Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Contents lists available at ScienceDirect

Journal of Wind Engineering & Industrial Aerodynamics

journal homepage: www.elsevier.com/locate/jweia

Impact of structural design criteria on the comfort assessment of tall

buildings author names and affiliations
Ferrareto A. Johann *
Escola Polit
ecnica da Universidade de S~
ao Paulo, Department of Structural and Foundation Engineering, Avenida Professor Almeida Prado, trav. 2, 83, Cidade
Universit

aria, S~ao Paulo, SP, 05508-900, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: Assessing tall building oscillation due to wind-induced motion is a multidisciplinary task that involves knowledge
Tall buildings from several fields of study, including: structural engineering, wind engineering, reliability, and even human
Structural analysis physiology. With the modern high strength structural materials and the latest tendencies in tall buildings con-
Dynamic analysis struction, new structural systems have become slender and new buildings have reached greater heights as time
Comfort assessment passes. This context leads to a situation where these slender structures become sensitive to the dynamic effects of
Design criteria
wind loads, case in which the human comfort is often the prevailing criterion for the structural design. This paper
Wind loads
Wind tunnel
addresses criteria from finite element modelling, modal truncation, wind directionality, and comfort assessment
applied to two building studies (buildings A and B) subjected to wind tunnel testing. Then, the impact of structural
design criteria on many different disciplines is exposed, establishing a comparison between different criteria. This
investigation intends to bring precision to the procedure, while creating a reliable set of criteria to perform an
assessment of the dynamic response from the wind tunnel testing of tall buildings.

1. Context, introduction and reasons for the study
1.1. Context
In today's context of big cities, the category of tall building con-
struction has quickly gained ground due to environmental and economic
issues (Ali and Moon, 2007; Drew et al., 2014). These new constructions
require extensive and multidisciplinary knowledge to make them
feasible, leaving a great deal of responsibility to a multidisciplinary group
of areas of study: structural engineering, wind engineering and comfort
assessment. This paper is focused on the understanding of the set of
criteria of each discipline, on the use of these data to perform a tall
building's motion assessment, and on the impact of each criterion on the
final motion assessment.
Latest advances in structural materials, including 65psi (450MPa)
high strength steel, high strength concrete, and new composite structures
allow for a great reduction in the use of material in tall buildings (Rosa
et al., 2012; Sarkisian, 2012). These improvements enable both slender
structures and slender structural systems, which lead to an overall
reduction of the building stiffness. These slender structural systems are
commonly used in tall building design and often present fundamental
modes of vibration with a behavior very similar to a cantilever beam (Wu
et al., 2007; Sarkisian, 2012).
Moreover, these structural systems often present important torsional
modes of vibration and a greater number of natural frequencies under
1:0Hz, making them more susceptible to dynamic effects of wind loads
(Rosa et al., 2012). These circumstances emphasize the importance of
service limit state (SLS) studies on tall buildings for comfort assessment
when compared to ultimate limit state (ULS), due to: higher modal
contribution, torsional acceleration and cantilever behavior of the
structural system (Hansen et al., 1973; ISO10137, 2007; Kim et al., 2009;
Rosa et al., 2012).
1.2. Introduction: structural data, wind tunnel testing and comfort
assessment
The structural data have a clear importance in the assessment of the
response of wind-induced motions in tall buildings. For a specific
approach to structures of multi-story tall buildings subjected to wind
tunnel testing (WTT), these data can be summarized as: natural fre-
quencies; mode shapes or mode deflection shapes; mass matrix; and
damping of the system.
Finite-element (FE) models are not always as precise as they might
look when it comes to finding out the natural frequencies of a building.

* Tel.: +55 27 99237 8696.

E-mail address: andrade.ferrareto@gmail.com.

https://doi.org/10.1016/j.jweia.2018.07.006
Received 14 February 2018; Received in revised form 16 May 2018; Accepted 6 July 2018

0167-6105/© 2018 Elsevier Ltd. All rights reserved.

F.A. Johann
Kim et al. (2009) performed field measurements in three buildings to
acquire the first three natural frequencies of each building, and found
discrepancies of up to 33% between the measured values and the results
of the finite element model. The results showed a great underestimation
of the natural frequencies for the FE models, where the authors investi-
gated the phenomena through several axes of investigation, among
which the most relevant ones for this paper were the flexural stiffness of
floor slabs and the increase in the modulus of elasticity of structural
members due to concrete ageing. Structural data are gathered and
analyzed for both buildings (A and B) for different FE models and for
different sets of criteria, concrete ageing, and floor slab modelling. Then,
dynamic responses are analyzed for the different models created.
The increased modulus of elasticity is the Young's modulus for
“t → ∞” in eq. (1) of the Brazilian concrete code NBR6118-2014. This
equation shows the increase in the elasticity modulus with the increase of
the concrete age “t”:
n nh ioo0:5
ECi;∞ ¼ limECi;28 exp s 1  ð28=tÞ0:5 ¼ ECi;28 expðs=2Þ (1)
t→∞
where:
 ECi;28 is the Young's Modulus of the concrete after 28 days, according
to NBR6118-2014;
 s is a coefficient depending on the category of cement: in the tall
buildings analyzed in this paper, this coefficient has the value of 0:25;
 ECi;∞ stands for the Young's Modulus of the matured concrete,
referred to here as probable E.
The schematics of the categories of concrete strength for each struc-
tural element are given in Fig. 1 for both buildings (buildings A and B).
The studies of Kim et al. (2009) showed a sensitive increase in the natural
frequencies of buildings due to concrete ageing (up to 12%), which lead
to an important effect in the final acceleration assessed on the top of the
building.
Intended for the scope of wind effects on tall buildings, a lumped mass
system approach was used to model the dynamic behavior of the struc-
ture (NBR6123, 1988; Rosa et al., 2012). Based on the rigid floor dia-
phragm hypothesis, this approach neglected in-plane floor deformations
and the restricted motion of each floor to three degrees of freedom
(DOF): translations on x and y-axes and rotation around the z-axis of the
building (Rosa et al., 2012).
As for the damping ratio, there are several types of damping that
might contribute to the control of a tall building's motion, including:
structural damping “ζs ”; damping ratios “ζd ” originated by dampers; and
aerodynamic damping “ζa .” In the case studies conducted in this paper,
the overall damping value will be equal to 1:25% for building A and
Fig. 1. Concrete strength for each building: A (left) and B (right).
232
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
1:00% for building B, which is consistent with the results obtained by Wu
et al. (2007) for the overall damping during SLS winds, and with the
Brazilian wind code NBR6123, 1988.
The WTT of both buildings used the high frequency pressure inte-
gration (HFPI) method. Along with the building's structural data, this test
can evaluate overall forces at the base (background and resonant), and
modal loads acting on each mode of vibration. In addition, due to the
assessment of precise loads over the building's height, this test allows for
a better evaluation of higher modal loads, i.e., for modes of vibration
after each fundamental sway/torsional mode (Irwin et al., 2013).
Moreover, this test provides a detailed time history of loads distribution
on the building's façade, enabling a precise time domain analysis. These
features make the HFPI a powerful tool to evaluate the responses of tall
buildings to wind-induced loads.
Finally, the users’ comfort during motion in this paper was evalu-
ated by the acceleration at the floor of interest (in the case studies it was
the highest occupied floor). Lateral drift, angular velocity, angular ac-
celeration (yaw), derivative of acceleration (jerk), and frequency of
movement are important parameters, as well as age, body posture, and
quality of insulation, among other physiological and psychological
features. These parameters and features were extensively discussed by
Ferrareto et al. (2015), from where we gathered the compilation of
comfort criteria used in this paper for the current approach to human
comfort. This compilation represents the current and most frequently
used assessment criteria, according to several national/international
standards.
1.3. Reasons for the study
Nowadays, most of the responsibility for the post-treatment of WTT's
results lies mainly on the hands of the wind tunnel facility, with the
exception of structural data. With the set of criteria studied in this paper,
WTT's results may achieve more accurate results and so may structural
engineers.
As a final point, this paper brings knowledge about the impact of
structural design throughout different disciplines and intends to bring
better understanding and verification tools for the whole procedure to all
fields of study that take part in the WTT. This paper provides tools to
enable control and responsibility increasing the role of the structural
engineer during the WTT's analysis of results.
2. Case study and methodology
The choice of the tall buildings studied in this chapter is justified by
their representative features when it comes to their technical context and
location. Together, the buildings represent two of the most used
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Fig. 2. Building's structure: a) blueprint; b) construction photo; c) 3D model.

