See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/236029946

Experimental verification of novel 3DOF model for grandstand crowd-structure

dynamic interaction

Conference Paper · January 2008

CITATIONS READS

11 217

2 authors:

Aleksandar Pavic Paul Reynolds

University of Exeter University of Exeter
165 PUBLICATIONS   2,482 CITATIONS    168 PUBLICATIONS   2,130 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

- Developing IMU-based wearable systems for medical-grade human movement monitoring in real-life environment (3D kinematics and kinetics) View project

Realistic approaches to vibration serviceability assessment of office floors. View project

All content following this page was uploaded by Paul Reynolds on 18 January 2014.

The user has requested enhancement of the downloaded file.

Experimental verification of novel 3DOF model for grandstand crowd-
structure dynamic interaction

Aleksandar Pavic and Paul Reynolds

Vibration Engineering Section

Department of Civil and Structural Engineering, The University of Sheffield
Sir Frederick Mappin Building, Mappin Street, Sheffield, S1 3JD, UK

ABSTRACT
The dynamic performance of sports stadia is becoming of concern due to their increasing slenderness combined
with the more onerous dynamic excitation being provided by crowds. One of the reasons for the more onerous
dynamic loading is the use of sports stadia for non-sporting events, such as concerts, where the presence of
musical beat and visual stimuli serve to synchronise the activities of a crowd. This paper presents a case study
with the aim to verify experimentally a novel 3DOF mathematical model which will feature in the new UK design
guidance on crowd dynamic loading of grandstands. The results obtained agree well with experimental
observations and hence indicate that the proposed methodology will be useful in practice.

1 INTRODUCTION
Problems with vibration serviceability in modern sports stadia are becoming more common. Three main reasons
for this trend are (Joint Working Group 2001):
1. sports stadia structures are becoming more slender with improvements in materials and construction
techniques and modern trends in architectural design,
2. crowds are in general becoming more lively than previously was the case, and
3. there is an increased use of stadia for pop/rock concert events, where crowd activities can become better
synchronised due to the presence of auditory and visual stimuli.
Therefore, there is a need to develop a predictive methodology which can predict reliably vibration responses of
structures occupied and dynamically excited by crowds due to jumping and/or bouncing. This methodology should
take into account the full effects of human-structure interaction i.e. the fact that the presence of people changes
the modal properties of the structure they occupy (natural frequencies and modal damping ratios, in particular)
(Ellis & Ji 1997; Sachse, Pavic, & Reynolds 2003) and also that the motion of the structure has the potential to
modify dynamic forces imparted by people due to jumping/bouncing (Yao et al. 2004; Yao et al. 2006). Moreover,
there is a considerable effect of crowd synchronisation, or lack of it, during jumping/bobbing (Ebrahimpour & Sack
1992). The change of modal properties, the dependency of the human-induced forcing function during
jumping/bouncing on the amount of structural motion and the crowd synchronisation are three key problems
which need to be addressed by modern design guidelines if they are to represent realistically what is actually
happening during crowd structure dynamic interaction.
However, these issues are quite complex and not particularly well researched and understood, so they are
therefore difficult to conceptualise and formulate mathematically in a coherent manner. Worldwide, no design
guideline dealing with crowd dynamic loading takes all three issues into account consistently and reliably. For
example, British standard BS6399 (British Standards Institution 1996) proposed a reduction factor of 2/3 to take
into account the effects of synchronisation during jumping, but for a large number of people present on
grandstands (often measured in thousands) this still yields extraordinarily high vibration responses calculated by
assuming that perfect resonance has been achieved. Another example is the Canadian Design Code (Canadian
Commission on Building and Fire Codes 2006) which takes into account the presence of people by suggesting an
increased damping value which does not depend on the dynamic properties of the structure analysed. This is a
good attempt to take into account the likely effects of the crowd’s presence on the structure, but lacks the
necessary refinement considering that the level of damping strongly depends on the properties of the empty
structure (Sachse, Pavic, & Reynolds 2004; Brownjohn 2001). Therefore, it can be concluded that explicit
treatment of all three key aspects of human-structure interaction has not been previously considered in design
guidance worldwide for crowd dynamic loading. Results of calculations have shown that, for practical grandstands
and dense crowd loading, it is necessary to take account of this interaction if the response near resonance is not
to be significantly overestimated. Brownjohn, Pavic, & Reynolds 2006
To address this important issue, Dougill et al. 2006 proposed recently a revolutionary new calculation model
which explicitly takes into account self generated dynamic action from a crowd. This model has been adopted for
use in the final draft recommendations of the Joint Working Group (Joint Working Group 2007). The aim of this
paper is to outline this new mathematical model and describe its verification using rare real-life vibration response
data from a full-scale grandstand measured during a lively concert event. In doing this it has been necessary to
take account of both active and passive crowd behaviour as observed from video records taken during the course
of the actual event.

