Engineering Review of Wind Induced Torsional Moment and Response

State-of-the-Art Review
Engineering Review of Wind-Induced Torsional

Moment and Response of Buildings
Hamidreza Alinejad 1; Thomas H.-K. Kang, M.ASCE 2; Seung Yong Jeong 3; and Byeonguk Ahn 4
Abstract: For building service and strength design under wind load, torsional moment and response (or torsional-wind load) is a main
component. Several factors contribute to torsional-wind load, including asymmetric distribution of wind pressure on the building façade
(aerodynamic source), dynamic torsional vibration, and the contribution of resonant components of along-wind and across-wind loads in
presence of mass-stiffness eccentricity. In addition, adjacent building influence can be considerable. For low (1 or 2 stories) to midrise (less
than 10 stories) buildings, the main component in torsional-wind load typically is aerodynamic source. For tall buildings, contribution due to
This work is made available under the terms of the Creative Commons Attribution 4.0 International license.
all sources is significant. In this article, theoretical background and procedures to calculate torsional-wind load, with a focus on the concept of
equivalent eccentricity and equivalent static wind load, are discussed. Then, procedures and perspectives in several international standards,
including ASCE 7-22, Architectural Institute of Japan-recommendations for loads on buildings (AIJ-RLB-2015), Australian and New
Zealand Standard (AS/NZS 1170.2-2011), Korean Design Standard (KDS 41-2019), and ISO 4354-2009, are introduced to clarify differing
points of view and how components of torsional-wind load are included. Finally, the main parameters of each standard are compared with
wind tunnel test results. The results confirm consistency between the wind tunnel test results and those based on the standards with con-
sideration of their covered ranges. DOI: 10.1061/JSENDH.STENG-12546. This work is made available under the terms of the Creative
Commons Attribution 4.0 International license, https://creativecommons.org/licenses/by/4.0/.
Author keywords: Torsional-wind load; Aerodynamic; Eccentricity; Interference; Architectural Institute of Japan-recommendations for
loads on buildings (AIJ-RLB); ASCE; Australian and New Zealand standard (AS/NZS); Korean design standard (KDS); ISO.
Introduction along-wind and across-wind directions of structures with mass-

stiffness eccentricity.
Torsional-wind load has a pronounced effect on building perfor- Aerodynamic components typically bear pivotal influence in
mance under both service and strength loads. The magnitude of short and medium-rise buildings. The role of induced-dynamic
torsional-wind load and its action, whether static or dynamic, is af- torsional component becomes significant for tall buildings, as
fected by several factors. Given that human perception is more sen- the natural frequency of the building approaches the high-energy
sitive to angular motion than translational, torsional acceleration can frequency content of the background load, where angular accel-
significantly affect building serviceability (Tallin and Ellingwood eration due to torsional vibration becomes noticeable. In addition,
1984; Kareem 1985). In addition, loading on peripheral members the existence of mass-stiffness eccentricity can considerably inten-
and cladding systems can be increased significantly, when torsional- sify the load.
wind load is considerable (Zhou et al. 2002). Moreover, neighboring These effects can be directly measured using an accurate finite
buildings can influence change of pressure distribution and, conse- element model and performing time-history analysis for aerody-
quently, torsional load. namic wind load obtained by wind tunnel tests. However, the wind
When wind flow passes a building, three torsional moments can tunnel test results are not available at the early stage of design, and
be applied to the structure: (1) torsional moment due to asymmetric performing time-history analysis is not convenient for required trial
distribution of wind pressure on building surfaces (aerodynamic and error at the design stage. Static analysis using equivalent static
component), (2) torsional moment (resonant component) due to in- wind load (ESWL) is a more efficient approach for design. ESWL
duced torsional vibration in response to aerodynamic component, for torsion (hereinafter referred to as torsional-wind load) should be
and (3) torsional moment due to contribution of inertial forces in determined so as to be representative of all mentioned components.
1
In wind engineering, the contribution of aerodynamic load is intro-
Postdoctoral Researcher, Institute of Engineering Research, Seoul National duced using mean and background components, and the effect of
Univ., 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Korea. ORCID: https://orcid
.org/0000-0002-4616-2131. Email: [email protected]
dynamic response is introduced by resonant component. Moreover,
2
Professor, Dept. of Architecture and Architectural Engineering and Insti- depending on the applicable range in different standards, torsional-
tute of Engineering Research, Seoul National Univ., 1 Gwanak-ro, Gwanak-gu, wind load is defined in quite different ways.
Seoul 08826, Korea (corresponding author). Email: [email protected] This article is part of a study on torsional-wind load, with the
3
Postdoctoral Researcher, Institute of Engineering Research, Seoul National purpose of identifying key parameters that should be considered
Univ., 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Korea. Email: [email protected] in determining torsional-wind load. Provisions and procedures
4
Ph.D. Student, Dept. of Architecture and Architectural Engineering, in relevant standards include ASCE 7-22 (ASCE 2022), AS/NZS
Seoul National Univ., 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Korea.
1170.2-2011 (AS/NZS 2011), AIJ-RLB-2015 (AIJ 2015), ISO
Email: [email protected]
Note. This manuscript was published online on September 12, 2023. 4354-2009 (ISO 2009), and KDS 41-2019 (KDS 2019); they are
Discussion period open until February 12, 2024; separate discussions must hereinafter referred to as ASCE 7-22, AS/NZS, AIJ, ISO, and KDS,
be submitted for individual papers. This paper is part of the Journal of respectively. In addition, components of torsional-wind load were
Structural Engineering, © ASCE, ISSN 0733-9445. calculated based on results of wind tunnel tests for a set of buildings
© ASCE 03123001-1 J. Struct. Eng.
J. Struct. Eng., 2023, 149(11): 03123001

peaks. Note that based on observation by Marukawa et al. (1992),
the PSD of torsional moment has two peaks at almost all side ratios,
and the frequencies of the two peaks approach each other with in-
crease in side ratio.
Studies by Blessmann and Riera (1985) and Blessmann (1992)
identified that torsional-wind load can be highly affected by the
interference effect. Research results by Zhang et al. (1995), Yu et al.
(2016), and Hui et al. (2017) also indicate that torsional-wind load
can be significantly larger than that of a building in an isolated con-
dition. The effect of adjacent buildings, which may either decrease
Fig. 1. Instantaneous pressure distribution on the walls of buildings at or increase aerodynamic forces, primarily depends on the geometry
the moment of maximum aerodynamic: (a) along-wind load; (b) across- and arrangement of these structures, incident angle of wind flow,
wind load; and (c) torsional-wind load. (Reprinted with permission and upstream terrain conditions (Khanduri et al. 1998).
from Tamura et al. 2008.)
Dynamic Response Effect and Role of

with a square-shaped plan and different heights, and comprehen- Mass-Stiffness Eccentricity
sively compared with the values from the standards.

If the natural frequency of torsional mode is close to the high-energy
frequency content of the aerodynamic load, the response can be sig-
Aerodynamic Load Effect nificantly intensified due to inertial moment. The effect of dynamic
response is introduced by resonant component in wind load. Fig. 2
Aerodynamic torsional moment is due to asymmetric distribution
illustrates the aerodynamic torsional moment and torsional response
of pressure on building surfaces (Fig. 1). The background compo-
of the structure (e.g., rotation and base or story torsional moment)
nent of along-wind load is primarily derived from the turbulence in
in the time domain. Aerodynamic torsional moment is composed of
(or gustiness of) the atmospheric boundary layer (ABL). The back-
mean and background components. The total response is composed
ground component of across-wind load depends on turbulence in
of mean (static response to mean load) and fluctuating components,
both ABL and the local boundary layer (LBL) formed between wind
where the fluctuating component is composed of background com-
flow and the sidewalls, especially due to vortex shedding (Tamura
ponent (static response to fluctuating load) and resonant component
et al. 1996). Mean and background components of aerodynamic tor-
sional moment are due to the contribution of pressures acting on all (dynamic response to fluctuating load).
walls, and thus, is a combination of turbulence in both ABL and The mean and background component of torsional-wind load is
LBL. Typically, the mean value of aerodynamic torsional moment the summation of the torsional moment of pressure on the building
for buildings with a symmetric plan is negligible when the wind surface with respect to a reference point (RP). The point can be
direction is parallel to the symmetric line (Zhou and Kareem 2000); considered at any arbitrary point of the floor plan, but is usually
it is not necessarily true when the plan shape is not symmetrical or chosen at mass center (Cm ), geometric center (Cg ) (geometrical
when the wind direction is not parallel to the symmetric line. centroid of plan area), or elastic stiffness center (Cs ). Regardless
Several factors may be considered as key factors included in the of the point selected, resonant components of along- and across-
aerodynamic source, such as geometry of building, the wind flow wind loads always form at mass center.
incidence angle, mean wind speed, turbulence in ABL and LBL, and If Cm , Cg , and Cs are not the same, additional components are
interference of adjacent building(s). induced. Where Cm and Cs are not the same (i.e., mass-stiffness
Across-wind load and torsional-wind load on buildings with large eccentricity), lateral and torsional modes of vibration are coupled.
pffiffiffiffiffiffiffi
aspect ratios (H= BD) are mainly induced by vortex shedding, Although other cases of internal eccentricity (i.e., between Cm
where B, D, and H are width (normal to wind direction), length (par- and Cg , or between Cs and Cg ) do not result in coupling of lateral
allel to wind direction), and height of the building, respectively. and torsional vibrations, their effect can also be significant for tall
From the experimental results of Marukawa et al. (1992) and Ha buildings. Study results by Tallin and Ellingwood (1985b), Kareem
et al. (2007), the separated vortices reattached to buildings when the (1985), Saiful Islam et al. (1992), and Chen and Kareem (2005)
side ratio (D=B) was larger than 3. Due to reattachment, the power showed that mass-stiffness eccentricity can change the overall load
spectral density (PSD) of across- and torsional-wind loads has two substantially.
Fig. 2. (a) Aerodynamic torsional moment; and (b) torsional response.
J. Struct. Eng., 2023, 149(11): 03123001