structural materials/systems currently used in Brazil (and very often used
worldwide): composite structure with concrete core (building A) and
concrete structure with concrete core (building B). Both have absolute
heights and/or relative heights when compared with their surroundings
to be considered as a tall building, and therefore require a WTT (Irwin
et al., 2013). Building B was designed to be the tallest building in S~ao
Paulo when its construction started. Both buildings have office use on the
typical floors.
2.3. FE modelling
FE models were made using a Brazilian commercial structural anal-
ysis software: TQS™. These models were later validated by Robot™ and
Strap™ models. For each building presented (A and B), two different
structural models were made, generating four structural models: A1, A2,
B1, and B2.
The first structural model for building A (model A1) incorporates
usual criteria for FE modelling for tall building's structures, while the

2.1. General information on building A

The building is 137:30m high, with five basement levels, which re-
sults in a structure 157:50m high. The building has 29 typical floors, with
a floor-to-floor height of 4.28 m each, and two technical levels on the top,
including a heliport. The total number of structural levels of this tall
building is 40.
The structural design of the building is based on gravity columns near
its façade, with a stiff concrete core in the center (see Fig. 2). This con-
crete core works for both for vertical loads and horizontal loads (wind
loads). The structure supporting the floor slab is a composite structure
with a steel deck and built-up steel sections with a 8:0cm concrete screed.
The live loads acting on the typical floors is 5:0kN=m2 .

2.2. General information on building B

Building B (see Fig. 3) is 178:00m high, with seven basement levels,

which results in a structure with a height of 203:60m. The building has
35 typical floors, with a floor-to-floor height of 4:00m  5:00m, and four
technical levels on the top, including a heliport. The whole extent of
structural levels of this tall building is 51.
This building structural design is quite similar to that of building A,
based on gravity columns near its façade, with a stiff concrete core in the
center. The main difference lies on the material used, which in this
building is reinforced concrete, instead of the composite structure of
building A. Fig. 3. Studied building (render).

