You are on page 1of 13

Generalized Wind Loading Chain: Time-Frequency

Modeling Framework for Nonstationary


Wind Effects on Structures
Ahsan Kareem, Dist.M.ASCE 1; Liang Hu 2; Yanlin Guo, A.M.ASCE 3; and Dae-Kun Kwon, M.ASCE 4
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

Abstract: This study proposes a generalized wind loading chain to describe a complete relationship among wind, force, and response
induced by nonstationary wind events such as tropical storms or downbursts to complement the Davenport wind loading chain. In the pro-
posed chain, nonstationary winds are represented as a nonstationary model in terms of the time-varying mean and nonstationary fluctuating
wind components similar to a stationary model involving the mean and stationary fluctuating wind components in the Davenport’s stationary
wind loading chain. Specifically, the five chain components of the fluctuating wind in the Davenport’s chain such as gustiness of wind,
aerodynamic transfer/admittance, aerodynamic force, structural transfer/admittance, and response statistics are recast as time-dependent
counterparts in the time-frequency domain to capture nonstationary winds effects on structures. These components are formulated using
the evolutionary power spectral density (EPSD) as a form of time-frequency representation that captures salient features of nonstationary
tropical storm and downburst winds. As an alternative, a wavelet-based representation is also offered. A numerical example demonstrates the
estimation of nonstationary response using the proposed generalized chain. For possible codification of nonstationary wind effects on struc-
tures, a preliminary examination suggests that the extreme nonstationary response computed by the generalized chain may be conveniently
meshed with the gust front factor or its generalized version frameworks. This may facilitate a rapid shift in the current design approach from
stationary to nonstationary winds. DOI: 10.1061/(ASCE)ST.1943-541X.0002376. © 2019 American Society of Civil Engineers.
Author keywords: Generalized wind loading chain; Nonstationary winds; Evolutionary power spectral density (EPSD); Wavelet
transform; Extreme response; Gust front factor (GFF).

Introduction synoptic winds. Field measurements have shown that the mean
wind as well as the spectral content of fluctuating components of
Over the years, it has been noted that some extreme wind events nonstationary winds may vary with time considerably (Wang et al.
recorded, while exhibiting dominant wind speeds, may not fall 2016; Wang and Kareem 2004; Xu and Chen 2004). Accordingly,
broadly in the category of stationary synoptic winds, e.g., tropical the nonturbulent component of nonstationary winds needs be char-
storms (typhoons, hurricanes, cyclones), downbursts, etc. With acterized by the time-varying mean wind estimated or modeled
strong vertical components of vorticity and convection, the chang- based on full-scale measurements (Bendat and Piersol 2010; Chen
ing kinematics and dynamics of these exceptional winds modify et al. 2007; Chen and Letchford 2007), whereas the turbulent
or vitiate the boundary layer structure and develop an organized (or fluctuating) wind component may be captured by the time-
structure in the horizontal plane, resulting in nonstationary and frequency tools such as evolutionary power spectral density
even transient nature of winds. Impacts of nonstationary features of (EPSD) (e.g., Chay et al. 2006; Chen and Letchford 2004a; Chen
such exceptional winds on the wind resistant design of civil struc- 2008, 2015; Kwon and Kareem 2009, 2013; Priestley 1965; Wang
tures thus need further examination. and Kareem 2004), empirical modal decomposition (EMD) (Huang
Nonstationary winds involve time-dependent statistical prop- et al. 2016; Kareem et al. 1999), and wavelet transform (Gurley and
erties, in contrast to the time-invariant properties of stationary Kareem 1999; Spanos and Failla 2005), etc. Furthermore, these
nonstationary characteristics may amplify structural response in
1
Robert M. Moran Professor, NatHaz Modeling Laboratory, Dept. of consideration of the transient aerodynamics and structural dynam-
Civil and Environmental Engineering and Earth Sciences, Univ. of Notre ics (Kareem 2009). Based on the nonstationary modeling, the tran-
Dame, Notre Dame, IN 46556. Email: kareem@nd.edu sient aerodynamic effects of nonstationary winds on structures have
2
Ph.D. Student, NatHaz Modeling Laboratory, Dept. of Civil and
Environmental Engineering and Earth Sciences, Univ. of Notre Dame,
been investigated through wind tunnel tests (Butler et al. 2010) and
Notre Dame, IN 46556 (corresponding author). Email: peettr@gmail.com numerical simulations (Chay et al. 2006), which highlighted that
3
Assistant Professor, Dept. of Civil and Environmental Engineering, research needs to evaluate the nonstationary wind-induced re-
Colorado State Univ., Fort Collins, CO 80523. Email: yanlin.guo@ sponse (e.g., Chen and Letchford 2004b; Chen 2008, 2015; Hu
colostate.edu et al. 2013; Kawai 2000; Kwon and Kareem 2009, 2013). Conse-
4
Research Assistant Professor, NatHaz Modeling Laboratory, Dept. of quently, nonstationary winds may significantly affect the design
Civil and Environmental Engineering and Earth Sciences, Univ. of Notre wind loading, rendering the application of the widely used gust
Dame, Notre Dame, IN 46556. Email: dkwon@nd.edu
loading factor approach in codes and standards problematic, as
Note. This manuscript was submitted on June 19, 2018; approved on
January 15, 2019; published online on July 16, 2019. Discussion period it is based on the stationary assumption of synoptic winds.
open until December 16, 2019; separate discussions must be submitted Based on past studies regarding nonstationary winds and their
for individual papers. This paper is part of the Journal of Structural impacts on structures in recent decades, we are at an appropriate
Engineering, © ASCE, ISSN 0733-9445. juncture to reflect on recent advances, reassess their advantages

© ASCE 04019092-1 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


and shortcomings, and identify the need for embarking on general where Uðz; tÞ = the nonstationary wind; Ūðz; tÞ = the time-varying
modeling philosophies and paradigms. The current paradigms con- mean wind; and uðz; tÞ = the nonstationary fluctuating wind.
cerning the wind resistant design are based on the well-established Assuming that the nonstationary wind field may be factorized into
Davenport’s wind loading chain (Davenport 1967; Isyumov 2012), a product of time and space components, the time-varying mean
which was proposed for synoptic winds and primarily focused wind may be expressed as
on stationary winds. Implicit or partial extension of Davenport’s
chain in the context of nonstationary winds has been attempted Ūðz; tÞ ¼ Ū 0 · Ū H ðzÞ · Ū T ðtÞ ð2Þ
for tropical storm and thunderstorm/downburst wind events (Chen
2008, 2015; Hu et al. 2013; Kwon and Kareem 2009; Solari et al. where Ū 0 = the reference horizontal mean wind at a specified
2015). However, a more comprehensive framework for shaping the height H0 (e.g., 10 m above the ground); Ū H ðzÞ = the vertical
design practice to encapsulate the generic nature of nonstationary profile of nonstationary winds; and Ū T ðtÞ = the time function
winds that seamlessly reduces to Davenport’s loading chain in normalized with respect to Ū 0 . Among these three components,
the case of stationary winds is missing. To this end, this paper Ū 0 describes the intensity of nonstationary winds events, whereas
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

