You are on page 1of 10

Received: 4 December 2015 Revised: 12 September 2016 Accepted: 7 March 2017

DOI: 10.1002/tal.1376

RESEARCH ARTICLE

Wind effects on Shenzhen Zhuoyue Century Center: Field


measurement and wind tunnel test
Haoran Pan1 | Zhuangning Xie1 | An Xu2 | Li Zhang3

1
School of Engineering and Transportation,
Guangzhou, China Summary
2
Engineering Technology Research and Field measurements of wind effects on Zhuoyue Century Center were conducted during 4
Development Center for Structural Wind typhoon events in the recent 5 years, during which the field data such as wind speeds, wind direc-
Resistance and Health, Monitoring in tions, and acceleration responses were simultaneously and continuously measured. On the basis
Guangdong Province, Guangzhou University,
of field measured data, dynamic characteristics of this super‐tall building were determined by
Guangzhou, China
3 recently developed fast Bayesian fast Fourier transform method. Using full‐scale measurement
Shenzhen Key Laboratory of Severe Weather
in South China, Meteorological Bureau of data under 4 typhoons and breezy conditions for modal identification, one could observe a rela-
Shenzhen Municipality, Shenzhen, China tively wide scatter in the identified modal damping ratios, and the damping ratios do not appear
Correspondence to have an obvious nonlinear relationship with vibration amplitude. The average damping ratios of
Zhuangning Xie, 381 Wushan Road, School of
the first 2 modes were 0.70% and 0.73%, respectively. Serviceability of the super‐tall building
Civil Engineering and Transportation,
Guangzhou 515041, China. under wind action was analyzed on the basis of the field measured response. Finally, the mea-
Email: znxie@scut.edu.cn sured wind‐induced acceleration responses were further compared with those obtained from
the wind tunnel test to evaluate the accuracy of the model test results.
Funding information
National Natural Science Foundation of China, KEY W ORDS
Grant/Award Number: 512782904
full‐scale measurement, modal parameter identification, tall building, wind effect, wind tunnel test

1 | I N T RO D U CT I O N comparisons between the wind tunnel test results and full‐scale mea-
surement have been rarely made for super‐tall buildings. Therefore, it
For a super‐tall building with height of over 200 m, wind load is the is always desirable to make such comparisons to evaluate the accuracy
controlling factor in the design stage. Full‐scale measurement is of the model test results as well as the adequacy of the techniques
considered to be the most reliable way to obtain wind load and applied in tests.
wind‐induced response. Successful implements of full‐scale monitoring Zhuoyue Century Center (ZCC) located in a typhoon prone city—
systems in terms of structural wind‐induced effect and investigations Shenzhen, has a height of 280 m. Its investigation of wind‐induced
of modal identification have been widely reported since 1980s.[1–4] effects and dynamic characteristic can provide valuable information
However, in previous studies of full‐scale surveys, many results were for wind‐resistant design. In this paper, the fast Bayesian fast Fourier
obtained from buildings in the vicinity of 40 storeys or shorter in transform (FBFFT) approach was adopted to identify modal parame-
a small amplitude region, it is more imperative to investigate the ters with measured data of this building under four typhoon events
structural dynamic characteristics of super‐tall buildings (building in recent 5 years. The identified parameters and their uncertainties
height > 200 m) under strong wind. Besides, identified damping ratios were analyzed and discussed. Comparisons between response of wind
show a great dispersion in the practice of Random Decrement tunnel tests and full‐scale measurement were also presented to evalu-
Technique (RDT)—a conventional identification method used in wind ate the reliability of wind tunnel tests.
engineering but the accuracy of identified results cannot be judged.
Wind tunnel testing is an effective method to obtain the dynamic
wind load and to evaluate serviceability of a super‐tall building. 2 | D E S C R I P T I O N OF T H E I N S T R U M E N T E D
Nevertheless, because of the uncertainties in structural damping and B U I LD I N G
the discrepancy between the actual field conditions, that is, incident
turbulence and terrain characteristics, and those in wind tunnel tests, ZCC located in Shenzhen, China, has 68 stories and is about 280 m
the model test results are still far from convincing. At present, from the ground. On its 68th floor, a wireless accelerometer (Type

Struct Design Tall Spec Build. 2017;e1376. wileyonlinelibrary.com/journal/tal Copyright © 2017 John Wiley & Sons, Ltd. 1 of 10
https://doi.org/10.1002/tal.1376
2 of 10 HAORAN ET AL.

