INTER-NOISE 2016

ANALYSIS OF WIND-INDUCED VIBRATIONS

IN HIGH-RISE BUILDINGS
Peter PERSSON1; Per-Erik AUSTRELL2; Poul Henning KIRKEGAARD3;
Lars Vabbersgaard ANDERSEN4; Fredrik STEFFEN5
1, 2, 5
Department of Construction Sciences, Lund University, Sweden
3
Department of Engineering, Aarhus University, Denmark
4
Department of Civil Engineering, Aalborg University, Denmark

ABSTRACT
Buildings are getting taller due to increased urbanisation and densification of cities. More
advanced construction methods and the desire to construct impressive buildings is also supporting the
trend. Due to their inherent slenderness resulting in low eigenfrequencies, these buildings are
susceptible to wind-induced vibrations which can be highly disturbing for occupants. Already barely
perceivable acceleration levels within the low-frequency range relevant to whole-body vibrations can
cause nausea and discomfort, while high acceleration levels can cause alarm and fea r amongst the
occupants. The paper summarises acceptable acceleration levels in high -rise buildings (here referred
to as buildings over 200 m in height) stated in different building codes and previous work on the
subject. Accelerations of a high-rise building subjected to wind-loads are evaluated using a full
numerical model and one reduced with Ritz-vectors and the results are compared. The paper focusses
on wind-load dynamics in early stages of the design process, with an intent to give an indication of
the dynamic properties of a building.

Keywords: High-rise buildings, wind-load response, Ritz-vectors

I-INCE Classification of Subjects Numbers: 22.3, 42, 49.1, 49.3.1, 75.3, 76.9.

1. INTRODUCTION
With advancement in material and construction sciences, buildings have been getting taller and
taller during the last century. High-strength steel, lighter cladding and modern construction techniques
have resulted in tall buildings, thus also giving more slender buildings with lower and lower natural
frequencies. Structures exposed to stochastic loading such as earthquake or wind load have a tendency
to vibrate in the first natural frequency, i.e. the fundamental frequency. For high-rise buildings, here
assumed to be around or over 200 m in height, the fundamental frequency can easily be below 1 Hz
due to the slender shape of these buildings. A rule of thumb to estimate the fundamental frequency of
a building of height H (in m) is f = 46/H, where the frequency f is given in Hz. The estimate is given
in the Eurocode (1) and is applicable to buildings with a height large r than 50 meters (1).
In designing buildings, serviceability criteria are often harder to fulfil than the survivability ones.
High acceleration levels may cause nausea and even alarm among residents. The acceptable
acceleration levels in terms of comfort for the residents have been studied by several researchers, as
described below, showing that these accelerations vary with frequency. As an example, at 0.1 Hz the
allowable acceleration is around 15 mg (15·10 -3 ‫ ڄ‬9.81m/s 2) giving a peak-to-peak range of the
displacement of 0.7 m if assuming a sinusoidal motion. For a building resonating at 1 Hz, the

1
peter.persson@construction.lth.se
2
per-erik.austrell@construction.lth.se
3
phk@eng.au.dk
4
la@civil.aau.dk
5
fredrik.steffen@gmail.com

7650
 INTER-NOISE 2016

allowable acceleration is around 5 mg giving a peak-to-peak displacement range of about 3 mm.

Accelerations of a high-rise building subjected to wind loads are evaluated in this work using a
numerical model. A high-rise building is modelled with beam elements using the finite-element (FE)
method to construct a model with large number of degrees of freedom (dofs). The full system is then
compared to a greatly reduced model using Ritz-vectors formed from static deflections. The two
models are compared in terms of natural frequencies and obtained accelerations.
The work is focused on wind-load dynamics in early stages of the design process in order to give
an indication of the dynamic properties of the modelled building.

2. VIBRATIONS IN HIGH-RISE BUILDINGS

Since tall buildings with low natural frequencies can vibrate significantly due to wind excitation,
it is important to understand how this effects the tenants. Here, human response to low-frequency
accelerations and acceptable accelerations in high-rise buildings are described.

