ARTICLE IN PRESS

Journal of Wind Engineering

and Industrial Aerodynamics 96 (2008) 229–256
www.elsevier.com/locate/jweia

A simplified approach to bridge deck flutter

Gianni Bartoli, Claudio Mannini
Department of Civil Engineering, University of Florence, Via S. Marta 3, 50139 Florence, Italy
Received 1 October 2005; received in revised form 27 March 2007; accepted 6 June 2007
Available online 21 August 2007

Abstract

Flutter instability of flexible bridge decks is investigated in the framework of the semi-empirical
Scanlan’s approach based on flutter derivatives. The equations of the eigenvalue problem of stability
are manipulated and simplified thanks to the analysis of a large number of dynamic and aerodynamic
data. Simple expressions for critical reduced wind speed and coupling frequency are obtained. The
major interest of this approximate method is the fact that only three or even two flutter derivatives
are required to calculate the flutter instability limit. In addition, these aerodynamic functions are
known to be quite reliable and the easiest to be identified through wind-tunnel tests. The proposed
formulas seem to give accurate results in a wide range of cases, unless the frequency ratio is very close
to unity. The method is shown to apply also in case of cross sections prone to torsional flutter.
Finally, the paper offers a valuable insight into the flutter behavior of several types of bridges and
investigates the role in the instability onset played by various structural parameters, such as
frequency ratio and structural damping.
r 2007 Elsevier Ltd. All rights reserved.

PACS: 46.40.Jj

Keywords: Bridges; Aeroelasticity; Flutter; Flutter derivatives; Wind-tunnel tests; Simplified method

1. Introduction

Flutter instability is one of the problems of major concern in the design of long-span
flexible bridges as it can lead to the collapse of the structure. Therefore a reasonable safety
margin against flutter has to be carefully assured.
Corresponding author. Tel.: +39 055 4796326.
E-mail address: claudio.mannini@dicea.unifi.it (C. Mannini).

0167-6105/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2007.06.001
 ARTICLE IN PRESS
230 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

The flutter critical wind speed can be determined directly by wind-tunnel tests observing
the behavior of aeroelastic models which respect as many similitude parameters as
possible. Alternatively, a model with only geometrical similitude can be employed and
eight empirical functions (or even eighteen, if drag force and sway motion are not
neglected), called flutter derivatives (Scanlan and Tomko, 1971), have to be identified in
order to define the self-excited load. These functions are supposed to be dependent only on
cross-sectional geometry and reduced wind speed, a dimensionless parameter involving the
mean wind speed and the frequency of oscillation. Nevertheless, it is well known that in
many instances these aeroelastic coefficients strongly depend on mean angle of attack as
well (e.g. Diana et al., 1999, 2004, 2005; Mannini, 2006). Assuming pure harmonic motion
with the same frequency for heaving and pitching degrees of freedom, the critical wind
speed is obtained by solving an eigenvalue problem of stability (e.g. Simiu and Scanlan,
1996). Alternatively a time-domain analysis can be performed (e.g. Caracoglia and Jones,
2003): indicial functions are identified usually from flutter derivatives and then pre-critical,
critical and post-critical conditions can be simulated.
In the case of very-long-span bridges, aeroelastic problems and in particular flutter are
often the most concerning points in the design. Consequently, extensive wind-tunnel tests
have to accompany all the stages of the project from the beginning. The cross-section
aerodynamics is optimized by modifying shape and non-structural details, but also some
dynamic parameters, such as the frequency ratio, can be changed in order to increase the
flutter critical wind speed. Afterwards, additional tests are needed to properly evaluate the
wind loads for the chosen solution and, for very important structures, a full-aeroelastic
model is finally built to validate the design.
The aforementioned experimental campaign is obviously expensive and time-consuming
but at the present state of the art is absolutely necessary for the design of long-span
suspension bridges. Conversely, for medium-span bridges and conventional geometries,
flutter instability is usually less concerning although it cannot be excluded a priori. For
bridge engineers it could be very useful to dispose of an instrument able to put in evidence
or exclude with a certain degree of reliability the risk of flutter, without performing wind-
tunnel tests or postpone them as final validation step. In addition, this tool could be useful
in general at the preliminary bridge design stages in order to quickly estimate the flutter
critical wind speed and take some first decisions before starting the wind-tunnel tests.
Unfortunately, this instrument is not provided by national or international codes or
recommendations, which always refer to experiments for flutter assessment. Indeed,
empirical formulas for classical flutter and galloping were given in the ECCS re-
commendations (ECCS, 1987), which, as far as the writers know, was the most courageous
normative document at this regard, but afterwards they were not reported in the following
documents of the Eurocode 1 (e.g. ENV1991-2-4, 1995). The present work should be seen
as a first step in setting-up an approximate procedure of flutter assessment.
In the following the flutter equations are manipulated on the basis of a large number of
dynamic and aerodynamic data in order to find simplified analytical expressions for critical
frequency and reduced wind speed, depending only on three (H 1 , A2 and A3 ) or even two
(H 1 and A2 ) aeroelastic functions instead of the usual eight. It is worth noting that the
retained coefficients are known to be quite reliable and the easiest to be identified through
wind-tunnel tests and those for which more data are available in the literature. In addition,
by means of these formulas the calculation of the coupled flutter critical wind speed is
possible also without measuring cross-flutter derivatives (H 2 , H 3 , A1 , A4 ), i.e. adopting a
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 231