Fig. 3. Location of RP for an arbitrary plan shape: (a) at the center (the coincidence of Cg , Cm , and Cs ); (b) at Cm ; (c) at Cg ; and (d) at Cs .
To illustrate possible eccentricities, four cases shown schemati- accidental eccentricity exists due to the nonuniform arrangement
cally in Fig. 3 for an arbitrary plan shape are considered. Here, total of mechanical, electrical and architectural components, construction
torsional-wind load (M TT ) acts at the center of stiffness and includes error, etc. For this reason, consideration of accidental eccentricity in
M T (the summation of aerodynamic and dynamic torsional-wind seismic design is recommended by codes and standards (e.g., 5%
moments) and the product of components FD and FL and their cor- based on ASCE 7-22) where the load is essentially inertial force.
responding eccentricity. For the forces, the first subscripts D and L With regard to wind design, the resonant component (which is
refer to along- and across-wind loads, respectively, and second sub- inertial force) in the case of short and medium-rise buildings is rel-
scripts R and S refer to resonant and the summation of mean and atively small, and noninclusion of accidental eccentricity is reason-
background (FDS ¼ F̄D þ FDB and FLS ¼ F̄L þ FLB ), respec- able. For tall buildings, the resonant component is generally large,
tively. Eccentricities between centers of mass and stiffness are de- especially for across-wind load, and even a small eccentricity can
noted by eDR and eLR , and eccentricities between centers of stiffness amplify the torsional-wind load. Hence, consideration of accidental
and geometry are denoted by eDS and eLS . eccentricity, as a minimum value for mass-stiffness eccentricity if
A summary of the terms that should be considered in each case inherent mass-stiffness is zero, for tall buildings can result in a safer
for calculation of M TT is reported in Table 1. It is worth mentioning design (Alinejad and Kang 2020b).
that floor behavior can be considered as a rigid diaphragm, which is
common in analytical models of tall buildings. Hence, although the
resonant component of M T is also formed at the mass center, it is Evaluation of Torsional-Wind Load by
not necessary to decompose it into resonant component (M TR ) and Wind Tunnel Tests
aerodynamic component (MTS ).
Mass-stiffness eccentricity may be inherent and accidental. In Current wind tunnel tests can be divided into two categories: aero-
practice, even when a building is designed to have zero inherent elastic and aerodynamic. Aeroelastic wind tunnel tests are based on
eccentricity (zero eccentricity between centers of mass and stiffness), scaling of flow characteristics, geometry, and structural system; it is
considered the most accurate available technique for the calculation
of wind load, since all components including mean, background,
Table 1. Contribution of components of FD and FL in MTT due to and resonant are measured directly, and possible aeroelastic effects
eccentricity can be observed within the test. However, this test is expensive and
Extra terms preparation of the model is difficult and complex.
Location Aerodynamic wind tunnel tests are typically used in practice if
of RP Along-wind load Across-wind load the flow is not considerably affected by the building deformation
Centera — — and aeroelastic effects are not significant. Aerodynamic wind tun-
Cm eDR FD eLR FL nel tests are based on scaling of flow characteristics and geometry,
Cg eDS FDS , eDR FDR eLS FLS , eLR FLR but structural properties are not included, and a rigid model is used.
Cs eDR FDR eLR FLR In the test, only aerodynamic components (mean and background)
a
Cg ¼ Cm ¼ Cs . of the loads are measured, using high-frequency force balance or
J. Struct. Eng., 2023, 149(11): 03123001

pressure integration techniques, and resonant component and con- moment in wind tunnel test is based on uniform mode shape
tribution of mass-stiffness eccentricity should be obtained by per- (β ¼ 0), while the load is based on linear mode shape (β ¼ 1).
forming numerical analysis. In common practice, a value between 0.6 and 0.7 is used for
For the calculation of torsional-wind load based on the result of ϕT0 in Marukawa et al. (1992), ASCE 49-21 (ASCE 2021), AIJ
aerodynamic wind tunnel tests, two approaches are considered. (2015), ISO (2009), and KDS (2019).
In case of short and medium-rise buildings, across-wind load and To calculate RT , σMTR can be obtained by time-history analysis
resonant component of torsional-wind load are negligible. The mean or by applying mechanical transfer function to the PSD of the base
and background components of along- and torsional-wind loads moment. In a case where a smooth PSD is available and the as-
(FDS and M TS , respectively) are measured from the wind tunnel test sumptions that the externalp load can be considered as white noise,
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
as base shear and torsional moment with respect to center of geom- σMTR can be estimated by πf1T SMTB ðf 1T Þ=4ξ 1T , where f 1T and
etry. Then, equivalent eccentricity (eS ) is representative of aerody- ξ 1T are the frequency and damping of the first torsional mode, re-
namic source of torsional-wind load and calculated as the ratio of spectively, and SMTB ðf1T Þ is the value of PSD of base torsional
M TSmax =ðFDSmax BÞ, where B is projected width of building in the moment at frequency of f1T (Lin 1967; Tallin and Ellingwood
direction normal to wind direction, and FDSmax and M TSmax are the 1985b; Boggs and Dragovich 2006). The general procedure for
maximum values of FDS and MTS , respectively (Boggs et al. 2000; obtaining equivalent static torsional-wind load based on wind tun-
Isyumov and Case 2000; Elsharawy et al. 2012). If mass-stiffness nel test results is reported in Appendix I.
eccentricity (eR ) is not zero, it should be considered for FDR as In the above procedure, it is assumed that there is no coupling
an added term for torsional-wind load. To view the correlation be- between lateral and torsional mode shapes, and thus mass-
tween along- and torsional-wind loads, the load ratio is calculated as stiffness eccentricity is zero. Several models were proposed to
FDS ðMTSmax Þ=FDSmax , where FDS ðMTSmax Þ is the value of FDS at consider coupled mode shapes and the contribution of mass-
the moment where M TS ¼ M TSmax . This method is implemented in stiffness eccentricity to torsional-wind load. Cui and Caracoglia
ASCE 7-22. Although the general procedure to obtain equivalent (2018) proposed a universal formula based on a novel generalized
eccentricity is the same, sometimes slightly different assumptions aerodynamic formulation for calculating the coupled dynamic re-
are used (e.g., using FD instead of FDS for calculation of equivalent sponse of a tall building with consideration of aerodynamic damp-
eccentricity). ing and stiffness.
For tall buildings, the condition is more complicated and both
translational vibration in across-wind direction and torsional vibra-
tion are significant. In this condition, the procedure based on equiv- Exploring Torsional-Wind Load in Standards
alent eccentricity is not suitable. Equivalent static torsional-wind
load, comprising mean, background, and resonant components, can In general, standards that include torsional-wind load can be di-
be a more realistic representation of wind-induced torsional effects. vided in two categories, where the first group is based on equiv-
Because of the complexity associated with distribution of forces alent eccentricity (ASCE 7-22 and AS/NZS) and the second
and torsional moment along with height, base forces and moments, group is based on computational procedure for calculation of
in general, are measured and distributed based on fundamental mode equivalent static torsional-wind load (AIJ 2015; ISO 2009;
shapes. Some models do not explicitly include turbulence intensity KDS 2019). These procedures are explained in detail in the fol-
of ABL (Tamura et al. 1996; Li et al. 2014), while others include it lowing subsections.
(Liang et al. 2004; Katsumura et al. 2014).
Despite differences in the derivation of the equivalent static Procedure in ASCE 7-22 and AS/NZS
torsional-wind load, the load should be representative of aerody-
namic load (mean and background loads) and dynamic response. The provisions for wind design in ASCE 7-22 are provided mainly
The equivalent static load for the dynamic response (resonant load) for short and medium-rise buildings with heights less than 122 m
is typically obtained based on either peak deformation or accelera- (150 m in practice) and cases that are not subject to vortex shedding
tion (Ryu et al. 2020). In current practice, with assumptions of neg- and associated across-wind load. ASCE 7-22 does not provide an
ligible mean value and linear mode shape, equivalent static torsional- analytical equation for across-wind load by the vortex shedding. In
wind load based on measured torsional moment in wind tunnel tests addition, it is assumed that torsional vibration is not considerable.
is defined based on Eq. (1) (Tamura et al. 1996; Ha 2017; ISO 2009; Hence, torsional-wind load, M T , in ASCE 7-22 is defined based on
AIJ 2015; KDS 2019) contribution of aerodynamic torsional-wind load and resonant com-
ponent of along-wind load (if mass-stiffness eccentricity is not zero).
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
MT ðzj Þ ¼ 3gT qH ðϕT0 CT0 ÞBAj μðzj Þ 1 þ ϕT RT ð1Þ Therefore, it can be viewed that provisions in ASCE 7-22 are for
short and medium-rise buildings, with wind tunnel tests required
where zj = elevation of jth story; gT = peak factor; qH = velocity for tall buildings. Based on this code, the torsional-wind load for
pressure at building height, H; B = width of building normal to rigid and flexible buildings is defined by M T ¼ eQ BFD and M T ¼
wind direction; Aj ¼ Bhj is the projected area normal to wind eBFD , respectively, where B is the width of building normal to wind
direction; hj = height of jth story; CT0 ¼ σMTB =ðqH HB2 Þ is the direction, eQ is eccentricity for rigid buildings, and e is eccentricity
normalized σMTB , standard deviation of measured base torsional for flexible buildings. Based on results of aerodynamic wind tunnel
tests, the measured eQ for the rigid building is about 0.15.
moment in wind tunnel; RT ¼ ðσMTR =σMTB Þ2 is the normalized
Note that eQ is the equivalent eccentricity introduced in the sec-
σMTR , standard deviation of resonant response; and μðzÞ ¼ ðz=HÞβ
is the mode shape. Also, ϕT is the mode correction factor for non- tion “Evaluation of Torsional-wind Load by Wind Tunnel Tests.”
linear mode shape and is usually defined as the ratio of the stan- Eccentricity for the flexible building is calculated by using the fol-
dard deviation or PSD of generalized load based on different mode lowing equation:
shapes (Tallin and Ellingwood 1985a; Holmes 1987; Kijewski and qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Kareem 1998; Zhou and Kareem 2000; Zhou et al. 2002); and ϕT0 eQ þ 1.7I z̄ ðgQ QeQ Þ2 þ ðgR ReR Þ2
e¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2Þ
is the mode correction factor for generalized base torsion and is
1 þ 1.7I z̄ ðgQ QÞ2 þ ðgR RÞ2
applied to CT0 . This is used because measured base torsional
J. Struct. Eng., 2023, 149(11): 03123001

Table 3. Covered range by AIJ, ISO, and KDS
Item Covered range
Plan section Uniform rectangular
H
Height to plan aspect ratio 3 ≤ pffiffiffiffiffiffiffi ≤ 6
BD
D
Plan aspect ratio 0.2 ≤ ≤ 5
B
VH VH
Wind speed and natural frequency pffiffiffiffiffiffiffi ≤ 10 and pffiffiffiffiffiffiffi ≤ 10
f 1L BD f 1T BD
Note: H = building height; B and D = width of building normal and parallel
to wind direction, respectively; V H = wind speed at building height; and f 1L
and f 1T = fundamental natural frequency in across and torsional modes of
vibration, respectively.
Fig. 4. Variation of eccentricity in ASCE 7.

