Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract: Although many models of pedestrian dynamic loading have been proposed, possible bidirectional interactions between the
walker and the excited structure are generally ignored, particularly for vertical vibrations. This shortcoming has arisen from scarcity of data
Downloaded from ascelibrary.org by University of Bristol on 11/27/13. Copyright ASCE. For personal use only; all rights reserved.
on gait-adaptation strategies used in the presence of structural motion and, as a consequence, the absence of a credible fundamental pedestrian
model capable of capturing the underlying relations between the two dynamic systems. To address this inadequacy of current approaches,
a biomechanically inspired inverted-pendulum pedestrian model has been applied to the human-structure interaction problem. The behavior
of the model is studied when subjected to vertical motion of the supporting structure, in particular, in relation to potential self-excited forces
that can be generated. A mechanism has been identied by which the timing of pedestrian footsteps can be altered subtly, giving a net damp-
ing effect on the structure, without necessarily involving full synchronization. It has been found that depending on the ratio between the
bridge vibration frequency and pedestrian pacing frequency, walkers can effectively act as positive or negative dampers to the structural
motion, but it is expected that for a group of pedestrians with distributed parameters, their action is, on average, to add damping and mass.
DOI: 10.1061/(ASCE)BE.1943-5592.0000490. 2013 American Society of Civil Engineers.
CE Database subject headings: Bridges; Human factors; Pedestrians.
Author keywords: Bridges; Human-structure interaction; Biomechanics; Gait; Inverted-pendulum model; Vertical pedestrian forces.
oscillating structure to investigate the effect of vertical bridge motion two-degree-of-freedom SMD body model in ISO 5982:1981 (ISO
on human gait and hence on the interaction forces. A simple foot- 1981) for analysis of pedestrian-induced vertical bridge vibra-
placement control law satisfying kinematic constraints is adopted to tions. A single-degree-of-freedom SMD body model was used by
ensure forward progression. The results from the model in terms of the Brownjohn et al. (2004), Tavares da Silva and Pimentel (2011), and
equivalent added damping and mass to the bridge are compared with the Caprani et al. (2011). Alexander (2006) added an actuator to the two-
available data from full-scale bridges and experimental investigations. degree-of-freedom SMD crowd-structure system model, although
pedestrians were not treated there as discrete entities. However, there
is a clear shortcoming of these models for walking pedestriansthey
Modeling of Human-Structure Interaction
are not able to fully capture the inuence of the feedback from the
Biomechanically inspired modeling of pedestrian action in the context vertical motion of the walking surface on the kinematics and kinetics of
of engineering structures is a relatively novel direction of research. Its human gait. Specically, the changes in gait parameters resulting from
emergence is the result of serviceability problems experienced by two disturbance to the trajectory of the center of mass (COM) are neglected,
high-prole bridges: the Solferino Footbridge in Paris, opened in 1999 which can potentially modify the pedestrian loading. This originates in
(Danbon and Grillaud 2005), and the previously mentioned LMF the fact that these models ignore the bipedal nature of human locomotion.
(Dallard et al. 2001). When subjected to crowd loading, unforeseen Humans use two lower limbs for normal (i.e., nonpathologic and
large divergent amplitude lateral vibrations developed, causing the unconstrained) walkinga form of locomotion termed bipedalism.
authorities to make decisions of close and later retrot of these bridges The implications of this feature on human gait are profound, inuencing
with damping devices. Since then, many other bridges susceptible to qualities such as speed, acceleration, and maneuverability, endurance,
pedestrian-induced vibrations have been identied, and much effort economy, energy, and at last but not least, stability (Alexander 2003).
has been spent on studying the cause of this phenomenon. Most of the The last of these qualities is possibly the most essential to a functional
proposed explanations focus on synchronization of the pedestrians gait because while different legged animals have gaits that differ in
frequencies with the structural vibration frequency and adjustments of patterns of movement (e.g., walking, running, hopping, trotting, and
the pedestrians phases so as to increase bridge motion. This, perhaps galloping), their gaits share a common characteristicthey are rst and
the most intuitive, solution could not, however, explain instability of foremost stable. The reason for this is simplethe energy losses as-
two other bridges: the Changi Mezzanine Bridge at Singapore Airport sociated with lack of stability are too costly (e.g., recovery of potential
(Brownjohn et al. 2004) and the Clifton Suspension Bridge in energy and buildup of kinetic energy level after a fall). This also reveals
Bristol, United Kingdom (Macdonald 2008), where no evidence of why no credible model of human walking exists in the eld of bio-
synchronization was observed during periods of instability. The mechanics that ignores bipedalism. Therefore, in contrast with other
measurements from these bridges indicated that some other models of vertical human-structure interaction for walking pedestrians,
mechanism is required, conceivably similar to Barkers (2002) a bipedal model is considered here in which the bridge motion perturbs
model, in which the lateral force was obtained from the action of the gait and challenges its stability. In line with the statement that
a mass moving along the bridge in a straight line, supported by an simplicity promotes understanding validated in this context by McGeer
inclined leg alternately switching from right to left. Although this (1990) and Alexander (1992), the focus of this paper is identication of
simple model was not scientically justied, an important discovery the fundamental relations between a pedestrian modeled as a minimal
was madepedestrians can effectively input energy into a vibrating biped and a vertically oscillating bridge. The motion of the bridge is
bridge even if their walking frequencies are far from the frequency shown to modify the passive motion of the pedestrians COM and thus
of the vibrating structure. Direct empirical evidence of this was cause the emergence of self-excited forces on the structure.