proposes a generalized wind loading chain framework that extends Ū H ðzÞ and Ū T ðtÞ represent the spatial and temporal variation of
Davenport’s wind loading chain to appropriately model a nonsta- nonstationary winds that may differentiate themselves from con-
tionary wind-force-response relationship in time-frequency represen- ventional stationary winds.
tations. This framework reduces to the Davenport’s formulation for 0
Reference Mean Wind U
stationary winds and thus enables researchers to capture both sta-
The reference mean wind speed at a specific height is defined as the
tionary and nonstationary wind effects on structures, and therefore
wind speed averaged over a specified time period (T 0 ) of concern.
has been referred to as the generalized loading chain. In this study,
Note that the time scales of tropical storms and downbursts are sig-
the generalized chain is formulated using evolutionary power
nificantly different. The tropical storms usually have a much larger
spectral density and wavelet transform as time-frequency represen-
time scale (usually 1–2 days) compared to that of downburst
tations for capturing salient features of nonstationary tropical
(about 5–30 min) (e.g., Letchford and Chay 2002) and this differ-
storm and downburst winds. Its efficacy is demonstrated through
ence may contribute to the discrepancy of models. Variable T 0 also
a numerical example. Finally, a preliminary examination of its fea-
should be selected with reference to empirical time periods em-
sibility that the extreme nonstationary response computed by the
ployed for characterizing winds commonly in the past. In this study,
generalized chain may be conveniently coupled with the gust front
the values of 10 min and 1 h are used for downbursts and tropical
factor (GFF) or generalized gust front factor (G-GFF) frameworks,
storms, respectively. For downbursts, Ū 0 has usually been taken as
which may enable to rapidly shift the current design paradigm from
the maximum of the short time-averaged mean wind speed within
stationary to nonstationary winds, has been made.
T 0 , while for tropical storms Ū 0 takes the average wind speed
over T 0.
It is desirable that Ū 0 at a specified site is provided according to
Modeling of Nonstationary Winds an occurrence probability of the event. Data-fitted models of Ū 0 at
one site may be obtained from long-term measurements of nonsta-
Modeling of nonstationary wind events is the most critical but also
tionary wind events, under the assumption that such wind events
the most delicate step. Universal parametric models for stationary
can envelop extreme wind conditions at the site (e.g., hurricanes
winds (e.g., mean wind, wind spectrum of fluctuating wind com-
for Miami, Florida). Insufficient local measurements may be sup-
ponent, wind profile, etc.) have been refined over decades based on
plemented by the Monte Carlo simulation of the events [e.g., down-
full-scale observations and advances in modeling. Using the prop-
bursts (Aboshosha et al. 2017) and tropical storms (Emanuel et al.
erties of ergodicity and homogeneity, these models are applicable
2006; Huang and Xu 2013; Vickery et al. 2009a)].
anytime and anywhere with adjustable site-specific parameters.
However, developing such models for nonstationary winds is quite  H z
Vertical Wind Profile U
challenging. Both the forms and parameters of nonstationary mod-
Wind profiles of tropical storms and downbursts have reached solid
els are contingent on temporal and spatial features and the mecha-
suggestions on their shapes. It has been reported in the literature
nism of nonstationarity (e.g., distinct characteristics of tropical
that the vertical profile of tropical storms is similar to stationary
storms and downbursts), i.e., the inherent lack of ergodicity and
winds. Based on full-scale observations, the logarithmic law has
inhomogeneity of nonstationary winds as compared to stationary
been shown to be efficient and universal (Vickery et al. 2009b),
ones. The paucity of measurement data under such wind events
although the power law may also be applicable (Song et al. 2016).
further compounds the difficulty of universal models. Currently,
A logarithmic profile of typhoon winds is presented in Fig. 2. In
the development of universal models for nonstationary wind char-
consideration of local complex terrains, the CFD simulation may
acteristics (e.g., time-varying mean and time-frequency wind spec-
be used to determine the site-specific vertical profile (Huang and
trum, vertical wind profile, etc.) is still in progress and far from
Xu 2013).
maturation, which although it suggests multiple research opportu-
Conversely, a limited number of full-scale measurements of
nities, is beyond the scope of this paper.
downburst wind events (e.g., Fujita 1981; Gunter and Schroeder
2015; Hjelmfelt 1988; Lombardo et al. 2014) has suggested a
Time-Varying Mean Wind nose-shape profile as shown in Fig. 2. Some analytical or empirical
models have been developed to describe the salient feature of the
In this study, similar to the stationary wind model comprised of downburst profile as a function of several parameters (e.g., Abd-
the sum of mean and fluctuating wind components, a nonstationary Elaal et al. 2014; Chay et al. 2006; Vicroy 1992; Wood et al. 2001).
wind model involving time dependency is utilized, which is ex-
pressed as the sum of time-varying mean and zero-mean nonsta- Normalized Time Function U  T t
tionary fluctuating (or turbulent) wind components (Fig. 1) The time function [Ū T ðtÞ] is one of the essential characteristics
of nonstationary winds. A variety of methods have been proposed
Uðz; tÞ ¼ Ūðz; tÞ þ uðz; tÞ ð1Þ to extract Ū T ðtÞ from time histories, such as moving average

© ASCE 04019092-2 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


Generalized Nonstationary Wind Loading Chain
Non-Gaussian
Non-linear Non-linear

Time-varying Time-varying Time-varying


mean velocity mean force mean response
Non-
Nonstationary
stationary
Response
wind
Fluctuating Fluctuating Fluctuating
gust velocity wind force response

Time- Instantaneous Time- Instantaneous Time-


Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

frequency aerodynamic frequency transfer frequency


gust transfer force function of response
spectrum function spectrum structure spectrum

EPSD

Wavelet

Drop time variable

Davenport Wind Loading Chain


Velocity Force Response

Time-
domain

Gust Spectrum Aero Admittance Force Spectrum Mechanical Admittance Response Spectrum

Frequency
-domain

Fig. 1. Schematic diagram of generalized wind loading chain (in contours, brighter color: higher values, darker color: lower values).

(Choi 2000; Hu et al. 2013), empirical modal decomposition shown as dash lines. In the figure, T 1 was selected as 10 and 1 min
(Kareem and Kijewski 2002; Xu and Chen 2004), and wavelet for the tropical storm and downburst winds, respectively.
analysis (Chen and Letchford 2005; Wang and Kareem 2004). Modeling of Ū T ðtÞ for tropical storms are rather limited.
Appropriateness and efficiency of these methods depend on the Although Monte Carlo simulation of tropical storms may generate
characteristics of data (Su et al. 2015). In this paper, only the mov- Ū T ðtÞ at arbitrary locations, its accuracy is constrained by the usual
ing average method is presented due to its simplicity and popularity 6-h averaged physical parameters (Vickery et al. 2009a), which
Z thus may not be suitable for the response evaluation in this paper
1 tþT 1 =2 UðtÞ
Ū T ðtÞ ¼ dt ð3Þ
T1 t−T 1 =2 Ū 0 Ū H ðzÞ
300
Downburst
where the averaging time interval T 1 can be a constant or variable. Tropical storm
Height (m)

The variable T 1 may be determined by the trial-and-error method 200


(Bendat and Piersol 2010), by testing the randomness and tur-
bulence characteristics of the fluctuating turbulent component
(McCullough et al. 2014), or with reference to widely-accepted 100
values for stationary winds (Hu et al. 2013). The typical averaging
time for downbursts has been reported in the range of 30–60 s 0
using the moving average filter. A summary of past studies regard- 10 15 20
ing the averaging time of downbursts based on full-scale data can Wind speed (m/s)
be found in Solari et al. (2015). In Figs. 3(a) and 4(a), time-varying
Fig. 2. Example of vertical profiles of nonstationary winds.
mean winds extracted from the downburst and typhoon data are

© ASCE 04019092-3 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


105
Time-varying mean Time-varying mean 0.02 Time-varying mean
30 2

Wind speed (m/s)

Displacement (m)
Force (N)
20 0.01
1

10 0
0

0 -0.01
-1
100 200 300 400 500 600 100 200 300 400 500 600 100 200 300 400 500 600
(a) Time (s) (b) Time (s) (c) Time (s)
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Time histories of downburst: (a) wind; (b) force; and (c) displacement.

105
Time-varying mean Time-varying mean
3
Wind speed (m/s)

Displacement (m)
30 0.02
Force (N)

2
20 0.01
1
10 0
0
0 Time-varying mean -0.01
0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
(a) Time (s) (b) Time (s) (c) Time (s)

Fig. 4. Time histories of tropical storm: (a) wind; (b) force; and (c) displacement.

concerning a shorter time period (normally up to 1 h). Furthermore, corresponding force and finally to the response) into a nonstation-
an analytical form for Ū T ðtÞ over multiple measured wind events is ary format in the time-frequency domain. The first box of the third
currently not available. row describes the nonstationary gust spectrum in terms of a time-
For downburst events, attempts to establish analytical models of frequency representation, e.g., evolutionary power spectral den-
Ū T ðtÞ has been made through limited full-scale data (e.g., Holmes sity or wavelet transform format, which will be discussed in the
and Oliver 2000; Abd-Elaal et al. 2014; Aboshosha et al. 2016; next sections. The wind spectrum is related to the time-frequency
Chay et al. 2006) or CFD simulation (Chay et al. 2006; Kim and force spectrum (third box) through the instantaneous aerodynamic
Hangan 2007; Li et al. 2012). A simple form of a half-sinusoid transfer/admittance function (second box), exhibiting the wind-
function was even utilized to describe the largest peak among force relationship for the nonstationary fluctuating wind compo-
two distinct peaks shown in typical downburst events (Kwon and nent. Similarly, the force-response relationship is achieved by
Kareem 2009, 2013), which has been confirmed in Solari et al. invoking the structural transfer function (mechanical admittance
(2015) recently that the half-sine shape was close to the inner function) (fourth box). In this manner, the time-frequency response
envelope of full-scale data. spectrum (the last box) is obtained through a product of the force
spectrum and the structural transfer function. Accordingly, a com-
plete wind-force-response relationship for nonstationary winds can
Generalized Wind Loading Chain: Overview be established throughout the proposed chain. Finally, statistics
of extremes of structural response may be estimated by combining
A generalized wind loading chain is proposed in this paper to the time-varying mean and the nonstationary fluctuating responses,
encompass the effects of nonstationary winds, recasting the well- which will also be discussed in a later section. This generalized
established stationary wind loading chain (Davenport 1967) into a wind loading chain of nonstationary fluctuating wind component
nonstationary format involving time-dependent nature. Invoking a may be introduced by substituting different types of time-frequency
nonstationary wind model, a nonstationary response is comprised representations to EPSD (fourth row) or wavelet transform (fifth
of two components, namely time-varying mean and nonstationary row), thus providing a toolbox with multiple options.
fluctuating responses, which correspond to the mean and fluctuat- A comparison between the proposed generalized chain and the
ing responses in a stationary response. Davenport’s stationary chain (lower half of Fig. 1) highlights this
A schematic diagram of the proposed generalized nonstationary key feature of analogy. Both chains are statistical methods based
wind loading chain is shown in Fig. 1. The first row represents the on the random vibration theory, pursuing probabilistic properties
wind-force-response relationship originated from the time-varying of wind-induced response by the wind-force-response relation-
mean component of nonstationary winds. The second and third ship connected through the aerodynamic and structural transfer/
rows represent the centerpiece of the proposed chain, i.e., rewriting admittance functions for wind-force and force-response chains, re-
the fluctuating part of the stationary chain (from the gust to the spectively. Conversely, the major difference between the two is also

