You are on page 1of 19

Buffeting Analysis of Long-Span Bridges under Typhoon

Winds with Time-Varying Spectra and Coherences


Tianyou Tao, A.M.ASCE 1; You-Lin Xu, F.ASCE 2; Zifeng Huang 3;
Sheng Zhan 4; and Hao Wang, M.ASCE 5
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Typhoon winds near the external eye wall of a strong typhoon are extremely nonstationary and disastrous due to its vortex and
convective origin as well as its high wind speeds. However, typhoon-induced nonstationary responses of long-span bridges are currently
analyzed with the assumption of time-invariant wind coherence. By taking the Stonecutters cable-stayed bridge in Hong Kong under Typhoon
Hato in 2017 as an example, this paper presents the time-varying wind spectra and coherences for the bridge based on the wind data recorded
by the multiple anemometers installed on the bridge. The analytical framework for nonstationary buffeting analysis of long-span bridges
under typhoon winds is enhanced by considering time-varying wind coherence and applied to the Stonecutters Bridge under Typhoon Hato.
The computed responses are compared with the field measurement responses. The comparative results demonstrate the feasibility and
accuracy of the enhanced framework for buffeting analysis of long-span bridges with time-varying wind spectra and coherences. Comparative
results with the other two cases of time-invariant coherence functions also highlight the importance and necessity of considering the
time-varying wind coherence in the nonstationary buffeting analysis of long-span bridges under typhoon winds. DOI: 10.1061/(ASCE)
ST.1943-541X.0002835. © 2020 American Society of Civil Engineers.
Author keywords: Long-span bridges; Typhoon winds; Non-stationary buffeting analysis; Time-varying wind spectra; Time-varying wind
coherences.

Introduction vibrations of long-span cable-supported bridges become a critical


concern to engineering communities (Professional Standard PRC
A typhoon, also known as a hurricane or cyclone, is one of the most 2018).
devastative natural hazards in the world and causes catastrophic Many studies have been conducted in the past decades to ex-
damage to human beings with extensive economic losses and casu- plore typhoon wind characteristics and their effects on long-span
alties every year (Kareem 1986; Li et al. 2015; Nakamura et al. bridges (Powell et al. 2003; Wang et al. 2016, 2018; Xu et al.
2016; Tao et al. 2017; Yang et al. 2018). As a vital element in a 2000b). Various models were proposed to characterize typhoon
transportation system, long-span cable-supported bridges, because winds, including but not limited to turbulence intensity, gust factor,
of their superior spanning capability, work as an indispensable turbulence integral scale, power spectral density (PSD), and coher-
medium to cross wide rivers, ocean straits, and deep valleys (Ge ence function (Cao et al. 2009; Choi 1983; He et al. 2013; Hui et al.
and Xiang 2011; Gimsing 1983; Tao et al. 2018). Long-span cable- 2009a, b; Li et al. 2015; Solari and Piccardo 2001; Xu et al. 2000b).
supported bridges are sensitive to wind actions due to its low fre- Some of these models have been accepted by codes and used for
quency and low damping (Hao and Wu 2018; Nagai et al. 2004; design purposes (ASCE 2010; National Research Council 2005;
Wang et al. 2011, 2013; Xu 2013). Therefore, typhoon-induced Professional Standard PRC 2018).
By assuming that wind excitation and structural response are sta-
1
Research Fellow, Dept. of Civil and Environmental Engineering, tionary processes, a number of methods governed by the Davenport
Hong Kong Polytechnic Univ., Hong Kong, China; Assistant Professor, wind loading chain (Davenport 1961b; Isyumov 2012) were devel-
School of Civil Engineering, Southeast Univ., Nanjing 211189, China. oped for buffeting analysis of long-span bridges in the past decades
ORCID: https://orcid.org/0000-0002-0922-8736. Email: tytao@seu.edu.cn (Chen et al. 2000; Chen and Kareem 2002; Davenport 1962; Jain
2
Chair Professor, Dept. of Civil and Environmental Engineering, et al. 1996; Scanlan 1978; Xu et al. 2000a, 2019). The assumption
Hong Kong Polytechnic Univ., Hong Kong, China (corresponding author). used in these methods implies that the mean wind speed is constant
ORCID: https://orcid.org/0000-0002-1460-082X. Email: ceylxu@polyu
and the PSD and coherence of wind turbulence are frequency de-
.edu.hk
3
Postdoctoral Fellow, Dept. of Civil and Environmental Engineering, pendent (Chen 2008; Tao et al. 2016, 2017; Xu and Chen 2004).
Hong Kong Polytechnic Univ., Hong Kong, China. Email: zifengh@tongji These methods work well for long-span bridges under synoptic
.edu.cn winds or typhoon winds far away from typhoon eye walls. However,
4
Research Fellow, Dept. of Civil and Environmental Engineering, these methods may not be acceptable for long-span bridges under
Hong Kong Polytechnic Univ., Hong Kong, China. Email: zhan.sheng@ downbursts, thunderstorms, and typhoon winds near typhoon eye
polyu.edu.hk walls because these winds exhibit strong nonstationary features
5
Professor, School of Civil Engineering, Southeast Univ., Nanjing (McCullough et al. 2014; Wang et al. 2016; Xu and Chen 2004).
211189, China. Email: wanghao1980@seu.edu.cn
Hence, the transition from stationary analysis to nonstationary
Note. This manuscript was submitted on December 25, 2019; approved
on June 16, 2020; published online on September 16, 2020. Discussion analysis was advocated by Kareem (2008) and Xu (2013). In a
period open until February 16, 2021; separate discussions must be sub- nonstationary perspective, a wind speed is treated as the superpo-
mitted for individual papers. This paper is part of the Journal of Structural sition of a time-varying mean wind speed plus three zero-mean
Engineering, © ASCE, ISSN 0733-9445. nonstationary fluctuating components (Chen and Letchford 2005;

© ASCE 04020255-1 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Wang and Kareem 2004; Tao et al. 2017; Xu and Chen 2004). are extended from their stationary counterparts, are proposed and
The time-varying mean speed could be extracted from the mea- used to fit the measured ones. The analytical framework for non-
sured wind speed using the empirical mode decomposition, wavelet stationary buffeting analysis of long-span bridges proposed by
transform, or moving-average method (Hu et al. 2013; Su et al. the second author and his colleagues is then extended to include
2015; Tao et al. 2017; Xu and Chen 2004). To characterize non- time-varying complex coherence functions and applied to the
stationary turbulent winds, Priestley (1965) introduced the concept Stonecutters cable-stayed bridge under Typhoon Hato. The pre-
of evolutionary power spectral density (EPSD) instead of the tradi- dicted nonstationary buffeting responses are finally compared with
tional PSD. The spectral features of nonstationary turbulent winds the measured responses as well as the two other cases that consider
are then expressed in a time-frequency domain rather than the fre- time-invariant coherence functions to examine the feasibility and
quency domain. There are some empirical models developed to de- accuracy of the enhanced framework.
scribe the EPSD of typhoon winds (Huang et al. 2015; Tao and
Wang 2019). For the coherence of turbulent winds, however, it
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

is described by a function of frequency only in the Priestley’s evolu- Typhoon Hato and Stonecutters Bridge
tionary spectral theory (Priestley and Tong 1973). Chen (2015) also
proposed a general frequency domain framework for predicting Typhoon Hato emerged as a tropical depression over the southeast
multimode-coupled buffeting response of long-span bridges to non- of Taiwan on August 19, 2017. It then moved to the north of the
stationary winds by introducing a frequency-dependent linear time- South China Sea as a typhoon on August 22. With a rapid intensi-
variant system. The time-independent coherence function is indeed fication on August 23, Hato became a strong typhoon and moved
widely used in the current nonstationary buffeting analysis of long- toward the South China coast. Hato finally made landfall in Zhuhai
span bridges (Chen 2008, 2015; Hu et al. 2013, 2017; Kareem et al. and dissipated over Guangxi Province, China, on August 24, 2017.
2019; Solari et al. 2015). However, this assumption may not re- The path of Typhoon Hato is shown in Fig. 1. The shortest dis-
present the real coherence of nonstationary typhoon winds. Whether tance between the center of Hato and the Stonecutters Bridge is ap-
this assumption will affect the accuracy of the prediction of nonsta- proximately 65 km. It is one of the strongest typhoons passing by
tionary buffeting responses of long-span bridges is unknown. Peng Hong Kong in the past 50 years.
et al. (2018) proposed a time-varying coherence model for down- The Stonecutters Bridge in Hong Kong, as shown in Fig. 2, has a
burst winds and applied it in the nonstationary response analysis of total length of 1,596 m and a main span of 1,018 m. The girder of
tall buildings. Nevertheless, the research on the time-varying coher- the bridge in the main span is made of steel while the girder in the
ence of typhoon winds and its effect on the buffeting responses of two side spans is in concrete but with the transition of 49.74-m steel
long-span bridges is still lacking. girder from the tower to the concrete girder. The girder is the twin-
In this regard, this paper presents a comprehensive study on the box girder supported by stay cables every 18 m in the main span
buffeting analysis of long-span bridges under typhoon winds with and supported by stay cables and piers in the side spans. The tower
time-varying wind spectra and coherences. The nonstationary spec- is of a single column with a reinforced concrete structure from the
tral features of typhoon winds recorded by the multiple anemom- base level to level þ175 m and then a composite steel and concrete
eters installed on the Stonecutters cable-stayed bridge during structure. The height of the two towers is nearly 300 m. The towers
Typhoon Hato are first investigated to obtain the measured time- are founded on piled foundations.
varying spectra and coherences of typhoon winds. The analytical A structural health monitoring system, including twelve
expressions of time-varying wind spectra and coherences, which Gill R3-50 triaxial ultrasonic anemometers, was installed in the

Fig. 1. Path of Typhoon Hato in 2017. (Map data © 2020 Google, Imagery © 2020 TerraMetrics.)

© ASCE 04020255-2 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


+298.0 +298.0

NAVIGATION CHANNEL: 900m

69.25m 70m 70m 79.75m 1018m 79.75m 70m 70m 69.25m


Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

Anemometer
1 2 3 4 5

Fig. 2. Stonecutters Bridge and layout of anemometers.

