted sea fetches. HKO operates an automatic weather station
that includes an anemometer at a height of 82.1 m above mean sea level.
Anemometer measurements have been seen to be influenced by Waglan Island’s
local surroundings, and a wind tunnel model study has been undertaken previously
to determine the direction-dependent topographical effects and to obtain correction
factors to 200 m in the freestream [2]. For this paper, these correction factors have
been applied to all wind speed data collected from Waglan Island.
The wind speeds measured at Waglan Island are scalar measurements recorded on
the minute. In other words, the hourly mean wind speed is the average of 60 samples,
which were taken instantaneously on the minute, at one-minute intervals. The largest
measured, hourly mean wind speed during typhoon Imbudo was approximately
22 m/s.
Typhoon wind records collected at Waglan Island from 1953 to 1999 have been
position corrected and the resulting hourly mean wind speeds at 200 m analyzed by
Holmes et al. [3] using the peaks-over-threshold approach. Based on the peaks-over-
threshold parameters, the return period of typhoon Imbudo was estimated to be
slightly less than 1 yr.

3.2. Typhoon Dujuan

The HKO raised typhoon standby signal #1 at 22:15, 1 September, 2003 in


response to typhoon Dujuan. The eye of typhoon Dujuan came to within 30 km of
the HKO headquarters during the night of 2 September as illustrated in Fig. 5. The
highest typhoon signal during typhoon Dujuan was #9, raised at 20:10, 2 September.
The lowest instantaneous mean sea-level pressure was 972.1 hPa. This pressure was
recorded at 21:11 and 21:12, 2 September at the HKO weather monitoring station
ARTICLE IN PRESS
466 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

113E 114E 115E 116E


24N 24N

23N 23N

22N 22N

21N 21N
113E 114E 115E 116E

Fig. 5. Track of typhoon Dujuan (after HKO).

30
50m
55m
25
Wind Speed (m/s)

20

15

10

0
9/2/03 9/2/03 9/2/03 9/2/03 9/3/03 9/3/03 9/3/03
12:00 15:00 18:00 21:00 0:00 3:00 6:00
Date/Time

Fig. 6. 10-min wind speed record taken at Stonecutters Island during typhoon Dujuan.

closest to the path of typhoon Dujuan. All typhoon signals were lowered at 03:20, 3
September with the passage of typhoon Dujuan [4].
The 10-min mean wind speeds measured at Stonecutters Island for typhoon
Dujuan are presented in Fig. 6. The maximum 10-min mean wind speed measured
was 25.9 m/s at 20:59, 2 September. During typhoon Dujuan’s approach, the
maximum 3-s gust wind speed was 32.5 m/s measured at a time of 20:23, 2
September.
Fig. 7 shows the wind direction record measured at Stonecutters Island during
typhoon Dujuan. The wind direction varied from 1401 to 3501, which is consistent
with the eye of typhoon Dujuan passing to the north of Hong Kong. The dominant
wind direction varied widely between 1101 and 3001 with a smaller range of
2301–2751 during the hour corresponding to the maximum measured mean wind
speed.
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 467

360
50m
55m
300

240
Heading (˚)
180

120

60

0
9/2/03 9/2/03 9/2/03 9/2/03 9/3/03 9/3/03 9/3/03
12:00 15:00 18:00 21:00 0:00 3:00 6:00
Date/Time

Fig. 7. 10-min wind direction record taken at Stonecutters Island during typhoon Dujuan.

The methodology used to calculate the return period of typhoon Imbudo was also
employed for typhoon Dujuan. The maximum hourly mean wind speed determined
at Waglan Island during typhoon Dujuan was approximately 23 m/s. Based on the
peaks-over-threshold parameters [3], typhoon Dujuan was estimated to have a return
period of 1 yr.

