Vous êtes sur la page 1sur 11

Engineering Structures 36 (2012) 228238

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Pedestrian-induced torsional vibrations of suspended footbridges: Proposal


and evaluation of vibration countermeasures
Luca Bruno , Fiammetta Venuti, Vittorio Nasc
Politecnico di Torino, Department of Architecture and Design, Viale Mattioli 39, I-10125 Torino, Italy

a r t i c l e

i n f o

Article history:
Received 6 July 2011
Revised 28 November 2011
Accepted 2 December 2011
Available online 9 January 2012
Keywords:
Suspension footbridge
Pedestrian dynamic load
Vibration countermeasures

a b s t r a c t
The aim of the paper is to propose and evaluate different design arrangements addressed to mitigate the
in-service pedestrian-induced torsional vibrations of lightweight suspended footbridges. All the proposed
structural countermeasures are characterised by the addition of few elements to the original structure,
i.e. punctual masses and/or additional cables located in specic sections to reduce vertical and longitudinal relative displacements of the suspension cables, in order to preserve the bridge lightness and slenderness. An application to a test case footbridge is provided. Its structural nonlinear analysis is carried out
by means of numerical simulations. On the basis of a preliminary modal analysis, a criterion for the localization of the vibration countermeasures along the span is proposed. The effectiveness of the localization
criterion and of the proposed design arrangements is evaluated by comparing the structural responses
obtained through step-by-step dynamic analyses with and without the countermeasures. The analysis
of the results allows the design arrangements to be discussed and the best suited ones to be selected.
2011 Elsevier Ltd. All rights reserved.

1. Introduction
The suspended bridge structural type has been widely
employed from ancient times to nowadays in building longer,
lighter and slender footbridge structures with limited resources
to meet aesthetic and functional demands. Among the various
examples (Fig. 1), let us recall the historical footbridges built over
the Sesia river in Italy [1], the Rhone river in France or the Hida river in Japan [2], the over 500 rescue bridges built by Toni Ruttimann
during the last 25 years in developing countries [3] or the recent
M-bridge in Japan [4].
The dynamic response under service loads is one of the key performances of these footbridges, usually characterised by a thin
plate-like deck, lightweight suspension cables and very low damping. In particular, their dynamic response is highly sensitive
to the pedestrian added mass, and signicantly driven by higher
modes, for which resonant conditions with the pedestrian loads
are expected. Besides vertical and lateral vibrations, which are
well-known to affect these kind of structures (e.g. [4]), the extremely
low torsional stiffness of the deck and the small main cable spacingto-span length ratio make them highly prone to torsional vibrations
induced by the pedestrian loading.
The structural measures commonly adopted to reduce vibrations
on footbridges can be classied according to two different approaches [5]: (i) increasing the damping; (ii) increasing the stiffness.
Corresponding author. Tel.: +39 011 090 4870; fax: +39 011 090 4999.
E-mail address: luca.bruno@polito.it (L. Bruno).
0141-0296/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.12.012

Damping is usually increased by means of passive control strategies,


e.g. through the installation of tuned mass dampers [6] or slotted
bolted connection elements that couple frictional dampers with
rods [7]. The use of extra damping devices is very effective, but requires constant maintenance. Among the arrangements adopted
on suspension footbridges to increase their deck stiffness, let us recall the installation of reverse proled cables in both the vertical and
horizontal planes [9,10], or stiffer handrails [11], or the addition of
inclined stays, connecting the top of the tower with the deck
(Fig. 1c). It is worth noting that the last arrangement has also been
widely adopted in historical long-span suspension bridges, like the
Brooklin bridge [8].
To the authors knowledge, there is no evidence in literature of
structural measures belonging to the second category specically
employed to mitigate torsional vibrations on suspended footbridges. On the contrary, different structural countermeasures have
been proposed in the case of extra-long and long-span suspended
bridges in order to improve their aeroelastic behaviour face to single-mode or coupled utter (e.g. [1216]). Coherently, the effectiveness of these countermeasures is evaluated via the utter critical
wind speed at ultimate limit state. Many of these arrangements
are characterised by the addition of few elements to the original
structure, i.e. punctual masses [13,14] and/or additional cables located in specic sections [15,16]. Therefore, they are well-suited to
be adopted in lightweight suspended footbridges, where the preservation of the aesthetics and slenderness are key objectives. Moreover, the low costs implied by these arrangements make them
suitable to be used on a number of medium and short span existing

L. Bruno et al. / Engineering Structures 36 (2012) 228238

229

Fig. 1. (a) Baraggiolo footbridge (1890, Varallo, Italy) [1]. (b) Gaietta footbridge (1914, Millesimo, Italy). (c) St. George footbridge after the renovation in 1998 (Lyon, France,
Photocredit Dark Gaybeul). (d) M-bridge (1999, Maple Valley, Japan) [4].