Stonecutters Bridge (Hui and Wong 2009). The ten anemometers Decomposition of Skew Winds
were installed on the bridge deck via the cantilever booms and the
As the wind struck Stonecutters Bridge with a yaw angle, the wind is
two anemometers were installed at the top of the west and east
decomposed into three components, which are the longitudinal
tower, respectively, as shown in Fig. 2. The distance between the
component UðtÞ perpendicular to the bridge span in the horizontal
anemometer and the edge of the bridge deck via the cantilever
plane, the lateral component VðtÞ along the bridge alignment, and
boom is 7 m. The height of the anemometers installed in the bridge
the vertical component WðtÞ perpendicular to the bridge span in the
deck is 73.5 m above the sea level. The anemometers with No. 1 to
vertical plane. Since the yaw angle is small, only the longitudinal
No. 5 are on the northeast side and the others are on the southwest
and vertical components are considered for the buffeting analysis of
side. Except that the distance between No. 2 and No. 3 anemom-
Stonecutters Bridge while the lateral component is ignored (Scanlan
eters is 21.5 m, the distance between other two adjacent anemom-
1993). Fig. 3 shows a typical example of the decomposed longi-
eters is 18 m. This layout ensures that these anemometers are able
tudinal and vertical wind velocities at the middle anemometer. Con-
to capture wind spatial coherences. The X-, Y-, and Z-axes of each
sidering nonstationary typhoon winds, both the longitudinal and
anemometer face the north, west, and upward direction, respec-
vertical wind components are separated into the time-varying mean
tively (Fig. 1). The measurement range of wind speed of the ane-
wind speed and the nonstationary turbulence based on Eqs. (1)
mometers spans from 0.01 m=s to 50 m=s and the sampling
and (2). The time-varying mean wind speed is determined via the
frequency is 50 Hz. With the 0-degree pointing north for wind di-
wavelet-based approach by Tao et al. (2017), and the turbulence can
rection, wind direction increases counterclockwise as viewed from
be obtained by subtracting the time-varying mean wind speed from
the top (Fig. 1). The anemometers in the structural health monitor-
the longitudinal or vertical component
ing system successfully recorded wind data and wind-induced
structural response data of the bridge during Typhoon Hato.
~
UðtÞ ¼ UðtÞ þ uðtÞ ð1Þ

Nonstationary Characteristics of
~
WðtÞ ¼ WðtÞ þ wðtÞ ð2Þ
Measured Typhoon Winds
~
where UðtÞ ~
and WðtÞ = longitudinal and vertical time-varying mean
Typhoon Wind Data
wind speeds, respectively; and uðtÞ and wðtÞ = nonstationary turbu-
The hourly typhoon wind data recorded by the anemometers from lences in the longitudinal and vertical components.
9:40 a.m. to 10:40 a.m. on August 23, 2017 are used in this study Because the time-varying mean wind speeds recorded by the
because during this period, Typhoon Hato is most close to the five anemometers are almost the same in each direction, an average
bridge and wind directions are almost perpendicular to the bridge of the five cases is employed to represent the longitudinal and ver-
alignment. Since the wind direction is from the northeast to the tical time-varying mean wind speeds for Typhoon Hato during the
southwest during this period, the wind data recorded by the ane- time period concerned. To facilitate the nonstationary buffeting
mometers (No. 1 to No. 5) installed on the northeast side of the analysis of Stonecutters Bridge, the integrated time-varying mean
bridge deck are selected as natural wind data for investigating wind speed and corresponding attack angle are calculated based on
time-varying spectra and coherences of typhoon winds. Based on Eqs. (3) and (4). The results used for the subsequent time-varying
the hourly wind samples of typhoon winds recorded by the middle static response analysis are presented in Fig. 4
anemometer (No. 3 in Fig. 2), the hourly mean wind direction is
approximately calculated as 124°. As shown in Fig. 1, the bridge qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
alignment is tilted 42° relative to due north, and therefore the wind U~ m ðtÞ ¼ U~ 2 ðtÞ þ W ~ 2 ðtÞ ð3Þ
struck Stonecutters Bridge with a small yaw angle about 8°. This
facilitates a comparison of the subsequent computed buffeting re- ~
WðtÞ
sponses with the measured responses of the bridge under small skew αðtÞ ¼ arctan ð4Þ
winds. ~
UðtÞ

© ASCE 04020255-3 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


36

24

U (m/s)
12

0
0 600 1200 1800 2400 3000 3600
Time (s)
16

8
W (m/s)

0
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

-16
0 600 1200 1800 2400 3000 3600
Time (s)

Fig. 3. Decomposed longitudinal and vertical wind speeds.

20
Um (m/s)

17
~

14
0 600 1200 1800 2400 3000 3600
Time (s)

1.1

0.8

0.5
0 600 1200 1800 2400 3000 3600
Time (s)

Fig. 4. Time-varying mean wind speed and corresponding attack angle.

Time-Varying Power Spectra time-varying features of typhoon winds could not be completely
revealed. It can also be seen from Fig. 5(a) that the maximum en-
Based on the nonstationary turbulence records from the five ane-
ergies in the longitudinal and vertical fluctuating wind speeds do not
mometers, the longitudinal and vertical EPSDs are estimated for
emerge simultaneously. The most prominent actions of the longitu-
each record using the Priestley’s method (Priestley 1965). In the
dinal and vertical turbulences are thus not synchronous in the buf-
EPSD estimation, the window length of the general filter is taken
feting responses of long-span bridges.
as 540 s to guarantee the lower limit of frequency, and the window
Since the shape of EPSD of the measured winds at each time
length for the general weight function is selected as 1,080 s. The
instant is similar to that of a stationary PSD, it could be thus de-
details on the EPSD estimation method can be found in Appendix I. scribed using an extended stationary model (Hu et al. 2017; Tao and
Due to the high similarity in the estimated EPSDs of the five re- Wang 2019). By extension from the stationary PSD of turbulence
cords, the five EPSDs in both the longitudinal and vertical direc- following the Kolmogrov hypothesis (Li et al. 2012), the EPSD
tions are averaged to establish an empirical EPSD model for the model of nonstationary longitudinal or vertical turbulence can be
bridge under Typhoon Hato. expressed as
The averaged measured longitudinal and vertical EPSDs are
shown in Fig. 5(a). In Fig. 5(a), Su ðω; tÞ and Sw ðω; tÞ, in which ω nSðn; tÞ AðtÞfðtÞ
¼ ð5Þ
represents the circular frequency and t denotes the time, are the u2 ðtÞ 5
½1 þ BðtÞfðtÞυðtÞ 3υðtÞ
EPSDs of longitudinal and vertical turbulence, respectively. It can
be seen from Fig. 5(a) that the time-varying features in the measured where Sðn; tÞ = EPSD; n = frequency of turbulence; u ðtÞ ¼
EPSDs appear clearly within a time duration of 3,600 s, demonstrat- 0.4U~ m ðtÞ= lnðz=z0 Þ = time-varying friction wind speed, in which
ing the nonstationarity of typhoon winds. These evolutionary spec- z denotes the height and equals to 73.5m in this study; z0 = rough-
tra also indicate that if a short duration, such as 600 s, is used, the ness length and taken as 0.01 for the sea terrain (Professional

© ASCE 04020255-4 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Sw ( , t )
Su ( , t )
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

Time (s) Time (s)


(a)
Su ( , t )

Sw ( , t )

Time (s) Time (s)


(b)

Fig. 5. Comparison of measured and fitted EPSDs of nonstationary turbulence: (a) measured results; and (b) fitted models.

Standard PRC 2018); fðtÞ ¼ nz=U~ m ðtÞ = time-dependent Monin assumption, such as the Sigma-oscillatory process (Battaglia 1979;
coordinate; and AðtÞ, BðtÞ, and υðtÞ = undetermined parameters Conte and Peng 1997). Therefore, a critical step that should be
dependent on the time. taken is to examine whether the coherence of typhoon winds com-
The longitudinal and vertical EPSD models expressed by Eq. (5) plies by the time-invariant and phase-ignored hypothesis. The co-
are used to fit the measured EPSDs of Typhoon Hato. The detailed herence of the nonstationary fluctuating wind speed in this study is
fitting procedure can be found in Tao and Wang (2019). The fitted estimated using Eq. (29) in Appendix I as a complex function in-
EPSDs are shown in Fig. 5(b) while the fitted time-dependent stead of using the modulus of the coherence function (Davenport
parameters are presented in Fig. 6. The comparison between the 1961a; Solari and Piccardo 2001; Tobin and Chamorro 2018). The
measured and fitted EPSDs shown in Fig. 5 indicates that the EPSD cross-EPSDs (CEPSD) of nonstationary fluctuating wind speeds
models expressed by Eq. (5) can well describe the measured EPSDs recorded by the five anemometers are estimated and the coherences
in either longitudinal or vertical direction. between the turbulent winds measured at any two of the five meas-
From Fig. 6, it can be seen that all the fitted parameters contin- urement points are calculated. Typical measured coherence func-
uously change with time, in which the parameter AðtÞ represents tions are shown in Fig. 7, where the two subscript numbers in the
the envelope appearance of the EPSD. The fitted parameters for coherence function denote the numbers (locations) of the two ane-
the longitudinal turbulence are different from those for the vertical mometers shown in Fig. 2. To unify the description of frequency
turbulence. The empirical EPSD models expressed by Eq. (5) and between coherence and spectrum, the natural frequency n (n ¼
fitted using the measured data can be used to conduct the sub- ω=2π) instead of the circular frequency ω is used in the frequency
sequent nonstationary buffeting analysis of the bridge. coordinate.
As illustrated in Fig. 7, the typhoon wind coherence varies with
time significantly in both real and imaginary parts. Although the
Time-Varying Coherence Function real part of typhoon wind coherence changes moderately, promi-
In addition to the EPSD function, the turbulence coherence func- nent peaks are observed in the imaginary part of typhoon wind
tion is another important consideration in the buffeting analysis of coherence within a certain frequency range. The magnitude of
long-span bridges. The coherence function characterizes the corre- the imaginary part can reach 0.5, which is compatible with the
lation of turbulence at different locations. magnitude of the real part. The phenomenon observed above dem-
In the current nonstationary buffeting analysis framework, the onstrates significant time-varying features of typhoon wind coher-
coherence function of turbulence is often treated as a real time- ence. Therefore, the time-invariant and phase-ignored hypothesis is
invariant function with the phase difference ignored. In reality, not adequate to describe these typhoon winds. More detail discus-
not all nonstationary processes follow the time-invariant coherence sions on time-varying typhoon wind coherences can be found in

© ASCE 04020255-5 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


700
Longitudinal
Vertical

A(t)
350

0
0 1200 2400 3600
Time (s)

100
Longitudinal
Vertical
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

B(t)

50

0
0 1200 2400 3600
Time (s)

3
Longitudinal
Vertical
2
(t)

0
0 1200 2400 3600
Time (s)

Fig. 6. Fitted parameters in EPSD models for longitudinal and vertical turbulence.