233
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
second model (model A2) incorporates the structural modelling criteria γ
g ¼ 2lnðνðnÞ TÞ þ p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
(3b)
presented in chapter one, which includes the probable modulus of elas- 2
2lnðνðnÞ TÞ
ticity due to concrete ageing, and floor slab modelling. The organization
of building B's models is analogous to that of building A (models B1 and where “γ” stands for the Euler's constant, equal to 0:5772, “T” stands for
B2). the duration of the event; and:
This part of the methodology aims at more accurate FE models for the sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R ∞ 2ðnþ1Þ
subsequent dynamic analysis. The contribution of non-structural walls ðnÞ f Sz ðfÞdf
ν
2
¼ R0 ∞ 2ðnÞ (3c)
(NSW) was not included in this study, since none of the studied buildings f Sz ðfÞdf
0
present in-filled frames with NSW that could work in strut models, as
presented by Cavaleri and Papia (2003), nor were the NSWs continuously where “f” stands for the frequency of motion, and the value of “n” refers
placed on all building stories. to the derivative of the displacement related to time: “n ¼ 0” for
The floor slabs were modelled with shell elements in order to provide displacement; “n ¼ 1” for velocity; “n ¼ 2” for acceleration.
the coupling between two shear walls and between shear walls and the Finally, when all components of acceleration (“ax ”, “ay ”, and “aθ R”)
external columns, discretized by a 50cm  50cm mesh of rod elements to are correctly evaluated, the resultant peak acceleration “aPeak ” is calcu-
simulate a grid. lated using the maximum value between “aPeak;x ”, “aPeak;y ” and
“aPeak;xþy ”, from eq. (4a), (4b) and (4c), respectively (Chen and Huang,
2009).
2.4. BLWT loads, duration of numerical simulations and domain of
dynamic analyses qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  
aPeak;x ¼ ðax Þ2 þ aθ ey  2 ey ax aθ ρaxθ (4a)
The BLWT data for building A correspond to a time-history of 222 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pressure taps over 4920s of duration for each azimuth multiple of 10 ,  2  
aPeak;y ¼ ay þ ðaθ ex Þ2 þ 2 ex ay aθ ρayθ (4b)
starting from the north. Building B has a time history of 363 pressure taps
during 13894s, also for each azimuth multiple of 10 . The angle gaps qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  ffi
chosen for this study follow the same interval as the BLWT's studies, since aPeak;xþy ¼ 0:8 ðax Þ2 þ ay; þ ðaθ RÞ2 þ 2 ex ay aθ ρayθ  ey ax aθ ρaxθ (4c)
a tall building's response can greatly vary between two 10 azimuths
(Irwin et al., 2005). where “ρaxθ ” and “ρayθ ” are the correlation coefficients between trans-
The overall duration of each of the WTT was reduced to smaller du- lational components and the torsional component, adapting Chen and
rations “T” and analyzed separately. Each duration “T” was reduced to Huang's (2009) approach to a greater amount of modes of vibration (see
increase the precision for the Fourier transform, achieving lower values section 2.5). The terms “ex ” and “ey ” stand for the eccentricities of the
for the harmonic frequencies. However, this period was high enough for user's location regarding the axis of reference for the project and they
both buildings to allow the ergodic hypothesis used for the frequency 0;5
hold the relation “ðe2x þ e2y Þ ¼ R”.
domain analysis.
The time domain analysis was performed using a simple 4th order
The overall damping ratio is 1:25% for building A, and the first
Runge-Kutta (Chapra and Canale, 2008). For the time domain analysis,
natural frequency for the first structural model (A1) is 0:2957Hz. For
once the three accelerations components (“ax ”, “ay ” and “aθ R”) are
model B1, these values are 1:00% and 0:2060Hz, respectively. The first
structural model was used to define the minimum period of integra- assessed for each time step, the vectorial sum was made more elaborated
tion, since it is the least stiff one. Considering the given natural fre- in order to compose the complex movement. At each time-step, the ac-
quencies “f”, overall damping ratios “ζ”, and the constants related to celeration was calculated at eight different locations of the building floor
(see Fig. 4). The final acceleration for each time-step was calculated ac-
time “τ ¼ 1=2πfζ” (Jeary, 2003), the minimum time span for the
cording to eq. (5).
Fourier transform is 215s for building A and 434s for building B
  !
(considering a total duration of 5 constants). A 889s period was used j!
a Peak j ¼ !axþ!
ayþ !aθ R  (5)
for both analyses, covering both requirements. The differences be-
tween buildings A and B can be explained by the height difference and The acceleration is calculated at the eight locations of the floor,
structural materials used in their constructions. The 889s period retaining the highest value of these locations. This criterion will be dis-
allowed a precision of 0:0011Hz for the frequency step in the Fourier cussed with more detailing in section 3.3.2.
transform, where 3186 harmonics were obtained to perform the pos- Since the length of time for each test (converted to real scale) was
terior frequency domain analysis. much higher than 889s, several samples were analyzed for each azimuth.
Frequency domain results for rms modal accelerations “σY;j
€ ” were
The peak results obtained from each sample were gathered and final
obtained through a Fourier analysis of each modal load, where each
acceleration component was calculated by:

X
n
 2
σ2ðx;y;θÞ ¼ Φðx;y;θÞ;ℓj σY;j
€ (2)
j¼1

where “Φðx;y;θÞ ” stands for the “jth ” mode shape at the “ℓth ” floor for the
chosen direction (x and y translations or rotation θ), and “n” is the overall
number of modes of vibration used in the analysis.
Some standards allow the direct use of rms acceleration for comfort
assessment (ISO6897, 1984), while others require the peak acceleration
(ISO10137, 2007). In order to transform the rms “σðx;y;θÞ ” into peak ac-
celeration “aðx;y;θÞm
ax ”, a peak factor “g” was applied:

aðx;y;θÞm
ax ¼ gσðx;y;θÞ (3a)

Fig. 4. Locations for resultant acceleration assessment.

234
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

results calculated using their average. acceleration level for the sought return period.
Finally, a sector-by-sector method (Simiu and Filliben, 2005) is car-
2.5. Higher modal contribution ried to provide comparison with the up-crossing method. The sectors
method chosen is the Multi-sector method (Holmes and Bekele, 2015)
For tall building designs, the usual number of modes of vibration used with a sector size of 22:5 , following the available meteorological data.
in a dynamic analysis is 3 (two sway modes and one torsional mode).
However, current investigations indicate non-negligible differences if 3. Results and discussion
higher modal contributions are taken into consideration (Huang and
Chen, 2007; Rosa et al., 2012; Ferrareto et al., 2014). The overall 3.1. Structural modelling criteria
building peak acceleration might increase up to 10%, according to Fer-
rareto et al. (2014), while the torsional acceleration showed an increase 3.1.1. Natural frequencies
of 36%, according to Huang and Chen (2007). Building A's natural frequencies of vibration showed a significant
It is important to state that the increase in the resultant acceleration increase due to the structural modelling criteria (see Table 1). The
for the user's comfort is relevant for the final assessment. The motion modelling of the floor slab and concrete ageing generated the structural
perception by the user, including higher modes and the shape of move- model A2. In the research by Kim et al. (2009), these were steps two and
ments, will be discussed in section 3.2, along with the results of such four and, when combined, represented an overall increase between 10%
increase. and 23% in the natural frequencies of the translational natural modes of
vibration. For the torsional mode, it represented an overall increase be-
2.6. Wind climate modelling tween 16% and 23% in the natural frequency. For model A2, the trans-
lational modes presented an increase between 9:63% (2nd mode) and
Initially, the worst-case method was used to calculate the final re- 15:21% (4th mode, also a torsional mode). For this same building, the
sponses. This approach allows for one to properly observe the directional torsional modes presented an increase of 16:98% on average. This result
dependency of the building's response and compare the three different is consistent with the third building in the research of Kim et al. (2009),
structural models for each building. which presents a structural system that is similar to that of building A,
For the worst-case method, the pressure taps data were normalized by studied in this paper.
the dynamic pressure of the wind tunnel at “1:52m” for building A. In There is a significant increase in the natural frequencies of torsional
order to find the pressure in the real scale, the final recorded value from modes when compared with the translational ones. According to Kim
each tap needed to be multiplied by the dynamic pressure of the wind et al. (2009), the warping rigidity of the diaphragm is related to its
speed at a height of “1:52m  400 ¼ 608m” (since the scale model is out-of-plane rigidity, and the floor slab modelling (which generated
1:400). This pressure must also take into consideration the “fetch” of the building A2) presents an important increase in the stiffness of the vertical
approaching terrain roughness for each azimuth. For building B, the direction of the floor slab. This feature (floor slab modelling) is respon-
reference dynamic pressure is at 500m for the standard open exposure. sible for the higher average increase in natural frequencies of torsional
The methodology described by Irwin et al. (2005) is used in the modes for building A2. Together, the criteria presented in section 2.3
up-crossing analysis. Differently from the sector methods, which attri- indicated overall increases in the natural frequencies of building A,
bute a different response to each direction, the up-crossing method looks varying from 9:63% to 19:66%.
for the asked magnitude of the response, regardless of the direction. To Different modelling criteria allowed for the assessment of natural
develop a better understanding of the procedure, one might observe the frequencies for model B2 (see Table 2). The overall increase in the nat-
wind speed associated with a given building's response magnitude (for ural frequencies varied between 4% and 7%. The results were slightly
example, a 5:0milli  g overall acceleration) for each sector of the
compass. This curve is denominated “response boundary,” which is
assessed for several required pressures, corresponding to several wind
speeds. Then, rate “R” at which the magnitude of the response will Table 1
Comparison of natural frequencies for structural models A1 and A2.
happen is:
Mode fn (Hz) Type
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X u 2  2  k 1  
N
1 _  u jα_ j dVi Ai ki V i  CV ki A1 A2 i1-2
R¼ V i t1 þ  i2 e i (6)
i¼1
2 V_  dα Ci Ci 1 0.2957 0.3262 þ 10.31% Translational
i
2 0.3032 0.3324 þ 9.63% Translational
3 0.4282 0.5124 þ 19.66% Torsional
where “Vi ” is the wind speed, “α” is the azimuth, and “Ai ;”. “Ci ;” and “ki ,” 4 0.9409 1.0840 þ 15.21% Translational with torsional
are parameters of the Weibull distribution of wind speeds for each sector 5 1.0920 1.2250 þ 12.18% components
 
