Wind and Structures, Vol. 11, No.

1 (2008) 1-18 1
DOI: http://dx.doi.org/10.12989/was.2008.11.1.001

Interference effects in a group of tall buildings

closely arranged in an L- or T-shaped pattern
J. G. Zhao and K. M. Lam
Department of Civil Engineering, University of Hong Kong, Pokfulam Road, Hong Kong
(Received December 26, 2006, Accepted January 11, 2008)

Abstract. Interference effects in five square tall buildings arranged in an L- or T-shaped pattern are
investigated in the wind tunnel. Mean and fluctuating shear forces, overturning moments and torsional
moment are measured on each building with a force balance mounted at its base. Results are obtained at
two values of clear separation between adjacent buildings, at half and a quarter building breadth. It is
found that strong interference effect exists on all member buildings, resulting in significant modifications
of wind loads as compared with the isolated single building case. Sheltering effect is observed on wind
loads acting along the direction of an arm of the L or T on the inner buildings. However, increase in
these wind loads from the isolated single building case is found on the most upwind edge building in the
arm when wind blows at a slight oblique angle to the arm. The corner formed by two arms of buildings
results in some wind catchment effect leading to increased wind pressure on windward building faces.
Interesting interference phenomena such as negative drag force are reported. Interference effects on wind
load fluctuations, load spectra and dynamic building responses are also studied and discussed.
Keywords: interference effect; wind loads; sheltering; channeling.

1. Introduction

Wind-induced interference effect in closely spaced buildings has been found to result in
significant modifications of mean and dynamic wind loads on a building from the isolated single
building situation (e.g. Khanduri, et al. 1998 for a review). Most studies have been carried out in
the wind tunnel in which wind loads are measured on the test building model with the interfering
building placed in different relative positions. Interference effects on mean wind pressure and wind
forces were studied with rigid building models (Blessman and Riera 1979, Saunders and Melbourne
1979, English 1985). Aeroelastic models and the base-balance technique were later used to assess
interference effects on dynamic behaviour of tall buildings (Bailey and Kwok 1985, Taniike 1992,
Zhang, et al. 1994, Thepmongkorn, et al. 2002). A number of interference mechanisms have been
reported. Sheltering effect generally leads to reductions in mean wind load on the downstream
building but wake buffeting may cause additional fluctuations in wind loads and thus increase the
dynamic response of the downstream building. Flow channelling often occurs through the gap

Graduate Student
Associate Professor, Corresponding Author, E-mail: [email protected]
2 J. G. Zhao and K. M. Lam

between two closely spaced buildings and this leads to highly negative pressure on the relevant
building faces.
In addition to an understanding of the interference mechanisms, recent studies attempted to seek
design guidelines for interference effects on wind loads on buildings. Cheng and Lin (2005), and
Xie and Gu (2005) commented that the large number of parameters affecting interference effects
such as the many possible building arrangements render it impractical to derive simple empirical
formulae for interference effect estimation. Data assimilation techniques such as artificial neural
network and expert system were recommended as a more sensible approach (English and Fricke
1999, Xie and Gu 2007). A large database of wind loads under interference effects is required for
the data assimilation techniques and there are a number of recent wind tunnel measurements
towards this effort (Xie and Gu 2004, Huang and Gu 2005, Cheng and Lin 2005). In particular, Xie
and Gu (2004) studied interference among three tall buildings and those data of interference effects
were recently quantitatively described by a proposed envelope interference factor in Xie and Gu
(2007).
Many previous studies report interference between two buildings with a clear building separation
between one and a few building width. In metropolitan cities such as Hong Kong, there are many
residential developments which comprise a number of similar shaped tall buildings located closely
in a group. Building members are often arranged in a row or in an L- or T-shaped pattern and the
clear separation between neighbouring buildings is usually less than one building width. In this
paper, interference effects on wind loading of five closely-spaced tall buildings arranged in an L- or
T-pattern are studied in the wind tunnel from base-balance measurements. The effects of wind
incidence angles and spacing separation between neighbouring buildings are investigated. The
experimental data will be useful in estimating design wind loads for a group of buildings in close
proximity.

