Vous êtes sur la page 1sur 9

1196

ARTICLE
Modeling of crowd load in vertical direction using biodynamic
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

model for pedestrians crossing footbridges


Felipe Tavares da Silva, Halane Maria Braga Fernandes Brito, and Roberto Leal Pimentel

Abstract: In the analysis of vibration of footbridges in vertical direction, for crowd situations, there is evidence in the literature
that the whole effect of pedestrian action is not well modeled when applying current force-only models to represent such an
action. In these models, the action of each pedestrian is represented by a pulsating force applied on the structure. In this paper,
a crowd load model is proposed for sparse and dense crowds (with densities up to around 1.0 pedestrian/m2) in which biodynamic
models are added to represent the whole action of pedestrians. The focus of the investigation is on vibration effects in vertical
direction. Comparisons with measurements on a prototype footbridge were carried out and made it possible to identify
differences in the structural response when applying force-only and force-biodynamic models to represent the pedestrian action.
The latter (proposed) model resulted in a better agreement with the measurements.

Key words: pedestrian bridges, crowd load, vibration, biodynamic model.

Résumé : La littérature montre que l’effet complet de l’action des piétons dans l’analyse des vibrations verticales des passerelles
à piétons n’est pas bien modélisé lorsque l’on applique les modèles actuels uniquement de force pour représenter une telle
action. Dans ces modèles, l’action de chaque piéton est représentée par une force pulsatoire appliquée à la structure. Le présent
article propose un modèle de chargement de foule pour des foules éparses et denses (avec des densités pouvant atteindre environ
1,0 piéton/m2); des modèles biodynamiques ont été ajoutés pour représenter toute l’action des piétons. L’enquête L’étude porte
For personal use only.

principalement sur les effets des vibrations verticales. Des comparaisons à des mesures prises sur un prototype de passerelle à
piétons ont permis d’identifier les différences dans la réponse structurale lors de l’application des modèles de force seulement
et de force-biodynamique afin de représenter l’action des piétons. Ce dernier modèle (proposé) a montré une corrélation
supérieure aux mesures. [Traduit par la Rédaction]

Mots-clés : passerelles à piétons, charge d’une foule, vibration, modèle biodynamique.

Introduction may lead to responses that differ from the actual behavior of the
structure.
Vibration of footbridges due to human-induced loads has been
Indeed, being a structure designed for the conveyance of pedes-
appointed as a design problem to be tackled. A compilation of case
trians, there is an interaction between a footbridge and the pedes-
reports of excessive vibration in footbridges (Pimentel et al. 2001)
trians crossing it since the latter vibrate together with the former.
indicated occurrence of vibration problems in vertical as well as
Živanović et al. (2010) compared the responses obtained from sim-
in lateral direction. As for the vertical direction, the analysis of
ulations employing several force models available in the litera-
vibration performance have evolved from adopting load models
ture for modeling multiperson traffic in footbridges in vertical
that employed a single pedestrian crossing the structure (Blanchard direction to experimental data collected on two structures. In one
et al. 1977; Rainer et al. 1988; Živanović et al. 2007; Tubino and of the test footbridges, there was an overestimation of the re-
Piccardo 2008; Pimentel and Fernandes 2009) to models in which sponses from all simulations while compared to experimental
a stream of pedestrians is taken into account (Barker and Mackenzie data. Živanović et al. (2010) considered as a possible explanation
2008; Charles 2008; Butz et al. 2008; Piccardo and Tubino 2009). that this could be a consequence of human-structure interaction
Milestone codes of practice in terms of vibration serviceability of not accounted for in the simulations. They explored the strategy
footbridges, such as the BS 5400 (1978) and Ontario code (1983), of an arbitrary increase of damping of the system to account for
were based on the action of a single pedestrian whereas recent such an effect and pointed out a strong need for further research
guidelines and codes, such as the one from the Service d'Études on quantification of human-structure interaction.
Techniques des Routes et Autoroutes (SÉTRA) (2006) and the UK There have also been reports of increases on damping and (or)
National Annex to Eurocode 1 (UK NA to EN1) (2008), consider the changes on the natural frequencies of structures when they are
presence of several pedestrians simultaneously on the structure. occupied by standing people (Sachse et al. 2004; Butz et al. 2008) or
In all aforementioned codes of practice, the action of pedestri- else due to a stream of pedestrians walking along (Barker and
ans is considered as forces applied to the structure. However, Mackenzie 2008). References can be found in which spring-mass-
when the action of a stream of pedestrians is taken into account, damper (SMD) models were employed to represent the dynamic
by disregarding the dynamics of the pedestrians, this procedure vertical behavior of a whole crowd standing (Sachse et al. 2004)

Received 16 December 2011. Accepted 14 July 2013.


F.T. da Silva and H.M.B.F. Brito. Postgraduate Research Programme in Mechanical Engineering, Federal University of Paraíba, Campus Universitário,
sn 58051-900 João Pessoa-PB, Brazil.
R.L. Pimentel. Department of Civil and Environmental Engineering, Federal University of Paraíba, Campus Universitário, sn 58051-900 João Pessoa-PB,
Brazil.
Corresponding author: Roberto Leal Pimentel (e-mail: r.pimentel@uol.com.br).