“i”. Parameters “jα_ j” and “V_ ” can be calculated through empirical ex- 6 1.2470 1.4850 þ 19.09%

pressions as functions of wind speed “V”, and are suitable for extra-
tropical winds (Irwin et al., 2005):
  Table 2
V_  ¼ 0:065V þ 0:5e0:252V (7a) Comparison of natural frequencies for structural models B1 and B2.

  Mode fn (Hz) Type

jα_ j ¼ 6:5 1 þ 3:3e0:252V (7b) B1 B2 i1-2
The return period “m,” associated with the magnitude of the chosen 1 0.2060 0.2207 þ 7.13% Translational y dir.
response is calculated by (Irwin et al., 2005): 2 0.2187 0.2274 þ 3.99% Translational x dir.
3 0.3012 0.3186 þ 5.77% Torsional
1 4 0.5821 0.6072 þ 4.31% Torsional
m¼ (8) 5 0.6812 0.7199 þ 5.69% Tors. with transl. in y dir.
8760R
6 0.7380 0.7698 þ 4.31% Tors. with transl. in x dir.
Several pairs of acceleration “am ” vs. return period “m” are obtained 7 1.0581 1.0971 þ 3.69% Torsional
in a tabular form and plotted into a chart, where a regression curve takes 8 1.2327 1.2547 þ 1.78% Translational x dir.
9 1.2737 1.3197 þ 3.61% Tors. with transl. in y dir.
place to interpolate these values and provide the final resultant