2. Wind tunnel experiments

The object of study was a group of tall buildings arranged in an L- or T-shaped pattern. As shown
in Fig. 1, five building models were used to achieve the minimum configuration of three member
buildings along one arm of the L or T. All building models were 50cm tall with a square plan
form of breadth 10cm. With the target geometric scale at 1:300, they represented typical high-rise
residential buildings with height at 150 m and height-to-breadth ratio at 5:1. Two different values of
clear building separation were used at S = 0.25B and 0.5B, B being the building breadth. The
experiments were carried out in the boundary layer wind tunnel of the Department of Civil
Engineering at the University of Hong Kong. It had a 3.0 m wide and 1.8 m tall working section.
With the installation of triangular spires and 8 m long fetch of floor roughness elements, wind
characteristics in the tunnel simulated natural wind of the open land terrain type. The mean wind
speed profile followed the power law with exponent 0.15 (Fig. 2a). The turbulence intensity varied
from a value of about 0.20 near ground to about 0.08 near the roof height (Fig. 2b). These wind
characteristics corresponded approximately to those specified for strong winds over the general
terrain type in the 1983 version of Hong Kong Wind Code (BDD 1983).
Fluctuating wind loads on a building model, including base shear forces Fx, Fy, base overturning
moments, Mx, My, and torsion Mz, were measured with a six-component force balance (JR3 Inc.)
mounted at the model base. From the load signals, mean force and moment coefficients are
computed with equations such as:
 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 3

Fig. 1 A group of 5 tall buildings arranged in (a) L-pattern; (b) T-pattern

Fig. 2 Wind characteristics in the wind tunnel: (a) mean wind speed profile; (b) turbulence intensity profile

Fx
C Fx = ---------------------------
-
1/ U 2B H
2 H

My
C My = ------------------------------ (1)
1/ U 2B H 2
2 H

Mz
C Mz = -----------------------------
-
1/ U 2B2 H
2 H

where F x is the mean shear force along the x- direction body axis of the building, M y is the mean
4 J. G. Zhao and K. M. Lam

moment about the y axis of the building and M z is the torsion. The building height is H and U H is
unobstructed mean wind speed at the building roof height. Standard deviation values of loading
coefficients are computed with equations such as:
Fy
CFy = --------------------------- -
1/ U 2B H
2 H
(2)
Mx
C Mx = ------------------------------
1/ U 2B H 2
2 H

where Fy and Mx are the standard deviations of the base shear force Fy and the base overturning
moment Mx, respectively.
Wind load measurements were made at all wind incidence angles to the group of buildings. The
definition of wind incidence angle is shown in Fig. 1 for the L- and T-patterns of buildings.
Making use of the symmetry of the L- or T-pattern and with tests performed at all wind angles
o o
between 0 360 at 10o intervals, wind load measurements were made on Buildings A, B and
C of one arm of the L only. The wind load information on the other two buildings, D and E,
could be deduced from the data on Buildings B and C, respectively. For the T-pattern, wind loads
were measured on all buildings except Building J whose information could be derived from that on
Building I.

3. Results and discussion

3.1. Mean wind loads on buildings of L-pattern

For Buildings A, B and C arranged in one arm of the L-pattern, wind angle variations of the two
mean force coefficients, C Fx and C Fy , and mean torsion coefficient, C Mz , are shown in Fig. 3 for
the two building separations. Results for the isolated single building case are also shown for com-
parison. The overturning moment coefficients, C Mx and C My , have the same wind angle variations as
the corresponding force coefficients C Fy and C Fx and the results are not shown. Since
measurements have been made at all wind angles from 0o to 360o, wind load behavior of buildings
on the other arm of the L-pattern can be obtained from those of Buildings A, B and C by symmetry.
It is evident that the two building separations lead to very similar modifications of wind loads from
the isolated single building case, with more significant interference effects at the smaller separation
value S/B = 0.25.
Between the two shear forces, much larger load reductions from the single isolated building case
are observed on C Fx which acts along the direction of the arm. This is due to the sheltering effect
by upstream building or buildings. The inner Building B experiences the largest reduction in C Fx
from sheltering effect at nearly all wind directions. The largest magnitude of Fx on Building B is
o
C Fx 0.6 (occurring at 120 ) and this represents a 40% reduction from the largest C Fx value at
1.0 for the isolated single building case. At wind angles between 0o = 180o and 270o, buildings on
the other arm of the L provide total shelter to Building B providing almost zero values of C
within this quadrant of wind incidence. Building C at the edge of the arm is under strong sheltering
effect along the x-direction at wind angles between = 90o and 270o and thus Fig. 3 shows very
small values of C Fx for Building A at the corner of the L, sheltering along the x-direction occurs
 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 5