© ASCE 04019092-4 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


apparent; the proposed chain introduces the time-dependent component of stationary winds in frequency domain. The EPSD
statistics of wind, force, and response. Particularly, the frequency is defined by accompanying a time variable with the frequency var-
domain formulation of fluctuating component in the Davenport’s iable of PSD, therefore the EPSD-based generalized chain is readily
chain is extended to the time-frequency domain representation, reduced to its stationary counterpart by dropping the time variable,
enabling us to capture the time-dependent effects. The treatment of thus the word generalized.
such effects, although it requires more technical and computational
efforts, is essential for nonstationary winds due to a potential Time-Frequency Gust Spectrum
amplification of response in comparison with stationary case. The EPSD (and also other time-frequency representations) of
Note that the generalized chain can be seamlessly reduced to the nonstationary fluctuating winds may be estimated from full-scale
Davenport’s chain by dropping the time variable, which highlights measurement of nonstationary tropical storms or downbursts in the
the generality of the proposed nonstationary chain. absence of universal models. Assuming that a nonstationary fluc-
In the following, time-frequency modeling of the proposed wind tuating wind is described as a zero-mean evolutionary random
loading chain and its components is discussed in detail. Note that process (Priestley 1965), the auto-EPSD of the fluctuating wind
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

when illustrating the formulation of the generalized chain, this Su~ u~ ðf; tÞ may be expressed as a product of a time-frequency modu-
study primarily concerns tropical storms and downbursts; however, lation function Aðf; tÞ and the PSD Su~ u~ ðf; tÞ of a zero-mean sta-
the proposed chain can also accommodate other types of nonsta- tionary process uðtÞ~
tionary winds events with proper modeling.
Suu ðf; tÞ ¼ jAðf; tÞj2 Su~ u~ ðfÞ ð7Þ
Time-Varying Mean Wind-Force-Response where Aðf; tÞ satisfies slowly-varying conditions (Priestley 1965).
To elaborate, a generic structural system subjected to winds is con- Estimation of EPSD from sample time histories has been
sidered without the loss of generality. The structure is modeled as a extensively investigated, resulting in methods such as windowed
N degree-of-freedom (DOF) dynamic system with its N × N mass, Fourier transform (Priestley 1965, 1966), time-varying correlation
and damping and stiffness matrices (M, C, and K, respectively). function-based method (Benowitz et al. 2015), time-varying AR
The proposed chain aims at evaluating the structural response XðtÞ model (Chen 2006), and wavelet-based methods (Huang and Chen
under the nonstationary wind field UðtÞ. 2009; Spanos and Failla 2004). However, all of these methods face
Similar to the treatment in the stationary wind model, the two an inherent difficulty in practice, i.e., the inability to establish
components of nonstationary wind model [Eq. (1)] are transformed the ensemble average of EPSD because sample time histories of
to the time-varying mean F̄ðtÞ and nonstationary fluctuating dy- nonstationary events are rare and only one sample time history is
namic forces Fb ðtÞ, eventually leading to the time-varying mean usually available (Wang et al. 2014). The lack of ensemble average
may lead to large random errors in the ESPD. Therefore, all these
X̄ðtÞ and nonstationary fluctuating responses xðtÞ of the structure
estimation methods should be used with careful consideration of
(see the first two rows in Fig. 1)
uncertainties, even if special procedures have been incorporated to
FðtÞ ¼ F̄ðtÞ þ Fb ðtÞ; XðtÞ ¼ X̄ðtÞ þ xðtÞ ð4Þ alleviate the random error. In this paper, the estimating method by
Priestley (1965) is briefly described, which temporally smooths the
The time-varying mean response, as depicted in the first row of EPSD estimation by means of a short time window
Fig. 1, is induced by the time-varying mean wind. For the case
of buffeting, the time-varying mean wind is first transferred to ζðf p ; tr Þ ¼ ½gðtr Þ  ½uðtr Þ expð−i2πfp tr Þ ð8Þ
along-wind force
Suu ðf p ; tr Þ ¼ ½Gðtr Þ  ½jζðfp ; tr Þj2  · 2ðΔtÞ3 · 2π ð9Þ
F̄ðtÞ ¼ 0.5ρAT CD ðtÞŪðtÞ2 ð5Þ
where the symbol ∗ = the discrete convolution; tr ¼ ðr − 1ÞΔt and
where ρ = air density; CD ðtÞ = instantaneous aerodynamic drag fp ¼ ðp − 1ÞΔf = discrete time and frequency series with time
force coefficient; and AT = tributary area. Note that if F̄ðtÞ varies step (Δt) and frequency step (Δf), respectively; and gðtÞ and
slowly with time that no transient dynamics is involved, then the GðtÞ = window functions with properly selected forms and lengths,
time-varying mean response X̄ðtÞ may be determined by solving respectively. The corresponding random error and condition to
the following static equilibrium equation achieve an unbiased estimation have been formulated in Priestley
KX̄ðtÞ ¼ F̄ðtÞ ð6Þ (1966), underpinning this method. It is thus rigorous, but also stable
and convenient due to avoiding the solution of delicate equations
at each time instant t. Possible static nonlinearity can be readily associated with other methods (e.g., wavelet-based methods).
dealt with by an iteration procedure (e.g., Cheng et al. 2002; Hu Alternatively, parametric modeling of EPSD of nonstationary
et al. 2013). tropical storm and downburst winds has been attempted based
on limited full-scale data. Observations reported in the literature
(Chen and Letchford 2006; Hu et al. 2013; Huang and Chen 2009;
Nonstationary Fluctuating Wind-Force-Response:
Huang et al. 2015; Lombardo et al. 2014; Wang et al. 2016) have
EPSD-Based Approach
suggested that in many cases, wind spectrum models of stationary
Modeling of the nonstationary fluctuating wind-force-response re- winds may be extended to describe the characteristics of nonsta-
lationship in the time-frequency domain is central to this proposed tionary winds by plugging in a time dimension in order to describe
wind loading chain (the third row in Fig. 1). Various time- the time-dependent turbulence spectral contents [e.g., the evolu-
frequency representations may be employed, e.g., EPSD, wavelet tionary Karman wind spectrum in Hu et al. (2013)]. This type of
transform, Chirplet transform (Wang et al. 2002), Yeh-Wen instan- fully nonstationary model may be further simplified as an ampli-
taneous spectra (Yeh and Wen 1990), response spectrum (Solari tude modulation function.
and De Gaetano 2018), and time-varying autoregressive (AR) The EPSD-based model also allows incorporation of other
model. Among them, the EPSD is a natural extension of the essential characteristics of nonstationary winds. For example, Hu
power spectral density (PSD) used in modeling fluctuating wind et al. (2017a) recently considered the time-dependent boundary of

© ASCE 04019092-5 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


inertial subrange especially in the high-frequency range of EPSD, structure, the instantaneous aerodynamic transfer/admittance func-
as the fundamental frequency of large civil structures may fall tion [χðf; tÞ] may be obtained from incorporating the time-varying
within the range. This improved EPSD model encompasses the mean wind into the aerodynamic transfer/admittance function
conceptual spectral model for tropical storms developed by Li et al. widely used in stationary buffeting analysis. Note that the quasi-
(2015), which has been based on theoretical consideration and veri- stationary assumption may be appropriate when the instantaneous
fied by full-scale measurements. statistical properties of nonstationary winds evolve with time so
After the auto-EPSD of wind speed component at different slowly that transient effects of the fluctuating wind components
locations is established, the cross-EPSD between each pair of on aerodynamic interaction are not significant. Once further obser-
components can be defined analogous to the cross-PSD through vations suggest an advanced model, the formulation of the transient
a coherence function γ u ðfÞ, which is time-invariant because of aerodynamic function can be updated in Eq. (13). Eq. (12) indicates
the constraint by EPSD theory (Mélard and Schutter 1989; Priestley that Fb ðtÞ is an evolutionary vector process as inherited from the
1966) fluctuating wind speeds.
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