LAC‐I; see Figure 1a) has been installed to monitor the structural accel- the prediction error that accounts for the discrepancy between mea-
eration response along the two main axes of the building at a sampling sured response and theoretical model response; N is the sampling
rate of 25 Hz, as shown in Figure 1b,c. Acceleration signals are number; n is the number of measured DOFs.
recorded continuously by LAC‐I at a measuring range of ±128 milli‐g The FFT of y€ j is defined as
and a resolution of less than 2 μg. This type of accelerometer is espe-
rffiffiffiffiffiffiffiffi N  

cially designed for the measurement of wind‐induced response of 2Δt ðk−1Þðj−1Þ
Fk ¼ y€ j exp −2πi ; (2)
structures, and its appropriate measuring range guarantees recorded N j¼1 N
data with a high SNR. Having been recorded, the data will be sent to
the main computer in South China University of Technology through where i2 = − 1, Δt is the sampling interval, k = 1 , . . . , Nnqy with
a GPRS network. Nnqy = int (N/2) + 1, int[] denotes the integer part of its augment.
Let Zk = {ReFk; ImFk} ∈ R2n be an augmented vector of the real and
imaginary part of Fk. In practice, only the FFT data confined to a
3 | A N A L YS I S M E TH O DO LO GY
selected frequency band dominated by the target modes is used for
Bayesian inference. Let such collection be denoted by {Zk}. Using
A FBFFT method was used to identify the modal parameters using the
Bayes' Theorem and assuming no prior information, the posterior
field‐measured data. The theory was briefly described in this section.
PDF of θ given the data is given by
The original formulation is based on Yuen and Katafygiotis,[5] and the
fast algorithms that allow practical implementation can be referred to pðθjfZk gÞ ∝ pðfZk gjθÞ: (3)
Au[6] and Au[7,8]. One major merit of this method is that not only can
it identify modal properties but also quantify their associated uncer- For long duration of data and small sampling interval, it can be
tainties. The basic idea is that for a linear elastic classically damped proved that the FFT data at different frequencies are asymptotically
structure under broad‐banded excitation, the real and imaginary parts independent and their real and imaginary part follow a Gaussian distri-
of the fast Fourier transform (FFT) of the structural response follow a bution. The likelihood function is given by
multidimensional Gaussian distribution, which can be characterized  
1 1
analytically in terms of the modal parameters. By maximizing the poste- pðfZk gjθÞ ¼ ∏ exp − ZTk Ck −1 Zk ; (4)
rior probability density function (PDF) of the modal parameters given k ð2πÞn ð detCk Þ1=2 2

the FFT data, the most probable modal parameters can be determined.
where det(.) is the determinant, and Ck denotes the covariance
The acceleration time history measured at n degrees of freedom
matrix of Zk:
(DOFs) of a structure is modeled as
  " #
1 Φ ReHk − ImHk T
Φ Sη
y€ j ¼ x€ j ðθÞ þ ηj ; j ¼ 1; 2; :::; N; (1) Ck ¼ þ I2n : (5)
2 Φ ImHk ReHk ΦT 2

where x€ j ∈Rn is the model acceleration response of a structure that Φ = [φ1, φ2, ... , φm] ∈ Rn × m is the structural mode shape matrix
depends on the set of modal parameters θ to be identified; ηj ∈ Rn is with φi (i = 1, ... , m) being the ith mode shape; I2n ∈ R2n denotes the

FIGURE 1 Accelerometer and instrumented building (a) LAC‐I wireless accelerometer developed by the project team, (b) accelerometer installed
along the two main axes of ZCC, (c) and image of the building
HAORAN ET AL. 3 of 10

identity matrix; and Hk ∈ Rm × m is the transfer matrix whose (i, j) response was selected as the triggering level for each sample segment.
element is given by Actually, the RDT results have been reported in a recent publication,[9]
and further details can be found for the steps of adopting RDT.
Sij
Hk ði; jÞ ¼ h  ih  i; (6) The characteristics of typhoon‐generated wind and wind‐induced
2
βik −1 þ 2iζ i βik βik 2 −1 −2iζ i βik responses of this building were also presented and discussed.