4. Dynamic response of buildings C and E

4.1. Typhoon Imbudo

Acceleration responses at the roof level above the highest occupied floor of
buildings C and E during typhoon Imbudo are presented in Figs. 8–11. The gaps in
the data in Figs. 8–11 occur because of an interruption in the initial equipment setup
as typhoon Imbudo approached Hong Kong. Recorded maximum accelerations are
summarized in Table 1. As expected, the response of buildings C and E increase with
increasing wind speed. The peak resultant accelerations were 2.3 and 2.4 mg for
buildings C and E, respectively, and were recorded over the period of strongest wind
during typhoon Imbudo. Also included in Table 1 are the largest component peak
accelerations and the largest standard deviation (SD) values from the acceleration
records along the X and Y axes. It should be noted that the peak accelerations and
the peak SDs do not necessarily occur during the same 10 min record. The small
magnitudes of the correlation coefficients ðrxy Þ in Table 1 demonstrate the lack of
correlation between the orthogonal components of acceleration.
A sample 10-min acceleration time history is presented in Fig. 12, which contains
the maximum peak acceleration recorded for building E during typhoon Imbudo.
Resultant accelerations such as those shown in Fig. 13 for building C and building E
correspond to the 10-min duration during which the maximum accelerations were
recorded. Arrows indicated on these figures represent the 10-min mean measured
wind direction during which these time-histories were acquired and were obtained
ARTICLE IN PRESS
468 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

5.0

S.D. Acceleration
4.0

Acceleration (milli-g)
Peak Acceleration

3.0

2.0

1.0

0.0
7/23/03 7/23/03 7/23/03 7/23/03 7/24/03 7/24/03 7/24/03 7/24/03 7/25/03
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00
Date/Time

Fig. 8. Building C acceleration response during typhoon Imbudo: Y-direction (N–S).

5.0

S.D. Acceleration
4.0
Acceleration (milli-g)

Peak Acceleration

3.0

2.0

1.0

0.0
7/23/03 7/23/03 7/23/03 7/23/03 7/24/03 7/24/03 7/24/03 7/24/03 7/25/03
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00
Date/Time

Fig. 9. Building C acceleration response during typhoon Imbudo: X-direction (E–W).

5.0

S.D. Acceleration
4.0
Acceleration (milli-g)

Peak Acceleration

3.0

2.0

1.0

0.0
7/23/03 7/23/03 7/23/03 7/23/03 7/24/03 7/24/03 7/24/03 7/24/03 7/25/03
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00
Date/Time

Fig. 10. Building E acceleration response during typhoon Imbudo: Y-direction (N–S).
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 469

5.0

S.D. Acceleration
4.0

Acceleration (milli-g)
Peak Acceleration

3.0

2.0

1.0

0.0
7/23/03 7/23/03 7/23/03 7/23/03 7/24/03 7/24/03 7/24/03 7/24/03 7/25/03
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00
Date/Time

Fig. 11. Building E acceleration response during typhoon Imbudo: X-direction (E–W).

Table 1
Acceleration response at roof level of buildings C and E during typhoon Imbudo

Building Resultant peak Component peak Component maximum SD

(milli-g) rxy a (milli-g) rxy a (milli-g) rxy a Axis

C 2.3 0.16 2.2 0.11 0.5 0.18 X


2.2 0.16 0.5 0.18 Y
E 2.4 0.07 2.2 0.07 0.6 0.21 X
2.0 0.35 0.5 0.41 Y
a
Correlation coefficient between X and Y acceleration signals.

5
Acceleration (milli-g)

2.5 Maximum = 2.2 milli-g

-2.5

-5
0 100 200 300 400 500 600
Time (s)

Fig. 12. Acceleration time history of building E during typhoon Imbudo: Y-direction (Beginning at 02:38
7/24/03).

from readings taken at Stonecutters Island anemometer. These mean wind directions
are indicative values which are used to assess the response characteristics of the
buildings. The acceleration traces exhibit little correlation between the components
ARTICLE IN PRESS
470 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

Mean Mean
N Wind N Wind

5 5
Y (~N-S) Acceleration: (milli-g)

Y (~N-S) Acceleration: (milli-g)


2.5 2.5

0 0

-2.5 -2.5

(rxy = 0.12) (rxy = 0.08)


-5 -5
-5 -2.5 0 2.5 5 -5 -2.5 0 2.5 5
X (~E-W) Acceleration: (milli-g) X (~E-W) Acceleration: (milli-g)

Fig. 13. Maximum resultant acceleration traces during typhoon Imbudo.

of acceleration and are not aligned with the alongwind and crosswind directions
according to the prevailing wind direction.