Can. J. Civ. Eng. 40: 1196–1204 (2013) dx.doi.org/10.1139/cjce-2011-0587 Published at www.nrcresearchpress.com/cjce on 16 July 2013.
da Silva et al. 1197

or the human vertical dynamics in situations such as standing Fig. 1. Single degree of freedom models for walking.
(ISO 5982 1981; Matsumoto and Griffin 2003), walking (Miyamori
et al. 2001; Kim et al. 2008; Caprani et al. 2011) or running (Nigg
and Liu 1999). Different values for the biodynamic parameters of
the models were reported according to the type of locomotion.
The works related to biodynamic models for walking are worth
discussing in more details.
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

Miyamori et al. (2001) proposed a three-degree of freedom


(3-DOF) biodynamic model to represent a walking pedestrian in
the vertical direction. Some parameters of their model were ob-
tained through the response of a footbridge crossed by a single
pedestrian. The application of the model to represent crowd ac-
tion as well as the differences between force-only and biodynamic
models for such action were not discussed in their paper.
to measurements points out in this direction. This means that,
Kim et al. (2008) proposed a 2-DOF model aiming to represent
from a design point of view, consideration of the interaction leads
the dynamics of the human body in the vertical direction while
to a more realistic model of the phenomenon. However, the inter-
walking. However, the parameters of their biodynamic model
action is accounted for by modeling the pedestrians and structure
were obtained from the ISO 5982 code (1981), which is applicable
altogether. It will be shown that this will imply in better agree-
to standing people. In addition, results obtained by applying their
ment with experimental results.
model for crowds crossing footbridges were not checked against
experimental data. Biodynamic model for a walking pedestrian
Caprani et al. (2011) proposed a SMD model to represent each
pedestrian crossing the structure in addition to a force model for The SDOF model employed to represent a walking pedestrian in
the same pedestrian. In the absence of own experimental data to the vertical direction was proposed by Silva and Pimentel (2011).
obtain the parameters of the SMD model, they adopted values for Herein, more details are provided and the features of the model
the spring stiffness and damping coefficient taken from other related to the development of the crowd load model are high-
studies and, as for the mass, the full mass of the pedestrian was lighted.
considered in the SMD model. Simulations were carried out to In essence, the whole body of a walking person was conceived as
evaluate the repercussion of changes in such parameters, by com- a SDOF model (Fig. 1). In this figure, the left side represents a
For personal use only.

paring the response of a beam-like simply supported footbridge model for walking on rigid surfaces whereas on the right side
crossed by a single pedestrian, the latter being represented as a there is a model for walking on vibrating surfaces. While walking,
SMD model and also as a force-only model. The responses away the centre of gravity of the body presents an up and down oscilla-
from resonance were similar, the difference being more remark- tory motion (Inman et al. 1994). An oscillating force resulting from
able close to resonance. In this case, the response using the SMD this motion is applied to the ground. Obtaining such a force while
model was much reduced, however depending on the values ad- walking on rigid surfaces has been comprehensively researched
opted for the parameters of the model. (Rainer and Pernica 1986; Pernica 1990; Kerr and Bishop 2001;
Simulations were also carried out by Caprani et al. (2011) to Brownjohn et al. 2004a; Sahnaci and Kasperski 2005). The model
evaluate the response for crowd action. To do that, they adopted a consists of a modal mass m, a spring of elastic constant k and also
substitute force model to represent the SMD action of each a viscous damper the constant of which is c. The dynamic param-
pedestrian, by multiplying the expression of the force due to each eters m, c, and k of the model do not have a direct relationship
pedestrian (from a force-only model) by a coefficient. This coeffi- with parts of the human body. Instead, such values were sought in
cient, in turn, was obtained from the ratio between the responses such a way that measured accelerations at waist level could be
of the beam-like simply supported footbridge when crossed by a reproduced by the model. As the contact between the pedestrian
single pedestrian represented as a SMD model and as a force-only and surface is always maintained, by isolating the pedestrian, the
model. However, since the action of a SMD model would depend equation that represents the motion of the centre of gravity of the
on the level of vibration of the structure being investigated, pedestrian is
use of coefficients obtained for a specific condition would lack
generality. (1) müt ⫹ cu̇t ⫹ kut ⫽ PGnd
Aiming to overcome the limitation of the force-only model to
represent the pedestrian action, particularly in crowd situations, In eq. (1), ut is the displacement of the degree of freedom with
and also to provide further insight into pedestrian–structure in- respect to a fixed reference point, a dot on top of it indicating a
teraction, a crowd load model is proposed in this paper, making derivative with respect to time, and PGnd is the ground reaction
use of a biodynamic model to represent the action of each pedes- force while walking on a rigid surface. Such a force also varies
trian. with time.
In a previous paper, Silva and Pimentel (2011) proposed a single On the other hand, while walking on a vibrating surface (e.g.,
degree of freedom (SDOF) biodynamic model to represent the while crossing a footbridge), by assuming that the level of vibra-
dynamics of a walking pedestrian in vertical direction. This model tion is such that the biodynamic parameters of the model re-
is employed in this paper to construct the crowd load model. The mained reasonably unchanged, the equation to represent the
adequacy of this biodynamic model and the procedure in which motion of the pedestrian is
the load was applied was checked by comparing the numerical
results obtained using this model and also using a force-only (2) 
mü1t ⫹ c(u̇1t ⫺ u̇g) ⫹ k(u1t ⫺ ug) ⫽ PGnd
model, with measurements on a prototype footbridge, for pedes-
trian densities up to 0.9 pedestrians/m2.
It is expected that, by considering the interaction between pe- In eq. (2), ug is the displacement of the ground with respect to the
destrians and structure, a reduction of the response is going to same fixed reference in which u1t is evaluated. The latter is the
occur, due to the fact the human body is a highly damped system. displacement of the degree of freedom and is represented here by
The strategy adopted by Živanović et al. (2010) of increasing the a symbol different from that employed in eq. (1), to emphasize
damping of the system to match results from crowd force models that it is not interchangeable between the equations. It should be