simpler one-degree-of freedom (1-DoF) test set-up. The validity and the limits of this
approach are investigated through a wide range of data and its applicability results to be
surprisingly wide, being the degree of approximation usually very good. In particular, the
method is valid not only in the case of classical coupled flutter but also of torsional flutter.
As future development, an approximate general expression for the three retained flutter
derivatives could be sought for some well-defined classes of conventional deck cross-
sectional geometry (Mannini, 2006), in order to define a simplified method of flutter
assessment not requiring wind-tunnel tests.
The writers have recently found out that Nakamura (1978) proposed a series of
approximate equations which can be easily transformed into the formula proposed here for
the critical reduced wind speed, provided that structural damping is neglected. It seems a
strong argument for the validity of the present approach the fact that the two results are, at
least partially, identical despite they are obtained in completely independent ways and
following different principles. A section of this paper is devoted to the comparison between
the two results.
As previously remarked, the proposed simplified approach cannot fully replace more
sophisticated flutter stability analyses but it is also important for the insight it offers into
the flutter equations and the instability mechanism. In particular, it highlights the key role
played by the flutter derivatives H 1 , A2 and A3 and confirms the idea that Scanlan’s model
for self-excited forces, with eight empirical aeroelastic functions, is somehow redundant
(e.g. Matsumoto et al., 1995, 1996, 1997; Matsumoto, 1996, 2005; Borri et al., 2005). The
effect on the flutter instability of structural dynamic parameters, such as damping and
frequency ratio, is also investigated.
Moreover, it is the first time, as far as the writers know, that such a wide range of
dynamic and aerodynamic data about bridge decks are combined and analyzed in order to
study the flutter mechanism and to check the validity of an approximate formula.
In the second section of this paper the simplified approaches to bridge flutter available in
the literature are briefly discussed. In the third section the mechanical model to deal with
flutter is introduced and then the simplification procedure is outlined. Afterwards, a
possible inter-relation between the retained flutter derivatives is proposed and discussed in
order to further simplify the problem. In Section 6 the degree of approximation of the
proposed formulas is analyzed for a few reference cases and then results are reported for
several combinations of bridge dynamics and aerodynamics. In the eighth section the
particular case of torsional flutter is taken into account. Finally, in Section 9 the
comparison between the proposed simplified formulas and Nakamura’s formulas is
discussed.

2. Simplified approaches to flutter: a short overview

The most common simplified method to assess flutter instability outlines the calculation
of the critical wind speed for a dynamically equivalent thin flat plate (Theodorsen, 1934),
which is then corrected by introducing an empirical coefficient (Aerodynamic Stability
Performance Index) depending on the actual cross-sectional geometry (e.g. Frandsen,
1966; Gimsing, 1997; Dyrbye and Hansen, 1997; Righi, 2003; Bartoli and Righi, 2006). The
advantage of this method consists in its simplicity although, summarizing in one coefficient
all the information about the aerodynamics of the deck section, nothing can be said about
the actual mechanism which leads to instability (for instance if classical or torsional flutter
 ARTICLE IN PRESS
232 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

arises). In addition, for some cross sections this index is not completely independent of the
dynamic properties of the structure.
Another simplified method is the quasi-steady approach (e.g. Bisplinghoff et al., 1996;
Miyata et al., 1995; Minh et al., 1999), in which the knowledge of the static aerodynamic
coefficients and their derivatives are required instead of flutter derivatives. The advantage
of this method is that static coefficients are much easier to be measured than flutter
derivatives. Nevertheless, this approach can be applied only when the quasi-steady
assumption holds, that is at high reduced wind speed U R (low reduced frequency of
oscillation), while many bridges experience instability much before this condition is
reached. In addition, it is worth pointing out that even at high reduced wind speed the
expected error neglecting the fluid-memory contribution can be significant, as suggested by
the example reported in Table 1 (Mannini, 2006), where the dynamic parameters of
Messina Strait Bridge (D’Asdia and Sepe, 1998) are combined with the aerodynamic
behavior of a thin flat plate and the critical reduced wind speed is progressively increased
by reducing the deck chord B. Comparing the unsteady and the quasi-steady results it can
be concluded that the second approach introduces a non-negligible approximation even for
U R ffi 50. This result is not surprising according to Bisplinghoff et al. (1996), who showed
that for a thin airfoil the phase shift between, for instance, self-excited lift and heaving
velocity is maximum for a reduced wind speed of about 16 (k ¼ p=U R ffi 0:2) and even for
U R 430 the fluid-memory effects on flutter instability can be remarkable. Although the
translation of these results from the theoretical case of a flat plate to real bridge deck cross
sections is not automatic, the considerations above should warn about a straightforward
use of the quasi-steady theory for flutter instability calculations, even in the few cases in
which the critical reduced wind speed is high enough. Nakamura and Mizota (1975) and
Nakamura (1979) pointed out that the quasi-steady results could be even ‘‘qualitatively’’
uncorrect in the case of cross sections prone to torsional flutter, playing the fluid-memory
effects an essential role also at high reduced wind speed. In addition, it is worth noting that
applying the quasi-steady theory to real bridge sections, the point which should be taken as
reference for the angular velocity motion is not well defined (e.g. Diana et al., 1993;
Mannini, 2006; Chen, 2007). This point corresponds to the three-quarter and midchord
position respectively for lift and moment in the theoretical case of a flat plate, while it is
assumed to be coincident with the aerodynamic center both in Miyata et al. (1995) and
AFGC (2002), with the leading edge according to Farquharson (1949) or with a point
placed at a chord distance upstream with respect to the midchord in Diana et al. (2004).