4.2. Typhoon Dujuan

The acceleration responses at the roof level above the highest occupied floor of
buildings C and E measured over the course of typhoon Dujuan are presented in
Figs. 14–17. The maximum accelerations recorded for buildings C and E during
typhoon Dujuan are summarized in Table 2. There is a noticeable difference between
the maximum accelerations experienced by building C and building E, with those of
building C being considerably larger. The acceleration response of both buildings C
and E increased as the wind speed increased with the approach of typhoon Dujuan.
The maximum resultant accelerations were 4.3 and 2.3 mg for buildings C and E,
respectively, and were recorded during the period of strongest wind during typhoon
Dujuan. Building C experienced significantly larger accelerations than building E
largely due to its more exposed position to south-westerly winds, and partly due to
its greater height. Once again, the small magnitudes of the correlation coefficients
ðrxy Þ in Table 2 demonstrate the lack of correlation between the orthogonal
components of acceleration.
A sample 10-min acceleration time-history is presented in Fig. 18, which contains
the maximum peak acceleration recorded for building C during typhoon Dujuan.
The resultant accelerations for buildings C and E are illustrated in Fig. 19 and
correspond to the 10-min duration during which the maximum accelerations were
recorded. Arrows indicated on these figures represent the 10-min mean measured
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 471

5.0

S.D. Acceleration
4.0

Acceleration (milli-g)
Peak Acceleration

3.0

2.0

1.0

0.0
9/2/03 9/2/03 9/2/03 9/2/03 9/3/03 9/3/03 9/3/03
12:00 15:00 18:00 21:00 0:00 3:00 6:00
Date/Time

Fig. 14. Building C acceleration response during typhoon Dujuan: Y-direction (N–S).

5.0

S.D. Acceleration
4.0
Acceleration (milli-g)

Peak Acceleration

3.0

2.0

1.0

0.0
9/2/03 9/2/03 9/2/03 9/2/03 9/3/03 9/3/03 9/3/03
12:00 15:00 18:00 21:00 0:00 3:00 6:00
Date/Time

Fig. 15. Building C acceleration response during typhoon Dujuan: X-direction (E–W).

5.0

S.D. Acceleration
4.0
Acceleration (milli-g)

Peak Acceleration

3.0

2.0

1.0

0.0
9/2/03 9/2/03 9/2/03 9/2/03 9/3/03 9/3/03 9/3/03
12:00 15:00 18:00 21:00 0:00 3:00 6:00
Date/Time

Fig. 16. Building E acceleration response during typhoon Dujuan: Y-Direction (N–S).
ARTICLE IN PRESS
472 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

5.0

S.D. Acceleration
4.0

Acceleration (milli-g)
Peak Acceleration

3.0

2.0

1.0

0.0
9/2/03 9/2/03 9/2/03 9/2/03 9/3/03 9/3/03 9/3/03
12:00 15:00 18:00 21:00 0:00 3:00 6:00
Date/Time

Fig. 17. Building E acceleration response during typhoon Dujuan: X-direction (E–W).

Table 2
Acceleration response at roof level of buildings C and E during typhoon Dujuan

Building Resultant peak Component peak Component maximum SD Axis

(milli-g) rxy a (milli-g) rxy a (milli-g) rxy a

C 4.3 0.12 4.3 0.12 1.3 0.12 X


3.6 0.01 1.0 0.01 Y
E 2.3 0.19 2.2 0.03 0.6 0.19 X
1.9 0.19 0.6 0.19 Y
a
Correlation coefficient between X and Y acceleration signals.

5
Acceleration (milli-g)

2.5

-2.5

Maximum = -4.3 milli-g


-5
0 100 200 300 400 500 600
Time (s)

Fig. 18. Acceleration time history of building C during typhoon Dujuan: X-direction (Beginning at 20:58
9/02/03).

wind direction during which these time-histories were acquired and were obtained
from readings taken at the Stonecutters Island anemometer. As for typhoon
Imbudo, the acceleration traces recorded for typhoon Dujuan exhibit little
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 473

N N

5 5
Mean Mean
Wind

Y (~N-S) Acceleration: (milli-g)


Y (~N-S) Acceleration: (milli-g)

Wind
2.5 2.5

0 0

-2.5 -2.5

(rxy = 0.12) (rxy = -0.19)


-5 -5
-5 -2.5 0 2.5 5 -5 -2.5 0 2.5 5
X (~E-W) Acceleration: (milli-g) X (~E-W) Acceleration: (milli-g)

Fig. 19. Maximum resultant acceleration traces during typhoon Dujuan.

correlation between the components of acceleration and are not aligned with the
alongwind and crosswind directions according to the prevailing wind direction.

