Engineering Structures 167 (2018) 308–326

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Monitoring of wind effects on 600 m high Ping-An Finance Center during T

Typhoon Haima
⁎
Yinghou Hea,c, Xuliang Hanb, Qiusheng Lib,c, , Hongping Zhua, Yuncheng Hed
a
School of Civil Engineering and Mechanics, Huazhong University of Science and Technology, Wuhan 430074, China
b
Architecture and Civil Engineering Research Center, Shenzhen Research Institute, City University of Hong Kong, Shenzhen, China
c
Department of Architecture and Civil Engineering, City University of Hong Kong, Kowloon, Hong Kong
d
Joint Research Center for Engineering Structure Disaster Prevention and Control, Guangzhou University, Guangzhou 510006, China

A R T I C LE I N FO A B S T R A C T

Keywords: This paper presents analyzed results of field measurements of wind-induced pressures on building claddings and
Skyscraper structural responses of 600 m high Ping-An Finance Center (PAFC) in Shenzhen, China during the passage of
Typhoon Typhoon Haima. The characteristics of cladding pressures during the typhoon were investigated with highlights
Field measurement on energy distribution, non-Gaussian probability and coherence features. Then, the natural frequencies, damping
Structural dynamic property
ratios and mode shapes of the skyscraper were presented. Based on the field measurements, the amplitude-
Wind-induced response
dependent characteristics of the modal parameters were discussed and the range of selection of damping ratio for
Wind effect
Cladding pressure the wind-resistant design of super-tall buildings was recommended. Finally, the serviceability of the skyscraper
under typhoon condition was assessed. This paper aims to further understand the wind effects on super-tall
buildings and provide useful information for the design of skyscrapers in tropic cyclone prone regions.

1. Introduction
In recent years, numerous skyscrapers, such as 828 m high Burj
Khalifa, 632 m high Shanghai Tower and 600 m high Ping-An Finance
Center (PAFC), have been built. These modern skyscrapers with in-
creasing building height, relatively low natural frequency and damping
ratio become more sensitive to wind excitations than those in the past.
Hence, wind effects on these skyscrapers have become a major concern
in their wind-resistant structural design. Though the assessment of
wind-effects on high-rise structures can be performed via wind tunnel
tests or numerical simulations by Computational Fluid Dynamics (CFD)
techniques, there are some issues or limitations to be improved or
overcome, such as the properly modelling of approaching wind fields
for wind tunnel tests and the validity and accuracy of CFD simulation
results need to be verified or improved [4,30]. Field measurement has
been regarded as the most reliable means to evaluate the wind effects
on high-rise buildings.
Field investigations of structural dynamic properties and wind-in-
duced responses have been performed on a number of high-rise build-
ings over the last decade, such as three skyscrapers in Chicago [12,13],
more than ten tall structures in Hong Kong, mainland China and
Taiwan, including Central Plaza Tower and Di Wang Tower [17], CITIC
Plaza Tower and Jin-Mao Building [18], Guangzhou West Tower [6] ,
Taipei 101 Tower [19], Canton Tower [5,9]. These studies have en-
hanced the understanding of structural modal parameters and wind-
induced vibrations of high-rise structures. However, with the growing
trend of continuously increasing number of skyscrapers with higher
altitudes and varied appearances and complex configurations in recent
years, the performance of skyscrapers during windstorms such as tro-
pical cyclones are still needed to be further investigated by field mea-
surement or monitoring.
This paper presents a field investigation of the wind effects on
600 m high PAFC located in Shenzhen, China, during Typhoon Haima.
Detailed analysis of the field measurements during the passage of the
typhoon, including wind speed and direction, wind-induced pressures
on cladding and structural acceleration responses, is carried out to in-
vestigate the structural dynamic properties and wind-induced responses
of the skyscraper under typhoon condition. The objective of this study is
to further understand the wind effects on super-tall buildings and
provide useful information for the wind-resistant design of future sky-
scrapers. The paper is organized as follows: Section 2 introduces
structural system of PAFC, field measurement arrangement and Ty-
phoon Haima. Section 3 presents the observations of wind speed and
direction during the passage of the typhoon. The following section fo-
cuses on the analysis of the wind-induced pressures on building clad-
dings. Section 5 discusses structural acceleration responses, structural

⁎
Corresponding author at: Department of Architecture and Civil Engineering, City University of Hong Kong, Kowloon, Hong Kong.
E-mail address: bcqsli@cityu.edu.hk (Q. Li).