obtained from tests during which pedestrians walked on a laterally
oscillating instrumented treadmill by Pizzimenti and Ricciardelli
Minimal Biped Model
(2005) and later by Inglfsson et al. (2011). A rigorous explanation
was provided by Macdonald (2009), who built a biomechanically The simplest possible stepping model of human gait is a rigid inverted
inspired biped inverted-pendulum model in which a stepping pendulum with instantaneous transfer from one leg to the other. This
strategy was adopted based on the requirement of maintaining model is especially valid for the single support phase of the gait,
balance (Hof 2008). Unlike in Barkers model, motion of the mass which, depending on step rate, comprises around 7095% of the
of pedestrians was included, therefore allowing for full pedestrian- whole gait cycle (Inman et al. 1981). In this period, the CoM moves
bridge dynamic interaction. Bocian et al. (2012) found that the along the arc determined by the length of the stance limb, here the
output of this model is in quite good agreement with the experi- pendulum leg. An advantage of this process is that as for a normal
mental investigations and measurements on full-scale bridges. pendulum, energy is conserved. There is transfer between gravita-
A new trend in modeling vertical pedestrian loading has origi- tional potential energy and kinetic energy, but once initiated, no
nated only recently, prompted by observations that the presence of mechanical work is required to sustain motion. Similar efciencies can
u 21 g zcos u (2)
l Self-Excited Forces
where u 5 support-leg inclination angle; l 5 equivalent inverted- To consider the inuence of vertical bridge motion on the interaction
pendulum length; g 5 gravitational acceleration; z 5 vertical dis- forces, sinusoidal motion of the supporting surface was assumed for
placement of the bridge; and dots over the symbols represent all the following analysis, given by:
derivatives with respect to time. Because Eq. (2) is nonlinear, the
zt A sinvb t (4)
MathWorks MATLAB ode45 numerical ordinary differential equa-
tion solver, incorporating the Runge-Kutta algorithm, was used to
where A 5 amplitude of the bridge motion; and vb 5 angular
obtain time-history solutions. Variable time steps were allowed for the
solver for computation efciency, with the upper bound set to 0.01 s. frequency.
The effect of the bridge motion on the pedestrian modies the
The results were then resampled into time series with regular 0.001 s
sampling intervals. The vertical force on the bridge F is given by interaction force [see Eqs. (2) and (3)], which can potentially in-
clude components at the bridge frequency and which from the point
of view of the bridge equation of motion (not addressed here per se)
F mp g z l
u cos u 2 lu_ sin u
2
(3)
can be considered as equivalent to added mass and added damping
of the structure. Although there is no fundamental difculty in
where mp 5 pedestrian mass. modeling the structure as well, this would involve additional
Clearly, the motion of the bridge z perturbs the motion of the parameters and would detract from the main goal with no real
pedestrian COM u, as given by Eq. (2), which thus modies the benet. From the point of view of the individual studied here, the
interaction force, in accordance with Eq. (3). This is a direct con- bridge is forced by the crowd or other external loading, and the
sequence of the bipedalism and has not been captured by any individuals inuence on bridge motion is minor. The pedestrian
previous model of vertical human-structure interaction.
effectively experiences imposed motion of the walking surface,
For simplicity, the transfer from one leg to the other is taken to be
which would be dominated by one or more natural frequencies of
instantaneous, and the dynamics of the swing leg are omitted. The
latter simplication is made based on the observation by Pandy
(2003) that the vertical ground reaction force during the single-
support phase is mainly a result of the stance-leg dynamics. To
account for the change in vertical velocity necessary at the transfer,
Simulation Period
To determine the magnitude of the long-term average equivalent added
damping and mass, the simulations were run for at least 600 pedestrian
steps. Provision was made for any initial transients to decay; hence the
Downloaded from ascelibrary.org by University of Bristol on 11/27/13. Copyright ASCE. For personal use only; all rights reserved.