where βik = fi/fk; fi is the natural frequency of the ith mode; fk is the FFT
frequency abscissa; Sij is the cross spectral density between the ith and 4.1 | Strong wind events and data
jth modal excitation. In practice, it is more convenient to work with the The vibration data investigated in this study were measured during
negative log‐likelihood function: four strong wind events, namely, Typhoon Nesat in 2011, Typhoon
h i Doksuri and Typhoon Vicente in 2012, and Typhoon Usage in 2013.
LðθÞ ¼
1
2
∑ ln det Ck ðθÞ þ Zk T Ck ðθÞ−1 Zk ; (7) The 1‐hr‐average‐recorded wind speed and direction data were
k
obtained from the weather stations that are located at the upwind
so that direction of ZCC. As seen in Figure 2, Nanshan Station is located in
the northwest of ZCC at a distance of 9 km, and Luofang Station is
pðθjfZk gÞ∝ exp½−LðθÞ: (8) located in the northeast of ZCC at a distance of 10 km. Besides, wind
data taken at the Yantian Harbor Station (as shown in Figure 2) during
Thus, the most probable values (MPV) of θ can be obtained by
Vicente were used to compare the wind tunnel results calculated with
minimizing L(θ). However, it can be shown that the minimization
data at Luofang Station. This will be discussed further in the following
process is ill conditioned because of large number of variables
section. Limited by space, only the wind speed and direction history
involved. The reader is referred to Au (2011) and Au (2012a,b) for fast
during Vicente are plotted here, as shown in Figure 3. Table 1 gives
algorithms that allow the MPV and covariance matrix to be computed
an overview of the wind condition of four typhoon events. The wind
efficiently for well‐separated modes or in general.
direction is measured clockwise from the north.
One can evaluate the uncertainties of identified parameters
In order to obtain reliable and accurate damping ratios, a long
through the “coefficient of variation” (COV), defined as the ratio of
enough record is usually required in the practice of RDT. For compar-
the posterior standard derivation to the MPV. The posterior standard
ison purposes, a 2‐hr time history segment containing the maximum
deviation can be obtained from the square root of the corresponding
acceleration response of ZCC measured during each typhoon's
diagonal element of the posterior covariance matrix.
passage was applied to modal identification through FBFFT. Coinci-
dentally, during each typhoon segment, there were no significant var-
4 | A N A L Y S I S O F B E N C H M A R K BU I L D I N G iations of wind direction. Therefore, the mean wind direction can be
regarded as constant during each period. Table 2 summarizes acquisi-
In this section, FBFFT was applied to modal identification of recorded tion time of acceleration data of each typhoon event, and Figure 4
acceleration data from ZCC under four strong wind events in recent shows the acceleration time history of 2‐hr time window measured
5 years. Comparison was made between FBFFT results and the RDT during Typhoon Vicente. The recorded maximum wind‐induced accel-
results. The RDT approach was carried out with level‐crossing trigger- eration response was approximately 11.0 milli‐g during Typhoon
ing condition, in which one standard deviation of each acceleration Vicente in 2012. As mentioned in Xu et al.,[9] such large peak

FIGURE 2 Location of weather stations


4 of 10 HAORAN ET AL.