235
F.A. Johann
Fig. 5. Building B floor plan.
reduced when compared with building A. This is due to two main factors:
first, the building structural system is composed of moment frames in
both orthogonal directions (see Fig. 5); second, the “stiffness increase
divided by the mass” ratio is considerably smaller for this building when
compared with building A (as the average density is nearly 56% higher
for building B due to its structural material).
The “moment frame” and/or “coupled shear walls” structural systems
in both directions lead to a reduced participation of the floor slab as a
structural element, as previously discussed in the first building's case
study and showed in the study of Kim et al. (2009).
3.1.2. Mode deflection shapes
The structural systems (moment frames and coupled shear walls) in
both directions for building B led to well-behaved mode deflection
shapes in well-defined directions (see Figure A.2, in the Appendix, where
translations in the x and y axes have little or no torsional components)
and no differences between these shapes for the assessed models (B1, and
Fig. 6. Translational mode shapes for models A1 and A2 (modes 1 and 2).
236
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
B2).
The mode deflection shapes for models A1 are presented in Appendix
A: Figure A.1. In Fig. 6, it is possible to verify that the sway component of
the mode deflection shapes for the first two natural modes of vibration
have a considerable difference.
The increase in the natural frequencies of model A2 and the change in
its mode shapes for the sway modes can be explained by two main effects
(Kim et al., 2009):
 Transfer of axial forces from the exterior frame, generating a resistant
moment on the core walls in an effect analogous to an outrigger
system (Choi et al., 2012), but considerably less substantial than the
actual system;
 Increased coupling between the shear walls through the floor slab
(Kim et al., 2009).
Both effects tend to approximate the mode deformation shapes of the
structure to the ones of a moment resisting frame (shown with more
details in Appendix B).
Only the sway components of the mode deflection shapes for the two
first modes of vibration are presented in Fig. 6. Torsional components of
these modes are very similar when we compare models A1 and A2.
The remaining mode deflection shapes are either purely torsional
(3rd mode) or very much stiffer than the first two modes (4th to 6th
modes), with a natural frequency ratio higher than 3:1 between the 4th
mode and the 1st mode, and up to 4:55 between the 6th mode and the 1st
mode. The stiffer modes are much less sensitive to smaller stiffness in-
creases, such as the outrigger effect or the shear wall coupling, and do not
present any significant changes in the mode deflection shapes, neither for
the x axis nor the y axis. These modes (3rd to 6th), however, are purely or
highly torsional and very sensitive to the warping rigidity increase and
present minor changes in the torsional components.
These changes in the mode shapes are responsible for a change in the
modal loads, while the changes in the natural frequencies bring the
resonant response to different values of the PSD of the wind loads. When
combined, these features change the modal responses and their sensi-
tivity concerning the azimuth, and even change the critical direction of
F.A. Johann
the building's response. These changes will be presented in section 3.1.3,
and their impacts in the wind climate modelling are shown in section
3.3.1.
3.1.3. Results from structural modelling criteria
Resultant accelerations using the presented methodology for the
frequency domain analysis are given for all azimuths and two structural
models per building (A1 vs. A2, and B1 vs. B2) in Fig. 7. The rms results
were obtained through eq. (2) and the modal combination was made
using the square route of sum squares (SRSS) rule.
As expected, pattern “A1>A2” is respected for most of the azimuths,
including the maximum acceleration. The critical direction for building
A1 is 220 , where the peak acceleration on the floor of interest is
1:61mili  g. For building A2, these results are 130 and 1:36mili  g.
The way in which the wind climate will be introduced to the pro-
cedure will define how this change in the critical direction can impact the
final acceleration assessment. For the worst-case method, the results are
given and a reduction of 11:4% is observed in the results, due to the use
of the structural modelling criteria presented in this paper.
These and other features of the wind climate associated with the
structural modelling criteria are explored in section 3.3.1. This is an
evidence of how the structural modelling criteria can influence the
Davenport's chain of wind loading.
Resultant accelerations using the same methodology as those used for
building A were calculated for building B using a 1-year return period
(BLWT-normalized wind pressures for 26m=s wind speed at 500m of
altitude), and a total of three modes of vibration. The modal truncation
criteria for this building will be discussed in section 3.2.
As previously observed in Table 2, the structural modelling criteria
play a minor role in building B when compared with building A. Differ-
ently from building A, there is no change in the critical direction for
building B, as there is no change in the mode deflection shapes due to
different modelling criteria.
3.2. Dynamic analysis criteria
The normalized PSD for the components and for the resultant accel-
erations for the 300 azimuth are given in Fig. 8, from which it is plau-
sible to infer that the sixth mode already has very little participation
when compared with the fifth mode. The incremental variation of the
Fig. 7. Resultant accelerations for models A1, A2, B1, and B2.
237
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
sixth mode compared to the fifth mode of vibration for this azimuth is
0:36%. The PSD was normalized with the variance of the respective
component acceleration “σR ” using all modes of vibration available (6 for
building A and 9 for building B). This azimuth was chosen for being the
most critical regarding higher modal contribution, as it will be discussed
later in this chapter.
These results support the justification to study higher modal trunca-
tion in the fifth mode of vibration in the current case study of building A.
However, studies carried out with other tall buildings might need even
higher modal truncation to provide reliable results. This is the case of
building B, which required an analysis with three extra modes (up to the
ninth mode) and results with higher modal truncation for the seventh
mode, as shown in Fig. 9.
For building B, the incremental variation in the resultant acceleration
originated from using modes fourth to sixth compared to the use of the
first three fundamental modes of vibration is 25:2%, whilst the use of the
seventh mode increases the previous response in 3:91% and the use of the
eighth and ninth modes represented increases of less than 0:01%. The 10
azimuth was chosen for being one of the most critical regarding higher
modal contributions for building B.
The rms of the time domain response is lower and very close to the
rms of frequency domain response for nearly all azimuths (building A,
model A2). This behavior corroborates the steady-state theory (Jeary,
2003), as it was also observed in the results found by Wu et al. (2007).
The final accelerations calculated in both time and frequency domains for
all azimuths for building A2 are given in Fig. 10, where it can be observed
that some of the azimuths present responses that are bigger for the time
domain, when compared with the frequency domain (300 , 320 and
70 ).
The correlation factor between the modes in this study is nearly 0:20
between the two first natural modes of vibration, and 0:00 for all other
combinations of modes of vibration for all azimuths. Huang and Chen's
(2007) results for their second situation of study have a similar behavior
between different modes of vibration, and present joint-action factors
around 0:88.
Fig. 11 presents higher modal contribution to the overall acceleration
for building A (model A2).
The higher modal contribution presented the following increases in
the maximum value for each acceleration component:
F.A. Johann
Fig. 8. Normalized PSD of the components and of the resultant accelerations for the 300 azimuth (Building A).
 X axis component (210 and along-wind direction for this sector):
1:86%;
 Y axis component (300 and cross-wind direction for this sector):
32:6%;
 Torsional component: 1:05%.
The overall resultant acceleration presented an increase of 9:47% in
the maximum value, mainly due to a higher modal contribution from Y
axis for the 300 azimuth. This result can also be observed in the PSD
presented in Fig. 8, where modes 4 and 5 contribute with a significant
part of the final response of the building for this azimuth.
Results of higher modal contribution are presented for building B
(model B2) in Fig. 12, where a noticeable increase in the maximum peak
resultant acceleration can be observed. The most perceptible increase of
47% occurs in the 10 sector. In that sector, components x, y, and
torsional presented results of 19%, 23%, and higher than 50%, respec-
tively, due to the contribution of modes 4  7. In this case study, the
contribution of higher modes was equally important for both along-wind
and cross-wind loads, while the torsional component showed to be far
more relevant than both translational components. A closer look into the
PSD of the response of 10 azimuth shows significant contribution from
these modes.
The explanation of such importance to higher modes is the slender-
ness of Brazilian buildings. This slenderness is a direct result of no sig-
nificant seismic activity and quite low wind speeds (30–40 m/s) in the
country, leading to lower natural frequencies of its buildings. This is clear
in Table 2, where a 178:00m high building has six modes with natural
238
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
frequencies under 1:00Hz and the seventh mode with nearly 1:00Hz.
Leading to several modes of vibration (seven at the total) with natural
frequencies highly sensitive to dynamic effects from wind loads, from
which only three were initially used in the WTT.
Resultant acceleration presented an overall increase of 26% as a result
of contributions from modes 4  7 (from 25milli  g in sector 350 to
31:7milli  g in sector 10 ). This increase is quite significant for the user
comfort, just as the 9:47% increase for building A.
However, assembling seven different oscillations (or five, in case of
building A) makes room for doubt regarding the user's perception of such
a movement. In order to verify the perception, the shape of the move-
ment is presented in Fig. 13, where the highest values of displacement
where highlighted in black for the most critical acceleration/displace-
ment time lapse. It is important to note that the movement is nearly oval
with a very elongated shape, which means that despite the complex
combination of five modes of vibration, the user will perceive an uniaxial
movement (Tamura et al., 2006).
3.2.1. Results from dynamic analysis criteria
The resultant acceleration for each model from building A, using the
main dynamic analysis criteria, is presented in Fig. 14, where the worst-
case method was applied and the results for 10 and 50-year return pe-
riods were obtained.
This figure shows that the most severe acceleration occurs in model
A1 in the frequency domain, which happens to be the most common
criteria used by BLWT and structural engineers around the world. A
reasonable reduction was obtained by improving the structural models.
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Fig. 9. Normalized PSD of the components and of the resultant accelerations for the 10 azimuth (Building B).

Fig. 10. Final acceleration results (Model A2, 5 modes of vibration).

