You are on page 1of 29

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

net/publication/228644791

Mitigation of Pedestrian-induced Vibrations in Suspension Footbridges via


Multiple Tuned Mass Dampers

Article in Journal of Vibration and Control · April 2010


DOI: 10.1177/1077546309350188

CITATIONS READS

46 664

3 authors, including:

Walter Lacarbonara Fabrizio Vestroni


Sapienza University of Rome Sapienza University of Rome
319 PUBLICATIONS 5,531 CITATIONS 216 PUBLICATIONS 4,943 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Walter Lacarbonara on 18 September 2014.

The user has requested enhancement of the downloaded file.


Mitigation of Pedestrian-induced Vibrations in
Suspension Footbridges via Multiple Tuned Mass
Dampers

NICOLA CARPINETO
WALTER LACARBONARA
FABRIZIO VESTRONI
Dipartimento di Ingegneria Strutturale e Geotecnica, University of Rome La Sapienza, via
Eudossiana, 18, Rome 00184, Italy (walter.lacarbonara@uniroma1.it)
(Received 9 January 20081 accepted 4 June 2009)

Abstract: The dynamic response of suspension footbridges to pedestrian-induced excitations and its passive
mitigation, via multiple tuned mass dampers (TMDs), are investigated. First, the nonlinear equations of mo-
tion are obtained assuming finite planar motions of the suspension bridge. A suitable approximate version of
the equations of motion is shown to be in agreement with existing theories and its linearization is then em-
ployed in the structural dynamics analyses. A Galerkin discretization is exploited to calculate both the free
and forced dynamic response towards the design of the vibration control system. First, the leading character-
istics of the bridge dynamic response are outlined. Resonant vibrations induced by the passage of pedestrians
are shown to be effectively reduced using viscoelastic TMDs. As the frequencies of the lowest two modes
in suspension footbridges can be very close in the proximity of the crossover phenomenon, three different
design scenarios are considered: below, near and above the crossover. In particular, the influence of these
scenarios on the passive control architecture is investigated.

Key words: Suspension footbridge, multiple tuned-mass dampers, pedestrian excitation, vibration absorber.

1. INTRODUCTION

The dynamic behavior of suspension footbridges with moderate spans is characterized by


the fact that the dominant frequencies are around the frequencies of the pedestrian-induced
excitations and typical earthquake motions. The dynamic aspects of the response need to be
addressed suitably in the design cycle as well as in the safety and comfort assessment. In
addition to vertical vibrations induced by the dynamic forces generated by walking crowds
of pedestrians, severe lateral vibrations may also be excited on suspension footbridges as
occurred in the London Millennium Footbridge (Dallard et al., 2001) on its opening day. The
movement of the Millennium Bridge was caused by a substantial lateral loading effect which
had not been anticipated during the design and had rarely been investigated before. Fujino
et al. (1993) studied the human-induced large-amplitude lateral vibrations of a pedestrian

Journal of Vibration and Control, 16(5): 749–776, 2010 DOI: 10.1177/1077546309350188


1
12010 SAGE Publications Los Angeles, London, New Delhi, Singapore
750 N. CARPINETO ET AL.

bridge under an extremely congested condition. The loading effect has been found to be due
to the synchronization of lateral footfall forces within a large crowd of pedestrians on the
bridge. This arises because it is more comfortable for pedestrians to walk in synchronization
with the natural swaying of the bridge, even if the degree of swaying is initially very small.
This instinctive behavior ensures that footfall lateral forces are applied at the resonant fre-
quency of the bridge lateral bending mode, and with a phase that amplifies the motion of the
bridge continuously.
There are at least two open fronts of research in suspension footbridge dynamics1 one is
related to the derivation of suitable distributed-parameter models, and the second is related
to devising feasible control strategies to make these lively structures safe and comfortable.
The use of distributed-parameter (DP) models of suspension footbridges, already pro-
posed by Brownjohn (1997) and Brownjohn et al. (1994) in linearized form, offers greater
flexibility and computational enhancement with respect to finite-element (FE) procedures,
since DP models allow extensive parametric investigations of the static and elastodynamic
response. However, when the amplitude of vibration increases near resonance, the need for
nonlinear analysis arises. Abdel-Ghaffar (1980, 1982) and Abdel-Ghaffar and Rubin (1983)
presented a general theory and analysis of nonlinear free coupled vertical-torsional vibra-
tions of suspension bridges. More recently, Cevik and Pakdemirli (2005) addressed nonlin-
early coupled vertical-torsional vibrations of suspension bridges employing the method of
multiple scales.
Regarding the forcing conditions, it is possible to analyze the response to pedestrian
walking by representing the walking-induced forces in a Fourier series based on the walk-
ing pace and its integer multiples (Bachmann, 1995). The presence of an arbitrary number
of pedestrians is introduced scaling the force relating to a single pedestrian with the square
root of the total number of pedestrians (Matsumoto et al., 1978). The forced response analy-
sis could be carried out, for the linearized equations of motion, in closed form following
Lacarbonara and Colone (2007) and Vestroni and Vidoli (2007).
When the response analysis predicts vibration levels high enough to make the foot-
bridge uncomfortable to users or even unsafe, vibration mitigation systems become nec-
essary. However, these control architectures should be designed so that they do not alter the
global flexibility characteristics of the footbridge. The use of tuned mass dampers (TMDs)
(Jones et al., 1981), given their versatility, is useful in the design of new structures but also
in the retrofitting of existing structures1 moreover, the maintenance costs are reasonable. Re-
cently, linear viscoelastic TMDs have been installed in the Simone de Beaviour footbridge in
Paris, a rather intentionally flexible bridge, so as to make it react pleasantly to the wind and
pedestrian movement. The literature on vibration absorbers is wide and ranges from classi-
cal viscoelastic TMDs to various kinds of nonlinear absorbers. Law et al. (2004) presented
a vibration mitigation study using frictional dampers on a suspension footbridge. Slotted
bolt connection elements had been used previously to dissipate energy by providing hys-
teretic damping and nonlinear stiffness to the structure although no theoretical background
on its damping mechanism has been presented. Lacarbonara and Vestroni (2002) proposed
the concept of a hysteretic TMD and showed its feasibility in the case of low-rise multi-story
buildings. Poovarodom et al. (2003) pursued the objective of reducing pedestrian-induced
vibrations in a footbridge using tuned liquid dampers with nonlinear quadratic damping.
The optimal absorber parameters are determined by comparing the control performances of
nonlinear multiple TMDs with those of linear multiple TMDs and single TMDs.
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 751

Within the context of linear devices, they have often been employed in multiple configu-
rations in order to increase their robustness1 Xu and Igusa (1992), Yamaguchi and Harnporn-
chai (1993), and Abe and Fujino (1994) showed how the sensitivity to detuning could be
suppressed successfully by placing several absorbers, while their tuning frequencies were
equally spaced around the natural frequencies of the structure. This topic has been ad-
dressed recently by Li and Ni (2007) who developed a parametric gradient-based optimiza-
tion method, removing the constraint of uniformly distributed dampers.
In most cases, the structures to be controlled exhibit a dynamic response prevalently
characterized by one mode. In such cases, the design of a control system based both on sin-
gle TMDs and multiple TMDs can be carried out using a single-degree-of-freedom (SDOF)
model of the structure, and adding n degrees of freedom (DOFs), one for each damper. How-
ever, SDOF models are not suitable for vibration control of structures with closely spaced
natural frequencies (Abe and Igusa, 1995), whenever these are excited. In fact, multiple
TMDs, with a minimum number equal to the number of closely spaced frequencies, are re-
quired to control the vibratory responses successfully.
In this paper, a DP model is presented with the aim of describing accurately planar
motions of suspension bridges. The model is based on a geometrically exact formulation
(Antman, 2005) for both the suspended cables and the bridge deck, incorporating also the
extensibility of the hangers. It is further shown how this nonlinear model is simplified by
enforcing the inextensibility of the hangers in the form of an internal kinematic constraint.
Subsequently, an approximate version of the general nonlinear model is sought under a few
kinematic assumptions (Irvine, 19841 Brownjohn et al., 1994). Moreover, the linearization
of the approximate model allows the natural frequencies of the Singapore Suspension Foot-
bridge to be calculated which are compared with those obtained by a 3D FE model of the
bridge. The analytical model allows the influence of the characteristic nondimensional para-
meters to be investigated on the sequence of the eigenmodes. Thereafter, the forced response
to pedestrian loading is obtained in closed form, modeling each pedestrian as a moving pul-
sating load. High levels of vibration are predicted by the analysis, thus providing an oppor-
tunity to test the effectiveness of several TMD control architectures, from single TMD to
multiple TMD configurations. It is shown how the choice of the most suitable architecture
depends on the fact that the footbridge has its two lowest frequencies below or above their
crossover (when the frequencies intersect transversally).