4.3. Characteristics of dynamic response

In Hong Kong, the effects of typhoon winds on tall buildings are complicated by
the complex terrain, the density of the built environment and the directional
characteristics of the winds. As anticipated, for both typhoon Imbudo and typhoon
Dujuan, there was a general increase in building response with increasing wind
speed, which can be seen from comparing Fig. 3 with Figs. 8–11 and Fig. 6 with Figs.
14–17. As the typhoons tracked past Hong Kong, changes in mean wind direction
affected the dynamic responses of buildings C and E by altering the wind exposure of
the structures. With typhoon Imbudo passing on the south, winds originated
generally from the northeast to southeast sector while the northerly passage of
typhoon Dujuan produced generally northwesterly winds. Consequently, building E
was more exposed than building C during typhoon Imbudo, whereas building C was
more exposed than building E during typhoon Dujuan, which can be readily deduced
from Fig. 1. Although building C is about 12 m taller than building E, the exposure
of building E during typhoon Imbudo is greater than building C. These two factors
counterbalance each other resulting in buildings C and E having acceleration records
with similar peak and SD accelerations during typhoon Imbudo. However, the winds
measured during typhoon Dujuan provided building C with increased exposure and
with its greater height, produced significantly higher accelerations than those
experienced by building E.
ARTICLE IN PRESS
474 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

Buildings C and E are similar in terms of construction, planform, and height. It is


therefore, expected that damping values of building C and E are also similar and
hence not the cause of different responses experienced during typhoon Imbudo and
typhoon Dujuan.
As indicated in Fig. 1, the core (stiffness) axes are aligned approximately 451 to the
geometric axes of the structure. With this alignment, the wind-induced acceleration
responses were found to occur in combination predominantly in the fundamental
modes of vibration, aligned closely with the core axes. This observation is consistent
with findings reported by Isyumov et al. [5] in a wind tunnel study of high-rise
buildings with similar arrangement of stiffness and geometric axes. Envelopes of the
resultant acceleration traces for buildings C and E illustrate the complex nature of
their wind-induced response. These envelopes are generally of an elliptical shape
implying that the component accelerations are largely independent of each other.

4.4. Natural frequencies of vibration

Natural frequencies of vibration of buildings C and E were determined through


experimental, computational, and empirical techniques. Experimentally, natural
frequencies of vibration were obtained from power spectra of digital acceleration
records measured during the two typhoons. Natural frequencies of structures
determined by computational methods are often based on the eigenvalues solved
from an idealized mass, damping, stiffness, and force model. Typically, empirical
formulas are obtained from regression analyses of results from full-scale measure-
ment databases. Empirical formulas derived this way may be general in nature or
may be applied to specific classes of buildings. A summary of the results of these
analyses are presented in Table 3. Categorization of modal direction has been made
with respect to the predominant response direction. Therefore, the first mode of
vibration for each building is classified as X  1; Y  1 or y  1, corresponding to a
predominant response in the X-, Y-, or torsional-directions, respectively.
Natural frequencies of vibration from on-site measurements were obtained from
analyzing power spectral density plots of acceleration records. These acceleration
records were almost two hours in duration, spanning periods containing the
maximum peak accelerations for building C and building E during both typhoon
Imbudo and typhoon Dujuan. Correspondingly, the normalized bias error of these
measurements is about 130  106 and the normalized random error for all
measurements is 0.43 [6].
Torsion in the measured response was found from power spectra by noting the
presence of a significant frequency found in the X-direction that was not found in the
Y-direction. This finding suggests that the input axis of the Y-direction accelerometer
and the axis of rotation lay approximately within the same plane. Support of this
result is obtained from the orthogonal accelerometers being in a geometric position
within the building core that would anticipate such a finding. Sample spectra given in
Fig. 20 highlight the dominant frequencies of vibration for buildings E and C during
typhoon Imbudo and typhoon Dujuan, respectively. The absence of significant
spectral power at the 0.582 and 0.538 Hz frequencies in the Y-direction spectra
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482
Table 3
Predicted and measured natural frequencies of vibration for building C and building E (Hz)