239
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Fig. 13. Shape of movement (model A2, time domain, five modes).

Fig. 11. Higher modal contribution (Model A2, 3 and 5 modes of vibration).

Fig. 14. Worst case method resultant peak accelerations vs. return period
(Building A).

case, the time domain analysis with a higher modal contribution for
model A2 presented resultant accelerations higher than the resultant
accelerations in the frequency domain for the same model. The worst-
case method was used in this analysis and generated the results pre-
sented in Fig. 14. However, different results are expected with the use of
not only the most severe acceleration, but also the contribution from each
sector to the final response. The results of the up-crossing method will be
presented in section 3.3.1.
Fig. 12. Higher modal contribution, peak acceleration (model B2, Analogously, for building B, the same set of criteria is presented in
time domain). Fig. 15. In this figure, structural modelling criteria show a reduction of
3:6%. Higher modal contribution (model B2, time domain, seven modes),
on the other hand, presented an increase of nearly 15:8%, leading to an
overall increase in the resultant acceleration of 8.5%, when compared
The three-mode analysis in the time domain showed one of the highest
with model B1 in the frequency domain for three modes.
levels of reduction. However, higher modal contribution showed to be
The dispersion of results is clearly great for this type of analysis and,
quite relevant in these resultant accelerations, and neglecting it would
therefore, consolidating criteria between structural engineers and WTT
lead to less reliable results.
facilities should be a priority when it comes to structural modelling and
The most precise set of criteria did not present lower results. In this

240
F.A. Johann
Fig. 15. Worst case method resultant peak accelerations vs. return period
(Building B).
dynamic analysis. Structural modelling criteria, along with dynamic
analysis criteria, may represent a noticeable increase or decrease in the
building's final response, and should be taken into consideration for
greater precision in the assessment of building motion using WTT data.
3.3. Wind and comfort
3.3.1. Up-crossing method
Previous results of the resultant acceleration were presented using
non-directional wind climate for both buildings, so as to analyze struc-
tural and dynamic analysis criteria purely, with no interference from
other disciplines. In this section, the wind climate modelling is intro-
duced to evaluate its impact on the final response from buildings A and B.
The results from models A1 and A2, for both frequency and time domain,
are presented in Fig. 16.
For the worst-case method, there is a significantly smaller result
(nearly 7% smaller) for model A2 in the time domain (5:52milli  g, 1-
year return period), when compared with model A1 in the frequency
Fig. 16. Up-crossing, Sector-by-sector and Worst case methods (Building A).
241
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
domain (5:92milli  g, 1-year return period) for five modes of vibration.
However, for the up-crossing method, the acceleration response of
building A2 is quite close to those of model A1. This means that the
decrease in the overall acceleration response from model A2 (37%) is
lower than that of model A1 (42%) when the up-crossing method is
applied.
The explanation for this behavior is that model A2 has a critical di-
rection equal to the critical direction of the wind climate, while these
directions are different for model A1 (see Fig. 17).
Fig. 17 shows the probability of exceedance of the wind speed for
Congonhas Airport, in S~ao Paulo (Brazil). The curves from the most
central to the most external positions represent the probabilities of
exceeding that velocity (these probabilities are, respectively, 0:001%,
0:010%, 0:100%, and 1:000%).
In this figure, we can also read the resultant acceleration response for
building A1 in the frequency domain, and for building A2 in the time
domain for five modes of vibration. By overlapping these curves, we can
verify that the stronger contributions of model A2 to the final response
are in azimuths 40 , 190 , and 290 , where the wind speed of the
probability curves reaches its maximum values (or where they are closer
to their maximum values). In contrast, the stronger contribution of
building A1 to the final response is between azimuths 210  260 ,
where the wind speed of these curves has its minimum values. Hence, the
contribution of model A1's responses between azimuths 130  190
(one of the critical azimuth gaps for wind climate) are much smaller than
building A2's contribution.
This is one of the results of the change in the modal deflection shapes
discussed in section 3.1.2. The change in the modal-deflection-shape led
to a change in the critical direction of the building and in the building's
response distribution over the different sectors, which, in turn, led to
different results in the application of the up-crossing method.
Building B's responses are given in Fig. 18, where the application of
the up-crossing method presented the same decrease for models B1 and
B2 (45%). A similar impact from the up-crossing method on the resultant
acceleration was expected for this building, as there was no significant
change in mode deflection shapes or in the natural frequencies due to the
structural modelling criteria, leading to no difference in the critical di-
rection of the building.
Finally, a closer look at figures Figs. 16 and 18 allows for comparison
with the Up-crossing method and the Sector-by-sector method. Results
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Fig. 17. Critical directions (Building A): wind climate and building's responses.

Fig. 18. Up-crossing, Sector-by-sector and Worst case methods results (Building B).

show consistency with the case study presented by Burton et al. (2015),
where an example is shown with a directional building in a directional
wind climate, prevailing the same critical direction both for the wind and
for the building's response. In this case, the Sector-by-sector method is
more conservative than the up-crossing method, as shown in Fig. 19 for
1-year return period.
3.3.2. Comfort assessment
The peak acceleration analysis for time domain is presented in
Fig. 20, where the most simple and conservative set of criteria (models
A1 and B1, frequency domain, 3 modes of vibration, and worst case
method) are compared with the most precise set of criteria, including
wind modelling (models A2 and B2, time domain, higher modal contri-
bution, up-crossing method). Looking at this figure, it is important to
restate the small consistency of comfort criteria around the world. This
lack of consistency can still be justified by the different tolerance to
motion in different countries. The user of a building in Chicago, tradi-
tionally a very windy city of high-rises, may have more tolerance to
motion than the user of a building in S~ao Paulo, a city with a recent
culture of high-rises and with much lower wind speeds in design.
Therefore, the comfort thresholds in the building codes of these two
cities/countries could be different as well.

242
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Fig. 19. Up-crossing, Sector-by-sector and Worst case methods results (Buildings A and B).

Fig. 20. Peak acceleration analysis (Buildings A and B).

Fig. 21. Peak and rms responses at different floor locations (300 sector).

ISO10137 (2007) and NBR6123, 1988 are related to perception,
while the criteria of NBCC (1990) and CTBUH are related to comfort. The
verification of comfort in tall buildings in Brazil should more focused on
the perception thresholds for the next years, due to its recent history with
high-rises. NBR6123, 1988 is currently under revision to integrate more
recent and precise thresholds, as well as dynamic analysis criteria. Thus,
it is reasonable to pursue comfort through thresholds closer to perception
of motion until this practice is better developed in Brazil. The most severe
peak acceleration of the floor will be used, based on the location of the
building (Brazil) and on the type of assessment (perception).