2 3DOF MODEL DESCRIPTION

The proposed model is in fact a 3DOF system shown in Figure 1.

x3(t) x2(t)
m3 m2

k3 c3 k2 c2
x1(t) m1

k1 c1

Figure 1: 3DOF crowd-structure interaction model.

The 3DOF system comprises three SDOF systems representing:
1. A dominant mode of vibration of the empty structure (associated with modal mass m1 , stiffness k1 and
damping coefficient c1 ) having motion is described by DOF x1 ( t ) .

2. Aggregate effects of all passive people (associated with their aggregate modal mass m2 , stiffness k 2 and
damping coefficient c2 ) who are just present on the structure but do not participate in the dynamic excitation
and whose motion is described by DOF x2 ( t )
3. Aggregate effects of all active people (associated with their aggregate modal mass m3 , stiffness k 3 and
damping coefficient c3 ) who are both present on the structure and dynamically excite the structure and whose
motion is described by DOF x3 ( t ) .

The rationale for the development of the concept of internal forces within the SDOF system representing an
aggregate effects of all active human bodies is presented in a recent paper by Dougill, Wright, Parkhouse, &
Harrison published in 2006. The whole of the 3DOF system is energised by a pair of equal but opposite internal
harmonic forces P ( t ) acting on DOFs 3 and 1 (Figure 1) which magnitudes need to be chosen to match the
observed activity of the crowd. The values of the forcing function for human bouncing given by Dougill, Wright,
Parkhouse, & Harrison ( 2006) are based on the data published by Parkhouse & Ewins in 2006.
2.1 Equations of motion of 3DOF system
The motion of the 3DOF crowd-structure system shown in Figure 1 is described by a matrix equation of motion
given in Equation (1).

⎡ m1 0 0 ⎤ ⎧ x1 ⎫ ⎡c1 + c2 + c3 −c 2 −c3 ⎤ ⎧ x 1 ⎫ ⎡ k1 + k 2 + k 3 −k 2 − k 3 ⎤ ⎧ x 1 ⎫ ⎧ −P ( t ) ⎫

⎢ ⎥⎪ ⎪ ⎢ ⎥⎪ ⎪ ⎢ ⎥⎪ ⎪ ⎪ ⎪
⎢0 m2 0 ⎥ ⎨ x2 ⎬ + ⎢ −c2 c2 0 ⎥ ⎨ x 2 ⎬ + ⎢ −k 2 k2 0 ⎥ ⎨ x2 ⎬ = ⎨ 0 ⎬ (1)
⎢⎣ 0 0 ⎥ ⎪ ⎪
m3 ⎦ ⎩ x3 ⎭ ⎣ ⎢ −c3 0 ⎥ ⎪
c3 ⎦ ⎩ x3 ⎭ ⎣⎪ ⎢ −k 3 0 k 3 ⎥⎦ ⎩⎪ x 3 ⎭⎪ ⎪⎩+P ( t ) ⎪⎭