12 ( n, t ) 14 ( n, t ) 12 ( n, t ) 14 ( n, t )

1.0 1.0 1.0 1.0


0.7 0.7 0.7 0.7
Real part

Real part
Real part

Real part

0.4 0.4 0.4 0.4


0.1 0.1 0.1 0.1
-0.2 -0.2 -0.2 -0.2
10-2 3600 10-2 3600 10-2 3600 10-2 3600
n (Hz) 10-1 2400 n (Hz) 10-1 2400 n (Hz) 10-1 2400 n (Hz) 10
-1 2400
100 1200 100 1200 Time (s) 100 1200 Time (s) 100 1200 Time (s)
0 Time (s) 0 0 0

0.5 0.6 0.5 0.6


Imaginary part

Imaginary part
Imaginary part

Imaginary part

0.4
0.3 0.3 0.3
0.2
0.1 0 0.1 0
-0.2
-0.1 -0.4 -0.1 -0.3
10-2 10-2 3600 10-2 3600 10-2 3600
3600 -1 2400 -1 2400
n (Hz) 10
-1
n (Hz) 10
-1 2400 n (Hz) 10
2400 n (Hz) 10 1200 1200 Time (s)
100 1200 Time (s) 100 1200 Time (s) 100 0 Time (s) 100 0
0 0

(a) (b)

Fig. 7. Measured coherence functions of Typhoon Hato: (a) longitudinal; and (b) vertical.

 
Huang et al. (2020). Hence, the time-varying coherence with phase nr
difference should be taken into account to conduct a more accurate γ jk ðnÞ ¼ exp −c þ iθ ð6Þ

nonstationary buffeting analysis of long-span bridges.
In the nonstationary buffeting analysis of long-span bridges, a where c = nondimensional attenuation parameter; r = separation
time-varying coherence model is essential to consider the spatial distance between point j and point k along p theffiffiffiffiffiffibridge deck; Ū =
correlation of turbulent winds at any two locations of the bridge mean wind speed over a given period; i ¼ −1; and θ = phase
deck. The Davenport coherence model taken as an exponential angle to separate the coherence into the two parts: co-coherence
form (Davenport 1961a; CEN 1994) is widely used in practice (real part) and quadrature coherence (imaginary part). However, the

© ASCE 04020255-6 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


quadrature coherence is often taken as zero for simplicity. Clearly, Eq. (9) is now used to fit the measured time-varying coherence
the Davenport coherence model is a time-invariant coherence of Typhoon Hato at each instant. The fitted coherence functions are
model. shown in Fig. 8 for both longitudinal and vertical turbulences. The
Although Eq. (6) is easy to implement, this model gives a full corresponding time-dependent parameters are presented in Fig. 9.
correlation of low-frequency wind components irrelevant to dis- It can be seen from Fig. 9 that the variation trend of dðtÞ or CðtÞ is
tance (Krenk 1996; Hansen and Krenk 1999). In fact, the turbu- similar for both the longitudinal and vertical components, but the
lence integral scale of typhoon winds cannot cover the entire trend of the scale length parameter LðtÞ is not similar between the
bridge span over 1,000 m and the correlation of low-frequency longitudinal and vertical turbulence components.
wind components (large-scale eddies) decreases as the separation The comparison between Figs. 7 and 8 can verify the effective-
distance increases. Therefore, Krenk (1996) proposed a coherence ness of Eq. (9) in characterizing the time-varying features of ty-
model below phoon wind coherence. A detailed slice comparison between the
    measured and fitted coherence functions is shown in Fig. 10. It
1n r n r 2πnr
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

γ jk ðnÞ ¼ 1 − x C exp − x C þ i d ð7Þ can be seen from Fig. 10 that the fitted coherence function is close
2 Ū Ū Ū to the measured coherence for the longitudinal turbulence compo-
nent in general, which proves the suitability of the extended Krenk
where
model. There are some discrepancies at the low-frequency region
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  between the fitted and measured coherences of vertical turbulence
Ū 2
nx ¼ n2 þ ð8Þ component. Fortunately, the vertical turbulence coherence below
2πL 0.01 Hz contributes little to the buffeting responses as the first ver-
tical natural frequency of a long-span bridge is usually much larger
d, L, and C = parameters to be fitted. than 0.01 Hz.
The Krenk coherence model is also a time-invariant model. To To assess the effect of time-varying coherence on the prediction
represent the time-varying features of coherence observed in Fig. 7, of nonstationary buffeting responses, three cases are considered:
the Krenk model is extended in this study into a time frequency– time-varying complex coherence function (Case 1); time-invariant
varying model. The suitability of this extension is supported by the coherence function with phase difference included (Case 2); and
similarity between the time-varying coherence at each time instant time-invariant coherence function with phase difference ignored
and its stationary counterpart (Case 3). For Case 1, the time-varying complex coherence function
    expressed by Eq. (9) is used. For Case 2, the measured coherence is
1 nx ðtÞr nx ðtÞr 2πnr
γ jk ðn; tÞ ¼ 1 − CðtÞ exp − CðtÞ þ i dðtÞ first averaged over time and then fitted by the time-invariant model
2 U~ m ðtÞ U~ m ðtÞ U~ m ðtÞ
expressed by Eq. (7). The fitted parameters are listed in Table 1. For
ð9Þ Case 3, the Davenport coherence model without phase angle is
used. The fitted parameters for the longitudinal and vertical turbu-
where lent winds are 7.34 and 4.89, respectively.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
2 U m ðtÞ 2
nx ðtÞ ¼ n þ ð10Þ
2πLðtÞ Theoretical Framework for Nonstationary Buffeting
Analysis
Different from Eq. (7), dðtÞ, LðtÞ, and CðtÞ in Eq. (9) are the
time-dependent parameters. The time-varying mean wind speed The nonstationary buffeting analysis framework has been estab-
is also introduced in Eq. (9). The time-varying features of typhoon lished by Chen (2015), Hu et al. (2013), Kareem et al. (2019),
wind coherence are represented by the four parameters and the and Xu (2013). Similar to the nonstationary wind model given
phase difference is included in terms of the parameter dðtÞ. by Eqs. (1) and (2), the associated total wind force acting on the

12 ( n, t ) 14 ( n, t ) 12 ( n, t ) 14 ( n, t )

1.0 1.0 1.0 1.0


0.7 0.7 0.7 0.7
Real part

Real part

Real part

Real part

0.4 0.4 0.4 0.4


0.1 0.1 0.1 0.1
-0.2 -0.2 -0.2 -0.2
10-2 3600 10-2 3600 10-2 3600 10-2 3600
n (Hz) 10-1 2400 n (Hz) 10-1 2400 n (Hz) 10-1 2400 n (Hz) 10
-1 2400
100 1200 Time (s) 100 1200 Time (s) 100 1200 Time (s) 100 1200 Time (s)
0 0 0 0

0.4 0.5 0.5


Imaginary part

Imaginary part

Imaginary part

Imaginary part

0.4
0.2 0.3 0.3
0.2
0 0 0.1 0.1

-0.2 -0.2 -0.1 -0.1


10-2 3600 10-2 3600 10-2 3600 10-2 3600
-1 2400 -1 2400 -1 2400 -1 2400
n (Hz) 10 1200 Time (s) n (Hz) 10 1200 Time (s) n (Hz) 10 1200 Time (s) n (Hz) 10 1200 Time (s)
100 0 100 0 100 0 100 0
(a) (b)

Fig. 8. Fitted time-varying coherence functions of Typhoon Hato: (a) longitudinal; and (b) vertical.

© ASCE 04020255-7 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


1.2
Longitudinal
Vertical
0.6

d (t)
0.0

-0.6
0 1 200 24 00 3 600
Time (s)

900
Longitudinal
Vertical
600
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

L (t)

300

0
0 1 200 24 00 3 600
Time (s)

5
Longitudinal
Vertical
4
C (t)

2
0 1 200 24 00 3 600
Time (s)

Fig. 9. Fitted time-varying parameters for longitudinal and vertical turbulences.

1.0 1.0
Measured Measured
Fitted Fitted
0.8 0.8
Real[ 12(n, t)] Real[ 12(n, t)]
Real[ 14(n, t)]
0.6 0.6 Real[ 14(n, t)]
Coherence

Coherence

0.4 Imag[ 12(n, t)] 0.4


Imag[ 12(n, t)]
Imag[ 14(n, t)]
Imag[ 14(n, t)]
0.2 0.2

0.0 0.0
t = 3000s t = 2000s
-0.2 -0.2
10-2 10-1 100 10-2 10-1 100
(a) n (Hz) (b) n (Hz)

Fig. 10. Slice comparisons of measured and fitted time-varying coherence functions: (a) longitudinal; and (b) vertical.