2.1 Performed Studies and Experiments

Throughout the years, several studies have been conducted on the human perception of vibrations
in buildings. The studies usually fall within one of three types of studies: field experiments and
surveys of occupants in tall building; motion simulator and shaker table experiments; and field
experiments in artificially excited buildings. The results of several studies have been summ arised by
Kwok et al. (2) and some of those, which fall within the first two categories mentioned above, are
discussed below. Few, if any, field experiments in artificially excited building s with natural
frequencies below 1 Hz (which is usually associated with very tall buildings) have been done and will
therefore not be reviewed.
Field experiments and surveys can be divided into two categories. The first category is surveys
conducted in buildings during or after a passing storm. The survey results are late r on compared to
data measured from wind-tunnel testing of that particular building. Result that is more useful have
been derived from buildings that are known to have complaints from occupants regarding disturbing
vibrations. This has allowed testers to properly install equipment to be able to register the actual
vibrations, such that these vibrations can be compared to the experience of the occupants.
Denoon et al. (3-5) conducted experiments in three airport towers equipped with accelerometers
and anemometers. With help of the test it was concluded that the average threshold of perception
varied with the natural frequency of the building. It is noteworthy that even though two of the
buildings experienced accelerations that were acceptable according to the ISO 6897:1984 (6) which
applied back then, one of the buildings received more complaints. It is believed that the tower was
exposed to winds during a longer period of time, suggesting that exposure durations effect the
perception of vibration. This is an important consideration, since the characteristics of windstorms
differ greatly across the world. Recently, Lamb et al. (7) conducted surveys in Wellington, New
Zeeland—a notoriously windy city. That, combined with the city’s high earthquake risk that forces
designers to allow more structural flexibility, makes the buildings prone to dynamic response. It was
noted that there is a significant increase among employees experiencing nausea and dizziness when
building motion was possibly or definitely perceivable. The study showed that 60-70% detected
building motion by feeling the sensation of motion. The rest detected it by sound and visual cues.
Symptoms such as nausea and dizziness did not differ significantly when motion was possible or
definite, suggesting the accelerations that are barely perceiver, or not perceived at all, might still affect
occupants.
To circumvent the uncertainties associated with field experiment, a series of controlled-motion
simulators and shaker-table experiments have been performed. The early experiments focused on
subjects exposed to sinusoidal vibrations and on vibration perception without the subject being
distracted by a task. Even though this may not be a realistic comparison to random wind-induced
building vibration, some of the conclusions drawn from the study still hold interest. One of the most
referenced study to date is by Chen and Robertson (8). It concluded that the frequency of oscillation,
body movement and expectancy of vibrations are factors that in large degree effect the vibration
perception threshold. Irwin and Goto (9) continued the study by Irwin (10) and concluded that
frequencies below 1 Hz, which are common for high-rise buildings, resulted in more nausea and
abdominal discomfort, while frequencies above 1 Hz were more disturbing when subjects tried to
perform manual tasks.

7651
INTER-NOISE 2016

2.2 Recommendations and Codes

Data gathered from some of the studies mentioned above led to the development of ISO 6897:1987
(6). The limits were set at the maximum standard deviation of acceleration from the worst 10 minutes
of windstorms occurring every 5 years. The standard was later modified by Melbourne and Cheung
(11) to specify a peak acceleration instead of the mean deviation as used in ISO 6897:1987 (6). As the
ISO norms, the acceleration criteria depend on frequency and were given for different return periods.
In 1993, Isyumov (12) suggested ranges of accelerations that were incorporated in the National
Building Code of Canada (NBCC) in the 1995 (13). Instead of a frequency-dependent criterion,
intervals were suggested depending on the building’s usage: 5–7 mg for residences, 7–9 mg for hotels
and 9–12 mg for offices (14). The Architectural Institute of Japan (AIJ) (15) has gone in another
direction. Instead of giving thresholds for human comfort, the accelerations for 10, 30, 50, 70, and
90% probabilities of perception have been suggested. This lets owners decided for themselves what
accelerations at an annual return period can be considered acceptable. ISO 6897:1984 has later been
revised and superseded by ISO 10137:2008 (16), shown in Figure 1. The acceptability curve for
resident buildings lies close to that of AIJ’s 90% probability curve. The criterion for residential
buildings is 2/3 of that acceptable in offices. A number of proposed criteria and perception threshold s
have been summarised by Burton et al. (14) and are shown in Figure 2.