To solve Equation (1), firstly it is necessary to calculate the stated equation coefficients.
For calculating the relevant dynamic properties of the empty seating deck ( m1 , k1 and c1 ), modal analysis of the
structure is required. This is typically performed via finite element (FE) analysis yielding structural natural
frequencies fs and modal mass m1 from which modal stiffness can be obtained, as follows:

k1 = m1 ( 2πfs )
2
(2)

For the purpose of verifying the analytical model, once the mode shapes had been determined, it was merely
necessary to identify which mode was dominant in the observed behaviour of the stand in service.
The properties of the crowd SDOF mass-spring-damper elements depend on whether people are predominantly
standing or sitting. For predominantly seated passive crowd it is proposed that the natural frequency fp of the
SDOF system representing them is 5 Hz and that the damping ratio corresponding to this SDOF isolated is 40%.
The contributing mass of the passive people m2 can be obtained by scaling the physical mass of the present
people using mode shape amplitudes of the relevant mode of the empty structure at the location where the
passive people are. This is a considerable assumption but recent work at the University of Sheffield (Ibrahim
2006) demonstrated that it is a reasonably correct assumption. Therefore:
Np
m2 = ∑m
i =1
p,i ⋅ φs2,i , (3)

where N p is the total number of lumped physical masses mp,i of passive people distributed over the structure
and φs,i is the mode shape amplitude of the relevant mode of vibration of the empty structure at the location of
mass mp,i . Moreover,

( ) = m2 ( 2π ⋅ 5 )
2 2
k 2 = m2 2πfp (4)

c2 = 0.4 ⋅ 2 k2 m2 (5)

Similarly, for active people, mass m3 can be obtained by scaling physical mass of people by the relevant mode
shape of the empty structure but using dynamic properties for active people of 2.3 Hz (and the corresponding
SDOF damping ratio 25%) as determined by Dougill, Wright, Parkhouse, & Harrison 2006). Therefore, the
relevant coefficients are:
Na
m3 = ∑m
i =1
a,i ⋅ φs2,i , (6)

where Na is the total number of lumped physical masses ma,i of passive people distributed over the structure
and φs,i is the mode shape amplitude of the relevant mode of vibration of the empty structure at the location of
mass ma,i . Moreover,

k 3 = m3 ( 2πfa ) = m3 ( 2π ⋅ 2.3 )
2 2
(7)

c3 = 0.25 ⋅ 2 k3 m3 (8)

Finally, as previously mentioned, the harmonic forcing function P ( t ) relate to the type and intensity of activity in
the crowd and is given in terms of the so called generated load factors (GLFs) (Joint Working Group 2007), as
opposed to the usually used dynamic loading factors (DLFs). The concept of GLFs was introduced by Dougill,
Wright, Parkhouse, & Harrison in 2006 to describe the internal set of forces P ( t ) whereas DLFs are traditionally
used to describe the interface force between the human body and the structure. In the context of the analysis, the
SDOF body units for active and passive people, including the energising couple of forces P ( t ) , replace
completely all of the forces that have commonly been prescribed in design guides worldwide to represent crowd
dynamic loading. It should be noted that the use of the body units leads to a combined structure-crowd dynamic
system that is not subject to external dynamic loads. The motion of the system is caused solely by the forces
P ( t ) generated within the system itself.

2.2 Solution of equations of motion of 3DOF system

The 3DOF system which needs solving has non-proportional damping and to solve it and find response to the
given couple of harmonic forces P ( t ) , a SIMULINK model was developed as shown in Figure 2.

Gain 14 Gain 15
-K- -K-

Scop e

1 1 du /dt du/d t
s s
I nteg rator Inte grator1 Deriva tive Derivative 1 Scope6

-K- -K-
Ga in Gain1

-K- -K-
Gain 9 Ga in10

Gain 2 Ga in3
-K- -K-

-K- 1 1
s s
Sin e Wave Ga in6 I nteg rato r2 Inte gra tor3 Scope 1

-K- -K-
Ga in7 Gain 8

Gain4 Gain 5
-K- -K-

-K- 1 1
s s
Gain11
Int egra tor4 I nteg rator5 Sco pe5

-K- -K-
Gain 12 Gain13

Figure 2: SIMULINK model for solving the 3DOF system.