Table 1. Fitted parameters for the time-invariant coherence functions in ~


XðtÞ ¼ XðtÞ þ xðtÞ ð12Þ
Case 2
~
where FðtÞ = force vector; FðtÞ = time-varying mean wind force
Parameter d L C
vector; fðtÞ = fluctuating aerodynamic force vector, which is the
Longitudinal 0.40 691.26 4.49 sum of buffeting force vector and self-excited force vector; and
Vertical 0.48 79.11 2.56
XðtÞ = structural displacement vector, which contains the time-
varying mean displacement vector XðtÞ~ and the zero-mean nonsta-
tionary buffeting displacement vector xðtÞ. The expressions for the
bridge in the global structural coordinate can be divided into the time-varying mean wind force, nonstationary buffeting force, and
time-varying mean wind force and nonstationary fluctuating aero- self-excited force can be found in Appendix II. The relationship
dynamic force, which lead to the time-varying mean response and between the lift force, drag force, and pitching moment provided
nonstationary buffeting response, respectively in Appendix II and the time-varying mean wind force vector, non-
stationary buffeting force vector, and self-excited force vector can
~ þ fðtÞ
FðtÞ ¼ FðtÞ ð11Þ be found in Hu et al. (2013).

© ASCE 04020255-8 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Time-Varying Mean Wind Response The CEPSD matrix of nonstationary longitudinal or vertical tur-
bulence can be given by
Since both the nonstationary buffeting force and self-excited force
depend on the instantaneous mean wind attack angle, a time-varying 2 η 3
mean wind response analysis needs to be first conducted (Chen S11 ðω; tÞ Sη12 ðω; tÞ · · · Sη1M ðω; tÞ
6 η 7
2015; Hu et al. 2013, 2017; Kareem et al. 2019). The time-varying 6 S ðω; tÞ Sη ðω; tÞ · · · Sη ðω; tÞ 7
6 21 22 2M 7
mean wind speed shown in Fig. 5 evolves with time very slowly, and Sηη ðω; tÞ ¼ 6
6 .. .. .. ..
7 η ¼ u; w
7
therefore the transient dynamic effect of the mean wind force can be 6. . . . 7
4 5
ignored. Thus, a static analysis can be performed at each time instant η η η
SM1 ðω; tÞ SM2 ðω; tÞ · · · SMM ðω; tÞ
to determine the time-varying mean wind response
~ ~ ð18Þ
KXðtÞ ¼ FðtÞ ð13Þ

where K = stiffness matrix of the bridge. where M = number of elements by which the bridge deck is
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

discretized.
It is well known that the CEPSD matrix Sηη ðω; tÞ is Hermitian
Nonstationary Buffeting Response and nonnegatively definite. Any element in this matrix can be ex-
pressed as
The governing equation for nonstationary buffeting analysis of a
long-span bridge under typhoon winds can be expressed as qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Sηjk ðω; tÞ ¼ Sηjj ðω; tÞSηkk ðω; tÞγ ηjk ðω; tÞ j; k ¼ 1; 2; : : : ; M ð19Þ
MẌðtÞ þ CẊðtÞ þ KXðtÞ ¼ Fb ðtÞ þ Fse ðtÞ ð14Þ
where M and C = mass and damping matrices of the bridge; Ẍ = where Sηjk ðω; tÞ = CEPSD function; Sηjj ðω; tÞ and Sηkk ðω; tÞ denote
acceleration response vector; Ẋ = velocity response vector; Fb = EPSD of nonstationary turbulence at point j and point k, respec-
buffeting force vector; and Fse = self-excited force vector. tively; and γ ηjk ðω; tÞ = time-varying complex coherence function
Eq. (14) cannot be directly solved in the time domain because the expressed by Eq. (9).
aerodynamic admittance functions, EPSDs, and the coherences of If the EPSD of longitudinal or vertical turbulence is assumed to
buffeting forces, as well as the self-excited forces, are all dependent be invariant along the bridge deck, Eq. (19) can be further simpli-
on both time and frequency. There are two approaches available to fied as
solve Eq. (14). One is the approach proposed by Chen (2015) that
simplifies the solution by using an equivalent frequency-dependent Sηjk ðω; tÞ ¼ Sη ðω; tÞγ ηjk ðω; tÞ j; k ¼ 1; 2; : : : ; M ð20Þ
linear time-variant system. The other is a time-frequency approach
using the pseudoexcitation method (Hu et al. 2013; Lin 2004). The
where Sη ðω; tÞ represents the EPSD of nonstationary longitudinal
latter is employed to solve Eq. (14) in this study. The details on how
or vertical turbulence.
to solve Eq. (14) in the time-frequency domain can be found in Hu
For the cases considering the time-invariant coherence, γ ηjk ðω; tÞ
et al. (2017) but the time-varying coherence should be taken into
consideration. in Eq. (19) can be simply replaced by the time-invariant coherence
For the time-frequency solution, the critical step is to determine function.
the CEPSD matrix SFb Fb ðω; tÞ of the nonstationary buffeting forces.
The nonstationary buffeting force vector can be expressed as
Case Study
Fb ðtÞ ¼ Ts ðtÞχðn; tÞCb ðtÞΘðtÞ ð15Þ

where Ts ðtÞ = transfer matrix to convert the element forces to the Dynamic and Aerodynamic Characteristics of
nodal forces in the global coordinate, and this matrix is time- Stonecutters Bridge
dependent as it relates to the instantaneous mean wind attack angle; The Stonecutters Bridge in Hong Kong under Typhoon Hato is
χðn; tÞ = time-varying aerodynamic admittance function matrix; taken as a case study for the buffeting analysis of long-span bridges
Cb ðtÞ = time-varying aerodynamic force coefficient matrix; and under typhoon winds with time-varying spectra and coherences us-
ΘðtÞ ¼ ½uðtÞT wðtÞT T = vector of nonstationary fluctuating wind ing the enhanced framework. A 3D finite element (FE) model of the
speeds. bridge is established by the software ANSYS. The FE model is then
Since the variations of Ts ðtÞ, χðn; tÞ, and Cb ðtÞ with time are updated using the measured modal properties to make sure the FE
relatively slow (Hu et al. 2013), the CEPSD of Fb ðtÞ depends on model can best represent the real bridge (Hu et al. 2013). The com-
mainly the CEPSD of ΘðtÞ. In such a case, the CEPSD matrix of puted first 40 modes of vibration, including modal frequencies and
the nonstationary buffeting forces can be expressed as mode shapes of the bridge, are extracted and used for the nonsta-
tionary buffeting analysis. Fig. 11 shows the first mode of vibration
SFb Fb ðω; tÞ ¼ Ts ðtÞχðn; tÞCb ðtÞSΘΘ ðω; tÞCTb ðtÞχT ðn; tÞTTs ðtÞ in each direction, and the values in the bracket denote the natural
ð16Þ frequency. The damping ratio for each mode of vibration is taken as
0.5% for simplicity. As the mean wind velocity of typhoon winds is
in which slowly varying, nonstationary aerodynamic forces can be modeled
  with force parameters from smooth flow (Chen 2015). The force
Suu ðω; tÞ 0 parameters of the bridge deck section were thus obtained from
SΘΘ ðω; tÞ ¼ ð17Þ
0 Sww ðω; tÞ the wind tunnel tests of the section models under uniform flow con-
ditions (Hui et al. 2006; Zhu and Xu 2014; Zhu et al. 2016). The
where Suu ðω; tÞ = CEPSD matrix of nonstationary longitudinal tur- steady-state force coefficients are presented in Table 2 for three an-
bulence; and Sww ðω; tÞ = CEPSD matrix of nonstationary vertical gles of attack, and the flutter derivatives for the two motions are
turbulence. shown in Fig. 12.

© ASCE 04020255-9 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Fig. 11. First mode of vibration of the bridge in each direction: (a) lateral (0.161 Hz); (b) vertical (0.212 Hz); and (c) torsional (0.445 Hz).

Table 2. Steady-state force coefficients and their derivatives results are compared with the computed results in terms of EPSDs.
α CL CD CM CL0 CD0 CM0 The measured accelerations at the midspan of the bridge is shown
in Fig. 15, in which aL , aV , and aT denote the lateral, vertical, and
−3° −0.2704 0.0458 −0.0044 4.0027 −0.1386 0.6524 torsional acceleration response, respectively. The measurement data
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

0° −0.1003 0.0426 0.0295 2.8741 −0.0490 0.8643


are synchronous with the wind samples presented in Fig. 3. Since
3° 0.1038 0.0484 0.0703 3.4827 0.2553 1.2044
the cut-off frequency in the computed results is 1.32 Hz, a digital
filter with the same frequency range is applied to the measured data
to ensure a reasonable comparison.
Typhoon-Induced Time-Varying Mean Response Based on the EPSD estimation method in Appendix I, the
EPSDs of the measured acceleration responses shown in Fig. 15
Based on the time-varying mean wind speed and attack angle are estimated and plotted in Fig. 16. By comparing Fig. 16 with
shown in Fig. 4, the time-varying mean wind response of the Stone- Fig. 14, one can find similar time-varying trends. To make a de-
cutters Bridge under Typhoon Hato is calculated using the nonsta- tailed comparison, some typical slices of the measured EPSDs are
tionary static force model. The mean displacement responses at the presented in Fig. 17 in comparison with the computed EPSDs for
midspan of the bridge are shown in Fig. 13. the same time instants. It can be seen from Fig. 17 that the com-
puted acceleration EPSD is in good agreement with the measured
Typhoon-Induced Nonstationary Buffeting Responses EPSD in each of the three directions for different time instants. The
satisfactory comparative results of EPSDs validate the feasibility
The instantaneous mean attack angle is obtained by adding the and accuracy of the enhanced nonstationary buffeting analysis
time-varying mean torsional displacement response at each node of framework.
the bridge deck to the attack angle of typhoon winds shown in Fig. 4. The RMS values of the computed nonstationary acceleration
Then, by following the enhanced framework, a nonstationary buf- responses of the bridge are calculated at each instant according
feting analysis of the bridge is conducted with consideration of the to Eq. (21)
developed time-varying wind spectra and coherence functions. The
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z ffi
EPSD of bridge response is calculated for each discretized location ∞
of the bridge. The EPSDs of the acceleration responses at the mid- σi ðtÞ ¼ Si ðω; tÞdω i ¼ L; V; T ð21Þ
span of the bridge are shown in Fig. 14 for the three directions. It can 0
be seen that several modes of vibration contribute to the acceleration
responses of the bridge, but the lowest mode of vibration or the first where σi ðtÞ = time-varying RMS value of acceleration response in
two modes of vibration contribute most. The magnitude of the the lateral (i ¼ L), vertical (i ¼ V), or torsional (i ¼ T) direction;
EPSD at the dominated mode of vibration changes with the time and Si ðω; tÞ denotes the EPSD of acceleration response. When the
significantly, demonstrating the nonstationary buffeting responses EPSD in Eq. (21) is replaced by the PSD of the stationary response,
under the nonstationary typhoon winds. σi ðtÞ is reduced to a constant value over time.
To verify the accuracy of the nonstationary buffeting responses To illustrate the effects of time-varying wind spectra, the buf-
computed using the enhanced framework, the field measurement feting analysis of the Stonecutters Bridge is also performed using

2 2

1
1 0
Flutter derivatives

Flutter derivatives

-1
0
-2

-3
-1
H1* -4 H2*
H4* H3*
-2 -5
A1* A2*
A4* -6 A3*

-3 -7
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18
B/U B/U

Fig. 12. Flutter derivatives of the bridge deck.