footbridges, which suffer vibration problems. In applying the


arrangements proposed for long-span bridges to footbridges, some
important differences between the two structural types and the dynamic loads which they undergo should be taken into account and
properly evaluated:
in extra long-span bridges the mass of the suspension cables is
predominant with respect to the total mass of the bridge, while
the same does not hold on footbridges;
serviceability and not ultimate limit state mostly affects the
dynamic performance of footbridges;
the relative displacements between the suspension cables due
to utter in long-span bridges are by several orders of magnitude higher than the ones induced by pedestrian loading on
footbridges;
in long-span bridges utter usually mainly excites the rst
modes of vibration while in suspended footbridges, characterised by very low natural frequencies, the pedestrian dynamic
load usually excites higher modes of vibration.
The scope of the paper is to propose and evaluate different
design arrangements some of which inspired by ones proposed
for long-span suspended bridges to mitigate the pedestrianinduced torsional vibration of suspended footbridges. Section 2 is
devoted to the description of the proposed arrangements and of
a localization criterion to identify the best position of vibration
countermeasures along the footbridge span. In Section 3 the proposed approach is applied to a test case footbridge and dynamic
analyses are performed in order to compare the peak acceleration
obtained without and with the vibration countermeasures. Finally,
some conclusions are outlined in Section 4.
2. Proposed approach
2.1. Description of vibration countermeasures
Several vibration countermeasures are proposed with the following objectives:

reduce torsional vibrations under pedestrian dynamic loading,


that is, reduce the relative vertical and/or longitudinal displacements between the suspension cables;
preserve the original footbridge aesthetic characterised by its
lightness and slenderness;
guarantee cheap, easy to install and on-site tuneable
countermeasures.
Most of the proposed arrangements are inspired by the ones
proposed in literature for long-span suspended bridges facing to
the wind load. In fact, some mechanical and geometrical similarities between this structural type and lightweight suspension footbridges can be observed [17].
The rst three arrangements are characterised by the addition
of structural elements in given cross sections of the footbridge:
a1. The rst one is based on the addition of eccentric masses ma
(Fig. 2a) on the deck, as already proposed in [12,13]. The role
of the added masses is to increase the deck dead load, in
order to increase in turn the tension of the suspension cables
and their geometric stiffness. This arrangement will be coupled with the following a2.
a2. The second one, inspired by the proposal in [15,16], is characterised by the addition of a horizontal truss between the suspension cables that prevents their relative horizontal
displacements and diagonal hangers, instead of the vertical
ones, that connect the suspension cables to the deck
(Fig. 2b). The diagonal hangers are arranged in order not to
interfere with the passage of the pedestrians and their pretension is given by the added masses on the deck.
a3. the last one is characterised by the substitution of the vertical hangers with a rigid frame. Three alternative arrangements are analysed: a3a (Fig. 2c) is a p-shaped frame, with
vertical columns pinned at both ends and diagonal prestressed cables that connect the horizontal beam to the
deck; a3b (Fig. 2d) is characterised by a frame pinned to
the deck transversal beam; a3c (Fig. 2e) is the same as a3b,

230

L. Bruno et al. / Engineering Structures 36 (2012) 228238

Fig. 2. Proposed arrangements a1 (a), a2 (b) and a3 (c)(d)(e).

but columns are xed at both ends. Arrangements a3 imply a


stronger impact on the bridge aestethics with respect to the
other ones.
Other two arrangements are characterised by the addition of
structural elements along the span of the footbridge:
a4. The rst one (Fig. 3a) is intended to reduce longitudinal relative displacements and is characterised by the addition of
cross stays and trusses connecting the two suspension
cables around the sections where the maximum longitudinal
relative displacement is attended [16].
a5. The second one is based on the addition of inclined stays
from the footings of the towers to the suspension cables
(negative stays [8]). The arrangement, that has been
adopted in the past on some historical footbridges (Fig. 4a)
or in the more recent ones over the river Segre in Spain
(Fig. 4b), was mainly conceived to reduce the static deection. It is expected to be effective to decrease both vertical
and longitudinal relative displacements and is alternative
to the more classical one in which the stays connect the
top of the tower to the deck longitudinal girders (positive
stays, Fig. 1c).
It is worth pointing out that in long-span suspended bridges the
rst antisymmetric torsional mode is usually the one excited by
the wind load. For this reason, the countermeasures of type a1 to
a3 are usually located at one quarter and three quarter of the span
[16], where the excited mode shape has its antinodes. In the case of
suspended footbridges excited by the pedestrian load, the best
location of the vibration countermeasures along the span will be
determined on the basis of the modes which are expected to be
excited and with the criterion described in the next section.

Fig. 4. (a) St. George footbridge over the Saone river, Lyon (France) before the
renovation in 1998 (Photocredit Gilbert Lamboley [24]). (b) Reula footbridge over
the Segre river (1986, Alt Urgell, Spain) [8].