https://doi.org/10.1016/j.engstruct.2018.04.021
Received 16 September 2017; Received in revised form 6 April 2018; Accepted 7 April 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
Y. He et al.
(a) Building exterior
Fig. 1. Building exterior and location of PAFC.
modal parameters and presents a serviceability assessment of the sky-
scraper. Finally, the main findings and conclusions of this field mon-
itoring study are summarized in Section 6.
2. Introduction of PAFC and Typhoon Haima
2.1. PAFC
PAFC shown in Fig. 1 is located at the central business district of
Futian in Shenzhen where is a coastal city in the south of China. With its
main structure erecting to 600 m high from the ground, PAFC is cur-
rently the fourth highest building in the world and the second tallest
structure in China. The skyscraper consists of 118 floors above the
ground and has a footprint of 63 × 63 m. To enhance the resistance to
lateral and vertical loads for such a high and slender building, PAFC
comprises a mega-frame core tube structure including: eight steel-re-
inforced concrete mega-columns, reinforced concrete core tube, and
four outrigger trusses that connect the core tube with the mega-frame.
2.2. Measurement arrangement
In this study, measurement devices at four selected floors in PAFC
were deployed during Typhoon Haima to record wind-induced pres-
sures on building claddings, and structural acceleration responses. The
layout of the measurement devices is depicted in Fig. 2.
a. Anemometers
Wind speed and direction records monitored by two cup-type an-
emometers at two automatic weather stations in Shenzhen shown in
Fig. 1(b) during the passage of Typhoon Haima were used in this paper
to demonstrate the typhoon-generated wind characteristics. One
weather station is located at Futian (FT) nearby PAFC (about 1.3 km
away along the south direction) with altitude of 120 m atop an office
building. The other is located at Bijiashan (BJS) about 4.3 km away
from PAFC along the northeast direction with altitude of 31 m atop a
factory building. The two stations provide 5-min mean wind speed and
direction records.
b. Pressure transducers
As indicated in Fig. 2, pressure transducers were installed on the
309
Engineering Structures 167 (2018) 308–326
(b) Building location
building claddings at the 81st floor of PAFC. To obtain the pressure
measurements at representative locations on the building surfaces, the
centers of the four building sides and four corners at the 81st floor were
equipped with Setra268 pressure transducers (Setra systems, Inc. Box-
borough, Massachusetts), as shown in Fig. 2(b). The full-scale range
(FS) and accuracy of the transducers are 2.5 kPa and 0.25% FS (or,
6.25 Pa), respectively. The dynamic pressure of the exterior wind was
imported into the transducer through a plastic tube while the reference
static pressure was derived from a box installed inside the building.
During the pressure measurements, analog signals collected by the
transducers were then acquired and digitized at a sampling frequency of
20 Hz by a set of high-resolution digital data logger (National Instru-
ments Corporation, Austin, Texas), including cDAQ-9174 (USB chassis)
and 9234 (signal acquisition module).
c. Accelerometers
As shown in Fig. 2, accelerometers were installed along two or-
thogonal directions (east–west and north–south) at four floors: 49th,
81st, 100th and 118th. The two directions are denoted to be “A-1”
(east–west) and “A-2” (north–south) in the following analysis. To en-
sure the acceleration measurements with desirable accuracy, each of
the four floors was equipped with two pairs of low-frequency oriented
accelerometers with different detection ranges and accuracy at nearby
locations. Moreover, in order to detect the information of torsional
modes as much as possible, the accelerometers were placed specifically
at corners of the building. The locations and details of the detection
range and accuracy of the accelerometers and pressure transducers used
in this study are listed in Table 1. The acceleration signals were re-
corded in the same manner with that for the wind pressure signals.
2.3. Typhoon Haima
Typhoon Haima was one of few tropic cyclones that have made
landfall in Shenzhen at a typhoon strength level over the last decade.
According to the information released by Cooperative Institute for
Meteorological Satellite Studies (CIMSS) Tropical Cyclones, as shown in
Fig. 3, Haima formed as a tropical depression over the Western North
Pacific at 18:00 of 14 October 2016 (Coordinated Universal Time
(UTC); Beijing Time = UTC + 08:00), then intensified into a typhoon
on 15 October and followed a northwesterly track towards the South
China Sea, then made landfall over the coastal area in Shenzhen on 21
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 2. Layout of measurement devices. “A” and “P” denote accelerometer and pressure transducer, respectively. “A-1” and “A-2” denote the east–west and
north–south direction, respectively.

Table 1 3. Wind speed and direction

Information of the sensors adopted in this study.
Sensor Location Type No. Range Accuracy
Variations of 5-min mean wind speed and direction measured at the
two weather stations in Shenzhen during the passage of Typhoon
Accelerometer 49/F B&W 2 ± 5g; 0.06–1 k Hz 10 μg Haima are shown in Fig. 4. As can be seen, both the wind speed and
AS 2 ± 2g; DC-250 Hz 2 μg direction show consistent changing trend at these two weather stations.
81/F B&W 2 ± 5g; 0.06–1 k Hz 10 μg
In Fig. 4(a), during the first few hours of the wind records, wind speed
GeoSIG 2 ± 1g; DC-100 Hz < 60 μg
100/F B&W 2 ± 5g; 0.06–1 k Hz 10 μg increased continuously when the typhoon approached to the weather
Jwell 2 ± 1g; DC-200 Hz < 10 μg stations, reached to the peak value of about 12 m/s at about
118/F B&W 2 ± 5g; 0.06–1 k Hz 10 μg 11:00–13:00, and then decreased gradually when the typhoon moved
GeoSIG 2 ± 1g; DC-100 Hz < 60 μg
continually along the northwest direction. Fig. 4(b) shows the varia-
Pressure 81/F Setra268 8 ± 2.5 kPa ± 0.25%FS tions of wind direction at the two stations during the typhoon, in which
transducer the wind direction varied along anticlockwise direction from about 30°
to near 210°, while the north direction is defined as 0° in this paper.
October and finally dissipated rapidly as it moved further inland. Ac-
cording to the information such as the track of Haima approaching to
4. Cladding pressures
Shenzhen, the field measurements of wind-induced pressures on clad-
dings and building acceleration responses commenced at 11PM on 20
The locations of the pressure taps on the building sides for mea-
October and ended at 8:20 PM on 21 October 21 (Beijing Time) lasting
surement of cladding pressures and the north direction of the building
about 21 h, which covered the main passage process when Haima came
are illustrated in Fig. 2(b). In this section, the measured wind pressures
close to Shenzhen.
on building cladding during the passage of the typhoon will be analyzed
and discussed.