Fig. 3 Mean wind load coefficients: variation with wind angles and effects of building separation (L-pattern)

at the opposite wind incidence angles between = 90o (that is 270o) and 90o.
A small increase in C Fx from the isolated single building case is observed on Building C at
o
30 . A similar observation has been made for the edge building in a row of tall buildings (Lam
and Zhao 2006). At this slightly oblique wind angle, wind hits on the windward x-face of Building
C in the same way as on a single building, producing similar levels of positive pressure on that
face. However, at the leeward side, wind is channeled to flow through the gap between this building
and Building B at high speeds (Fig. 1). Highly negative pressure is induced on the leeward x-face of
Building C. This results in a higher value of Fx than a single building. Similar gap flow is believed
to be responsible for the higher on Building A at = 180o. If Building A were alone, wind would
separate at its windward corners with diverging separation streamlines and the negative pressure on
its leeward face will depend on the distance to the separation streamlines (which entrains air from
the face). In the L-pattern, wind flows through the gap between Buildings A and D at high speeds
6 J. G. Zhao and K. M. Lam

(Fig. 1). The fast air stream is very close to the leeward face and this results in more negative wind
pressure on that face. This leads to high magnitudes of Fx. It is clear in Fig. 3 that the increases in
C Fx for the two cases become more significant at the narrower building separation. This provides
an indirect evidence to the role of the gap flow in the interference mechanism.
Another notable observation on C Fx in Fig. 3 is the negative drag on Building B at = 180o.
The shear force Fx is positive, that is acting in the upwind direction. It is speculated that at this
normal wind incidence, wind separates at Building A, and Buildings B and C are inside its wake.
Pressure at a point inside the building wake depends on the distance from the separation streamlines
extending from Building A. The windward x-face of Building B is nearer to the separation
streamlines than its leeward x-face and thus more negative wake pressure is induced on the former
face. The result is a small drag force Fx acting upwind.
For the shear force along the y-direction, little interference effects occur on Buildings B and C at
wind angles between = 0o and 180o. Wind blows freely onto this arm of the L and Fy acts
o
perpendicular to the arm (Fig. 1). There is some very slight increase in C Fy near 90 during
which high-speed channeled flow through the building gaps causes more negative pressure on the
leeward y-faces. Between = 180o and 270o, the other arm of the L provides some degree of
sheltering to Buildings B and C. This reduces the magnitudes of negative pressure on their leeward
y-faces and results in small values of C Fy . In the fourth quadrant, 270o < < 360o, the L catches
more wind to slow down on windward faces of all buildings. Higher stagnation effects on these
faces are believed to cause increased values of C Fx , especially on Building B.
Building A at the corner of the L is shielded by Buildings D and E at wind angles between
180o < < 360o. At = 180o, Building A is under positive Fy at C Fy 0.4 rather than zero force for
an isolated single building. This is because its two y-faces are exposed to different types of flow. On
the free side, wind flows around the building the same way as for a single building but on the other
side, wind flows through the gap with Building D at highspeed, inducing highly negative pressure
on the building faces. In the fourth quadrant of wind incidence, 270o < < 360o, positive values of
C Fy are found at the narrower building separation of S/B = 0.25 instead of negative values for the
single building. The cause for this negative drag behavior is different from that described above
for C Fx . The windward y-face on Building A is under highly negative pressure due to channeled
gap flow while the more exposed leeward y-face is under less negative wake pressure.
Fig. 4 shows that the mean drag coefficient CD and lift coefficient CL on buildings in the L-pattern
for the wider building separation S/B = 0.5. They are computed from the shear force and act in the
along-wind and lateral directions. For the isolated single building, CD is always positive and has a
value from 0.9 to 1.2 while CL changes between 0.3 and 0.3. The phenomenon of negative drag is
evidently observed on Building B at = 10o and 180o. Building C has almost zero drag at = 10o
and 180o at this wider S and negative drag occurs when the separation is halved to S/B = 0.25 (Fig.
3). Similarly for Building A, the drag has small magnitudes in the fourth quadrant and will turn
negative at the narrower building separation. Building C at the end of the arm experiences peak
mean drag about C D 1.4 which is about 15% higher than that on an isolated single building. This
occurs at = 330o and is caused by increases in both Fx and Fy as described above. The range of
variations of lift force is widened significantly for all three buildings, Buildings A, B and C. On
Building A, CL varies between 0.6 and 0.5. The maximum and minimum lift forces on Building C
are about 0.4 and 0.45, respectively. The largest magnitude of peak lift is found on the inner
Building B at C L 0.7 occurring at = 350o. This magnitude is more than double the single
building value.
 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 7