Source of natural frequency estimate First translational mode First torsional mode

Predictor (Hz)a Predictor (Hz)a

ARTICLE IN PRESS
Building Building

C E C E

X1 Y1 X1 Y1 y1

Experimental — 0.313–0.314 0.306–0.308 0.355–0.359 0.367 — 0.538–0.541 0.582–0.586


Computational — 0.17 0.17 0.18 0.19 — 0.14 0.15
Ellis [8] 46/H 0.21 0.22 72/H 0.33 0.35
Lagomarsino [9] 55/H 0.25 0.27 78/H 0.36 0.38
Tamura et al. [10] 67/H 0.31 0.33 91/H 0.42 0.44
Su et al. [11] b 77/H 0.35 0.37 — — —
a
H denotes the height of the building in metres.
b
Torsional natural frequency predictor not available.

475
ARTICLE IN PRESS
476 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

Fig. 20. Sample power spectral densities of responses.

compared to those in the X-direction suggests that these values represent torsion.
The first measured translational frequency of building C is Y  1, while that of
building E is X  1. This difference is suspected to arise from a different position of
shear walls associated with the slightly different plan forms between building C and
building E. The relative contribution of torsion in the X-direction of response is
approximately two orders of magnitude smaller than that of translation and is
therefore considered negligible.
In general, building frequencies calculated by computational means are
significantly lower than measured values. This difference is observed to range from
about 45% in translation to 75% in torsion. The cause of the absolute difference
between predicted and measured values may be due to a systemic error in the model
(e.g. approximation of boundary conditions, discretization of distributed parameter
systems, or estimation of physical properties of structural materials), evolving
building designs (e.g. new construction methods), or a combination thereof [7].
Accounting for the source of this difference may improve subsequent computational
estimations and is addressed more thoroughly in the model updating literature. A
few discrepancies may be somewhat reconciled when noting that the lowest natural
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 477

frequencies in the computational model were estimated for nominally torsional


modes of vibration whereas the first detectable, measured frequency corresponded to
a predominantly translational mode. Nevertheless, natural frequencies estimated
computationally may offer insight as to the relationship of modes with respect to
each other in the form of frequency ratios. The computationally estimated frequency
ratios, X  1 to Y  1, are 1.0 and 0.96 for buildings C and E, respectively. The
corresponding measured frequency ratios are about 1.03 and 0.98 for buildings C
and E, respectively. This result illustrates that the measured frequency ratios of the
first two translational frequencies are within 3% of the computationally predicted
frequency ratios.
Empirical formulas for predicting natural frequencies of vibration may serve as a
check against computational methods for assessing building designs. Built structures
checked against these formulas may then be monitored and added to the overall
database of measured buildings. Thus, a cycle is formed by including new, monitored
buildings in the database, and over time may reveal trends in design or construction.
Several authors have analyzed comprehensive databases of full-scale measurement
results for purposes of developing formulas typifying natural frequencies of
vibration of buildings [8–11]. These predictors are given in Table 3. Natural
frequency predictors are usually given for the first translation and torsional modes of
vibration. Some database analyses differentiate between orthogonal translational
modes of vibration due to nonsymmetrical structural form. In this study no
differentiation was made due to the relative symmetry of buildings C and E.
Structural descriptions have been specified for some predictors and have been used,
where applicable, for buildings C and E.
It is noteworthy that the predictor developed by Ellis [8] has received support by
inclusion into the Australian and New Zealand Standard AS/NZS 1170.2 [12] and
the Eurocode ENV1991-2-4 [13]. However, it is apparent from the results presented
in Table 3, and further supported by Tamura et al. [10] and Su et al. [11], that this
formula may not be the best suited to tall, RC buildings built to resist typhoon winds
and that are typical of Hong Kong.
From the range of predicted translational frequencies, the measured values are
more similar to those from Tamura, et al. [10]. Measured frequencies in the X- and
Y-directions for building C were approximately equal to those from Tamura’s
predictors. The measured frequencies in the X- and Y-directions for building E were
larger than Tamura’s predictors by about +7% to +10%. Measured torsional
frequencies of vibration are about 25% higher than the largest predicted values from
Tamura et al. [10].
In comparison to the frequency predictor provided by Su et al. [11], the measured
frequency values for building C yields differences of 12% and 14% in the X- and
Y-directions, respectively. The measured frequency values for building E are just
slightly smaller than the values from the frequency predictor by Su et al. [11], in both
X- and Y-directions. Unfortunately, torsional frequency predictors for buildings in
Hong Kong are not offered. Measured frequencies were generally greater than those
predicted numerically and empirically. These results support the notion that Hong
Kong’s RC buildings are stiffer than their international equivalents [11].
ARTICLE IN PRESS
478 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