Published by NRC Research Press


1198 Can. J. Civ. Eng. Vol. 40, 2013

Table 1. Mean and standard deviation (SD) of pedestrian data and Fig. 2. Correlations of the biodynamic parameters: modal mass.
single degree of freedom parameters.
fp (Hz) M (kg) m (kg) c (N·s/m) k (N/m)
Mean 1.87 63.82 45.82 867.06 16684.66
SD 0.14 9.88 3.70 66.41 1765.29
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

noted that the ground reaction force in the right hand side of the
equation may change when compared to the force while walking
on a rigid surface, due to potential changes in walking pattern.
A simplifying approach is to consider each pedestrian repre-
sented as a biodynamic model attached to the structure (that is, as
Fig. 3. Correlations of the biodynamic parameters: modal damping.
a SDOF moving oscillator vibrating together with the structure)
and simultaneously apply on the structure the force PGnd obtained
from walking on rigid surfaces. This is the same of considering the
right hand sides of eqs. (1) and (2) as being equal. The validity of
this simplification is checked by comparing results from simula-
tions and measurements on a vibrating structure.
The first step, however, is obtaining the biodynamic parameters
of the model. They were obtained by solving a system of three
nonlinear equations (eq. (3)), which is the expression of the acce-
lerance frequency response function of a SDOF model described
by eq. (1). Equation (3) represents the stationary response and is
consistent with the condition of steady walking adopted in the
tests to obtain the biodynamic parameters.

␻i2PGnd0(␻i) Fig. 4. Correlations of the biodynamic parameters: modal stiffness.


(3) A(␻i) ⫽
ⱍk ⫺ ⱍ
For personal use only.

␻i2m ⫹ j␻ic

In eq. (3), the parameters m, c, and k are the unknowns, j is the


square root of (−1), and PGnd0(␻i) and A(␻i) are the input values,
being respectively the amplitudes of the harmonic components of
the ground reaction force and of the measured acceleration at
waist level for the frequency ␻i, with i varying from 1 to 3. The
force was represented by means of a Fourier series and, for its
main first three harmonic components, expressions for the dy-
namic load factors proposed by Kerr and Bishop (2001) were ad-
opted, as discussed later on (eqs. (6)). It should be noted that only
the first three harmonic components of force and acceleration
were employed in the formulation since these were the most
prominent components in the expression of the force and mea- tions (see the curves in Figs. 2 to 4) and enabling the parameters to
sured accelerations at waist level, respectively. be obtained as a function of the total body mass M and pacing rate
Twenty test subjects took part in the experimental campaign, fp. These expressions are given by
being eleven men and nine women. Each test subject walked at
free will along a rigid surface, with an accelerometer attached at m ⫽ 97.082 ⫹ 0.275M ⫺ 37.518 fp (kg)
waist level. The mean pacing rate fp observed was 1.87 Hz, which is (4) c ⫽ 29.041 m0.883 (Ns/m)
in agreement with results for free walking presented in the liter-
ature (Pachi and Ji 2005). Mean and standard deviations (SD) of the k ⫽ 30351.744 ⫺ 50.261 c ⫹ 0.035 c2 (N/m)
biodynamic parameters obtained are presented in Table 1, to-
gether with related data. It can be noticed that the modal mass m In Figs. 2 to 4, the parameters p-value and R2 are statistical
is a fraction of the total body mass M, which means that only a metrics to evaluate the effectiveness of the fitting of data by the
fraction of the body mass would be considered to provide the regression function. The p-value is a measure of the similarity
inertia forces generated by a walking pedestrian when using such between the regression function and the data, through a hypoth-
a model. It is also worth mentioning that the mass, damping, and esis test. The p-value is related to the significance of the null
stiffness values obtained differed from the values for SDOF biody- hypothesis, that is, if the p-value has low values, typically lower
namic models proposed in the literature for standing people than 0.05 (Triola 2008), the null hypothesis can be rejected and it
(Matsumoto and Griffin 2003). can be concluded that the data are well fitted by the function. On
A way to model the walking pedestrians would be to randomly the other hand, R2 is a measure of the errors of the fitting, being
assign independent values for the parameters m, c, and k of each related to the proportion of the variation of the data points that can
pedestrian, for example, by picking values in a range of two SDs be explained (predicted) by the fitting; R2 values close to 1 indicate
around the mean values presented in Table 1. However, it was a function that better fits the data points. As can be seen from the
observed that a correlation existed among the parameters of the values obtained in Figs. 2 to 4, the strongest fit obtained was for
model. The strongest correlations are shown in Figs. 2 to 4. Ex- the relation between modal damping and modal mass.
pressions were then proposed (Silva and Pimentel 2011) for the Regarding Figs. 3 and 4, it can be seen that the R2 slightly fa-
biodynamic parameters so as to take into account such correla- vours the selection of the nonlinear functions for both cases

Published by NRC Research Press


da Silva et al. 1199

(eq. (4)), whereas the p-value slightly favours the linear relation in Fig. 5. Test structure.
Fig. 4. However, differences among the values are small and, in
general, the nonlinear relations resulted in better fit or else it is
not conclusive to replace them by linear fits. If anything, expres-
sions for the linear fits are presented below as modeling alterna-
tives:
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

c ⫽ 107.455 ⫹ 16.208 m (Ns/m)


(5)
k ⫽ 5758.441 ⫹ 11.103 c (N/m)

Values for M (in kg) and fp (in Hz) were then randomly assigned
according to the respective distribution each of these parameters
followed (e.g., see Table 1) and, from these, values for the biody-
namic parameters of each pedestrian were obtained by applying
eq. (4).

Crowd model
The proposed crowd model is based on non-syncronized move-
ment of pedestrians, as observed in the experimental campaign
(see the description of the tests). Although synchronization has
been observed in some footbridges prone to vibrate excessively in
the lateral direction (Fujino et al. 1993; Dallard et al. 2001; Danbon
and Grillaud 2005), studies to verify whether this phenomenon
occurs for footbridges prone to vibrate excessively in the vertical
direction are scarce. Indeed, the few experimental results avail- conducted to obtain the mode shapes of the structure in the ver-
able from tests in long span footbridges evidenced no synchroni- tical direction (Brasiliano et al. 2007).
zation in the vertical direction, at least for the vibration level A video camera was placed close to one of the footbridge
observed in the reported structures (Brownjohn et al. 2004b; supports and aligned in relation to its longitudinal axis. It was
For personal use only.