310
Y. He et al.
Fig. 3. Track of Typhoon Haima: CATEGORY 1–5 denote low-to-high typhoon strength.
Fig. 4. 5-Min mean wind records at the two weather stations in Shenzhen.
4.1. Time histories of cladding pressures
Fig. 5 depicts time histories of wind-induced pressures on building
claddings, in which P1–P4 denote the pressure transducers at the four
corners and P5–P8 represent those at the central points on the four
building sides, respectively, as shown in Fig. 2(b). Fig. 5 also plots
boxplots for each 10-min data segment. For a boxplot, the central mark
denotes the median, box edges correspond to the upper and lower
quartile, and line edges limit the distribution range of normal values.
According to the wind records shown in Fig. 4, wind flows at PAFC
evolved along anticlockwise direction from north-northeastern wind to
southwest wind. During the passage of Haima, it can be seen from Fig. 5
that the wind pressures at P-5 showed negative values most of the time
(such as those on building sides or leeward face), the wind pressures at
P-6 and P-7 changed from negative to positive (changed from leeward
face or building lateral sides to windward face) while those at P-8
converted from positive to negative (changed from windward face or
building lateral sides to leeward face). It was observed that the evolu-
tions of the wind pressures at the corners were relatively complex. The
wind pressures at P-1 had consistently changing trend with those at P-8
311
Engineering Structures 167 (2018) 308–326
(north side). The similar observation was found between the measure-
ments at P-3 and P-6 (south side) and also between those at P-4 and P-7
(west side). Although the pressure measurements at P-2 were distinct
from those at P-5 (south side), it is noteworthy that the evolving process
were similar between those at P-2 and P-8, which was attributed to that
these two measurement points had the same orientation towards to the
incident winds during Typhoon Haima. Though the corners of the
building were adjacent to the corresponding side faces, the measured
wind pressure characteristics at these corners demonstrated significant
differences from those at the central points on the building sides, which
illustrated that there were different mechanisms for generation of the
wind pressures at the two locations. In addition, it is noted that the peak
positive pressure occurred at a central point was 497 Pa, while the peak
negative pressure appeared on a building corner was −866 Pa. The
peak negative pressure on the building corner (−866 Pa) was 74% and
49% larger than the maximum positive and negative pressure (497 Pa
and −580 Pa) at a central point, respectively, implying that the max-
imum peak negative pressures likely occur at building corners or con-
cave regions.
Y. He et al. Engineering Structures 167 (2018) 308–326

P-1 P-2
0.6 0.6
0.4 0.4
0.2 0.2
Presure (kPa)

0 0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00
P-3 P-4
0.6 0.6
0.4 0.4
0.2 0.2
Presure (kPa)

0 0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00

P-5 P-6
0.6 0.6
0.4 0.4
0.2 0.2
Presure (kPa)

0 0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00
P-7 P-8
0.6 0.6
0.4 0.4
0.2 0.2
Presure (kPa)

0 0
-0.2 -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 -1
00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00
Fig. 5. Time histories of cladding pressures (blue lines) overlaid by boxplots (green color) for each 10-min data segment.

4.2. Non-Gaussian probabilistic characteristics of cladding pressures used to characterize the non-Gaussian feature of a probability dis-
tribution. The skewness (s) and kurtosis (k) can be determined by Eqs.
The boxplots shown in Fig. 5 reflect that there exists a heavy tail (1) and (2), respectively:
toward the peak value side in the negative pressure regions. Skewness is
a measure of the asymmetry and kurtosis is a measure of peakedness of
a probability distribution. And the skewness and kurtosis are widely

312
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 6. Correlations among skewness, kurtosis and mean and peak cladding pressures for each 10-min data segment.

1 n 1 n
n
∑i = 1 (x i−x )3 n
∑i = 1 (x i−x ) 4
s= 3
k= 2
1 n
1
∑ (x
n
−x )2 2 ⎡ ∑ (x −x )2⎤
⎡
⎣ n i=1 i ⎤
⎦ (1) ⎣ n i=1 i ⎦ (2)

and where xi is sample data, x is mean value and n is sample number.

Particularly, values of skewness and kurtosis for Gaussian distribution is

313
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 7. Normalized PSDs of wind pressures under the north wind.

0 and 3, respectively [8]. In this section, the discussions about the
skewness and kurtosis of the pressure measurements are based on 10-
min samples. Fig. 6 presents the interrelation between skewness, kur-
tosis and mean pressure, peak pressure for each 10-min data segment. It
can be found that at the central locations on the four building sides
(P5–P8), the positive wind pressures appeared to follow the normal
distribution while the negative pressures revealed obvious non-Gaus-
sian characteristics which became more evident with the increase of
negative pressure amplitude. Similarly, the similar phenomenon was
found at the building corners (P-2, P-3 and P-4), except for the pressure
measurements at P-1. The location of P-1 was subjected to both positive
and negative pressures which showed distinct non-Gaussian char-
acteristics.

314
Y. He et al.
Fig. 8. Normalized PSDs of wind pressures under the west wind.
4.3. Spectral analysis of cladding pressures
4.3.1. Normalized spectra
The spectral analysis techniques are widely used to identify the
energy distribution of signals in frequency domain. Three frequently
used method, i.e., Welch method [28], multi-taper method (MTM) [14],
and Yule-Walker method [24], are adopted herein to analyze the
spectral characteristics of the measured wind pressures on building
315
Engineering Structures 167 (2018) 308–326
claddings. Figs. 7 and 8 plot normalized power spectral densities (PSDs)
of the wind pressure measurements associated with two wind direc-
tions: north wind (360°) and west wind (270°). The corresponding block
length in the PSD estimators is set as 214 (820 s) so that a resolution
finer than 0.002 Hz in frequency domain can be achieved. It is noted
that these three methods provide consistent spectral distributions.
Strouhal number (St) is defined as:
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 9. Coherences of wind pressures at different measurement points when PAFC was subjected to the north winds.