2. NONLINEAR SUSPENSION BRIDGE MODEL

The reference configuration of the cable by itself is the catenary under the action of gravity,
1 2 3y1x2 2 xe1 4 y1x2e2 3 x 5 [03 4]6 where 1O3e1 3e2 3e3 2 is a fixed right-handed reference
frame. The catenary and the cable tension are given by
1 2 34 2 3
4 5 1 8 1 8
y27 cosh 7 cosh 5 7x 3 N 2 H cosh 5
c c
7x 3 (1)
5 2 2 2

where 4 is the span of the cable, x 8 2 x643 5 :2 7 Ac g46H c is a nondimensional (geomet-


ric flexibility) parameter obtained as solution of the compatibility condition sinh15 622 2
752 N. CARPINETO ET AL.

15 62280 with 80 :2 L c 643 L c is the initial total length of the undeformed cable, H c is the
horizontal projection of the cable tension N c (Lacarbonara et al., 2007). The sag-to-span ra-
tio can be calculated as d :2 y 11622 64 9 1cosh 5 62 7 12 65 . The unit vector tangent to 1
is denoted by ac1 2 cos 9 c e1 4 sin 9 c e2 where cos 9 c 2 11 4 yx2 27162 and tan 9 c 2 yx , where
the subscript x denotes differentiation with respect to x. We also occasionally refer to the
arclength coordinate s along the cable axis in 1.
When the cables and the horizontal bridge deck are linked by the vertical hangers and
allowed to be subject to the dead loads, the ensuing equilibrium state is the pre-stressed
bridge configuration denoted by 10 2 3p0 1x2 2 x1x24 u0 1x23 1b01 1 0 23b02 1 0 223 q0 1x2 2
y1x24 v0 1x23 x 5 [03 4]6 where p0 and q0 are the position vectors of the beam and cable axes,
respectively, and u0 2 u 01 e1 4 u 02 e2 and v0 2 10 e1 4 20 e2 the associated displacement vectors1
1b01 3b02 2 indicates the deck cross-section fixed basis rotated by an angle 0 with respect to
1e1 3e2 2. The deck strains in 10 are the stretch ( 0b ), the shear strain (80b ), and the bending
curvature ( 0b ), respectively, given by (Nayfeh and Pai, 20041 Antman, 20051 Lacarbonara
and Yabuno, 2006)

0
b 2 11 4 u 01x 2 cos 0
4 u 02x sin 0
3 80b 2 711 4 u 01x 2 sin 0
4 u 02x cos 0
3 0
b 2 x
0
(2)

On the other hand, the cable and hanger stretches are


5 5
0
2 cos 9 c 11 4 1x 2
0 2
4 1yx 4 2x 2 3
0 2 0
h 2 16 h 1 0
1 7 u 01 22 4 1h 4 0
2 7 u 02 22 3 (3)

where h1x2 represents the length of the hanger at x. Let n0 1x2 2 N 0 1x2a01 1x2 be the cable
tension in 10 collinear with the unit vector a01 2 cos 9 0 e1 4 sin 9 0 e2 in the cable tangential
direction, and f 0c 2 7 f 0c e2 2 717 A2c sec 9 c e2 be the gravity force per unit reference length
x acting on the cable. Further, let r0 1x2 be the force exerted by the hangers on the bridge deck
per unit reference length x, let t0 1x2 and m0 1x2 be the contact force and contact couple that
two adjoining bridge deck sections mutually exert on each other, and let f 0b 1x2 2 7 f 0b e2
be the force per unit reference length applied on the bridge. The equilibrium equations, in
vectorial form, read

n0x 7 r0 4 f 0c 2 o3 t0x 4 2r0 4 f 0b 2 o3 m0x 4 p0x t0 2 o3 (4)

where the multiplicative factor 2 in front of r0 accounts for the two identically deformed
suspension cables during a planar deformation. We let r0 2 R 0 h0 be the hanger force per
unit reference length in the direction of the displaced hangers, given by the unit vector h0 2
h 01 e1 4h 02 e2 , and t0 2 P 0 b01 4S 0 b02 be the deck contact force decomposed into the deck tension
P 0 and shear force S 0 3 respectively. Thus, the equilibrium equations, in componential form,
become

1N 0 cos 9 0 2x 7 R 0 h 01 2 03 1N 0 sin 9 0 2x 7 R 0 h 02 7 7 Ac g sec 9 c 2 03 (5)

1Px0 7 b S 2 cos
0 0 0
7 1Sx0 4 b P 2 sin
0 0 0
4 2R 0 h 01 2 03 (6)
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 753

1Px0 7 b S 2 sin
0 0 0
4 1Sx0 4 b P 2 cos
0 0 0
4 2R 0 h 02 7 f 0b 2 03
Mx0 4 0 0
bS 7 80b P 0 2 0 (7)

The following geometric relationships hold:


11 4 1x 2
0
1tan 9 c 4 2x 2
0
cos 9 0 2 cos 9 c 0
3 sin 9 0 2 cos 9 c 0
3
0
7 u 01 h4 0
7 u 02
h 01 2 1
3 h 02 2 2
 (8)
h 0h h 0
h

The current cable tension N 0 in 10 can be expressed as N 0 2 N c 4 N where N is the


incremental tension arising in the change of configuration from 1 to 10 . The constitutive
equations are expressed, for the incremental axial load N 3 as N 1x2 2 N 1 0 23 for the deck
tension and shear force, as P 0 1x2 2 P 0 1 0b 2, S 0 1x2 2 S 0 180b 23 and, for the bending moment,
as M 0 1x2 2 M 0 1 0b 2.
By enforcing the inextensibility of the hangers, the reactive hanger tension, can be ob-
tained from (5)2 as R 0 2 [1N 0 sin 9 0 2x 7 7 Ac g sec 9 c ]6 h 02 . We substitute the latter into (5)1 3
(6) and (7) thus obtaining, besides (7)2 that is unchanged, the following modified equilibrium
equations of the suspension bridge in its static configuration:

1N 0 cos 9 0 2x 7 [1N 0 sin 9 0 2x 7 7 Ac g sec 9 c ]h 01 6 h 02 2 03 (9)


1Px0 7 b S 2 cos
0 0 0
7 1Sx0 4 b P 2 sin
0 0 0
4 2[1N 0 sin 9 0 2x 7 7 Ac g sec 9 c ]h 01 6 h 02 2 03
1Px0 7 b S 2 sin
0 0 0
4 1Sx0 4 b P 2 cos
0 0 0
4 2[1N 0 sin 9 0 2x 7 7 Ac g sec 9 c ] 7 f 0b 2 0 (10)

2.1. Equations of Motion

The application of dynamic loads, such as those induced by the passage of pedestrians, causes
a change of configuration from the prestressed configuration 10 to the actual configuration
1. The position vectors of the deck and the cable are p 2 p0 4 u and q 2 q0 4 v where u and v
are the associated incremental displacements, respectively. The deck total strains, the stretch,
the shear strain, and the bending curvature, are

b 2 11 4 u 01x 4 u 1x 2 cos 4 1u 02x 4 u 2x 2 sin 3 (11)

8b 2 711 4 u 01x 4 u 1x 2 sin 4 1u 02x 4 u 2x 2 cos 3 b 2 0


x 4 x3 (12)

where 2 0 4 . On the other hand, the cable total stretch is 2 qs  2 0


where is
the incremental stretch from 10 to 1 given by
5
cos 9 c
2 0
11 4 0
1x 4 1x 2
2 4 1tan 9 c 4 0
2x 4 2x 2 
2 (13)

Further, by letting 9 2 9 0 4 9 denote the angle that the cable axis in 1 makes with e1 , we
obtain
754 N. CARPINETO ET AL.