Case 2 and Case 4 are for the combination of along- and torsional-
wind loads. Case 3 and Case 4 identified in the table are for the
where Q and R = background and resonant factors, respectively; I z̄ = maximum load effect scenario when the wind load acts in a diagonal
turbulence intensity at equivalent height (60% of building height); direction. Based on wind tunnel tests of a rigid building, the peak
gQ and gR = peak factors for background and resonant forces, re- torsional-wind load is equivalent to the case where 15% eccentricity
spectively; and eR = eccentricity between mass and stiffness center. exists for a portion of peak along-wind load. It occurs at about 75%
It is worth noting that accidental eccentricity can be simply in- of the peak along-wind load. Thus, a reduction factor of 0.75 is in-
cluded in ASCE 7-22 by directly adding its value to eR . Based on cluded for Case 2 and Case 4.
the results of Alinejad and Kang (2020a), variation of e as a func- Based on AS/NZS, torsional-wind load should be included
tion of building time period, damping ratio, eR , and wind speed is based on 20% eccentricity with respect to the center of geometry
shown schematically in Fig. 4. of the building on the along-wind load for rectangular buildings
When R is zero (i.e., for rigid buildings), Eq. (2) is reduced sim- taller than 70 m. It is also required where the plan aspect ratio is
ply to eQ . However, for flexible buildings (with nonzero R) with larger than 1.5, due to the influence of across- and torsional-wind
eR ¼ 0, the value of e is smaller than eQ . The reason is that M T loads. In contrast to ASCE 7-22, the eccentricity in AS/NZS is ap-
plied to all components of along-wind load (mean, background, and
is calculated based on FD (including its mean, background, and res-
resonant components). In addition, the eccentricity should be used
onant components). However, mean and background are multiplied
for medium-rise to tall buildings, and thus includes torsional vibra-
by eQ B and resonant component by eR B. Hence, when the resonant
tion (resonant component). Based on the evaluation by Alinjead et al.
component is not zero but eR is zero, e is smaller to adjust the re-
(2020), the ratio of torsional-wind load (including resonant compo-
quired correction for exclusion of the resonant component. Thus, the
nent) and along-wind load for buildings with height of 80 to 200 m
value of e is smaller for a larger R component (due to higher wind
and different wind speeds varies between 0.11 to 0.24. Therefore, a
speed, longer period of the structure, and smaller damping ratio).
value of 0.2 for eccentricity can be considered a logical value for the
For cases with non-zero eR but less than eQ , the value of e is also
estimation of torsional-wind load.
less than eQ . By subtracting eQ from the numerator, the multiplier of
R would be eR =eQ , with the multiplier in the denominator being 1.
Hence, the value of e is always less than eQ , unless eR is larger than Procedure in AIJ, ISO, and KDS
eQ . When eR is larger than eQ for a larger R component, the value of In addition to wind load provisions for short and medium-rise
e is larger. For better understanding of the above observation, the buildings, AIJ, ISO, and KDS provide procedures to calculate
procedure to derive Eq. (2) is provided in Appendix II.
To consider the maximum impact of the wind load on the struc-
ture, four load cases are defined by ASCE 7-22, as shown in Table 2.
Table 4. Definition of background and resonant component in AIJ, ISO,
and KDS
Table 2. Combination of along- and torsional-wind loads in ASCE 7-22 Background

Code components (M TB ) Resonant component (MTR )
Eccentricity, β
ea zj zj pffiffiffiffiffiffi
Case FD MT AIJ 1.8gT CT0 qH BAj 1.8gT CT0 qH BAj ϕT RT
H H
1 FDX — — β β
zj zj pffiffiffiffiffiffi
FDY — — ISO 3gT CT0 qH BAj 3gT CT0 qH BAj K RT
0.75FDX 0.75eX FDX eX ¼ eBX H H
2
0.75FDY 0.75eY FDY eY ¼ eBY zj zj p ffiffiffiffiffiffi
KDS 1.8gT CT0 qH BAj 1.8gT CT0 qH BAj RT
3 0.75FDX þ 0.75FDY — — H H
4 0.563FDX þ 0.563FDY 0.563eX FDX þ 0.563eY FDY eX ¼ eBX ,
Note: qH = pressure velocity at building height; B and H = width and height
eY ¼ eBY
of the building, respectively; zj , hj , and Aj = elevation, height, and
Note: FDX and FDY are the wind force FD corresponding to x and y area normal to wind direction (¼ Bhj ) of jth story, respectively; CT0 =
directions, respectively; BX and BY are the widths of building normal to normalized standard deviation of torsional moment fluctuation; gT =
x and y axes, respectively; eX and eY are the eccentricities for x and y peak factor in torsional-wind load; RT = resonant response factor for
axes, respectively. torsional-wind load; K and ϕT = mode shape correction factors; and β =
a
eQ is used for rigid building in lieu of e. exponent of power law for the first mode vibration shape.
J. Struct. Eng., 2023, 149(11): 03123001

across- and torsional-wind loads for tall buildings with a rectan- consideration of background component (M TB ) and resonant com-
gular plan for a range of wind speeds and building properties ponent p(M TR ), torsional-wind load is defined in the form of
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(Table 3). M T ¼ M 2TB þ M 2TR . The definition of M TB and M TR in each code
Torsional-wind loads in these standards are obtained based on is provided in Table 4.
the procedure explained in the section “Evaluation of Torsional- Peak factor in torsional-wind load, gT , based on ISO, is defined
Wind Load by Wind Tunnel Tests” and by the format of Eq. (1) (Ha by Eq. (3), and for AIJ and KDS by Eq. (4). The values from both
2017). With assumption of negligible mean component and only equations are almost the same
Table 5. Definition of FT and its related parameters in ISO and KDS

Parameter Content
0.14K 2T ðV T Þ2βT DðB2 þ D 2 Þ2
FT ¼ V T ≤ 4.5 or 6 ≤ V T ≤ 10
π L2DB B3
Coefficients of spectrum in torsional vibration FT ðV T ¼ 4.5Þ ¼ F4.5 and FT ðV T ¼ 6Þ ¼ F6

F6 V
FT ¼ F4.5 exp 3.5 ln ln T 4.5 < V T < 6
F4.5 4.5
8
>
> −1.1ðD=BÞ þ 0.97
>
< þ 0.17 V T ≤ 4.5
ðD=BÞ2 þ 0.85ðD=BÞ þ 3.3
Amplification factor for torsional vibration spectrum KT ¼
>
> 0.077ðD=BÞ − 0.16 0.35
>
: þ þ 0.095 6 ≤ V T ≤ 10
ðD=BÞ2 − 0.96ðD=BÞ þ 0.42 ðD=BÞ
8
>
> ðD=BÞ þ 3.6 0.14
>
< þ þ 0.14 V T ≤ 4.5
ðD=BÞ2 − 5.1ðD=BÞ þ 9.1 D=B
Exponential factor for the normalized wind speed for torsional vibration KT ¼
>
> 0.44ðD=BÞ2 − 0.0064
>
: þ 0.2 6 ≤ V T ≤ 10
ðD=BÞ4 − 0.26ðD=BÞ2 þ 0.1
VpHffiffiffiffiffi
Normalized wind speed for the torsional vibration V T ¼ f BD
1T
Note: B and D = widths of building normal and parallel to wind direction, respectively; LDB = greater value of D and B; V H = wind speed at building height;
and f 1T = fundamental natural frequency in torsional mode of vibration.
Table 6. Definition of FT and its related parameters in AIJ

Parameter Content
Coefficients of spectrum in torsional vibration FT ¼ 0.8FB þ v1 FV þ w1 FW
18f m
Component of FT due to approach flow and turbulence created by the buildinga FB ¼
ð1 þ 0.46ð18f m Þ1.8 Þ2.3

1 lnðf s =v3 Þ þ 0.5v22 2
Component of FT due to vortex sheddinga FV ¼ pffiffiffiffiffiffi exp −0.5
v2 2π
v2

1 lnðf m =w3 Þ þ 0.5w22 2
Component of FT due to reattachmenta FW ¼ pffiffiffiffiffiffi exp −0.5
w2 2π w2
2
−2ððD=BÞ − ðD=BÞÞ D=B ≤ 0.9 ð 1.0Þ
Amplification factor for FV v1 ¼
0.1ðD=BÞ−1 D=B > 0.9 ð 1.0Þ

0.56I H ðD=BÞ−1 D=B ≤ 0.9 ð 1.0Þ
Factor related to FV v2 ¼
2I H ðD=BÞ0.5 D=B > 0.9 ð 1.0Þ
−0.2
0.8ðD=BÞ D=B ≤ 0.9 ð 1.0Þ
Factor related to FV v
3 ¼
0.18ðD=BÞ D=B > 0.9 ð 1.0Þ

0.04ðD=BÞ−0.5 D=B ≤ 0.9 ð 1.0Þ
Amplification factor for FW w1 ¼
0.06 D=B > 0.9 ð 1.0Þ
pffiffiffiffiffi
1.75 I H D=B ≤ 0.9 ð 1.0Þ
Factor related to FW w2 ¼ pffiffiffiffiffi
0.7 I H D=B > 0.9 ð 1.0Þ
Factor related to FW w3 ¼ 0.24ðD=BÞ−0.4
f B
Nondimensional frequency for FB and FW fm ¼ 1T
VH
8.3f1T Bð1 þ 0.38ðD=BÞ1.5 Þ0.89
Nondimensional frequency for FV fs ¼
VH
Note: B and D = widths of building normal and parallel to wind direction, respectively; V H = wind speed at building height; I H = turbulence intensity at
building height; and f 1T = fundamental natural frequency in torsional mode of vibration.
a
Definitions were taken from Katsumura et al. (2014).
J. Struct. Eng., 2023, 149(11): 03123001

Table 7. Definition of CT0 in AIJ, ISO, and KDS Table 9. Combination factor for across- and torsional-wind loads, κ, of
ISO and KDS
Standard Content
D=B n1 B=V H κ
AIJ CT0 ¼ 0.04ðD=BÞ2 þ 0.02
KDS CT0 ¼ f0.0066 þ 0.015ðD=BÞ2 g0.78 ≤0.5 0.1 0.55
ISO CT0 ¼ f0.0034 þ 0.0078ðD=BÞ2 g0.78 0.2 0.65
0.6 0.8
Note: B and D = widths of building normal and parallel to wind direction,
respectively. 1 0.1 0.55
0.3 0.55
0.6 0.65
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0.577 ≥2 — 0.55

gT ¼ 2 lnð600f1T Þ þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3Þ
2 lnð600f1T Þ Note = For intermediate values of D=B and n1 B=U H , linear interpolation
can be used. B and D = widths of building normal and parallel to wind
direction, respectively; V H = wind speed at building height; and n1 =
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi smaller of natural frequency in first translational mode in across-wind
gT ¼ 2 lnð600f 1T Þ þ 1.2 ð4Þ
direction (f1L ) and first torsional mode (f1T ).
where f1T = natural frequency of the first mode of torsional