mass per pedestrian was 56 kg. Lack of evidence of increased dominant component of the average self-excited forces in the
damping on the RCF was explained by Zivanovi c et al. (2010) to be integrals in Eqs. (5) and (6). An example of the case where fb is
caused by two factors: insufcient acceleration levels and time of slightly lower than fp is shown in Fig. 6(a), whereas an example of
exposure as a result of the short lengths of the spans on which the case where fb is slightly higher than fp is shown in Fig. 6(b). In
measurements were taken. Another possible explanation for the these plots, densication of the phase angles occurs in the vicinity of
discrepancy between the measured and expected added damping p [Fig. 6(a)] and in the vicinity of 0 or 2p [Fig. 6(b)]. For com-
might be its dependence on the walking frequency that is captured by parison, an example where fb is not close to fp or its positive-integer
the current model (Fig. 5). However, this issue requires further multiples is shown in Fig. 6(c). The phase angles in this case are
investigation based on empirical data. evenly distributed throughout the whole range.
The presented model is also reasonably consistent with the The plots presented in Fig. 6 reveal the reason for the different
results of laboratory investigations on the behavior of a slab sub- effective added damping from the pedestrians in different bridge
jected to the action of 2, 4, 6, and 10 walking people reported by frequency ranges. For bridge frequencies far from integer multiples
Zivanovi c et al. (2009). During testing, the slab with a mass of of the (unperturbed) pedestrian pacing frequency [e.g., Fig. 6(c)], the
approximately 15,000 kg vibrated with a frequency of approxi- step periods are affected little by the bridge motion, and there is
mately 4.45 Hz and a displacement amplitude of approximately a steadily progressing phase difference between the bridge and
2 mm (based on acceleration amplitudes equal to peak acceleration pedestrian, giving an even distribution over all phase angles. Hence,
levels of 1:85, 1:6, 1:46, and 1:19 m=s2 achieved during testing). It on some steps, the work done by the pedestrian on the bridge will be
was identied for all tests that pedestrians acted as a source of positive, but equally on others steps, it will be negative, so on
additional benecial damping to the structure, increasing its damping average overall there is little work done on the bridge, and the ef-
ratio approximately proportionally to the number of walkers from fective added damping coefcient is close to zero [Fig. 5(a);
0.72% in the case of an empty slab to 2.86% in the case of 10 fb 5 0:5 Hz).
pedestrians present on the slab. The simulations were run for the For bridge frequencies slightly above the (unperturbed) pedes-
average vibratory conditions measured during testing (A 5 2 mm and trian pacing frequency [e.g., Fig. 6(b)], for phase angles close to zero
fb 5 4:45 Hz). Given the lack of details regarding the behavior of the (i.e., bridge velocity upward at the beginning of the pedestrian step),
pedestrians and their parameters, model parameters had to be chosen. the bridge is generally accelerating downward for the rst half of the
C. T. Georgakis (personal communication, 2012) found that during step (u , 90; see Fig. 1) and upward for the second half of the step
similar tests on a footbridge set in a laboratory environment, the (u . 90). Both these bridge accelerations tend to accelerate the
preferred pacing rate of pedestrians decreased to 1.38 Hz, and he pedestrian motion [see Eq. (2)]. Hence the pedestrian arrives at
attributed this behavior to physical and psychological constraints the end of the step earlier than for unperturbed walking. Therefore,
associated with a narrow deck and a lack of handrails. It was assumed the relative phase angle between the bridge and pedestrian pro-
herein that pedestrians acted in the same manner during tests con- gresses more slowly than otherwise expected. Conversely, for phase
ducted by Zivanovi c et al. (2009) and that the average pedestrian angles close to p at the beginning of the step, the pedestrian motion is
walking frequency matched that measured by Georgakis. Based on decelerated by the bridge acceleration, and the relative phase angle
these assumptions and using data from Bertram and Ruina (2001), the progresses more rapidly. Hence there is a concentration of phase
corresponding average pedestrian parameters were established angles close to zero. The principal dynamic load from the pedestrian
as uTD 5 74 and x_ 0n 5 1:065 m=s. The estimated average equivalent is the impulse at the foot transition. For the favored phase re-
added damping per pedestrian was 283 Ns=m, and the average lationship, i.e., close to zero, the impulse (downward on the bridge)
equivalent added mass per pedestrian was 61 kg. More detailed occurs when the bridge velocity is upward; hence it has a damping
comparisons would require further speculations as to the behavior effect on the bridge motion, so the net DC is positive [Fig. 5(a);
of pedestrians during testing and therefore are omitted. fb 5 1:9 Hz; for x_ 0n 5 1:55 m=s, fp 1:86 Hz]. Similarly, for bridge
Unfortunately, the authors are not aware of any experimental data frequencies slightly below the (unperturbed) pedestrian pacing
in the frequency ranges where the model predicts equivalent nega- frequency [e.g., Fig. 6(a)], the bridge acceleration modies the
tive damping for further validation of the model. pedestrian step periods so as to concentrate the relative phases near
Note that the peak damping coefcients (Fig. 5) are much larger p, giving net negative effective damping to the bridge [Fig. 5(a);
in magnitude than the value of 2300 Ns=m proposed by Arup for fb 5 1:8 Hz; for x_ 0n 5 1:55 m=s, fp 1:86 Hz].