© ASCE 04020255-10 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


-2
2.4 ×10

Displacement (m)
2.0

1.6

1.2
0 600 1200 1800 2400 3000 3600
(a) Time (s)

-0.4
Displacement (m)

-0.8
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

-1.2

-1.6

-2.0
0 600 1200 1800 2400 3000 3600
(b) Time (s)
-4
4.8 ×10
Displacement (rad)

4.2

3.6

3.0

2.4
0 600 1200 1800 2400 3000 3600
(c) Time (s)

Fig. 13. Time-varying mean displacements at the midspan of the bridge: (a) lateral; (b) vertical; and (c) torsional.

Torsional EPSD (rad2/s3)


Vertical EPSD (m2/s3)
Lateral EPSD (m2/s3)

n (Hz) n (Hz) n (Hz)


Time (s) Time (s) Time (s)

Fig. 14. EPSDs of calculated acceleration responses at the midspan of the bridge.

the stationary approach with time-invariant spectra and coherence coherence functions. As mentioned in the previous section, Case 2
functions derived from the measured typhoon winds. The RMS val- considers the time-invariant coherence function with phase differ-
ues of the stationary acceleration responses at the midspan are ob- ence included and Case 3 takes account of the time-invariant co-
tained and compared with the nonstationary results, as shown in herence function with no phase difference. The RMS values of the
Fig. 18. It can be seen that the nonstationary response is larger than acceleration responses are also calculated for Cases 2 and 3 as well
the stationary response at a certain time period, which means that as the measured results according to Eq. (21).
the buffeting responses would be underestimated by the stationary In Case 2, the measured coherence is first averaged over time
approach. This observation is consistent with that reported in Hu and then fitted by the time-invariant model expressed by Eq. (7),
et al. (2013). while Eq. (9) used in Case 1 is actually extended from Eq. (7).
Therefore, to facilitate the comparison among the three cases,
the RMS value is divided into two parts as follows
Effect of Time-Varying Wind Coherence
σi ðtÞ ¼ σ̄i þ σ~ i ðtÞ ð22Þ
To better understand the effect of time-varying wind coherence on
non-stationary buffeting responses, the computed results presented where σ̄i denotes the average value of the time-varying RMS re-
above for Case 1, which considers time-varying complex coher- sponse over the time; and σ~ i ðtÞ represents the variation of the
ence, are compared with the other two cases with time-invariant time-varying RMS response with respect to the average value.

© ASCE 04020255-11 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


0.08
0.04

aL (m/s2)
0
-0.04
-0.08
0 600 1200 1800 2400 3000 3600
Time (s)
0.4

aV (m/s2) 0.2
0
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

-0.2
-0.4
0 600 1200 1800 2400 3000 3600
Time (s)
0.01
aT (rad/s2)

0.005
0
-0.005
-0.01
0 600 1200 1800 2400 3000 3600
Time (s)

Fig. 15. Measured acceleration responses at the midspan of the bridge.

×10-2 ×10-5

Torsional EPSD (rad2/s3)


Lateral EPSD (m2/s3)

Vertical EPSD (m2/s3)

n (Hz) n (Hz) n (Hz)


Time (s) Time (s) Time (s)

Fig. 16. EPSDs of measured acceleration responses at the midspan of the bridge.

The average RMS values of the acceleration responses computed On the other hand, the errors of the computed average RMS
from the three cases are compared with the measured average RMS values of the vertical acceleration responses of the bridge deck
value of the acceleration response, and the comparative results at the midspan are 17.92%, 22.72%, and 32.16% for Cases 1, 2,
for the midspan of the bridge deck are given in Table 3. It can and 3, respectively. The errors of the computed average RMS val-
be seen that the errors of the computed average RMS values of the ues of the torsional acceleration responses of the bridge deck at the
lateral acceleration responses of the bridge deck at the midspan are midspan are 0.00%, 8.31%, and 16.67% for Cases 1, 2, and 3, re-
−28.28%, −26.26%, and −20.20% for Cases 1, 2, and 3, respec- spectively. The computed average RMS values are in general larger
tively. Fig. 19 shows the measured and computed EPSDs of the lat- than the measured ones with the results of Case 1 most close to the
eral acceleration responses of the bridge deck at the midspan at the measured ones. The difference of the average RMS values between
two different time instants. It can be seen from Fig. 19 that the three Case 1 and Case 2 is much smaller than that between Case 1 and
computed EPSDs are very similar to each other with Case 1 having Case 3. Thus, the use of the modulus of coherence function in Case 3
slight large amplitudes. Some peaks in the measured EPSD miss without consideration of phase difference overestimates the correla-
in the computed EPSDs, and the magnitudes of some dominated tion of typhoon wind speeds between two locations. The use of the
frequencies in the measured EPSD are larger than those of the com- time-varying complex coherence function as used in Case 1 is the
puted EPSDs. This is why the computed average RMS values are all most appropriate.
smaller than the measured one, and the error of the computed aver- Fig. 20 shows the comparison of variation part of the measured
age RMS value of Case 3 is the smallest among the three cases. The and computed acceleration RMS values. It can be seen that the varia-
errors between the measured and computed average RMS values in tion parts of Case 1 match the measured ones quite well. It can also
the lateral direction may be attributed to the unavailability of the be seen from Fig. 20 that the variation parts of Case 2 in the three
flutter derivatives related to the lateral motions of the bridge deck. directions are similar to those of Case 3. This is because both Case 2

© ASCE 04020255-12 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


10-1 100 10-3
Measured Measured Measured
Calculated Calculated Calculated

Torsional EPSD (rad2/s3)


10-3 10-2 10-5

Vertical EPSD (m2/s3)


Lateral EPSD (m2/s3)

10-5 10-4 10-7

10-7 10-6 10-9

t=1000s t=1000s t=1000s


10-9 10-8 10-11
0.0 0.5 1.0 0.0 0.5 1.0 0.0 0.5 1.0
n (Hz) n (Hz) n (Hz)
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

10-1 100 10-3


Measured Measured Measured
Calculated Calculated Calculated

Torsional EPSD (rad2/s3)


10-3 10-2 10-5

Vertical EPSD (m2/s3)


Lateral EPSD (m2/s3)

10-5 10-4 10-7

10-7 10-6 10-9

t=3000s t=3000s t=3000s


10-9 10-8 10-11
0.0 0.5 1.0 0.0 0.5 1.0 0.0 0.5 1.0
n (Hz) n (Hz) n (Hz)

Fig. 17. EPSD comparisons of measured and computed acceleration responses.

-3
10 ×10
L (m/s )
2

Stationary
Non-stationary
2
0 600 1200 1800 2400 3000 3600
Times (s)
-2
10 ×10
(m/s2)

6
V

Stationary
Non-stationary
2
0 600 1200 1800 2400 3000 3600
Times (s)
-4
18 ×10
(rad/s2)

11
T

Stationary
Non-stationary
4
0 600 1200 1800 2400 3000 3600
Times (s)

Fig. 18. Comparison of stationary and nonstationary buffeting responses.

© ASCE 04020255-13 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Table 3. Average RMS values of nonstationary acceleration responses of bridge deck at midspan
Case 1 Case 2 Case 3
Direction Measured Value Error (%) Value Error (%) Value Error (%)
2
Lateral (m=s ) 0.0099 0.0071 −28.28 0.0073 −26.26 0.0079 −20.20
Vertical (m=s2 ) 0.0625 0.0737 17.92 0.0767 22.72 0.0826 32.16
Torsional (rad=s2 ) 0.0012 0.0012 0 0.0013 8.33 0.0014 16.67
Note: Error = (computed-measured)/measured.

Measured Measured
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

10-3 Case 1 10-3 Case 1


Case 3 Case 2
Lateral EPSD (m2/s3)

Lateral EPSD (m2/s3)


Case 2 Case 3

10-5 10-5

10-7 10-7

t=2000s t=3000s
10-9 10-9
0.0 0.5 1.0 0.0 0.5 1.0
n (Hz) n (Hz)

Fig. 19. Slice comparison of the measured and computed EPSDs of the lateral acceleration responses.

-3
3 ×10
(m/s2 )

0
Measured
Case 1
L
~

-3
Case 2
Case 3
-6
0 600 1200 1800 2400 3000 3600
Times (s)
-2
2 ×10
(m/s2 )

0
Measured
Case 1
V

-2
~

Case 2
Case 3
-4
0 600 1200 1800 2400 3000 3600
Times (s)
×10-4

3
(rad/s2 )

0
Measured
Case 1
T

-3
~

Case 2
-6 Case 3
0 600 1200 1800 2400 3000 3600
Times (s)

Fig. 20. Comparison of variation parts of the measured and computed acceleration responses.