11 4 0
4 1x 2 1tan 9 c 4 0
4 2x 2
cos 9 2 cos 9 0 1x
3 sin 9 2 cos 9 0 2x
 (14)

The current force exerted by the hangers on the bridge deck per unit reference length x is
denoted by r1x3 t23 the current contact force and contact couple in the bridge are t1x3 t2 and
c b
m1x3 t21 f 1x3 t2 2 f 0c 1x24 f c 1x3 t2 and f 2 f 0b 1x24 f b 1x3 t2 are the current forces per unit
reference length applied on the cable and the bridge, respectively. Then, by enforcing the
balance of linear and angular momentum, we obtain the following equations of motion:
c
nx 7 r 4 f 2 7 Ac sec 9 c vtt 3 (15)
b
tx 4 2r 4 f 2 7 Ab utt 3 Mx 4 bS 7 8b P 2 7 J b tt 3 (16)

where the subscript t indicates differentiation with respect to time and 7 J b is the bridge mass
moment of inertia around e3 .
The present nonlinear model can be simplified by incorporating the assumption that the
hangers are inextensible no matter what the loading conditions are. Under such conditions,
the internal kinematic constraint ie enforced through p 7 q 2 h, and r 2 R1h 1 e1 4 h 2 e2 2
is a reactive force that must be eliminated from the equations of motion. It is obtained from
c
the projection of (15) in the e2 direction as R 2 1nx 4 f 7 7 Ac sec 9 c vtt 2 e2 6h 2 . The reac-
tive force R is substituted into the projection of (15) in the e1 direction and into (16)1 thus
obtaining
c
1 N cos 92x 7 R h 1 4 f  e1 2 7 Ac sec 9 c vtt  e1 3 (17)
b
1 Px 7 b S2 cos 7 1 Sx 4 b P2 sin 4 2 R h 1 4 f  e1 2 7 Ab utt  e1 3
b
1 Px 7 b S2 sin 4 1 Sx 4 b P2 cos 4 2 R h 2 4 f  e2 2 7 Ab utt  e2  (18)

The balance of angular momentum (16) is unchanged.

2.2. Approximate Mechanical Model


We consider small-amplitude purely transverse motions of the bridge deck and moderately
large-amplitude displacements of the cables1 we further consider the hangers to be inex-
tensible. Consequently, the longitudinal motion of the deck is neglected and the kinematic
unknowns reduce to the two displacements, 1u 0 3 0 23 having let 0 2 u 02 9 20 and u 0 2 10 .
The hanger forces are considered to be vertical in the static configuration 10 . Further, in
the balance equations, we let 0b  1 and 80b 2 03 then the moment balance of the deck yields
the shear force as the negative gradient of the bending moment, S 0 2 7Mx0 3 while the two
combined balance equations for the cable and the deck are

1N 0 cos 9 0 2x 2 03 21N 0 sin 9 0 2x 4 Sx0 2 f 0b 4 2 f 0c 3 (19)

The first equation implies H 0 2 N 0 cos 9 0 2 const, therefore, the only governing equilib-
rium equation becomes

Mx0x 7 2H 0 1tan 9 0 2x 2 7 f 0b 7 2 f 0c 3 (20)


MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 755

where tan 9 0 2 1yx 4 x 2611


0
4 u 0x 2 and
6 4 6 4
Hc 11 4 u 0x 2 1 11 4 u 0x 21 0
7 12
H0 2 dx 4 E Ac cos 9 c dx (21)
4 0
0 4 0
0

Under the assumption of small longitudinal cable displacement gradients, u 0x   13 and


0
 13 then H 0 2 H c 4 H where H 2 N cos 9 0 2 N cos 9 c 114u 0x 26 0  N cos 9 c . The
constitutive law for N is N 2 N 2 E Ac 1 0 712. The second-order Taylor expansion of the
7 8162
stretch is 0 2 q0s  2 1 4 1u 0s 22 4 1 s0 22 4 2xs u 0s 4 2ys s0  141621 s0 22 4 xs u 0s 4 ys s0 3
having neglected the second-order horizontal displacement gradient 1u 0s 22 . Then, by using
d6ds 2 cos 9 c d6dx3 the leading part of the elongation is 0 7 1 2 cos2 9 c [u 0x 4 yx x0 4
1
1 0 22 ]. Consequently, H 61E A2c sec3 9 c 2 u 0x 4 yx x0 4 12 1 x0 22 . Integrating both members
2 x
and enforcing the boundary conditions (u 0 102 2 u 0 142 2 0) yields
6 41 4 6 4
E Ac 1 0 2
H 2 e yx x0 4 1 x 2 dx3 L 2 e
sec3 9 c dx (22)
L 0 2 0

Further, by considering the cable equilibrium equation in 1, H c yx x 2 f 0c 3 and the constitu-


tive equation for the bending moment in the bridge deck, M 0 2 E J b x0x 3 and by substituting
the latter with (22) into (20), we obtain the following equilibrium equation under the dead
load f 0b :
c 6 41 4
0 EA 1
E Jb 0
xxxx 7 2H c 0
xx 7 21yx x 4 xx 2 yx 0
x 4 1 x0 22 dx 2 7 f 0b  (23)
Le 0 2

On the other hand, the equations of motion, in componential form, can be obtained from
(17) and (18) by letting n 2 N a1 with a1 2 cos 9e1 4 sin 9e2 3 t 2 Se2 , and by neglecting the
bridge rotatory inertia. The balance of angular momentum becomes Mx 4 S 2 0 that, solved
for the shear force S 2 7 Mx and substituted into the balance of vertical linear momentum,
furnishes

1 N cos 92x 2 17 Ac sec 9 c 2u tt 3 (24)

21 N sin 92x 7 Mx x 7 2 f c 7 f b 2 127 Ac sec 9 c 4 7 Ab 2 tt  (25)

Since the dynamic loads are prevalently vertical, it is reasonable to expect a negligible lon-
gitudinal acceleration of the cable, u tt  0. Hence, the first balance equation implies that
H 2 N cos 9 2 const, that, substituted in turn into the second balance equation, yields

2 H 1tan 92x 7 Mx x 7 2 f c 7 f b 2 127 Ac sec 9 c 4 7 Ab 2 tt  (26)

By further assuming the smallness of the gradient of the dynamic horizontal cable displace-
ment, u x   1, we obtain tan 9 2 1yx 4 x0 4 x 2611 4 u 0x 4 u x 2  yx 4 x0 4 x .
Moreover, cos 9 2 cos 9 0 11 4 u 0x 4 u x 26  cos 9 0 putting  1. Therefore, H 2
N cos 9 2 1N 4 N 2 cos 9  1N 4 N 2 cos 9 2 H 4 H where H 2 N cos 9 0 is the
0 0 0 0
756 N. CARPINETO ET AL.