lateral bridge vibrations from the behavior of the LMF (Dallard et al. The effect of the phase angle between the beginning of step and
2001). Hence the estimated effective damping for vertical vibrations the beginning of bridge vibration cycle [occurring when z 5 0 and z_
could have a major inuence on bridge behavior. Also note that, on is positive; see Eq. (4) and Fig. 1], in relation to the bridge vibration
average, the pedestrians potential to add damping to the bridge is period, on the pedestrian motion is presented in more detail in Fig. 7.
stronger than the potential to effectively subtract damping. This All results were obtained for a model with representative pedestrian
parameters uTD 5 70(u 2 huTD ; 180+ 2 uTD i), x_ 0n 5 1:55 m=s (giv-
ing a mean velocity of 1:33 m=s and fp 5 1:86 Hz for stationary
ground), A 5 3 mm, and fb 5 1:8 Hz, corresponding to Fig. 6(a). uz
Fig. 6. Phase angle between timing of instantaneous transfer from one presented in the bottom subplots is the component of the angular
leg to the other and beginning of the nearest preceding bridge vibration acceleration of the COM that depends on the bridge acceleration
cycle [occurring when z 5 0 and z_ is positive; see Eq. (4) and Fig. 1] in (presented in the middle subplots), resolved by the cosine of the leg
relation to bridge vibration period (after conservatively discarding rst inclination angle u (presented in the top subplots) in the equation of
300 steps, allowing the transients to decay) for a model with repre- motion [uz 5 21=lz cos u; see Eq. (2)]. Therefore,
uz modies the
sentative pedestrian parameters uTD 5 70; x_ 0n 5 1:55 m=s (giving mean dominant component of the angular acceleration, which depends on
walking velocity of 1:33 m=s and fp 5 1:86 Hz for stationary ground); gravity [
ug 5 21=lg cos u; see Eq. (2)], which denes the behavior
A 5 3 mm and (a) fb 5 1:8 Hz; (b) fb 5 1:9 Hz; (c) fb 5 0:5 Hz of the model in the case of stationary ground. Fig. 7(a) reveals that for
a step starting at a bridge phase angle of p [hence acceleration
z 5 2A sinvb ts 1 p, where ts is the time from the beginning of the
step, as in the middle subplot of Fig. 7(a); see Eq. (4)], the vertical
jects, support conditions, such as knee postures and whether one or Barker, C. (2002). Some observations on the nature of the mechanism that
two feet are in contact with the ground, have been found to inuence drives the self-excited lateral response of footbridges. Proc., Foot-
the resonant frequency of the apparent mass. The main resonance of bridge 2002: 1st Int. Conf., Paris.
a standing human exposed to 1 m=s2 (RMS) vibrations occurs at Bertram, J. A. E., and Ruina, A. (2001). Multiple walking speed-frequency
5.5 Hz and decreases to 3.75 Hz in the case of standing on one leg and relations are predicted by constraint optimization. J. Theor. Biol.,
to 2.75 Hz for a posture in which both legs are bent and in contact 209(4), 445453.
with the ground (Matsumoto and Grifn 1998). Such effects could Bocian, M., Macdonald, J. H. G., and Burn, J. F. (2012). Biomechanically
be represented by changes in the stiffness and damping of the SMD inspired modelling of pedestrian-induced forces on laterally oscillating
human models (Matsumoto and Grifn 2003). Hence, to fully structures. J. Sound Vibrat., 331(16), 39143929.
capture the damping effect of walking pedestrians on the supporting British Standards Institution (BSI). (2003). UK national annex to Eurocode
structure, it may be necessary to include such features along with the 1: Actions on structures. 2: Trafc loads on bridges, London.