Table 2
Pressure coefficients on PAFC under the west wind.
Pressure coefficient (Cp) Corners Leeward Lateral Windward Lateral

P-1 P-2 P-3 P-4 P-5 P-6 P-7 P-8

Mean (Code ) *
0.56 −1.48 −0.49 −0.07 −0.34 (−0.3) −0.83 (−0.7) 0.80 (0.80) −0.69 (−0.7)
Standard deviation 0.14 0.46 0.28 0.39 0.19 0.27 0.17 0.29
Peak 0.99 −3.85 −2.06 −1.51 −1.68 −2.04 1.42 −2.20

* Code denotes GB50009-2012: windward, 0.8; leeward, -0.3; lateral sides, -0.7.

D nD
St = = s
Ts v v (3)
where Ts is vortex shedding period (s), ns is vortex shedding frequency
(Hz), D is the width of the building that faces the approaching wind
(m), v is the mean wind speed of the approaching wind (m/s). As in-
fluenced by interference caused by construction equipment and
building components atop PAFC during the passage of Typhoon Haima,
anemometers installed above the roof of PAFC failed to provide valid
information on the local wind field. Hence, this study used the mean
wind speed records at FT weather station which is the nearest weather
station to PAFC as reference wind information. These wind measure-
ments were converted from 120 m height to those at higher altitudes
based on the wind profiles stipulated in the load code of China [7].
As shown in Fig. 7, when the northern wind was normal to P-8 on
the windward face, two peaks in the PSDs can be observed clearly at
both P-5 and P-7 located on the building lateral sides: the low-fre-
quency peak centered at 0.02 Hz and the higher one is located at
0.29 Hz. Meanwhile, a peak centered at 0.02 Hz is also found at all the
four corners (P1–P4). Herein, the converted wind speed from the FT
weather station was 8 m/s, the width of the building is 56.4 m (shown
in Fig. 2(b)), then the Strouhal number (St) is determined as St = 0.14,
which is within the range of 0.12 ∼ 0.16 and consistent with that of
prisms with a square section. Consequently, it can be concluded that the
peak at 0.02 Hz was attributed to vortex shedding from the building.
The analogical features of spectral distributions can also be found when
the approaching wind came from west as illustrated in Fig. 8. In this
case, P-7 was on the windward face while P-6 and P-8 were situated on
the building lateral sides. The peaks of PSDs of the wind pressures on
the two building lateral sides occurred at frequency of 0.04 Hz and
0.60 Hz, respectively. Expectedly, the peak at 0.04 Hz related to vortex
shedding also appeared at the four building corners. It is interesting to
observe that the ratios between the second peak and the first peak are
all around 15 for those two approaching wind cases (i.e., 0.29/
0.02 = 14.5 for the north wind in Fig. 7, and 0.60/0.04 = 15 for the
west wind in Fig. 8), especially, which is consistent with the ratio
(∼12) of the characteristic width of PAFC and the length of the bulge
shape shown in Fig. 2. Thus, the peaks at 0.29 Hz (in Fig. 7) and 0.60 Hz
(in Fig. 8) might be caused by the high-frequency eddies correlated with
the bulged shape of the building corners. Clearly, the spectra provide
the information of vortex shedding and influence of building shape on
the wind pressures on cladding.
4.3.2. Coherence of wind pressures
Generally, larger eddies associated with low frequencies could lead
to stronger correlations because their large sizes cover a wider scope
while smaller eddies related to higher frequencies might only produce
local impacts. Fig. 9 shows the coherence of the pressure measurements
at different building sides when PAFC was mainly subjected to the north
wind. It can be observed that the coherent amplitudes are relatively
larger in the range below 0.1 Hz than those in the range between 0.1 Hz
and 1 Hz, especially for P-5 & P-7 (two lateral faces). Furthermore, it’s
interesting to note that the coherence between P-5 and P-7 featured

316
Y. He et al.
Fig. 10. Time histories of acceleration responses: Ai, acceleration amplitude; i, direction; “mark” at the top right corner denotes floor number and accelerometer type.
with a local amplitude peak around 0.02 Hz (vortex shedding). Around
this frequency, the coherent phase was about 180°, which indicated that
vortex shedding alternately occurred from the two lateral sides of the
building.
4.4. Pressure coefficients
Pressure coefficient Cp is generally expressed as
p−ps
Cp = 1
ρv 2
2 ref
where P is the measured pressure on the building (Pa), Ps is the static
reference pressure (Pa), ρ is the air density (kg/m3), and vref is the mean
wind speed at the reference height (m/s) which is the building top
height in this paper. As mentioned in Section 4.3.1, based on the wind
profile models stipulated in the load code of China [7], when the ap-
proaching west wind was normal to P-7 on the windward face, the
reference mean wind speed converted from the wind records at the
nearest weather station was 16 m/s (i.e., νref = 16 m/s), corresponding
to the 10-min period of 15:00–15:10. The pressure coefficients (Cp)
determined based on the pressure measurements during this period are
listed in Table 2. It is found that the maximum positive mean pressure
coefficients occurred on the corners and building lateral sides were 0.56
Engineering Structures 167 (2018) 308–326
and 0.80, respectively, while the maximum negative mean pressure
coefficients on the corners and building lateral sides were −1.48 and
−0.83, respectively, indicating that the values of the mean pressure
coefficient were distinct between the building corners and the side
faces. Similar phenomenon was also found for the standard deviation
pressure coefficients, the maximum standard deviation pressure coef-
ficients were 0.46 on the corners and 0.29 on the building sides, re-
spectively. For the peak pressure coefficient, the maximum and
minimum values of the positive and negative peak pressure coefficients
were found as 0.99 and −3.85 on the corners and 1.42 and −2.20 on
(4) the building sides, respectively. The peak pressure coefficients showed
more distinct differences between the building corners and the side
faces. In particular, it is noted that the mean pressure coefficients on the
windward face and leeward side were 0.80 (P-7) and −0.34 (P-5), re-
spectively, which are in good agreement with the values (0.8 and −0.3)
stipulated in the load code of China [7]. For the mean pressure coef-
ficients on the lateral sides, the measured value of −0.69 at P-8 is very
close to that (−0.7) stipulated in the code while there is difference of
15.7% at P-6 between the field measurement and that recommended in
the code.
317
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 11. PSDs of accelerations.