Fig. 4 Variation of mean drag and lift coefficients with wind angles. S/B = 0.5 (L-pattern)

Fig. 3 also shows the variation of torsion coefficient C Mz with wind angle. Torsion on a square
building is caused by the combined effects of asymmetrical pressure distribution on all four building
faces. For normal incidence on the isolated single building, separation occurs at the upwind corners
and negative pressure on the side face will have large magnitudes near the windward edge and
decreasing magnitudes towards the leeward edge. This face produces a significant contribution to
torsion. At oblique incidence, positive pressure on the windward face also has an asymmetric
distribution and contributes to torsion. It is observed in Fig. 3 that C Mz on the inner Building B is
much smaller than the isolated building case at wind angles 0o < < 180o. It is believed that
channeled gap flow always occurs on its two side faces and their contributions to torsion of
opposite signs are balanced out. Between = 225o to 270o, torsion on Building B is positive as
contrary to the negative torsion on a single isolated building. The negative torsion in the single
building case at these wind angles is mainly caused by the asymmetry distribution of positive
pressure on the windward building face, that is the right face of Building B in Fig. 1. In the L-
pattern, this face is sheltered by the other arm of the L and there is thus little contribution to
negative torsion. It is expected that flow channeling effect is more intense through the gap on the
upper side of Building B than the gap on the lower side. The upper face makes positive contribution
to torsion, that is in the counter-clockwise direction.
At most wind angles, torsion on Building C varies within the same range of variation as that of
the isolated single building. The exception is around = 330o when the torsion reaches C Mz < 0.12,
a peak value about 1.5 times the single building value. Similar high values of peak torsion is found
on Building A around = 90o and 180o. This can be explained by the gap flow on one side face of
Building A only. The highly negative pressure on this face with higher levels near the entrance side
makes a dominant contribution to torsion.
The peak positive and negative values of mean wind loading coefficients of the Buildings in the L-
pattern are summarized and compared with those on an isolated single building in Table 1. The data of
all five wind load components are listed for the narrower building separation S/B = 0.25 at which the
interference effects are generally larger. For the wind loads along the x-direction, Building A at the
corner of the L and Building C at the end of the arm experience increased peak mean loads which
can be 15% higher than the single building values but the increase occurs only over a very narrow
range of wind angles (Fig. 3). In many quadrants of wind incidence, wind loads are reduced as a result
8 J. G. Zhao and K. M. Lam