vibration.
In contrast to KDS, the mode shape correction factor for nonlin- Table 10. Correlation coefficient for across- and torsional-wind loads,
ρLT , of AIJ
ear mode shape is included in AIJ and ISO (K and ϕT , respectively); pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
K and ϕT are defined by 0.27β þ 0.73 and I T1 =I T ð1 − 0.4 ln βÞ, ρLT 2 þ 2ρLT − 1
respectively, where β is the exponent of power-law for the first tor- D=B n1 B=V H ζ ¼ 1.0 ζ ¼ 1.1 ζ ≥ 1.4 ζ ¼ 1.0 ζ ¼ 1.1 ζ ≥ 1.4
sional vibration mode, I T is the generalized mass of building for
≤0.5 ≤0.1 0.9 0.7 0.5 0.95 0.84 0.73
torsional vibration, and I T1 is the value of I T for linear mode shape
0.2 0.3 0.6 0.5 0.61 0.79 0.73
(i.e., β ¼ 1). Values of K and ϕT are very close to each other for β in 0.3 0.4 0.6 0.6 0.67 0.79 0.79
the range of 0.1 to 2. 0.6 0.6 0.6 0.6 0.79 0.79 0.79
Resonant response factor for torsional vibration, RT , is defined ≥1 0.7 0.7 0.7 0.84 0.84 0.84
by πFT =ð4ξ 1T Þ, where ξ 1T is the damping ratio of the first mode of
torsional vibration and FT is the coefficient of spectrum in torsional 1 ≤0.1 0.8 0.5 0.2 0.90 0.73 0.55
0.2 0.6 0.5 0.5 0.79 0.73 0.73
vibration.
≥0.3 0.5 0.5 0.5 0.73 0.73 0.73
Procedures to calculate FT in ISO and KDS are the same
(Table 5), while in AIJ the procedure differs (Table 6). ≥2 ≤0.05 0.6 0.4 0.3 0.79 0.67 0.61
CT0 is equal to σMTB =ðqH HB2 Þ, where σMTB is the standard 0.1 0.6 0.2 0.2 0.79 0.55 0.55
deviation of base torsional moment fluctuation, and its definitions ≥0.2 0.2 0.2 0.2 0.55 0.55 0.55
based on listed standards are provided in Table 7. Note = For intermediate values of D=B, ζ, and n1 B=V H , linear interpolation
Derivation of procedures in these standards is more-or-less similar can be used. B and D = widths of building normal and parallel to wind
to the method described in the section “Evaluation of Torsional-Wind direction, respectively; V H = wind speed at building height; n1 = smaller
Load by Wind Tunnel Tests.” In these standards, torsional-wind load of natural frequency of first translational mode in across-wind direction
is calculated based on linear mode shape with a correction factor for (f 1L ) and first torsional mode (f 1T ); and ζ ¼ f 1L ≥ f 1T ζ ¼ f 1L =f1T ,
generalized base torsion, ϕT0 , equal to 0.6. In ISO, the standard f1L < f 1T ζ ¼ f 1T =f1L .
deviation is multiplied by the factor and directly included in
CT0 , while in AIJ and KDS it is included in the multiplier of
the load (3 × 0.6 ¼ 1.8). The values of CT0 in ISO (after applying Overall, AIJ considers larger load combination factors for buildings
ϕT0 ¼ 0.6) and KDS are similar to that in the former version of with smaller plan aspect ratios (D=B).
AIJ (AIJ 2004), whereas for AIJ, a fitted function developed by In the derivation of the load combinations in these standards, it is
Katsumura et al. (2014) is introduced for CT0 . assumed that the response can be expressed by the normal distribu-
Load case combinations for along-, across-, and torsional-wind tion, which can be attributed to the significant contribution of res-
loads based on AIJ, ISO, and KDS are listed in Table 8. onance response for tall buildings with small natural frequencies.
The values of factors for combining across- and torsional-wind Responses in two directions, M x and M y , are expressed through
loads (κ and ρLT ) shown in Table 8 are listed in Tables 9 and 10. a 2D normal distribution, where M x and My can be base overturning
Table 8. Combination of along-, across-, and torsional-wind loads in AIJ, ISO, and KDS
FD FL MT
Load case AIJ ISO KDS AIJ ISO KDS AIJ ISO KDS
1 1 0.4 0.4
0.6 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 0.4 þ 1 2 þ 2ρLT − 1 κ
GD
0.6 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 0.4 þ 2 þ 2ρLT − 1 κ 1
GD
Note: FD , FL , and MT = along-, across-, and torsional-wind loads, respectively; GD = gust-effect factor (mean dynamic response factor for FD in ISO); and
κ and ρLT = wind load combination factors and correlation coefficient for across- and torsional-wind load.
J. Struct. Eng., 2023, 149(11): 03123001

If damping ratio, ξ, is sufficiently small, the steady-state re-
sponse of structure, uðtÞ, with natural frequency, fn , under external
sinusoidal force (with the frequency of f) can be calculated by
Eqs. (6)–(8) (Chopra 2017)
uðtÞ ¼ ust0 Rd sinð2πft − ϕÞ ð6Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi −1
Rd ¼ ½1 − ðf=f n Þ2 2 þ ½2ξðf=f n Þ2 ð7Þ
2ξðf=fn Þ
ϕ ¼ tan−1 ð8Þ
1 − ðf=f n Þ2
where ðust Þo = maximum value of static deformation; Rd = defor-

Fig. 5. Schema of AIJ load combination in consideration of response mation response factor; and ϕ = phase lag. Regardless of natural
correlation. frequency, the responses can be considered as sinusoidal functions
with frequency of load. Aerodynamic wind loads are wideband
processes and can be decomposed into a summation of sinusoidal
functions with different frequencies. The total response of a linear

system to aerodynamic wind load is the combination of the re-
or torsional moments, respectively. In this condition, the equiva- sponses to harmonic loads corresponding to each frequency content
lence line of probability becomes an elliptical line using the corre- of the load. If resonant response is very significant and the gov-
lation coefficient of the response, ρ, as shown in Fig. 5 (Solari et al. erning response, the contribution of response from the harmonic
1998; Solari and Pagnini 1999). Given that the probability of any load with the frequency equal to natural frequency, will be signifi-
design point (a combination of M x and M y ) outside of the ellipse is cantly larger than other frequencies. In this condition, overall re-
small, all points on the elliptical line (solid line) can be considered as sponse can be assumed as a single harmonic response with the
a load combination. However, it is impractical to consider all load natural frequency, especially for across-wind and torsional response.
combinations. Therefore, load combinations, based on the model Consider two resonant responses XðtÞ ¼ AX cosð2πfX tÞ and
proposed by Asami (2000), are defined with eight points on the YðtÞ ¼ AY cosð2πfY t þ θÞ with natural frequencies of fX and fY ,
apexes of the enveloping octagon. In this model, the moment in respectively, where AX and AY are amplification factors for the two
y-direction (M yc ) to be combined with the maximum moment in resonant responses, respectively, and θ is phase difference between
x-direction (M xmax ) is defined by XðtÞ and YðtÞ. Here, XðtÞ and YðtÞ may be across- and torsional-
wind responses. It can be shown that covariance between XðtÞ and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi YðtÞ can be derived by using Eq. (9)
M yc ¼ M̄ y þ mymax 2 þ 2ρ − 1 ð5Þ
CovðXðtÞ; YðtÞÞ ¼ E½ðAX cosð2πfX tÞÞðAY cosð2πf Y t þ θÞÞ
where M̄ y and mymax = mean and the maximum of fluctuating com- AX AY
¼ ðE½cosð2πðf X − f Y Þt − θÞ
ponent of the moment in the y-direction; and ρ = correlation coef- 2
ficient between M x and M y . Regarding the components of wind þ E½cosð2πðf X þ f Y Þt þ θÞÞ ð9Þ
load, ρ is defined as ρDL (correlation coefficient between along- and
across-wind loads), ρDT (correlation coefficient between along- and where E½· = the expected value. Because f X and f Y are positive
torsional-wind loads), and ρLT (correlation coefficient between values, the right expectation is equal to zero. However, the left ex-
across- and torsional-wind loads). pectation is equal to zero if f X and fY are not equal. Thus, unless
The correlation factor for across- and torsional-wind loads fX and f Y are the same, the covariance between XðtÞ and YðtÞ is
(κ and ρLT ) is defined based on (1) co-coherence (correlation co- zero, and the resultant correlation coefficient is also zero. In prac-
efficient for each frequency) between along- and across-wind loads tice, natural frequency in torsional mode of vibration is not the
and also between along- and torsional-wind loads being negligible, same as that of translational modes. Natural frequency of torsional
i.e., ρDL and ρDT are zero; and co-coherence between across- and mode in ISO was assumed to be around 1.3 to 1.4 times those of
torsional-wind loads not being zero, where the absolute value of the translation modes. Thus, if torsional mode is not coupled with trans-
correlation coefficient of response, ρLT , is defined based on wind lational modes, it can be assumed that the correlation coefficient
tunnel tests. between torsional-resonant response and both along- and across-
Factors for along-wind load, FD , in load cases 2 and 3 are ob- resonant responses is zero. On the other hand, for background com-
tained as follows. Because ρDL and ρDT are zero, the value inside ponents, if original two aerodynamic wind loads are correlated at a
the parenthesis in Eq. (5) is equal to roughly 0.4. By replacing M yc specific frequency, the background responses are still correlated,
0 provided that contribution of phase lags, ϕ, by natural frequencies
with components of FD (¼ F̄D þ FDmax , i.e., summation of mean
and maximum fluctuation component, respectively), Eq. (5) can be is not significant. However, amplitudes of background responses are
0
written as FD ¼ F̄D þ 0.4FDmax . Substituting F 0 Dmax by FD − F̄D negligible compared with resonant responses. Consequently, a value
and considering the definition of along-wind load, FD ¼ GD F̄D , it of 0.4 can be used instead of an exact value of κ.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
is further modified as FD ð0.4 þ 0.6=GD Þ. The values of κ ð¼ 2 þ 2ρ − 1Þ in ISO and KDS are also cal-
In case of across- (FL ) and torsional-wind (MT ) loads, the mean culated following a similar procedure.
term in Eq. (5) is zero. Therefore, the factors of 0.4 for both of FL Values of ρLT in Table 10 for AIJ are obtained based on time-
and M T in load case 1 are obtained by considering zero ρDL and history analysis (AIJ 2015; Somekawa et al. 2014), while the val-
ρDT . Similarly, the factors for load cases 2 and 3 are obtained by ues of κ in ISO and KDS are calculated by the statistical analysis
replacing ρ with ρLT . method (Asami 2000). In all cases, analysis is done under a specific
J. Struct. Eng., 2023, 149(11): 03123001