Figure 2 shows that the source of the excitation is a sine wave acting on DOFs 1 and 3 and the displacement
responses are calculated for all three DOFs whereas accelerations are also calculated for DOF 1. To solve the
linear non-proportionally damped 3DOF system, modal analysis can be performed yielding three pairs of complex
conjugate eigenvalues and the corresponding conjugate pairs of complex mode shapes.
2.3 3DOF model verification strategy
Having developed this analysis framework, an attempt was made to verify it by comparing simulation results with
vibration response measurements made on a real-life grandstand during a lively concert. Modal properties of the
empty structure ( m1 , k1 and c1 ) were obtained by a combined experimental and analytical approach. Natural
frequency fs and modal damping ratio of the relevant mode of vibration was obtained by ambient vibration survey
(AVS). Due to very short timescales to perform the tests just before the concert, the AVS also yielded amplitudes
of the relevant unscaled mode shape over only a very limited number of points. As it is not possible to determine
modal mass from an unscaled mode shape, a full 3D FE model was developed and manually tuned to match the
AVS-obtained natural frequencies and mode shapes as much as possible. This FE model was then used to obtain
modal mass m1 using a unity-scaled mode shape of the relevant mode of vibration. The distribution of the
passive and active crowd was estimated using video recording of the crowd behaviour during the concert. Finally,
based on the mode-shape of the tuned FE model and the estimated distribution of the active and passive crowd,
masses m2 and m3 were estimated from Equations (3) and (6), and used to calculate the remaining properties of
the two SDOF body units using Equations (4), (5), (7) and (8).
For the observed behaviour of the crowd an appropriate generalised load factor was selected. This, together with
the mode shapes of the empty structure and the assumption of the number and location of active people, yielded
the internal forces P ( t ) obtained by summing individual physical forcing functions Pi ( t ) scaled by the mode
shape of the relevant mode of the empty structure, as follows:
Na
P (t ) = ∑ P (t ) ⋅ φ
i =1
i s,i , (9)

Obviously, DOF 1 response obtained, either in terms of displacement x1 ( t ) or accelerations x1 ( t ) , is a modal
response which needs scaling back into physical space by multiplying it by the corresponding mode shape
amplitude at the location of interest. In the case of the antinode point for the relevant mode of vibration of the
empty structure, if unity scale mode shapes are used, then the scaling factor is obviously 1.0, so the modal
response x1 ( t ) or x1 ( t ) is the same as the physical response at the antinode.