Table 1 Peak positive and negative values of mean wind loading coefficients over all wind angles. Building
separation at S/B = 0.25.
x-direction loads y-direction loads Torsion
CF x
CM y
* CF y
CM y
* CM z
*
Single building 1.01 0.60 0o, 180o 1.01 0.59 90o, 270o 0.08 90no15o
L-Pattern
A 1.15 0.70 180o 1.31 0.72 90o 0.12 90o
20o( 30 ) 180o~360o 180o
o
0.60 0.32 0.53 0.30 0.13
0.65 0.39 0o~180o 1.10 0.61 100o 0.04 20o
B
0.21 0.15 0o90o 0.99 0.55 180o~360o 0.04 250o
1.13 0.73 20o 1.12 0.65 70o 0.12 350o
C o o o
0.23 0.03 90 ~270 1.19 0.64 310 0.09 100o
T-Pattern
180o 0o
o
F 1.02 0.61 0.52 0.29 110 0.12
0o 100o
o
0.38 0.22 0.56 0.32 240 0.09
0.23 0.15 120o 0.98 0.55 100o 0.09 110o
G
0.22 0.13 0o90o 1.03 0.56 260o 0.10 240o
o o
0.26 0.06 90 ~270 1.15 0.65 70
o
0.12 340o
H o
1.21 0.75 30o 1.24 0.66 310 0.09 10o
1.08 0.64 170o 1.38 0.79 90o 0.13 90o
I
1.04 0.58 0o~90o 0.19 0.08 180o~360o 0.11 190o
*Note: This column lists the wind angle at which the peak mean load occurs or the range of over which the peak
mean load occurs.

of sheltering. Building B inside the arm of the L is under severe sheltering at nearly all wind angles.
All buildings can experience at least 10% increase in mean wind loads in the y-direction at some
particular wind incidence angles. For Building A, up to 30% increase can occur.

3.2. Mean wind loads on buildings of T-pattern

Interference effects on mean wind loads on Buildings F, G, H, I in the T-pattern are shown in Fig.
5. Since this pattern of buildings has a plane of symmetry about the line of = 0o or 180o,
Buildings F, G and H have their C Fx ( ) symmetric about = 180o (and = 0o) while their C Fy ( )
and C Mz ( ) are anti-symmetry about = 180o. For Building H at the long end of the T, wind
angle variations of all three wind load components are almost identical to those of Building C at the
end of the L. This is due to similar wind flow patterns and same interference mechanisms. The
only notable difference is that Building H does not experience the sheltering effect on C Fx between
= 180o and 270o which exists on Building C due to the other arm of the L. For Building G, the
behavior of wind loads is very similar to that of the second half part of the load curves for Building
o
B at 180o < < 360o in Fig. 3. Around 240 , there is a lower degree of shielding offered by
Building J in the T than the other arm of Buildings D and E in the L. Hence, both shear forces
 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 9

Fig. 5 Mean wind load coefficients: variation with wind angles and effects of building separation (T-pattern)

on Building G are less sheltered than on Building B. Wind load patterns for Building F are similar
to those for Building A at 180o < < 360o. Some differences are observed near = 180o because
10 J. G. Zhao and K. M. Lam

the interference effects on Fx and Fy found on Building A due to different wind flow behavior
around its free and congested faces do not occur on Building F.
For Building I, modification of its C mainly occurs in the fourth quadrant at which its windward
x-face is almost totally sheltered by other building members of the T. This results in near zero
values of Fx. Similarly, at 180o < < 360o, its windward y-faces are sheltered resulting in very small
values of Fy . At 45o < < 90o, highly negative pressure is expected to occur on its leeward y-face
when wind flows through the narrower building gap.
The coefficient becomes larger than the single building value and reaches a peak value at 1.4. The
asymmetric distribution of this highly negative pressure on the gap face also produces a peak
torsion much larger than the single building peak value.
Table 1 summarized the increase or decrease in the peak positive and negative values of mean
wind loading coefficients of the buildings due to building interference as compared to the case of an

Fig. 6 Fluctuations of wind loads: RMS wind load coefficients (L-pattern)

 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 11

isolated single building.