Table 11. Wind tunnel information To compare the results based on wind tunnel tests with ASCE
Averaged wind speed at 7-22, equivalent eccentricity and load ratio were calculated for
Model dimensions building height (m=s) buildings with different aspect ratios and for angles of attack of 0°,
H (m) B ¼ D (m) α ¼ 1=4 α ¼ 1=6
as shown in Fig. 6. The results show that both equivalent eccen-
tricity and load ratio decrease with increasing building aspect ratio.
0.2 0.1 10.97 10.87 In general, the value of equivalent eccentricity for α ¼ 1=4 is con-
0.3 10.75 11.15
siderably larger than that which can be attributed to the higher
0.4 11.00 11.37
0.5 11.14 11.25
turbulence in comparison with α ¼ 1=6, but the load ratio for α ¼
1=4 is less than α ¼ 1=6. Overall, the average value of equivalent
Note: H = building height; B = building width; D = building depth; and eccentricity is between 0.1 and 0.15, which is comparable to the
α = exposure factor.
value of eQ ¼ 0.15 in ASCE. For load ratio, the average value is
between 0.7 and 0.9, which is generally compatible with the load
ratio of 0.75 in ASCE.
range of damping ratios (e.g., 0.01–0.05) and the assumption that Normalized coefficient of base torsional moment, CMT
the building has no coupling vibration mode (i.e., zero eccentric- (obtained by dividing the component by ðqH HB2 Þ), of buildings
ity). Hence, provisions are not applicable when the damping factor with different aspect ratio and exposure factor is shown in Fig. 7.
is not in the range, or if the vibration mode of the building is sig- According to the figure, the effect of exposure type is not consid-
nificantly coupled (i.e., nonzero eccentricity). erable on the values of CMT , especially for mean and standard
Based on ISO, if the building has some eccentricity, a more de- deviation. In addition, the mean value is relatively small in com-
tailed examination should be made for the wind-load combinations; parison with standard deviation and maximum value, and is almost
whereas, in AIJ, buildings with an eccentricity (eccentric distance/ zero in the 0° angle of attack.
radius of rotation) of 0.15 or less (i.e., eR ≤ 0.15) are subject to the Fig. 8 shows the value of CT0 [¼ σTB =ðqH HB2 Þ] for the build-
estimation equation. However, the wind load on a building for which ings to compare with the listed standards. It can be seen that the
eccentricity cannot be ignored needs to be calculated based on the value of CT0 for all buildings is maximum at the angle of attack of
results of wind tunnel tests or computational fluid dynamics (CFD). 0°, and its value decreases by increasing angle of attack. The effect
of exposure type is also negligible.
The observations, including negligible value of the mean com-
Comparison with Results of Wind Tunnel Tests ponent and the largest value of CT0 at an angle of attack of 0°, and
recalling the fact that maximum along- and across-wind loads usu-
Tokyo Polytechnic University (TPU) provides an open-access aero- ally occur at the angle of attack of 0° for square-shape plan, confirm
dynamic database for high-rise buildings (Tamura 2012), which the consistency of the standards with the wind tunnel test results.
includes results of pressure integration tests on 1/400 scaled rigid According to Fig. 8, the value of CT0 at 0° angle of attack is be-
rectangular prism models with different heights, side ratios (D=B), tween 0.03 and 0.05 for buildings with different aspect ratios,
exposure types, and angles of attack. Sampling frequency and du- whereby increasing the aspect ratio, the difference between the val-
ration of all tests are 1,000 Hz and 32.768 s, respectively. For the ues of CT0 becomes smaller. Note that buildings with an aspect ratio
purpose of this study, wind pressure data were extracted for build- of 2∶1∶1 are not in the covered range of the listed standards (Table 3).
ings with aspect ratio (height∶breadth∶depth) of 2∶1∶1, 3∶1∶1, 4∶1∶1, Based on test results, the values of CT0 at 0° angle of attack for build-
and 5∶1∶1 (which correspond to buildings with width of 40 m and ings with aspect ratios of 3∶1∶1, 4∶1∶1, and 5∶1∶1 are 0.037, 0.045,
height of 80, 120, 160, and 200 m in real scale); the sample included and 0.048, respectively.
urban and suburban areas with wind speed profile power (exposure The listed standards provide CT0 as a function of plan aspect ratio
factor), α, of 1/4 and 1/6, respectively, and 0° to 45° angle of attack. (D=B) and independent of building aspect ratio. The value CT0 for
Information regarding dimensions of specimens and wind speeds in the building with D=B ¼ 1, based on AIJ, is equal to 0.06 and,
the wind tunnel test is summarized in Table 11. based on KDS and ISO, is equal to 0.05 (Table 7). Note that the
Below, wind tunnel results are compared with ASCE 7-22, AIJ, value of ISO is multiplied by 0.6 (the mode shape correction) to be
and ISO (which is similar to KDS). comparable with KDS and AIJ. The value from ISO almost fits with
Fig. 6. (a) Equivalent eccentricity; and (b) load ratio based on TPU.
J. Struct. Eng., 2023, 149(11): 03123001

Fig. 7. Normalized coefficient of base torsional moment, CMT .
Fig. 9. Normalized PSD of base torsional moment based on wind

tunnel test, AIJ, and ISO (5∶1∶1 and α ¼ 1=6).
Fig. 8. Normalized standard deviation of base torsional moment, CT0 .
of less than 0.1. The results illustrate a good agreement between

the standards and wind tunnel results. However, the model by AIJ
the value from wind tunnel tests for buildings with an aspect ratio of can reproduce the step part of the PSD (in the range of normalized
5∶1∶1, while the value from AIJ is about 20% larger. frequency between about 0.1 and 0.2) better than ISO. Although
Normalized PSD of base torsional moment, fSMTB ðfÞ=σ2MTB , of it is not depicted here, the results for other buildings show similar
buildings with an aspect ratio of 5∶1∶1 was calculated based on TPU consistency.
results and compared with AIJ and ISO, as shown in Fig. 9. Note To compare equivalent static torsional-wind load, the first PSD
that the formulation in ISO does not cover the normalized frequency of torsional response was calculated for buildings 200-m high,
J. Struct. Eng., 2023, 149(11): 03123001

exposure C (α ¼ 1=6), and angle of attack of 0°. Damping ratio, ξ,
and wind speed at building height, V H , were assumed to be equal to
1% and 45 m=s, respectively. The response was calculated for nor-
malized natural frequency of structure, fT1 B=V H , in the range of
0.1 to 0.5. Corresponding value of f T1 was calculated and the PSD
of torsional response based on wind tunnel results, AIJ, and ISO
were obtained as the product of corresponding PSD and mechanical
admittance function (Fig. 10). As shown in the figure, the PSD of
the response based on the wind tunnel test results is in a good agree-
ment with AIJ and ISO.
In the next step, the standard deviation of the total response,
σMT , was obtained by integration of the area below the PSD of the
response. The standard deviation of resonant response, σMTR , was
calculated by extracting background components, σMTB (which is
equal to the standard deviation of the base torsional moment ob-
tained from wind p tunnel test) fromffi the standard deviation of total
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
response ðσMT ¼ σ2MTB þ σ2MTR Þ. Then, RT [¼ ðσMTR =σMTB Þ2 ]
was calculated and compared with the values of AIJ and ISO (refer
Fig. 10. Normalized PSD of torsional response based on wind tunnel to Tables 4–6), as shown in Fig. 11. It can be seen that the results
test, AIJ, and ISO (5∶1∶1, fT1 B=V H ¼ 0.5, and α ¼ 1=6). from AIJ, ISO, and wind tunnel tests are comparable in most ranges.
Values of RT based on ISO for normalized frequency of 0.3 and
larger match well with wind tunnel test results, and for normalized
frequency less than 0.2 are more conservative than TPU. However,
the results based on AIJ fit more closely with the trend based on
wind tunnel test results, especially for normalized frequencies of 0.3
and less.
The equivalent static load is a function of both CT0 and RT .
Therefore, the equivalent static torsional-wind load at building
height, M T (H), was calculated based on AIJ, ISO, and wind tunnel
tests, as shown in Fig. 12. M T (H) based on wind tunnel test results
was calculated based on Eqs. (1) and (4), and ϕT0 ¼ 0.6 was used
to be compatible with AIJ and ISO. Based on the results, for nor-
malized frequencies of about 0.3 and larger, the load calculated
from the wind tunnel test is comparable with AIJ and ISO. In ad-
dition, the values of AIJ and ISO are very close for normalized
frequencies of less than about 0.2. The main reason for differences
of AIJ and ISO for the range of 0.2 to 0.3 the different modeling for
the step-shape part of the PSD. Overall, the value of both AIJ and
ISO is larger than that based on the wind tunnel test for normalized
Fig. 11. Comparison of RT based on wind tunnel test, AIJ, and ISO
frequencies of less than 0.3 (AIJ is between 20% and 70%, with an
(5∶1∶1 and α ¼ 1=6).
average of 50% larger, and ISO is between 2% and 80%, with an
average of 34% larger). In addition to general conservatism in es-
timation of the load in comparison with wind tunnel test results, one
reason can be that, based on AIJ, the provision is applicable to the
structure with mass-stiffness eccentricity smaller than 0.15.
To evaluate the load case factors, a series of linear time-history
analyses have been performed using a single degree of freedom
(SDOF) system. For simplification, evaluation was done only for
a building with an aspect ratio of 5∶1∶1. Also, 0° angle of attack was
chosen to be consistent with the listed standards. Properties of the
SDOF system were defined using generalized mass and mass mo-
ment of inertia, stiffness, and damping with assumption of linear
mode shape. Damping ratio, ξ, of 1% and 2% were examined as
representative values for wind design.
Time-histories were made by dividing the full length of ex-
tracted data into five parts (with duration of 600 s in scaled-up
time-histories), and ensemble averages were reported. To examine
the results for different normalized frequencies (fB=V H ), two ap-
proaches can be taken: (1) fix f and change V H , or (2) fix V H and
change f. With the change of V H , time scale and length of time-
histories vary. Because the correlation between the applied loads
Fig. 12. Comparison of equivalent static torsional moment at building
themselves is also important, particularly in higher values of nor-
height, MT (H), based on wind tunnel test, AIJ, and ISO (5∶1∶1 and
malized frequency where resonant components are small, it is cru-
α ¼ 1=6).
cial to avoid overlapping between the parts. For this reason, V H was
J. Struct. Eng., 2023, 149(11): 03123001