© ASCE 04020255-14 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Table 4. Correlation coefficients of variation parts of the measured and the lateral direction, while the correlation coefficients in the other
computed RMS values two directions both are below 0.8. The results show that the varia-
Direction Case 1 Case 2 Case 3 tion parts of the time-varying RMS value can be well represented
Lateral 0.984 0.942 0.942
by considering the time-varying complex coherence function. If the
Vertical 0.941 0.638 0.638 time-varying feature of the coherence function is ignored, the maxi-
Torsional 0.938 0.703 0.703 mum in the variation part will be noticeably underestimated. With a
comprehensive consideration of both the average and the variation
parts of the time-varying RMS value, Case 1 provides the most
accurate estimation of the nonstationary responses of the Stonecut-
and Case 3 consider the time-invariant coherence functions. This ters Bridge.
result also indicates that the phase difference between Case 2 and Apart from the aforementioned nonstationary acceleration re-
Case 3 affects the average RMS value rather than the variation part. sponses, the RMS values of the nonstationary displacement re-
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

To quantify the difference of the variation part between the three sponses of the bridge at the midspan are also computed. Fig. 21
cases and the measurement, the correlation coefficients are calcu- shows the RMS comparison of the computed displacements with
lated using the following equation time-varying and time-invariant coherence functions. It can be seen
RT c that the RMS values of Case 2 are more close to those of Case 1.
0 σ ~ i ðtÞσ~ m i ðtÞdt
r ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
RT c 2 R ffi i ¼ L; V; T ð23Þ Due to the overestimation of the spatial correlation by using the
2
0 ½σ ~ i ðtÞ dt · 0T ½σ~ m i ðtÞ dt
modulus of coherence function, the RMS values of Case 3 are
much larger than those of Case 1. The difference in the lateral, ver-
where r = correlation coefficient; and σ~ ci ðtÞ and σ~ m
i ðtÞ = variation
tical, and torsional directions can reach 30.5%, 29.1%, and 29.6%,
parts of the computed and measured time-varying RMS values, respectively. It can be concluded that the consideration of a time-
respectively. varying complex coherence can make a more accurate prediction
The correlation coefficient calculated using Eq. (23) are listed in of nonstationary buffeting responses of long-span bridges under
Table 4. It can be seen from Table 4 that the correlation coefficients typhoon winds. For engineering application purposes, the enhanced
of Case 1 are all larger than 0.9, demonstrating a high correlation nonstationary buffeting analysis framework by considering time-
between the computed and measured results of variation parts. For varying coherence would provide a refined guidance for the wind-
Cases 2 and 3, the correlation coefficients are only acceptable for resistant design of long-span bridges.

×10-3
8
L (m)

4
Case 1
Case 2
Case 3
0
0 600 1200 1800 2400 3000 3600
Time (s)

×10-2
8
(m)

4
Case 1
V

Case 2
Case 3
0
0 600 1200 1800 2400 3000 3600
Time (s)

×10-4
8
(rad)

4
Case 1
T

Case 2
Case 3
0
0 600 1200 1800 2400 3000 3600
Time (s)

Fig. 21. Comparison of computed displacements with time-varying and time-invariant coherences.

© ASCE 04020255-15 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Conclusions Z ∞
Sxy ðω; tÞ ¼ Wðτ Þ½U x ðω; t − τ ÞU y ðω; t − τ Þdτ ð26Þ
The time-varying features of wind spectra and coherences and their −∞

effects on the nonstationary buffeting response of long-span bridges


where Sxy ðω; tÞ = CEPSD of xðtÞ and yðtÞ; τ denotes the lag time;
under typhoon winds have been investigated in this study. The field pffiffiffiffiffiffi
measurement data of Typhoon Hato in 2017 and the responses of the the symbol * denotes a conjugate indicator; i ¼ −1; gðtÞ and
Stonecutters cable-stayed bridge in Hong Kong recorded during the WðtÞ = window functions chosen as
typhoon were analyzed. The analytical framework for nonstationary  pffiffiffiffiffiffi
buffeting analysis of long-span bridges proposed by the second au- 1=2 hπ jtj ≤ h
gðtÞ ¼ ð27Þ
thor and his colleagues has been extended to include time-varying 0 jtj > h
complex coherence functions and applied to the Stonecutters cable-
stayed bridge. The major conclusions can be drawn from this study 
1=T 0 −T 0 =2 ≤ t ≤ T 0 =2
as follows:
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

WðtÞ ¼ ð28Þ
1. Significant time-varying features exist in wind spectra and coher- 0 otherwise
ences of the measured typhoon winds. The amplitude of the
imaginary part of the coherence function is in the same order as where h and T 0 = window lengths.
the real part. The time-varying complex coherence function It is noted that the CEPSD, Sxy ðω; tÞ, will be reduced as the
should be considered for typhoon winds near its external eye EPSD, Sxx ðω; tÞ, if xðtÞ ¼ yðtÞ. The integrals in Eqs. (24)–(26)
wall. can be calculated by convolutions. The essence of this estimation
2. The time-invariant complex coherence function proposed by method is to identify the local spectral information at the time in-
Krenk (1996) has been extended to the time-varying complex stant t by using a window function gðtÞ. The window WðtÞ is used
coherence function. The time-varying wind spectrum model as a weight function to smooth the estimated values along time to
(Hu et al. 2017) and the extended time-varying complex coher- reduce the estimation error.
ence model could well fit the measured time-varying wind spec- After the CEPSD and EPSDs of xðtÞ and yðtÞ are obtained, the
tra and coherences. time-varying coherence function can be calculated as
3. The analytical framework for nonstationary buffeting analysis
of long-span bridges proposed by Hu et al. (2013) has been ex- Sxy ðω; tÞ
γ xy ðω; tÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð29Þ
tended to include the time-varying complex coherence functions Sxx ðω; tÞSyy ðω; tÞ
and applied to the Stonecutters cable-stayed bridge. The time-
varying mean responses and the nonstationary buffeting re- As the CEPSD between two different locations is complex, the
sponses considering the three cases of different coherence func- time-varying coherence function is also complex and can be ex-
tions are computed and compared with the measured ones. pressed as
4. The average RMS values of the acceleration responses computed
from the three cases are compared with the measured average γ xy ðω; tÞ ¼ jγ xy ðω; tÞje−iθxy ðω;tÞ ð30Þ
RMS value of the acceleration response. The comparative re-
sults shows that the use of the modulus of coherence function in where |·| = operator to take the modulus; and θxy ðω; tÞ =
Case 3 without consideration of phase difference overestimates phase angle.
the correlation of typhoon wind speeds between two locations.
The use of the time-varying complex coherence function as used
in Case 1 is most appropriate.
Appendix II. Nonstationary Force Models
5. The variation parts of the time-varying RMS value can be well
represented by considering the time-varying complex coherence For the buffeting analysis of a long-span bridge, the wind forces
function. If the time-varying feature of the coherence function is acting on the bridge deck is the most important concern. The
ignored, the maximum in the variation part will be noticeably nonstationary force models mainly cover the time-varying mean
underestimated. wind forces, nonstationary buffeting forces, and nonstationary
6. With a comprehensive consideration of both the average RMS self-excited forces.
value and the variation part of the time-varying RMS value, The time-varying mean wind forces on the bridge deck per unit
Case 1 using the time-varying complex coherence functions pro- length can be expressed as
vides a most accurate estimation of the nonstationary responses
of the bridge. 1
Lm ðtÞ ¼ ρU~ 2m ðtÞCL ðαðtÞÞB ð31Þ
2

Appendix I. Estimation Method for EPSD, CEPSD, 1


and Time-Varying Coherence Dm ðtÞ ¼ ρU~ 2m ðtÞCD ðαðtÞÞB ð32Þ
2
Suppose xðtÞ and yðtÞ are the two nonstationary processes, the
1
CEPSD of xðtÞ and yðtÞ can be calculated based on the following M m ðtÞ ¼ ρU~ 2m ðtÞCM ðαðtÞÞB2 ð33Þ
formulas (Priestley 1965; Priestley and Tong 1973) 2
Z ∞ where Lm , Dm , and M m = time-varying mean lift force, drag force,
U x ðω; tÞ ¼ gðτ Þxðt − τ Þ exp½−iωðt − τ Þdτ ð24Þ and pitching moment, respectively; ρ = air density; CL , CD , and
−∞
CM = steady-state force coefficients measured from wind tunnel
Z tests; B = width of the bridge deck; and αðtÞ = instantaneous mean

U y ðω; tÞ ¼ gðτ Þyðt − τ Þ exp½−iωðt − τ Þdτ ð25Þ wind attack angle, which is the sum of the attack angle of wind and
−∞ the torsional displacement of the bridge.