Su;ij ðf; tÞ ¼ Su;i ðf; tÞSu;j ðf; tÞγ u;ij ðfÞ Nonstationary Fluctuating Response: Instantaneous
¼ jAu;i ðf; tÞj2 jAu;j ðf; tÞj2 ½Su;i Structural Transfer Function and Time-Frequency
~ ðfÞSu;
~ j ðfÞγ u;ij ðfÞ ð10Þ
Response Spectrum
The time-invariant coherence function γ u ðfÞ relates the coher- Given the EPSD of buffeting force SFb Fb ðf; tÞ in the previous sec-
tion, the EPSD of fluctuating responses Sxx ðf; tÞ may be estimated
ence between PSDs Su;i ~ ðfÞ and Su;j ~ ðfÞ to model the stationary
processes u~ i ðtÞ and u~ j ðtÞ in Priestley’s model [Eq. (7)], while the by solving the following equations of motion in the time-frequency
domain:
nonstationarity is modeled by the time-frequency modulation func-
tions Au;i ðf; tÞ and Au;j ðf; tÞ. Measurements of coherence function
MẍðtÞ þ CẋðtÞ þ KxðtÞ ¼ Fb ðtÞ ð14Þ
of nonstationary winds is rare (e.g., Chen and Letchford 2005). Thus,
the format of coherence function introduced by Davenport for sta-
tionary winds may be utilized for nonstationary winds The solution of the equations of motion for nonstationary pro-
cess has been extensively examined in various fields of researches
 1
½C2z ðzi − zj Þ2 þ C2y ðyi − yj Þ2 2 (e.g., wind, earthquake, and aerospace engineering). The fully-
γ ij ðfÞ ¼ exp −f 1
ð11Þ nonstationary theoretical solution reported in Howell and Lin
2 Ū 0 ½Ū H ðzi Þ þ Ū H ðzj Þ
(1971) was derived by extending the stationary solution in terms
where yi = the lateral coordinates; and Cz and Cy = decay co- of PSD, which can give an accurate solution of Sxx ðf; tÞ, but is
efficients. More full-scale data may suggest better models. For computationally inefficient. To enhance computational efficiency,
example, recently the time-invariant coherence function has been up- approximate closed-form solutions have been proposed with par-
graded to the time-varying counterpart γ ij ðf; tÞ, by modeling ui ðtÞ ticularly designed analytical approximations of EPSD to take
as the summation of multiple time-frequency modulated stationary advantage of separable time and frequency modulation functions
processes (Peng et al. 2018). In terms of the auto- and cross-EPSDs, (Conte and Peng 1996; Li et al. 2016; Li and Kareem 1991). For
the EPSD matrix of multivariate nonstationary winds Suu ðf; tÞ can reference, closed-form solutions of SDOF system concerning sim-
then be established. ple amplitude modulating functions can also be found in Jangid
(2004). In solving the problem, some common issues may arise
Instantaneous Aerodynamic Transfer/Admittance Function from structural characteristics requiring special treatment and have
and Time-Frequency Force Spectrum been addressed, such as nonclassical damping (Chen 2015), non-
Time-dependent properties of nonstationary tropical storms or linearity (Canor et al. 2016; Kareem and Wu 2013), time-dependent
downbursts may cause time-varying characteristics of wind-structure structural parameters (Hu et al. 2013; Zhang et al. 2010), and un-
interactions. Specially designed wind tunnel tests (e.g., Butler et al. certainty (Li and Chen 2009). Conversely, there are still exclusive
2010; Cao et al. 2002; Chay and Letchford 2002; Jesson et al. issues of nonstationary response analysis deserving attention as
2015) have been carried out to study the transient relationship be- well, including sensitivity to initial conditions (Li and Chen 2009)
tween nonstationary wind-induced force and simulated downburst and the lag-time effects (Chen 2008; Hammond 1973; Igusa 1989;
winds (e.g., by multiple-fan wind tunnels or impinging jets) and Langley 1987). In the following, the fully nonstationary solution in
have revealed the nonstationary wind-force relationship different terms of both original and modal coordinates, as well as an approxi-
from its stationary counterpart. mated solution are briefly discussed for general interests. However,
The EPSD of nonstationary wind-induced buffeting force may one may always adopt other methods in the generalized chain if
be expressed as (e.g., Hu et al. 2013, 2017a) special issues discussed emerge.
Fully nonstationary solution can be achieved by the direct
SFb Fb ðf; tÞ ¼ Ξðf; tÞSuu ðf; tÞΞðf; tÞT ð12Þ integration method (Howell and Lin 1971; Li and Chen 2009)

where Ξðf; tÞ can be expressed as Sxx ðf; tÞ ¼ Pðf; tÞSFb Fb ðfÞPðf; tÞH ð15Þ
Ξðf; tÞ ¼ Ts ðtÞχðf; tÞCb ðtÞ ð13Þ Z t
where Ts ðtÞ, χðf; tÞ, and Cb ðtÞ are the matrices of instantaneous Pðf; tÞ ¼ hðt − τ ÞAFb Fb ðf; tÞ expð−i2πfτ Þdτ ð16Þ
0
coordinate transformation coefficients, aerodynamic admittance
function, and aerodynamic coefficients, respectively. Moreover, the where the superscript H denotes conjugate transpose; and hðtÞ =
horizontal/lateral joint acceptance function jJ y ðfÞj2 may also be impulse response function of dynamic system in Eq. (14) deter-
included in Eq. (12) (e.g., Chen 2008; Kwon and Kareem 2009). mined under the impulse loading matrix IðtÞ ¼ diag½δðtÞ and zero
Under the quasi-stationary assumption (Langley 1987), both Ts ðtÞ initial conditions. Although this integration-based method fully
and Cb ðtÞ are coefficients slowly varying with structural motion accounts for the effects of the transient dynamics, it may be com-
due to the time-varying mean wind loading. For a point-like putationally prohibitive for a system with a large amount of

© ASCE 04019092-6 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


DOFs due to the time-consuming integration in Eq. (16). A more solution takes the advantage of multiplication against the time-
computationally efficient solution can be achieved via modal super- consuming time domain integration, thus leading to higher com-
position and the stochastic decomposition (Li and Kareem 1995) or putational efficiency. However, it ignores the transient lag-time
pseudo excitation method (Lin 2004). In this manner, Eq. (14) can effects of structural dynamic transfer function, and thus may over-
be decomposed into modal coordinates estimate response (Chen 2008; Langley 1987). This overestima-
tion decreases with the increase in natural frequencies and/or
~ Ẏ j ðf; tÞ þ KY
Ÿ j ðf; tÞ þ C ~ b;j ðf; tÞ
~ j ðf; tÞ ¼ Q ð17Þ damping ratios of the system, while increases with the increase
in the changing rate of EPSD of nonstationary winds. Because
where Y j ðf; tÞ (j ¼ 1 : : : N S ) = the pseudo modal response, which EPSD of tropical storms usually changes in a lower rate, it is
can be solved using the step-by-step integration in time domain; expected that this quasi-stationary solution would be more accu-
N S = the number of modes; C ~ ¼ ΦT CΦ and K ~ ¼ diag½ð2πf j Þ2  = rate for tropical storm winds than for downburst winds (Langley
the generalized modal damping and stiffness matrices, respectively; 1987).
f j = the natural frequency of the jth mode; Φ = the mode shape
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

matrix normalized with respect to the mass matrix (ΦT MΦ ¼ Extreme Response
~ b;j ðf; tÞ ¼ BQ ;j ðf; tÞei2πft = pseudo modal excitation;
IN S ×N S ); Q Extreme response is of critical concern in structural design and reli-
b
ability analysis, and has been examined for nonstationary random
BQb ;j = the jth column vector of the matrix BQb ¼ ΦT AFb Fb
processes (Corotis and Vanmarcke 1975; Corotis et al. 1972;
ðf; tÞBFb Fb ðfÞ; and BFb Fb ðfÞ = the lower triangular matrix from
Michaelov et al. 2001). In the proposed chain, the probabilistic
Cholesky decomposition of SFb Fb ðfÞ. Using the pseudo response characteristics of the extreme response under nonstationary winds
solved in modal coordinates, the EPSD matrix of xðtÞ is ex- may be assessed by an approximation reported in Hu and Xu
pressed as (2014) incorporating the aforementioned time-varying mean and
X
NS time-frequency representation of fluctuating responses.
Sxx ðf; tÞ ¼ ½ΦY j ðf; tÞ · ½ΦY j ðf; tÞT ð18Þ Considering the nonstationary response XðtÞ ¼ mðtÞ þ xðtÞ
j¼1 [let mðtÞ ¼ X̄ðtÞ for the sake of brevity], the cumulative distribu-
tion function (CDF) of its extremes X max within the time period of
where the superscript * denotes conjugate. The solution obtained
concern (0, T 0 ) may be approximated in terms of the time-varying
through Eqs. (17) and (18) reserves the transient effects almost
mean response [mðtÞ] and the EPSD of fluctuating response
the same with Eqs. (15) and (16) but with a reasonable reduction
[Sxx ðf; tÞ]
in computation. To further improve computational efficiency, an
approximate solution of Eq. (17) can be achieved under the quasi-  Z 
stationary assumption, that is T0
FXmax ða; T 0 Þ ¼ exp − ½ηþ
~
Xð−mÞ
ða; tÞ þ ηþ
XðmÞ ða; tÞdt ð20Þ
0
~ b;j ðf; tÞ
Y~ j ðf; tÞ ¼ HðfÞQ ð19Þ

where HðfÞ ¼ ½K ~ −1 = the steady-state trans-


~ − ð2πfÞ2 þ i2πf C The instantaneous up-crossing rate ηþ
X ða; tÞ is computed by (Hu
fer function of the dynamic system in Eq. (17). This approximate and Xu 2014)

  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
q2X ðtÞ−ρ2XẊ ðtÞ

1 − exp −π a−mðtÞ
σX ðtÞ 1−ρ2XẊ ðtÞ
fϕ½ξ V
ðtÞ þ ξ V
ðtÞΦ½ξ V
ðtÞg
1 σẊ ðtÞ 1
ηVþ
XðmÞ ða; tÞ ¼ 2π σ ðtÞ ½1 − ρX Ẋ ðtÞ ·
2 h  2 i ð21Þ
X exp 12 a−mðtÞ
σX ðtÞ −1

where ϕðxÞ and ΦðxÞ denote the probability density function (PDF) 1996; Hu et al. 2017b; Liang et al. 2007), and many other can-
and CDF of standard Gaussian distribution; qX ðtÞ, ρXẊ ðtÞ, ξ V ðtÞ = didates]. Next, the time histories of nonstationary fluctuating wind
nongeometric spectral parameters; and details can be found in force Fb ðtÞ] can be obtained by passing the wind time histories
Hu and Xu (2014). Once FXmax ða; T 0 Þ is given, the mean, standard through a filter with time-dependent parameters designed to re-
deviation, and percentile of the extreme response X max can be present the instantaneous aerodynamic transfer function. When
readily determined. the aforementioned quasi-stationary assumption retains, a reason-
able design procedure of this filter carries the rational function
Time-Domain Realization of Generalized Wind Loading approximation (Chen and Kareem 2002) separately for each time
Chain instant. Applying the superposition of the time-varying mean F̄ðtÞ
The generalized wind loading chain is proposed in terms of time- and Fb ðtÞ on the target structure, the sample time histories of
frequency representation of statistical properties of nonstationary structural response can be obtained through the structural dy-
winds, force, and response, but it can also be realized in the time namics problem in Eq. (14) using time-integration methods
domain in a way similar to the time domain representation of (e.g., Newmark method). Finally, the statistics of nonstationary
Davenport’s stationary chain. The procedure of time-domain re- response may be computed by estimation over an ensemble aver-
alization may include the following steps. Sample time histories age of the sample response time histories. This time-domain reali-
of nonstationary fluctuating winds uðtÞ targeted at the time- zation procedure has been applied to calculate the response of
frequency gust spectrum can be generated using nonstationary long-span bridges to nonstationary winds (e.g., Chen 2015; Hao
winds simulation methods [e.g., spectral representation (Deodatis and Wu 2017; Hu et al. 2017a).