2.2. Description of the localization criterion


This section is devoted to the proposal of a criterion that allows
the best suited position n 2 0; L along the footbridge span to be

selected for the location of the vibration countermeasures. It is


worth recalling that the latter are conceived to minimise the relative displacements between the two suspension cables both in the
vertical and longitudinal direction. Hence, the best suited spanwise
position of the measures is expected to be the one at which the
maximum relative displacement occurs. The proposed criterion is
inspired by these remarks and developed from the results of the
mode extraction only.
Let us consider the modal equation of motion for mode n:

n t 2fn q_ n t x2n qn t
q

Fig. 3. Proposed arrangements a4 (a) and a5 (b).

F n t
;
Mn

where qn t is the principal coordinate, fn is the nth modal damping


ratio, xn 2pfn is the nth natural circular frequency being fn the
natural frequency, F n t is the modal force and Mn the modal mass.
The modal mass and force can be expressed as [18]:

231

L. Bruno et al. / Engineering Structures 36 (2012) 228238

Mn

ms x/2n xdx;

F n t

Fx; t/n xdx;

where L is the length of the footbridge span, /n is the nth unityscaled mode shape, ms the structural mass per unit length. The time
dependent force exerted by a single pedestrians is expressed by the
simplied model

Fx; t F sinxpv t;

where F and xpv are the amplitude and circular frequency of the
vertical component of the force, respectively.
The contribution of each mode to the structural response is
evaluated by means of a weight function:

ctot;n cxr;n cMn ;

where the two functions cxr;n and cMn are dened as follows.

cxr;n

Dxr;n
Dn 2fn ;
Dxr;n 1

where xr;n xpv =xn is the frequency ratio and Dn 1  x


2fn xr;n 2 0:5 is the nth dynamic amplication factor. This function weights the contribution of each mode on the basis of its
distance from the resonance condition (xr;n 1).

Mn
;
M

where M is the total mass of the structure. This function weights


each mode on the basis of its modal mass. Indeed, it is evident from
Eq. (1) that the larger Mn , the lower the modal (and hence physical)
responses and vice versa [19]. For the structural type object of this
study, characterised by extreme lightness and sine-like mode
shapes, it is expected that cxr;n plays a more crucial role than
cMn in weighting each mode contribution.
Let us introduce a general localization function Kx conceived
in a way that the higher its value along the span, the more suited
the corresponding spanwise position to install the vibration countermeasures, the more effective the countermeasures. The function
is dened as:

Kk

j
X

D/0k;n ;

n1

where

D/0k;n ctot;n jD/k;n j ctot;n j/k;n;A  /k;n;B j;

maxK  maxloc K
;
maxK

0 6 g 6 1;

where maxK and maxlocK are the absolute and local maxima of the
function, respectively. The local maxima for which g is less than a
 are selected: the lower g
 , the lower the number
threshold value g
of selected local maxima, and the closer each other their values. Other
non structural selection criteria can be adopted in the second step, if
multiple alternative locations are selected in the rst step. In this
study the selection criterion aims at: (i) minimising the interference
with the passage of pedestrians (specically in arrangement a2); (ii)
minimising the number of cross stays and trusses in arrangement a4
and maximising the slope of the negative stay in arrangement a5. It
follows that the position closest to the footbridge tower is always
preferred among the alternative locations.
3. Application and results
3.1. Description of the test case footbridge

5
2 2
r;n

cMn 1 

with /k;n;A and /k;n;B the kth component (k x longitudinal, k y


vertical) of the nth mode shape for the suspension cables A and B,
respectively. It is worth pointing out that the number j of modes
to be retained to evaluate K can be dened according to different
criteria. For instance, every mode with frequency below a given
threshold value, the modes with the highest value of ctot or the
modes within the maximum risk of resonance according to the classication proposed in [20] could be selected.
It is worth noting that the localisation function can have multiple local maxima with close values in 0; L, because it results from
the weighted sum of periodic functions (Eqs. 7, 8). From an
engineering point of view, this means that several spanwise locations candidate to host the countermeasures showing analogous
effectiveness. In this case, a two-steps strategy is proposed to select the abscissa n of the countermeasure location among the
abscissa of the local maxima argmaxloc K. The rst step involves
a selection of the local maxima based on the localisation function
only. Local maxima with values close enough to the absolute
maxima are selected by introducing the ratio g:

The test case suspended footbridge, symmetrical with respect to


the xy and yz planes, is represented in Fig. 5 and its main geometrical properties are summarised in Table 1. Its geometric and
mechanical features have been set to be representative of lightweight footbridges (e.g. the ones in Fig. 1): the mass of the deck
per spanwise unit length is equal to 106 kg/m. Specically, the
deck consists of main transversal steel beams (UPN140), which
support two secondary longitudinal stringers (UPN100) covered
with steel gratings. The 10 m-high A-shaped towers are made of
steel tubes and are pinned at the footing.
The 3D m-dofs discrete model is implemented and analysed
using the Finite Element Method (FEM) through the commercial
code Ansys. The full description of the adopted numerical techniques is out of the scope of this paper. Nevertheless, the main features of the numerical approach are briey outlined in the
following. Each suspension cable and each stay is modelled by multiple truss elements, with tensioncompression capabilities, while
the hangers are resistant to tension only. The deck transversal and
longitudinal beams and the towers are modelled by beam elements,
with tensioncompression, torsion and bending capabilities. The
non-structural components (paving, handrails) are modelled by nodal masses located at the intersection between longitudinal and
transversal beams. The element types and properties are summarised in Table 2. A damping ratio f 0:005 is considered and taken
into account by a classical proportional damping model, being the
damping matrix assumed as a linear combination of the stiffness
and mass matrix (Rayleigh damping, [18]). The towers are pinned
at the footings. The deck abutments are restrained to translation
along the y and z axes, while elastic constraints (with stiffness equal
to 1e-5 times the axial stiffness of the deck longitudinal beam, that
is, approximately equal to 20 N/m) are imposed in the longitudinal
direction x in order to fullll symmetrical conditions.
3.2. Footbridge dynamic behaviour without vibration
countermeasures
In this section, the dynamic behaviour of the test case footbridge in its initial conguration a0, i.e. without the vibration countermeasures, is analysed. First, a modal analysis is performed to
extract mode shapes and frequencies; hence, the dynamic response
of the footbridge under two different load conditions is calculated
by means of step-by-step dynamic analysis.
3.2.1. Modal analysis
Due to the characteristic nonlinearity of suspension bridges, the
modal analysis is performed after a nonlinear static analysis of

232

L. Bruno et al. / Engineering Structures 36 (2012) 228238

Fig. 5. Geometry of the test case suspended footbridge.

Table 1
Geometrical data for the test case suspended footbridge.
Main span
Sag
Hangers spacing
Towers height
Distance between suspension cables
Distance between longitudinal beams
Deck width

L 100 m
s 8:5 m
d 2:5 m
H 10 m
2be 2:5 m
2bi 1:5 m
2bd 1:6 m

Table 2
Element properties for the test case suspended footbridge.

Deck trans.
beam
Deck long. beam
Tower
Cable
Hanger
Nodal masses

A m2

J z m4

Jy m4

E
(GPa)

Mass (kg/m kg)

0.0020

6:05e6

6:25e7

210

16

0.0014
0.0083
0.0013
0.0002

2:05e6
7:15e5

2:92e7
7:15e5

210
210
159
210

10.6
68.4
10.4
1.6
68

the structure, subjected to dead load and prestress, in order to


determine the geometric tangent stiffness matrix to be used in
the modal analysis. The frequencies of the rst 80 modes extracted
are plotted versus the mode number in Fig. 6 and classied according to their predominant component. It can be observed that there
is a signicant number of local (cable) modes and that natural frequencies are really close to one another, as typically happens in
this kind of structures. The mode frequency axis is mapped according to the risk of resonance for vertical vibrations as proposed in
the Setra/AFGC guideline [20]. The cumulative number of vertical
and torsional modes belonging to each class is graphed in Fig. 7.
No torsional modes fall into the maximum risk range, while four
modes (10, 19, 27, 28) fall in the medium risk range. It is worth

pointing out than two of them (27 and 28) have frequencies that
are close to the boundary value between maximum and medium
risk. Fig. 8 reports the mode shape and frequencies of the rst six
torsional modes. It is worth pointing out that the rst torsional
mode is one-node antisymmetric: this is typical of suspension
bridges with extremely exible deck, so that the bridge has a mode
shape close to the one of the cable alone [21].

3.2.2. Transient analysis


Nonlinear step-by-step dynamic analyses are performed on the
FE model in order to obtain the maximum accelerations on the footbridge in the initial conguration (subscript 0) under a realistic,
exploratory service load condition. The load condition is conceived
in order to: (i) compare the effectiveness of the proposed countermeasures under common, exploratory external loads; (ii) induce a
structural response characterised by signicant contributions from
multiple modes, allowing the effectiveness of the proposed multimode localisation criterion to be evaluated; (iii) model a realistic
service load which induces uncomfortable vertical accelerations of

Fig. 6. Frequencies in Hz of the rst 80 modes.

233

L. Bruno et al. / Engineering Structures 36 (2012) 228238

Fig. 7. Classication of the torsional (a) and vertical (b) modes according to the risk of resonance [20].