horizontal projection of the dynamic incremental load N 2 E Ac 1 7 12 and indicates


the incremental dynamic cable stretch (13). By following steps similar to those undertaken
for the static equilibrium, we first obtain the expansion of the incremental elongation as
7 1  1cos 9 c 6 0 22 [u x 4 1yx 4 x0 2x x 4 12 x2 ]. The cable horizontal incremental load is then
expressed as
6 2 3 6 4
E Ac 4 1
H2 d 1yx 4 x 2 x 4 1 x 2 dx3 L 2
0 2 d
1sec 9 0 21sec2 9 c 21 0 22 dx (27)
L 0 2 0

Further, by considering the equilibrium in 10 and the constitutive law in bending, M 2


M 0 4 E J b x x 3 and by substituting the latter with (27) into (26), we obtain

127 Ac sec 9 c 4 7 Ab 2 4 E J b x x x x 7 2H 0 x x
tt
6 1 4
E Ac 4 1
7 21yx x 4 0
xx 4 xx 2 1yx 4 x 2 x 4 1 x 2 dx 2 7 12 f c 4 f b 23
0 2
(28)
Ld 0 2

where f c 1x3 t2 and f b 1x3 t2 are the incremental dynamic loads applied on the cable and the
bridge deck, respectively.
In view of the subsequent analysis, the problem is suitably nondimensionalized. The
space coordinate x and the displacements are made nondimensional with respect to the bridge
span 4 and time t with respect to the characteristic time tb 2 16b 2 17 Ab 44 6E J b 2162 . The
static and dynamic nondimensional balance laws become
6 12 3
1 0 2
x x x x 7  x x 7 1yx x 4 x x 2
0 c 0 0
yx x 4 1 x 2 dx 2 7 f 0b 3
0
(29)
0 2
11 4 72 tt 4 7 0 x x
xxxx
6 11 4
1
7 1yx x 4 0
xx 4 xx 2 1yx 4 x2 x
0
4 1 x 2 dx 2 7 12 f c 4 f b 23
2
(30)
0 2

where L d 2 L e .
The resulting dimensionless parameters turn out to be: 80 , the initial cable length-to-span
ratio1 5 , the geometric flexibility parameter in the catenary equilibrium of the cable by itself1
the geometric-to-elastic bending stiffness ratios  c 2 2H c 42 6E J b and  0 2 2H 0 42 6E J b 1
the elastic axial-to-bending stiffness ratio  2 21E Ac 6E J b 2143 6L e 2 the mass ratio 7 2
217 Ac 2617 Ab 2. In addition, the nondimensional vertical dead load applied on the deck, f 0b 2
pE J b 643 3 acts as a nondimensional parameter for the subsequent elastodynamic problem in
that it mostly affects the final prestress level in the bridge. Given the geometric compatibility
condition 80 2 1265 2 sinh15 6223 along with  0 2  c 4  (the expression of the incremental
geometric stiffness ratio is given in the Appendix by Equation (42)), the overall independent
bridge parameters are five1 namely, 15 ,  c , , f 0b , 7).
The eigenvalue problem for the frequencies and mode shapes is obtained by linearizing
(30) and dropping the forcing terms. The resulting linear boundary-value problem is
6 1
7 8
11 4 72 tt 4 x x x x 7  x x 7 1yx x 4 x x 2
0 0
1yx 4 x0 2 x dx 2 03 (31)
0
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 757

supplemented with the kinematic and mechanical boundary conditions given by 103 t2 2
113 t2 2 0 and x x 103 t2 2 x x 113 t2 2 0.

3. COMPUTATIONAL APPROACH AND VALIDATION

A Galerkin discretization is employed to solve Equations (29) and (30). By introducing


7 8
the vector of generalized coordinates, q1t2 2 q1 1t23 q2 1t23    3 q M 1t2 , the vertical dis-
placement field can be expressed as 1x3 t2 2 1 1x2q1t23 where 11x2 is the vector of trial
functions, here sinusoidal (so that they satisfy both the kinematic and mechanical bound-
ary conditions). Thus 11x2 2 [sin1 x23 sin12 x23    3 sin1M x2] 3 where  denotes the
transpose. Here M is the number of discretizing functions, taken equal to 20. After substi-
tuting the discretized displacement field 1x3 t2 2 1 1x2q1t2 into the equation of motion
(31), the minimization of the residuals, according to the Galerkin method, is performed pre-
multiplying the residuals by 1 and integrating the result over the problem domain. The
following vector-valued second-order differential equation governing the bridge free oscilla-
tions is obtained:
9 9
Mb 4 Mc q 4 KbE 4 KcE 4 KcG q 2 03 (32)

where Mb and Mc are the deck and cable mass matrices, respectively, given by
6 1 6 1
Mb 2 11x21 1x2dx3 Mc 2 11x217 sec 921 1x2 dx (33)
0 0

On the other hand, the deck and cable elastic stiffness matrices, KbE and KcE 3 and the cable
geometric stiffness matrix, KcG 3 are expressed as
6 1
K bE
2 1x x 1x21x x 1x2 dx3 (34)
0
26 1 3 26 1 3
K cE
2 11x2q20 x x 1x2 dx q20 x 1x21x 1x2 dx 3 (35)
0 0
6 1
K cG
2 7 0
11x21x x 1x2 dx3 (36)
0

where q20 1x2 2 y1x2 4 0 1x2. The bridge natural frequencies and mode shapes are found by
solving the eigenvalue problem associated with Equation (32).
In this study, the Singapore Footbridge is taken as an exemplary footbridge. A planar
model is justified since the lateral and torsional modes can be neglected herein. In fact, from
the experimental data reported by Brownjohn (1997), the frequency of the first lateral bend-
ing mode (1.250 Hz) is not within the range of the pedestrian-induced lateral frequencies
[0.6–1.2] Hz1 moreover, the mode is pointed out as being heavily damped, so that the effects
on the footbridge response are not expected to be appreciable. The contributions to torsional
758 N. CARPINETO ET AL.

Table 1. Natural frequencies in Hertz of the Singapore Suspension Footbridge: finite-


element results, experimental data, and closed-form results.
Mode FEM (Hz) Experimental DP
Lateral symmetric 1.446 1.250 —
Torsional skew-symmetric 1.652 1.840 —
Vertical skew-symmetric 1.911 2.072 2.067
Vertical symmetric 2.143 2.151 2.157
Torsional symmetric 2.664 2.520 —
Lateral skew-symmetric 3.632 — —
Torsional skew-symmetric 4.242 — —
Vertical symmetric 4.311 4.288 4.668

Figure 1. Schematic geometry of the Singapore Suspension Footbridge.

modes due to eccentric moving loads on the bridge deck are not further considered, since the
pedestrian-induced torsional couples are not important, given the limited width of the walk-
way section. All of these statements are corroborated by a 3D FE model of the footbridge,
useful to investigate the spacing and sequence of modes (see Table 1), hence, their poten-
tial participation to the bridge dynamic response. Since the pedestrian-induced loading is
assumed to be resonant with the natural frequencies of the vertical bending modes, the high-
est reduction of the ensuing vertical vibrations is sought by a collocation of multiple TMDs
along the deck centerline. For a more general loading condition involving bending-torsional
vibrations, these could be effectively controlled by additional TMDs, symmetrically collo-
cated with respect to the deck centerline. Clearly, in this scenario, a more general mechanical
model, including the bending-torsional motions, should be constructed.
The lowest six mode shapes of the Singapore Footbridge are shown in Figure 3. The
associated frequencies are in close agreement with the experimental measurements docu-
mented in Brownjohn (1997), as shown in Table 1. It is worth observing the close spacing
of the lowest two in-plane frequencies: the fundamental mode is skew-symmetric and the
second mode is symmetric without nodes. Hence, both modes are likely to contribute to the
response of the footbridge, potentially giving rise to a beating phenomenon.
To better understand the dynamic behavior of such a structure, Figures 4(a) and (b) show
the variation of the frequencies of the lowest four modes with the axial-to-bending stiffness
ratio  and the sag-to-span ratio d 0 , respectively. In the first figure, once  c , f 0b and 7 are
set to the values characterizing the Singapore Footbridge, only the parameter  is varied. At
each step, the sag-to-span ratio in the static configuration 10 is given by d 0 2 d c 4 0 11622,
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 759

Figure 2. Planar model of the suspension footbridge: the dashed lines indicate the undeformed geometry
1, the solid thin lines represent the static deformed geometry 10 , and the solid thick lines denote the
actual geometry 1.