© ASCE 04019092-7 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


Nonstationary Fluctuating Wind-Force-Response: E½W jx ðW jx ÞT 
Wavelet-Based Approach 2 3
Sxx ðfj ; t1 Þ 0
In lieu of the EPSD-based approach discussed in the previous sec- 6 7
6 Sxx ðfj ; tk Þ 7
tions, an alternative formulation to deal with nonstationary fluctu- ðn − mÞ 6
6
7
7
¼ ..
ating wind components using wavelet transform is presented in this T0 6 6 .
7
7
paper. Unlike EPSD that compromises frequency resolution due to 4 5
the fixed time window size, wavelet transform adopts flexible time 0 Sxx ðfj ; tN T Þ
window adaptive to frequency resolution requisite and thus can
achieve the best time-frequency resolution allowed by Heisenberg’s ð25Þ
uncertainty principle for different frequency bands (e.g., Kijewski-
Correa and Kareem 2003). Therefore, wavelet transform has be- A similar relationship also exists for the buffeting force W jFb .
come a popular time-frequency analysis tool, whereas the wavelet
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

scale can be directly related to the frequency in the context of


Fourier transform. Numerical Example
To represent the generalized loading chain using the wavelet-
based approach, the periodic generalized harmonic wavelet In this section, the proposed wind loading chain is demonstrated
(PGHW) is employed in this study, in which the wavelet function using an example building structure subject to nonstationary
is defined as (Kong and Li 2015; Spanos et al. 2016) winds. A high-rise building of 120 m is modeled as a 3-degree-
of-freedom (3DOF) system (Fig. 5). The natural frequencies of
the system are 0.6, 1.28, and 1.94 Hz, and the modal damping
nj −1
X −iq2πΔf ratios are assumed to be 0.01 for all three modes. Two nonsta-
1 kT
0
ψG;per
ðmj ;nj Þ;k ðtÞ ¼ e t−n−m
ð22Þ tionary winds are utilized in this example: (1) a downburst event
ðn − mÞ q¼m modeled by analytical time-varying mean function (Chen 2008),
j
analytical profile (Kwon and Kareem 2009), and amplitude-
modulated Karman spectrum; and (2) a tropical storm event
so that wavelet coefficients can be obtained by the wavelet whose detailed information can be found in Hu et al. (2017a).
transform The time period of concern is 600 s for the downburst and
3,600 s for the tropical storm. The vertical profile, time-varying
mean wind, and EPSD of nonstationary fluctuating components
Z for the two nonstationary winds are displayed in Figs. 2, 3(a),
n−m T0
W G;per
ðmj ;nj Þ;k ðxÞ ¼ xðtÞψ̄G;per
ðmj ;nj Þ;k ðtÞdt ð23Þ and 4(a), and Figs. 6(a) and 7(a), respectively. The coherence
T0 0 function in Eq. (11) with the decay parameter Cz ¼ 8 is used
for both events. Aerodynamic parameters include drag force co-
efficient Cd ¼ 1.3, and tributary area AT ¼ 800 m2 for Nodes 1
where T 0 ¼ N 0 Δt = the time period of concern (with reference and 2, and 400 m2 for Node 3.
to Section “Time-Varying Mean Wind”); j ¼ 1; : : : ; N f , N f ¼ In the following, the numerical results are presented using
N 0 =ðn − mÞ = the jth wavelet scale and the number of scales; time-frequency representations of the proposed chain such as the
and k; l ¼ 0; : : : ; N T − 1, N T ¼ ðn − mÞ ¼ nj − mj = the wave- EPSD-based and wavelet-based approaches to evaluate structural
let time shift and the number of shifts. The equations of motion in response. The quasi-stationary assumption is applied to the EPSD
Eq. (14) may be solved in terms of the PGHW, resulting in the model of wind speeds, aerodynamic transfer/admittance function,
wavelet scalogram of response and solution of dynamic equations in Eq. (19). Figs. 3(b) and 4(b)
show the time-varying mean wind force, whereas Figs. 3(c)
and 4(c) show the time-varying mean wind response (as indicated
T0 2 by the dashed lines). Fig. 6 (downburst) and Fig. 7 (tropical storm)
E½W jx ðW jx ÞT  ¼ ½ðAj Þ−1 fE½W jFb ðW jFb ÞT g½ðAj Þ−1 T illustrate the proposed generalized chain and its subcomponents
n−m
ð24Þ

ð2Þ
where Aj is a matrix comprised of block matrix Ajk;l ¼ Cj;k;l M þ
ð1Þ ð0Þ ð°Þ
Cj;k;l C þ Cj;k;l K; Cj;k;l = wavelet connection coefficients; W jx and
W jFb = vectors of wavelet coefficients of nonstationary fluctuating
response and buffeting force in Eq. (14), respectively; and de-
tailed information can be found in Kong and Li (2015) and Spanos
et al. (2016). Eq. (24) implies a wavelet-based representation
of second-order structural dynamics system, therefore by analogy
it may be extended to higher order linear systems. Accordingly,
the aerodynamics transfer relationship between nonstationary
fluctuating winds and attendant force may also be represented
in terms of wavelet coefficients, when this relationship can be
approximated by a high-order linear system with sufficient accu-
Fig. 5. Illustration of the 3-DOF system used in the numerical
racy. The wavelet scalogram of buffeting response may be trans-
example.
formed to EPSD by

© ASCE 04019092-8 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


10 -7
8

Transfer function
6

0.5 1 1.5
Frequency (Hz)

(a) (b) (c) (d) (e)


Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

(f) (g) (h) (i) (j)

Fig. 6. Illustration of generalized wind loading chain of nonstationary fluctuating wind of downburst event and its components at Node 3:
(a) EPSD of fluctuating wind; (b) instantaneous aerodynamic transfer/admittance function; (c) EPSD of buffeting force; (d) structural transfer
function; (e) EPSD of fluctuating response; (f) wavelet scalogram of fluctuating wind; (g) wavelet-based instantaneous aerodynamic transfer/
admittance function; (h) wavelet scalogram of buffeting force; (i) wavelet-based structural transfer function; and (j) wavelet scalogram of
fluctuating response (in contours, brighter color: higher values, darker color: lower values).

10-7
8

Transfer function 6

0.5 1 1.5
Frequency (Hz)

(a) (b) (c) (d) (e)

(f) (g) (h) (i) (j)

Fig. 7. Illustration of generalized wind loading chain of nonstationary fluctuating wind of tropical storm event and its components at Node 3:
(a) EPSD of fluctuating wind; (b) instantaneous aerodynamic transfer/admittance function; (c) EPSD of buffeting force; (d) structural transfer
function; (e) EPSD of fluctuating response; (f) wavelet scalogram of fluctuating wind; (g) wavelet-based instantaneous aerodynamic transfer/
admittance function; (h) wavelet scalogram of buffeting force; (i) wavelet-based structural transfer function; and (j) wavelet scalogram of fluctuating
response (in contours, brighter color: higher values, darker color: lower values).

of nonstationary fluctuating winds, in which the first row of each method is utilized to simulation nonstationary winds speed time
figure from the left to the right displays the EPSD-based ap- histories. The simulated sample time histories of wind speed,
proach such as the EPSD of gust spectrum, aerodynamic transfer/ wind-induced force, and resulted displacement response at Node 3
admittance function, EPSD of force spectrum, structural transfer are shown in Figs. 3 and 4 together with their time-varying mean
function, and EPSD of displacement response. In addition, the values, respectively.
second row of each figure displays the same wind loading chain Using the time-varying mean and the time-frequency represen-
but represented by the wavelet-based approach. Finally, the time tation of fluctuating responses, the extreme displacement re-
domain realization of the proposed chain as discussed earlier is sponses are estimated from the method discussed earlier and
also demonstrated in Figs. 3 and 4. The spectral representation listed in Table 1, in comparison with the extreme responses

© ASCE 04019092-9 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


Table 1. Comparison of mean extreme displacement response (m) A generalized version of the gust front factor was later intro-
Downburst Tropical storm duced by Kwon and Kareem (2013), named generalized gust front
factor, which was intended not only to encapsulate dynamic load
Node EPSD Wavelet Simulationa EPSD Wavelet Simulation
effects associated with nonstationary winds independent of any
1 0.00494 0.00513 0.00526 0.00648 0.00669 0.00697 reference design code/standard but also to highlight other general
2 0.00890 0.00920 0.00955 0.0121 0.0126 0.0131 features like those found in the conventional gust loading factor
3 0.0119 0.0122 0.0127 0.0173 0.0178 0.0186 scheme. More detailed information concerning the GFF/G-GFF
a
Simulation in time domain. can be found in Kwon and Kareem (2009, 2013).

