Engineering Structures 138 (2017) 91–104

Contents lists available at ScienceDirect

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

Performance of wind-excited linked building systems considering the

link-induced structural coupling
Gang Hu a, K.T. Tse c, Jie Song b,c,⇑, Shuguo Liang b
a
CLP Power Wind/Wave Tunnel Facility, Hong Kong University of Science and Technology, Hong Kong, China
b
Department of Engineering Mechanics, School of Civil Engineering, Wuhan University, Wuhan, China
c
Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates the performance of wind-excited linked building systems (LBSs) considering the
Received 20 May 2015 structural coupling caused by horizontal links that connect two adjacent buildings. An analytical evalu-
Revised 11 October 2016 ation model of the LBS and wind force data acquired from wind tunnel tests were used to calculate the
Accepted 2 February 2017
wind-induced response, to assess the performance. After determining the critical wind directions, the
Available online 14 February 2017
effects of link properties (e.g., mass, stiffness, and location) on the acceleration response of LBSs were
comprehensively examined for these wind directions. Results show that the extra link mass tends to
Keywords:
decrease the acceleration response, although it usually increases the displacement response. The trans-
High-rise building
Linked building system
lational acceleration responses of the LBS decrease with increasing link stiffness and location, whereas
Structural coupling the torsional acceleration response of the LBS is usually larger than that in the associated unlinked case.
Wind-induced response As a result, in some cases the resultant acceleration of the LBS exceeds that of the associated unlinked
case, although in many other cases the resultant acceleration is decreased in LBSs. Therefore, researchers
and practicing engineers should exercise caution when designing LBSs to avoid unfavorable acceleration
responses.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction duce inter-building structural coupling for vibrations in the linked

In recent decades, high-rise buildings in metropolises have
often been built near each other because of limited available land.
In turn, these high-rise buildings are increasingly being designed
as linked building systems (LBSs), i.e., systems comprising several
buildings connected by horizontal links such as skybridges, sky-
pools and skygardens. They are usually built to great heights in
order to achieve a grand appearance, so wind-resistance is one of
primary concerns in design practice, especially in typhoon-prone
cities such as Hong Kong and Tokyo.
Compared to a single isolated building, wind-resistant design
for an LBS is relatively complicated because of the interference of
effect on wind buildings in close proximity and the existence of
the link that connects adjacent buildings. The interference effect
will distort wind forces on building surfaces [1,2] and hence the
commonly-used wind force model in the codes of practice may
not applicable for LBSs. The link can provides an additional passage
and an evacuation route in the event of a fire or other emergencies
[3]. Furthermore, from a structural point of view, the link can intro-
⇑ Corresponding author at: Department of Engineering Mechanics, School of Civil
Engineering, Wuhan University, Wuhan, China.
E-mail address: jsongaa@connect.ust.hk (J. Song).
buildings [4,5] and hence affect the modal properties and wind-
induced responses of the LBS [6,7]. Because of this complexity, only
a limited amount of literature can be found on the performance of
wind-excited LBSs [4,8–10].
Apart from the aforementioned LBSs where adjacent buildings
are connected by structural links such as skybridges, skypools
and skygardens, there are a number of other coupled building sys-
tems that are connected by nonstructural elements. These building
systems share considerable similarities to LBSs, so it is worth
reviewing studies on coupled building systems. For example, Klein
and Healey [11], initially, proposed a cable system to connect two
buildings in order to reduce wind-induced oscillations of the two
buildings. Another type of the coupled building systems consisting
of two adjacent buildings connected by control devices was widely
discussed in the field of seismic design of tall buildings. These dis-
cussions proposed using passive control devices [12–17], active
control devices [18,19], or semi-active control devices [20–22] to
connect two adjacent buildings in order to improve seismic resis-
tance. These studies provided an adequate understanding of struc-
tural coupling due to such control devices and their effects on the
structural performance. It should be mentioned, however, that
most of the above studies focus on seismic responses and perfor-
mance of the coupled building system. Results and conclusions

http://dx.doi.org/10.1016/j.engstruct.2017.02.007
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
92 G. Hu et al. / Engineering Structures 138 (2017) 91–104

from the above studies about seismic response and performance of erties) connected by several horizontal links. To highlight the link-
coupled building systems cannot be directly applied to assessing induced structural coupling, the internal structural coupling of
the performance of wind-excited LBSs. This is because one- each building caused by the eccentricities between the mass and
dimensional seismic excitation was adopted in these studies, stiffness centers is eliminated by assuming that the mass center
whereas one distinct characteristic of wind forces on tall buildings of each tower floor coincides with the associated stiffness center.
is that there simultaneously exists along-wind, cross-wind, and tor- Each individual building in the LBS is modeled as a linear multiple
sional force components. Moreover, wind forces vary along the struc- DOF system, in which each floor has three degrees of freedom, i.e.,
tural height, rather than external loads merely at the structural base. two horizontal translations and one rotation about the vertical axis
To evaluate the performance of wind-excited LBSs, Lim’s group [30], as shown in Fig. 1. Each link is regarded as a beam rigidly con-
[8,9] proposed a simplified 3D evaluation model with six degrees nected to the twin towers, although other types of connection,
of freedom for twin buildings connected by a skybridge. Similar such as semi-rigid and hinged connections, are possible [31]. Com-
to the derivation given by Christenson et al. [23], this model was pared to the structural damping of each building, link damping is
created based on assumed structural mode shapes (or trial func- usually insignificant. In addition, the rigid connection will not pro-
tions). The wind-induced responses of the LBSs can then be easily vide considerable damping for the system. Therefore, the damping
calculated using the traditional HFBB based approach requiring of the rigidly-connected link is ignored in this study. The case with
mode shape correction. Their studies have offered a simple and significant link damping is out the scope of this study, as in this
useful formulation of wind-exited LBSs. However, whereas the case the LBS is not a classical damping system.
(first) mode shape of an isolated building is generally simple and
can be estimated to a certain degree, the mode shape of an LBS is 2.2. Formation of the analytical model
usually difficult to estimate as they can be significantly interre-
lated with various link properties, resulting in complicated shapes Each building in the LBS has m floors and the two buildings are
[6,7]. As a result, the inherent assumption about mode shapes and interconnected by n (n 6 m) links at n arbitrary floors. The LBS can
the use of mode shape correction factors may introduce uncertain- be regarded as a 6m degrees of freedom system and the equations
ties for the predicted generalized forces and structural responses of motion for the LBS when subjected to external wind loads can be
[24–29]. Recently, Song’s group developed a 3D analytical evalua- written as
tion model that did not require any assumptions of mode shapes € þ CD
_ þ ½K þ KL D ¼ F
[6] and can allow for the effects of all link properties (i.e., the mass,
½M þ ML D ð1Þ
location, and stiffness). The effects of the link properties on the where M, C, and K are the mass, damping, and stiffness matrices of
modal properties such as frequencies and mode shapes were com- the twin buildings, respectively; ML and KL are additional mass and
prehensively examined using this model to shed light on the effect stiffness matrices due to the link; F = {F1x, F1y, F1h, F2x, F2y, F2h}T is the
of the link-induced structural coupling. external wind force vector, in which Fgs (g = 1, 2; s = x, y or h) is the
Compared to modal properties, the wind-induced responses wind force acting on tower g in the s direction; D = {D1x, D1y, D1h,
(e.g., displacement, acceleration, base moment, stress in an ele- D2x, D2y, D2h}T is the displacement response vector of the LBS, in
ment, etc.) are clearer and more straightforward in illustrating which Dgs is the displacement vector of tower g in the s direction.
the performance of tall buildings subjected to severe wind excita- The details of each matrix are listed as follows:
tions and hence are of great importance and practical concern for