Table 1
Unsteady and quasi-steady (QS) flutter results for a thin flat plate (Mannini, 2006)

B (m) kc U Rc U Rc (QS) DU Rc (%)

25 0.1407 22.3 13.8 38:3

15 0.0625 50.2 37.6 25:1
10 0.0313 100.3 84.5 15:7
3 0.0033 960.5 934.5 2:7
1 0.0004 8444 8410 0:4

U Rc ¼ critical reduced wind speed, kc ¼ p=U Rc ¼ critical frequency, B ¼ deck chord. The subscript c stands for
‘‘critical’’.
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 233

In order to overcome the aforementioned limits of the quasi-steady approach, the quasi-
steady-modified theory has been developed at the Politecnico di Milano (e.g. Diana et al.,
1993, 1999; Zasso, 1996), in which the pure quasi-steady formulation is corrected by means
of coefficients which account for the unsteadiness of the flow. These coefficients are close
relatives of flutter derivatives and in the same way they have to be identified through wind-
tunnel tests.
An important research work about the simplification of the flutter problem has been
conducted by Matsumoto and co-workers (Matsumoto et al., 1995, 1996, 1997;
Matsumoto, 1996, 2005). On the basis of wind-tunnel test results for rectangular cylinders
with various chord-to-thickness ratios, several flutter derivative inter-relations were
verified, showing that only two functions (for instance H 1 and A2 ) can be considered as
independent to describe the aeroelastic behavior of these cross sections. This mutual
dependence of aerodynamic derivatives is due to the similarity of unsteady pressure
characteristics per unit angle of attack between heaving and pitching motions and the fact
that the equivalent Wagner function, which is defined as the inverse Fourier
transformation of the equivalent Theodorsen function, expressed by means of the
empirical aeroelastic coefficients, has to be a real-time function (Matsumoto, 1996).
Scanlan et al. (1997) showed that some bridge sections cannot be completely characterized
from the aeroelastic point of view by only two derivatives. Nevertheless, the results for
rectangular cylinders strongly suggest the redundancy of the classical model with eight
flutter derivatives.
Furthermore, Matsumoto proposed an iterative method to deal with flutter instability,
as an alternative to the more common complex eigenvalue analysis, called step-
by-step flutter analysis (Matsumoto et al., 1995, 1996, 1997; Matsumoto, 1996, 2005),
in which heaving and torsional branches are separately analyzed in 2-DoF systems. This
method has the advantage to highlight the stabilizing or destabilizing role played by
each flutter derivative. In particular, concerning coupled flutter Matsumoto shows
the major destabilizing effect of the flutter derivatives A1 and H 3 . On the other hand,
H 1 and A2 have a key stabilizing role, respectively if heaving-branch or torsional-
branch flutter are concerned (Matsumoto, 2005). Conversely, in the case of torsional
flutter the critical destabilizing role of A2 is put in evidence in Matsumoto et al.
(1997), while the contribution of A1 H 3 can be either stabilizing or destabilizing,
according to the degree of bluffness of the cross section. The fact that the pitch
damping can directly control the stability of a bridge deck section even in case
of classical flutter is underlined also in Nakamura (1978), in sharp contrast to most
aircraft flutter, in which the in-phase aerodynamic coefficients have dominant contri-
butions.
Recently, the writers have noted that a closed-form solution to the bimodal flutter
problem was proposed by Chen and Kareem (2006a) under the only assumption of low-
level damping. Afterwards, this solution was further simplified in Chen and Kareem
(2006b) and Chen (2007), leading to an approximate formula for the critical wind speed,
retaining only the flutter derivatives H 3 , A1 , A2 and A3 . This approach also allows
to condensate in just one parameter the influence of the cross-section aerodynamics
on the flutter wind speed. The fact that the considered flutter derivatives are different
from those emphasized in this work should not be seen as a contradiction, in view
of consolidate flutter derivative inter-relations which will be discussed in Section 4.2
[see Eqs. (15)–(17)].
 ARTICLE IN PRESS
234 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

3. Mechanical model

The approximate formulas that are going to be introduced are based on the bimodal
approach to flutter. This is equivalent to assume that only a vertical bending and a
torsional mode contribute to the instability phenomenon and that the coupling is not
significantly affected by structural mode shapes. This is a commonly accepted framework
but for instance it does not fully apply in the presence of imperfect dynamic similarity,
typical of erection stages of a bridge. In these cases the more sophisticated multimode
approach has to be followed (e.g. Jain et al., 1996). In another paper (Bartoli and Mannini,
2005a) all the assumptions that have to be verified in order to be able to apply the bimodal
approach without significant errors, have been studied in details by the writers.
For a two-degree-of-freedom system undergoing heaving and pitching vibrations, the
equations of motion can be written in the following way, common in the aeronautical field
(Fung, 1993):
m½h€ þ ð1 þ igh Þo2h h ¼ Lh , (1)

I½€a þ ð1 þ iga Þo2a a ¼ M a , (2)

where h and a denote respectively the heaving displacement and the pitching rotation
(Fig. 1), m and I the mass and mass moment of inertia per unit length, oh and oa the
circular frequencies of heaving and pitching modes (in still air), Lh and M a the lift and
moment per unit length, i is the imaginary unit and the dot denotes derivative with respect
to time. gh and ga are the coefficients of rate-independent damping for heaving and
pitching modes and, in the case of a sinusoidal motion with circular frequency o, they can
be related to the more common ratio-to-critical damping coefficients zh and za through the
simple expressions:
o
gh ¼ 2zh , (3)
oh
o
ga ¼ 2za . (4)
oa

Fig. 1. Reference system for cross-section displacements and self-excited forces.