3.3. Fluctuation wind loads and load spectra

Fig. 6 shows the coefficients of standard deviation of wind loads on buildings in the L-pattern.
These coefficients CFx , CFy and CMz are computed from the time varying wind load signals
measured by the base balance. They represent wind load fluctuations acting on the static building
models. On an isolated square building, peak values of coefficients occur at normal wind
incidences, with turbulence buffeting responsible for load fluctuations in alongwind direction and
vortex excitation for the across-wind actions. These mechanisms are evident from the wind moment
spectra shown in Fig. 7 At = 90o, for instance, the spectrum of the across-wind moment My shows
a sharp spectral peak at nB/U H 0.1 . This is due to vortex shedding from the building and leads to
the peak in CFx ( ) in Fig. 6. The along-wind spectrum of Mx is broad-banded without any sharp
spectral peaks. This is caused by turbulence buffeting and high values of CFy ( ) occurs over a

Fig. 7 Wind moment spectra at normal wind incidence. S/B = 0.5 (L-pattern)
12 J. G. Zhao and K. M. Lam

wider range of wind angles around = 90o.

When buildings are placed in an L-pattern, modifications of standard deviations of wind loads

Fig. 8 Fluctuations of wind loads: RMS wind load coefficients (T-pattern)

 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 13

from the single building case generally follow the same trends as the mean wind loads (Fig. 6 as
compared with Fig. 3). At those wind angles when C Fx ( ) is reduced due to the sheltering effect
(e.g. on Building C at 90o < < 270o in Fig. 3), the standard deviations CFx ( ) are also found to
be reduced. Fig. 7 shows that at = 0o, sheltering effect lowers the broad-banded along-wind
moment spectra of My on Buildings A and B at most frequencies but there are little changes or even
increased spectra power around frequencies nB/U H 0.2 . Since this frequency is about double the
vortex shedding frequency of a square building, this suggests that the sheltered building is under the
effect of vortex excitation from the upwind building. However, the sharp vortex shedding peak is
not evidently found in the across-wind spectra of Mx at = 0o for all buildings in the L-pattern,
except a much broader and lower energy peak remaining on the upwind Building C. When wind
blows normal to the arm of buildings, around = 90o and 270o, turbulence buffeting is not affected
or even enhanced by presence of neighboring buildings and thus similar or larger values of C in
Fig. 6. For the across-wind force and moment, Fx and My, the vortex shedding peak disappears
completely. This implies that vortex excitation is greatly reduced on buildings located side-by-side.
Fig. 7 shows that the spectra peak around nB/U H 0.1, which occurs in the torsion spectra of an
isolated square building, disappears in all buildings in a group. At higher frequencies nB/U H 0.2,
the spectral levels and shape remain similar. These two observations suggest that modifications in
torsion spectra come mainly from increased turbulence in the along-wind actions and lack of vortex-
excitation in the across-wind actions.
In Fig. 6, increase in standard deviations of wind loads from the isolated single building case can
occur on Building A, B or C at some wind angles. At these wind angles, mean wind loads are also
higher than the single building values (Fig. 3) and they are due to channeled flow through building
gaps or increased blockage presented by buildings arranged side-byside. Fig. 7 shows that the
increase is generally over all spectral frequencies of the broadbanded load spectra.
Fig. 8 shows the coefficients of CFx ( ) , C Fy ( ) and C Mz ( ) on Buildings F, G, H, I in the T-
pattern. The same as for the mean wind loads, the behavior of these standard deviation values of
wind loads on these buildings follow similar trends as those buildings in the L-pattern at similar
relative positions in the group (e.g. Building H and Building C, both on the edge of an arm). For
Building I located at the end of the horizontal arm of the T, the variation of CFx ( ) is similar to
Building A of the L-pattern for 180o < < 360o and to Building C for 0o < < 90o. The variation of
CFy ( ) has the same pattern as that on Building A but even higher values are found at
45o < < 90o. The moment spectra at normal wind incidences are shown in Fig. 9. When compared
to Fig. 7, it is noted that a weaker sharp spectral peak can be found in the across-wind spectra of
Mx on Building I at = 0o, and on Building H at = 90o or 270o. It seems that some weak form of
vortex shedding can occur from these buildings which are located at the ends of the symmetric T-
pattern.