obtained in the time-domain with 1,000 samples of simulations. It Concluding Remarks


is observed that the results obtained from different approaches
This paper deals with the changing dynamic of wind field charac-
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

match closely, which demonstrates the efficacy of the proposed


generalized chain. teristics to address the challenges posed by nonstationary nature of
extreme wind events such as tropical storms and downbursts on the
structural response. To this end, a generalized wind loading chain
for evaluating structural response under nonstationary winds is pro-
Codification of Nonstationary Wind Effects on
posed in this study by recasting the Davenport’s chain into a non-
Structures: Gust Front Factor Approach
stationary format in terms of time-frequency representation and
In this paper, a preliminary examination of the feasibility for nonstationary models. Using both the EPSD and wavelet transform
codification of nonstationary wind effects on structures is made. as the representation of nonstationary fluctuating wind component,
This kind of research is rather limited except for the following two sets of expressions of the proposed wind loading chain are in-
two approaches: the gust front factor framework (Kwon and troduced in detail and demonstrated by a numerical example. The
Kareem 2009, 2013) and the thunderstorm response spectrum tech- solution of nonstationary wind-induced response in the time-
nique (Solari and De Gaetano 2018). From the two approaches, it is frequency domain using the quasi-stationary assumption enables
observed that the extreme nonstationary response determined by high computational efficiency, but sacrifices the accuracy of cap-
the generalized wind loading chain may be conveniently meshed turing transient effects. Realization of the proposed chain in the
with the concept of the gust front factor (Kwon and Kareem time domain is also discussed. It should also be emphasized that
2009). In particular, the GFF framework has been aimed at estimat- despite the versatility of the time domain approach (e.g., nonlinear
ing the design load induced by gust-front winds (originated from problems exclusively achievable), the time-frequency represented
thunderstorm/downburst), i.e., equivalent static wind load (ESWL), chain with its computational advantage is effective not only for
which most wind codes and standards for stationary winds suggest. the efficient estimation of extreme response, but also for imple-
Accordingly, the fusion of the generalized chain and the GFF may menting universal models of nonstationary winds effects akin to
enable to rapidly shift the current design paradigm from stationary the widely used stationary counterpart.
to nonstationary winds. The generalized wind loading chain aims at facilitating seam-
The GFF is defined as the relationship of the structural re- less response analysis under nonstationary as well as stationary
sponses under nonstationary and stationary winds wind events. More importantly, it offers a time-frequency domain
framework to make a case for a shift in the current design para-
max ðz; T 0 Þ
xGen digm that is attentive to the nature of nonstationary winds. An
GG−F ¼ ð26Þ immediate application of the proposed wind loading chain is a
max½xB−L ðz; tÞ
possible codification of nonstationary wind effects on structures
where xB−L ðz; tÞ = displacement response according to Daven- in the context of a gust front factor and a generalized gust front
port’s stationary wind loading chain prescribed in most wind design factor frameworks to estimate design wind loading. These are
codes and standards, thus one can easily calculate the response briefly examined in this paper but will be detailed in a future pub-
max ðz; T 0 Þ = mean or
through the code/standard of interest; and xGen lication. Note that the models for the nonstationary winds used in
percentile of extreme displacement response under nonstationary this paper are for the sake of establishing the chain, which can
winds, which can be computed by the proposed generalized wind conveniently incorporate more reliable models when these be-
loading chain. For the case of ASCE 7 (ASCE 2016) standard, the come available. Indeed, additional research efforts are needed
GFF leads to a design wind loading FDesign using the following to develop robust, reliable, and well-accepted models for nonsta-
relationship: tionary winds.
Finally, at this juncture it is inspiring to see the formulation of
FDesign ¼ FASCE7 · K z;G−F · GG−F ð27Þ the proposed chain that promises to provide a roadmap for further
studies on the effects of nonstationary winds as well as the wind
where FASCE7 denotes the design wind loading under stationary resistant design of structures under nonstationary tropical storm or
boundary layer winds specified in ASCE 7, which involves the gust downburst wind events.
loading/effect factor; and K z;G−F accounts for the effect of nonsta-
tionary wind profile (Kwon and Kareem 2009). Note at this junc-
ture that FASCE7 is usually based on the 3-s gust wind speed and Acknowledgments
therefore the transformation between different averaging times (3 s
and T 1 in this paper) may be necessary (Kwon and Kareem 2009). The authors wish to acknowledge the financial supports from the
For reference, the GFF framework has also been implemented as a US National Science Foundation (CMMI 1462076), and the
web-enabled module to facilitate expeditious utilization in design National Natural Science Foundation of China (No. 51308244).
practice, which is available online (NatHaz Modeling Laboratory Any opinions and concluding remarks presented in this paper
2007). are entirely those of the authors.

© ASCE 04019092-10 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


References Chen, X. 2015. “Analysis of multimode coupled buffeting response of long-
span bridges to nonstationary winds with force parameters from station-
Abd-Elaal, E. S., J. E. Mills, and X. Ma. 2014. “Empirical models for ary wind.” J. Struct. Eng. 141 (4): 04014131. https://doi.org/10.1061
predicting unsteady-state downburst wind speeds.” J. Wind Eng. Ind. /(ASCE)ST.1943-541X.0001078.
Aerodyn. 129 (Jun): 49–63. https://doi.org/10.1016/j.jweia.2014 Chen, X., and A. Kareem. 2002. “Advances in modeling of aerodynamic
.03.011. forces on bridge decks.” J. Eng. Mech. 128 (11): 1193–1205. https://doi
Aboshosha, H., A. Elawady, A. El Ansary, and A. El Damatty. 2016. .org/10.1061/(ASCE)0733-9399(2002)128:11(1193).
“Review on dynamic and quasi-static buffeting response of transmis- Cheng, J., J. J. Jiang, R. C. Xiao, and H. F. Xiang. 2002. “Nonlinear aero-
sion lines under synoptic and non-synoptic winds.” Eng. Struct. static stability analysis of Jiang Yin suspension bridge.” Eng. Struct.
112 (Apr): 23–46. https://doi.org/10.1016/j.engstruct.2016.01.003. 24 (6): 773–781. https://doi.org/10.1016/S0141-0296(02)00006-8.
Aboshosha, H., T. Mara, and P. Case. 2017. “New framework for estimating Choi, E. C. C. 2000. “Wind characteristics of tropical thunderstorms.”
thunderstorm design speed.” In Proc., 13th Americas Conf. on Wind J. Wind Eng. Ind. Aerodyn. 84 (2): 215–226. https://doi.org/10.1016
Engineering (13ACWE). Gainesville, FL: Univ. of Florida. /S0167-6105(99)00054-9.
ASCE. 2016. Minimum design loads for buildings and other structures. Conte, J. P., and B. F. Peng. 1996. “An explicit closed-form solution for
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