M1 0 C1 0 K1 0
structural engineers. Although a simple case study was conducted M¼ C¼ K¼
in [6] to show the effects of a link on the wind-induced response of 0 M2 6m6m 0 C2 6m6m 0 K2 6m6m
LBSs, the full effects have not yet been examined comprehensively. ð2Þ
Therefore, this paper extends previous research in the analytical
model to examine the effects of link properties on the performance
of wind-excited LBSs due to differences in the link’s mass, axial 2 3 2 3
Mx 0 0 Kxx 0 Kxh
stiffness, bending stiffness, and location. 6 7 6 7
In this study, the 3D analytical model and a pressure measure- M1 ¼ M 2 ¼ 4 0 My 0 5 K1 ¼ K2 ¼ 4 0 Kyy Kyh 5
ment wind tunnel test were briefly described first. After examining 0 0 J 3m3m
Khx Khy Khh 3m3m
the key characteristics of wind pressure on the LBSs, the perfor- ð3Þ
mance of wind-excited LBSs was then assessed in terms of the
acceleration response. The effects of link properties, including the where Mg, Cg, and Kg are the mass, damping, and stiffness matrices
mass, axial stiffness, bending stiffness, and location, on the acceler- of tower g, respectively; Mx, My, and J are the mass and mass
ation response were investigated comprehensively, to provide moment of the inertia sub-matrices of each tower; Krs (r, s = x, y
guidance for initial design of LBSs. A simple example was pre- or h) is the stiffness sub-matrix of each tower.
sented to show the application of the results. The main findings Matrices ML and KL in Eq. (1) are the products of the links and
were summarized in the concluding section. can be derived from the associated structural-property matrices of
each link. The stiffness matrix of each link against the deformations
2. Analytical model of wind-excited linked building system
at the two link ends (i.e., dL1x,p, dL1y,p, dL1h,p, dL2x,p, dL2y,p, and dL2h,p
shown in Fig. 1) can be expressed as [9,32]
Two adjacent tall buildings are horizontally connected by struc-
tural links to form an LBS, as shown in Fig. 1. A 3D analytical model 2 3
ka 0 0 ka 0 0
in matrix form is developed to reproduce the salient features of the 6 0 2 2 7
LBS. This section, summarized from [6], briefly explains the forma- 6 12kb =l 6kb =l 0 12kb =l 6kb =l 7
6 7
tion of the analytical model of the LBS. 6 0 6kb =l ð4 þ bÞkb 0 6kb =l ð2  bÞkb 7
K0l ¼ 6
6 k
7
7
6 a 0 0 ka 0 0 7
6 7
2.1. Assumptions of 3D analytical model 4 0 12kb =l
2
6kb =l 0 12kb =l
2
6kb =l 5
0 6kb =l ð2  bÞkb 0 6kb =l ð4 þ bÞkb
The LBS consists of two identical buildings (with given struc-
ð4Þ
tural properties such as height, mass, stiffness, and damping prop-
 G. Hu et al. / Engineering Structures 138 (2017) 91–104 93

dL1y,p
dL1θ,p dL2y,p
dL2θ,p
dL1x,p dL2x,p
3. p-th
link
d1y,j
d2y,j
d1θ,j
j-th floor
d1x,j d2θ,j j-th floor
z y d2x,j

x Tower 1
Tower 2

Fig. 1. Schematic of an LBS and the associated degrees of freedom.

where ka = EA/l is the axial stiffness of the link; kb = EI’/l is the hor- where kl,p(d, f) (p = 1, 2, . . ., n) is the element of the stiffness matrix
izontal in-plane bending stiffness of the link, in which I’ = I/(1 + b) is for the p-th link calculated by Eq. (5).
the effective area moment of inertia of the link, and b is the shear After obtaining the 6n  6n stiffness matrix KLinks for n links,
deformation constant; A and I are the cross-sectional area and the matrix KLinks needs to be converted to a larger 6m  6m matrix
area moment of inertia of the link, respectively; E is Young’s mod- KL having the same size as K in Eq. (1). Meanwhile, the location
ulus and l is the link span. of each link should be taken into account when assembling the var-
In order to assemble K’l (formed with respect to the DOF at the ious matrices. Therefore, a location indicator matrix E is intro-
two link ends) with the stiffness matrix K of the twin towers duced, through which KL can be determined as:
(formed at the mass center of each floor), K’l should be transformed
KL ¼ EKLinks ET ð9Þ
to the mass center of the related floor [33]:
where E is defined as
Kl ¼ ½kl ðd; f Þ66 ¼ HT K0l H ð5Þ
E ¼ diagfE1 ; E1 ; E1 ; E2 ; E2 ; E2 g6m6n ð10Þ
where Kl is the resulting transformed stiffness matrix of the link,
and kl(d, f) is the element of matrix Kl at the d-th row and the f- here, Eg is a location indicator sub-matrix to show the link location.
th column. Matrix H can be expressed as follows: Taking tower 1 for example, E1 = [e1, e2,. . ., ep,. . ., en]mn and each
2 3 2 3 column of E1 is a vector of

 1 0 0 1 0 0
H1 0 6 7 6 7
H¼ ; H1 ¼ 4 0 1 r 1 5; H2 ¼ 4 0 1 r 2 5 ep ¼ f 0;    0; 1; 0    0 gT ð11Þ
0 H2
0 0 1 0 0 1 where all the elements of vector ep are zero, except the i-th element
ð6Þ that equals one, suggesting that the p-th link is connected to the i-th
floor of Tower 1.
where Hg (g = 1 or 2) is the transformation sub-matrix for tower g,
Making use of the same location indicator matrix E, the addi-
and rg is the distance between the mass center of the tower floor
tional mass matrix ML in Eq. (1) can be determined as
and the link end.
Similarly, for n links at n arbitrary stories, n stiffness matrices, ML ¼ EMLinks ET ð12Þ
denoted as Kl, p (p = 1, 2, . . ., n), can be obtained. These n matrices
where MLinks is the 6n  6n mass matrix for n links, calculated by
are then packaged together in line with the DOF sequence of K in
lumping equally the mass of each link at the two mass centers of
Eq. (1)
the two connecting tower floors:
X L1 Y L1    HL2
2 3 1
KL11 KL12    KL16 MLinks ¼  diagð1; 1; r 21 ; 1; 1; r 22 Þ  Ml ð13Þ
6 7 2
..
KLinks ¼ 6
6 KL21 KL22 . 7
7 ð7Þ where Ml = diag(ml,1, ml,2,. . ., ml,n), in which ml,p (p = 1, 2, . . ., n) is
6. . . .. 7
6. 7 the mass of the p-th link.
4. . . 5
Once all structural-property matrices of the twin buildings and
KL61     KL66 6n6n links have been obtained, Eq. (1) can then be decoupled to several
where KLdf (d, f = 1, 2, . . ., 6) is an n  n diagonal sub-matrix, whose single degree of freedom equations by the classical mode superpo-
diagonal elements are the set of corresponding elements kl, p(d, f) of sition method [28,29,34]
the stiffness matrix Kl,p: F j ðtÞ
€j ðtÞ þ 2nj xj q_ j ðtÞ þ x2j qj ðtÞ ¼
q ð14Þ
KLdf ¼ diag½kl;1 ðd; f Þ; kl;2 ðd; f Þ;    ; kl;n ðd; f Þ ð8Þ M j
94 G. Hu et al. / Engineering Structures 138 (2017) 91–104