3.4. Dynamic building responses to wind excitation

With the base moment signals measured, wind-induced dynamic deflections of the buildings in the
L-pattern or T-pattern are computed with the base-balance technique (Tschanz 1982). With the
assumption of linear mode shapes in the sway directions and constant mode shape in the torsional
direction, the base moment spectra in Fig. 7 gave the spectra of the generalized wind forces. These
spectra are multiplied by the mechanical admittance function of the building to obtain the spectra of
building deflections from which the standard deviations of deflections at the building top floor can
14 J. G. Zhao and K. M. Lam

Fig. 9 Wind moment spectra at normal wind incidence. S/B = 0.5 (T-pattern)

be found. Dynamic properties of a full-scale 52-storey reinforced concreted residential building of

similar height and breadth as the test square building are used for analysis. Same values of
fundamental natural frequency and mode shapes are set for the two sway modes: n0, x = n0, y = 0.238 Hz
and mode shape, (z) = [0, 0.0015, 0.0040, 0.0070, 0.0101] at z/H = [0, 0.25, 0.5, 0.75, 1]. The
natural frequency for torsion is n0, = 0.415 Hz and a damping ratio at 1% of critical is assumed.
The use of this lower value of damping is intended to obtain larger resonant effects which enable
clearer observation of the behavior of building responses at different reduced velocities. A damping
ratio at 1% is commonly used for assessing wind-induced building accelerations for serviceability
purposes.
Figs. 10 and 11 show, for buildings in the L-pattern and T-pattern, respectively, the standard
deviations of wind-induced deflections along the sway directions x, y and the torsional direction
at different reduced velocity at building roof height. Results are shown for the building separation
S/B = 0.5 only. The data for S/B = 0.25 exhibit similar response behavior with reduced velocity and
are thus not shown for brevity. For an isolated square building, along-wind responses increases
 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 15

Fig. 10 Wind-induced translational and torsional responses at different reduced wind velocities. S/B = 0.5 (L-
pattern)

approximately with the square of reduced velocity while across-wind responses show resonant
excitation at reduced velocities around 10. This value is the reciprocal of the non-dimensional
frequency of the vortex shedding peak in the across-wind moment spectra. Resonant torsion
responses are found at reduced velocities around 5 and 10.
For all buildings in the L-pattern, across-wind responses are largely reduced from the single
building case with the disappearance of resonant excitation at reduced velocities around 10 (Fig.
10). This is due to the suppression of vortex shedding and vortex excitation for closely spaced
buildings. When wind blows along the arm of Buildings A, B and C at = 0o, along-wind responses
are reduced for the downwind building but x, of the windward Building C is slightly increased,
especially at reduced velocities around 5. At = 90o, when wind blows normal to the arm from the
free side of the L, there is little interference effect on turbulence buffeting and the along-wind
responses, y, of Buildings A, B and C are very similar to those of an isolated building. When wind
blows from the opposite direction at = 270o, Building A is sheltered and exhibits largely reduced
alongwind responses of y. For Building B, the responses become higher than the single building
16 J. G. Zhao and K. M. Lam

Fig. 11 Wind-induced translational and torsional responses at different reduced wind velocities. S/B = 0.5 (T-
pattern)

case at reduced velocities around 5.

Similarly, for buildings in the T-pattern, dynamic responses are generally reduced from the single
building case (Fig. 11). At = 0o, the moment spectra in Fig. 9 shows some form of vortex
shedding from Building I. This explains its resonant across-wind responses of y near reduced
velocities 10. Increased y-direction wind loads are found on Building I near = 90o and thus the
along-wind responses of y at = 90o are higher than those of an isolated single building.

4. Conclusions

In this paper, wind tunnel experiments are carried out to investigate interference effects on five
tall buildings arranged in an L- or T-shaped pattern. The buildings are square in plan and have a
height-to-breadth ratio at 5. The buildings are closely spaced and two values of clear separation
between adjacent buildings are tested, one at 0.5B and the other at 0.25B. Mean and fluctuating
wind forces and moments on a building are measured with a base balance for all possible wind
 Interference effects in a group of tall buildings closely arranged in an L- or T-shaped pattern 17