ASCE 7. Reston, VA: ASCE. linear systems subjected to nonstationary random excitation.” Probab.
Bendat, J. S., and A. G. Piersol. 2010. Random data: Analysis and Eng. Mech. 11 (1): 37–50. https://doi.org/10.1016/0266-8920(95)
measurement procedures. New York: Wiley. 00026-7.
Benowitz, B. A., M. D. Shields, and G. Deodatis. 2015. “Determining evolu- Corotis, R. B., and E. H. Vanmarcke. 1975. “Time-dependent spectral
tionary spectra from non-stationary autocorrelation functions.” Probab. content of system response.” J. Eng. Mech. Div. 101 (5): 623–636.
Eng. Mech. 41 (Jul): 73–88. https://doi.org/10.1016/j.probengmech Corotis, R. B., E. H. Vanmarcke, and C. A. Cornell. 1972. “First passage of
.2015.06.004. nonstationary random processes.” J. Eng. Mech. Div. 98 (2): 401–414.
Butler, K., S. Cao, A. Kareem, Y. Tamura, and S. Ozono. 2010. “Surface Davenport, A. G. 1967. “Gust loading factors.” J. Struct. Eng. 93 (3):
pressure and wind load characteristics on prisms immersed in a simu- 11–34.
lated transient gust front flow field.” J. Wind Eng. Ind. Aerodyn. Deodatis, G. 1996. “Non-stationary stochastic vector processes: Seismic
98 (6–7): 299–316. https://doi.org/10.1016/j.jweia.2009.11.003. ground motion applications.” Probab. Eng. Mech. 11 (3): 149–167.
Canor, T., L. Caracoglia, and V. Denoël. 2016. “Perturbation methods in https://doi.org/10.1016/0266-8920(96)00007-0.
evolutionary spectral analysis for linear dynamics and equivalent stat- Emanuel, K., S. Ravela, E. Vivant, and C. Risi. 2006. “A statistical deter-
istical linearization.” Probab. Eng. Mech. 46 (Oct): 1–17. https://doi.org ministic approach to hurricane risk assessment.” Bull. Am. Meteorol.
/10.1016/j.probengmech.2016.07.001. Soc. 87 (3): 299–314. https://doi.org/10.1175/BAMS-87-3-299.
Cao, S., A. Nishi, H. Kikugawa, and Y. Matsuda. 2002. “Reproduction of Fujita, T. T. 1981. “Tornadoes and downbursts in the context of generalized
wind velocity history in a multiple fan wind tunnel.” J. Wind Eng. Ind. planetary scales.” J. Atmos. Sci. 38 (8): 1511–1534. https://doi.org/10
Aerodyn. 90 (12): 1719–1729. https://doi.org/10.1016/S0167-6105(02) .1175/1520-0469(1981)038<1511:TADITC>2.0.CO;2.
00282-9. Gunter, W. S., and J. L. Schroeder. 2015. “High-resolution full-scale mea-
Chay, M. T., F. Albermani, and R. Wilson. 2006. “Numerical and analytical surements of thunderstorm outflow winds.” J. Wind Eng. Ind. Aerodyn.
simulation of downburst wind loads.” Eng. Struct. 28 (2): 240–254. 138 (Mar): 13–26. https://doi.org/10.1016/j.jweia.2014.12.005.
https://doi.org/10.1016/j.engstruct.2005.07.007. Gurley, K., and A. Kareem. 1999. “Applications of wavelet transforms in
Chay, M. T., and C. W. Letchford. 2002. “Pressure distributions on a cube earthquake, wind and ocean engineering.” Eng. Struct. 21 (2): 149–167.
in a simulated thunderstorm downburst. Part A: Stationary downburst https://doi.org/10.1016/S0141-0296(97)00139-9.
observations.” J. Wind Eng. Ind. Aerodyn. 90 (7): 711–732. https://doi Hammond, J. K. 1973. “Evolutionary spectra in random vibrations.” J. R.
.org/10.1016/S0167-6105(02)00158-7. Stat. Soc. Ser. B (Methodol.) 35 (2): 167–188. https://doi.org/10.1111/j
Chen, J., M. C. H. Hui, and Y. L. Xu. 2007. “A comparative study of sta- .2517-6161.1973.tb00950.x.
tionary and non-stationary wind models using field measurements.” Hao, J., and T. Wu. 2017. “Nonsynoptic wind-induced transient effects on
Boundary Layer Meteorol. 122 (1): 105–121. https://doi.org/10.1007 linear bridge aerodynamics.” J. Eng. Mech. 143 (9): 04017092. https://
/s10546-006-9085-1. doi.org/10.1061/(ASCE)EM.1943-7889.0001313.
Chen, L. 2006. “Vector time-varying autoregressive (TVAR) models and Hjelmfelt, M. R. 1988. “Structure and life cycle of microburst outflows
their application to downburst wind speeds.” Ph.D. dissertation, Dept. observed in Colorado.” J. Appl. Meteorol. 27 (8): 900–927. https://doi
of Civil and Environmental Engineering, Texas Tech Univ. .org/10.1175/1520-0450(1988)027<0900:SALCOM>2.0.CO;2.
Chen, L., and C. W. Letchford. 2004a. “A deterministic-stochastic hybrid Holmes, J. D., and S. E. Oliver. 2000. “An empirical model of a downburst.”
model of downbursts and its impact on a cantilevered structure.” Eng. Struct. 22 (9): 1167–1172. https://doi.org/10.1016/S0141-0296
Eng. Struct. 26 (5): 619–629. https://doi.org/10.1016/j.engstruct.2003 (99)00058-9.
.12.009. Howell, L. J., and Y. K. Lin. 1971. “Response of flight vehicles to nonsta-
Chen, L., and C. W. Letchford. 2004b. “Parametric study on the along-wind tionary atmospheric turbulence.” AIAA J. 9 (11): 2201–2207. https://doi
response of the CAARC building to downbursts in the time domain.” .org/10.2514/3.50026.
J. Wind Eng. Ind. Aerodyn. 92 (9): 703–724. https://doi.org/10.1016/j Hu, L., and Y. L. Xu. 2014. “Extreme value of typhoon-induced non-
.jweia.2004.03.001. stationary buffeting response of long-span bridges.” Probab. Eng.
Chen, L., and C. W. Letchford. 2005. “Proper orthogonal decomposition of Mech. 36 (Apr): 19–27. https://doi.org/10.1016/j.probengmech.2014
two vertical profiles of full-scale nonstationary downburst wind speeds .02.002.
[lzcl].” J. Wind Eng. Ind. Aerodyn. 93 (3): 187–216. https://doi.org/10 Hu, L., Y. L. Xu, and W. F. Huang. 2013. “Typhoon-induced non-stationary
.1016/j.jweia.2004.11.004. buffeting response of long-span bridges in complex terrain.” Eng.
Chen, L., and C. W. Letchford. 2006. “Multi-scale correlation analyses of Struct. 57 (Dec): 406–415. https://doi.org/10.1016/j.engstruct.2013
two lateral profiles of full-scale downburst wind speeds.” J. Wind Eng. .09.044.
Ind. Aerodyn. 94 (9): 675–696. https://doi.org/10.1016/j.jweia.2006 Hu, L., Y. L. Xu, Q. Zhu, A. Guo, and A. Kareem. 2017a. “Tropical storm-
.01.021. induced buffeting response of long-span bridges: Enhanced nonstation-
Chen, L., and C. W. Letchford. 2007. “Numerical simulation of extreme ary buffeting force model.” J. Struct. Eng. 143 (6): 04017027. https://
winds from thunderstorm downbursts.” J. Wind Eng. Ind. Aerodyn. doi.org/10.1061/(ASCE)ST.1943-541X.0001745.
95 (9–11): 977–990. https://doi.org/10.1016/j.jweia.2007.01.021. Hu, L., Z. Xu, Y. L. Xu, L. Li, and A. Kareem. 2017b. “Error analysis of
Chen, X. 2008. “Analysis of alongwind tall building response to transient spatially varying seismic ground motion simulation by spectral repre-
nonstationary winds.” J. Struct. Eng. 134 (5): 782–791. https://doi.org sentation method.” J. Eng. Mech. 143 (9): 04017083. https://doi.org/10
/10.1061/(ASCE)0733-9445(2008)134:5(782). .1061/(ASCE)EM.1943-7889.0001282.

© ASCE 04019092-11 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


Huang, G., and X. Chen. 2009. “Wavelets-based estimation of multivariate Li, J., and J. Chen. 2009. Stochastic dynamics of structures. Singapore:
evolutionary spectra and its application to nonstationary downburst Wiley.
winds.” Eng. Struct. 31 (4): 976–989. https://doi.org/10.1016/j.engstruct Li, L., A. Kareem, J. Hunt, Y. Xiao, C. Zhou, and L. Song. 2015.
.2008.12.010. “Turbulence spectra for boundary-layer winds in tropical cyclones:
Huang, G., Y. Su, A. Kareem, and H. Liao. 2016. “Time-frequency analysis A conceptual framework and field measurements at coastlines.” Boun-
of nonstationary process based on multivariate empirical mode decom- dary Layer Meteorol. 154 (2): 243–263. https://doi.org/10.1007/s10546
position.” J. Eng. Mech. 142 (1): 04015065. https://doi.org/10.1061 -014-9974-7.
/(ASCE)EM.1943-7889.0000975. Li, Y., J. P. Conte, and M. Barbato. 2016. “Influence of time-varying
Huang, G., H. Zheng, Y. Xu, and Y. Li. 2015. “Spectrum models for non- frequency content in earthquake ground motions on seismic response
stationary extreme winds.” J. Struct. Eng. 141 (10): 04015010. https:// of linear elastic systems.” Earthquake Eng. Struct. Dyn. 45 (8):
doi.org/10.1061/(ASCE)ST.1943-541X.0001257. 1271–1291. https://doi.org/10.1002/eqe.2707.
Huang, W. F., and Y. L. Xu. 2013. “Prediction of typhoon design wind Li, Y. S., and A. Kareem. 1991. “Simulation of multivariate nonstationary
speed and profile over complex terrain.” Struct. Eng. Mech. 45 (1): random-processes by FFT.” J. Eng. Mech. 117 (5): 1037–1058. https://
1–18. https://doi.org/10.12989/sem.2013.45.1.001. doi.org/10.1061/(ASCE)0733-9399(1991)117:5(1037).
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