FIGURE 3 Wind data during typhoon Vicente

TABLE 1 Recorded maximum 1‐hr mean wind speed and the corresponding wind direction during the passage of four typhoons
Maximum1‐hr Corresponding wind
Typhoon Recorded date Recorded Station mean wind speed direction (°)

Nesat 2011.09.29 Luofang 12.0 90


Doksuri 2012.06.29–30 Luofang 11.1 90
Vicente 2012.07.23–24 Luofang 17.1 90
Yantian Harbor 26.1 110
Usage 2013.9.22–23 Nanshan 15.1 270

acceleration value of super‐tall buildings subjected to wind loads is sel- Table 3. Frequency bands (Hz) used for identification of the first
dom reported in past papers, which means that the measured data in three modes were [0.161 0.198], [0.210 0.238], and [0.328 0.39].
this study can provide valuable knowledge in the investigation of the For the posterior uncertainties, it can be found that the posterior
dynamic characteristics of super‐tall buildings under strong wind. COVs of the natural frequencies are less than 0.2%, which implies
the uncertainties of the identified parameters are small. Meanwhile,
4.2 | Analysis results the identified frequency values under windless ambient excitations
were presented to serve as a reference in this table. As the third
4.2.1 | Natural frequency
mode (torsional mode) was quite insignificant, only the first two
Figure 5 shows the PSD of the acceleration response of ZCC during
translational modes were given. The identified frequencies under dif-
typhoon Vincente. The spectrum was obtained from the measured
ferent conditions were very close, indicating that they can be
acceleration data from LAC‐I. The blue and red lines denote the mea-
regarded as constant under wind actions. Table 3 also lists the RDT
sured data in the sensor x and y directions, respectively. It can be seen
results under the same excitations. As shown in the table, excellent
from the figure that the wind‐induced responses of the building are
consistency between these two different approaches is observed
primarily in the two fundamental sway modes of vibration, and the
for frequency identification.
response of the first torsional mode (about 0.35 Hz) is insignificant.
Table 4 gives mean values of identified frequency under different
From the spectrum, one can also observe the first two modes are
wind conditions using FBFFT and the calculated result from the
closely spaced. This is ascribed to the approximate symmetric struc-
computational model by the finite element method. It is shown that
tural design in the two orthogonal directions.
there are about 6.14–33.43% differences between the calculated and
Using the data in each typhoon, the MPV and posterior COV of
measured natural frequencies for the first three modes of ZCC. This
natural frequency were calculated through FBFFT, as presented in
finding is consistent with the result in literature review,[10,11] which

TABLE 2 Acquisition time of acceleration data implies the natural frequencies are usually a little conservative and
hence the structural response would be overestimated in the design
Maximum
Typhoon Acquisition time acceleration (mg) stage of tall buildings.

Nesat x dir 2011.09.29 00:44:35–02:44:35 2.20


y dir 3.20
4.2.2 | Damping
Doksuri x dir 2012.06.29 21:36:14–23:36:14 0.90
y dir 2.60 The MPV and posterior COV of damping were obtained through
Vicente x dir 2012.07.24 01:31:09–03:39:09 3.00 FBFFT given the acceleration data in each typhoon, as shown in
y dir 11.0 Table 5. The values of damping were less than 1%, and they were iden-
Usage x dir 2013.09.22 22:37:46–00:37:46 1.90 tified with higher uncertainties than the natural frequencies. While the
y dir 2.95
frequencies had a COV of 0.1%, the damping ratios had a COV in the
HAORAN ET AL. 5 of 10

FIGURE 4 Acceleration response during typhoon Vicente

TABLE 5 Identified damping ratio (%) for the first two modes
Nesat Doksuri Vicente Usage

FBFFT Mode 1 0.74(12.2%) 0.73(12.5%) 0.65(12.1%) 0.87(11.5%)


(COV) Mode 2 0.63(12.1%) 0.84(10.5%) 0.79(11.0%) 0.75(11.5%)
RDT Mode 1 0.66 0.79 0.75 0.92
Mode 2 0.73 0.88 0.83 0.87

Note. COV = coefficient of variation; FBFFT = fast Bayesian fast Fourier


transform; RDT = random decrement technique.