Figure 1 – Acceptable accelerations from ISO 10137:2008 (16) in 1: offices, 2: residences. A is in m/s2.

Figure 2 – Proposed criterion and perception curves summarised by Burton et al. (14).

7652
 INTER-NOISE 2016

2.3 Wind-Tunnel Testing

For designers to gain insight into a building’s dynamic behaviour, wind tunnel testing if often
performed. One major step forward in wind engineering was the development of the boundary-layer
wind tunnel. While many different test setups are available, one of the most common is the High-
Frequency-Force-Balance Method. A rigid foam model with a high natural frequency is placed on a
sensitive five-component balance. By performing several tests, an average response spectrum of the
base bending moment is obtained. The response spectrum is only valid for the particular building
shape and environmental exposure tested, while being normalised to be independent of wind velocity
and structural parameters such as stiffness, damping and mass (17). Figure 7 shows the basic setup of
a balance model.

Figure 3 – A base-pivoted building model [21].

3. NUMERICAL MODELLING AND SIMULATIONS

This chapter discusses numerical simulation of a building subjected to excitation by wind. It begins
with some basic theory about wind loading, followed by how the considered building has been
modelled. Results for the full FE model and for a model reduced by Ritz-vectors are compared.

3.1 Wind Loading

The most common analytical procedure for evaluating the response of a building was developed
by Davenport in 1967 (18). A peak response can be calculated from the mean response using a gust
loading factor giving a dynamic amplification. The method can be used to calculate a number of
responses such as forces, acceleration and displacement.
Vortex shedding is a phenomenon that has a high impact on acceleration levels in tall and slender
buildings. The oncoming wind is shed alternately on each side of the building, resulting in a force
acting perpendicular to the wind direction. Eurocode (1) gives the critical velocity where the vortex
shedding frequency coincides with the natural frequency of a building. The critical velocity can be
described as
ܾ‫݊ڄ‬
‫ݒ‬௖௥௜௧ ൌ (1)
ܵ‫ݐ‬
where b is the width of the building perpendicular to the oncoming wind, n is the natural frequency
of the building, and St is the Strouhal number, defined by Figure 4.

7653
INTER-NOISE 2016

Figure 4 – Strouhal number for rectangular cross sections with sharp corners [Eurocode].

The numerical method used here employs another approach using spectra from wind-tunnel testing.
The wind load will be applied in two directions: the along-wind and the across-wind directions. A
subroutine creates a target spectrum in the along-wind and across-wind directions for the idealised
building. Then an inverse Fast Fourier Transform (FFT) is used to generate two time series from the
spectrum. Finally, an FFT on the time-series is performed to confirm that the generated time series
include the same frequencies as the target spectrum. Another subroutine generates the equivalent static
wind load used by the main program to calculate a static deflection which is later used as a Ritz-vector
to reduce the system.

3.2 FE Model
The FE model has a symmetric cross section with centre of mass and moment of inertia coinciding
on each floor. The system will be reduced using the fundamental frequency of the building, higher
modes being neglected. The building is modelled in CALFEM, a finite element toolbox developed at
the Department of Construction Sciences at Lund University using MATLAB.
The code developed consist of a main program and two subroutines. The main program generates
the mass and stiffness matrix for an idealised building. The building mass is equally distributed to the
nodal points applying the concept of lumped mass. The stiffness is adjusted to provide a fundamental
frequency close to that of real high-rise buildings.
The model building with 85 floors and a rectangular cross section of 50u50 m2 is shown in Figure 5.
The building has a floor height of 3.5 m providing a total height of 297.5 m. The density was set to
180 kg/m3 . The lowest natural frequency is 0.1574 Hz and the associated mode can be described as
global bending in the horizontal x-direction. The total system has of 3360 nodes and 8160 dofs. The
buildings is modelled with beam elements fully tied to eachother. Fixed boundary conditions were
applied at height z=0, see Figure 5.

Figure 5 – Structure modelled in CALFEM.