243
F.A. Johann
The inconsistency in the overall criteria, conversely, is more con-
cerning. Both buildings A and B, in the most conservative set of criteria,
do not pass the comfort verification using the threshold defined in
NBR6123, 1988. On the other hand, using more precise parameters, and
comparing them with codes with higher thresholds, like the NBCC
(1990), the resultant accelerations achieve a ratio where they are only
29% to 64% of the comfort threshold for the same return period (10
years). Hence, all sets of criteria pass in the comfort verification for a
smaller return period (1 year), with the exception of the worst-case
method for model B1, which passes the comfort verification (Sarkisian,
2012), but doesn't pass the SLS verification based on perception
(ISO10137, 2007).
Comparing three modal responses with time domain (model A2)
and frequency domain (model A1), the time domain response is even
lower and should lead to a more economic design (allowing further
structural optimization). However, considering higher modal contri-
butions, the final acceleration for model A2 in the time domain is quite
similar to the responses for model A1 in the frequency domain. This
means that this could be the case where “two wrongs made a right,”
and the different response behavior throughout the azimuths and
higher modal contributions (which lead to higher responses) balanced
the reduced response with the use of structural modelling criteria and
time domain analysis (which lead to lower responses).
Fig. 20 shows the worst peak resultant acceleration, considering the
most critical location of the user on the highest occupied floor. The
subsequent analysis will explore the aspect of this location in order to
evaluate the human comfort during wind-induced motions. The other
peak and rms accelerations for the other top occupied floor positions are
shown in Fig. 21.
In this figure, we can observe that the rms accelerations do not pre-
sent a significant variation over the floor positions, while the peak ac-
celerations do. The rms accelerations adopt an almost constant value,
with an average of 1:03milli  g throughout the positions, while the peak
accelerations vary between 3:49milli  g and 5:52milli  g.
This variation occurs as a result of the torsional components of both
torsional and translational modes of vibration, which is particularly
significant in this project. The constructive interference between the
translational and the torsional components in one position of the floor is
a destructive interference in the opposite direction, generating these
discrepancies.
As a result, the peak factors for this floor also vary greatly. The peak
factor presented for sector 300 (critical sector in the final analysis of
building A2) in the previous sections is 5:34 and corresponds to the
highest peak acceleration, while the average peak factor for this floor is
4:31.
To correctly evaluate the comfort level of the user, one must define
whether the perception of movement or the annoyance levels will be
assessed. If the annoyance levels are chosen, the average value and
probably the rms acceleration should be used. However, if the perception
of movement is chosen, then there is a chance that the occupants in the
region with the most severe result will perceive the motion and alarm the
other occupants on the floor. In this case, the peak acceleration is prob-
ably the most suitable way to evaluate comfort.
4. Conclusions
First, the modelling criteria are very relevant for comfort and for
the SLS analysis. They represent a substantial increase in the building's
natural frequencies (up to 19:66%, as observed in building A). In
addition, there were visible changes in mode deflection shapes,
Appendix A. Supplementary data
Supplementary data related to this article can be found at https://doi.org/10.1016/j.jweia.2018.07.006.
244
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
Table 3
Summary of results of each set of criteria for buildings A and B.
Building Structural Higher modal Wind Comfort Total
modelling contribution climate criteria
modelling
A 11.4% 6.4% 39.8% 36.9% ¡66.9%
B 3.6% 15.8% 44.7% 36.9% ¡62.1%
generating relevant deviations in the probabilistic wind climate
analysis.
Time domain analyses present several paybacks when compared with
the frequency domain analyses. The assessment of complex motion is one
of them, where one can precisely calculate the exact vector acceleration
at any location on the floor for all time steps. This feature allows for a
precise modal combination (including higher modes) and peak re-
sponses, without the need for any modal combination rule or statistical
peak factors, and also allows for the assessment of the exact values
through the rms acceleration.
Higher modes presented an important contribution to the final ac-
celeration (up to 15:8%). Part of this increase in response, however, is
largely compensated by a decrease in other criteria, as it can be observed
in Table 3.
The wind climate criteria presented the most relevant impact in the
final response, decreasing the overall response by nearly 37  45%.
Different mode shapes that result from structural modelling criteria also
contributed to the wind climate analyses, since these mode shapes
changed the critical direction of the building.
Table 3 shows the summary of variation results for each set of criteria
studied in this paper through an exhaustive analysis of Figs. 10, 14, 15
and 20.
Structural modelling criteria were evaluated by a comparison of the
results of models A1 vs. A2, and B1 vs. B2 in the frequency domain for
three modes of vibration (worst-case method). Higher modal contribu-
tion is evaluated by comparison of the results from building A2 (B2) for
three and five (seven) modes in the time domain (worst case method).
Wind modelling criteria could be evaluated through the variation of the
results of model A2 (B2) for the worst-case method and for the up-
crossing method.
The comfort criteria were evaluated using the ratios of a specific set of
criteria (A2/B2, 5/7 modes, time domain, up-crossing method) and two
comfort thresholds: ISO 10137 (2007) and Sarkisian's (2012) for offices.
For building A, these ratios were, respectively: 3:53=8:20 ¼ 0:43, and
3:53=13:0 ¼ 0:27, and its variation was  36:9%. This calculation works
analogously for building B.
This result shows that the impact of design criteria may represent
response reductions up to 66:9% (for building A) when compared with
the results obtained from a conservative analysis and a conservative
threshold. A reduction through different sets of criteria lead to more
efficient designs, which, in turn, lead to the reduced use of material,
lower costs, and lower embodied energy of the structure.
Acknowledgements
A special thanks to the following companies and universities, without
which the results obtained in this paper wouldn't be possible: Escola
Polit
ecnica da Universidade de S~
ao Paulo, Alan G. Davenport Wind Engi-
neering Group, França & Associados Projetos Estruturais, Rowan Williams
Davies and Irwin Inc.
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Appendix A. Mode deflection shapes

Fig. A.1. Mode shapes for structural model A1.

245
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Fig. A.2. Mode shapes for structural models B1 and B2.