Natural frequency predictors have been shown to be most appropriate when


formulated with building height [9] but, as such, their accuracy and applicability may
be questionable when applied to structures with height considerably beyond the
range of the databases from which they were derived. Overall, there is a lack of
readily available information concerning the dynamic characteristics of tall RC
structures, and further development of a comprehensive database is needed.
Furthermore, it appears that there are certain regional factors that also contribute
to the discrepancies between empirical predictors and measured natural frequencies.

4.5. Peak factors of building response

^ of a building
It has been shown that the average peak wind-induced response (x)
over a duration T may be conveyed as a function of the mean (x̄), peak factor ðgf Þ
and the SD ðsx Þ of the process (x), as shown in Eq. (1) [14].
x^ ¼ x̄ þ gf sx . (1)
From the statistical derivation of the peak acceleration, the peak factor may be
approximated deterministically in the form given by Eq. (2). This result allows
prediction of peak factors based on knowledge of a mean value crossing rate (n) and
the length of time over which it is recorded (T) of a process.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0:5772
gf ¼ 2 loge nT þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi . (2)
2 loge nT
Wind-induced structural vibrations are predominantly attributed to the excitation
energy within the narrow bands of frequencies about the building’s fundamental
frequencies of vibration. The majority of excitation energy of the wind is found to be
lower in frequency than fundamental frequencies of vibration, typical of tall
buildings. Furthermore, the amount of energy contained within the wind decreases
with increasing frequency. For this reason, response in the fundamental modes of
vibration is generally the most significant for tall buildings. Designers concern
themselves with the greatest response of these modes in the form of the maximum
value, or excursion. Generally, the wind excitation and structural response of tall
buildings under normal strong wind conditions have a probability distribution that
can be considered approximately Gaussian, which is attributable to the Central
Limit Theorem. The peak factor of a Gaussian process over 10 min is 3.3 which is
roughly the value predicted by Eq. (2). In practice, deviations from a Gaussian
distribution can occur and their effect can help to better characterize the true nature
of the excitation mechanism. Quantitatively, the parameter signifying the largest
probable excursion is the peak factor, which represents the number of SDs above the
mean value.
For a lightly damped dynamic system, such as that of a structure oscillating about
its fundamental frequency, it may be more convenient to express probabilities of
exceedence in terms of frequency, or upcrossings, than on a time basis. An
upcrossing frequency of 1/200 defines the point at which the probability of a single
maximum value exceeds a limiting value of once in 200 cycles, corresponding to
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 479

approximately 10 min based on the mean value crossing rates of buildings C and E.
This analysis procedure follows a Weibull distribution (i.e. a Rayleigh distribution
with k ¼ 2) and defines the probability of upcrossing exceedence as [15]
     
x^ a k
P 4a ¼ exp  , (3)
sx c

pffiffiffisx is the SD of x, a ¼ x=sx , the so-called reduced variate, k ¼ 2, and


where
c ¼ 2 sx .
A Gaussian process provides an appropriate dividing line between two extremes:

a process in which subsequent values depend heavily on previous values (e.g. a


sinusoidal process); and
a process in which intermittent characteristics become significant (a random
process).