Igusa, T. 1989. “Characteristics of response to nonstationary white noise: Li, Y. S., and A. Kareem. 1995. “Stochastic decomposition and application
Theory.” J. Eng. Mech. 115 (9): 1904–1918. https://doi.org/10.1061 to probabilistic dynamics.” J. Eng. Mech. 121 (1): 162–174. https://doi
/(ASCE)0733-9399(1989)115:9(1904). .org/10.1061/(ASCE)0733-9399(1995)121:1(162).
Isyumov, N. 2012. “Alan G. Davenport’s mark on wind engineering.” Liang, J., S. R. Chaudhuri, and M. Shinozuka. 2007. “Simulation of non-
J. Wind Eng. Ind. Aerodyn. 104–106 (May–Jul): 12–24. https://doi stationary stochastic processes by spectral representation.” J. Eng.
.org/10.1016/j.jweia.2012.02.007. Mech. 133 (6): 616–627. https://doi.org/10.1061/(ASCE)0733-9399
Jangid, R. S. 2004. “Response of SDOF system to non-stationary (2007)133:6(616).
earthquake excitation.” Earthquake Eng. Struct. Dyn. 33 (15): 1417–1428. Lin, J. H. 2004. Pseudo excitation method in random vibration. Beijing:
https://doi.org/10.1002/eqe.409. Science Press.
Jesson, M., M. Sterling, C. Letchford, and M. Haines. 2015. “Aerodynamic Lombardo, F. T., D. A. Smith, J. L. Schroeder, and K. C. Mehta. 2014.
forces on generic buildings subject to transient, downburst-type winds.” “Thunderstorm characteristics of importance to wind engineering.”
J. Wind Eng. Ind. Aerodyn. 137 (Feb): 58–68. https://doi.org/10.1016/j J. Wind Eng. Ind. Aerodyn. 125 (Feb): 121–132. https://doi.org/10
.jweia.2014.12.003. .1016/j.jweia.2013.12.004.
Kareem, A. 2009. “The changing dynamics of aerodynamics: New fron- McCullough, M., D. K. Kwon, A. Kareem, and L. Wang. 2014. “Efficacy
tiers.” In Proc., 7th Asia–Pacific Conf. on Wind Engineering of averaging interval for nonstationary winds.” J. Eng. Mech. 140 (1):
(APCWQ-VII). Taiwan, Republic of China: Tamkang Univ. 1–19. https://doi.org/10.1061/(ASCE)EM.1943-7889.0000641.
Kareem, A., and T. Kijewski. 2002. “Time-frequency analysis of wind Mélard, G., and A. H. D. Schutter. 1989. “Contributions to evolutionary
effects on structures.” J. Wind Eng. Ind. Aerodyn. 90 (12–15): spectral theory.” J. Time Ser. Anal. 10 (1): 41–63. https://doi.org/10
1435–1452. https://doi.org/10.1016/S0167-6105(02)00263-5. .1111/j.1467-9892.1989.tb00014.x.
Kareem, A., T. Kijewski, and C. E. Smith. 1999. “Analysis and perfor- Michaelov, G., L. D. Lutes, and S. Sarkani. 2001. “Extreme value of re-
mance of offshore platforms in hurricanes.” Wind Struct. Int. J. sponse to nonstationary excitation.” J. Eng. Mech. 127 (4): 352–363.
2 (1): 1–23. https://doi.org/10.12989/was.1999.2.1.001. https://doi.org/10.1061/(ASCE)0733-9399(2001)127:4(352).
Kareem, A., and T. Wu. 2013. “Wind-induced effects on bluff bodies in NatHaz Modeling Laboratory. 2007. “NatHaz gust-front factor.” Accessed
turbulent flows: Nonstationary, non-Gaussian and nonlinear features.” August 14, 2018. http://gff.ce.nd.edu.
J. Wind Eng. Ind. Aerodyn. 122 (Nov): 21–37. https://doi.org/10.1016/j Peng, L., G. Huang, X. Chen, and Q. Yang. 2018. “Evolutionary spectra-
.jweia.2013.06.002. based time-varying coherence function and application in structural re-
Kawai, H. 2000. “Response of structure during a typhoonwind.” In Proc., sponse analysis to downburst winds.” J. Struct. Eng. 144 (7): 04018078.
8th ASCE Specialty Conf. on Probabilistic Mechanics and Structural https://doi.org/10.1061/(ASCE)ST.1943-541X.0002066.
Reliability. Reston, VA: ASCE. Priestley, M. B. 1965. “Evolutionary spectra and non-stationary processes.”
Kijewski-Correa, T., and A. Kareem. 2003. “Wavelet transforms for system J. R. Stat. Soc. Ser. B Stat. (Methodol.) 27 (2): 204–237. https://doi.org
identification in civil engineering.” Comput. Aided Civ. Infrastruct. Eng. /10.1111/j.2517-6161.1965.tb01488.x.
18 (5): 339–355. https://doi.org/10.1111/1467-8667.t01-1-00312. Priestley, M. B. 1966. “Design relations for non-stationary processes.” J. R.
Kim, J., and H. Hangan. 2007. “Numerical simulations of impinging jets Stat. Soc. Ser. B Stat. (Methodol.) 28 (1): 228–240.
with application to downbursts.” J. Wind Eng. Ind. Aerodyn. 95 (4): Solari, G., and P. De Gaetano. 2018. “Dynamic response of structures to
279–298. https://doi.org/10.1016/j.jweia.2006.07.002. thunderstorm outflows: Response spectrum technique vs time-domain
Kong, F., and J. Li. 2015. “Wavelet-expansion-based stochastic response of analysis.” Eng. Struct. 176 (Dec): 188–207. https://doi.org/10.1016/j
chain-like MDOF structures.” J. Sound Vib. 359 (Dec): 136–153. .engstruct.2018.08.062.
https://doi.org/10.1016/j.jsv.2015.09.011. Solari, G., P. De Gaetano, and M. P. Repetto. 2015. “Thunderstorm
Kwon, D. K., and A. Kareem. 2009. “Gust-front factor: A new framework response spectrum: Fundamentals and case study.” J. Wind Eng.
for wind load effects on structures.” J. Struct. Eng. 135 (6): 717–732. Ind. Aerodyn. 143 (Aug): 62–77. https://doi.org/10.1016/j.jweia
https://doi.org/10.1061/(ASCE)0733-9445(2009)135:6(717). .2015.04.009.
Kwon, D. K., and A. Kareem. 2013. “Generalized gust-front factor: A com- Song, L. L., W. C. Chen, B. L. Wang, S. Q. Zhi, and A. J. Liu. 2016. “Char-
putational framework for wind load effects.” Eng. Struct. 48 (Mar): acteristics of wind profiles in the landfalling typhoon boundary layer.”
635–644. https://doi.org/10.1016/j.engstruct.2012.12.024. J. Wind Eng. Ind. Aerodyn. 149 (Feb): 77–88. https://doi.org/10.1016/j
Langley, R. S. 1987. “On quasi-stationary approximations to non-stationary .jweia.2015.11.008.
random vibration.” J. Sound. Vib. 113 (2): 365–375. https://doi.org/10 Spanos, P. D., and G. Failla. 2004. “Evolutionary spectra estimation using
.1016/S0022-460X(87)80222-5. wavelets.” J. Eng. Mech. 130 (8): 952–960. https://doi.org/10.1061
Letchford, C. W., and M. T. Chay. 2002. “Pressure distributions on a cube /(ASCE)0733-9399(2004)130:8(952).
in a simulated thunderstorm downburst. Part B: Moving downburst ob- Spanos, P. D., and G. Failla. 2005. “Wavelets: Theoretical concepts and
servations.” J. Wind Eng. Ind. Aerodyn. 90 (7): 733–753. https://doi.org vibrations related applications.” Shock Vib. Digest 37 (5): 359–375.
/10.1016/S0167-6105(02)00163-0. https://doi.org/10.1177/0583102405055441.
Li, C., Q. S. Li, Y. Q. Xiao, and J. P. Ou. 2012. “A revised empirical model Spanos, P. D., F. Kong, J. Li, and I. A. Kougioumtzoglou. 2016. “Harmonic
and CFD simulations for 3D axisymmetric steady-state flows of down- wavelets based excitation-response relationships for linear systems:
bursts and impinging jets.” J. Wind Eng. Ind. Aerodyn. 102 (Mar): A critical perspective.” Probab. Eng. Mech. 44 (Apr): 163–173. https://
48–60. https://doi.org/10.1016/j.jweia.2011.12.004. doi.org/10.1016/j.probengmech.2015.09.021.

© ASCE 04019092-12 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092


Su, Y., G. Huang, and Y.-l. Xu. 2015. “Derivation of time-varying mean Wang, L., and A. Kareem. 2004. “Modeling of non-stationary winds in
for non-stationary downburst winds.” J. Wind Eng. Ind. Aerodyn. gust-fronts.” In Proc., 9th ASCE Joint Specialty Conf. on Probabilistic
141 (Jun): 39–48. https://doi.org/10.1016/j.jweia.2015.02.008. Mechanics and Structural Reliability. CD-ROM. Reston, VA: ASCE.
Vickery, P. J., F. J. Masters, M. D. Powell, and D. Wadhera. 2009a. Wang, L., M. McCullough, and A. Kareem. 2014. “Modeling and simu-
“Hurricane hazard modeling: The past, present, and future.” J. Wind lation of nonstationary processes utilizing wavelet and Hilbert trans-
Eng. Ind. Aerodyn. 97 (7–8): 392–405. https://doi.org/10.1016/j.jweia forms.” J. Eng. Mech. 140 (2): 345–360. https://doi.org/10.1061
.2009.05.005. /(ASCE)EM.1943-7889.0000666.
Vickery, P. J., D. Wadhera, M. D. Powell, and Y. Chen. 2009b. “A hurricane Wood, G. S., K. C. S. Kwok, N. A. Motteram, and D. F. Fletcher. 2001.
boundary layer and wind field model for use in engineering applica- “Physical and numerical modelling of thunderstorm downbursts.”
tions.” J. Appl. Meteorol. Climatol. 48 (2): 381–405. https://doi.org/10 J. Wind Eng. Ind. Aerodyn. 89 (6): 535–552. https://doi.org/10.1016
.1175/2008JAMC1841.1. /S0167-6105(00)00090-8.
Vicroy, D. D. 1992. “Assessment of microburst models for downdraft es- Xu, Y. L., and J. Chen. 2004. “Characterizing nonstationary wind speed
timation.” J. Aircr. 29 (6): 1043–1048. https://doi.org/10.2514/3.46282. using empirical mode decomposition.” J. Struct. Eng. 130 (6): 912–920.
Wang, H., T. Wu, T. Tao, A. Li, and A. Kareem. 2016. “Measurements and https://doi.org/10.1061/(ASCE)0733-9445(2004)130:6(912).
Downloaded from ascelibrary.org by Nottingham Trent University on 07/17/19. Copyright ASCE. For personal use only; all rights reserved.

analysis of non-stationary wind characteristics at Sutong bridge in Yeh, C. H., and Y. K. Wen. 1990. “Modeling of nonstationary ground mo-
Typhoon Damrey.” J. Wind Eng. Ind. Aerodyn. 151 (Apr): 100–106. tion and analysis of inelastic structural response.” Struct. Saf. 8 (1–4):
https://doi.org/10.1016/j.jweia.2016.02.001. 281–298. https://doi.org/10.1016/0167-4730(90)90046-R.
Wang, J., L. Fan, S. Qian, and J. Zhou. 2002. “Simulations of non- Zhang, Z. C., J. H. Lin, Y. H. Zhang, Y. Zhao, W. P. Howson, and
stationary frequency content and its importance to seismic assessment F. W. Williams. 2010. “Non-stationary random vibration analysis for
of structures.” Earthquake Eng. Struct. Dyn. 31 (4): 993–1005. https:// train-bridge systems subjected to horizontal earthquakes.” Eng. Struct.
doi.org/10.1002/eqe.134. 32 (11): 3571–3582. https://doi.org/10.1016/j.engstruct.2010.08.001.

© ASCE 04019092-13 J. Struct. Eng.

J. Struct. Eng., 2019, 145(10): 04019092

You might also like