order of 10%. It is shown that the accuracy in the damping ratios gov-
erns the data length requirement.[12] To reduce the uncertainty in the
damping ratios, a longer data length for identification is required, as the
COV generally follows an inverse square root law with the data length.
However, an extended period is not recommended in practice, because
wind loads and structural response amplitude may change significantly
during this period, which would weaken the stationary assumption in
FIGURE 5 Power spectral density of acceleration in typhoon Vincente the stochastic model excitation. The RDT results were also provided in
Table 5. It can be found that these results are a little different from
TABLE 3 Identified natural frequency (Hz)
those evaluated by FBFFT. This may be relevant to the dispersion of
Ambient the damping ratios.
Nesat Doksuri Vicente Usage excitations
More field data during the four typhoons and breezy conditions
FBFFT Mode 1 0.181 0.181 0.177 0.176 0.178 were utilized to observe the relationship between damping and the
(COV) (0.10%) (0.08%) (0.11%) (0.09%) (0.10%) vibration amplitude along x and y directions. Figure 6a,b shows the
Mode 2 0.222 0.221 0.214 0.216 0.219
(0.09%) (0.11%) (0.14%) (0.10%) (0.10%) scatter plot of the MPV of damping ratios versus the root mean square
Mode 3 0.373 0.367 0.352 0.354 — (RMS) of acceleration for the first sway modes in each direction. In
(0.13%) (0.11%) (0.14%) (0.13%)
these figures, different symbols denote the identified results from dif-
RDT Mode 1 0.181 0.180 0.177 0.177 0.178
Mode 2 0.222 0.223 0.214 0.216 0.220 ferent typhoons or breezy cases and each point is centered at the MPV
Mode 3 0.376 0.371 0.354 0.355 — with an error bar covering ± one posterior standard derivation of the

Note. COV = coefficient of variation; FBFFT = fast Bayesian fast Fourier modal parameter. This interval includes a probability of 68.3% for the
transform; RDT = random decrement technique. modal damping ratio falling in this range because the posterior PDF
is approximately Gaussian. The identified values of damping ratio show

TABLE 4 Comparison between the FEM and the identified results a relatively wide scatter in the plots of Figure 6a,b, ranging from 0.40%
to 1.15% along x direction and from 0.46% to 1.05% along y direction.
FEM (Hz) Identified (Hz) Differencea (%)
Almost all the posterior COVs of damping ratio were larger than 10%,
Mode 1 0.168 0.179 6.14 reflecting large uncertainties.
Mode 2 0.170 0.218 22.01 Next, RDT was employed to evaluate the amplitude‐dependent
Mode 3 0.241 0.362 33.43 damping ratios of ZCC. The result is shown in Figure 7. The trends in
Note. FEM = finite element method. both Figures 6 and 7 are similar, showing that the damping ratios do
a
Difference = (Mean of identified‐FEM)/Mean of identified. not appear to have an obvious nonlinear relationship with vibration
6 of 10 HAORAN ET AL.

FIGURE 6 Relationship between identified damping ratios using FBFFT = fast Bayesian fast Fourier transform and the root mean square (RMS) of
acceleration ( : Usage : Doksuri : Nesat : Vicente : Breezy conditions)

FIGURE 7 Relationship between identified damping ratios using random decrement technique and the root mean square (RMS) of acceleration
( : Usage : Doksuri : Nesat : Vicente : Breezy conditions)

amplitude, which is totally different from the reported results in some TABLE 6 Peak acceleration at the top of the building
previous studies.[13,14] The average damping ratios through FBFFT
Measured Occupancy comfort
along x and y directions were 0.73% and 0.70%, respectively. Typhoon peak value (milli‐g) criterion (milli‐g)
Currently, there is a lack of theoretical approach for estimating Nesat x‐dir 2.20 25
damping ratios of buildings prior to construction. In the wind tunnel y‐dir 3.20
tests at the design stage, structural damping ratios are generally Vicente x‐dir 3.00
y‐dir 11.0
assumed to be 1.5% or higher for evaluating the serviceability of build-
Doksuri x‐dir 0.90
ings. From the above identified results, it appears that the damping
y‐dir 2.60
ratios of these buildings may be a little overestimated.
Usage x‐dir 1.90
y‐dir 2.95
4.2.3 | Serviceability analysis
For modern flexible super‐tall buildings, wind‐induced serviceability
issues are always the main concern in the design. Although codes for Note that the maximum 1‐hour mean wind speed recorded during
the design of tall buildings vary in different countries or regions, it is Typhoon Vicente was 26.1 m/s (recorded in Yantian Harbor, at the
widely accepted that building acceleration is the most appropriate elevation of 10 m), and the design wind speed with 10‐year return
response component for establishing checking procedure for structural period for Shenzhen by the design codes and standards of China is
serviceability requirements under wind action. In Chinese technical 26.8 m/s, 10‐minute average.[16] As the mean wind speeds depend upon
[15]
specification for the design of tall buildings, only the criterion of the average time, the 10‐min mean wind speed would exceed the hourly
occupancy comfort for the case of design wind with 10‐year return mean. In this regard, Typhoon Vicente can be regarded as 10‐year return
period is given, which is 25 milli‐g for office buildings/hotels and period design wind. As shown in Table 6, all peak values measured atop
15milli‐g for apartments. Table 6 gives the comparison between the the building are below the occupancy comfort criterion specified in the
occupancy comfort criterion and the peak value at the top of ZCC China design code. Therefore, the wind‐induced response of ZCC
during four typhoon events. satisfactorily meets the occupancy comfort criterion for this case.
HAORAN ET AL. 7 of 10