The upcrossing analysis is appropriate for a narrow band response which, in this
case, arises in majority from the fundamental mode. A band-pass filter centered on
the fundamental natural frequency was used to isolate this frequency before
commencing the analysis. These records were acquired when wind speed and

X-Direction Y-Direction
Reduced Variate (a) Reduced Variate (a)
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
3 3

Sinusoid Sinusoid
2.5 2.5
loge(-loge(Probability of Exceedence (P))

loge(-loge(Probability of Exceedence (P))

Gaussian Gaussian
← Process Process ←
2 2
1/200 1/200

1.5 1.5

1 1

0.5 0.5

0 Peak Factor = 2.9 0 Peak Factor = 3.3

-0.5 -0.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
loge(Reduced Variate (a)) loge(Reduced Variate (a))

Fig. 21. Upcrossing analysis for building E acceleration during typhoon Imbudo.
ARTICLE IN PRESS
480 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

X-Direction Y-Direction
Reduced Variate (a) Reduced Variate (a)
1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 910
3 3

Sinusoid Sinusoid
2.5 2.5

loge(-loge(Probability of Exceedence (P))


loge(-loge(Probability of Exceedence (P))

Gaussian Gaussian
← Process ← Process
2 2
1/200 1/200

1.5 1.5

1 1

0.5 0.5

0 Peak Factor = 2.9 0 Peak Factor = 2.9

-0.5 -0.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
loge(Reduced Variate (a)) loge(Reduced Variate (a))

Fig. 22. Upcrossing analysis for building C acceleration during typhoon Dujuan.

direction were relatively constant. Sample graphs generated from the upcrossing
analysis for buildings C and E, for both typhoons Imbudo and Dujuan, are
presented in Figs. 21 and 22, respectively. The upcrossing analysis was performed to
determine the peak factors of six 10-min data records from buildings C and E that
occur over the hour of greatest response during typhoons Imbudo and Dujuan. The
mean, maximum and minimum values of the peak factors are summarized in Table 4
along with values predicted by Eq. (2). The results in Table 4 demonstrate that peak
factors for buildings C and E during the peak response periods of both typhoon
Imbudo and typhoon Dujuan are similar to that for a Gaussian process.

5. Conclusions

The dynamic characteristics of two high-rise residential buildings subjected to


typhoons Imbudo and Dujuan have been presented. According to wind records from
the Stonecutters Island anemometer, typhoon Imbudo was recorded as having a
maximum 3-s gust wind speed of 27.3 m/s and a maximum 10-min mean wind speed
of 17.0 m/s. For typhoon Dujuan, the Stonecutters Island anemometer measured a
ARTICLE IN PRESS
S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482 481

Table 4
Peak factors (gf) of 10-min acceleration records

Building Axis Predicted: Eq. (2) Measured: Imbudo Measured: Dujuan

00:30–01:30 20:30–21:30

24 July, 2003 2 September, 2003

Mean Max Min Mean Max Min

C X 3.4 3.5 4.2 2.9 3.0 3.4 2.5


Y 3.4 3.2 4.7 2.5 3.2 4.3 2.8
E X 3.4 3.6 4.4 2.9 3.1 3.5 2.8
Y 3.4 3.6 4.3 3.0 3.1 3.4 2.7

Probability of upcrossing exceedence for 1/200 cycles.

maximum 3-s gust wind speed of 32.5 m/s and a maximum 10-min mean wind speed
of 25.9 m/s. Based on the peaks-over-threshold approach, estimates of return periods
are: slightly less than 1 yr for typhoon Imbudo, and 1 yr for typhoon Dujuan. The
measured maximum dynamic building resultant accelerations were 2.4 and 4.3 mg
during typhoon Imbudo and typhoon Dujuan, respectively. It was found that the
motion experienced by the buildings depends heavily on wind speed, wind direction
and exposure that are complicated by the complexity of Hong Kong’s built
environment. The peak factors of the acceleration response of both buildings were
similar to that for a Gaussian distribution.
The first translational frequency of vibration occurred in the Y-direction for
building C and the X-direction for E. This is believed to arise from a different
configuration of shear walls due to the slight change of plan form between building C
and building E.
Frequencies of vibration were determined through experimental, computational
and empirical techniques. Generally, frequencies calculated by computational means
are significantly lower than measured values. However, Frequency ratios calculated
by computational means are very similar to those found during full-scale
measurements. A range of empirical frequency predictors were compared to
measured values and it was found that for the subject buildings, values were most
similar to those from Tamura et al. [10]. The survey of relevant and common
empirical frequency predictors has shown that, generally, Hong Kong buildings are
stiffer than international buildings of similar height.