 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 235

The advantage of using rate-independent damping (i.e. complex stiffness) instead of the
more familiar notation will result clear later.
Lift and moment can be expressed as the sum of steady forces due to mean wind,
buffeting and self-excited forces but only the latter have to be retained when flutter
stability is concerned. Self-excited forces can be expressed in the Scanlan’s form (Simiu and
Scanlan, 1996), as follows:
" #
1 2  h_  B_a 2  2  h
Lse ¼ rU B KH 1 ðKÞ þ KH 2 ðKÞ þ K H 3 ðKÞa þ K H 4 ðKÞ , (5)
2 U U B
" #
1 2 2  h_  B_a 2  2  h
M se ¼ rU B KA1 ðKÞ þ KA2 ðKÞ þ K A3 ðKÞa þ K A4 ðKÞ , (6)
2 U U B

where r is the air density, U is the mean wind speed, B is the deck chord, K ¼ oB=U is the
reduced frequency of oscillation and the functions H j and Aj (j ¼ 1; . . . ; 4) are the flutter
derivatives, which are supposed to depend on reduced frequency of oscillation and deck
geometry only and have to be measured through wind-tunnel tests.
Substituting Eqs. (5) and (6) into Eqs. (1) and (2), imposing harmonic motion with
coupled frequency (critical condition) and the singularity of the resulting system, as
explained for instance in Simiu and Scanlan (1996), two flutter equations are then
obtained. In particular, following the approach outlined here, they can be written as
follows (Fung, 1993):
   
r2 m2 1 m
ð1  gh ga Þ a 2 X 2  r2a m2 1 þ 2 þ 2 ðA3  gh A2 Þ þ r2a mðH 4  ga H 1 Þ X
go go go
þ r2a m2 þ mA3 þ r2a mH 4 þ H 4 A3  H 1 A2  H 3 A4 þ H 2 A1 ¼ 0, ð7Þ
   
r2a m2 2 2 2 gh m   2  
ðgh þ ga Þ X  r m g þ þ ðg A þ A Þ þ r mðg H þ H 1 X
Þ
g2o a a
g2o g2o h 3 2 a a 4

þ mA2 þ r2a mH 1 þ H 4 A2 þ H 1 A3  H 3 A1  H 2 A4 ¼ 0, ð8Þ

where it is set:
rffiffiffiffiffiffiffiffiffi
2m I oa o2a
m¼ ; ra ¼ ; go ¼ ; X¼ .
rB2 mB2 oh o2
The parameter m represents the mass of the deck, that is proportional to the ratio of the
mass of the deck to the mass of the air moved by the deck; ra is the nondimensional radius
of inertia of the deck, go is the frequency ratio in still air between the torsional and bending
modes susceptible to couple and X is the unknown nondimensional frequency of
oscillation at flutter. It is now clear the advantage of expressing in Eqs. (1) and (2) the
damping as a complex stiffness, since it allows to deal with flutter equations of second
order in X instead of fourth and third order as if the usual notation is assumed (e.g.
Dyrbye and Hansen, 1997). The dynamic properties of the 2-DoF system relevant for
flutter instability are summarized by the parameters m, ra and go along with the damping
coefficients.
Eq. (7) is usually called ‘‘real equation’’ as it is obtained by imposing the vanishing of the
real part of the determinant of the aforementioned system of equations. For analogous
 ARTICLE IN PRESS
236 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

reasons Eq. (8) is called ‘‘imaginary equation’’. It is important to point out that from
Eqs. (7) and (8) it is possible to calculate the critical reduced wind speed (U Rc ¼ 2p=K c ,
where c stands for ‘‘critical’’), the coupling frequency and consequently the dimensional
flutter wind speed, which depend only on the aerodynamic properties (flutter derivatives)
and the previously mentioned nondimensional dynamic parameters (m, ra , go , gh , ga ).
An important limit of this simple approach is the effect of large-scale turbulence present
in the oncoming wind flow, which would require a stochastic stability analysis (Bucher and
Lin, 1988a, b, 1989; Bartoli et al., 1995) and which is neglected here. In addition, the
usually remarkable dependence of flutter derivatives on the mean angle of attack and the
presence of vertical gusts in natural wind would require a nonlinear analysis, such as those
outlined in Diana et al. (1999, 2005) or Chen and Kareem (2000). Nevertheless, concerning
flutter stability, the cited authors themselves claim that a linearized approach with respect
to the mean angle of twist due to static wind is reasonable. Albeit very important, these
issues are out of the scope of the present work and will not be accounted for in the
proposed simplified method.

4. Simplification procedure

In order to simplify the flutter equations it can be noted in Eq. (7) that gh ga 51, since
hardly the damping coefficients are higher than 0.02. Then, isolating in the left-hand side
of both equations the term r2a m2 X 2 =g2o and then equating the right-hand sides, an equation
of first order in X is obtained:
 2 2 2       
r a m g o gh þ ga mA3 ga mA2 1 2  gh
þ 2  2 gh þ þ ra mH 4
g2o gh þ ga g o gh þ ga go gh þ ga gh þ ga
   2 
1 mA r mH
r2a mH 1 ga þ X  r2a m2  mA3 þ 2
 r2a mH 4 þ a 1
gh þ ga gh þ ga gh þ ga
     
H A þ H A  H A  H 2 A4
 H 4 A3 þ H 1 A2 þ H 3 A4  H 2 A1 þ 4 2 1 3 3 1
¼ 0. ð9Þ
gh þ ga
In Eq. (9) the terms which contain a second power of the damping coefficients can be
neglected, giving:
 2 2 2     
r a m g o gh þ ga m ga A3 A2 2 gh H 4 H 1
þ  þ r a m  X
g2o gh þ ga g2o gh þ ga gh þ ga gh þ ga gh þ ga
mA2 r2 mH 
 r2a m2  mA3 þ  r2a mH 4 þ a 1  H 4 A3 þ H 1 A2 þ H 3 A4  H 2 A1
gh þ ga gh þ g a
H 4 A2 þ H 1 A3  H 3 A1  H 2 A4
þ ¼ 0. ð10Þ
gh þ ga
It is worth noting that Eq. (9) relies on the assumption that gh þ ga a0. The case gh ¼
ga ¼ 0 will be discussed in Section 4.2.