5 | WIND TUNNEL EXPERIMENTS The spectral density of the generalized displacement SX(f) is
given by
The acceleration response of the full‐scale measurement, its modal
identification result, and serviceability analysis were given in the above SX ðf Þ ¼ jHðf Þj2 SF ðf Þ; (10)
sections. The following section discussed the wind tunnel testing of
where SF(f) is the spectral density of the generalized force and the |H(f)|2
ZCC. The result serves to make a comparison with the actual perfor-
mance of this building. In this study, the high‐frequency force‐balance is the mechanical admittance ¼  1 , ζ0 is
½1−ðf=f 0 Þ2  þ4ζ 20 ðf=f 0 Þ2
2
K 2p
method was chosen for the wind tunnel tests, as it allows repeatable
the damping ratio.
response predictions based on the measured modal force spectra but
Accordingly, the power spectral density of the generalized force SF
considering different building dynamic characteristic without the
(f) can be obtained by the following equation:
requirement of additional wind tunnel testing.

1
SF ðf Þ ¼ Sm ðf Þ; (11)
5.1 | Experimental arrangements and definition of H2
parameters
where the power spectral density of the base moment Sm(f) (prototype
The wind tunnel experiment was carried out in the boundary scale) can be determined directly from the force balance test.
layer wind tunnel at South China University of Technology. The The peak acceleration response can be written as follows:
dimensions of the working section of the wind tunnel are 24 m
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(L) × 5.4 m(W) × 3 m(H). σF π 1 f 0 SF ðf 0 Þ
σa ¼ g ; (12)
On the basis of the categorization of the terrain surrounding the Mp 4 ζ 0 σ2F
sites and the China National Load Code (GB 50009–2012), a typical
where σF is the RMS generalized force; SF(f0) is the power spectrum of
boundary layer wind flow representing urban flow environment
the generalized force at the natural frequency f0; and g is the peak gust
(specified as terrain C in code) was simulated, which had a power law pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
exponent of α = 0.22 (corresponding to gradient height of 450 m) with factor ¼ 2 lnðf 0 T Þ þ 0:5772= 2 lnðf 0 T Þ , T is the averaging period
turbulence intensity at the top level of the model about 11%. The inte- equal to 600 s.
gral scale of turbulence Lu at the height of the model was 0.64 m. The
wind field was simulated by placing a barrier at the entrance of the 5.2 | Comparison between the field measurement
wind tunnel, and triangular‐shaped spires and arrayed cubic roughness and wind tunnel results
elements with different sizes on the tunnel floor upstream of the build-
It is always of considerable interest and practical use to compare the
ing model. Figure 8a,b gives the measured mean wind speed, turbu-
model tests results with the actual performance for a super‐tall build-
lence intensity at various heights, respectively. The normalized power
ing. Using the obtained data from wind tunnel tests and the identified
spectra of the fluctuating wind speed are presented in Figure 8c,
frequencies during typhoon Vicente, the peak acceleration responses
showing a good agreement with Karman‐type spectrum.
were calculated by Equation 12. Figure 10 shows the comparison
High‐frequency base tests were performed at a geometric scale of
between the model test and full‐scale measurements results for the
1:400 and a time scale of 1/154. Figure 9a shows a photo of the
peak acceleration responses along x and y directions. Peak acceleration
models mounted in the wind tunnel representing the existing sur-
responses calculated with different maximum 1‐hr mean wind speed
rounding conditions within a full‐scale radius of approximately 500 m
V10 and corresponding wind direction β taken from Luofang Station
from ZCC. In the wind tunnel test, wind direction is defined as β from
and Yantian Harbor Station are presented in these figures. The solid
the north along a clockwise direction varied from 0° to 360° with
marker in each figure denotes the field measured result, which is
increment of 10°, as shown in Figure 9b.
centered at the MPV of the identified damping ratio with an error
For clarity, the theory of high‐frequency force balance is briefly
bar covering ± one posterior standard derivation of the modal
presented here. This result is applicable for fundamental mode with a
parameter.
linear mode shape. Consider a building modeled as a multidegree‐of‐
As shown in Figure 10, the result of model tests indicates that the
freedom linear elastic structure subjected to aerodynamic forces satis-
resulting peak acceleration of the building decreases with increasing
fying the following modal equation of motion:
structural damping. Meanwhile, it can be shown that the acceleration