where qj, xj, nj, F⁄j and M⁄j are the normal coordinate, frequency, are connected by a three-story link (15 m long and 18 m wide).
damping ratio, generalized wind force, generalized mass of the j- The associated rigid model with a downscale length ratio of
th mode, respectively, and 1:400 is shown in Fig. 3a. Definitions of the wind direction (a),
coordinate, and face name are also shown in Fig. 3b.
X
m X X
F j ðtÞ ¼ F gs ðzi ; tÞUgs;j ðzi Þ ð15Þ The measurements were repeated from 0° to 350° at 10° incre-
i¼1 s¼x;y;hg¼1;2 ments; wind directions of 45°, 135°, 225°, and 315° were also
added, for a total of 40 wind directions. Wind pressure data on
m Xn
X h i o the faces of the LBS surfaces were collected synchronously. The
M j ¼ mðzi Þ U2gx;j ðzi Þ þ U2gy;j ðzi Þ þ Jðzi ÞU2gh;j ðzi Þ : ð16Þ layer-by-layer spatiotemporal wind force time histories Fgs(zi, t)
i¼1 g¼1;2
were then determined by integrating wind pressure over the corre-
In these equations, zi is the height of the i-th floor of the tower; Usg,j sponding effective surface area [27,35]. In order to examine the
is the j-th mode shape of tower g in the s direction. The frequency xj effect of link location on aerodynamic forces, six cases were con-
and mode shape Usg,j for the LBS are determined by solving the cor- sidered, as listed in Table 1.
responding characteristic equation of Eq. (1)
3.2. Wind pressure characteristics
ð½K þ KL   x2 ½M þ ML ÞU ¼ 0 ð17Þ
Although structural engineers are mostly concerned with
where x is the structural frequency; U = {U1x, U1y, U1h, U2x,U2y,
resulting wind-induced structural responses to evaluate the per-
U2h}T is the associated mode shape, in which Usg is the mode shape
formance of tall buildings, it is worthwhile to briefly investigate
of tower g in the s direction.
the aerodynamic characteristics of wind pressure buildings.
The normal coordinate qj in Eq. (14) can be solved in the time
The contours of mean wind pressure on the inner face of Tower
domain by the step-by-step method. The wind-induced responses,
1 (i.e., face E1 shown in Fig. 3b) are shown in Fig. 4 for all the six
such as displacement and acceleration, of tower g in the s direction
cases, when a = 0°. As can be seen, for all cases, link location affects
can be then calculated by summing all modal components
the wind pressure distribution on the face. The suction forces on
8X
>
> qj ðtÞugx;j ðzÞ the area of face E1 around the link are decreased, compared to
9 >
>
>
> j the cases without a link (i.e., case NL). The effect of link location,
Dgx ðz; tÞ >
= >
<X
X qj ðtÞugy;j ðzÞ however, is relatively local, since the size of the link is small. The
Dgy ðz; tÞ ¼ Dgs;j ¼ ð18aÞ existence of the link with small size is thus likely to have no signif-
>
; >
>Xj
Dgh ðz; tÞ j >
> icant effect on the global wind force parameters such as base shear
>
> qj ðtÞugh;j ðzÞ
>
: forces and moments. This can be illustrated by comparing the
j
spectra of based moments for all the six cases, as shown in Fig. 5.
8X It can be observed that the spectra of cases with a link do not devi-
>
> €j ðtÞugx;j ðzÞ
q
9 >
> ate much from that of the case without a link, especially for the
€
Dgx ðz; tÞ > >
> j
= X >
<X along-wind base moment spectrum. In view of this, in the subse-
€ € €j ðtÞugy;j ðzÞ
q quent analysis, wind forces on the case without a link (i.e., case
Dgy ðz; tÞ ¼ D gs;j ¼ ð18bÞ
>
; >
> j
NL) are used to represent those on the other five cases with a link.
€ gh ðz; tÞ j >
> X
D >
> €j ðtÞugh;j ðzÞ
>
: q Characteristics of wind forces on LBSs are then briefly discussed
j for two critical scenarios, i.e., a = 0° and 90°, to understand the
main differences between wind loads on a single isolated building
€ gs;j are the j-th modal displacement and accelera-
where Dgs,j and D and on an LBS.
tion response vectors of tower g in the s direction, respectively. The distribution of mean wind pressures on the faces of the
The accuracy of the above 3D analytical model has been vali- tested model is first examined for a = 0°, in terms of the contours
dated in [6], by comparing the modal properties calculated by of the pressure coefficients, as shown in Fig. 6. For this wind direc-
the analytical model and those provided by FEM model for a typical tion, the twin towers are in a side-by-side symmetric arrangement,
LBS. so in theory the distribution of mean wind pressures on the twin
towers are identical. This can be illustrated by Fig. 6, although
3. Wind tunnel test measurement slight differences can be found due to measurement errors. Similar
to pressures on a single isolated building [36], mean pressures on
3.1. Wind tunnel measurement the two windward faces of the LBS (i.e., S1 and S2) are positive

To determine wind-induced responses of the LBS, information

on the spatiotemporal wind forces F acting on the LBS is also nec- 1
essary. Therefore, a wind tunnel test, using a synchronous multi- wind speed
pressure measurement system (SMPMS), was carried out in the turbulence intensity
0.75
boundary layer wind tunnel at the CLP Power Wind/Wave Tunnel
AS/NZS 1170.2:2002, Cat.3
Facility at the Hong Kong University of Science and Technology
z /H

to obtain the associated wind force information. The mean velocity 0.5
of the approaching wind, which follows a power law function with
an exponent of 0.2 and represents largely an open terrain (i.e. Cat- 0.25
egory 3) defined in AS/NZS 1170.2:2002, was calibrated in the
wind tunnel. The mean wind speed and the longitudinal turbu-
lence intensity at the top of the building (160 m in prototype) were 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
42.8 m/s and 13.9%, respectively. The resulting calibrated
Normalized wind characteristics
approaching wind profiles are shown in Fig. 2. The LBS identical
to that in [6] is used in the wind tunnel. Each tower in the LBS is Fig. 2. Simulated mean wind speed and turbulence intensity profiles in the wind
30 m  30 m in plan and 160 m in height, and the two towers tunnel.
 G. Hu et al. / Engineering Structures 138 (2017) 91–104 95

(a)

(b)
N1 N2
E1
W1 Tower 1 Tower 2 E2
W2
S1 S2
y
θ
x wind

Fig. 3. (a) Downscale structure model in the wind tunnel and (b) the associated wind direction and coordinate.

Table 1 through the gap and then increases the pressures on the area of the
Configuration of six cases. windward faces close to the gap. As a result, the distribution of the
Cases Location of link wind pressure on the two windward faces is shifted slightly
NL Without a link, just twin buildings
inwards. Second, the pressure distribution on the two inner faces
L1 152–160 m (380–400 mm in model scale), floor 39–40 of the LBS (i.e., faces E1 and W2) differs significantly from that
L2 124–132 m (310–330 mm in model scale), floor 32–33 for a single building. For the single building, the gradient of the
L3 92–100 m (230–250 mm in model scale), floor 24–25 pressure is usually gentle. By contrast, the gradient of the pressure
L4 52–60 m (130–150 mm in model scale), floor 14–15
on the two inner faces of the LBS is dramatic, especially in the area
L5 24–32 m (60–80 mm in model scale), floor 7–8
near to the windward edge of the two faces. The severely asym-
metric distribution of the pressure will definitely cause large mean
torsion on the LBS. In addition, it can be observed that for the zone
and those on the other three faces are negative. It should be noted, near to the windward edges of the two inner faces, the absolute
however, that the distribution of the pressures on the LBS differs values of the mean pressure coefficients can be as large as 1.4,
that for the associated single building. First, the pressure on the which is much larger than those on the same zone of the single
windward faces (S1 and S2) is skewed, rather than symmetric, as building. In other words, the mean suction forces on the inner faces
shown in Fig. 6. This can be explained by the channeling effect of the LBS are more considerable than those on the side faces of the
caused by the inter-building gap, which accelerates the wind flows single building.