4.1. Range of variability of dynamic parameters

Now it is important to evaluate the range of variability of the nondimensional dynamic

parameters which appear in the flutter equations. For this reason they are reported in
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 237

Table 2
Main geometric and dynamic properties of the considered suspension bridges, cable-stayed bridges and
footbridges (assuming r ¼ 1:25 kg=m3 )

L (m) B (m) f h (Hz) f a (Hz) m (kg/m) I ðkg m2 =mÞ go (–) ra (–) m (–) r2a m (–) r2a m2 (–)

Suspended-span bridges
Tacoma 854 12.0 0.130 0.200 4250 177 730 1.54 0.539 47.2 13.7 647.6
Bosporus 1074 28.0 0.162 0.371 13 550 1 351 645 2.29 0.357 27.7 3.5 97.3
Akashi 1991 35.5 0.064 0.150 43 790 9 826 000 2.34 0.422 55.6 9.9 550.3
Messina 3300 60.4 0.0605 0.0796 55 000 28 000 000 1.32 0.374 24.1 3.4 81.2

Cable-stayed bridges
Indiano 189 22.4 0.573 1.179 17 401 545 274 2.06 0.250 55.5 3.5 192.3
Guamà 320 14.2 0.331 0.649 22 513 566 838 1.96 0.353 178.6 22.3 3984.8
Tsurumi 510 38.0 0.204 0.486 32 220 2 880 100 2.38 0.249 35.7 2.2 78.9
Normandy 856 23.8 0.220 0.500 13 700 633 488 2.27 0.286 38.7 3.2 122.2
Footbridges
Siena 59.4 3.255 1.735 4.839 1420 810 2.79 0.232 214.4 11.5 2475.8
Turin 156 6.91 0.510 0.590 3347 13 502 1.16 0.291 112.2 9.5 1062.7

Table 2 for some bridge decks with various characteristics. Four suspended-span bridges of
different dimensions and cross sections are taken into account: First Tacoma Narrows
Bridge, USA (Larsen, 1998), First Bosporus Strait Bridge, Turkey (Martini, 2004), Akashi
Kaikyo Bridge, Japan (Miyata and Yamaguchi, 1993) and the proposed Messina Strait
Bridge, Italy (D’Asdia and Sepe, 1998). Table 2 shows the main geometric and dynamic
properties, as well as the nondimensional parameters appearing in the flutter equations
(L ¼ main span length; f h , f a ¼ vertical bending and torsional frequencies susceptible to
couple). The same data can be found in the second part of Table 2 for four cable-stayed
bridges: Ponte all’Indiano, Italy (Augusti et al., 1980), Rio Guamà Bridge, Brazil (Rocha
et al., 2003), Tsurumi Fairway Bridge, Japan (Sarkar, 1992) and Normandy Bridge, France
(ONERA, 1989). Finally, in the third part of the table, two footbridges are considered:
Ruffolo footbridge near Siena, Italy (Bartoli and Procino, 2004) and the footbridge of the
Olympic Arch in the Olympic Village in Turin, Italy (Flamand, 2003).
From the analysis of the last three columns of Table 2 [in view of the fact that ra never
appears alone in Eqs. (7)–(10)], it seems reasonable to consider as opposite limit cases
Tsurumi Fairway Bridge and Rio Guamà Bridge, whose dynamic properties will be
assumed as reference in the following analyses. In particular, Rio Guamà Bridge is
characterized by a concrete deck and consequently by a remarkable mass as compared to
the chord B. In contrast, Tsurumi Fairway Bridge presents a quite light steel deck with a
large chord length. The nondimensional mass and radius of inertia of the proposed
Messina Strait Bridge are close to those of Tsurumi Bridge. Conversely flexible footbridges
seem to be characterized by values of m and ra rather similar to those of Rio Guamà
Bridge, since they usually present light but narrow decks.

4.2. Simplified formula for critical frequency

In order to draw as general as possible conclusions, it is important to consider a large

number of dynamic and aerodynamic data. Assuming that flutter derivatives depend on
 ARTICLE IN PRESS
238 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

reduced wind speed and cross-section geometry only, it is possible to combine the
aerodynamics of a bridge with the dynamic properties of a completely different one,
performing calculations on this ‘‘hybrid’’ structure. Obviously the results have nothing
to do with a real bridge, referring to a non-existent but possible structure. This obser-
vation is important because it allows to employ for calculations all the reliable
aerodynamic and dynamic data available to the authors and consider the two reference
dynamics highlighted in the previous section, denoted in the following by means of their m
and ra values.
From the aerodynamic point of view let us take into account the behavior of two cross
sections quite different from each other: the theoretical thin flat plate (Theodorsen, 1934;
Fung, 1993; Bisplinghoff et al., 1996) and a rectangular cylinder with a chord-to-thickness
ratio of 12.5 (R12.5) (Matsumoto, 1996). In particular, it is worth noting that R12.5 is the
bluffer rectangular cylinder studied in Matsumoto (1996) which does not show negative
values of the flutter derivative A2 , that is tendency to torsional flutter.
For several values of the reduced wind speed U R , the terms appearing in Eq. (10) are
compared in Tables 3 and 4 for the flat plate case and in Tables 5 and 6 for the rectangular
cylinder case, setting:

mA2 r2 mH 
S 1 ¼ r2a m2  mA3 þ þ a 1 , (11)
gh þ ga gh þ ga

S 2 ¼ r2a mH 4 , (12)

S 3 ¼ H 4 A3 þ H 1 A2 þ H 3 A4  H 2 A1 , (13)

H 4 A2 þ H 1 A3  H 3 A1  H 2 A4

S4 ¼ . (14)
gh þ ga
Since it can be claimed that gh and ga hardly take values larger than 0.02, damping
coefficients of 0.001 and 0.03 are taken into account in order to have as general as possible
results.