Acknowledgments

Kowloon Properties Co. Ltd., MTR Corporation Limited, Wong & Ouyang
(Civil-Structural Engineering) Ltd., and Hip Hing Construction Co. Ltd. must be
thanked for their cooperation in the testing of these buildings. Thanks also go to the
ARTICLE IN PRESS
482 S. Campbell et al. / J. Wind Eng. Ind. Aerodyn. 93 (2005) 461–482

staff of CLP Power Wind/Wave Tunnel Facility at HKUST, particularly Mr. H.Y.
Leung and Mr. C.W. Yu, whose expertise contributed much to this project.
Acknowledgment goes to the Highways Department of the Hong Kong, SAR.
who provided the wind data from the Stonecutters Island anemometer.
This research project is funded by the Research Grants Council of Hong Kong
(Project HKUST6222/01E). The financial support is gratefully acknowledged.

References

[1] Hong Kong Observatory, typhoon Imbudo (0307): 17–25 July 2003, http://www.hko.gov.hk/
informtc/imbudo/report.htm, 2003.
[2] P.A. Hitchcock, K.C.S. Kwok, C.W. Yu, A study of anemometer measurements at Waglan Island,
Hong Kong. Technical Report WWTF002-003, HKUST, 2003.
[3] J.D. Holmes, P. Hitchcock, K.C.S. Kwok, J.K.S. Chim, Re-analysis of Hong Kong typhoon wind
speeds using the ‘peaks over threshold’ approach, in: Proceedings of Fifth Asia-Pacific Conference on
Wind Engineering (Kyoto, Japan, 2001), Elsevier B.V.,Amsterdam, 2001, pp. 357–360.
[4] Hong Kong Observatory, typhoon Dujuan (0313): 29 August–3 September, 2003, http://
www.hko.gov.hk/informtc/dujuan/report.htm, 2003.
[5] N. Isyumov, A. Steckley, N. Amin, H. Fatehi, Effects of orientation of the principal axis of stiffness
on the dynamic response of a slender square building, J. Wind Eng. Ind. Aerodyn. 36 (1990) 769–778.
[6] J.S. Bendat, A.G. Piersol, Random Data: Analysis and Measurement Procedures, second ed, Wiley,
New York, 1986.
[7] N.M.M. Maia, J.M.M. e Silva (Eds.), Theoretical and Experimental Modal Analysis, Research
Studies Press Ltd., Somerset, 1997.
[8] B.R. Ellis, An assessment of the accuracy of predicting the fundamental natural frequencies of
buildings and the implications concerning the dynamic analysis of structures, Proc. Inst. Civ. Eng. 69
(pt 2) (1980) 763–776.
[9] S. Lagomarsino, Forecast models for damping and vibration periods of buildings, J. Wind Eng. Ind.
Aerodyn. 48 (1993) 221–239.
[10] Y. Tamura, K. Suda, A. Sasaki, Damping in buildings for wind resistant design, in: Proceedings of
the International Symposium on Wind and Structures (Cheju, 2000), Techno-Press, Korea, 2000,
pp. 115–130.
[11] R.K.L. Su, A.M. Chandler, P.K.K. Lee, A. To, J.H. Li, Dynamic testing and modelling of existing
buildings in Hong Kong, Hong Kong Inst. Eng. Trans. 10 (2) (2003) 17–25.
[12] AS/NZS1170.2, Australian/New Zealand Standard, Structural design actions, Part 2: Wind Actions,
Standards Australia & Standards New Zealand, 2002.
[13] Eurocode ENV1991-2-4, EUROCODE 1: Basis of Design and Actions on Structures, Part 2.4: Wind
Actions, CEN/TC 250/Sc1, 1994.
[14] A.G. Davenport, Gust loading factors, J. Struct. Div. Proc. ASCE 93(ST3) (1967) 10–34.
[15] W.H. Melbourne, Probability distributions associated with the wind loading of structures, I.E. Aust.
Civil Eng. Trans. (1977) 58–67.

You might also like