Figure 3. Lowest six mode shapes of the Singapore Footbridge: (a) first skew-symmetric mode
(2.067 Hz)1 (b) first symmetric mode (2.157 Hz)1 (c) second symmetric mode (4.668 Hz)1 (d) second
skew-symmetric mode (8.106 Hz)1 (e) third symmetric mode (12.645 Hz)1 (f) third skew-symmetric mode
(18.171 Hz).

where 0 1x2 is the solution of the static nonlinear problem (given by Equation (45) in the
Appendix) and d 0 is fixed to the sag-to-span ratio of the Singapore Footbridge, namely 0.1571
this constraint on d 0 allows us to determine the remaining parameter 5 . According to the
geometric and mechanical parameters, a realistic range of values for  is [2 102 –104 ]: a
crossover occurs when the axial-to-bending stiffness ratio is slightly below 103 , quite close to
760 N. CARPINETO ET AL.

Figure 4. Variation of the frequencies with (a) the cable elastic stiffness ratio  (in logarithmic scale) and
(b) the static sag-to-span ratio d 0 . Footbridge 1 (dashed line), Singapore Footbridge (dotted line), and
Footbridge 3 (dotted-dashed line).

the value characterizing the Singapore Footbridge, namely 134 103 . In the second figure,
the sag-to-span ratio d 0 is varied assuming constant the independent parameters (,  c , f 0b ,
7) and determining at each step 5 .
The presence of two natural frequencies, that are so closely spaced as in the Singapore
Footbridge, is not a common occurrence. To achieve more general conclusions, from the
point of view of the vibration analysis and control design procedures, different values of
the geometrical and mechanical parameters are taken into account, thus leading to a dif-
ferent footbridge referred to as Footbridge 1, whose lowest two frequencies are far from
the crossover point, yet maintaining the same modal sequence of the Singapore Footbridge.
In contrast, Footbridge 3 is obtained decreasing d 0 from 0.157 (the reference sag-to-span
ratio of the Singapore Footbridge) to 0.114, so that its lowest two frequencies are below
the crossover, hence, the fundamental mode is symmetric and the second is the first skew-
symmetric mode.

4. SCENARIOS OF PEDESTRIAN-INDUCED VIBRATORY RESPONSE

The dynamics of pedestrian walking, both in its vertical and lateral components, are widely
described in the literature, although there is a lack of thorough experimental characteriza-
tions. It is shown that the frequency content of the forces induced by each pedestrian is
within the range [1.2–2.4] Hz with regards to the vertical direction, while the frequencies of
the lateral forces are halved, that is, they fall within the range [0.6–1.2] Hz. For the investi-
gated case, the fact that the lateral natural frequencies do not fall within the mentioned range
made it possible to analyze the control problem of the vertical modes only, without loss of
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 761

generality. The structural damping, due to the dissipation in the constraints and the viscoelas-
tic material properties, is assumed to be of the proportional type1 further, the incorporation
of the external forces into the equations of motion furnishes

Mq 4 Cq 4 Kq 2 Q1t23 (37)

where C 2 Cb 4 Cc is the system damping matrix, direct summation of the deck and cable
damping matrices, respectively1 Q1t2 2 Qb 1t24Qc 1t2 is the vector of generalized forces. The
introduction of the modal transformation q1t2 2 231t2 in Equation (37), where 2 denotes
the modal matrix collecting the eigenvectors as column vectors, yields

 4 D31t2
31t2  4 431t2 2 P1t23 (38)

where D 2 diag32 j  j 6, 42 diag32j 6, for j 2 13    3 M, and P1t2 2 2 Q1t2 is the modal


force vector.
By approximating the pedestrian forces as pulsating loads, moving with constant speed
along the bridge deck, the force of a single pedestrian becomes F1x3 t2 2 F1t21x 7 p t2.
The pedestrian speed p depends on the walking frequency f p , that is, p 2 09 f p [m s72 ]
where f p is the pace frequency in Hertz. Expressing the amplitude F1t2 in a Fourier series
and retaining the first harmonic only, according to Zivanovic et al. (2005), yields F1t2 2
P[1 4  1 sin12 f p t2] 2 736 4 180 sin12 f p t2 (in Newtons). The presence of a crowd
walking along the bridge deck can be represented by summing up all of the pedestrian forces,
each characterized, in practice, by different speeds and spacing between the pedestrians. For
an arbitrary number of pedestrians, say N p 3 by considering the same average force for all
pedestrians, the overall pedestrian-induced force distribution can be simply expressed as

Np

F1x3 t2 2 F1t2 1x 7 pt 7  p 23 (39)


p21

where  p indicates the nondimensional distance from the leading pedestrian. By consider-
ing the kth modal force associated with the pth pedestrian, it turns out that it carries two
frequencies, namely, f p  k p 62. Hence, the pth single pedestrian induces a resonance in
the kth mode when the generated frequencies are close to the natural frequency of the kth
mode. Further, all of the pedestrians walking in the same direction generate a stronger reso-
nance when the assumed relative constant spacing,  p 3 is such that the ratio p 6 p is close
to the frequency of one of the bridge modes. In the following analysis, by assuming that the
pedestrians are constantly spaced 2 m apart and walk with the same speed, the considered
critical walking frequencies are f p 2 f 1 and f p 2 1 f 1 4 f 2 262, for all p 2 13    3 N p . The
first frequency is suggested by most of the codes1 the second critical frequency is selected
because the Singapore Footbridge exhibits two very closely spaced natural frequencies, with
the direct consequence that they can both be excited by the pedestrian walking.
According to the proposed modeling of pedestrian-induced loading, the passage of a
group of pedestrians can be effectively analyzed through the consideration of four different
loading phases. In the first phase, the pedestrians walk along the bridge giving their con-
tribution to both the first skew-symmetric and symmetric modal forces, until the mid-span
762 N. CARPINETO ET AL.