7654
 INTER-NOISE 2016

The damping matrix is determined from the mass matrix m and the stiffness matrix k using
Rayleigh damping (19):
ࢉ ൌ ܽ଴ ࢓ ൅ ܽଵ ࢑ (2)
where
ʹ߱ଵ ߱ଶ
ܽ଴ ൌ ߞ (3)
߱ଵ ൅߱ଶ
ʹ
ܽଵ ൌ ߞ  (4)
߱ଵ ൅߱ଶ
With a structural damping ratio of ] = 1% (20) for modes 1 and 2, and first and second natural
frequencies of f1 = 0.1574 Hz and f2 = 0.4777 Hz, respectively, the factors a 0 and a 1 become 0.0024
and 0.0315. The damping ratio includes energy dissipation such as material damping and friction in
joints.

3.3 Ritz-vectors
This section gives a brief overview of model reduction using Ritz-vectors. The theory is thoroughly
described by Chopra (19). The equation of motion for a system with N dofs can be described as
࢓࢛ሷ ൅ ࢉ࢛ሶ ൅ ࢑࢛ ൌ ࢖ሺ‫ݐ‬ሻ (5)

By selecting a number of Ritz vectors, ࣒௜ , i = 1, 2, …, J, representing expected mode shapes, a large

system can be reduced and solved, saving computational time:
௃

࢛ ൎ ෍ ‫ݖ‬௜ ሺ‫ݐ‬ሻ࣒௜ ൌ શࢠሺ‫ݐ‬ሻ (6)

௜ୀଵ

Here શ is a matrix storing the Ritz-vectors as columns, and ࢠሺ‫ݐ‬ሻ contains the coordinates ‫ݖ‬௜ ሺ‫ݐ‬ሻ, i = 1,
2, …, J, of the Ritz-vector basis. Substituting this into the original equations of motion (5), the
equations of motion can be rewritten as
࢓શࢠሷ ൅ ࢉશࢠሶ ൅ ࢑શࢠ ൌ ‫݌‬ሺ‫ݐ‬ሻ (7)
‫܂‬
Furthermore, pre-multiplying with શ yields the formulation
࢓ ෩ࢠ ൌ ࢖
෥ ࢠሷ ൅ ࢉ෤ࢠሶ ൅ ࢑ ෥ ሺ‫ݐ‬ሻ (8)
giving a reduced system where
෥ ൌ શ ࢀ ࢓શǡࢉ
࢓ ෩ ൌ શ ࢀ ࢑શǡ ƒ†࢖
෥ ൌ શ ࢀ ࢉશǡ࢑ ෥ ൌ શࢀ࢖ (9)
After solving the smaller system, detailed information about the full system can be back calculated
with Eq. (6).
The accuracy depends on the choice of Ritz-vectors. A good way of determining the Ritz-vectors
is by applying an equivalent static wind load and using the calculated deflection as a Ritz-vector. The
shape corresponds well with the first mode of vibration, thus making it a suitable approximation. In
this work, only two Ritz-vectors were needed to adequately reproduce the response of the model
building. It was found that one deflected shape in the along-wind direction together with a deflected
shape in the across-wind direction was enough to represent the building’s response to wind load.

3.4 Equivalent Static Wind Load

To determine the Ritz-vectors, a static deflection was calculated by applying a static wind-load.
Eurocode (1) does not provide a method of determining the Equivalent Static Wind Load (ESWL) on
a building. However, a method for the ESWL in the along-wind direction has been developed by Zhou
et al. (21). It is based on the principle of a Gust Loading Factor (GLF) originally developed by
Davenport and is more suitable for tall and slender buildings. Combining the two methods provides
the total equivalent static wind load on each floor according to Figure 6.

7655
INTER-NOISE 2016

Figure 6 – Equivalent static wind load on each floor.

Also for the across-wind direction, Eurocode provides no means for calculation of the EWSL.
While AIJ may be suitable for common high-rise buildings, it is not suitable for very slender buildings
where vortex shedding has a significant impact on the building response. Instead, a method developed
by Quan et al. (22) is used. The peak ESWL for each floor can be calculated according to Figure 6.