Table 3
Comparison of the weight of the terms appearing in Eq. (10) (dynamics: m ¼ 35:7, ra ¼ 0:249; aerodynamics: thin
flat plate)

UR gh ¼ ga ¼ 0:001 gh ¼ ga ¼ 0:03 H 1 A3 H 3 A1

S1 S2 S3 S4 S1 S2 S3 S4

0 80.7 3.5 0.1 0.0 80.7 3.5 0.1 0.0 0.00 0.00
2.5 3236.3 3.0 0.1 73.8 190.8 3.0 0.1 2.5 0.26 0.19
5.0 8036.8 2.0 0.8 70.0 367.0 2.0 0.8 2.3 1.90 1.76
7.5 14931.2 0.8 2.0 31.0 626.1 0.8 2.0 1.0 7.01 6.79
10.0 23864.9 0.4 3.9 225.1 967.7 0.4 3.9 7.5 18.29 17.97
15.0 47026.5 2.8 9.9 853.5 1874.1 2.8 9.9 28.5 72.11 71.58
20.0 75881.5 4.8 18.9 1734.5 3036.0 4.8 18.9 57.8 190.93 190.18
25.0 109134.7 6.6 30.9 2811.1 4411.8 6.6 30.9 93.7 404.12 403.13
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 239

Table 4
Comparison of the weight of the terms appearing in Eq. (10) (dynamics: m ¼ 178:6, ra ¼ 0:353; aerodynamics: thin
flat plate)

UR gh ¼ ga ¼ 0:001 gh ¼ ga ¼ 0:03 H 1 A3 H 3 A1

S1 S2 S3 S4 S1 S2 S3 S4

0 3993.7 35.0 0.1 0.0 3993.7 35.0 0.1 0.0 0.00 0.00
2.5 27206.2 30.3 0.1 73.8 4792.4 30.3 0.1 2.5 0.26 0.19
5.0 59957.5 20.0 0.8 70.0 5965.1 20.0 0.8 2.3 1.90 1.76
7.5 104416.9 7.9 2.0 31.0 7593.5 7.9 2.0 1.0 7.01 6.79
10.0 160026.2 4.5 3.9 225.1 9666.2 4.5 3.9 7.5 18.29 17.97
15.0 299667.3 28.0 9.9 853.5 14993.5 28.0 9.9 28.5 72.11 71.58
20.0 469410.4 48.5 18.9 1734.5 21652.3 48.5 18.9 57.8 190.93 190.18
25.0 662104.0 66.3 30.9 2811.1 29413.3 66.3 30.9 93.7 404.12 403.13

Table 5
Comparison of the weight of the terms appearing in Eq. (10) (dynamics: m ¼ 35:7, ra ¼ 0:249; aerodynamics:
rectangular cylinder R12.5)

UR gh ¼ ga ¼ 0:001 gh ¼ ga ¼ 0:03 H 1 A3 H 3 A1

S1 S2 S3 S4 S1 S2 S3 S4

0 78.9 0.0 0.0 0.0 78.9 0.0 0.0 0.0 0.00 0.00
7.4 11587.2 1.2 0.6 459.6 513.8 1.2 0.6 15.3 7.33 6.53
11.1 18811.5 1.1 0.0 170.6 819.3 1.1 0.0 5.7 30.49 29.59
14.8 25319.4 1.9 2.5 1137.9 1118.6 1.9 2.5 37.9 74.96 75.32
18.5 29898.6 4.7 6.8 5110.3 1378.3 4.7 6.8 170.3 156.95 163.73
22.2 41134.0 6.1 11.4 6563.7 1892.0 6.1 11.4 218.8 286.76 296.46

Table 6
Comparison of the weight of the terms appearing in Eq. (10) (dynamics: m ¼ 178:6, ra ¼ 0:353; aerodynamics:
rectangular cylinder R12.5)

UR gh ¼ ga ¼ 0:001 gh ¼ ga ¼ 0:03 H 1 A3 H 3 A1

S1 S2 S3 S4 S1 S2 S3 S4

0 3984.9 0.0 0.0 0.0 3984.9 0.0 0.0 0.0 0.00 0.00
7.4 89326.4 12.2 0.6 459.6 7086.3 12.2 0.6 15.3 7.33 6.53
11.1 148732.1 11.0 0.0 170.6 9390.2 11.0 0.0 5.7 30.49 29.59
14.8 203653.2 19.6 2.5 1137.9 11632.8 19.6 2.5 37.9 74.96 75.32
18.5 252944.4 47.0 6.8 5110.3 13811.8 47.0 6.8 170.3 156.95 163.73
22.2 334614.4 61.3 11.4 6563.7 17230.5 61.3 11.4 218.8 286.76 296.46

It is evident that the terms S2 , S 3 and S4 in Eq. (10) can be neglected with respect to S1 .
In addition, it is important to highlight that:
H 1 A3  H 3 A1 ffi 0. (15)
This fact is in agreement with two of the relationships between flutter derivatives discussed
in Matsumoto et al. (1995) and in Matsumoto (1996) for rectangular cylinders and for
 ARTICLE IN PRESS
240 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

three bridge decks in Scanlan et al. (1997), which seem not to be contradicted by the flutter
derivatives of a wide range of cross-sectional geometries:
H 1 ffi KH 3 , (16)

A1 ffi KA3 . (17)

Finally, an expression for the nondimensional frequency X is obtained, depending only on
the three flutter derivatives H 1 , A2 and A3 :
 2 2 2    
r a m g o gh þ ga m ga A3 A2 r2a mH 1
þ   X  r2a m2  mA3
g2o gh þ ga g2o gh þ ga gh þ ga g h þ ga
mA2 r2 mH 
þ þ a 1 ¼ 0, ð18Þ
gh þ ga gh þ ga

r2a mðgh þ ga Þ þ A3 ðgh þ ga Þ  A2  r2a H 1

X ¼ g2o . (19)
r2a mðg2o gh þ ga Þ þ ga A3  A2  g2o r2a H 1
If damping is neglected, the formula simplifies in the following way:
A2 þ r2a H 1
X ¼ g2o . (20)
A2 þ g2o r2a H 1
For cross sections prone to coupled flutter it generally results that H 1 o0, A2 o0 and
A3 40. As go 41, being the damping coefficient gh small enough, it is easy to show that, as
expected:
o2a
1pX ¼ pg2o . (21)
o2
The good degree of approximation of this formula can be appreciated by supposing to
know the critical reduced wind speed, at which the flutter derivatives are evaluated, and
comparing the results given by the rigorous eigenvalue analysis and by Eq. (19). Zero
damping and high damping ratios (zh ¼ za ¼ 1%) are assumed for both bending and
torsional modes, in order to better stress the role of this parameter. The results for the four
test cases previously analyzed are reported in Table 7, showing the accuracy of the given
formula in estimating the flutter critical frequency.