Charles and Bui 2005; Macdonald 2008). then possible to identify some of the characteristics of the pedes-
After conceiving the biodynamic model for a walking pedes- trian flows. The pedestrian densities investigated were 0.3, 0.7,
trian, there comes the task of coupling it with the model of the and 0.9 pedestrians/m2, in a condition of steady flow. With
structure. The strategy adopted was to apply the force as obtained regard to the number of pedestrians simultaneously on the
from walking in a rigid surface and add the biodynamic models to footbridge at a time for each density, they were respectively 5, 11,
the model of the structure, as previously discussed. By vibrating and 15 pedestrians. A total of 12 test subjects took part in the test
together with the structure, the biodynamic models will account for the density of 0.3 pedestrians/m2, 31 test subjects for the den-
for the changes in the dynamics of the system. sity of 0.7 pedestrians/m2, and 48 test subjects for the density of
Two approaches were adopted to add the biodynamic models. 0.9 pedestrians/m2. To assure a continuous stream of pedestrians
In the first one, called here ‘force and moving biodynamic mod- crossing the structure, the participants walked along a closed
els’, each biodynamic model moves along the structure together path, as evidenced in Fig. 5.
with the respective applied force. In this case, the analysis is non- All pedestrians walked at free will. From the processing of im-
linear because the elements of the vibrating system (in this case, ages, which involved observing the variation of phase angles be-
the biodynamic models) are changing positions during the analy- tween lateral movements of pairs of pedestrians (Araújo et al.
sis. The second approach consists of moving only the forces and 2009), no signs of synchronization among pedestrians were ob-
evenly distribute the fixed biodynamic models along the struc- served for the densities investigated. By processing the video im-
ture, according to a given pedestrian density. This second ap- ages, it was also possible to obtain the mean and SD of the pacing
proach is more easily implemented in terms of adapting existing rates of the pedestrians for each density. The mean pacing rate
reduced from 1.9 Hz for the density of 0.3 pedestrians/m2 to
algorithms that deal with moving force problems so as to include
1.76 Hz for the density of 0.9 pedestrians/m2. More details about
fixed biodynamic models. This approach is called here ‘force and
the processing of images and evaluation of synchronization can
fixed biodynamic models’ and would also benefit the computa-
be found elsewhere (Araújo et al. 2009).
tional analysis since the matrices of the system, consisting of
The pedestrians involved in these tests were mostly university
structure plus biodynamic models, would not change over time
students wearing rubber sole shoes, in conformity with the test
and the analysis would remain linear. Both approaches were
campaign in which the parameters of the biodynamic model
checked against experimental results.
were obtained. The acceleration time response of the structure
was measured at midspan by an Endevco accelerometer model
Test structure and measurements
7754-A, with a sensitivity of 1 V/g. For each density, ten signals of
The prototype footbridge available for this study (Fig. 5) had a approximately 40 s each were acquired. Root mean square (RMS)
natural frequency of approximately 4 Hz, being within the range acceleration was calculated for each acquired signal and the
of frequencies of the second harmonic of the pedestrian load for a Shapiro–Wilk test (Shapiro and Wilk 1965) was applied to confirm
condition of fast walk. This footbridge had a deck consisting of a that the ten samples of RMS accelerations followed a normal dis-
slab 10 cm thick and 1.80 m wide supported by two beams, with tribution. In sequence and adopting the t-distribution, it was ver-
a theoretical span of 11.30 m (Fig. 6). ified that the number of samples was such that a margin of error
Through impacts at one quarter of the span and at midspan and of 10% for the mean RMS acceleration was attained.
measuring the response with an accelerometer, the first vertical The signals were acquired by the spectrum analyser Data Phys-
vibration mode was identified, having a natural frequency of 4.27 Hz. ics SignalCalc and each signal had a total of 4096 points. A typical
The respective damping ratio was obtained from free decay tests, recorded signal is shown in Fig. 7. The first 10 s of each signal were
presenting a mean value of 1%. Full-scale modal tests were also discarded so as to consider only the part of the signals in which

Published by NRC Research Press


1200 Can. J. Civ. Eng. Vol. 40, 2013

Fig. 6. Footbridge plan (dimensions in centimetres).


Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

Fig. 7. Typical time history signal at mid-span of the footbridge. Table 2. Results of root mean square (RMS) accelerations at midspan.
Test aRMS03 (m/s2) aRMS07 (m/s2) aRMS09 (m/s2)
1 0.1347 0.0755 0.0931
2 0.1323 0.0744 0.1010
3 0.0921 0.0855 0.1212
4 0.1475 0.0886 0.0839
5 0.1008 0.0804 0.1000
6 0.1245 0.0989 0.0953
7 0.1309 0.0850 0.0954
8 0.0939 0.0924 0.0883
9 0.1098 0.0998 0.0965
10 0.0979 0.0731 0.1078
Mean 0.1164 0.0854 0.0983
the flow was steady. The RMS accelerations obtained from each SD 0.0199 0.0097 0.0104
measured signal are presented in Table 2 for each density, to-
gether with the respective mean and SDs.
It can be observed from Table 2 that the mean RMS acceleration With regard to the models representing the pedestrian action,
did not increase steadily with the increasing number of people on in the first model, the force applied by the crowd was considered
For personal use only.