3.5 Wind-Load Time Series

From wind-tunnel testing, the spectral representation of the base bending moment is given. In
certain building codes and articles, simplified expressions for the spectrum are available. In this paper,
the spectrum described in the Japanese building code (AIJ) (23) and by Qu et al. (24) is utilized to
calculate the spectral density of the base bending moment in the along-wind and across-wind
directions, respectively.

7656
 INTER-NOISE 2016

Figure 7 – Time series and frequency spectrum of along-wind base bending moment.

Figure 8 – Time series and frequency spectrum of across-wind base bending moment.

Using these spectra, an Inverse Fast Fourier Transformation (IFFT) procedure with some added
noise and random phase angles can be performed, creating time series of the wind load. By performing
an FFT on the generated time series, a spectral representation closely matching that of closed-form
expressions has been found. The IFFT was manually scaled. Figure 7 and 8 show the generated time
series and their spectral density compared to the target spectrum, for base bending moment.

7657
INTER-NOISE 2016

4. MODEL COMPARISON – RESULTS

By applying the static wind load calculated using the methods presented above and illustrated in
Figure 6, a deflection was determined. Figure 6 shows EWSL in the along-wind and across-wind
directions. By applying the static load to the structure and calculate the deflection, two Ritz-vectors
were defined. By reducing the system and solving the eigenvalue problem, the frequencies were
determined to be 0.1574 Hz and 0.1574 Hz in the along-wind and across-wind directions, respectively..
Table 1 shows a comparison of the natural frequencies between the full and the reduced systems.

Table 1 – Natural frequencies

Frequency along, Hz Frequency across, Hz
Full system 0.1574 0.1574
Reduced system 0.1574 0.1547
Difference 0.0% 0.0%

This shows that the Ritz-vector are accurate and can be used in further dynamic simulations, given
the proximity of the natural frequencies in the full and reduced systems. During a 600 seconds long
time-series simulation, two degrees of freedom at the top of the building were observed: one in the
along-wind direction and one in the across-wind direction. Table 2 shows the time necessary to
complete the analysis for the full and reduced systems.

Table 2 – Simulation time.

Time, s
Full system 3690
Reduced system 0.23

Figures 9 and 10 show the acceleration response of the full and the reduced system in along- and
across-wind directions respectively. Furthermore, RMS values were calculated to show the similarity
between the two systems. The results are shown in Table 3.

Table 3 – RMS acceleration.

Along-wind Across-wind
Full system 0.5599 0.1092
Reduced system 0.5570 0.1099
Difference 0.52% -0.60%

7658
 INTER-NOISE 2016

Figure 6 – Comparison between full and reduced systems – along-wind direction.

Figure 7 – Comparison between full and reduced systems – across-wind direction.

5. CONCLUDING REMARKS
Designers and building owners should be aware of the complexity associated with structural design
of high-rise building exposed to significant wind loads. Not only are there uncertainties in the
approximation of the load and of the dynamic response itself, but also people react very differently to
accelerations. Careful consideration has to be taken regarding: the building properties in terms of
stiffness, mass and damping; type of storms likely to occur at its location; and whether it is to be an
office building, residential building or a mix of both. Since removing all noticeable vibrations is
uneconomical from a building-owner point-of-view, the owner will have to decide how much vibration
can be tolerated without effecting the rental or sale plans.
In this paper, a time-efficient and accurate numerical procedure for determining accelerations in
high-rise buildings was developed. The simulation of a system reduced with Ritz-vectors showed a

7659
INTER-NOISE 2016

high correlation with the whole system, while saving substantial computational time. However, one
should be aware that with even taller and more slender buildings than the one used here, higher modes
become more influential. Neglecting them may result in an underestimation of the true response.
Further studies could try to determine under which prerequisites using only the first mode of
vibration is satisfactory. Using the static deflection from an equivalent static wind load works
sufficiently for estimating the first mode, but estimating higher modes may become challenging and
would also be suitable for further study.

ACKNOWLEDGEMENTS
The authors would like to show their gratitude to the European Union for financial support via the
Interreg V project “Urban Tranquillity”.