Figure 5. Time histories of the dimensional vertical acceleration recorded at the quarter-span and
mid-span sections of the Singapore Footbridge: the case of five pedestrians moving in the same
direction with f p 2 f 1 (top), f p 2 1 f 1 4 f 2 262 (middle), and f p 2 f 2 (bottom).

section is reached by the front-line pedestrian at time t1m 2 462 p . Once the front-line
pedestrian crosses the mid-span section its action turns out to be opposite to that exerted by
the remaining pedestrians and, similarly, this occurs for each pedestrian crossing the mid-
span of the bridge. The first skew-symmetric modal force decreases while the symmetric
modal force continues to increase, reaching the highest value when the center of the group
is at the mid-span of the bridge. The second phase ends when the last pedestrian reaches the
mid-span section, namely at time t Nmp 2 t1m 4 l N p 6 p , where l N p is the distance of the last
pedestrian from the front-line pedestrian. In the third stage, while the symmetric modal force
decreases, the skew-symmetric force increases once again until both modal forces vanish at
t Nf p 2 14 4 l N p 26 p . The last stage is characterized by free oscillations of the bridge.
For illustrative purposes, in Figure 5 we show the vertical acceleration at the quarter-
span (left) and the mid-span (right) sections of the Singapore Footbridge resulting from a
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 763

group of five pedestrians moving in the same direction with f p 2 f 1 (top), f p 2 1 f 1 4 f 2 262
(middle) and f p 2 f 2 (bottom). The vertical dashed lines indicate the mentioned times t1m ,
t5m 3 and t5f 3 respectively. It is worth noting that the highest response peak at the quarter-span
section is attained when the group of pedestrians is across the second half of the bridge, since
the structure is already vibrating for their passage in the first half.
The comparison of the footbridge response to the passage of pedestrians in different
configurations allows us to determine the most critical pedestrian-induced loading scenar-
ios for the structure. Here we considered the two critical walking frequencies for the three
representative bridges under various numbers of pedestrians walking in different orders and
directions. It is expected that a symmetric excitation (i.e. two pedestrians walking across
the footbridge in opposite directions) does not possess any modal projection onto the skew-
symmetric modes, thus giving its greatest contribution only to the vertical acceleration at
the mid-span section. On the other hand, two pedestrians walking across the footbridge in
the same direction may drive both the symmetric and skew-symmetric modes to resonance.
In this scenario, when the walking frequencies are taken equal to the average value of the
first two natural frequencies, the acceleration peaks at the mid- and quarter-span sections are
similar (Figures 6(b), (d), and (f)). Moreover, we observe the occurrence of the beating phe-
nomenon arising from the interaction of these frequencies. For structures with more widely
spaced natural frequencies such as Footbridges 1 and 3, the response is expected to be highly
amplified by the resonance when the walking frequency is equal to the frequency of the
lowest mode with the highest acceleration peak where the involved modal shape exhibits its
maximum (Figures 6(a)–(e)). The situation of Footbridge 3 is different since its lowest mode,
in contrast to the other bridges, is symmetric. Hence, the transit of pedestrians in opposite
directions is here the most severe loading scenario, since the resulting pedestrian-induced
force possesses projection onto the symmetric mode only.
In conclusion, it is clear that the vibration amplitudes induced at the quarter-span by the
passage of a group of pedestrians can be reduced by even a single pedestrian walking in the
opposite direction. This means that the worst loading scenario relates to the case when all of
the pedestrians walk in the same direction and are spaced by a fixed distance of 2 m. Under
these conditions, the vertical accelerations of the bridge deck can go well beyond the comfort
limit state prescribed by British Standards Institution (1978).

5. DESIGN OF MULTIPLE VISCOELASTIC TUNED MASS DAMPERS

The objective of minimizing the vibration levels, so as to ensure comfort to the pedestrians,
while maintaining the bridge global flexibility, has been pursued by installing viscoelastic
TMDs, which are tuned with the appropriate frequencies and are suitably collocated. Both
single and multiple TMDs have been investigated. In particular, the optimal design has been
performed analyzing the peaks of the transfer function of a prescribed harmonic base dis-
placement, with frequency , and defined as

V 1x3 2 2 1 1x2G12Fa 12 (40)


1
where Fa 12 2 716b 22 11 4 2 0 11 7 x211x2dx G12 2 172 M  4 iC  4 K2
 71  i
 C
is the imaginary unit1 M,  and K
 are the modified mass, damping, and stiffness matrices
764 N. CARPINETO ET AL.

Figure 6. Pedestrian loading with f p 2 f 1 (left) and f p 2 1 f 1 4 f 2 262 (right). Vertical acceleration for
different loading scenarios: (—) one pedestrian, (— —) two pedestrians in the same direction, (- - -) two
pedestrians in opposite directions, (  ) three pedestrians, two in the same direction, (—   —) three
pedestrians. (a),(b) Footbridge 11 (c),(d) Singapore Footbridge1 (e),(f) Footbridge 3.

of the structure, respectively. The motivation for employing the transfer function of the base
motion is in the fact that the TMD design and performance have been dealt with considering
pedestrian loading and earthquake excitations although the results documented in this work
are related to the pedestrian loading only. Since the goal consists of controlling also structures
with closely spaced natural frequencies, such as the Singapore Footbridge, a SDOF model is
not suitable to this end. With respect to the DP model proposed previously, the mechanical
formulation is modified to account for multiple TMDs.
The schematic geometry of the control architectures with multiple TMDs is portrayed
in Figure 7. The design values for each TMD, the absorber stiffness and damping ratio,
according to Den Hartog (1934) are a 2 m a 6m b , k a 2 m a 12 f 22 611 4 a 22 ,  a 2
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 765

Figure 7. Schematic geometry of the multiple TMD architecture.

f [15 a 611 4 a 23 ]162 for a given mass ratio a 2 m a 6m b , where f is the frequency of the
mode to control, m a is the mass of the absorber, and m b is the modal mass of the structure.
These optimal values do not account for the presence of damping in the primary structure.
For the optimal parameters in the case of damped primary structure, there are several semi-
empirical expressions which are related to specific excitation scenarios. However, they do
not account for the presence of multiple TMDs, for which a thorough ad hoc optimization
procedure should be adopted.

5.1. Control Architectures Away and Near Crossovers

Retrofitting by means of multiple TMDs has been applied to three different structures, in-
cluding the Singapore Footbridge, whose first two natural frequencies are close to the walk-
ing frequencies (2 Hz) and the acceleration levels are remarkably high. This bridge is well
known as being a lively structure and because of this it has become a tourist attraction. In
the following we consider different control architectures: the main criteria for the proposed
design is the collocation of the absorbers at the stationary points of the modal shapes whose
frequency the absorbers are tuned with. In all of the comparisons, for both the single and
multiple TMD configurations, the total mass of the absorbers is the same, 1650 m b . We con-
sider five control architectures, summarized in Table 3 with one, two, and three TMDs. In
particular, architectures 1 and 2 comprise a single TMD either at x 2 164 tuned with the low-
est skew-symmetric frequency and at x 2 162 tuned with the lowest symmetric frequency.
Architectures 3 and 4 are based on two absorbers. In the first case, they are positioned at
x 2 164 or at x 2 162 and are tuned, respectively, to the skew-symmetric and symmet-
ric modes. In the second case, (architecture 4), the absorbers are symmetrically placed at
x 2 164 and at x 2 364, both tuned to the lowest skew-symmetric mode. Finally, archi-
tecture 5 comprises three absorbers at x 2 164 and x 2 364, tuned to the lowest skew-
symmetric mode, and at x 2 162, tuned to the lowest symmetric mode.
In Figure 8 we compare the frequency-response functions of the uncontrolled and con-
trolled footbridges: architectures 1 (thick dashed line) and 4 (thick solid line) are expected to
give the same results in vibration control. It can be observed that this occurs for the structure
whose first two frequencies are not so closely spaced. In Figure 9, the comparison between
architectures 1 (thick dashed line) and 2 (thick dotted line) shows that a properly tuned
766 N. CARPINETO ET AL.

Table 2. Natural frequencies in Hertz of the in-plane modes of three bridges.


Mode 1 2 3 4 5
Footbridge 1 2.0682 2.5713 4.7273 8.1072 12.6503
Singapore Footbridge 2.0674 2.1566 4.6680 8.1064 12.6447
Footbridge 3 1.6325 2.0843 4.6375 8.1283 12.6648

Table 3. Parameters of the various TMDs architectures.