the footbridge at a time, as it would be expected when adopting as moving harmonic point loads traveling across the structure at
crowd load models in which only forces are considered to repre- a speed proportional to the respective pedestrian pacing rate. The
sent the pedestrian action (SÉTRA 2006, UK NA to EN1 2008). expressions defining the force applied by each pedestrian and the
respective speed are shown in eqs. (6)–(8).
Numerical models: structure and crowd
To investigate the effects of force-only and force-biodynamic
models to represent the action of pedestrians in crowd conditions,
(6) F(t) ⫽ G 兺 ␣ sin (2␲if t ⫺ ␾ )
i
i p i

a finite element (FE) model was developed for the test footbridge,
using beam elements. ␣1(fp) ⫽ ⫺0.27 fp3 ⫹ 1.32 fp2 ⫺ 1.76 fp ⫹ 0.76
This FE model was implemented using the computer program (7) ␣2 ⫽ 0.07
ANSYS (2010). A limitation of the finite element analysis software
␣3 ⫽ 0.06
implied that each biodynamic model should be added only on
nodes of the structure. Therefore, for the case of the force and
(8) v ⫽ 0.9 fp
moving biodynamic models, each biodynamic model moved from
node to node of the FE model, in time with the movement of the
respective applied force. As for the force, it moved continuously In eqs. (6)–(8), F(t) is the harmonic point force that represents the
and, for each position it stays within a beam element, equivalent fluctuating part of the load applied by a pedestrian on the surface,
nodal forces were determined. This can be visualized in the detail G is the pedestrian weight, ␣i is the dynamic load factor (DLF) of
of Fig. 8 for a single pedestrian, in which PA and PB indicates the ith harmonic, t is time, ␾i is the phase angle between the ith
equivalent nodal forces. harmonic and the first, and v is the pedestrian speed. The expres-
With regard to the modeling of the force and fixed biodynamic sion adopted for the latter was that proposed in the former UK
models, the force again moves continuously along the span while code BS 5400 (1978), which was in agreement with what was ob-
the biodynamic models remain stationary at nodes of the FE model, served during the tests.
being evenly distributed along the span. The FE model should With regard to the DLFs, the ones adopted were proposed by
have a number of nodes that makes it possible to place the biody- Kerr and Bishop (2001). The expression for the first DLF (␣1) was
namic models in distinct nodes of the FE model even in the case of taken as the lower bound among three expressions proposed by
the most crowded analysis (pedestrian density of 0.9 ped/m2), in those authors, the reason for this choice being related to the
which 15 people at a time were at the footbridge. Therefore, a total characteristics of the group of test subjects, all of them using
of 16 beam elements were adopted, resulting in an element size of rubber sole shoes and walking mostly in a relaxed way. As for the
0.7063 m. other two DLFs, a great dispersion on the values was pointed out
The flexural stiffness EI and elastic constant to rotation KSUP at by Kerr and Bishop (2001). The values adopted here are somewhat
the supports were obtained from correlation between the FE model smaller than other proposals in the literature, so as to follow the
and the test results, having values of 57 MNm2 and 4300 kNm/rad, line of thought adopted in the choice of the expression for the
respectively. The mass per unit length mL of the footbridge, in first DLF. As a comparison, a selection of values for DLFs from
turn, had a value of 620.65 kg/m and was obtained from the prop- several studies can be found elsewhere (Živanović et al. 2005).
erties of the materials employed to build the structure. The cali- Each pedestrian in a flow was assigned a pacing rate randomly
brated FE model presented a fundamental frequency of 4.28 Hz selected and following a normal distribution, according to the
and correlated well with the measurements, particularly for the respective mean and SD obtained from the tests, for the respective
mode of interest. density. As for the mass, it was distributed according to the mean

Published by NRC Research Press


da Silva et al. 1201

Fig. 8. Illustration of the crowd load model.


Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

and SD presented in Table 1. The phase angles ␾ among the har- Table 3. Sample sizes employed.
monics of the pedestrian load (eq. (6)) were randomly assigned, Pedestrian density
with values following a uniform distribution and varying from –␲ (ped/m2)
to +␲.
A group of point loads was conceived to provide a continuous Data set 0.3 0.7 0.9
pedestrian flow crossing the structure, during a time compatible measured 10 10 7
to that employed in the experimental campaign. This is described Force-only model 29 26 79
in the following stages and the test case of 0.3 pedestrians/m2 is Force and fixed biodynamic model 11 12 72
taken to exemplify the procedure (see also Fig. 8): Force and moving biodynamic model 15 15 15
(a) first, the extension of the group of point loads is determined.
Such an extension is equal to the distance that the pedestrian
For personal use only.

at the end of the queue has to cover until leaving the foot- loads were employed, in a quantity equal to number of pedestri-
bridge. Using the pedestrian mean speed of the test plus the ans at a time on the footbridge. For a given pair of values of M and
duration of the measured signal (30 s of stationary flow, dis- fp, the parameters of the biodynamic models were obtained by
regarding the time to cross the footbridge), the extension of applying eq. (4).
the group of point loads is obtained by multiplying the mean The Newmark method was employed to solve the equation of
speed by 30 s and adding to this result the span length of the motion and the following values of the parameters of the method
footbridge. For instance, for the test case of the density of were adopted to provide unconditional stability: ␦ as 0.50 and ␣ as
0.3 pedestrians/m2, the mean pacing rate was 1.90 Hz (Araújo 0.25. In addition, the time step of the analysis was 0.0025 s, being
et al. 2009) and the mean speed was 0.9 times this pacing rate approximately 1/100 of the fundamental period of vibration, so as
(eq. (8)); the extension obtained was 62.6 m. to obtain accuracy for the solution (Bathe 1996).
(b) the number of pedestrians (point loads) placed along the
aforementioned extension is calculated in such a way to ob- Comparison between measurements and
tain a pedestrian density similar to that of the respective den- simulations
sity of the test. Taking the same example, that is, the density
of 0.3 pedestrians/m2, five pedestrians were on the footbridge The results from the simulations presented more variability
at a time, leading to a mean distance of approximately 2.26 m than the respective measurements and therefore the number of
among them. Then, the number of 28 pedestrians in the samples differed from case to case. This possibly had to do with
queue is obtained by the ratio between the extension and the the application of the group of point loads, which may present
mean distance among pedestrians. much variability than the actual situation observed during the
(c) finally, for each run (simulation), the position of each load tests. However, the same number of pedestrians at a time was at
(pedestrian) in the queue is randomly defined by a value be- the footbridge in both sets of results (measurements and simula-
tween zero and the calculated extension, using an uniform tions) and the same duration of analysis was adopted. In addition,
distribution. It should be noted that, due to the different pac- consistency was observed when comparing the sets of results of
ing rates (and so the different speed each load moves, see different densities. Therefore, the differences in variability were
eq. (8)), the distances among the loads do not remain constant considered a minor effect and the final number of samples ad-
during each run and also among runs. Each run, thus, pro- opted in each test or simulation is presented in Table 3, being
duced different accelerations, requiring repetition for statis- defined to provide statistical significance for the respective mean
tical significance. value of each set. Still regarding the number of samples, some
fluctuation is observed and this was because: (a) data from each set
With regard to the inclusion of the biodynamic models, they was reviewed to discard outliers, according to Triola (2008); and
were placed on nodes of the FE model and the group of moving (b) the number of samples should also be such that hypotheses
point loads was also applied as previously described. In the case of tests confirm that differences in mean values were due to differ-
the force and moving biodynamic models, each biodynamic ences in modeling the load.
model was assigned a value of the body mass M and pacing rate fp Mean RMS acceleration and respective SDs for each analysis are
according to the adopted distribution of these parameters and presented in Table 4, in which the measured results are shown
moved together with the force, from node to node of the FE again to simplify the comparison with results from the simula-
model. In the case of the fixed biodynamic models, since they do tions. For the density of 0.9 pedestrians/m2, three samples within
not move together with the forces, the values of the body mass M the set of the measurements were discarded because they were
and pacing rate fp of the first pedestrians in the group of point found to be outliers. This is why the mean and SD for this density