-0.4 - -1.4 -1.4

-0.7 -0.7 1
-1.4 -0.1 -1
-1.4 -0.3 -1
-0.8 -1
140
-1.2 -0.3
-0.1
-1 -0.2 -1.2
-0.7 0.5
120 -0.4
-1 -1.2
-0.8
-0.4 -0.6
-0.2 0
100 -0.7 -0.1
Height

-0.7 -1.2
-0.3
-0.6 -1.2 -0.6
-0.5 -0.5
80 -0.5
-0.8
-0.6 -1 -0.7
-0.5 -0.4
-0.5 -1.2 -0.2
60 -0.8 -0.6
-0.1 -0.8 -0.5 -1
-0.8
-0.3
40 -0.4
-0.4 -0.2 -0.1
-0.4 -0.4 -1.5
-10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10
NL L1 L2 L3 L4 L5

Fig. 4. Pressure on face E1 for all cases when a = 0°.

96 G. Hu et al. / Engineering Structures 138 (2017) 91–104

-2 -2 -2
10 10 10

-3 -3 -3
10 10 10
fSM/(0.5ρ v2HBH2) 2

fSM/(0.5ρ v2HBH2) 2

fSM/(0.5ρ v2HB2H) 2
-4 -4 -4
10 10 10

NL
-5 -5 -5
10 L1 10 10
L2
-6 L3 -6 -6
10 10 10
L4
L5
-7 -7 -7
10 10 10
0.01 0.1 0.5 0.01 0.1 0.5 0.01 0.1 0.5
reduced frequency fB/vH reduced frequency fB/vH reduced frequency fB/vH

(a) Along-wind (b) Cross-wind (c) Torsion

Fig. 5. Spectra of base moments and torsion for all the six cases when a = 0°.

0.8
0.8 1
-1.4
-1.2 -0.5
140 -0.5
-1.2 -0.8
-1.2
-1 0.5
-0.8
120
0.8 -0.8 -1
0.8
-0.4
-0.8 0
100 -0.4
Height

-0.7

-0.6
0.7 -0.7
80 0.7 -0.5
-0.6
-0.7
-0.4
-0.5 -0.5
60 0.6 -1
0.6
-0.7 -0.4
-0.4
40 0.4 0.5
0.5 -0.4
0.5 -0.6 -0.4 -1.5

-10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10

S1 S2 E1 W2 W1 E2 N1 N2

Fig. 6. Distribution of mean pressure coefficient on the faces of the example LBS for a = 0°.

The distribution of mean wind pressures for another wind direc- 4. Performance of Wind-excited LBSs
tion a = 90° is shown in Fig. 7. When a = 90°, the two towers are in a
tandem setup. Tower 1 is the upstream building and the Tower 2 is 4.1. Measure of Wind-induced responses
the downstream building. For this wind direction, the approaching
wind directly impacts face W1. As a result, the mean pressure on An LBS under the excitation of external wind forces has many
the face is positive. Furthermore, the mean pressure on the face is response indices, such as displacement, acceleration, base moment,
rather even due to the symmetry, which is different from the skewed stress of elements, and so on. Flexible tall buildings originally
distribution of the pressure on faces S1 and S2 for a = 0°. Although designed to satisfy the customary static lateral displacement
face W2 is also normal to the approaching wind, the pressure on requirements could still oscillate excessively during a windstorm,
the face is totally negative, because the face is completely sub- thereby causing discomfort to occupants [37,38]. Therefore, human
merged in the upstream shear layer. Not surprisingly, the pressure comfort is a major part of a tall building’s wind-resistant design.
coefficients on the four side faces S1, N1, S2, and N2 are negative. Because human comfort is usually evaluated using acceleration
Whereas the magnitude of the mean pressures on side faces S1 responses [39], the standard deviation of acceleration response is
and N1 of Tower 1 is similar to that of a single building, the magni- calculated in this work to reflect the performance of an wind-
tude of the mean pressures on side faces S2 and N2 of Tower 2 is excited LBS. Allowing for the fact that there are two towers in
much smaller than that for a single building. This is because the vor- the LBS, the maximal standard deviation rmax of the acceleration
tex shedding rolls up from the two outside edges of Tower 1 response of any story of either building is used to show the perfor-
(upstream building) and decays gradually along the direction of mance of wind-excited LBS, i.e.,
the coming wind flow.
 G. Hu et al. / Engineering Structures 138 (2017) 91–104 97

-0.1
-0.5 -0.8 -0.2 1
0.9 -0.3
-0.2 -0.3
140
-0.3
-0.4 -0.2
-0.8 -0.8 -0.2 0.5
-0.8 -0.2
120
0.8 -0.4
-
-0.8 0
100
Height

-0.8
0.7 -0.7
-0.8 -0.7
-0.4 -0.5
80
0.6 -0.2 -0.2

60 0.6 -0.6
-0.7 -0.7 -1

-0.5 -0.2
-0.4 -0.2 -0.3
40 0.5 -0.6 -0.3
0.5 -0.2 -1.5
-0.5
-10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10
W1 E1 S1 N1 S2 N2 W2 E2

Fig. 7. Distribution of mean pressure coefficient on the faces of the example LBS for a = 90°.

rmax ¼ maxfrTower1;1 ; rTower1;2 ;    ; rTower1;40 ;
 rTower2;1 ; rTower2;2 ;    ; rTower2;40 g
where rTower1,i and rTower2,i are the standard deviations of accelera-
tion response of i-th floor of Tower 1 and Tower 2, respectively.
In order to highlight the effects of link-induced structural cou-
pling, the structural properties of the twin towers (without a link)
are fixed and identical to those in [6]. The first three vibration fre-
quencies of each individual tower are 0.244 Hz (x direction),
0.244 Hz (y direction), and 0.411 Hz (h direction), respectively. A
three-story link similar to that in [6] is still used to connect the
twin towers, but the link’s structural properties (i.e., the mass, axial
stiffness, bending stiffness, and location) are not fixed, but varying
in terms of the following ratios:
mlink ka kb hlink
km ¼ kka ¼ kkb ¼ kh ¼
mtower ke;b ke;t htower
where mlink is the mass of each link; hlink is the link elevation (here
is the elevation of the top of the three-story link); mtower and htower
are the tower mass and height, respectively; ke,b is the equivalent
lateral stiffness of the tower (i.e., the required lateral force at the
top of the tower to make the tower sway one unit in displacement
at the same elevation); ke,t is the equivalent torsional stiffness of the
tower (i.e., the required torque at the top to make the tower rotate
one degree at the top of the building) [34]. It is worth mentioning
that for the example LBS in [6] (i.e., the twin buildings connected
by a top three-story link), the values of km, kka and kkb are about
0.006, 800, and 45, respectively.
4.2. Critical wind directions
It is clear that wind forces for different wind directions are dif-
ferent and hence the performance of wind-excited LBS may vary
with wind direction. It is impractical to present the performance
of wind-excited LBS for all 40 wind directions here. Therefore,
before investigating the effects of the link’s properties on the
wind-induced responses for all wind directions, effects of wind
direction on the wind-induced responses are first investigated to
determine the critical wind direction. The example LBS (denoting
as kh = 1) in [6] is used to calculate the wind-induced response.
In addition, the same twin buildings with the same three-story link
at the middle elevation (denoting as kh = 0.5) is also considered to
avoid making one-sided conclusions. The damping ratio n has been
set to 2.5% for each mode of the LBS. The contribution of the first
ð19Þ six modes is taken into account to calculate the total responses.
The variations of the maximal standard deviation rmax with
wind directions are shown in Fig. 8a–c, for the responses in the x,
y and h directions, respectively. To maintain dimensional consis-
tency in the three directions, the torsional acceleration response
pffiffiffi
in the h direction has been multiplied by a length of 15 15 2m
(similarly hereinafter). As can be seen, the values of rmax in the x
direction for both cases (kh = 0.5, 1) show a maximum when
a = 0° (or 180°), whereas those in the y direction show a maximum
when a = 90° (or 270°). This suggests that for the translational
responses, a = 0° and 90° are two critical wind directions. Com-
pared to the translational response, the torsional response shown
in Fig. 8c is relatively small. Still, a maximum can be found when
a = 90° (or 270°) for the two cases, indicating that this wind direc-
ð20Þ
tion is critical for the torsional response.
Apart from the above directional acceleration responses at the
mass center, the resultant acceleration response at the corner
(i.e., the combination of responses in the three principal directions)
which undergoes the largest resultant acceleration response is
more important and straightforward for assessing the building ser-
viceability performance such as human comfort. Thus, variations of
rmax for the resultant acceleration calculated according to [40] are
also investigated for all wind directions, as shown in Fig. 8d. It can
be observed from the figure that for the two cases (i.e., kh = 0.5 and
1), rmax exhibits two obvious peaks at a = 0° (180°) and 90° (270°),
whereas rmax is relatively small when wind is oblique to the LBS, in
particular when a = 45° (or 135°, 225°, 315°).
From the above discussion, it is obvious that a = 0° (180°) and
90° (270°) are the two critical wind directions for the design of
LBSs. Furthermore, it should be mentioned that if there is a domi-
nant wind direction in the site of the proposed construction, and
the wind direction is known, it is favorable to design the orienta-
tion of an LBS as a = 45°. In this way, the LBS will likely have the
minimal wind-induced response.
4.3. Effect of the link-induced structural coupling on wind-induced
responses
4.3.1. Effect of link mass
Previous case studies [6] indicate that a link mass added onto an
LBS tends to lower the fundamental frequency of the LBS and hence
98 G. Hu et al. / Engineering Structures 138 (2017) 91–104