Table 7
Comparison between the critical frequency calculated through the eigenvalue analysis and with Eq. (19),
supposing to know the critical reduced wind speed

Aerodyn. Dynamics zh ¼ za ¼ 0 zh ¼ za ¼ 0:01

m (–) ra (–) U Rc (–) f eig

c (Hz)
f approx
c (Hz) Df c (%) U Rc (–) f eig
c (Hz)
f approx
c (Hz) Df c (%)

Flat plate 35.7 0.249 10.56 0.3178 0.3161 0.53 10.93 0.3123 0.3102 0.67
Flat plate 178.6 0.353 25.32 0.4649 0.4639 0.21 26.51 0.4551 0.4541 0.22
R12.5 35.7 0.249 7.80 0.3667 0.3752 +2.32 8.25 0.3589 0.3646 +1.59
R12.5 178.6 0.353 14.04 0.5841 0.5824 0.29 17.32 0.5580 0.5555 0.45
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 241

The previous formula is based on the assumption that gh þ ga a0, used to obtain
Eq. (9). Nevertheless, if gh ¼ ga ¼ 0 the imaginary flutter equation [Eq. (8)] largely
simplifies and the result of Eq. (20) can be derived directly. This means that the
information about the critical frequency is mainly contained in the imaginary flutter
equation, although it is not possible to obtain directly a closed-form solution for X , unless
damping is neglected.

4.3. Simplified formula for critical reduced wind speed

Since the result of the previous section is mainly derived from the imaginary flutter
equation, let us come back now to the real flutter equation [Eq. (7)], retaining only the
trivial simplification 1  gh ga  1 and rearranging it in the following way:
 
X X
½mA3  ra m ðX  1Þ  1  2  r2a mH 4 ðX  1Þ þ ðgh mA2 þ ga g2o r2a mH 1 Þ 2
 2 2
go go
þ H 4 A3  H 1 A2  H 3 A4 þ H 2 A1 ¼ 0. ð22Þ
By defining:
 
X
T1 ¼ ½mA3  r2a m2 ðX  1Þ  1  2 , (23)
go

T 2 ¼ r2a mH 4 ðX  1Þ, (24)

X
T 3 ¼ ðgh mA2 þ ga g2o r2a mH 1 Þ , (25)
g2o

T 4 ¼ H 4 A3  H 1 A2  H 3 A4 þ H 2 A1 . (26)

Eq. (22) can be written in the form:
T 1 þ T 2 þ T 3 þ T 4 ¼ 0. (27)
The term T 3 vanishes if zero damping is assumed. It is interesting to understand which is
the actual weight of the four terms in the equation. Therefore, each term and their sum are
plotted against U R in Fig. 2, for different values of X in the interval 1pX pg2o for the four
reference cases. Unrealistically high damping coefficients are assumed (gh ¼ ga ¼ 0:05) in
order to stress the role of the term T 3 . It is possible to observe that the terms T 2 , T 3 and T 4
are definitely negligible with respect to T 1 .
0 ¼ T 1 þ T 2 þ T 3 þ T 4 ffi T 1. (28)
In addition, it is also expected that X ago , that is oaoh , so that the critical condition
becomes:
A3  r2a mðX  1Þ ¼ 0. (29)
Finally, the expression for X given by Eq. (19) is substituted in this equation, leading to an
approximate formula for the flutter critical reduced wind speed:
ga ðA3 Þ2 þ ga r2a mA3 ð2  g2o Þ  A2 A3  g2o r2a H 1 A3 þ r2a mA2 ðg2o  1Þ  r4a m2 ga ðg2o  1Þ ¼ 0.
(30)
 ARTICLE IN PRESS
242 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

Flate plate − X = 1.20 (μ = 35.7 rα = 0.249) Flate plate − X = 1.80 (μ = 35.7 rα = 0.249)
600 450
400
500
350
400 300
300 250
200
200 150
100 100
50
0
0
−100 −50
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

R12.5 − X = 1.20 (μ = 35.7 rα = 0.249) R12.5 − X = 1.80 (μ = 35.7 rα = 0.249)

400 300
350 250
300
200
250
200 150
150 100
100
50
50
0 0

−50 − 50
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

Flate plate − X = 1.15 (μ = 178.6 rα = 0.353) Flate plate − X = 1.55 (μ = 178.6 rα = 0.353)
2500 1000

2000 500
1500
0
1000
−500
500

0 −1000

−500 −1500
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

R12.5 − X = 1.55 (μ = 178.6 rα = 0.353) R12.5 − X = 1.55 (μ = 178.6 rα = 0.353)

1200 200
1000 0
800 −200
600
−400
400
−600
200
−800
0
−200 −1000
−400 −1200
−600 −1400
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

Fig. 2. Comparison between the four terms composing the real flutter equation [Eq. (22)].
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 243

If damping is neglected, the equation simplifies as follows:

A2 A3 þ g2o r2a H 1 A3  r2a mA2 ðg2o  1Þ ¼ 0. (31)

Eqs. (30) and (19) constitute an approximate solution of the flutter problem which depends
only on dynamic parameters (ra , m, go , gh and ga ) and three aeroelastic functions (H 1 , A2 ,
A3 ), thus representing a remarkable simplification. In addition, this approximate solution
allows to calculate easily, practically by hand, the reduced critical wind speed [from
Eq. (30)] and flutter frequency [from Eq. (19)].