3 STRUCTURAL DESCRIPTION
The full-scale stand structure and the corresponding modal and response data used for this verification exercise
have been described in detail as Structure A by Reynolds & Pavic in 2005. For completeness, key facts will be
repeated here.
The structure is a stand within a major stadium facility in the UK. The stand is of modern construction and
consists of regularly spaced steel frames between which are spanning precast concrete seating deck elements. In
one of the corners of the stadium, there is short section of an upper seating deck of approximately 18.9 m in
length that is cantilevered approximately 7 m over the lower tier. A typical cross-section through the seating deck
section is shown in Figure 3. This cantilevered section was known from previous experience to be more lively
than the rest of the areas of seating deck and therefore the in-service monitoring was concentrated in this area.
Figure 3: Cross section Figure 4: Test grid.
From Figure 3 the cantilevered upper tier section can clearly be seen to contain 8 rows of seating. It is also
noteworthy that there are some structural members that were installed on the structure after the original
construction of the stand, in an attempt to stiffen the seating deck section and make it less susceptible to dynamic
excitation. The modifications include the horizontal struts that are attached between the cantilever tip and the
column behind it, and an additional section attached to the column to resist the horizontal reaction from the strut.
The stadium was used to host a concert by a well known contemporary artist and the stadium management and
their structural engineers were interested in the performance of the stand during the event. The authors therefore
were invited to carry out some in-service vibration monitoring during the concert event. The results of exercises
like this are rarely available in the public domain.
Preparations for the in-service monitoring took place at very short notice as the authors were only informed of the
opportunity to carry out the monitoring a few days in advance of the concert event. As a result, the monitoring was
limited only to the cantilever of primary interest. With additional preparation time, it would have been
advantageous to install transducers more widely over the stand, for example at the top/back of the stand to pick
up global modes of vibration. The short length of seating deck to be monitored contained 6 frames between which
spanned the precast concrete seating deck elements. Four of these spans were of equal length and the fifth was
shorter. It was decided to position transducers vertically at the tip of the cantilever at five of the frame locations
(test points (TPs) 01 to 05) as illustrated in Figure 4. It was not possible to install transducers on the sixth frame
as this was out of reach of the test personnel. The transducers used were Endevco Model 7754-1000
piezoelectric accelerometers. These were attached to aluminium mounting brackets that were then bonded to the
front of the cantilever using adhesive. The front of the cantilever was constructed of a precast concrete parapet
unit as can be seen in Figure 3. To prevent the accelerometers falling onto occupants of the lower tier in case the
adhesive failed, their cables were tied to the horizontal struts underneath the cantilever. Data acquisition was
carried out using two independent data acquisition systems. The first was a PC-based vibration and video data
acquisition system that was originally designed as a remote monitoring system for stadia dynamics (Reynolds,
Pavic, & Ibrahim 2004).
Data acquisition was carried out for several hours, including periods when the tier was empty and when it was
occupied during the support and main acts of the concert. A typical time-history from one of the accelerometers is
shown in Figure 3. The data presented were acquired using the remote monitoring data acquisition system
sampling at a rate of 80 Hz with antialiasing filters set at 30 Hz. Additionally, the time history had been weighted
using the BS6841 Wb weighting curve (British Standards Institution 1987), which takes account of the variation of
human perception of vibration at different frequencies. In addition to the vibration monitoring, a professional video
crew were employed to take video footage of the crowd behaviour during the concert. In this way it was possible
to correlate to some extent the vibration response of the stadium with the crowd activity.
4 AVS OF EMPTY STAND
Ambient modal parameter estimation was performed on the data acquired prior to crowd occupation. In this way it
was possible to determine the modal properties of the dominant cantilever modes. The data were analysed using
the Frequency Domain Decomposition algorithm (Brincker, Zhang, & Andersen 2000) in the ARTeMIS software
and the results are presented in Figure 5.

Figure 5: AVS estimated modes of vibration of the empty stand.

The first mode at 2.70 Hz appeared to be a global fore-and-aft mode of the whole stand, due to its low frequency
and the relatively low peak in the measured auto-spectral density (ASD) functions. It was also characterised by a
shape in which there was little bending of the cantilever and all test points were moving in phase with each other
indicating likely global motion of the stand. The next five modes appeared to be local modes that engaged mainly
the cantilever in vibration. The shapes are typical for a strongly orthotropic structure, in that a number of modes
with relatively close natural frequencies were observed with an increasing number of half-sine shapes along the
length of the cantilever.

5 VIBRATION RESPONSE MONITORING DURING CONCERT

Figure 6 shows the time history from TP05 for the entire event. This was the test point at which the maximum
response was observed throughout the event and it is clearly an antinode for mode 2 (Figure 5) . It is clear that
there was a relatively low level of response caused by the crowd during the support acts. Visual observations of
the crowd activity showed that occupants were not particularly lively during this time. However, there is a clear
increase in response during the main act when the crowd was more active. There was a particularly high level of
response during the period from 23:14 to 23:19 as can be seen from Figure 6.
Figure 6: Vertical vibration response monitoring at TP5 during the concert.
Figure 7 zooms in on this period of highest response and also shows the 1 s RMS trend calculated for this time
history.