€ ðtÞ þ Cp X_ ðtÞ þ K p XðtÞ ¼ F ðtÞ; response of model tests has a positive correlation with the maximum
Mp X (9)
1‐hr mean wind speed. The peak acceleration responses from the field

where X(t) is the generalized displacement, Mp is the generalized mass measurements in the across‐wind direction (along y direction) were
H much larger than those in the along‐wind direction (along x direction)
¼ ∫ 0 mðzÞφ
ðzÞdz, m(z) is weight of the floor at the height of z, the
2
during four typhoon events, which is consistent with the observations
z in the wind tunnel tests. It can also be found from the figures that
mode shape φðzÞ ¼ H , H = height of the structure; Cp is the generalized
damping; Kp is the generalized stiffness =(2πf0)2Mp; f0 is the natural fre- the calculated result using the wind data from Luofang Station
H (V10 = 17.1 m/s, β = 90°) agrees better with the field measured result.
quency; and F(t) is the generalized forces ¼ ∫ 0 f ðz; tÞφðzÞdz, f(z, t) is On the other hand, there were difference between the results of
the wind forces acting on the building at the height of z at time t. model tests and full‐scale measurements. One major reason of the
8 of 10 HAORAN ET AL.

FIGURE 8 Simulated wind parameters of the terrain C

FIGURE 9 The models in the wind tunnel test


HAORAN ET AL. 9 of 10

FIGURE 10 Comparison between the full‐scale measurements and wind tunnel test result