180 180
(a) 150 210 (b) 150 210

120 λh =0.5 240 120 240

λh =0.5

90 λh =1 270 90 λh =1 270
10 10
20 20
60 30 300 60 30 300

30 330 30 330
0 0

180 180
(c) 150 210
(d) 150 210
225
135
120 240 120 240
λh =0.5
λh =1
λh =0.5

90 270 90 270
10 λh = 1
10 20
20 30
60 300 60 300
30
45 315
30 330 30 330
(unit: milli-g)
0 0

Fig. 8. Variation of rmax with wind direction a for: (a) response in the x direction; (b) response in the y direction; (c) response in the h direction; and (d) resultant response.

has a tendency to increase the wind-induced displacement
response. It is important to note, however, that the link mass will
not necessarily lead to an increase in wind-induced acceleration
responses. Therefore, it is worthwhile and necessary to further
investigate the effect of link mass on the acceleration response of
LBSs.
The variations of rmax with mass ratio km are examined for four
distinct scenarios, although the link is always at the top in the four
scenarios. As shown in Fig. 9a and b, the wind direction is identical
to a = 0°, but link properties are totally different (i.e., the link stiff-
ness in Fig. 9a is insignificant, whereas that in Fig. 9b is consider-
able). In Fig. 9c and d, the link properties are the same as those
in Fig. 9b, while the wind directions differ that in Fig. 9a and b.
As can be seen from Fig. 9a, for a = 0°, the effect of link mass on
the two translational acceleration responses is marginal. This point
can be reinforced by Fig. 9b and c, where the value of rmax is not
very sensitive to link mass ratio km. However, it is a little surprising
that a slight decrease in the value of rmax can be found, although it
was reported that the displacement response increased with km.
This is because although link mass tends to decrease the structural
frequency x, thus increasing structural displacement response d,
the structural frequency x is also related to the acceleration
response. For simplicity, taking harmonic response for example,
the acceleration response can be expressed as a = x2d. Therefore,
the decrease in x can cause decrease in the acceleration, which
could cancel out the associated increase in d, or even be larger than
the increase. As a result, the acceleration response shows a minor
decrease with increasing km in Fig. 9a–c. More interestingly, the
decrease is significant in Fig. 9d for a = 90°, especially for the
cross-wind acceleration response accy. As can be seen, rmax for accy
decreases gradually from 15.7 milli-g to 10.7 milli-g (i.e., decreas-
ing by 32%) when km increases from 0.001 to 0.05. Therefore,
although the effect of link mass is not significant in some cases
(e.g., a = 0° and 45°), the additional link mass can cause a consider-
able decrease in the acceleration response in other cases (e.g.,
a = 90°).
4.3.2. Integrated effect of link stiffness and location on directional
acceleration
In this section, the integrated effects of the link’s axial stiffness,
bending stiffness, and location are investigated together, by chang-
ing kka, kkb, and kh while fixing km. The discussion in Section 4.3.1
shows that for most of cases, the link’s mass tends to decrease
the acceleration response of an LBS. To be on the safe side, a rela-
tive small value of km (km = 0.001) is used in this section. It should
be mentioned that although the link location affects the LBSs’ wind
loads, the effect is negligible as discussed in Section 3.2, as a small-
sized three-story link is used for all analysis. In view of this fact and
that the focus of this paper is link-induced structural coupling, for
each wind direction, LBSs with different link properties are then
assumed to be subjected to the same wind excitation, regardless
of link location. In this way, differences between responses of the
cases with different link properties can then be attributed to the
link-induced structural coupling alone. Allowing the findings to
be applicable to other similar LBSs, a response ratio kacc of the max-
imum standard deviation rmax of acceleration response of the LBS
to that of the associated unlinked case (i.e., two towers without a
link) is calculated.
 G. Hu et al. / Engineering Structures 138 (2017) 91–104 99

(a) 25 (b) 25
acc x
20 20
acc y
acc x

max σacc (milli-g)

max σacc (milli-g)

15 15
acc y

10 10
y
y
5 x
θ 5
x
θ
o
wind α=0 wind α=0
o

0 0
0.001 0.005 0.01 0.05 0.001 0.005 0.01 0.05
λm λm

(c) 25 (d) 25
acc x y acc x
y wind
20 acc y 20 x acc y
x
o θ
α = 90

max σacc (milli-g)

θ
max σacc (milli-g)

15 15
o
wind α = 45
10 10

5 5

0 0
0.001 0.005 0.01 0.05 0.001 0.005 0.01 0.05
λm λm

Fig. 9. Effect of link mass on the two translational acceleration responses of LBSs with top link for: (a) a = 0°, kka = 0.01, kkb = 0.001; (b) a = 0°, kka = 100, kkb = 10; (c) a = 45°,
kka = 100, kkb = 10; and (d) a = 90°, kka = 100, kkb = 10.