Fig. 8. Mode shape of the rst six torsional modes.

the deck without countermeasures. Bearing in mind these objectives, the simulations are performed by considering the travelling
load of a single pedestrian walking at constant speed v [22]:

Fx; t Ftdxp  v t;

10

where d is the Dirac operator, xp is the pedestrian position along the


bridge span and the time component Ft is expressed according to
Eq. (3), with F 280 N [20], xpv 4p rad/s (that is, fpv 2 Hz) and
v 1:5 m=s. The total time of simulation T is set equal to the time
needed to cross the footbridge length walking at velocity v, that is,
T L=v 66:67 s. The following load conditions (LC) are considered (Fig. 9):
LCa: one pedestrian walks keeping his/her right, so that he/she
excites both torsional and vertical motion of the deck;
LCb: two pedestrians walk at a distance equal to the width of
the walking path in opposition of phase, so that torsional vibrations only are induced.

The results of the simulations in terms of maximum value of the


vertical and lateral acceleration of the deck are reported in Table 3.
Load condition LCb is the one which induces the highest value of
the structural response, with a maximum vertical acceleration that
falls close to the boundary between minimum comfort (1

Fig. 9. Scheme of the load conditions.

6 2:5 m=s2 ) and discomfort classes (y


> 2:5 m=s2 ), according
<y
to the classication in [20]. Therefore, the vibration countermeasures will be tested at LCb only.
In order to evaluate the inuence of each mode on the structural response over the simulation time T as a function of the pedestrian position along the span, a criterion analogous to the
Table 3
Maximum vertical and lateral acceleration of the deck.

max;0 m=s2
y
x=L
zmax;0 m=s2
x=L

LCa

LCb

1.7803

2.482

0.175
0.0331

0.625
0.0756

0.6

0.6

234

L. Bruno et al. / Engineering Structures 36 (2012) 228238

Fig. 10. (a) MAC envelopes versus t=T and (b) mode shapes of the rst six modes
against x=L.

Modal Assurance Criterion (MAC) [23] is calculated for the rst six
modes and for LCb:

MACn t

 T

/ yt2
n

/Tn /n yT tyt

11

where /n is the eigenvector of the nth mode and yt is the vector of


vertical displacements of the footbridge deck. In Fig. 10a the envelopes of MACn t are plotted against the dimensionless time t=T,
while Fig. 10b recalls the eigenvectors against the dimensionless
space variable x=L. The following considerations can be done: mode
28 is the one which mostly contributes to the structural response,
since the value of MAC is higher than 0.9 for almost all the time
T; mode 27 is mainly excited when the pedestrian is around the
positions x=L 0:25 and x=L 0:85, which correspond to the mode
shape antinodes; modes 10 and 19 reach a value of MAC around 0.6
when the pedestrian crosses the bridge at x=L 0:1 and x=L 0:5,
respectively; the other modes have a minor inuence.
3.3. Footbridge dynamic behaviour with vibration countermeasures
In this section, the vibration countermeasures a1 to a5 are
applied to the test case footbridge. First, the localization criterion
is applied to nd the best location of the countermeasures on the
basis of the modal analysis performed in Section 3.2.1; hence,
step-by-step dynamic analyses are performed on the FE model in

Fig. 12. Weighting functions cxr;n (a), cMn (b) and ctot (c).

order to compare the maximum accelerations on the footbridge


in the initial conguration (Table 3) with the ones obtained when
the vibration countermeasures are considered. The simulations are
performed by applying load condition LCb only.

3.3.1. Location of the vibration countermeasures


The localization criterion described in Section 2.2 is applied to
the test case footbridge. Fig. 11 plots the vertical and longitudinal

Fig. 11. Vertical (a) and longitudinal (b) relative components of cable mode shapes.

235

L. Bruno et al. / Engineering Structures 36 (2012) 228238

Fig. 13. Vertical (a) and longitudinal (b) weighted relative components of cable mode shapes.

Table 4
 0:3.
Location n=L of the vibration countermeasures for g

Vertical
Longitudinal

All modes

Modes 27 28 36 53

Modes 27 28 19 10

0.15
0.075

0.125
0.125

0.15
0.15

Table 5
Maximum relative vertical and longitudinal displacements of the suspension cables
and their position along the span.

Fig. 14. Localization functions obtained considering different number of modes.

relative components of the rst six torsional mode shapes for the
suspension cables.
Fig. 12 graphs the weight functions. As expected, the function
cxr;n (Fig. 12a) has a more relevant role in weighting the contribution of each mode than cMn (Fig. 12b). It can be observed that the
two modes with the highest value of ctot (Fig. 12c) (modes 27 and
28) are also the ones with the highest values of MAC (see Fig. 10),
that is, the ones which mostly contributes to the bridge dynamic
response.
Fig. 13 plots the vertical and longitudinal weighted relative
components D/0x;n and D/0y;n of the torsional mode shapes for the
suspension cables for the rst six modes. The localization functions
plotted in Fig. 14 for half span have been obtained by considering
different number of modes according to the general criteria introduced in Section 2.2: all torsional modes with frequency below
5 Hz; the rst four modes with the highest value of ctot (# 27 28
36 53); the modes located in the maximum and medium risk of
resonance according to Setra/AFGC (# 27 28 19 10). It can be
observed that: (i) the function has a similar trend for the three
cases, but the position of the maxima slightly differs; (ii) despite