Figure 7: Time history at TP5 during period of highest Figure 8 ASD at TP5 during period of highest
response. response
It is notable from an examination of the video footage that there was not an appreciable increase in crowd activity
during this period of highest response. It is therefore reasonable to assume that the beat frequency of the
particular song that was being played coincided with a resonant frequency of the combined structure/crowd
system. This can be seen from Figure 8 in which an ASD of the period of highest response is presented. Two
main peaks are observable from this ASD. The first at 1.88 Hz corresponds with the beat frequency of the music,
which was confirmed from the recorded sound in the video footage. This, therefore, was also the fundamental
frequency of the crowd dynamic activity, which was a ‘bouncing’ activity in nature. The second much larger peak
at 3.75 Hz corresponded to the second harmonic of the crowd activity that appeared to coincide with the first
mode of vibration of the cantilevered tier occupied by humans resulting in a (near-)resonant response. This
frequency corresponds to Mode 2 in Figure 4, but with reduced natural frequency due to the presence of
spectators. Also clear from the ASD are other smaller peaks at integer multiples of 1.88 Hz, corresponding with
higher harmonics of the fundamental excitation frequency.

6 FE MODELLING
Figure 9 shows an FE model of the stand developed using ANSYS software.
Figure 9: Full 3D FE model based on available drawings and other design information.
Figure 10 shows the first global mode of vibration of the stand in which the structure oscillates to-and-from the
pitch and the tip of the cantilever moves up-and-down with very little bending along the cantilever. This mode
correlates very well with the experimentally estimated mode at 2.70 Hz (Figure 5).

Figure 10: First global mode of vibration (moving to-and-from the pitch) at 2.8 Hz.
Figure 11 shows first bending mode of the cantilever at 4.2 Hz which corresponds to the second experimental
mode at 4.34 Hz (Figure 5).
Figure 11: First (symmetric) vertical bending mode of the cantilever.
Finally, Figure 12 shows the second vertical bending mode of the cantilever at 5.9 Hz which is anti-symmetric.
This, clearly, correlates with the fourth experimental mode at 5.88 Hz (Figure 5).

Figure 12: Second (anti-symmetric) vertical bending mode of the cantilever.

By comparing experimentally and numerically obtained modal analysis results, it is clear that the FE mode at
4.2 Hz corresponds to the cantilever mode which was excited in resonance by the spectators. Using a unity-
scaled mode shape, the modal mass of this mode was estimated to be 82,811 kg.
7 3DOF CROWD-STRUCTURE MODEL IMPLEMENTATION
By examining the video recording corresponding to the highest levels of vibration response (Figure 13), it was
estimated that the ratio between active and passive people was 40:60 and that both groups were more or less
uniformly distributed on the stand which was sold out to its full capacity.

Figure 13: Video recording of the crowd behaviour during period of highest responses.
Approximately 1050 people were estimated to be on the stand and the mass of each spectator was assumed to
be 800 kg. Using an FE-calculated mode shape amplitudes for this mode of vibration and Equations (3) and (6),
parameters m2 and m3 of the 3DOF model were calculated.