© ASCE 04020255-16 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Based on the quasisteady theory, the nonstationary buffeting References
forces acting on the bridge deck per unit length are given by
ASCE. 2010. Minimum design loads for buildings and other structures.
1
Lb ðtÞ ¼ ρU~ m ðtÞBχL ðn; tÞf2CL ðαðtÞÞuðtÞ
ASCE 7-10. Reston, VA: ASCE.
2 Battaglia, F. 1979. “Some extensions of the evolutionary spectral analysis
of a stochastic process.” Boll. dell’Unione Matematica Ital. 16 (5):
þ ½CD ðαðtÞÞ þ CL0 ðαðtÞÞwðtÞg ð34Þ
1154–1166.
Cao, S. Y., Y. Tamura, N. Kikuchi, M. Saito, I. Nakayama, and Y.
1
Db ðtÞ ¼ ρU~ m ðtÞBχD ðn; tÞf2CD ðαðtÞÞuðtÞ Matsuzaki. 2009. “Wind characteristics of a strong typhoon.” J. Wind
2 Eng. Ind. Aerodyn. 97 (1): 11–21. https://doi.org/10.1016/j.jweia.2008
þ ½CD0 ðαðtÞÞ − CL0 ðαðtÞÞwðtÞg ð35Þ .10.002.
CEN (European Committee for Standardization). 1994. Basis of design and
1 actions on structures, part 2-4: Wind actions. Eurocode 1. Brussels,
M b ðtÞ ¼ ρU~ m ðtÞB2 χM ðn; tÞf2CM ðαðtÞÞuðtÞ þ CM0 ðαðtÞÞwðtÞg
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

Belgium: CEN.
2
Chen, L., and C. W. Letchford. 2005. “Proper orthogonal decomposition
ð36Þ of two vertical profiles of full-scale nonstationary downburst wind
speeds.” J. Wind Eng. Ind. Aerodyn. 93 (3): 187–216. https://doi.org/10
where Lb , Db , and M b = nonstationary lift force, drag force, and .1016/j.jweia.2004.11.004.
pitching moment, respectively; CL0 ¼ dCL =dα, CD0 ¼ dCD =dα, Chen, X. Z. 2008. “Analysis of alongwind tall building response to tran-
CM0 ¼ dCM =dα = first-order derivatives of steady-state coeffi- sient nonstationary winds.” J. Struct. Eng. 134 (5): 782–791. https://doi
cients with respect to the attack angle; and χL ðn; tÞ, χD ðn; tÞ, and .org/10.1061/(ASCE)0733-9445(2008)134:5(782).
χM ðn; tÞ = time-varying aerodynamic admittance functions. Chen, X. Z. 2015. “Analysis of multimode coupled buffeting response of
By the extension of the stationary self-excited forces, the non- long-span bridges to nonstationary winds with force parameters from
stationary self-excited forces acting the bridge deck per unit length stationary wind.” J. Struct. Eng. 141 (4): 04014131. https://doi.org/10
can be expressed as .1061/(ASCE)ST.1943-541X.0001078.
Chen, X. Z., and A. Kareem. 2002. “Advances in modeling of aerodynamic
Lse ðtÞ ¼ ρB2 ω½H 1 ḣðtÞ þ BH2 α̇ðtÞ þ ωBH 3 αðtÞ forces on bridge decks.” J. Eng. Mech. 128 (11): 1193–1205. https://doi
.org/10.1061/(ASCE)0733-9399(2002)128:11(1193).
þ ωH4 hðtÞ þ H 5 ṗðtÞ þ ωH6 pðtÞ ð37Þ
Chen, X. Z., M. Matsumoto, and A. Kareem. 2000. “Time domain flutter
and buffeting response analysis of bridges.” J. Eng. Mech. 126 (1):
Dse ðtÞ ¼ ρB2 ω½P1 ṗðtÞ þ BP2 α̇ðtÞ þ ωBP3 αðtÞ þ ωP4 pðtÞ 7–16. https://doi.org/10.1061/(ASCE)0733-9399(2000)126:1(7).
þ P5 ḣðtÞ þ ωP6 hðtÞ ð38Þ Choi, E. C. C. 1983. “Gradient height and velocity profile during ty-
phoons.” J. Wind Eng. Ind. Aerodyn. 13 (1–3): 31–41. https://doi.org/10
.1016/0167-6105(83)90126-5.
M se ðtÞ ¼ ρB3 ω½A1 ḣðtÞ þ BA2 α̇ðtÞ þ ωBA3 αðtÞ þ ωA4 hðtÞ Conte, J. P., and B. F. Peng. 1997. “Fully nonstationary analytical earth-
þ A5 ṗðtÞ þ ωA6 pðtÞ ð39Þ quake ground-motion model.” J. Eng. Mech. 123 (1): 15–24. https://doi
.org/10.1061/(ASCE)0733-9399(1997)123:1(15).
where Lse , Dse , and Mse = nonstationary self-excited lift force, Davenport, A. G. 1961a. “The spectrum of horizontal gustiness near the
drag force, and pitching moment, respectively; Hj , Pj , and Aj ground in high winds.” Q. J. R. Meteorol. Soc. 87 (372): 194–211.
https://doi.org/10.1002/qj.49708737208.
(j ¼ 1; 2; : : : ; 6) = flutter derivatives, which are the functions of the
Davenport, A. G. 1961b. “A statistical approach to the treatment of wind
reduced frequency ωB=U~ m ðtÞ as well as the attack angle; h, p, and
loading on tall masts and suspension bridges.” Ph.D. thesis, Dept. of
α = vertical, lateral, and torsional displacements; and the overdot Civil Engineering, Univ. of Bristol.
denotes the differentiation with respect to time. Eqs. (37)–(39) were Davenport, A. G. 1962. “Buffeting of a suspension bridge by storm winds.”
employed to identify the flutter derivatives of the Stonecutters J. Struct. Div. 88 (3): 233–270.
Bridge (Hui 2006) without 1/2 in front of the equations in compari- Ge, Y. J., and H. F. Xiang. 2011. “Extension of bridging capacity of cable-
son with the expressions of time-varying mean wind forces and buf- supported bridges using double main spans or twin parallel decks so-
feting forces. lutions.” Struct. Infrastruct. Eng. 7 (7–8): 551–567. https://doi.org/10
.1080/15732479.2010.496980.
Gimsing, N. J. 1983. Cable supported bridges concept and design.
Data Availability Statement New York: Wiley.
Hansen, S. O., and S. Krenk. 1999. “Dynamic along-wind response of sim-
Some or all data, models, or code used during the study were pro- ple structures.” J. Wind Eng. Ind. Aerodyn. 82 (1–3): 147–171. https://
vided by a third party. Direct requests for these materials may be doi.org/10.1016/S0167-6105(98)00215-3.
made to the provider as indicated in the Acknowledgements. Hao, J. M., and T. Wu. 2018. “Downburst-induced transient response of
a long-span bridge: A CFD-CSD-based hybrid approach.” J. Wind
Eng. Ind. Aerodyn. 179 (Aug): 273–286. https://doi.org/10.1016/j.jweia
.2018.06.006.
Acknowledgments
He, Y. C., P. W. Chan, and Q. S. Li. 2013. “Wind characteristics over differ-
The authors would like to acknowledge the supports from The ent terrains.” J. Wind Eng. Ind. Aerodyn. 120 (Sep): 51–69. https://doi
.org/10.1016/j.jweia.2013.06.016.
Hong Kong Polytechnic University (PolyU 4-ZVM9), the Natural
Hu, L., Y. L. Xu, and W. F. Huang. 2013. “Typhoon-induced non-stationary
Science Foundation of Jiangsu Province (Grant No. BK20190359), buffeting response of long-span bridges in complex terrain.” Eng.
and the National Natural Science Foundation of China (Grant Struct. 57 (Dec): 406–415. https://doi.org/10.1016/j.engstruct.2013
No. 51908125). The authors also want to thank the Hong Kong .09.044.
Highways Department for providing the authors with the monitor- Hu, L., Y. L. Xu, Q. Zhu, A. N. Guo, and A. Kareem. 2017. “Tropical
ing data of the Stonecutters Bridge for academic research. Any opi- storm-induced buffeting response of long-span bridges: Enhanced non-
nions and conclusions presented in this paper are entirely those of stationary buffeting force model.” J. Struct. Eng. 143 (6): 04017027.
the authors. https://doi.org/10.1061/(ASCE)ST.1943-541X.0001745.

© ASCE 04020255-17 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Huang, G., H. T. Zheng, Y. L. Xu, and Y. L. Li. 2015. “Spectrum models for Powell, M. D., P. J. Vickery, and T. A. Reinhold. 2003. “Reduced
nonstationary extreme winds.” J. Struct. Eng. 141 (10): 04015010. drag coefficient for high wind speeds in tropical cyclones.” Nature
https://doi.org/10.1061/(ASCE)ST.1943-541X.0001257. 422 (6929): 279–283. https://doi.org/10.1038/nature01481.
Huang, Z. F., Y. L. Xu, T. Y. Tao, and S. Zhan. 2020. “Time-varying power Priestley, M. B. 1965. “Evolutionary spectra and non-stationary processes.”
spectra and coherences of non-stationary typhoon winds.” J. Wind Eng. J. R. Stat. Soc. Ser.: B (Methodol.) 27 (2): 204–229. https://doi.org/10
Ind. Aerodyn. 198 (Mar): 104115. https://doi.org/10.1016/j.jweia.2020 .1111/j.2517-6161.1965.tb01488.x.
.104115. Priestley, M. B., and H. Tong. 1973. “On the analysis of bivariate non-
Hui, M. C. H. 2006. “Turbulent wind action on long span bridges with stationary processes.” J. R. Stat. Soc. Ser.: B (Methodol.) 35 (2):
separated twin-girder decks.” Ph.D. thesis, Dept. of Civil Engineering, 153–166. https://doi.org/10.1111/j.2517-6161.1973.tb00949.x.
Tongji Univ. Professional Standard PRC (People’s Republic of China). 2018. Wind-
Hui, M. C. H., Q. S. Ding, and Y. L. Xu. 2006. “Flutter analysis of Stone- resistant design specification for highway bridges. Beijing: China
cutters Bridge.” Wind Struct. 9 (2): 125–146. https://doi.org/10.12989 Communications Press.
/was.2006.9.2.125. Scanlan, R. H. 1978. “The action of flexible bridges under wind, II:
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