rmax;linked (along-wind direction) are more synchronous than those in the x

kacc ¼ % ð21Þ
rmax;unlinked
where rmax,linked is the maximum standard deviation of the acceler-
ation response of the LBS, and rmax,unlinked is the maximum standard
deviation of the associated unlinked case.
Fig. 10 illustrates the effect of link properties on kacc of LBSs, for
one critical wind direction a = 0°. For the two translational acceler-
ation responses (i.e., responses in the x and y directions shown in
Fig. 10a and b), the values of kacc are smaller than 100%, indicating
that installing the link can reduce the two translational accelera-
tion responses. To put it simply, values of kacc in the x and y direc-
tions decrease with increasing kka (or kkb) and kh, indicating that
the translational acceleration responses decrease with increasing
link stiffness and location. However, it should be noted that when
kka > 102 (or kkb > 10) and kh > 0.7, both translational acceleration
responses are no longer hypersensitive to link properties, suggest-
ing that further increasing kh and kka (or kkb) after those values has
no significant effect on the two translational acceleration
responses. Additionally, comparing Fig. 10a and b shows that the
maximal reduction of the acceleration response in the x direction
(i.e., the cross-wind response) is far larger than the maximal reduc-
tion in the y direction (i.e., the along-wind direction). This differ-
ence can be explained by the correlation between the associated
wind forces. For a = 0°, it is clear that the correlation between
the along-wind forces on the two towers is more significant than
that between the cross-wind forces. From a probabilistic point of
view, the structural vibrations of the two towers in the y direction
direction (cross-wind direction). For the case with large synchrony
between vibrations of the two towers, the relative deformation
between two ends of the link is relatively small, so the effect of
the link is marginal. On the contrary, the synchrony between
motions in the x direction (cross-wind direction) is small, so the
reduction of the acceleration response in the y direction is found
to be as much as 42%.
Effects of the link’s bending stiffness and location on the tor-
sional acceleration response for a = 0° are presented in Fig. 10c.
Similar to the two translational accelerations, the torsional acceler-
ation response ratio kacc, in general, decreases as the link is placed
higher and higher (i.e., larger kh), and with increasing bending stiff-
ness (kkb), for most of cases. However, it should be noted that in the
bottom left region encircled by the dashed lines, the torsional
acceleration response ratio kacc exceeds 100%. Similar increases in
the torsional displacement response were also reported in [6]
and the increase was attributed to the linked-induced coupled
motion between the y and h directions. This suggests that installing
a link with structural properties within the encircled region can
actually increase the torsional acceleration response of an LBS,
rather than reducing it, which clearly emphasizes the importance
of designing link with care. Furthermore, it should be mentioned
that when kkb is very large (10–100), the value of kacc does not
decrease when kh increases from 0.7 to 1. When kh increases from
0.9 to 1, the value of kh even increases. In other words, in terms of
the torsional response, placing the link at the very top of the build-
ing does not necessarily provide the best performance.
100 G. Hu et al. / Engineering Structures 138 (2017) 91–104

Max RMS acceleration x Max RMS acceleration y

(a) 1 (b) 1
90
92
0.8 0.8
58
94
60
0.6 0.6
96
λh

λh
70 y 98
0.4 0.4
80
x
θ 99
90
o
0.2 0.2 wind α=0
99

-2 -1 0 1 2 3 -3 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
λka λkb

Max RMS acceleration θ

(c) 1
60
80

0.8 99
58

0.6 70
λh

100
0.4 90

0.2

-3 -2 -1 0 1 2
10 10 10 10 10 10
λkb

Fig. 10. Integrated effect of link location and stiffness on the acceleration response (at mass center) in the three directions for a = 0°: (a) x direction; (b) y direction; and (c) h
direction.

The effects of link properties on the structural response for
another critical wind direction a = 90° are also investigated in
terms of the variation of kacc with kka, kkb, and kh, as shown in
Fig. 11. It can be observed that the two translational acceleration
responses (shown in Fig. 11a and b) in the upper right region above
the dashed line decrease with increasing link stiffness (kka and kkb)
and location (kh), which is similar to that shown in Fig. 10. In addi-
tion, when kka > 100 and kkb > 10, increasing kh above 0.8 or
increasing kka and kkb will not cause significant further reduction
in the two translational accelerations, which echoes that shown
in Fig. 10a and b. However, it should be noted that the values of kacc
for the two translational acceleration responses exceed 100% on
the bottom left region below the dashed line, indicating that in
these cases, the translational acceleration responses of the LBSs
are larger than those in the associated unlinked case. This reiter-
ates the need for caution in the design of LBSs, because under cer-
tain design conditions, the structural performance of the LBS can
be worse than that of the associated unlinked cases.
The torsional acceleration responses for a = 90° (shown in
Fig. 11c) can illustrate this point more clearly. As can be seen, all
the values of kacc shown in Fig. 11c are larger than 100%. When
the link is at the top and has a moderate stiffness (kkb = 0.1), the
value of kacc can be as high as 140%. This suggests that the torsional
acceleration in the LBS can be increased significantly, compared to
that of the associated unlinked case. Similarly, the increase is
caused by the linked-induced coupling between the y and h direc-
tions. Therefore, the link stiffness, in particular the bending stiff-
ness, should be carefully designed to avoid undesirable
acceleration responses. Furthermore, it worthwhile to mentioned
that unlike the variations shown in Fig. 10c for a = 0°, Fig. 10c
shows that for virtually all cases, the torsional acceleration
response decreases with increasing link location. This actually
has already been illustrated in Fig. 6c. As can be seen from Fig. 6c,
for a = 90°, the case with top link (kh = 1, the blue line) has larger
response than that with middle link (kh = 0.5, the purple line).
In addition to the two critical wind directions a = 0° and 90°,
discussion in Section 4.2 demonstrates that a = 45° (or 135°,
225°, 315°) is also interesting since this wind direction can provide
minimal structural responses. For a = 0° and 90°, the approaching
wind is normal to the building face, whereas for a = 45°, the
approaching wind is oblique to the building. Clearly, characteristics
of wind forces on the LBS for a = 45° differ significantly from those
for a = 0° and 90°. Accordingly, the effects of the link properties for
a = 45° could differ from those for a = 0° and 90° and deserve to be
investigated.
The variations of kacc for a = 45° with kka, kkb, and kh are shown
in Fig. 12, for the responses in three directions. Similar to the vari-
ations of kacc for a = 0°, values of kacc for the two translational
responses decrease significantly with increasing kka (or kkb) and
kh. Furthermore, the decrease in kacc for the acceleration in the x
direction is much more considerable than that in the y direction,
because of the marked difference between the wind forces in the
x direction. Similarly, for all kh, the values of kacc in the x and y
directions are not very sensitive to kka and kkb, when kka > 100
 G. Hu et al. / Engineering Structures 138 (2017) 91–104 101

Max RMS acceleration x Max RMS acceleration y

(a) 1 (b) 1
50
70
66
48
0.8 0.8
60
80
0.6 70
0.6

λh
λh

80 90

0.4 90 0.4 100

100 103
102
0.2 0.2

-2 -1 0 1 2 3 -3 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
λ ka λ kb

Max RMS acceleration θ

(c) 1
140

135
0.8
125
y
o
0.6 115 α = 90
λh

x
θ
120 wind
105 1
0.4

101
105
0.2

-3 -2 -1 0 1 2
10 10 10 10 10 10
λ kb

Fig. 11. Integrated effect of link location and stiffness on the acceleration response (at mass center) in the three directions for a = 90°: (a) x direction; (b) y direction; and (c) h
direction.