Published by NRC Research Press


1202 Can. J. Civ. Eng. Vol. 40, 2013

Table 4. Mean root mean square acceleration and standard deviation (SD) from measured and
simulated results.
Force and fixed Force and moving
Measured Force-only biodynamic biodynamic
(m/s2) model (m/s2) model (m/s2) model (m/s2)
Density
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

(ped/m2) Mean SD Mean SD Mean SD Mean SD


0.3 0.1164 0.0199 0.1397 0.0355 0.1060 0.0089 0.1154 0.0184
0.7 0.0854 0.0097 0.1130 0.0191 0.0920 0.0055 0.0906 0.0083
0.9 0.0984 0.0050 0.1122 0.0211 0.1053 0.0128 0.0918 0.0078

Fig. 9. Mean spectra for a density of 0.3 pedestrians/m2.


For personal use only.

differed between the results presented in Table 4 to the respective (2011), in which differences due to the dynamics of the body were
results presented in Table 2. noticed only around the resonance of the system. Such a differ-
It can be seen in Table 4 that the results of the force-biodynamic ence is related to the level of response of the structure for a
models (both moving and fixed) are in closer agreement with the particular frequency, since structural motion is needed for the
measured results than the results from the force-only model. In effects of force-only and force-biodynamic models to differ be-
addition, the results of the force-biodynamic models followed the tween each other, which occurs to a greater extent around the
trend of variation of mean RMS acceleration observed in the mea- resonance of the system.
surements. It can be seen in Figs. 9 to 11 that, around the resonance, the
By taking the mean, SD, and number of samples of each set of measured and both force-biodynamic models response peaks oc-
simulations, a hypothesis test (Triola 2008) was carried out for curred at a frequency lower than that observed for the results of
each pair of mean accelerations of same density, to verify whether the force-only model. The peak measured frequency continued to
there was sufficient evidence to state that the mean values from decrease with the increase of the number of pedestrians simulta-
each test were statistically distinct, that is, whether the mean neously at the footbridge (that is, with density) and this effect
values were actually affected by the distinct models employed to was followed by the results in which both force-biodynamic mod-
represent the pedestrian action. This was applied between the els were employed. However, this did not occur when applying
force-only model and each of the force-biodynamic models. A sig- the force-only model to represent the action of the pedestrians.
nificance level of 0.05 was adopted and all tests confirmed that The response of the footbridge excited by the force-only model
there was sufficient evidence to support the claim that the choice showed a sharp peak at a frequency of approximately 4.2 Hz for all
of model affected the value of mean accelerations. densities, which corresponded to the natural frequency of the
However, more clarifying results were observed by compar- empty structure.
ing the mean spectra (measurement and simulations), shown A reduction of the value of the peak when applying the force-
in Figs. 9 to 11 for each density investigated. With regard to the only model was observed as the density increased. This is due to
response around the first harmonic of the walking load, the ex- the fact that the mean pacing rate of pedestrians reduced with
perimental results were somehow lower than those respectively increased density, and thus the pacing rates were getting farther
obtained by both simulations. This is possibly due to some inade- away from the natural frequency of the empty footbridge as the
quacy of the DLFs adopted to represent the first harmonic of the density increased. Therefore, a consistent reduction of peak val-
walking load in the force models or else a limitation of the force ues were observed as the density increased.
model itself. Another observation was that the response did not It can also be observed that the responses obtained by employ-
differ much between the simulated results for that range of fre- ing both force-biodynamic models were more damped around the
quencies. This is consistent with the observations of Caprani et al. resonance than the responses obtained by employing the force-

Published by NRC Research Press


da Silva et al. 1203

Fig. 10. Mean spectra for a density of 0.7 pedestrians/m2.


Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

Fig. 11. Mean spectra for a density of 0.9 pedestrians/m2.