Dymax;0 m
x=L
Dxmax;0 m
x=L

LCa

LCb

0.0162
0.15
0.0037
0.125

0.0317
0.15
0.0073
0.125

the considered number of modes, the localization function Ky


shows two local maxima with almost the same value. The best
location n=L is nally determined by identifying, among the local
 , the one closest to the
maxima that satisfy the condition g < g
tower. Fig. 15 shows the values of n=L obtained with increasing
number of considered modes (ordered for decreasing value of
ctot ) and with two different values of g . It is worth recalling that
 closest to zero forces the selection of the absolute
the value of g
maximum of K, whichever the number of considered modes: this
causes the location n=L to oscillate between the two design
solutions as the cumulative number of modes increases. On the
 allows the selection
contrary, the choice of the higher value of g
of the relative maximum nearest to the tower, causing the convergence of n=L for increasing cumulative number of modes.
 0:3 are retained and
Therefore, the results obtained for g
summarized in Table 4.
In order to evaluate a posteriori the applied criterion, Table 5 reports the maximum relative vertical and longitudinal displacements
of the suspension cables obtained in the initial conguration:

.
Fig. 15. Values of n=L for vertical (a) and longitudinal (b) localization functions and different values of g

236

L. Bruno et al. / Engineering Structures 36 (2012) 228238

 n.
Fig. 16. Peak values of vertical (a) and lateral (b) acceleration with arrangements a1 and a2 for different values of 4ma =M

Fig. 17. a3: peak values of vertical (a) and lateral (b) acceleration for different values of the shortening ratio.

Fig. 18. a4 (a) and a5 (b): peak values of vertical acceleration for different values of the shortening ratio SR.

Dymax;0 maxjDyj and Dxmax;0 maxjDxj. It can be observed that the


position along the span where the values of Dymax;0 are attained is the
same obtained through the localization criterion when all modes or
the ones according to the Setra/AFGC criterion are selected. On the
contrary, Dxmax;0 is attained at x=L 0:125, which corresponds to
the position obtained with the localization criterion when the rst
four modes with the highest values of ctot are considered (see Table

4). The position n=L 0:15 (and its symmetric along the span
n=L 0:85) is retained in the following.
3.3.2. Transient analysis
The maximum vertical accelerations obtained in the initial conguration are compared with the ones attained in the presence of
the vibration countermeasures a1 to a5 described in Section 2.1

L. Bruno et al. / Engineering Structures 36 (2012) 228238

237

Fig. 19. Arrangement a5: mode shape, mode number and frequency of the rst torsional (a)(b)(c) and vertical (d)(e)(f) modes for different values of the shortening ratio
SR in the negative stays.

and located along the span according to the results obtained in previous section. Specically, the following properties of the added
structural elements are adopted:
the cable elements in arrangements a2 to a5 have the same
properties as the vertical hangers;
the horizontal truss in arrangement a2 is a steel prole with circular hollow section (CHS 76.1  3.2 mm);
the frame elements in arrangements a3 are CHS steel proles
with Sections 96  7 mm, except for the columns in a3b that
are 96  3 mm CHS;
the horizontal truss in arrangement a4 is a steel prole with circular hollow section (CHS 48.3  1.2 mm).
As far as arrangements a1 and a2 are concerned, a sensitivity
study on their performance to the added mass ma is carried out. In
particular, the total added mass ( 4ma ) varies between zero and
 n . The results of the
20% of the average modal mass of the bridge M
simulations are summarised in Fig. 16. The arrangement a1 allows
a reduction of the vertical acceleration of 52.6 % with ma 112 kg.
The rate of reduction is higher for lower values of the added mass
and decreases as the added mass increases. The arrangement a2
without added mass produces a signicant reduction of the vertical
acceleration of about 33%, while the rate of reduction remains almost constant and low for increasing values of the added mass.
For ma higher than 56 kg, arrangement a1 results more effective
 n , the
than arrangement a2. Within the explored range of 4ma =M
vertical accelerations attained with both arrangements are still in
the class of minimum comfort [20], but almost one half of the initial acceleration. The arrangement a1 is also applied in a different
location along the bridge span than the one selected through the
localization criterion, in order to test the effectiveness of the latter.
The adopted alternative location is at x 0:25L (and its symmetrical x 0:75L), that is, in correspondence of the antinodes of the
rst torsional mode, which is the one usually accounted for in