TMDs Number of Tuning
architecture absorbers Position (x) frequencies Mass ratio
1 1 1/4 f skw 1/50
2 1 1/2 f sym 1/50
3 2 1/4, 1/2 f skw , f sym 1/75, 1/150
4 2 1/4, 3/4 f skw , f skw 1/100, 1/100
5 3 1/4, 1/2, 3/4 f skw , f sym , f skw 1/150, 1/150, 1/150

absorber can successfully reduce the peak response of the deck at the section where it is
positioned. However, to obtain the best control authority, a multiple TMD architecture is
necessary, with a number of TMDs at least equal to the number of closely spaced frequen-
cies. In the same figure, note that architecture 5 (thick solid line) exhibits the best perfor-
mance in vibration suppression, both at quarter-span and mid-span sections. A comparison
between 2-TMD and 3-TMD architectures is finally proposed in Figure 10 which shows that
a symmetric configuration with three TMDs gives better results than the architecture with
two TMDs below the frequency crossover.
Figure 11 shows a comparison between the time histories of the vertical accelerations
of the Singapore Footbridge in the uncontrolled and controlled configurations. In both cases
five pedestrians are assumed to walk with a frequency f p 2 1 f 1 4 f 2 262. A multiple TMD
architecture with three absorbers symmetrically placed at each quarter-length of the bridge
allows a suppression of the vertical accelerations, whose highest peaks at the quarter-span
and mid-span sections are reduced by 80% and 75%, respectively. As to the quarter-span sec-
tion, it is already shown that the highest acceleration peak in the uncontrolled configuration
is attained when the pedestrians occupy the second half of the deck. On the other hand, in the
controlled configuration, this peak is reduced because of the higher damping in the structure,
so that the highest value is reached when the pedestrians occupy the first half of the bridge.
Finally, free oscillations of the bridge deck for t  t 5f are largely reduced. As to the mid-span
section of the bridge, when the last pedestrian leaves the deck, the vibration amplitudes are
almost entirely reduced to a negligible level, whereas at the quarter-span section they die out
after a few cycles.
The comparison of the behaviors of the Singapore Footbridge, both in the uncontrolled
and controlled configurations, resulting from the passage of an increasing number of pedes-
trians, walking in the same direction at a distance of 2 m with f p 2 f 1 , f p 2 1 f 1 4 f 2 262,
and f p 2 f 2 are shown in Figures 12 and 13. The accelerations at the quarter- and mid-
span sections of the deck, in the controlled configuration, become at most one-quarter of
the accelerations exhibited in the uncontrolled case. Similarly, Figures 14 and 15 portray the
results for Footbridges 1 and 3, respectively. It can be seen that there is an overall reduction
of acceleration peak, although in some cases there is a deterioration of control effectiveness,
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 767

Figure 8. Frequency-response functions comparison: without TMDs (thin line), architecture 1 (thick
dashed line), and architecture 4 (thick solid line). (a),(b) Footbridge 11 (c),(d) Singapore Footbridge1
(e),(f) Footbridge 3.
768 N. CARPINETO ET AL.

Figure 9. Frequency-response functions comparison: without TMDs (thin line), architecture 1 (thick
dashed line), architecture 2 (dotted line), and architecture 5 (thick solid line). (a),(b) Footbridge 11 (c),(d)
Singapore Footbridge1 (e),(f) Footbridge 3.
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 769

Figure 10. Frequency-response functions comparison: without TMDs (thin line), architecture 3 (thick
dashed line), and architecture 5 (thick solid line). (a),(b) Footbridge 11 (c),(d) Singapore Footbridge1
(e),(f) Footbridge 3.
770 N. CARPINETO ET AL.

Figure 11. Time histories of the dimensional vertical acceleration at the quarter- and mid-span sections of
the Singapore Footbridge in the uncontrolled (thin line) and 3-TMD controlled (thick line) configurations:
the case of five pedestrians moving in the same direction with f p 2 1 f 1 4 f 2 262.

Figure 12. Singapore Footbridge: vertical displacement in the uncontrolled (thin line) and controlled
(thick line) configurations observed at x 2 164 and x 2 162: (a),(b) f p 2 f 1 1 (c),(d) f p 2 1 f 1 4 f 2 2621
(e),(f) f p 2 f 2 . The TMD architecture is that labeled 5 in Table 3.
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 771

Figure 13. Singapore Footbridge: vertical acceleration in the uncontrolled (thin line) and controlled (thick
line) configurations observed at x 2 164 and x 2 162: (a),(b) f p 2 f 1 1 (c),(d) f p 2 1 f 1 4 f 2 2621 (e),(f)
f p 2 f 2 . The TMD architecture is that labeled 5 in Table 3.

when the walking frequency is away from the tuning frequency (see Figures 14(b), 13(c),
15(d), and 15(f)), whereby the structural response is of small amplitude anyway.

6. CONCLUSIONS

Flexural vibrations induced in lightly damped suspension footbridges by walking pedestrians


may be greatly amplified near resonance. It is shown that the highest vibrational levels are
attained at the quarter-span of the bridge when all pedestrians walk in the same direction with
a walking frequency close to the frequencies of the lowest modes and when they are as many
772 N. CARPINETO ET AL.

Figure 14. Footbridge 1: vertical acceleration in the uncontrolled (thin line) and controlled (thick line)
configurations observed at x 2 164 and x 2 162: (a),(b) f p 2 f 1 1 (c),(d) f p 2 1 f 1 4 f 2 2621 (e),(f)
f p 2 f 2 . The TMD architecture is that labeled 5 in Table 3.

as to fill half of the bridge span. As to the mid-span section, the highest vibratory response
peak is obtained when the number of pedestrians is such to fill the whole bridge span with
a given spacing. Although the pedestrians walking frequency, their weight, spacing and dis-
tribution are random in nature, here the modeling has been conducted within the framework
of reasonable deterministic assumptions. Under the discussed severe loading scenarios, the
accelerations can go well beyond the comfort limit state prescribed by the established design
codes and the ensuing bridge dynamic response may become fairly nonlinear. This has moti-
vated the development of a fully nonlinear mechanical model for the planar dynamics of the
suspension bridge. The nonlinear geometrically exact model, under suitable hypotheses, has
been proved to lead to a reduced mechanical model in agreement with existing theories.
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 773

Figure 15. Footbridge 3: vertical acceleration in the uncontrolled (thin line) and controlled (thick line)
configurations observed at x 2 164 and x 2 162: (a),(b) f p 2 f 1 1 (c),(d) f p 2 1 f 1 4 f 2 2621 (e),(f)
f p 2 f 2 . The TMD architecture is that labeled 5 in Table 3.

The overall vibrations and forces can be successfully reduced by installing single or mul-
tiple TMDs. Given the significant reductions in oscillation amplitudes, the linearized model
associated with the nonlinear reduced model has been employed. The use of the distributed-
parameter model is suitable for control design, in that it allows us to describe the behavior
of these structures with closely spaced natural frequencies thus overcoming the limitation of
SDOF representations.
In the most general configurations, the use of the transfer functions of the modified
system, joined with some suitable optimization schemes, can yield the optimal tuning and
collocation conditions. For the single TMD, the tuning frequency and collocation depend
on the selected mode to be controlled. The best collocation is where the peak of the modal
774 N. CARPINETO ET AL.

response is attained. On the other hand, suspension footbridges are likely to have the lowest
two frequencies closely spaced and, differently from beam-like structures, more than one
mode could be involved in the vibrational response. In such cases, multiple dampers are
necessary, one for each mode to be controlled. Some heuristic rationale can be developed,
however, no general conclusions can be drawn a priori on the collocation of the multiple
TMDs. The main issue is the interaction between the TMDs.
In the present case study, bridges with two widely-to-closely spaced natural frequencies
are addressed. It is shown that, by collocating two TMDs at the quarter- and mid-span sec-
tions and tuning them to the skew-symmetric and symmetric modes, respectively, the best
control performance is obtained. A configuration with three devices symmetrically placed
at each quarter-length of the bridge is the best choice although detuning phenomena may
modify the response of the footbridge. The effectiveness of the control system in damping
vibrations and enhancing the user comfort is demonstrated. Moreover, the proposed architec-
ture certainly increases the in-service lifetime of these structures, since the fatigue behavior
can be significantly improved by reducing the response amplitude during the forced phase
and both the amplitude and the number of cycles in free oscillations.
Acknowledgment. This work was partially supported under an FY 2005-2006 PRIN Grant (Research Projects of Na-
tional Interest).