For personal use only.

only model, as the former presented peaks of lesser magnitude another in which a number of biodynamic models equal to the
than the latter. In other words, an increase in damping due to the number of pedestrians that were on the footbridge at a time were
presence of pedestrians on the structure occurred and this is ob- added to the model of the structure, being evenly distributed
served only when adopting the force-biodynamic models to rep- along the span in fixed positions. In addition, forces were applied
resent the pedestrian action. moving along the structure. The expression adopted to represent
Finally, a comparison between the results from the moving and such forces was taken from the literature, being obtained when
fixed biodynamic models showed that the latter can also be rea- walking in rigid surfaces. The combination of such forces with
sonably employed to represent the effects of reduction on natural
biodynamic models provided the net effect of a pedestrian walk-
frequency and increase on damping.
ing in a vibrating surface.
Conclusions A prototype footbridge was employed and pedestrian tests were
carried out, with densities ranging from 0.3 to 0.9 pedestrians/m2,
In this paper a crowd load model is proposed to represent the
so as to check the proposed crowd model. By comparing the re-
action of pedestrians walking along a footbridge, applicable to
pedestrian densities up to approximately 1.0 pedestrian/m2. This sponses obtained from measurements with simulations employ-
is thought as a step further in modeling the action of pedestrians ing the model, it was observed that the results with the added
on footbridges, while comparing to force models adopted in cur- biodynamic models were in better agreement with the measure-
rent codes of practice. In the proposed model, SDOF biodynamic ments than those results obtained representing the pedestrian
models are added to represent the action of pedestrians. Two action by a model in which only forces moving along the structure
approaches were adopted in this task: one in which the biody- were applied. When including the biodynamic models as part of
namic models moved from node to node along the structure, and the structural system, a reduction on the predominant frequency

Published by NRC Research Press


1204 Can. J. Civ. Eng. Vol. 40, 2013

of vibration and an increase in damping were noticed, as evi- pedestrian bridge. Earthquake Engineering and Structural Dynamics, 22(9):
741–758. doi:10.1002/eqe.4290220902.
denced in the literature and observed in the measurements.
Inman, V.T., Ralston, H.J., and Todd, F. 1994. Human locomotion. In Human
Both approaches employing biodynamic models to represent walking. Edited by J. Rose and J.G. Gamble. Williams & Wilkins, Baltimore,
the pedestrian action, that is, with moving or fixed biodynamic Maryland. pp. 1–22.
models, produced results in close agreement with the measure- International Organization for Standardization. 1981. ISO 5982- Vibration and
Shock-Mechanical driving point impedance of the human body.
ments, the advantage in using the fixed biodynamic models being Kerr, S.C., and Bishop, N.W.M. 2001. Human induced loading on flexible stair-
the simplicity of a linear analysis and a much easier adaptation of cases. Engineering Structures, 23: 37–45. doi:10.1016/S0141-0296(00)00020-1.
Can. J. Civ. Eng. Downloaded from cdnsciencepub.com by Bangladesh Univ of Engineering & Technol - BUET (PERI) on 05/01/22

existing moving force algorithms to include biodynamic models. Kim, S.-H., Cho, K.-I., Choi, M.-S., and Lim, J.-Y. 2008. Development of human
Beyond confirming that walking crowds contributed to change body model for the dynamic analysis of footbridges under pedestrian in-
duced excitation. Steel Structures, 8: 333–345.
in the dynamic properties of the system, the results also corrobo- Macdonald, J.H.G. 2008. Pedestrian-induced vibrations of the Clifton suspension
rated that the biodynamic model employed improved the model- bridge, UK. Proceedings of the Institution of Civil Engineers Bridge Engineer-
ing of the pedestrian action. Its inclusion made the results from ing, 161(2): 69–77.
the simulations much closer to the measurements, thus avoiding Matsumoto, Y., and Griffin, M.J. 2003. Mathematical models for the apparent
masses of standing subjects exposed to vertical whole-body vibration. Jour-
conservative estimates obtained when applying force-only models nal of Sound and Vibration, 260: 431–451. doi:10.1016/S0022-460X(02)00941-0.
to represent the action of pedestrians in crowd situations. Miyamori, Y., Obata, T., Hayashikawa, T., and Sato, K. 2001. Study on identifica-
tion of human walking model based on dynamic response characteristics of
Acknowledgements pedestrian bridges. In Proceedings of the Eighth East Asia-Pacific Conference
on Structural Engineering and Construction, Paper No. 1066.
The first author acknowledges the financial support of CNPq – Nigg, B.M., and Liu, W. 1999. The effect of muscle stiffness and damping on
Brazil, in terms of a scholarship to carry out his doctoral studies. simulated impact force peaks during running. Journal of Biomechanics, 32:
849–856. doi:10.1016/S0021-9290(99)00048-2. PMID:10433427.
References Ontario Highway Bridge Design Code. 1983. Ministry of Transportation and
ANSYS. 2010. ANSYS v. 12.1 Reference Manual. Communications, Highway Engineering Division, Ontario, Canada.
Araújo, M.C., Jr., Brito, H.M.B.F., and Pimentel, R.L. 2009. Experimental evalua- Pachi, A., and Ji, T. 2005. Frequency and velocity of people walking. The Struc-
tion of synchronization in footbridges due to crowd density. Structural En- tural Engineer, 83(3): 36–40.
gineering International, 19(3): 298–303. doi:10.2749/101686609788957784. Pernica, G. 1990. Dynamic load factors for pedestrian movements and rhythmic
Barker, C., and Mackenzie, D. 2008. Calibration of the UK National Annex. In exercises. Canadian Acoustics, 18(2): 3–18.
Footbridge 2008 - Proceedings of the Third International Conference, Porto, Piccardo, G., and Tubino, F. 2009. Simplified procedures for vibration service-
Portugal. ability analysis of footbridges subjected to realistic walking loads. Comput-
Bathe, K.J. 1996. Finite Element Procedures. Prentice Hall, New Jersey, USA. ers and Structures, 87: 890–903. doi:10.1016/j.compstruc.2009.04.006.
Blanchard, J., Davies, B.L., and Smith, J.W. 1977. Design criteria and analysis for Pimentel, R.L., and Fernandes, H.M.B. 2009. Simplified expressions for the vibra-
dynamic loading of footbridges. Symposium on Dynamic Behaviour of tion serviceability of beam-like footbridges. Bridge Structures, 5(1): 3–16. doi:
For personal use only.