Fig. 13. Correlation coefficient between responses: (a) (f D ∶f T ¼ 1∶1); (b) (f D ∶fT ¼ 1∶1.1); (c) (fL ∶fT ¼ 1∶1); and (d) (f L ∶f T ¼ 1∶1.1).
fixed to preserve the inherent correlation between the applied loads along- and torsional-wind responses, ρDT , is very small in all ranges
themselves. Recalling the values of length scale and duration of of normalized frequencies and for both cases of (fD ∶fT ¼ 1∶1) and
measurement, a value of 45 m=s was chosen for V H , at which the (f D ∶f T ¼ 1∶1.1). Therefore, it is logical to assume that along- and
total time of five time series is almost equal to the total length of torsional-wind responses are uncorrelated.
available data. For the case of (f L ∶fT ¼ 1∶1), the correlation coefficient be-
In addition to the values of natural frequencies, the studies by tween across- and torsional-wind responses, ρLT , is very large at
Somekawa et al. (2014) and Jeong and Kang (2021) showed that a small normalized frequency around 0.1, and decreases by increas-
the ratio of frequencies is also important to calculate load case fac- ing the normalized frequency. For the case of (f L ∶f T ¼ 1∶1.1),
tors. They examined different cases of ratio of frequencies (in the there is no large correlation at small normalized frequencies. At
range of 1 to 1.5), and found that if the frequencies are not equal higher normalized frequency of about 0.3 or larger, ρLT increases
(even for ratio of frequencies equal to 1.1), the load case factors can for both cases of (fL ∶f T ¼ 1∶1) and (f L ∶f T ¼ 1∶1.1). Because the
be significantly smaller than the case with equal frequencies. In resonant component of responses is smaller at higher normalized
contrast to ISO, the combination factors for directional load in AIJ frequency, large values of ρLT are due to the correlation between
are provided based on the frequency ratio (refer to Tables 9 and 10). across- and torsional-wind loads itself rather than the responses.
SDOF systems were investigated with respect to natural frequen- According to the results, the values provided by ISO are close to
cies, including natural frequency ratios of (fD ∶f L ∶fT ¼ 1∶1∶1) and the average value of correlation coefficients in the examined range
(fD ∶f L ∶fT ¼ 1∶1∶1.1) as representative of cases with equal and non- of normalized frequency, and the values provided by AIJ are based
equal frequencies. on an envelope to the analysis results with some degree of con-
Normalized frequencies were set in the range of 0.1 to 0.5 with servatism. For the case of (fL ∶f T ¼ 1∶1) at smaller values of nor-
an increment of 0.01, and for each value a set of five analyses for malized frequency, the correlation coefficients provided by AIJ fit
along-, across-, and torsional-wind directions were performed for better with the analysis results than those provided by ISO. For the
two natural frequency ratios and two damping ratios. Because the case of (f L ∶f T ¼ 1∶1.1), large values of ρLT at small normalized
values of V H (45 m=s) and B (40 m) were constant, natural fre- frequency are not observed. These observations imply that, with
quency was defined by manipulating corresponding normalized a little difference between the natural frequencies, the correlation
frequency and associated stiffness and damping coefficient of the coefficient is very small for small normalized frequency.
SDOF system. In the next step, the value of a component when the other
Absolute values of correlation coefficient between the fluc- component is the maximum was extracted from the results of
tuation components of responses for along- and torsional-wind di- time-history analysis, and scaled by the maximum value of that
rections and across- and torsional-wind directions are shown in time-history (i.e., mYX =mYmax ). Ensemble average of these ratios
Fig. 13. It can be seen that the correlation coefficient between was calculated for each set of five time-histories, and is denoted
J. Struct. Eng., 2023, 149(11): 03123001

Fig. 14. Comparison of values of AXY and κ based on analysis and ISO: (a) (fD ∶f T ¼ 1∶1); (b) (fD ∶f T ¼ 1∶1.1); (c) (fL ∶f T ¼ 1∶1); and
(d) (fL ∶fT ¼ 1∶1.1).
by AYX (ratio to the maximum). Here, Y and X are D for along- Summary and Conclusion
wind, L for across-wind, and T for torsional-wind (for instance,
ADT is equal to mDT =mDmax ). It is worth mentioning that for every The purpose of the present study was to discuss perspectives behind
two components, two cases including AXY and AYX can be calcu- the source of torsional-wind load and the current code-based pro-
lated, and the maximum of two is shown in Fig. 14. Additionally, cedures. Characteristics of torsional moment due to asymmetric
values of κ based on correlation coefficients shown in Fig. 13, wind pressure distribution on building surfaces (aerodynamic source)
uncorrelated cases for along-wind and torsional-wind (ρDT ¼ 0, and possible torsional vibration in response to the load were ex-
similar to the assumption of ISO and AIJ), and the provided values plained, which are primary components of torsional-wind load. The
of ISO and AIJ for across-wind and torsional-wind are calculated role of mass-stiffness eccentricity in calculation of torsional load was
and shown in Fig. 14. clarified, and it was explained that the contribution of mass-stiffness
It can be seen that values of κ for the exact value of ρDT and for eccentricity can be significant and should be carefully considered in
ρDT ¼ 0 are very close, especially for (f D ∶fT ¼ 1∶1.1). The values the calculation of the load. Accordingly, it was noted that for even
of ADT and κ are well matched in all the ranges of normalized buildings with no inherent eccentricity between centers of mass and
frequencies. Thus, the value of 0.4 for κ in AIJ and ISO, with the stiffness, a minimum eccentricity (accidental eccentricity) may need
assumption that along-wind and torsional-wind responses are un- to be considered for more reliable design. The concept is similar
correlated, seems logical. to that commonly used in seismic design to address uncertainties in
As shown in Fig. 14, the value of κ based on wind tunnel tests is both dead and live loads, and quantitative research on this aspect is
matched with that based on AIJ and ISO, except for the small nor- being conducted.
malized frequency around 0.1 where the underestimation by ISO is Although procedures in wind standards for calculating torsional-
quite large. However, for the case with (fL ∶f T ¼ 1∶1.1), the accu- wind load from a theoretical point of view should be the same, clear
racy is higher and the values are closer. It is worth mentioning that differences can be seen in the formulation among various codes. To
natural frequency of tall buildings in translational and torsional clarify these differences and perspectives, procedures in ASCE 7-22,
modes are usually different in practice, and thus the application AS/NZS 1170.2-2011, AIJ-RLB-2015, ISO 4354-2009, and KDS
of the case with (f L ∶fT ¼ 1∶1) is limited. 41-2019 were explained in detail. In the next step, the main
J. Struct. Eng., 2023, 149(11): 03123001

parameters of the standards were comprehensively compared with IðzÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the results of wind tunnel test for several case studies. In general, M T ðzÞ ¼ gT σMTB μðzÞ 1 þ RT ð14Þ
I
a good consistency was observed between results of the standards
and wind tunnel tests for the covered range of each standard. By assuming a linear mode shape and IðzÞ ¼ I, the ratio of
It is expected that the study will lead to a deeper under- IðzÞ=I is equal to 3=H.
standing of the critical parameters that should be considered in Where resonance response is measured as peak acceleration
the calculation of torsional-wind load, which can result in a more [θ̈max ðzÞ], the general procedure is the same, and standard deviation
logical and accurate design. of angular acceleration (σθ̈ ) can be estimated as σθ =ð2πf 1T Þ2 .
Appendix I. Procedure for Equivalent Static Appendix II. Derivation of Eccentricity for Flexible
Torsional-Wind Load (AIJ, ISO, and KDS) Building (ASCE)
The general procedure to obtain equivalent static torsional-wind Eq. (2) is obtained by modifying gust-effect factor in ASCE 7-22.
load based on peak rotation is as follows. The equivalent static Along-wind load in ASCE 7-22 is defined by
torsional-wind load is defined as the product of peak factor for
FD ¼ q3s Cd AG ð15Þ
torsion, gT , and standard deviation of the equivalent static load,
σMT ðzÞ, at the height z, that is, M T ðzÞ ¼ gT σMT ðzÞ. It can be de-
where q3s = 3-second gust velocity pressure based on basic wind

fined as M T ðzÞ ¼ K θ ðzÞθmax ðzÞ, where K θ ðzÞ is torsional stiffness speed, which is 3-second gust wind speed; A = projected area
at height z; and K θ ðzÞ can be estimated as IðzÞð2πf1T Þ2 , where IðzÞ normal to wind direction; Cd = drag force coefficient; and G =
is mass moment of inertia at height z, and f 1T is the frequency of gust-effect factor, which is defined by
first torsional mode (Ha 2017). The peak rotation is defined by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
θmax ðzÞ ¼ gT σθ μðzÞ, where σθ is standard deviation of general- 1 þ 1.7I z̄ ðgQ QÞ2 þ ðgR RÞ2
ized rotation, θ , and μðzÞ is the mode shape. Given the above, G ¼ 0.925 ð16Þ
torsional-wind load can be defined by 1 þ 1.7gV I z̄ gV
M T ðzÞ ¼ IðzÞð2πf 1T Þ2 gT σθ μðzÞ ð10Þ A multiplier of 0.925 is included to match the value of G with
its values in former versions of the code (Kwon and Kareem 2013;
ASCE 1995). The numerator and denominator of Eq. (16) are gust-
Given the dynamic properties of the building, σθ are determined displacement factor, GX , and gust-pressure factor, Gq , respectively.
from postprocessing of the results from wind tunnel tests. The basic In general, gust load factor (GLF) is defined by GX, which is the
parameters from the test are standard deviation of background tor- ratio of peak displacement to mean displacement. Recalling the nu-
sional moment at the base, σMTB , and its power spectral density merator of Eq. (16), the terms 1.7I z̄ gQ Q and 1.7I z̄ gR R are called
(PSD), SMTB ðfÞ, where f is frequency. The generalized externally background and resonant components of GX , which are shown by
applied torsional-wind load, M TB ðtÞ, is defined by GXB and GXR , respectively. GX can be written as
Z qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
H
M TB ðtÞ ¼ M TB ðz; tÞμðzÞdz ð11Þ GX ¼ 1 þ G2XB þ G2XR ð17Þ
0
In ASCE 7-22, duration for averaging to obtain mean and peak

where M TB ðz; tÞ = external torsional moment at height z and time t. wind speeds is 1 h and 3 s, respectively. However, rather than mean
Assuming uniform mode shape, that is, μðzÞ ¼ 1, M TB ðtÞ and mea- hourly wind speed, wind speed maps in ASCE 7-22 are based on
sured base torsion from the test are equal. Therefore, the PSD of gust wind speed (3-s average). As a result, the velocity pressure is
M TB ðtÞ, SMTB ðfÞ, is equal to the PSD of base torsion, SMTB ðfÞ, also the peak value (gust velocity pressure), i.e., q3s . Thus, it is
and so the PSD of θ , Sθ ðfÞ, can be defined by reduced to mean hourly velocity pressure, q1h , by using Gq , which
is the ratio of peak pressure to mean pressure (q3s =q1h ). In this way,
SMTB ðfÞjHðfÞj2 G in ASCE 7-22 is defined as the ratio of GX =Gq to modify
Sθ ðfÞ ¼ ð12Þ
K 2
θ
Eq. (17) as shown in Eq. (18) (Solari and Kareem 1998; Kwon and
Kareem 2013)
where jHðfÞj2 ¼ ð½1 − ðf=f1T Þ2 2 þ ½2ξ 1T ðf=f1T Þ2 Þ−1 is the me- GX
chanical transfer function; ξ 1T is the damping ratio in first torsional FD ¼ q3s Cd A ¼ q1h Cd AGX ð18Þ
Gq
mode; K θ ¼ I ð2πf1T Þ2 is the generalized torsional stiffness; and
I ¼ ∫ H 2
0 IðzÞμ ðzÞdz is generalized mass moment of inertia. The This approach is used wherever peak (or gust) wind speed is
value of σθ isqobtained by integrating
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi the PSD of torsional re- provided instead of mean wind speed. A similar concept is also im-
sponse, σθ ¼ ∫∞
0 Sθ ðfÞdf . Substituting Sθ ðfÞ from Eq. (12)
plemented in AS/NZS (2011) and the peak approach in ISO 4354
(ISO 2009). However, in standards such as AIJ (2015), KDS 41
and calculation of the integral, and then decomposing σθ into res-
(KDS 2019), the mean approach of ISO 4354 (ISO 2009), where
onant and background components, σθ can be expressed by
basic wind speed is the mean value (10-min. mean wind speed), this
σMTB pffiffiffiffiffiffiffiffiffiffiffiffiffiffi modification is not included.
σ θ ¼ 1 þ RT ð13Þ Mean (hourly) along-wind load, F̄D , can be obtained as the ratio
ð2πf1T Þ2 I
of FD =GX . Eq. (18) can be rewritten as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where RT = normalized standard deviation of resonant response. By
FD ¼ F̄D þ F2DB þ F2DR ð19Þ
substituting σθ from Eq. (13) into Eq. (10), M T is defined by
J. Struct. Eng., 2023, 149(11): 03123001