discrepancy can be attributed to the large uncertainties (in the order of dispersion and do not appear to have an obvious nonlinear
10%, as mentioned previously) of identified damping ratios. Besides, relationship with vibration amplitude. The identified average
the reference wind speed and wind direction atop the building would damping ratios through FBFFT along x and y directions were
have a significant effect on the model test results. Whereas the model 0.73% and 0.70%, respectively.
reference wind speed and direction in the test were measured directly 3. The measured acceleration responses atop the building were
in front of the top of the building, the prototype counterparts were cal- below the acceleration comfort criterion.
culated using the data from the weather stations, bringing the result
4. The field measurements of acceleration data were roughly consis-
less accurate. Also, it is worthy to note the scale effect in the study.
tent with the wind tunnel study prediction. In fact, the consis-
As a large reduced scale model was employed in the test, some small
tency is acceptable for engineering applications, thus verifying
components on the building surface useful for the wind‐resistant
the accuracy of the model test results and illustrating that the
design may be omitted, which would also enlarge the difference
wind tunnel test can satisfactorily predict the wind‐induced
between results of full‐scale measurement and test.
vibrations of the super‐tall building under typhoon conditions.
In view of the proceeding factors, the agreement between the
wind tunnel model prediction and the field measured values observed
is acceptable, thus illustrating that the wind tunnel test can provide
RE FE RE NC ES
satisfactory predictions of wind‐induced response of the super‐tall
[1] T. Ohkuma, H. Marukawa, Y. Niihori, N. Kato, J. Wind Eng. Ind. Aerod.
building under typhoon conditions. 1991, 38(2), 185.
[2] B. R. Ellis, J. wind Eng. Ind. Aerod. 1996, 59(2), 365.
[3] S. Campbell, K. C. S. Kwok, P. A. Hitchcock, J. Wind Eng. Ind. Aerod.
6 | C O N CL U S I O N 2005, 93(6), 461.
[4] T. Kijewski‐Correa, J. Kilpatrick, A. Kareem, D. K. Kwon, R. Bashor, M.
The paper presents selected results from full‐scale measurements of
Kochly, B. S. Young, A. Abdelrazaq, J. Galsworthy, N. Isyumov, D.
wind‐induced effect on ZCC during four typhoon events. Structural Morrish, J. Structural Eng. 2006, 132(10), 1509.
dynamic characteristics of this building under strong wind were [5] K. V. Yuen, L. S. Katafygiotis, Adv. Struct. Eng. 2003, 30(2), 81.
analyzed and discussed in this paper. Meanwhile, wind tunnel experi-
[6] S. K. Au, J. Eng. Mech. 2011, 137(3), 214.
ment for ZCC was conducted through force balance model test to
[7] S. K. Au, Mech. Syst. Signal Process. 2012, 26(1), 60.
investigate wind effects on the building. Some conclusions from this
study are summarized as follows: [8] S. K. Au, Mech. Syst. Signal Process. 2012, 26(1), 76.
[9] A. Xu, Z. N. Xie, M. Gu, J. R. Wu, Wind and Structures 2015, 21(2), 159.
1. The uncertainties of identified natural frequencies from field [10] Q. S. Li, J. R. Wu, S. G. Liang, Y. Q. Xiao, C. K. Wong, Eng. Structures the
measured data were small, and their values under different J. Earthq. Wind & Ocean Eng. 2004, 26(12), 1779.

conditions were very close. The measured natural frequencies of [11] J. Y. Fu, J. R. Wu, A. Xu, Q. S. Li, Y. Q. Xiao, Eng. Struct. 2012, 35(1), 120.
the two fundamental translational modes and the fundamental [12] S. K. Au, Mech. Systems & Signal Process. 2014, 48(1–2), 15.
torsional mode were compared with those determined from the [13] Y. Tamura, A. Yoshida, American Society of Civil Engineers 2008, 1.
computational models of ZCC. There were about 6.14–33.43%
[14] J. Yi, J. W. Zhang, Q. S. Li, J. Wind Eng. Ind. Aerod. 2013, 121(5), 116.
differences between the measured and calculated natural
[15] J186–2010. Technical specification for concrete structures of tall build-
frequencies.
ing, China Architecture & Building Press, Beijing 2010 [in Chinese].
2. The damping ratios were identified with large uncertainties. Both [16] GB50009–2012. Load code for the design of building structures, China
the identified damping through FBFFT and RDT show a great Architecture & Building Press, Beijing 2012 [in Chinese].
10 of 10 HAORAN ET AL.

Haoran Pan is currently a PhD student at Tokyo University. He Li Zhang is an engineer of Meteorological Bureau of Shenzhen
received BS and MS from South China University of Technology. Municipality, Shenzhen, China. She received BS and MS from
His research interests include structural health monitoring and Nanjing University of Information Science & Technology. Her
structural dynamic. research interests include meteorologic observation and wind
climate.
Zhuangning Xie is a professor of South China University of
Technology, Guangzhou, China. He received BS and MS from Xian
Jiaotong University and Ph.D from Tongji University. His research
How to cite this article: Pan H, Xie Z, Xu A, Zhang L. Wind
interests include engineering random vibration, wind engineering,
effects on Shenzhen Zhuoyue Century Center: Field measure-
and so on.
ment and wind tunnel test. Struct Design Tall Spec Build. 2017;
An Xu is an assistant research professor of Guangzhou University, e1376. https://doi.org/10.1002/tal.1376
Guangzhou, China. He received MS from Shantou Universuin and
Ph.D from South China University of Technology. His research
interests include structural engineering and wind engineering.

You might also like