318
Y. He et al.
Table 3
Natural frequencies of PAFC (Hz).
Mode 1st swaying 1st torsion
Direction A1 A2 A1(A2)
Measured 0.12 0.12 0.28
FEM (ABAQUS) 0.115 0.116 0.288
Difference (%) 4.2 3.3 −2.9
Note: A1 and A2 denote the acceleration response in direction 1 and 2, respectively. Difference = (measured − FEM)/measured.
5. Acceleration responses
5.1. Time histories of acceleration responses
At each level of the 49th, 81st, 100th and 118th floors, there were
two pairs of accelerometers installed to detect the acceleration re-
sponses of PAFC (listed in Table 1). Time histories of the measured
acceleration signals in the two orthogonal directions at those floors are
depicted in Fig. 10. It can be found that at each floor, the acceleration
responses measured by two types of accelerometers show the same
changing trend and close acceleration amplitudes in each direction. It is
also observed that at different floor levels, these time history curves
show consistent fluctuation characteristics, i.e. peaks appeared around
the same moment, good agreement among these evolution curves
verifies the reliability and accuracy of these acceleration signals mea-
sured by different types of accelerometers. It is noted that the signals
detected by a B&W accelerometer in direction 2 at the 49th floor show
some unexpectedly sharp apexes which distinguished from other re-
cords. This error was caused by the fault of the accelerometer and its
records were excluded in this paper.
5.2. Natural frequencies
Both the Welch method and the Yule-Walker technique are used to
identify the natural frequencies of PAFC. The selected acceleration
signals included accelerations from 49F-B&W, 81F-B&W, 100F-B&W,
118F-B&W accelerometers in direction 1 and those from 49F-AS, 81F-
GeoSIG, 100F-Jwell, 118F-GeoSIG accelerometers in direction 2. A
stationary segment from each selected signal record was extracted for
spectral analysis. The obtained spectra are shown in Fig. 11. Through
comparison, it is observed that the peaks identified by these two
techniques were consistent for the first five modal components in the
two directions at the four floors. Judging from the frequencies corre-
sponding to the spectral peaks, it is identified that the frequencies of
0.12 Hz, 0.40 Hz and 0.80 Hz approximately correspond to the 1st, 2nd,
and 3rd swaying modes, respectively, while the other two frequencies
of 0.28 Hz and 0.69 Hz are associated with the 1st and 2nd torsional
modes, respectively. According to the natural frequencies predicted by
the commercial finite element software, i.e., ABAQUS [29], the natural
frequencies of the first few modes from both the field measurements
and the numerical predictions are listed in Table 3 for comparison
purposes. As shown in the table, the numerical predictions by the finite
element method (FEM) are in good agreement with the field measure-
ments within differences of ± 7%.
5.3. Contributions of modal components
The spectral analysis results illustrated that each acceleration re-
sponse record consisted of several vibration modes. To clarify the
contributions of different mode components, the acceleration signals
were filtered to separate each component. Fig. 12 shows 10-min stan-
dard deviation (σA) and peak response (PkA) of the separated modal
components, such as the 1st swaying mode and the 1st torsional mode,
from two directional acceleration signals on the four floors (49F-AS,
81F-GeoSIG, 100F-Jwell, 118F-GeoSIG). It can be observed directly that
319
Engineering Structures 167 (2018) 308–326
2nd swaying 2nd torsion 3rd swaying
A1 A2 A1(A2) A1 A2
0.41 0.41 0.69 0.80 0.83
0.429 0.437 0.714 – –
−4.6 −6.6 −3.5 – –
the standard deviation and peak response of each mode have con-
sistently relative scale of amplitudes among different floors. In addition,
by comparing the peak responses of the filtered modes with the un-
filtered signals shown in Fig. 10, it can be concluded that the 1st
swaying modal component dominated the acceleration responses and
the 1st torsional responses were generally one order smaller than those
of the 1st swaying mode. Moreover, Fig. 12 also illustrate that the ac-
celeration responses or the modal responses increased with the building
height (from 49th floor to 118th floor). However, it is noted that the
amplitudes of the 1st torsional mode on the 100th floor were larger
than those on the 118th floor. Actually, this anomaly was as a result of
the farther distance of the accelerometer location from the central line
on the 100th floor.
5.4. Dependence of modal parameters on response amplitude
Based on the filtered narrow bandwidth signals of each mode, the
random decrement technique (RDT) is used to estimate the modal
parameters of PAFC. The RDT was proposed by Cole [3], then got rapid
development and wide application in the fields of structural dynamics
and engineering vibrations, e.g., Tamura and Suganuma [26], Satake
et al. [22], Shi et al. [23], Yi et al. [31], He and Li [11], Huang and Gu
[10] and Li et al. [17,19,20,14,15]. This method is based on a process
with ensemble average of a number of equally-spaced segments with
displacement x ,̂ then the resulting response, namely the random de-
crement signature (RDS), is derived. Lastly, damping ratio and natural
frequency can thus be determined by fitting the signature curve. Ac-
cording to the RDT results, the dependence of the modal parameters of
PAFC on response amplitudes are illustrated and discussed herein.
Fig. 13 plots the variations of damping ratio (ξ) and natural fre-
quency (ƒ) with the acceleration response amplitudes for the 1st
swaying and torsional modes. Hence, the amplitude-dependent char-
acteristics of the modal parameters can be observed directly. The nat-
ural frequencies of both the 1st swaying and torsional modes display a
steady decrease and then about a constant as the vibration amplitude
becomes large. In other words, the natural frequencies decrease with
the increase of the response amplitudes, from about 0.123 to 0.118 Hz
for the 1st swaying mode and from about 0.284 to 0.276 Hz for the 1st
torsional mode. On the other hand, the damping ratios exhibit different
changing trends from natural frequencies even though both the swaying
mode and torsional mode also show a similar tendency. On the whole,
damping ratio curves show slight fluctuations and finally turn to be
relatively stable values. Actually it seems to decrease or fluctuate in the
low-amplitude range and then rises slowly in the relatively high-am-
plitude region to stabilize at about 0.7%. Basically, it can be observed
that the damping ratios identified with the RDT vary within the range
of 0.3% to 1.5%. The results of this field measurement study suggest
that the damping ratios of the fundamental swaying and torsional
modes of 0.5–1.2% appear reasonable for the wind-resistant design of
PAFC.
5.5. Mode shapes
As discussed by Bernardini et al. [2] and Li et al. [16], the ratio of
the mode shape coefficients between two points could be defined by the
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 12. Comparison of modal components of acceleration responses on different floors.