Hui, M. C. H., A. Larsen, and H. F. Xiang. 2009a. “Wind turbulence Buffeting theory.” J. Sound Vib. 60 (2): 201–211. https://doi.org/10
characteristics study at the Stonecutters Bridge site. Part I: Mean wind .1016/S0022-460X(78)80029-7.
and turbulence intensities.” J. Wind Eng. Ind. Aerodyn. 97 (1): 22–36. Scanlan, R. H. 1993. “Bridge buffeting by skew winds in erection stage.”
https://doi.org/10.1016/j.jweia.2008.11.002. J. Eng. Mech. 119 (2): 251–269. https://doi.org/10.1061/(ASCE)0733
Hui, M. C. H., A. Larsen, and H. F. Xiang. 2009b. “Wind turbulence char- -9399(1993)119:2(251).
acteristics study at the Stonecutters Bridge site. Part II: Wind power Solari, G., P. D. Gaetano, and M. P. Repetto. 2015. “Thunderstorm response
spectra, integral length scales and coherences.” J. Wind Eng. Ind. Aero- spectrum: Fundamentals and case study.” J. Wind Eng. Ind. Aerodyn.
dyn. 97 (1): 48–59. https://doi.org/10.1016/j.jweia.2008.11.003. 143 (Aug): 62–77. https://doi.org/10.1016/j.jweia.2015.04.009.
Hui, M. C. H., and C. K. Wong. 2009. “Stonecutters Bridge—Durability, Solari, G., and G. Piccardo. 2001. “Probabilistic 3-D turbulence modeling
maintenance and safety considerations.” Struct. Infrastruct. Eng. 5 (3): for gust buffeting of structures.” Probab. Eng. Mech. 16 (1): 73–86.
229–243. https://doi.org/10.1080/15732470601130337. https://doi.org/10.1016/S0266-8920(00)00010-2.
Isyumov, N. 2012. “Alan G. Davenport’s mark on wind engineering.” Su, Y. W., G. Q. Huang, and Y. L. Xu. 2015. “Derivation of time-varying
J. Wind Eng. Ind. Aerodyn. 104 (May): 49–66. https://doi.org/10 mean for non-stationary downburst winds.” J. Wind Eng. Ind. Aerodyn.
.1016/j.jweia.2012.02.007. 141 (Jun): 39–48. https://doi.org/10.1016/j.jweia.2015.02.008.
Jain, A., N. P. Jones, and R. H. Scanlan. 1996. “Coupled flutter and buf- Tao, T. Y., and H. Wang. 2019. “Modelling of longitudinal evolutionary
feting analysis of long-span bridges.” J. Struct. Eng. 122 (7): 716–725. power spectral density of typhoon winds considering high-frequency
https://doi.org/10.1061/(ASCE)0733-9445(1996)122:7(716). subrange.” J. Wind Eng. Ind. Aerodyn. 193 (Oct): 103957. https://doi
.org/10.1016/j.jweia.2019.103957.
Kareem, A. 1986. “Performance of cladding in Hurricane Alicia.” J. Struct.
Eng. 112 (12): 2679–2693. https://doi.org/10.1061/(ASCE)0733-9445 Tao, T. Y., H. Wang, and A. Q. Li. 2016. “Stationary and nonstationary
analysis on the wind characteristics of a tropical storm.” Smart Struct.
(1986)112:12(2679).
Syst. 17 (6): 1067–1085. https://doi.org/10.12989/sss.2016.17.6.1067.
Kareem, A. 2008. “Numerical simulation of wind effects: A probabilistic
Tao, T. Y., H. Wang, and T. Wu. 2017. “Comparative study of the wind
perspective.” J. Wind Eng. Ind. Aerodyn. 96 (10–11): 1472–1497.
characteristics of a strong wind event based on stationary and nonsta-
https://doi.org/10.1016/j.jweia.2008.02.048.
tionary models.” J. Struct. Eng. 143 (5): 04016230. https://doi.org/10
Kareem, A., L. Hu, Y. L. Guo, and D. K. Kwon. 2019. “Generalized wind
.1061/(ASCE)ST.1943-541X.0001725.
loading chain: Time-frequency modeling framework for nonstationary
Tao, T. Y., H. Wang, and T. Wu. 2018. “Parametric study on buffeting
wind effects on structures.” J. Struct. Eng. 145 (10): 04019092. https://
performance of a long-span triple-tower suspension bridge.” Struct.
doi.org/10.1061/(ASCE)ST.1943-541X.0002376.
Infrastruct. Eng. 14 (3): 381–399. https://doi.org/10.1080/15732479
Krenk, S. 1996. “Wind field coherence and dynamic wind forces.” In Proc., .2017.1354034.
IUTAM Symp. on Advances in Nonlinear Stochastic Mechanics, Solid
Tobin, N., and L. P. Chamorro. 2018. “Turbulence coherence and its
Mechanics and its Applications. Dordrecht, Netherlands: Springer. impact on wind-farm power fluctuations.” J. Fluid Mech. 855 (Nov):
Li, L. X., A. Kareem, Y. Q. Xiao, L. L. Song, and C. Y. Zhou. 2015. “A 1116–1129. https://doi.org/10.1017/jfm.2018.713.
comparative study of field measurements of the turbulence character- Wang, H., R. M. Hu, J. Xie, and T. Tong. 2013. “Comparative study on
istics of typhoon and hurricane winds.” J. Wind Eng. Ind. Aerodyn. buffeting performance of Sutong Bridge based on design and measured
140: 49–66. https://doi.org/10.1016/j.jweia.2014.12.008. spectrum.” J. Bridge Eng. 18 (7): 587–600. https://doi.org/10.1061
Li, L. X., Y. Q. Xiao, A. Kareem, L. L. Song, and P. Qin. 2012. “Modeling /(ASCE)BE.1943-5592.0000394.
typhoon wind power spectra near sea surface based on measurements Wang, H., A. Q. Li, and R. M. Hu. 2011. “Comparison of ambient vibration
in the South China Sea.” J. Wind Eng. Ind. Aerodyn. 104–106 (May): response of the Runyang Suspension Bridge under skew winds with
565–576. https://doi.org/10.1016/j.jweia.2012.04.005. time-domain numerical predictions.” J. Bridge Eng. 16 (4): 513–526.
Lin, J. H. 2004. Pseudo excitation method in random vibration. [In https://doi.org/10.1061/(ASCE)BE.1943-5592.0000168.
Chinese.] Beijing: Science Press. Wang, H., T. Y. Tao, Y. Q. Gao, and F. Y. Xu. 2018. “Measurement of
McCullough, M., D. K. Kwon, A. Kareem, and L. J. Wang. 2014. “Efficacy wind effects on a kilometer-level cable-stayed bridge during Typhoon
of averaging interval for nonstationary winds.” J. Eng. Mech. 140 (1): Haikui.” J. Struct. Eng. 144 (9): 04018142. https://doi.org/10.1061
1–19. https://doi.org/10.1061/(ASCE)EM.1943-7889.0000641. /(ASCE)ST.1943-541X.0002138.
Nagai, M., Y. Fujino, H. Yamaguchi, and E. Iwasaki. 2004. “Feasibility of a Wang, H., T. Wu, T. Y. Tao, A. Q. Li, and A. Kareem. 2016. “Measurements
1,400 m span steel cable-stayed bridge.” J. Bridge Eng. 9 (5): 444–452. and analysis of non-stationary wind characteristics at Sutong Bridge in
https://doi.org/10.1061/(ASCE)1084-0702(2004)9:5(444). Typhoon Damrey.” J. Wind Eng. Ind. Aerodyn. 151 (Apr): 100–106.
Nakamura, R., T. Shibayama, M. Esteban, and T. Iwamoto. 2016. “Future https://doi.org/10.1016/j.jweia.2016.02.001.
typhoon and storm surges under different global warming scenarios: Wang, L. J., and A. Kareem. 2004. “Modeling of non-stationary winds
Case study of typhoon Haiyan (2013).” Nat. Hazards 82 (3): 1645– in gust fronts.” In Proc., 9th ASCE Joint Specialty Conf. on Probabi-
1681. https://doi.org/10.1007/s11069-016-2259-3. listic Mechanics and Structural Reliability. Red Hook, NY: Curran
National Research Council. 2005. National building code of Canada. Associates.
Ottawa, ON: National Research Council. Xu, Y. L. 2013. Wind effects on long-span cable-supported bridges.
Peng, L. L., G. Q. Huang, X. Z. Chen, and Q. S. Yang. 2018. “Evolutionary Singapore: Wiley.
spectra-based time-varying coherence function and application in struc- Xu, Y. L., and J. Chen. 2004. “Characterizing nonstationary wind speed
tural response analysis to downburst winds.” J. Struct. Eng. 144 (7): using empirical mode decomposition.” J. Struct. Eng. 130 (6): 912–920.
04018078. https://doi.org/10.1061/(ASCE)ST.1943-541X.0002066. https://doi.org/10.1061/(ASCE)0733-9445(2004)130:6(912).

© ASCE 04020255-18 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255


Xu, Y. L., D. K. Sun, J. M. Ko, and J. H. Lin. 2000a. “Fully coupled buf- Yang, Q. S., R. Gao, F. Bai, T. Li, and Y. Tamura. 2018. “Damage to build-
feting analysis of Tsing Ma suspension bridge.” J. Wind Eng. Ind. Aero- ings and structures due to recent devastating wind hazards in East Asia.”
dyn. 85 (1): 97–117. https://doi.org/10.1016/S0167-6105(99)00133-6. Nat. Hazards 92 (3): 1321–1353. https://doi.org/10.1007/s11069-018
Xu, Y. L., Z. X. Tan, L. D. Zhu, Q. Zhu, and S. Zhan. 2019. “Buffeting- -3253-8.
induced stress analysis of long-span twin-box-beck bridges based Zhu, Q., and Y. L. Xu. 2014. “Characteristics of distributed aerodynamic
on POD pressure modes.” J. Wind Eng. Ind. Aerodyn. 188 (May): forces on a twin-box bridge deck.” J. Wind Eng. Ind. Aerodyn.
397–409. https://doi.org/10.1016/j.jweia.2019.03.016. 131 (Aug): 31–45. https://doi.org/10.1016/j.jweia.2014.05.003.
Xu, Y. L., L. D. Zhu, K. Y. Wong, and K. W. Y. Chan. 2000b. “Field meas- Zhu, Q., Y. L. Xu, and K. M. Shum. 2016. “Stress-level buffeting analysis
urement results of Tsing Ma suspension bridge during Typhoon Victor.” of a long-span cable-stayed bridge with a twin-box deck under distrib-
Struct. Eng. Mech. 10 (6): 545–559. https://doi.org/10.12989/sem.2000 uted wind loads.” Eng. Struct. 127 (Nov): 416–433. https://doi.org/10
.10.6.545. .1016/j.engstruct.2016.08.050.
Downloaded from ascelibrary.org by CORNELL UNIV LIBRARIES on 09/17/20. Copyright ASCE. For personal use only; all rights reserved.

© ASCE 04020255-19 J. Struct. Eng.

J. Struct. Eng., 2020, 146(12): 04020255

You might also like