REFERENCES

1. Eurocode 1: Actions on structures - Part 1-4: General actions - wind actions, SS-EN 1991-1-4:2005
2. Kwok KC, Hitchcock PA, Burton MD. Perception of vibration and occupant comfort in wind-excited tall
buildings. Journal of Wind Engineering and Industrial Aerodynamics. 2009 Oct 31;97(7):368-80.
3. Denoon RO, Letchford CW, KCS K, Morrison DL. Field measurements of human reaction to wind-
induced building motion. In Wind Engineering Into the 21st Century, Vol 1-3 1999 (Vol. 1, pp. 637-644).
AA Balkema.
4. Denoon RO, Roberts RD, Letchford CW, Kwok KC. Field experiments to investigate occupant perception
and tolerance of wind-induced building motion. Research report. University of Sydney. Department of
Civil Engineering. 2000.
5. Denoon R.O. Designing for wind-induced serviceability in buildings. PhD Thesis. The University of
Queensland. 2001.
6. International Standards Organization. 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 1.0
Hz) ISO6897:1984. Geneva, Switzerland. 1984.
7. Lamb S, Kwok KC, Walton D. A longitudinal field study of the effects of wind-induced building motion
on occupant wellbeing and work performance. Journal of Wind Engineering and Industrial Aerodynamics.
2014 Oct 31;133:39-51.
8. Chen, P. W., Robertson, L. E. Human perception thresholds of horizontal motion. Journal of the structural
division 1972;92(8):1681-1695.
9. Irwin AW, Goto T. Human perception, task performance and simulator sickness in single and multi-axis
low-frequency horizontal linear and rotational vibration. In Journal of sound and vibration.
1984;97(4):677-677.
10. Irwin AW. Perception, comfort and performance criteria for human beings exposed to whole body pure
yaw vibration and vibration containing yaw and translational components. Journal of Sound and Vibration.
1981 Jun 22;76(4):481-97.
11. Melbourne WH, Cheung JC. Designing for serviceable accelerations in tall buildings. In Proceedings of
the 4th International Conference on Tall Buildings, Hong Kong and Shanghai 1988:148-155.
12. Isyumov N. Criteria for acceptable wind-induced motions of tall buildings. In International Conference
on Tall Buildings 1993 May 17 (pp. 401-411).
13. National Buildings Code of Canada. Structural Commentaries (Part 4). National Research Council of
Canada. Ottawa, Canada. 1995.
14. Melissa Burton BM. Wind-Induced Motion of Tall Buildings: Designing for Occupant Comfort. In
International Journal of High-Rise Buildings. 2015 Mar:4(1).
15. Architectural Institute of Japan Recommendations. Guidelines for the evaluation of habitability to
building vibration, AIJES-V001-2004. 2004. Tokyo, Japan.
16. International Standards Organization. Bases for design of structures: Serviceability of buildings and
walkways against vibrations, ISO10137:2007. Geneva, Switzerland. 2007.
17. Tschanz T. The Base Balance Measurement Technique And Applications To Dynamic Wind Loading To
Structures.
18. Davenport AG. Gust loading factors. Journal of the Structural Division. 1967 Jun;93(3):11-34.
19. Chopra AK. Dynamics of structures. New Jersey: Prentice Hall; 1995 Jan.
20. The 21st Century Center of Excellent Program. Wind Effects on Buildings and Urban Environment:

7660
 INTER-NOISE 2016

Lecture of damping in buildings. Tokyo Polytech University. (Cited May 2016). Available from
http://www.wind.arch.t-kougei.ac.jp/info_center/ITcontent/tamura/10.pdf
21. Zhou Y, Kareem A. Gust loading factor: new model. Journal of Structural Engineering. 2001
Feb;127(2):168-75.
22. Quan Y, Gu M. Across-wind equivalent static wind loads and responses of super-high-rise buildings.
Advances in Structural Engineering. 2012 Dec 1;15(12):2145-56.
23. Architectural Institute of Japan Recommendations. Recommendations for loads on buildings. 2004.
Tokyo, Japan.
24. Gu M, Quan Y. Across-wind loads of typical tall buildings. Journal of Wind Engineering and Industrial
Aerodynamics. 2004 Nov 30;92(13):1147-65.

7661