and kkb > 10. Above kh = 0.8, the values of kacc almost level off when
kka > 100 and kkb > 1, suggesting that in these cases, raising the link
above kh = 0.8 will not provide significant reduction in the
responses. The variation of kacc for the torsional response, however,
is very similar to those for a = 90°. As can be seen from Fig. 12c, the
value of kacc in the most cases increases gradually with increasing
kh. Similar to that shown in Fig. 11c, a maximum of 133% occurs
when kkb = 0.1 and kh = 1, suggesting that the top link tends to
cause large torsional response.
4.3.3. Effects of link stiffness and location on resultant acceleration
As mentioned earlier, apart from the directional acceleration
responses discussed in the above section, the resultant acceleration
response is more important and straightforward for assessing the
building serviceability performance such as human comfort. Fur-
thermore, the effects of the link properties are not consistent for
the responses in the three directions, so it is difficult to obtain gen-
eral conclusions from the directional responses. Therefore, this sec-
tion investigates the effect of link stiffness and location on the
resultant acceleration response for the two critical wind directions
(a = 0° and 90°) and the special wind direction (a = 45°), as shown
in Fig. 13.
It can be seen from Fig. 13a and b that the variations of kacc are
very similar for a = 0° and 45°, albeit with minor differences
between the values of kacc. In general, a link with a higher location
and a larger stiffness tends to result in smaller resultant responses.
This trend almost follows that for the translational responses
shown in Fig. 12a and b, since translational responses are domi-
nant (as shown in Fig. 6). When kh > 0.7, a link with kka = 100 and
kkb = 10 can decrease the resultant acceleration response by about
30% and 40%, for a = 0° and 45°, respectively. This indicates the sig-
nificant effect of the link on reducing the resultant acceleration
response and improving the performance of LBSs. However, it
should be noted that when link stiffness exceeds a certain value
(i.e., when kka > 100, kkb > 10), kacc in Fig. 13a and b is no longer very
sensitive to kka and kkb. This means that in the design of an LBS,
there is no need for the link to have a very high stiffness.
The effects of link stiffness and location for a = 90° is illustrated
in Fig. 13c. As can be seen, values of kacc in the upper right region
above the dashed line (region I) in Fig. 13c decrease when values
of kh, kka, and kkb increase, which is similar to that for a = 0° and
45°. When kka > 100 and kkb > 10, value of kacc is not sensitive to
kka and kkb either. However, because all the directional acceleration
response ratios in the bottom left region below the dashed line are
larger than 100% (as shown in Fig. 9); it is unsurprising that the
values of the combined resultant acceleration response ratio kacc
in the bottom left region below the dashed line (area II) also exceed
100%. This suggests that installing a link with properties that lie
within this region will cause undesirable effects: the resultant
acceleration will be even larger than that in the associated
unlinked case. For instance, when kka < 0.1 and kkb < 102, the
resultant acceleration response of the LBS exceeds that of the asso-
ciated unlinked case, regardless of link location. Moreover, com-
pared to the resultant acceleration response of the unlinked case,
102 G. Hu et al. / Engineering Structures 138 (2017) 91–104

Max RMS acceleration x Max RMS acceleration y

(a) 1
(b) 1

80
0.8 43 0.8 81
45

0.6 85 0.6
55
λh

λh
65 85 83
95
95

0.4 0.4

75 99 90
0.2 99 0.2

-2 -1 0 1 2 3 -3 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
λka λkb

Max RMS acceleration θ

(c) 1
133
130
0.8 1
120 y

125
0.6 115 x
θ
λh

o
0.4 110 α = 45 wind

101
0.2 105
101
-3 -2 -1 0 1 2
10 10 10 10 10 10
λkb

Fig. 12. Integrated effect of link location and stiffness on the acceleration response (at mass center) in the three directions for a = 45°: (a) x direction; (b) y direction; and (c) h
direction.

a link having properties within the region encircled by the dotted
line (i.e., 0.2 < kh < 0.5, 1 < kka < 100, 102 < kkb < 10, area III) will
even increase the response by 7%. This finding reinforces that
researchers and engineers should exercise caution when designing
the link to avoid unfavorable building responses.
5. Application of the results
As mentioned in Section 4.1, human comfort is an important
element in a tall building’s wind-resistant design. The acceptable
minimal standard of human comfort is usually expressed by the
acceleration response of the building. Therefore, the results shown
in Section 4 can be directly applied to facilitate initial design of the
link. A simple example will be present to show how to apply the
results.
The example is for two separate identical tall buildings without
a link, the geometry of which is very similar to that of the tested
model. A preliminary dynamic analysis shows that the wind-
induced acceleration response of either tower exceeds the acceler-
ation limit by, say, 20%. In order to reduce the acceleration
response and improve human comfort, a three-story link is used
to connect the two buildings. As discussed earlier, a = 0° and 90°
are the two critical wind directions, only the resultant acceleration
responses for these two wind directions are necessary to be consid-
ered. It can be observed from Fig. 13a that, to achieve a 20% reduc-
tion in the resultant acceleration (i.e., kacc < 80%) for a = 0°, in
general, the stiffness and location of the link should meet kka > 1,
kka > 0.1, and kh > 0.5. For a = 90°, Fig. 13c shows that a 20% reduc-
tion in the resultant acceleration usually requires kka > 3, kka > 0.3,
and kh > 0.8. Considering these two requirements, the final link’s
properties should meet kka > 3, kka > 0.3, and kh > 0.8. Obviously,
these link parameters are outside areas II and III, so unfavorable
acceleration response will not occur for the LBS with these link
parameters. In addition, it can be seen from Fig. 13c that when
kh > 0.8, increasing link’s stiffness above kka = 50, kka > 5 has no sig-
nificant effect on the value of kacc. Therefore, it is useless to blindly
increase link’s stiffness excessively above these values. Increasing
stiffness above these values cannot improve human comfort signif-
icantly, but usually leads to the increase in material cost. In addi-
tion, it should be mentioned that these design parameters (i.e.,
kka > 3, kka > 0.3, and kh > 0.8) based on Fig. 13 may be slightly con-
servative. This is because a relatively light mass of the link (i.e.,
km = 0.001) is considered in Fig. 13, while the associated km for
these design parameters could be relatively large.
6. Concluding remarks
In order to understand the performance of wind-excited LBSs,
the effects of link properties such as mass, location, axial stiffness,
and bending stiffness on the wind-induced responses of LBSs have
been comprehensively examined through an evaluation model
together with measured wind force data from the wind tunnel test.
The main findings are summarized as follows:
 G. Hu et al. / Engineering Structures 138 (2017) 91–104 103

λkb λkb

(a) 10-3 10-2 10-1 100 101 102 (b) 10-3 10-2 10-1 100 101 102
1 1
70
90
0.8 0.8 60 59
68
85

0.6 99 0.6 65

λh
λh

95
80 70
y
0.4 y 0.4
75
x
x θ
θ o
0.2 0.2 α = 45 99
o
wind α=0 wind
-2 -1 0 1 2 3 -2 -1 0 1 2 3
10 10 10 10 10 10 10 10 10 10 10 10
λka λka

λkb
(c) 10 -3 -2
10 10 -1
100 101 102
1

79
80
0.8
I
85
90 y
o
0.6 95 α=90
λh

x
wind θ
100
0.4
II 105
III
107
0.2

-2 -1 0 1 2 3
10 10 10 10 10 10
λka

Fig. 13. Effect of link stiffness and location on the resultant acceleration response (at the worst corner) for: (a) a = 0°; (b) a = 45°; (c) a = 90°.