the evaluation of analogous vibration countermeasures for longspan bridges against utter (see Section 1). In this case the maximum acceleration has a slight decrease, conrming the importance
and effectiveness of the localization criterion for this kind of footbridges. Fig. 16b evidences that, while the lateral accelerations
zmax;1 obtained with arrangement a1 are very small, the inclined
hangers in arrangement a2 cause the lateral acceleration zmax;2 to
increase up to eight times the maximum lateral acceleration
zmax;0 of the deck without countermeasures (zmax;0 0:0756 m=s2 ,
see Table 3), with values that fall in the range of minimum comfort
(z > 0:3 m=s2 ). This fact conrms the best overall performances of
arrangement a1.
The performances of arrangements a3a, a4 and a5 are evaluated
for different values of the shortening ratio (SR) of the added cables,
that is, the ratio of the relative displacement between the cable
anchorages (e.g. imposed by means of a turnbuckle) and the length
of the cable.
The effects of arrangement a3 are plotted in Fig. 17: the peak
acceleration is plotted against the SR on the lower axis, while the
upper axis reports the percentage value of the mean cable stress
r0 in the dead load conguration with respect to the ultimate cable
stress ru 1450 MPa. Among the three alternative schemes, a3b is
certainly the less effective: indeed, it induces a small decrease
(11.27%) of the vertical acceleration, but the lateral acceleration
reaches the highest value (0:8 pedestrian=m2 ), which corresponds
to the limit of the discomfort range. With arrangements a3a and
a3c, the vertical acceleration signicantly decreases (37.8% and
51.6%, respectively). The SR in a3a has a minor effect on the structural response, which remains almost constant for values of the SR
higher than 0.76e-3. All a3 arrangements induce peak lateral accelerations falling in the range of minimum comfort.
Arrangement a4 (Fig. 18a) shows a signicant effectiveness in
reducing the structural vertical response without inducing large
lateral oscillations (0:053 6 zmax;4 6 0:084), always in the maximum comfort range [20], and almost constant versus SR. The peak

238

L. Bruno et al. / Engineering Structures 36 (2012) 228238

vertical acceleration in the case of null cable prestress is 30% of the


initial value and further decreases for increasing values of the SR,
reaching the range of mean comfort. It should, therefore, be considered that the high tension in the cross stays (up to 35% the ultimate
stress) induces high values of compression in the horizontal
trusses, which should be properly resized in a design scenario.
Finally, the peak values of the vertical acceleration obtained with
arrangement a5 are plotted in Fig. 18b, while lateral acceleration is
not graphed because its low values (0:023 6 zmax;5 6 0:073) almost
constant versus SR. Differently from the previous arrangements,
the a5 performances do not show a regular and monotonic trend versus the design parameter (i.e. SR or the prestress r0 =ru ). The
arrangement is highly effective for very low SR value (lower than
0.76e-3, i.e. r0 =ru < 1:38%), when the vertical acceleration decreases of around 62% reaching the range of mean comfort. As
the SR increases, the footbridge gradually undergoes changes in
its structural behaviour due to the stress stiffening effects induced
by the negative stays. In particular, the following effects take place:
(i) the stays provide an elastic constraint to the suspension cables
so that the suspension cable effective length tends to reduce to the
distance between the stays anchorages L1 (see Fig. 3); (ii) the sag to
span ratio of the suspension cables decreases; (iii) the main cable
stiffness increases. As a consequence, the trend of the peak vertical
acceleration versus the SR shows two transient regimes characterised by decreasing branches (tuning branches) followed by an
abrupt increase of the acceleration almost up to the initial value
max;0 (switching branches), which correspond to a sudden change
y
of the structural behaviour. This change is conrmed by looking at
the frequencies and mode shapes of the rst torsional and vertical
modes for the three values of SR corresponding to the response
local maxima (highlighted with circles in Fig. 18b). The reduction
of the suspension cable effective length is clearly visible in the vertical mode starting from the second case (Fig. 19ef) and in the torsional mode in the third case (Fig. 19c). The change in the
suspension cable behaviour also corresponds to a change in the
natural frequency, which is particularly evident in the torsional
mode. It is worth pointing out that this change does not involve
the vertical and torsional modes for the same value of the SR: this
is conjectured to be due to the fact that the stay added exural
stiffness is higher than the torsional one with respect to the global
exural and torsional stiffnesses of the footbridge. Once the footbridge has reached the new conguration for values of SR higher
than 4.75e-3, the vibration countermeasure shows regular performances, the trend of the peak acceleration monotonically
decreases and its value belongs to the mean comfort class for SR
values higher than 15e-3.
4. Conclusions
In this paper, ve structural countermeasures to reduce torsional human-induced vibrations on suspended footbridges have
been proposed and evaluated on a test case footbridge by means
of numerical simulations. All the proposed measures are characterised by the addition of only few elements to the structure, while
preserving the footbridge lightness and so allow the application
on existing footbridges at low costs. The elements added to the
structure are located along the bridge span according to a proposed
localization criterion based on the minimisation of the relative displacements between the suspension cables.
The results of the simulations allow the following considerations to be outlined:
among the different arrangements, considering their overall
performances and drawbacks (e.g. the relevant increase of the
lateral acceleration in a2 and a3 or the high compression in