Bridges, Transport and Road Research Laboratory, Crowthorne, UK. TRRL 10.1080/15732480902857348.
Supplementary Report 275. pp. 90–106. Pimentel, R.L., Pavic, A., and Waldron, P. 2001. Evaluation of design require-
Brasiliano, A., Fernandes, H.H.F., Pimentel, R.L., Doz, G.N., and Brito, J.L.V. 2007. ments for footbridges excited by vertical forces from walking. Canadian
Output-only modal testing of a laboratory test footbridge. In EVACES'07 Pro- Journal of Civil Engineering, 28(5): 769–777. doi:10.1139/l01-036.
ceedings of the Experimental Vibration Analysis for Civil Engineering Struc- Rainer, J.H., and Pernica, G. 1986. Vertical dynamic forces from footsteps. Cana-
tures Conference, Porto, Portugal. dian Acoustics, 14(Part 2): 12–21.
British Standards Institute. 1978. BS5400 British Standards - Steel, concrete and Rainer, J.H., Pernica, G., and Allen, D.E. 1988. Dynamic Loading and Response of
composite bridges: specification for loads, Part 2, Appendix C. London. Footbridges. Canadian Journal of Civil Engineering, 15(1): 66–71. doi:10.1139/
British Standards Institute. 2008. UK National Annex to Eurocode 1: Actions on l88-007.
structures – Part 2: Traffic loads on bridges. NA to BS EN 1991-2:2003. London. Sachse, R., Pavic, A., and Reynolds, P. 2004. Parametric study of modal properties
Brownjohn, J.M.W., Pavic, A., and Omenzetter, P. 2004a. A spectral density ap- of damped two-degree-of-freedom crowd-structure dynamic systems. Journal
proach for modeling continuous vertical forces on pedestrian structures due of Sound and Vibration, 274: 461–480. doi:10.1016/j.jsv.2003.08.052.
to walking. Canadian Journal of Civil Engineering, 31: 65–77. doi:10.1139/l03- Sahnaci, C., and Kasperski, M. 2005. Random loads induced by walking.
072. In EURODYN 2005 Proceedings of the 6th International Conference on
Brownjohn, J.M.W., Fok, P., Roche, M., and Omenzetter, P. 2004b. Long span steel Structural Dynamics, Paris, France. pp. 441–446.
pedestrian bridge at Singapore Changi Airport – part 2: Crowd loading tests Service d'Études Techniques des Routes et Autoroutes (SÉTRA). 2006. Foot-
and vibration mitigation measures. The Structural Engineer, 82: 28–34. bridges – Assessment of vibrational behavior of footbridges under pedestrian
Butz, C., Feldmann, M., Heinemeyer, C., Sedlacek, G., Chabrolin, B., Lemaire, A. loading. Technical guide, SETRA, Paris.
et al. 2008. EUR 23318 - Advanced load models for synchronous pedestrian Shapiro, S.S., and Wilk, M.B. 1965. An Analysis of Variance Test for Normality
excitation and optimised design guidelines for steel footbridges, Research (Complete Samples). Biometrika, 52(3–4): 591–611. doi:10.1093/biomet/52.3-4.
Fund for Coal and Steel, European Comission. 591.
Caprani, C.C., Keogh, J., Archbold, P., and Fanning, P. 2011. Characteristic verti- Silva, F.T., and Pimentel, R.L. 2011. Biodynamic walking model for vibration
cal response of a footbridge due to crowd loading. In EURODYN 2011 Proceed- serviceability of footbridges in vertical direction. In EURODYN 2011 Proceed-
ings of the 8th International Conference on Structural Dynamics, Leuven, ings of the 8th International Conference on Structural Dynamics. Leuven,
Belgium. pp. 978–985. Belgium. pp. 1090–1096.
Charles, P. 2008. Application of french guidelines in design. In Footbridge Triola, M.F. 2008. Elementary Statistics. Pearson Education, Inc., Boston, MA.
2008 -Proceedings of the Third International Conference, Porto, Portugal. Tubino, F., and Piccardo, G. 2008. Vibration serviceability of footbridges: a
Charles, P., and Bui, V. 2005. Transversal dynamic actions of pedestrians. Syn- closed-form solution. In Footbridge 2008 – Proceedings of the Third Interna-
chronization. Footbridge 2005 - Second International Conference, OTUA, tional Conference, Porto, Portugal.
Venice, Italy. Živanović, S., Pavić, A., and Reynolds, P. 2005. Vibration serviceability of foot-
Dallard, P., Fitzpatrick, A.J., Flint, A., Bourva, S., Low, A., Smith, R.M.R., et al. bridges under human-induced excitation: a literature review. Journal of
2001. The London Millennium footbridge. The Structural Engineer, 79: 17–33. Sound and Vibration, 279: 1–74. doi:10.1016/j.jsv.2004.01.019.
Danbon, F., and Grillaud, G. 2005. Dynamic behaviour of a steel footbridge - Živanović, S., Pavić, A., and Reynolds, P. 2007. Probability-based prediction of
Characterisation and modelling of the dynamic loading induced by a moving multi-mode vibration response to walking excitation. Engineering Struc-
crowd on the Solferino footbridge in Paris. In Footbridge 2005 - Second Inter- tures, 29(6): 942–954. doi:10.1016/j.engstruct.2006.07.004.
national Conference, OTUA, Venice, Italy. Živanović, S., Pavić, A., and Ingólfsson, E.T. 2010. Modeling Spatially Unre-
Fujino, Y., Pacheco, B.M., Nakamura, S., and Warnitchai, P. 1993. Synchroniza- stricted Pedestrian Traffic on Footbridges. Journal of Structural Engineering,
tion of human walking observed during lateral vibration of a congested 136(10): 1296–1308. doi:10.1061/(ASCE)ST.1943-541X.0000226.

Published by NRC Research Press

Vous aimerez peut-être aussi