320
Y. He et al.
Fig. 13. Amplitude dependence of damping ratio (ξ) and natural frequency (ƒ).
ratio of their cross-power spectral density to the auto-power spectral
density in terms of responses. This spectral-based method can be de-
fined as follows (denoted as “FM1” in this paper):
ϕki / ϕpi = Spk (ωi )/ Spp (ωi ) (5)
where ϕpi and ϕki are the mode shape coefficients of point p and point k,
respectively and i = 1, 2, 3… indicating the orders of modes; ωi is the
natural frequency of the ith mode; Spk(ωi) indicates the cross-power
321
Engineering Structures 167 (2018) 308–326
spectral density of responses between point p and point k for the ith
mode while Spp(ωi) indicates the auto-power spectral density of re-
sponse at point p.
According to the discussion about the relationship between the
standard deviation of structural response and mode shapes by Tse et al.
[25], the ratio of the standard deviation of acceleration responses is
used to estimate the ratio of the mode shape coefficients as follows
(denoted as “FM2” in this paper):
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 14. Identified mode shapes of PAFC.

ϕki / ϕpi = σki/ σpi (6) such as a mode shape determined by numerical prediction such as FEM
and that based on on-site measurements. NMD is defined as follows
where σki and σpi denote the standard deviation of acceleration re- [27,19]:
sponses at points p and k corresponding to the ith mode, respectively.
Load code for the design of building structures of the People’s 1−MAC (ϕi A,ϕiB )
NMD (ϕi A,ϕiB ) =
Republic of China [7] provides the following equation as the 1st mode MAC (ϕi A,ϕiB ) (8)
shape of high-rise buildings (denoted as “Standard” in this paper).
A B th
where ϕi and ϕi denote the i measured and predicted mode shape
6z 2H 2−4z 3H + z 4 vectors, respectively.
ϕ=
3H 4 (7) The modal assurance criterion (MAC) is also considered in this study
where ϕ is the 1st mode shape, z is height, H is the total height of a to assess the correlation of mode shapes obtained by different methods
building. [1]:
The first three swaying mode shapes and the first two torsional ((ϕi A)T ϕiB )2
mode shapes derived from the above mentioned methods are plotted in MAC (ϕi A,ϕiB ) =
((ϕi ) ϕi )((ϕiB )T ϕiB )
A T A
(9)
Fig. 14 for comparison purposes. In addition, these mode shapes are
also compared with those from the commercial finite element software, Specifically, MAC = 1 means a perfect correlation between two
i.e., Midas (denoted as “FEM” in this paper). vectors, while MAC = 0 represents uncorrelated vectors. In contrast, a
The normalized modal difference (NMD) is adopted herein to esti- smaller NMD value indicates a better correlation between two vectors.
mate the correlation of mode shapes obtained by different methods, Table 4 lists the values of MAC and NMD between the measured and

322
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 14. (continued)

Table 4
Correlation index of the measured and numerical predicted mode shapes.
Mode 1st swaying 1st torsion 2nd swaying 2nd torsion 3rd swaying

Direction A1 A2 A1(A2) A1 A2 A1(A2) A1 A2

MAC FE1 0.998 0.997 0.896 0.995 0.982 0.949 0.965 0.969
FE2 0.999 0.998 0.875 0.996 0.995 0.907 0.901 0.855

NMD (%) FE1 4.5 5.1 23.2 7.4 13.5 34.0 18.9 17.8
FE2 3.6 4.3 32.0 6.6 7.0 37.8 33.2 41.1

calculated mode shapes. It is observed that all the MAC values are
greater than 0.85, illustrating that the measured and calculated mode
shapes are in good agreement.
It can be observed that the mode shapes of the first two swaying
modes and the first torsional mode identified from the field measure-
ments and FEM show good agreement, though there are differences to
some extent for the higher modes. In addition, the Chinese design code
provides a satisfactory description of the 1st swaying mode shape of the
super-tall building. It can be concluded that there are good correlations
between the measured and calculated mode shapes, while the lower-
order mode shapes show better correlations than the higher-order ones.
5.6. Normalized amplitude dependence of modal parameters
In order to observe the amplitude-dependent characteristics of
structural modal parameters more clearly, the acceleration amplitudes

323
Y. He et al. Engineering Structures 167 (2018) 308–326

Fig. 15. Normalized amplitude-dependent damping ratios (ξ) and natural frequencies (ƒ).

of a mode at different height levels are normalized by its mode shape.
The normalized amplitude-dependent natural frequencies and damping
ratios are illustrated in Fig. 15. After the normalization, all the response
amplitudes at different height levels are adjusted to a uniform di-
mensionless scale, i.e., ‘1’. This is helpful for a direct observation about
the amplitude-dependent characteristics of the modal parameters. As
discussed in Section 5.4, the identified natural frequencies decreased
with the increase of the response amplitude while the damping ratios
showed a fluctuation with the response amplitude. Now, it is interesting
to observe that the modal parameters (damping ratio and natural fre-
quency) identified from the four floors equipped with accelerometers
show consistent variations and are in good agreement, which indicates
the independence and validity of the measured acceleration signals at
the different height levels and further demonstrates the accuracy and
reliability of the identified modal parameters.