246
F.A. Johann Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248

Appendix B. mode deflection shapes and Building A's critical wind direction

As explained in section 3.1 and presented in Fig. 6, the FE models A1 and A2, generated from building A, presented many different structural
features. Undeniably, mode deflection shapes from both models present important differences due to the modelling of the floor slab criterion. These
differences led to the results shown in Fig. 7, where it can be observed a sensitive difference of behavior or response pattern in a few sectors (mostly
around azimuths 160 , 200 and 260 , in order of significance).

Fig. B.1. Building A floor slab.

To give a better overview of the impact of structural modelling in the mode deflection shapes, one can observe Figure B.1, featuring the mode
deflection shape of the second mode of vibration, which is the mode of vibration more susceptible to benefit from frame behavior of the floor slab. In this
figure, it is possible to observe:

1) Floor slab deflection pattern following shear-wall connections in a straight line;

2) Cross section of the modelled floor slab, which for composite floors presents high stiffness due to the presence of the steel section.
3) Clamped portion of the floor slab to the external columns.

The frame behavior observed in FE model A2 is not present in model A1 and despite the very low values of bending moment acting on these floor
slabs, the cumulative effect on every building floor is very expressive and, for this building, it generates the differences in the mode deflection shapes
and in the natural frequencies presented respectively in Fig. 6 and in Table 1.

References Ferrareto, J.A., França, R.L.S., Mazzilli, C.E.N., 2014. The impact of comfort assessment
criteria on building design. In: Proceedings of Council of Tall Buildings and Urban
Habitat (CTBUH) – Future Cities, towards Sustained Vertical Urbanism. Shanghai,
Ali, M.M., Moon, K.S., 2007. Structural developments in tall buildings: current trends and
pp. 722–730.
future prospects. Architect. Sci. Rev. 50 (3), 205–223.
Ferrareto, J.A., França, R.L.S., Mazzilli, C.E.N., 2015. Wind-induced motion on tall
Burton, M., Li, Y.F., Cammelli, S., 2015. Wind directionality effects: revisiting an old
buildings: a comfort criteria overview. J. Wind Eng. Ind. Aerod. 142, 26–42.
conundrum. In: XIV International Conference on Wind Engineering (ICWE), Porto
Hansen, R.J., Reed, J.W., Vanmarcke, E.H., 1973. Human response to wind-induced
Alegre.
motion of buildings. J. Struct. Div., ASCE 99 (7), 1589–1605.
Cavaleri, L., Papia, M., 2003. A new dynamic identification technique: application to the
Holmes, J.D., Bekele, S.A., 2015. Directionality and wind-induced response – calculation
evaluation of the equivalent strut for infilled frames. Eng. Struct. 25, 889–901.
by sector methods. In: XIV International Conference on Wind Engineering (ICWE),
Chapra, S.C., Canale, R.P., 2008. M
etodos Num
ericos para Engenharia, fifth ed. McGraw-
Porto Alegre.
Hill Interamericana do Brasil.
Huang, G., Chen, X., 2007. Wind load effects and equivalent static wind loads of tall
Chen, X., Huang, G., 2009. Evaluation of peak resultant response for wind-excited tall
buildings based on synchronous pressure measurements. Eng. Struct. 29, 2641–2653.
buildings. Eng. Struct. 31, 858–868.
Irwin, P., Denoon, R., Scott, D., 2013. Wind Tunnel Testing of High-rise Buildings: an
Choi, H.S., Goman, H., Joseph, L., Mathias, N., 2012. Outrigger Design for High-rise
Output of the CTBUH Wind Engineering Working Group. Council of Tall Buildings
Building: an Output of the CTBUH Outrigger Working. Group Council of Tall
and Urban Habitat, Chicago.
Buildings and Urban Habitat, Chicago.
Irwin, P., Garber, J., Ho, E., 2005. Integration of wind tunnel data with full scale wind
Drew, C., Fernandez, K., Fanning, K., 2014. The Environmental Impact of Tall vs. Small: a
climate. In: Proceedings of the 10th Americas Conference on Wind Engineering.
comparative study. In: Proceedings of Council of Tall Buildings and Urban Habitat
Baton Rouge, LA.
(CTBUH) – Future Cities. Towards Sustained Vertical Urbanism. Shanghai,
pp. 418–424.

247
F.A. Johann
ISO6897, 1984. Guidelines for the Evaluation of the Response of Occupants of Fixed
Structures, Especially Buildings and Offshore Structures, to Low-frequency Horizontal
Motion (0.063 to 1 Hz). International Organization for Standardization, Geneva.
ISO10137, 2007. Bases for Design of Structures – Serviceability of Buildings and
Walkways against Vibrations. International Organization for Standardization,
Geneva.
Jeary, A.P., 2003. The effect of time-constants on the response of very large structures. In:
Eleventh International Conference on Wind Engineering.
Kim, Y.J., Yu, E., Kim, Y.D., Kim, S., 2009. Calibration of analytical models to assess wind-
induced acceleration responses of tall buildings in serviceability level. Eng. Struct.
31, 2086–2096.
NBCC, 1990. National Building Code of Canada. Part 4 Structural Design (Chapter 4)
Commentaries on Part 4 of the National Building Code.
NBR6118, 2014. Projeto de estruturas de concreto – Procedimento. Associaç~ao Brasileira
de Normas T
ecnicas, Rio de Janeiro, Brasil.
248
Journal of Wind Engineering & Industrial Aerodynamics 180 (2018) 231–248
NBR6123, 1988. Forças Devidas Ao Vento Em Edificaç~ oes. Associaç~ao Brasileira de
Normas T
ecnicas, Rio de Janeiro, Brasil.
Rosa, L., Tomasini, G., Zasso, A., Aly, A.M., 2012. Wind-induced dynamics and loads in a
prismatic slender building: a modal approach based on unsteady pressure
measurements. J. Wind Eng. Ind. Aerod. 107–108, 118–130.
Sarkisian, M.P., 2012. Designing Tall Buildings Structure as Architecture. Routledge, New
York.
Simiu, E., Filliben, J.J., 2005. Wind tunnel testing and the sector-by-sector approach to
wind directionality effect. J. Struct. Eng., ASCE 131 (7), 1143–1145.
Tamura, Y., Kawana, S., Nakamura, O., Kanda, J., Nakada, S., 2006. Evaluation
perception of wind-induced vibration in buildings. Proc. Inst. Civ. Eng. Struct. Build.
159 (SB5), 283–293.
Wu, J.R., Liu, P.F., Li, Q.S., 2007. Effects of amplitude-dependent damping and time
constant on wind-induced responses of super tall building. Comput. Struct. 85,
1165–1176.