Brownjohn, J. M. W., Fok, P., Roche, M., and Omenzetter, P. (2004). Long
bipedal behavior considered here. This could be achieved, for ex-
span steel pedestrian bridge at Singapore Changi Airport. 2: Crowd loading
ample, by adding a spring and a dashpot damper to the current model.
tests and vibration mitigation measures. Struct. Eng., 82(16), 2834.
However, importantly, the current model demonstrates that sig- Butz, C., et al. (2008). Advanced load models for synchronous pedestrian
nicant effective damping of the structure can arise as a result of the excitation and optimised design guidelines for steel footbridges. RFSR-
simple bipedal mechanics of walking, without considering the internal CT-2003-00019, European Commission, Brussels, Belgium.
dynamic characteristics of the human body. Therefore, any model of Caprani, C. C., Keogh, J., Archbold, P., and Fanning, P. (2011). Char-
the human-structure interaction from walking pedestrians should take acteristic vertical response of a footbridge due to crowd loading. Proc.,
into account the stepping behavior and its modication by bridge Eurodyn 2011: 8th Int. Conf. on Structural Dynamics, 978985.
motion. Furthermore, currently used models of human dynamic loading Collins, S., Ruina, A., Tedrake, R., and Wisse, M. (2005). Efcient bipedal
neglect this effect, which could signicantly affect the resulting struc- robots based on passive-dynamic walkers. Science, 307(5712), 1082
tural response amplitudes. 1085.
Dallard, P., et al. (2001). The London Millennium Footbridge. Struct.
Conclusions Eng., 79(22), 1733.
Danbon, F., and Grillaud, G. (2005). Dynamic behaviour of a steel foot-
bridge: Characterisation and modelling of the dynamic loading induced
This paper presents results of a study on interactions between a walking
by a moving crowd on the Solferino Footbridge in Paris. Proc.,
pedestrian and a vertically oscillating structure. It has been shown that
Footbridge 2008: 2nd Int. Conf., Porto, Portugal.
the pedestrian has a potential to act as a source of either positive or Grifn, M. J. (1990). Handbook of human vibration, Academic Press,
negative equivalent added damping and mass to the structure depending London.
on the system parameters. However, the net expected effect from a group Grundmann, H., Kreuzinger, H., and Schneider, M. (1993). Schwin-
of pedestrians with distributed parameters is likely to be an increase in gungsuntersuchungen fr Fugngerbrcken [Vibration tests of pe-
effective damping and mass. A possible subtle entrainment mechanism destrian bridges]. Der Bauingenieur, 68, 215225 (in German).
has been identied that, depending on the pedestrian-to-bridge fre- Hof, A. L. (2008). The extrapolated center of mass concept suggests
quency ratio, could be responsible for providing negative or positive a simple control of balance in walking. Hum. Mov. Sci., 27(1), 112125.
damping without necessarily involving full synchronization of the pe- Hof, A. L., Gazendam, M. G. J., and Sinke, W. E. (2005). The condition
destrian to the bridge frequency. The key nding is that signicant ef- for dynamic stability. J. Biomech., 38(1), 18.
fective damping and variations in effective mass of the structure resulting Inglfsson, E. T., Georgakis, C. T., Ricciardelli, F., and Jnsson, J. (2011).
from pedestrians can arise from their stepping behavior, which has not Experimental identication of pedestrian-induced lateral forces on
been considered previously. This is apart from any effects of resonances footbridges. J. Sound Vibrat. 330(6), 12651284.
Inglfsson, E. T., Georgakis, C. T., and Svendsen, M. N. (2008). Vertical
or internal damping of the human body. However, improvements to the
footbridge vibrations: Details regarding and experimental validation of the
model, based on experimental observations of human gait, are needed to
response spectrum methodology. Proc., Footbridge 2008: 3rd Int. Conf.,
build a more reliable fundamental pedestrian model.
Porto, Portugal.
Inman, V. T., Ralston, H. J., and Todd, F. (1981). Human walking, Williams
Acknowledgments & Wilkins, Baltimore.
Institution of Structural Engineers (ISE). (2008). Dynamic performance
This work was supported by an Engineering and Physical Sciences requirements for permanent grandstands subject to crowd action:
Research Council (EPSRC) Doctoral Training Account studentship Recommendations for management, design and assessment, London.
for Mateusz Bocian and an EPSRC Advanced Research Fellowship International Federation for Structural Concrete (b). (2005). Guidelines
for John H. G. Macdonald. for the design of footbridges. b Bull. 32, Lausanne, Switzerland.