324
Y. He et al.
5.7. Assessment of serviceability of PAFC
It has been widely accepted that wind-induced acceleration re-
sponse is a preferable index for serviceability assessment of high-rise
structures. Melbourne and Palmer [21] proposed a comfort criterion in
terms of the upper limit of acceleration response with consideration of
the return period of design wind speed:
Amax = 2ln(fT ) (0.68 + 0.2lnR)exp(−3.65−0.41lnf ) (10)
where Amax is the maximum of acceleration responses (cm/s ). ƒ is the
2 (5) By comparing the mode shapes derived from the field measure-
building fundamental natural frequency in Hz. T is the observational
duration in seconds, and R is the return period in years.
Taking ƒ = 0.12 Hz for PAFC, given an observation duration of
T = 600 s, values of Amax for different return periods predicted by Eq.
(10) are 12.3 cm/s2 (1-year return period) and 18.2 cm/s2 (5-yeas), and
20.7 cm/s2 (10-year). As shown in Fig. 10, the maximum of acceleration
responses was 4.4 cm/s2 in direction 1, 7.2 cm/s2 in direction 2 and
8.4 cm/s2 for combination of directions 1 and 2, which were measured
by the accelerometers on the top floor of PAFC. It is evident that the
measured acceleration responses of PAFC during Typhoon Haima were
well confined within the upper limits suggested by Melbourne and
Palmer [21]. Thus, the serviceability performance of the 600 m high
PAFC satisfactorily met the requirement of the comfort criterion during
Typhoon Haima.
6. Concluding remarks
This paper presented the field measurements of wind effects on
600 m high PAFC during Typhoon Haima. Wind-induced cladding
pressures and structural responses as well as modal parameters of the
skyscraper were presented and discussed based on detailed analysis of
the measurements. Main findings and conclusions are summarized
below:
(1) From the wind pressure measurements at four central points and
four corners at the 81st floor during Typhoon Haima, it was found
that the measured cladding pressures demonstrated significant
differences between the building central locations and corners. The
magnitudes of the maximum peak negative pressure appeared on a
building corner was 74% and 49% larger than the maximum posi-
tive and negative pressures at its adjacent central point, implying
that the maximum peak negative pressures likely occur at building
corners or concave regions. Moreover, the mean pressure coeffi-
cients obtained from the field measurements were in good agree-
ment with those stipulated in the load code of China [7].
(2) By examination of the skewness and kurtosis index of the measured
cladding pressures at both the building central points and corners,
the probability distribution of negative pressures illustrated distinct
non-Gaussian characteristics which became more evident with the
increase of negative pressure amplitude. On contrast, the measured
positive wind pressures appeared to follow the normal distribution.
(3) From spectral analysis of cladding pressures, the phenomenon of
vortex shedding occurred on lateral faces of the building was ob-
served, which also provided the information of the influence of
building shape on the cladding pressures, e.g., the bulged shape of
the building corners generated local vortex shedding at relatively
high frequencies. In addition, the coherence of wind pressures be-
tween two lateral sides of the building had a local amplitude peak
around the vortex shedding frequency when the approaching wind
was normal to a building face.
(4) By comparison of the structural modal parameters from the field
measurements and numerical predictions by FEM, the natural fre-
quencies identified from these two methods were in good agree-
ment within differences of ± 7%. The spectral analysis results of
the measured structural dynamic responses illustrated that the 1st
325
Engineering Structures 167 (2018) 308–326
swaying modal component dominated the acceleration responses of
PAFC and the 1st torsional responses were generally one order
smaller than those of the 1st swaying mode. According to the RDT
results, the identified natural frequencies decreased with the in-
crease of the response amplitude while the damping ratios showed a
fluctuation with the response amplitude. It is suggested that
damping ratios of the fundamental swaying and torsional modes of
0.5%-1.2% appeared reasonable for the wind-resistant design of
super-tall buildings similar to PAFC.
ments and predictions by FEM analysis, it was observed that there
were good correlations between the measured and calculated mode
shapes of the swaying and torsional components, and the lower-
order mode shapes showed better correlations than the higher-
order ones. In addition, on the basis of the identified mode shapes,
the normalized amplitude-dependent characteristics of the modal
parameters (damping ratio and natural frequency) determined from
different floors showed consistent variations and were in good
agreement.
(6) Based on the field measurements of acceleration responses, it was
concluded that the serviceability performance of the 600 m high
PAFC satisfactorily met the requirement of the comfort criterion
during Typhoon Haima.
Acknowledgements