where FDB (¼ F̄D GXB ) and FDR (¼ F̄D GXR ) are background and es = equivalent eccentricity for aerodynamic
resonant components of along-wind load, respectively. Multiplying torsional-wind load [¼ M TSmax =ðFDSmax BÞ;
each component of along-wind load (mean, background, and res- eX = eccentricity for flexible buildings for x-axis
onant) by its corresponding eccentricity, corresponding component (ASCE 7-22);
of torsional-wind load can be obtained. Recalling eccentricity for eY = eccentricity for flexible buildings for y-axis
mean and background (eQ ) and eccentricity for resonant compo- (ASCE 7-22);
nent (eR ), the mean, background, and resonant components of FB = component of FT due to the approach flow and
torsional-wind load are eQ BF̄D ; eQ BF̄D GXB ; and eR BF̄D GXR , turbulence created by the building (AIJ);
respectively. A modified form of gust-displacement factor for FD = along-wind load;
torsional-wind load, GX0 , simply can be defined by F̄D = mean component of along-wind load;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi FDB = background component of along-wind load;
GX0 ¼ eQ þ 1.7I z̄ ðgQ QeQ Þ2 þ ðgR ReR Þ2 ð20Þ FDR = resonant component of along-wind load;
FDS = summation of mean and background components of
Finally, e is defined as the ratio of GX0 =GX for calculation of along-wind load;
torsional-wind load directly based on calculated along-wind load, FDSmax = max of FDS ;
as shown in FL = across-wind load;
GX0 F̄L = mean component of across-wind load;

M T ¼ eBFD ¼ q C ABGX ¼ q1h Cd ABGX0 ð21Þ FLB = background component of across-wind load;
GX 1h d
FLR = resonant component of across-wind load;
FLS = summation of mean and background components of
Data Availability Statement across-wind load;
FLSmax = max of FLS ;
All data, models, and code generated or used during the study ap- FT = coefficient of spectrum in torsional vibration
pear in the published article. (AIJ, ISO, and KDS);
FV = component of FT due to vortex shedding (AIJ);
FW = component of FT due to reattachment (AIJ);
Acknowledgments f = frequency;
fn = natural frequency;
This research was supported by National Research Foundation
fm = normalized frequency for FB and
of Korea (No. 2021R1D1A1B06044752) and by the Institute of
FW (¼ f1T B=V H ) (AIJ);
Construction and Environmental Engineering at Seoul National
University. The views expressed are those of the authors, and do fS = normalized frequency for FV (AIJ);
not necessarily represent those of the sponsors. f1D , fD = natural frequency of the first lateral mode in
along-wind direction;
f1L , f L = natural frequency of the first lateral mode in
Notation across-wind direction;
f1T , fT = natural frequency of the first torsional mode;
The following symbols are used in this paper: GD = gust-effect factor (AIJ, KDS), mean dynamic
A = projected area normal to wind direction; response factor for FD (ISO);
AYX = ratio to the maximum (¼ mYX =mYmax ); gQ = peak factor for background response of along-wind
B = width of building normal to wind direction; load (ASCE 7-22);
BX = width of building normal to x-axis; gR = peak factor for resonant response of along-wind load
BY = width of building normal to y-axis; (ASCE 7-22);
Cg = geometric center; gT = peak factor for torsional-wind load;
CMT = normalized coefficient of base torsional moment; H = building height;
Cm = mass center; h = story height;
Cs = elastic stiffness center; I H = turbulence intensity at building height;
CT0 = normalized standard deviation of base torsional I T = generalized mass of building for torsional vibration;
moment [¼ σMTB =ðqH HB2 Þ]; I T1 = generalized mass of building for torsional vibration
D = width of building parallel to wind direction; for linear mode shape;
e = eccentricity for flexible buildings (ASCE 7-22); I z̄ = turbulence intensity at equivalent height (0.6H)
eDR = eccentricity between centers of mass and stiffness (ASCE 7-22);
for along-wind load; K = mode shape correction factor (ISO);
eDS = eccentricity between centers of geometry and K T = amplification factor for the torsional vibration
stiffness for along-wind load; spectrum (KDS, ISO);
eLR = eccentricity between centers of mass and stiffness LDB = greater value of D and B (ISO, KDS);
for across-wind load; M T = torsional-wind load, summation of aerodynamic and
eLS = eccentricity between centers of geometry and dynamic torsional-wind moments (exception:
stiffness for across-wind load; summation of aerodynamic and eccentric torsional-
eQ = eccentricity for rigid buildings (¼ es ) (ASCE 7-22); wind moments in ASCE 7-22);
eR = eccentricity between centers of mass and stiffness; M̄ T = mean component of torsional-wind load;
J. Struct. Eng., 2023, 149(11): 03123001

M TB = background component of torsional-wind load; Alinejad, H., S. Y. Jeong, and T. H.-K. Kang. 2020. “Comparative assess-
M TR = resonant component of torsional-wind load; ment of ASCE 7-16 and KBC 2016 for determination of design wind
loads for tall buildings.” Wind Struct. 31 (6): 575–591. https://doi.org
MTS = summation of mean and background components of /10.12989/was.2020.31.6.575.
torsional-wind load; Alinejad, H., and T. H.-K. Kang. 2020a. “Engineering review of ASCE 7-16
M TSmax = maximum of M TS ; wind-load provisions and wind effect on tall concrete-frame buildings.”
MTT = total torsional-wind load (including M T and J. Struct. Eng. 146 (6): 04020100. https://doi.org/10.1061/(ASCE)ST
moments due to eccentricity); .1943-541X.0002622.
mYmax = maximum value of Y(t); Alinejad, H., and T. H.-K. Kang. 2020b. “Importance of considering acci-
dental eccentricity in wind design of tall buildings.” In Proc., 2020
mYX = value of YðtÞ at the moment that XðtÞ is maximum;
World Congress on Advances in Civil, Environmental, and Materials
n1 = smaller of f 1L and f1T (AIJ, ISO, KDS); (ACEM20). Daejeon, South Korea: International Association of Struc-
Q = background response factor for along-wind load tural Engineering and Mechanics.
(ASCE 7-22); Asami, Y. 2000. “Combination method for wind loads on high-rise build-
qH = velocity pressure at building height; ings.” [In Japanese.] In Proc., 16th National Symp. on Wind Engineering,
R = resonant response factor for along-wind load 531–534. Tokyo: Japan Association for Wind Engineering.
ASCE. 1995. Minimum design loads and associated criteria for buildings
(ASCE 7-22);
and other structures. ASCE 7-95. Reston, VA: ASCE.
Rd = deformation response factor; ASCE. 2021. Wind tunnel testing for buildings and other structures. ASCE
RT = normalized standard deviation of resonant response 49-21. Reston, VA: ASCE.

for torsional vibration [¼ ðσMTR =σMTB Þ2 ] and ASCE. 2022. Minimum design loads and associated criteria for buildings
[¼ πFT =ð4ξ 1T Þ in AIJ, ISO, and KDS]; and other structures. ASCE 7-22. Reston, VA: ASCE.
S = power spectral density (PSD); AS/NZS (Australian/New Zealand Standard). 2011. Structural design
SMT = PSD of torsional response; actions—Part 2: Wind actions. AS/NZS 1170.2-2011. Sydney, NSW,
Australia: Standards Australia.
SMTB = PSD of base torsional moment;
Blessmann, J. 1992. “Neighbouring wind effects on two tall buildings.”
V H = wind speed at building height; J. Wind Eng. Ind. Aerodyn. 42 (1–3): 1041–1052. https://doi.org/10
V T = normalized wind
pffiffiffiffiffiffiffispeed for torsional vibration .1016/0167-6105(92)90111-M.
[¼ V H =ðf1T BDÞ] (ISO, KDS); Blessmann, J., and J. D. Riera. 1985. “Wind excitation of neighbouring tall
v1 , v2 , v3 = factors related to FV (AIJ); buildings.” J. Wind Eng. Ind. Aerodyn. 18 (1): 91–103. https://doi.org
/10.1016/0167-6105(85)90076-5.
w1 , w2 , w3 = factors related to FW (AIJ);
Boggs, D., and J. Dragovich. 2006. “The nature of wind loads and dynamic
z = elevation; response.” ACI Spec. Publ. 240 (Oct): 15–44.
α = exponent of power law for mean wind speed profile; Boggs, D., N. Hosoya, and L. Cochran. 2000. “Sources of torsional wind
β = exponent of power law for mode shape; loading on tall buildings: Lessons from the wind tunnel.” In Proc.,
β T = exponential factor for the normalized wind speed for Structures Congress and Exposition, Advanced Technology in Struc-
the torsional vibration (ISO, KDS); tural Engineering, 1–8. Reston, VA: ASCE.
Chen, X., and A. Kareem. 2005. “Coupled dynamic analysis and equivalent
κ = wind load combination factors for across- and
static wind loads on buildings with three-dimensional modes.” J. Struct.
torsional-wind loads (ISO, KDS); Eng. 131 (7): 1071–1082. https://doi.org/10.1061/(ASCE)0733-9445
μ = mode shape [¼ ðz=HÞβ ]; (2005)131:7(1071).
ϕ = phase angle; Chopra, A. K. 2017. Dynamics of structures: Theory and applications to
ϕT = mode shape correction factor; earthquake engineering. London: Pearson.
ϕT0 = mode shape correction factor for generalized base Cui, W., and L. Caracoglia. 2018. “A fully-coupled generalized model for
torsion; multi-directional wind loads on tall buildings: A development of the
quasi-steady theory.” J. Fluids Struct. 78 (Apr): 52–68. https://doi.org
ρ = correlation coefficient; /10.1016/j.jfluidstructs.2017.12.008.
ρDL = correlation coefficient between along- and Elsharawy, M., T. Stathopoulos, and K. Galal. 2012. “Wind-induced torsional
across-wind loads; loads on low buildings.” J. Wind Eng. Ind. Aerodyn. 104–106 (May):
ρDT = correlation coefficient between along- and 40–48. https://doi.org/10.1016/j.jweia.2012.03.011.
torsional-wind loads; Ha, Y.-C. 2017. “Evaluation of the across-wind load and response of high-
ρLT = correlation coefficient between across- and rise buildings.” [In Korean.] In MIDAS expert column. Seongnam,
torsional-wind loads; Korea: MIDAS Information Technology.
Ha, Y.-C., Y.-S. Kil, S.-K. Hwang, and D.-W. Kim. 2007. “Simplified for-
σ = standard deviation; mulas for estimating the across-wind induced response of rectangular
σMT = standard deviation of response; tall buildings.” [In Korean.] J. Archit. Inst. Korea Struct. Constr. 23 (6):
σMTB = standard deviation of measured base torsional 39–48.
moment; Holmes, J. D. 1987. “Mode shape corrections for dynamic response to
σMTR = standard deviation of resonant response; wind.” Eng. Struct. 9 (3): 210–212. https://doi.org/10.1016/0141-0296
ξ = damping ratio; and (87)90017-4.
Hui, Y., Y. Tamura, and Q. Yang. 2017. “Analysis of interference effects
ξ 1T = damping ratio of the first torsional mode. on torsional moment between two high-rise buildings based on pressure
and flow field measurement.” J. Wind Eng. Ind. Aerodyn. 164 (May):
54–68. https://doi.org/10.1016/j.jweia.2017.02.008.
References ISO. 2009. Wind actions on structures. ISO 4354. Geneva: ISO.
Isyumov, N., and P. C. Case. 2000. “Wind-induced torsional loads and
AIJ (Architectural Institute of Japan). 2004. Recommendations for loads on responses of buildings.” In Proc., Structures Congress, Advanced Tech-
buildings. AIJ-RLB-2004. Tokyo: AIJ. niques in Structural Engineering. Reston, VA: ASCE.
AIJ (Architectural Institute of Japan). 2015. Recommendations for loads on Jeong, S. Y., and T. H.-K. Kang. 2021. “Correlation of directional wind
buildings. AIJ-RLB-2015. Tokyo: AIJ. loads on high-rise buildings with square-shaped plan.” In Proc., 2021
J. Struct. Eng., 2023, 149(11): 03123001