incidence angles.
In general, wind loads acting along the direction of an arm of the L or T are reduced from
the isolated building case as a result of sheltering. These wind loads are found to be increased on
the upwind building at the end of the arm at a slight oblique wind incidence angle from parallel to
the arm. This is caused by wind being channeled to flow fast through the building gap behind the
building which produces highly negative pressure on the rear building face. This channeling effect
also leads to increase in torsion at some wind angles. When the mean wind loads on a building are
increased or reduced due to the interference effect, the fluctuating loads as described by the standard
deviations are modified in the same way. The load spectra show that while turbulence buffeting is
not very much affected by interference, vortex shedding and the related across-wind excitation are
largely suppressed when buildings are closely placed in a group. The computed wind-induced
dynamic responses are affected accordingly. In most cases, resonant across-wind responses which
occurs on an isolated square building at reduced velocities around 10 are not found on buildings in
a group.

Acknowledgement

The investigation is supported by a research grant (HKU7014/02E) awarded by the Research

Grants Council of Hong Kong.

References

Bailey, P. A., and Kwok, K. C. S. (1985), Interference excitation of twin tall buildings, J. Wind Eng. Ind.
Aerodyn., 21, 323-338.
Blessman, J. and Riera, J. D. (1979), Interaction effects in neighbouring tall buildings, Proceedings of 5th Int.
Conf. Wind Engineering, Fort Collins, 381-395.
Buildings Development Department, Hong Kong (BDD) (1983), Code of Practice on Wind Effects, Hong Kong.
Cheng, C. M. and Lin, Y. C. (2005), Interference effects on the design wind loads of tall buildings,
Proceedings of 6th Asia-Pacific. Conf. Wind Engng, Seoul, 586-598.
English, E. C. (1985), Shielding factors from wind-tunnel studies of mid-rise and high-rise structures,
Proceedings of 5th US Nat. Conf. Wind Engng, Lubbock, 4A, 49-56.
English, E. C. and Fricke, F. R. (1999), The interference index and its prediction using a neural network
analysis of wind-tunnel data, J. Wind Eng. Ind. Aerodyn., 83, 567-575.
Huang, P. and Gu, M. (2005), Experimental study on wind-induced dynamic interference effects between two
tall buildings, Wind Struct., 8(3), 147-161.
Khanduri, A. C., Stathopoulos, T., and Bedard, C. (1998), Wind-induced interference effects on buildings . A
review of the state-of-art, Eng. Struct., 20(7), 617-630.
Lam, K. M. and Zhao, J. G. (2006), Interference effects on wind loads on a row of tall buildings, Proceedings
of 4th Int. Sym. on Comput. Wind Engng., Yokohama, 817-820.
Saunders, J. W. and Melbourne, W. H. (1979), Buffeting effects of upwind buildings, Proceedings of 5th Int.
Conf. Wind Engng, Fort Collins, 593-605.
Taniike, Y. (1992), Interference mechanism for enhanced wind forces on neighbouring tall buildings, J. Wind
Eng. Ind. Aerodyn., 42, 1073-1083.
Thepmongkorn, S., Wood, G. S. and Kwok, K. C. S. (2002), Interference effects on windinduced coupled
motion of a tall building, J. Wind Eng. Ind. Aerodyn., 90, 1807-1815.
Tschanz, A. (1982), Measurement of total dynamic loads using elastic models with high natural frequencies,
Workshop on Wind Tunnel Modeling Criteria and Effects, N.B.S., Gaithersburg, Mass., 1982.
Xie, Z. N. and Gu, M. (2004), Mean interference effects among tall buildings, Eng. Struct., 26, 1173-1183.
Xie, Z. N. and Gu, M. (2005), A correlation-based analysis on wind-induced dynamic interference effects
18 J. G. Zhao and K. M. Lam

between two tall buildings, Wind Struct., 8(3), 163-178.

Xie, Z. N. and Gu, M. (2007), Simplified evaluation of wind-induced interference effects among three tall
buildings, J. Wind Eng. Ind. Aerodyn., 95, 31-52.
Zhang, W. J., Kwok, K. C. S. and Xu, Y. L. (1994), Aeroelastic torsional behaviour of tall buildings in wakes,
J. Wind Eng. Ind. Aerodyn., 51, 229-248.

CC