The work described in this paper was fully supported by a grant
from the Research Grants Council of Hong Kong Special Administrative
Region, China (Project No: CityU 11256416) and a grant from National
Natural Science Foundation of China (Project No. 51778554).
References
[1] Allemang RJ, Brown DL. A correlation coefficient for modal vector analysis. In:
Proceedings of the 1st international modal analysis conference, vol. 1. Orlando:
Union College Press; 1982., p. 110–6.
[2] Bernardini E, Spence SM, Kwon DK, Kareem A. Performance-based design of high-
rise buildings for occupant comfort. J Struct Eng 2014;141(10):04014244.
[3] Cole HA. On-line failure detection and damping measurement of aerospace struc-
tures by random decrement signatures. NASA CR-2205. VA: National Technical
Information Service; 1973.
[4] Cochran L, Erickson R. A physical modeler’s view of computational wind en-
gineering. J Wind Eng Ind Aerodyn 2011;99:139–53.
[5] Chen W, Lu Z, Lin W, Chen S, Ni YQ, Xia Y, et al. Theoretical and experimental
modal analysis of the Guangzhou New TV Tower. Eng Struct 2011;33(12):3628–46.
[6] Fu JY, Wu JR, Xu A, Li QS, Xiao YQ. Full-scale measurements of wind effects on
Guangzhou West Tower. Eng Struct 2012;35:120–39.
[7] GB 50009–2012. Load code for the design of building structures. Beijing: China
Building Industry Press; 2012.
[8] Gioffrè M, Gusella V, Grigoriu M. Non-Gaussian wind pressure on prismatic build-
ings. II: Numerical simulation. J Struct Eng 2001;127(9):990–5.
[9] Guo Y, Kareem A, Ni YQ, Liao W. Performance evaluation of Canton Tower under
winds based on full-scale data. J Wind Eng Ind Aerodyn 2012;104:116–28.
[10] Huang Z, Gu M. Envelope random decrement technique for identification of non-
linear damping of tall buildings. J Struct Eng 2016;142(11):04016101.
[11] He YC, Li QS. Dynamic responses of a 492-m-high tall building with active tuned
mass damping system during a typhoon. Struct Control Health Monit
2014;21(5):705–20.
[12] Kijewski-Correa T, Kilpatrick J, Kareem A, Kwon DK, Bashor R, Kochly M, et al.
Validating wind-induced response of tall buildings: synopsis of the Chicago full-
scale monitoring program. J Struct Eng 2006;132(10):1509–23.
[13] Kijewski-Correa T, Pirnia JD. Dynamic behavior of tall buildings under wind: in-
sights from full-scale monitoring. Struct Design Tall Special Build
2007;16(4):471–86.
[14] Li QS, Li X, He YC. Monitoring wind characteristics and structural performance of a
supertall building during a landfall typhoon. J Struct Eng 2016;142:04016097.
[15] Li QS, Li X, He Y, Yi J. Observation of wind fields over different terrains and wind
effects on a super-tall building during a severe typhoon and verification of wind
tunnel predictions. J Wind Eng Ind Aerodyn 2017;162:73–84.
[16] Li ZN, Pan YY, Sang C, Shi WH. Study on wind-induced vibration characteristics of
coastal high-rise building in Xiamen. J Build Struct 2013;34(6):22–9. (in chinese).
[17] Li QS, Xiao YQ, Wong CK. Full-scale monitoring of typhoon effects on super tall
buildings. J Fluids Struct 2005;20(5):697–717.
[18] Li QS, Xiao YQ, Wu JR, Fu JY, Li ZN. Typhoon effects on super-tall buildings. J
Sound Vib 2008;313(3):581–602.
Y. He et al.
[19] Li QS, Zhi LH, Tuan AY, Kao CS, Su SC, Wu CF. Dynamic behavior of Taipei 101
Tower: field measurement and numerical analysis. J Struct Eng
2010;137(1):143–55.
[20] Li QS, Zhi LH, Yi J, To A, Xie J. Monitoring of typhoon effects on a super-tall
building in Hong Kong. Struct Control Health Monit 2014;21(6):926–49.
[21] Melbourne WH, Palmer TR. Acceleration and comfort criteria for buildings un-
dergoing complex motions. J Wind Eng Ind Aerodyn 1992;44:105–16.
[22] Satake N, Suda KI, Arakawa T, Sasaki A, Tamura Y. Damping evaluation using full-
scale data of buildings in Japan. J Struct Eng 2003;129(4):470–7.
[23] Shi W, Shan J, Lu X. Modal identification of Shanghai World Financial Center both
from free and ambient vibration response. Eng Struct 2012;36:14–26.
[24] Thomson DJ. Spectrum estimation and harmonic analysis. Proc IEEE
1982;70(9):1055–96.
[25] Tse KT, Hitchcock PA, Kwok KCS. Mode shape linearization for HFBB analysis of
wind-excited complex tall buildings. Eng Struct 2009;31(3):675–85.
326
Engineering Structures 167 (2018) 308–326
[26] Tamura Y, Suganuma SY. Evaluation of amplitude dependent damping and natural
frequency of buildings during strong winds. J Wind Eng Ind Aerodyn
1996;59:115–30.
[27] Waters TP. Finite element model updating using measured frequency response
functions. Ph.D. thesis. UK: Univ. of Bristol; 1995.
[28] Welch P. The use of fast Fourier transform for the estimation of power spectra: a
method based on time averaging over short, modified periodograms. IEEE Trans
Audio Electroacoust 1967;15(2):70–3.
[29] Yang XQ, Fu XY, Huang YJ. Dynamic elastoplastic analysis of the Shenzhen Ping-An
Financial Center Tower. J Build Struct 2011;32:40–9. [in Chinese].
[30] Yan BW, Li QS. Large-eddy simulation of wind effects on a super tall building in
urban environment conditions. Struct Infrastruct Eng 2016;12:765–85.
[31] Yi J, Zhang JW, Li QS. Dynamic characteristics and wind-induced responses of a
super-tall building during typhoons. J Wind Eng Ind Aerodyn 2013;121:116–30.