World Congress on Advances in Structural Engineering and Mechanics Solari, G., and L. C. Pagnini. 1999. “Gust buffeting and aeroelastic be-
(ASEM21). Daejeon, South Korea: International Association of Struc- haviour of poles and monotubular towers.” J. Fluids Struct. 13 (7–8):
tural Engineering and Mechanics. 877–905. https://doi.org/10.1006/jfls.1999.0240.
Kareem, A. 1985. “Lateral-torsional motion of tall buildings to wind loads.” Solari, G., T. A. Reinhold, and F. Livesey. 1998. “Investigation of wind
J. Struct. Eng. 111 (11): 2479–2496. https://doi.org/10.1061/(ASCE) actions and effects on the Leaning Tower of Pisa.” Wind Struct. 1 (1):
0733-9445(1985)111:11(2479). 1–23. https://doi.org/10.12989/was.1998.1.1.001.
Katsumura, A., H. Kawai, H. Terazaki, Y. Tamura, and Y. Fugo. 2014. Somekawa, D., H. Kawai, and H. Nishimura. 2014. “Experimental study on
“Modeling of wind-induced torsional moment on tall buildings.” the correlation coefficient for combination of wind loads.” [In Japanese.]
[In Japanese.] In Proc., 23th National Symp. on Wind Engineering, In Proc., Summaries of Technical Papers of Annual Meeting, AIJ, Struc-
421–426. Tokyo: Japan Association for Wind Engineering. tures I, 149–150. Tokyo: Architectural Institute of Japan.
KDS (Korean Design Standard). 2019. Korean design standard. KDS Tallin, A., and B. Ellingwood. 1984. “Serviceability limit states: Wind in-
41-2019. Seoul: KDS. duced vibrations.” J. Struct. Eng. 110 (10): 2424–2437. https://doi.org
Khanduri, A. C., T. Stathopoulos, and C. Bédard. 1998. “Wind-induced in- /10.1061/(ASCE)0733-9445(1984)110:10(2424).
terference effects on buildings—A review of the state-of-the-art.” Eng. Tallin, A., and B. Ellingwood. 1985a. “Analysis of torsional moments on
Struct. 20 (7): 617–630. https://doi.org/10.1016/S0141-0296(97)00066-7. tall buildings.” J. Wind Eng. Ind. Aerodyn. 18 (2): 191–195. https://doi
Kijewski, T., and A. Kareem. 1998. “Dynamic wind effects: A comparative .org/10.1016/0167-6105(85)90097-2.
study of provisions in codes and standards with wind tunnel data.” Wind Tallin, A., and B. Ellingwood. 1985b. “Wind-induced motion of tall build-
Struct. 1 (1): 77–109. https://doi.org/10.12989/was.1998.1.1.077. ings.” Eng. Struct. 7 (4): 245–252. https://doi.org/10.1016/0141-0296
Kwon, D. K., and A. Kareem. 2013. “Comparative study of major (85)90004-5.
international wind codes and standards for wind effects on tall build-
Tamura, Y. 2012. “Aerodynamic database for high-rise buildings.” In
ings.” Eng. Struct. 51 (Jun): 23–35. https://doi.org/10.1016/j.engstruct
Tokyo: Global center of excellence program. Tokyo: Tokyo Polytechnic
.2013.01.008.
Univ.
Li, Y., J. W. Zhang, and Q. S. Li. 2014. “Experimental investigation of
Tamura, Y., H. Kawai, Y. Uematsu, H. Marukawa, K. Fujii, and Y. Taniike.
characteristics of torsional wind loads on rectangular tall buildings.”
1996. “Wind load and wind-induced response estimations in the rec-
Struct. Eng. Mech. 49 (1): 129–145. https://doi.org/10.12989/sem.2014
ommendations for loads on buildings, AIJ 1993.” Eng. Struct. 18 (6):
.49.1.129.
Liang, S., Q. S. Li, S. Liu, L. Zhang, and M. Gu. 2004. “Torsional dynamic 399–411. https://doi.org/10.1016/0141-0296(95)00121-2.
wind loads on rectangular tall buildings.” Eng. Struct. 26 (1): 129–137. Tamura, Y., H. Kikuchi, N. Pillai, and K. Hibi. 2008. “Pressure distributions
https://doi.org/10.1016/j.engstruct.2003.09.004. and flow fields imposing extreme wind force components.” In Proc.,
Lin, Y. K. 1967. Probabilistic theory of structural dynamics, 366. Int. Colloquium on Bluff Body Aerodynamics and Applications (BBAA
New York: McGraw-Hill. VI), 20–24. Amsterdam, Netherlands: Elsevier.
Marukawa, H., T. Ohkuma, and Y. Momomura. 1992. “Across-wind and Yu, X. F., Z. N. Xie, X. Wang, and B. Cai. 2016. “Interference effects be-
torsional acceleration of prismatic high rise buildings.” J. Wind Eng. tween two high-rise buildings on wind-induced torsion.” J. Wind Eng.
Ind. Aerodyn. 42 (1–3): 1139–1150. https://doi.org/10.1016/0167-6105 Ind. Aerodyn. 159 (Dec): 123–133. https://doi.org/10.1016/j.jweia.2016
(92)90121-P. .10.011.
Ryu, H.-J., D.-H. Shin, and Y.-C. Ha. 2020. “Serviceability evaluation Zhang, W. J., Y. L. Xu, and K. C. S. Kwok. 1995. “Interference effects on
methods for high-rise structures considering wind direction.” Wind aeroelastic torsional response of structurally asymmetric tall buildings.”
Struct. 30 (3): 275–288. https://doi.org/10.12989/was.2020.30.3.275. J. Wind Eng. Ind. Aerodyn. 57 (1): 41–61. https://doi.org/10.1016/0167
Saiful Islam, M., B. Ellingwood, and R. B. Corotis. 1992. “Wind-induced -6105(94)00098-X.
response of structurally asymmetric high-rise buildings.” J. Struct. Eng. Zhou, Y., and A. Kareem. 2000. “Torsional load effects on buildings under
118 (1): 207–222. https://doi.org/10.1061/(ASCE)0733-9445(1992) wind.” In Proc., Structures Congress, Advanced Techniques in Struc-
118:1(207). tural Engineering, 1–8. Reston, VA: ASCE.
Solari, G., and A. Kareem. 1998. “On the formulation of ASCE 7-95 gust Zhou, Y., A. Kareem, and M. Gu. 2002. “Mode shape corrections for wind
effect factor.” J. Wind Eng. Ind. Aerodyn. 77–78 (Sep): 673–684. load effects.” J. Eng. Mech. 128 (1): 15–23. https://doi.org/10.1061
https://doi.org/10.1016/S0167-6105(98)00182-2. /(ASCE)0733-9399(2002)128:1(15).
J. Struct. Eng., 2023, 149(11): 03123001

Engineering Review of Wind Induced Torsional Moment and Response

Uploaded by

Copyright:

Available Formats

Engineering Review of Wind Induced Torsional Moment and Response

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Engineering Review of Wind Induced Torsional Moment and Response

Uploaded by

Copyright:

Available Formats

State-of-the-Art Review

Engineering Review of Wind-Induced Torsional

Introduction along-wind and across-wind directions of structures with mass-

© ASCE 03123001-1 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

Dynamic Response Effect and Role of

sively compared with the values from the standards.

Fig. 2. (a) Aerodynamic torsional moment; and (b) torsional response.

© ASCE 03123001-2 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-3 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-4 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

Fig. 4. Variation of eccentricity in ASCE 7.

Table 2. Combination of along- and torsional-wind loads in ASCE 7-22 Background

© ASCE 03123001-5 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

Table 5. Definition of FT and its related parameters in ISO and KDS

Table 6. Definition of FT and its related parameters in AIJ

© ASCE 03123001-6 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0.577 ≥2 — 0.55

where f1T = natural frequency of the first mode of torsional

© ASCE 03123001-7 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

uðtÞ ¼ ust0 Rd sinð2πft − ϕÞ ð6Þ

where ðust Þo = maximum value of static deformation; Rd = defor-

functions with different frequencies. The total response of a linear

© ASCE 03123001-8 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-9 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

Fig. 7. Normalized coefficient of base torsional moment, CMT .

Fig. 9. Normalized PSD of base torsional moment based on wind

Fig. 8. Normalized standard deviation of base torsional moment, CT0 .

of less than 0.1. The results illustrate a good agreement between

© ASCE 03123001-10 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-11 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-12 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-13 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

where q3s = 3-second gust velocity pressure based on basic wind

In ASCE 7-22, duration for averaging to obtain mean and peak

© ASCE 03123001-14 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

GX0 F̄L = mean component of across-wind load;

© ASCE 03123001-15 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

RT = normalized standard deviation of resonant response 49-21. Reston, VA: ASCE.

© ASCE 03123001-16 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

© ASCE 03123001-17 J. Struct. Eng.

J. Struct. Eng., 2023, 149(11): 03123001

You might also like