the trusses of a4), a1 seems to be the most effective since it


reduces the vertical acceleration by one half with a relatively
small added mass (20% of the average modal mass). A similar
performance can be obtained with arrangement a5, for suitable
values of the initial prestress;
each arrangement should be tuned on the specic footbridge,
both to select its best location, which depends on the structure modal properties, and to calibrate the value of the free
design parameter, that is, the added mass in arrangements
a1 and a2, and the added cable prestress in arrangements
a3a, a4 and a5. Such a ne tuning could be performed by
means of further numerical simulations or in situ experimental tests on the basis of the general and comparative results
obtained in the present study. Moreover, the in situ experimental tuning can be easily performed because it involves
few accessible and lightweight structural elements, especially
in arrangements a1 and a5;
some of the arrangements could be coupled in order to increase
their effectiveness (e.g. a4 with a1 or a4 with a2).

References
[1] Re L. Cable suspended: the suspension bridges of the 19th century in Valsesia
(in Italian), Lindau, Turin; 1993.
[2] Leonhardt F. Brcken/Bridges: aesthetics and design. Stuttgart: Deutsche
Verlags-Anstalt; 1982.
[3] Adev B, Badoux M, Brhwiler E, Ruttimann T. Bridges making a difference.
Struct Eng Int 2000;10(1):669.
[4] Nakamura S. Field measurement of lateral vibration on a pedestrian
suspension bridge. Struct Eng 2003;81(22):226.
[5] Zivanovic S, Pavic A, Reynolds P. Vibration serviceability of footbridges
under human-induced excitation: a literature review. J Sound Vib
2005;279:174.
[6] Carpineto N, Lacarbonara W, Vestroni F. Mitigation of pedestrian-induced
vibrations in suspension footbridges via multiple tuned mass dampers. J Vib
Control 2010;16(5):74976.
[7] Law SS, Wu ZM, Chan SL. Vibration control study of a suspension footbridge
using hybrid slotted bolted connection elements. Eng Struct 2004;26:10716.
[8] Troyano LF. Bridge engineering: a global prospective. Cambridge: Thomas
Telford; 2003.
[9] Huang M-H. Dynamic characteristics of slender suspension footbridges, Ph.D.
thesis, Queensland University of Technology; 2006.
[10] Huang M-H, Thambiratnam DP, Perera NJ. Dynamic performance of slender
suspension footbridges under eccentric walking dynamic loads. J Sound Vib
2007;303(1-2):23954.
[11] Gentile C, Gallino N. Ambient vibration testing and structural evaluation of an
historic suspension footbridge. Adv Eng Softw 2008;39:35666.
[12] Diana G, Bruni S, Collina A, Zasso A. Aerodynamic challenges in super long span
bridge design. In: Proceedings international symposium on advances in Bridge
Aerodynamics; 1998.
[13] Nakazaki S, Yamaguchi H. Preliminary design of super long span suspension
bridges using the temporary additional mass method against a storm. J Wind
Eng Indust Aerodyn 1999;83:31726.
[14] Phongkumsing S, Wilde K, Fujino Y. Analytical study on utter suppression by
eccentric mass method on fem model of long-span suspension bridge. J Wind
Eng Indust Aerodyn 2001;89:51534.
[15] Astiz M, Casado C. Flutter stability of very long suspension bridges. J Bridge
Eng ASCE 1998;3(3):1329.
[16] Miyazaki M, Arai M, Kazama K, Kubota H. Stay-cable systems of long span
suspension bridges for coupled utter. In: Proceedings second European and
African conference on wind engineering; 1997.
[17] Bruno L, Venuti F, Scotti A. Limit of hanger linearity in suspension footbridge
dynamics: a new section model. J Sound Vib 2011. doi:10.1016/
j.jsv.2011.07.042.
[18] Clough R, Penzien J. Dynamics of structures. New York: McGraw-Hill; 1987.
[19] Brownjohn J, Pavic A. Experimental methods for estimating modal mass in
footbridges using human-induced dynamic excitation. J Sound Vib
2007;29(1):283343.
[20] Stra/AFGC, Assessment of vibrational behaviour of footbridges under
pedestrian loading (October 2006).
[21] Brownjohn JMW. Vibration characteristics of a suspension footbridge. J Sound
Vib 1997;202(1):2946.
[22] Kala J, Salajka V, Hradil P. Footbridge response on single pedestrian induced
vibration analysis. Eng Technol 2009;5:26980.
[23] Maia NMM, Montalvo e Silva JM. Theoretical and experimental modal
analysis. New York: Research Studies Press Ltd.; 1997.
[24] <http://www.art-ethistoire.com/blc.php?./seguin/pt7lyg.jpg;1340;887./hr/>.

Vous aimerez peut-être aussi