From the video recording, it was evident that the active crowd were predominantly engaged in bobbing, i.e.
maintaining continuous contact with the structure, with very few people jumping. Accordingly, a value of 0.57 was
used for the generated load factor as determined by Dougill, Wright, Parkhouse, & Harrison 2006) for groups of
50 or more people bobbing. The factor takes account the effects of imperfect synchronisation, as determined by
Parkhouse & Ewins in 2006 in synthesising DLFs for group loading from an extensive series of bobbing test
results with different individuals. Using the FE calculated mode shapes, the estimated total number of people of
1050 and the average weight of 800 N for each spectator, and assuming that the active people were uniformly
distributed over the stand, an internal harmonic force P ( t ) was calculated. This was the last information needed
for solving equations of motion (1).
7.1 Modal analysis
A modal analysis of the non-proportionally damping 3DOF system was performed using MATLAB yielding the
following three natural frequencies: 2.2 Hz, 3.8 Hz and 5.3 Hz. By inspecting the corresponding complex
conjugate mode shapes it was obvious that the second frequency corresponded to the mode of vibration
engaging the structural degree of freedom DOF 1. This realisation was quite encouraging considering that the
dominant frequency of the strong structural response observed experimentally (Figure 8) was 3.75 Hz as
previously mentioned. Therefore, it seems that the 3DOF model was capable of estimating the relevant natural
frequency of the crowd-structure system accurately. It is crucial that the natural frequency of the structural mode
of vibration in the 3DOF system is close to the dominant response frequency of the real structure which is likely to
correspond to the resonance by the second harmonic of jumping/bouncing performed rhythmically following the
song beat which was approximately half of this natural frequency, as previously mentioned. Moreover the modal
damping ratio of this structural mode went up from 3.3% for the empty full-scale structure as determined via the
AVS (Figure 5) to 11% as calculated using a 3DOF system for the mode at 3.8 Hz.
7.2 Response analysis
The results of the steady state resonant excitation by harmonic force P ( t ) set to have frequency at 3.8 Hz were
equally encouraging. Figure 14 shows a comparison of the measured and 3DOF simulated acceleration
responses. Limits of vertical axes on both graphs are set to plus or minus 1 m/s2 and they are shown to the same
scale in the figure.

0.6 m/s2 RMS

Figure 14: Comparison of the measured (left) and simulated (right) results at TP5 which is the antinode of the
cantilever bending mode at 3.8 Hz.
It can be seen that the simulated steady-state resonant peak acceleration of 0.85 m/s2 represent approximately
an upper bound of the measured responses which were occasionally reaching this level. In terms of RMS the
simulated steady state sinusoidal response of 0.6 m/s2 RMS also compare quite well with the calculated 1 s RMS
trend shown in Figure 14.
This indicates that the proposed methodology is quite viable. Further parametric study of the effects of the various
uncertain assumptions made indicated that the calculated response is not overly sensitive. For example, an
assumption that the ratio active: passive people is 45:55 instead of 40:60 increases the response to slightly over
0.9 m/s2 which is still quite acceptable.

8 CONCLUSIONS
This paper presents a case study with the aim to verify experimentally the novel 3DOF mathematical model which
will be featuring in the new UK design guidance on crowd dynamic loading of grandstands. A rare AVS and
vibration response data recorded during a concert were used in this verification exercise.
Firstly, a modal analysis of the 3DOF crowd-structure system developed using real-life structural and crowd
occupation data in conjunction with the new methodology yielded reduced natural frequency and considerably
enhanced damping of the relevant mode of vibration (when compared with the corresponding mode of the empty
structure) which was excited near resonance. These modal properties were fully in line with what was observed
on the full-scale structure.
Secondly, vibration response analysis was performed using the proposed methodology yielding steady state
acceleration levels which were an upper limit of what was observed during a real-life concert.
Thus, the results obtained agree well with experimental observations and hence indicate that the proposed novel
methodology is viable for prediction of crowd-induced dynamic responses of grandstand structures.

9 ACKNOWLEDGEMENTS
The authors would like to acknowledge the financial assistance provided by the UK Engineering and Physical
Sciences Research Council (EPSRC) through a research grant entitled “Dynamic Behaviour of Stadia Under
Human Occupation and Excitation” (Ref: GR/S73761/01). Also, thanks go to the owner of the monitored stadium,
for providing access to the structure during the described high-profile event.
Special thanks are due to Dr John Dougill, Chairman of the UK Joint Working Group which is responsible for new
UK design guidelines for crowd dynamic loading of stadia that are now being finalised.
Finally, the authors would like to acknowledge the contribution of Mr Nicola Perandin, a former MSc student at the
University of Sheffield who originally developed the stadium FE model utilised in this paper.
10 REFERENCES

Brincker, R., Zhang, L., & Andersen, P. (2000). "Output only modal analysis by frequency domain decomposition",
in 25th International Seminar on Modal Analysis (ISMA), Leuven, Belgium, pp. 717-723.