APPENDIX
A closed-form solution of the footbridge equilibrium solutions in 10 is obtained. By letting
 0 2  c 4  , the simplified equilibrium equation becomes

0
xxxx 7 0 0
xx2  yx x 7 f 0b 3 (41)
16 1 6 4
1 1 
0 2
 2 yx x dx 4
0
x dx  (42)
0 2 0

The general solution of Equation (41) is the summation of the solution of the associated
homogeneous equation and the particular solution1 namely,
9 9
0
h 2 c1 4 c2 x 4 c3 cosh  0 x 4 c4 sinh 0 x 3 (43)
1 2 34
1
0
p 2 b1 cosh 5 7x 4 b2 x 4 b3 x 2 3 (44)
2

where

 f 0b
b1 2 3 b2 2 03 b3 2 
5 15 2 4  0 2 2 0

By enforcing the boundary conditions, 0


102 2 0
112 2 0, x x 102
0
2 x x 112
0
2 0, the
constants (c1 3 c2 3 c3 3 c4 ) are obtained as
MITIGATION OF PEDESTRIAN-INDUCED VIBRATIONS 775

f 0b  cosh 15 622 f 0b
c1 2 4 3 c2 2 7 3
1 2
0 2 05 2 0
 
f 0b  5 cosh 15 622 f 0b  5 cosh 15 622 9
c3 2 7 0 2 4 0 0  3 c4 2 7 4  tanh  0 62 
1 2   4 52 1 0 22 0 0 4 5 2

The closed-form static equilibrium of the bridge is thus given by

 cosh 15 622
f 0b f 0b f 0b 2
0
1x2 2 4 7 x 4 x
1 0 22 0 5 2 0 2 0
1 2 34  0b 
9
 1 f  5 cosh 15 622
4 3 cosh 5 7 x 4 4  cosh 0 x
5 4 0 5 2 1 0 22 0 0 4 5 2
    
f 0b  5 cosh 15 6222 0 9
7 7  tanh sinh 0 x  (45)
1 0 22 0 0 4 5 2 2

REFERENCES
Abdel-Ghaffar, A. M., 1980, “Vertical vibration analysis of suspension bridges,” ASCE Journal of Structural Division
106, 2053–2075.
Abdel-Ghaffar, A. M., 1982, “Suspension bridge vibration: continuum formulation,” ASCE Journal of Engineering
Mechanics 108, 1215–1232.
Abdel-Ghaffar, A. M. and Rubin, L. I., 1983, “Nonlinear free vibrations of suspension bridges: theory,” ASCE
Journal of Engineering Mechanics 109, 313–345.
Abe, M. and Fujino, Y., 1994, “Dynamic characterization of multiple tuned mass dampers and some design formu-
las,” Earthquake Engineering and Structural Dynamics 23, 813–835.
Abe, M. and Igusa, T., 1995, “Tuned mass dampers for structures with closely spaced natural frequencies,” Earth-
quake Engineering and Structural Dynamics 24, 247–261.
Antman, S. S., 2005, Nonlinear Problems of Elasticity, Springer, New York.
Bachmann, H., 1995, “Dynamic forces for rhythmical human body motions,” in Vibration Problems in Structures:
Practical Guidelines, Birkhäuser, Basel.
Brownjohn, J. M. W., 1997, “Vibration characteristics of a suspension footbridge,” Journal of Sound and Vibration
202, 29–46.
Brownjohn, J. M. W., Dumanoglu, A. A., and Taylor, C. A., 1994, “Vibration characteristics of a suspension foot-
bridge,” Engineering Structures 16, 395–406.
British Standards Institution, 1978 “BS5400: Steel, concrete and composite bridges. Part 2: specification for loads.”
Cevik, M., Pakdemirli, M., 2005, “Non-linear vibrations of suspension bridges with external excitation”, Interna-
tional Journal of Non-Linear Mechanics 40, 901–923.
Dallard, P., Fitzpatrick, A. J., Flint, A., Le Bourva, S., Low, A., Ridsdill Smith, R. M., and Willford, M., 2001, “The
London Millennium Footbridge,” The Structural Engineer 79, 17–33.
Den Hartog, J. P., 1934, Mechanical Vibrations, McGraw-Hill, New York.
Fujino, Y., Pacheco, B. M., Nakamura, S., and Warnitchai, P., 1993, “Synchronization of human walking observed
during lateral vibration of a congested pedestrian bridge,” Earthquake Engineering and Structural Dynamics
22, 741–758.
Irvine, H. M., 1984, Cable Structures, Dover Publications, New York.
Jones, R. T., Pretlove, A. J., and Eyre, R., 1981, “Two case studies in the use of tuned vibration absorbers on
footbridges,” The Structural Engineer 59B, 27–32.
776 N. CARPINETO ET AL.

Lacarbonara, W. and Colone, V., 2007, “Dynamic response of arch bridges traversed by high-speed trains,” Journal
of Sound and Vibration 304, 72–90.
Lacarbonara, W., Paolone, A., and Vestroni, F., 2007, “Nonlinear modal properties of nonshallow cables,” Interna-
tional Journal of Non-Linear Mechanics 42, 542–554.
Lacarbonara, W. and Vestroni, F., 2002, “Feasibility of a vibration absorber based on hysteresis,” in Proceedings of
the Third World Congress on Structural Control, Como, Italy, April 7–12.
Lacarbonara, W. and Yabuno, H., 2006, “Refined models of elastic beams undergoing large in-plane motions: theory
and experiment,” International Journal of Solids and Structures 43, 5066–5084.
Law, S. S., Wu, Z. M., and Chan, S. L., 2004, “Vibration control study of a suspension footbridge using hybrid
slotted bolted connection elements,” Engineering Structures 26, 107–116.
Li, H. and Ni, X., 2007, “Optimization of non-uniformly distributed multiple tuned mass damper,” Journal of Sound
and Vibration 308, 80–97.
Matsumoto, Y., Nishioka, T., Shiojiri, H., and Matsuzaki, K., 1978, “Dynamic design of footbridges,” IABSE Pro-
ceedings, No P-17/78, 1–15.
Nayfeh, A. H. and Pai, P. F., 2004, Linear and Nonlinear Structural Mechanics, Wiley-Interscience, New York.
Poovarodom, N., Kanchanosot, S., and Warnitchai, P., 2003, “Application of non-linear multiple tuned mass dampers
to suppress man-induced vibrations of a pedestrian bridge,” Earthquake Engineering and Structural Dynam-
ics 32, 1117–1131.
Vestroni, V. and Vidoli, S., 2007, “Closed-form solutions for the structural response to train loads,” Journal of Sound
and Vibration 303, 691–706.
Xu, K. and Igusa, T., 1992, “Dynamic characteristics of multiple substructures with closely spaced frequencies,”
Earthquake Engineering and Structural Dynamics 21, 1059–1070.
Yamaguchi, H. and Harnpornchai, N., 1993, “Fundamental characteristics of multiple tuned mass dampers for
suppressing harmonically forced oscillation,” Earthquake Engineering and Structural Dynamics 22, 51–62.
Zivanovic, S., Pavic, A., and Reynolds, P., 2005, “Vibration serviceability of footbridges under human-induced
excitation: a literature review,” Journal of Sound and Vibration 279, 1–74.

View publication stats

You might also like