1. The characteristics of wind forces on the LBS differ from those
for a single building. For a = 0°, the pressure on the windward
face is not symmetric and the suction on the two inner faces
are significant, because of the channeling effect. For a = 90°,
the pressure on the windward face of the downstream building
of the LBS is negative as the face is completely submerged in the
upstream shear layer.
2. In terms of structural responses, a = 0° and 90° are the two
critical wind directions, whereas a = 45° (or 135°, 225°,
315°) appears to have the minimal resultant acceleration
response.
3. Although additional link mass decreases the structural fre-
quency and hence increases the displacement response of an
LBS, it does not enhance the acceleration. On the contrary, it
tends to decrease the acceleration response of the LBS, espe-
cially for when a = 90°.
4. Compared to the acceleration response of two buildings with-
out a link, a link with a large stiffness and a high location usu-
ally deceases the translational acceleration responses of the
associated LBS for all wind directions. However, when the stiff-
ness is strong (i.e., kka > 100, kkb > 10), increasing the link’s stiff-
ness or raising the link location above kh = 0.8 cannot produce
significant further reduction in the response. The torsional
acceleration responses of an LBS for a = 45° and 90° are larger
than that of the unlinked case, especially when the link location
is high.
5. For a = 0° and 45°, the resultant acceleration of an LBS decreases
with increasing link location and stiffness. The maximal reduc-
tion can be as much as 32% and 41%, for a = 0° and 45°, respec-
tively. For a = 90°, however, the resultant acceleration in some
LBS cases (region II) are even larger than that in the associated
unlinked case, although when kka > 1, kkb > 0.1, and kh > 0.5, the
resultant acceleration of an LBS decreases with increasing link
stiffness and location. Therefore, researchers and engineers
should exercise caution when designing LBSs to avoid unfavor-
able acceleration responses.
Acknowledgements
The work described in this paper was partly supported by
National Natural Science Foundation of China (Grant Number:
51608398) and the Research Grants Council of the Hong Kong Spe-
cial Administrative Region, China (Project No. 16205515). The CLP
Power Wind/Wave Tunnel Facility at HKUST is thanked for their
assistance in this project.
References
[1] Khanduri AC, Stathopoulos T, Bedard C. Wind-induced interference effects on
buildings—a review of the state-of-the-art. Eng Struct 1998;20:617–30.
[2] Song J, Tse K, Tamura Y, Kareem A. Aerodynamics of closely spaced buildings:
with application to linked buildings. J Wind Eng Ind Aerodyn 2016;149:1–16.
104 G. Hu et al. / Engineering Structures 138 (2017) 91–104
[3] Lee DG, Kim HS, Ko H. Evaluation of coupling–control effect of a sky-bridge for
adjacent tall buildings. Struct Design Tall Spec Build 2010;21:311–28.
[4] Lim J, Bienkiewicz B. Effects of structural and aerodynamic couplings on the
dynamic response of tall twin buildings with a skybridge. ASCE; 2009. p. 1–9.
[5] Westermo BD. The dynamics of interstructural connection to prevent
pounding. Earthq Eng Struct Dyn 1989;18:687–99.
[6] Song J, Tse K. Dynamic characteristics of wind-excited linked twin buildings
based on a 3-dimensional analytical model. Eng Struct 2014;79:169–81.
[7] Tse K, Song J. Modal analysis of a linked cantilever flexible building system. J
Struct Eng 2015;141.
[8] Lim J. Structural coupling and wind-induced response of twin tall buildings
with a skybridge [Doctor dissertation]: Colorado State University; 2008.
[9] Lim J, Bienkiewicz B, Richards E. Modeling of structural coupling for
assessment of modal properties of twin tall buildings with a skybridge. J
Wind Eng Ind Aerodyn 2011;99:615–23.
[10] Xie J, Irwin PA. Wind-induced response of a twin-tower structure. Wind Struct
Int J 2001;4:495–504.
[11] Klein RE, Healey MD. Semi-active control of wind induced oscillations in
structures. Structural control. Springer; 1987. p. 354–69.
[12] Gurley K, Kareem A, Bergman LA, Johnson EA, Klein RE. Coupling tall buildings
for control of response to wind. Struct Saf Reliab 1994:1553–60.
[13] Xu YL, He Q, Ko JM. Dynamic response of damper-connected adjacent
buildings under earthquake excitation. Eng Struct 1999;21:135–48.
[14] Zhang WS, Xu YL. Dynamic characteristics and seismic response of adjacent
buildings linked by discrete dampers. Earthq Eng Struct Dyn
1999;28:1163–85.
[15] Zhu H, Iemura H. A study of response control on the passive coupling element
between two parallel structures. Struct Eng Mech 2000;9:383–96.
[16] Kim J, Ryu J, Chung L. Seismic performance of structures connected by
viscoelastic dampers. Eng Struct 2006;28:183–95.
[17] Ok S-Y, Song J, Park K-S. Optimal design of hysteretic dampers connecting
adjacent structures using multi-objective genetic algorithm and stochastic
linearization method. Eng Struct 2008;30:1240–9.
[18] Mitsuta S, Okawa E, Seto K, Ito H. Active vibration control of structures
arranged in parallel. JSME Int J Ser C, Dyn, Control, Robot, Design Manuf
1994;37:436–43.
[19] Haramoto H, Seto K, Koike Y. Active vibration control of triple flexible
structures arranged in parallel. JSME Int J, Ser C 2000;43:712–8.
[20] Christenson RE, Spencer Jr BF, Johnson EA. Semiactive connected control
method for adjacent multidegree-of-freedom buildings. J Eng Mech
2007;133:290–8.
[21] Cundumi O, Suárez LE. Numerical investigation of a variable damping
semiactive device for the mitigation of the seismic response of adjacent
structures. Comput-Aid Civ Infrastruct Eng 2008;23:291–308.
[22] Zhu H, Wen Y, Iemura H. A study on interaction control for seismic response of
parallel structures. Comput Struct 2001;79:231–42.
[23] Christenson RE, Spencer Jr BF, Johnson EA, Seto K. Coupled building control
considering the effects of building/connector configuration. J Struct Eng
2006;132:853–63.
[24] Chen X, Kareem A. Dynamic wind effects on buildings with 3D coupled modes:
application of high frequency force balance measurements. J Eng Mech
2005;131:1115–25.
[25] Tschanz T, Davenport AG. The base balance technique for the determination of
dynamic wind loads. J Wind Eng Ind Aerodyn 1983;13:429–39.
[26] Tse KT, Hitchcock PA, Kwok K. Mode shape linearization for HFBB analysis of
wind-excited complex tall buildings. Eng Struct 2009;31:675–85.
[27] Bernardini E, Spence SM, Gioffrè M. Effects of the aerodynamic uncertainties in
HFFB loading schemes on the response of tall buildings with coupled dynamic
modes. Eng Struct 2012;42:329–41.
[28] Huang M, Tse KT, Chan CM, Kwok K, Hitchcock PA, Lou W. Mode shape
linearization and correction in coupled dynamic analysis of wind-excited tall
buildings. Struct Design Tall Spec Build 2011;20:327–48.
[29] Huang MF, Tse KT, Chan CM, Kwok K, Hitchcock PA, Lou WJ, et al. An integrated
design technique of advanced linear-mode-shape method and serviceability
drift optimization for tall buildings with lateral–torsional modes. Eng Struct
2010;32:2146–56.
[30] Wilson EL. Three-dimensional static and dynamic analysis of
structures. Berkeley: CSi Computers and Structures; 2002.
[31] Awkar J, Lui E. Seismic analysis and response of multistory semirigid frames.
Eng Struct 1999;21:425–41.
[32] Przemieniecki JS. Theory of matrix structural analysis. Courier Dover
Publications; 1985.
[33] Huang K. Static, seismic and wind-resistant analysis of linked two towers
[Doctor dissertation]. Zhejiang University; 2001.
[34] Clough RW, Penzien J. Dynamics of structures. New York: McGraw-Hill; 1993.
[35] Rossi RE, Laura P, Avalos DR, Larrondo H. Free vibrations of Timoshenko beams
carrying elastically mounted, concentrated masses. J Sound Vib
1993;165:209–23.
[36] Holmes JD. Wind loading of structures. CRC Press; 2001.
[37] Islam M Saiful, Ellingwood B, Corotis RB. Dynamic response of tall buildings to
stochastic wind load. J Struct Eng 1990;116:2982–3002.
[38] Tallin A, Ellingwood B. Serviceability limit states: wind induced vibrations. J
Struct Eng 1984;110:2424–37.
[39] Boggs D. Acceleration indexes for human comfort in tall buildings—peak or
RMS? CTBUH Monogr 1997.
[40] Tallin A, Ellingwood B. Wind induced lateral-torsional motion of buildings. J
Struct Eng 1985;111:2197–213.