British Standards Institution (1987). BS6841: 1987 Measurement and evaluation of human exposure to whole-
body mechanical vibration and repeated shock British Standards Institution, Milton Keynes.

British Standards Institution (1996). BS 6399: 1996 Loading for buildings British Standard Institution, London.

Brownjohn, J. M. W. (2001). "Energy dissipation from vibrating floor slabs due to human-structure interaction",
Shock and vibration, vol. 8, no. 6, pp. 315-323.

Brownjohn, J. M. W., Pavic, A., & Reynolds, P. (2006). "Assessing and Managing Vibration Serviceability of
Footbridges", in First International Conference on Advances in Bridge Engineering, Brunel University, London.

Canadian Commission on Building and Fire Codes (2006). User's Guide - NBC 2005: Structural Commentaries
(Part 4 of Division B) National Research Council Canada; Institute for Research in Construction, Ottawa.

Dougill, J. W., Wright, J. R., Parkhouse, J. G., & Harrison, R. E. (2006). "Human structure interaction during
rhythmic bobbing", The Structural Engineer, vol. November 2006, no. 22, pp. 32-39.

Ebrahimpour, A. & Sack, R. L. (1992). "Design live loads for coherent crowd harmonic movements", Journal of
Structural Engineering, vol. 118, no. 4, pp. 1121-1136.

Ellis, B. R. & Ji, T. (1997). "Human-structure interaction in vertical vibrations", Proceedings of the Institution of
Civil Engineers, Structures and Buildings, vol. 122, no. 1, pp. 1-9.

Ibrahim, Z. (2006). The Effects Of Crowds On Dynamic Characteristics Of Stadia Structures, PhD, The University
of Sheffield.

Joint Working Group (2001). Dynamic performance requirements for permanent grandstands subjected to crowd
action: Interim guidance on assessment and design Institution of Structural Engineers, London.

Joint Working Group (2007). Dynamic performance requirements for permanent grandstands: recommendations
for management design and assessment, Final draft for approval Institution of Structural Engineers, London.

Parkhouse, J. G. & Ewins, D. J. (2006). "Crowd-induced rhythmic loading", Structures & Buildings, vol. 159, no.
SB5, pp. 247-259.

Reynolds, P. & Pavic, A. (2005). "The dynamic performance of sports stadia under crowd dynamic loading at
concert events", in Sixth European Conference on Dynamics EURODYN 2005, Millpress, Rotterdam, pp. 473-
479.

Reynolds, P., Pavic, A., & Ibrahim, Z. (2004). "A remote monitoring system for stadia dynamics", Proceedings of
ICE: Structures and Building, vol. 157, no. 6, pp. 385-393.

Sachse, R., Pavic, A., & Reynolds, P. (2003). "Human-Structure Dynamic Interaction in Civil Engineering
Dynamics: A Literature Review", The Shock and Vibration Digest, vol. 35, no. 1, pp. 3-18.

Sachse, R., Pavic, A., & Reynolds, P. (2004). "Parametric Study of Modal Properties of Damped Two-Degree-of-
Freedom Crowd-Structure Dynamic Systems", Journal of Sound and Vibration, vol. 274, no. 3-5, pp. 461-480.
Yao, S., Wright, J. R., Pavic, A., & Reynolds, P. (2004). "Experimental study of human-induced dynamic forces
due to bouncing on a perceptibly moving structure", Canadian Journal of Civil Engineering, vol. 31, no. 6, pp.
1109-1118.

Yao, S., Wright, J. R., Pavic, A., & Reynolds, P. (2006). "Experimental study of human-induced dynamic forces
due to jumping on a perceptibly moving structure", Journal of Sound and Vibration, vol. 296, no. 1-2, pp. 150-165.

View publication stats