5. Further simplification

For a theoretical flat plate the three flutter derivatives involved in the proposed
approximate formulas can be expressed by means of the Theodorsen’s Circulatory
Function CðkÞ ¼ F ðkÞ þ iGðkÞ (Fung, 1993; Bisplinghoff et al., 1996; Simiu and Scanlan,
1996):

F ðkÞ
H 1 ðKÞ ¼ p , (32)
k
 
p 2GðkÞ
A2 ðKÞ ¼ 1  F ðkÞ  , (33)
16k k

 
p k2 kGðkÞ
A3 ðKÞ ¼ 2 þ F ðkÞ  , (34)
8k 8 2

where, as usual, K ¼ oB=U and k ¼ K=2. Since there are three functions expressed by
means of two independent ones, a relationship between them can be found:

p kA2 ðKÞ 4 þ k2 
A3 ðKÞ ¼    H 1 ðKÞ. (35)
64 2 32k
This relationship is exact for a theoretical flat plate but seems to apply more or less
reasonably also for a wide range of real deck sections. As a matter of fact, in order to check
its validity, the flutter derivatives of two rectangular cylinders characterized by chord-to-
thickness ratios of 12.5 and 20 (Matsumoto, 1996), Tsurumi Fairway Bridge (Singh et al.,
1995) and Akashi Kaikyo Bridge (Katsuchi et al., 1999: modified cross section) are taken
into account, as shown in Fig. 3. It is evident that the relationship between A3 , A2 and H 1 ,
found for the thin flat plate, is also valid for rectangular cross sections with large chord-to-
thickness ratio (R20 for instance) and for streamlined box-girder sections (Tsurumi
Fairway Bridge), whose aerodynamic behavior is supposed to be not very different from
that of an airfoil. The equation is less accurate for bluffer rectangular cylinders (R12.5)
and for truss-stiffened girders (Akashi Kaikyo Bridge) but the approximation seems to be
still acceptable for some applications. Eq. (35) with Eqs. (19) and (30) constitute a further
simplified system to calculate the flutter critical frequency and reduced wind speed,
retaining only the flutter derivatives H 1 and A2 .
 ARTICLE IN PRESS
244 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

R12.5 R20
20 18
∗
18 A∗3 A3
16
A∗3= f(H ∗1,A∗2) A∗3= f(H ∗1,A∗2)
16 14
14
12
12
10
10
8
8
6
6
4 4
2 2
0 0
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

Tsurumi Fairway Bridge Akashi Kaikyo Bridge

2.5 8
A∗3 A∗3
7 A∗3= f(H ∗1,A∗2)
A∗3= f(H ∗1,A∗2)
2
6

1.5 5

4
1 3

2
0.5
1

0 0
0 2 4 6 8 10 0 5 10 15 20 25 30
UR UR

Fig. 3. Degree of approximation of Eq. (35) in the case of two rectangular cylinders and two bridge sections.

6. Validation of the method

In order to understand the degree of approximation and the limits of the proposed
approach, the computation of the critical reduced wind speed is repeated for the previous
four case studies for different damping levels and frequency ratios go . The results of this
analysis are reported in Fig. 4. Some first conclusions can be drawn from these graphs. The
approximate formulas give results which are very close to those obtainable from the
rigorous solution of the eigenvalue problem of stability for a wide range of frequency
ratios both in case of zero and high structural damping. Conversely, the main limit of the
proposed approach consists in its inability to give reasonable results when the frequency
ratio go tends to unity. It is known that when the still-air torsional frequency is very close
to the vertical bending frequency, the critical reduced wind speed tends to become very
high (Dyrbye and Hansen, 1997). The simplified approach is not able to reproduce this
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 245

Aerodynamics: Flat plate Aerodynamics: Flat plate

Dynamics: μ = 35.7, rα = 0.249 Dynamics: μ = 178.6, rα = 0.353
18 45
16 40
14 35
12 30
10 25
URc

URc
8 20
ξh=ξα=0%; eig. sol. ξh=ξα=0%; eig. sol.
6 15
ξh=ξα=1%; eig. sol. ξh=ξα=1%; eig. sol.
4 ξh=ξα=0%; approx. sol. 10 ξh=ξα=0%; approx. sol.
ξh=ξα=1%; approx. sol. ξh=ξα=1%; approx. sol.
2 5
0 0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
γω γω
Aerodynamics: R12.5 Aerodynamics: R12.5
Dynamics: μ = 35.7, rα = 0.249 Dynamics: μ = 178.6, rα = 0.353
14 22
20
12
18
10 16
14
8 12
URc

URc

ξh=ξα=0%; eig. sol. 10 ξh=ξα=0%; eig. sol.

6 ξh=ξα=1%; eig. sol. ξh=ξα=1%; eig. sol.
8
ξh=ξα=0%; approx. sol. I ξh=ξα=0%; approx. sol. I
4 6
ξh=ξα=1%; approx. sol. I ξh=ξα=1%; approx. sol. I
ξh=ξα=0%; approx. sol. II 4 ξh=ξα=0%; approx. sol. II
2
ξh=ξα=1%; approx. sol. II 2 ξh=ξα=1%; approx. sol. II
0 0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
γω γω

Fig. 4. Comparisons between the solution of the eigenvalue problem of flutter stability (eig. sol.) and that of
Eqs. (19) and (30) (approx. sol. I) or Eqs. (19), (30) and (35) (approx. sol. II) for different values of the frequency
ratio go and two structural damping ratios (0% and 1%).

behavior, as it could have been expected from the analysis of the expression of T 1 in
Eq. (23): when go ! 1, X ! 1 as well, so that T 1 tends to vanish instead of being the
dominant term in the real flutter equation. Nevertheless, it seems that a frequency ratio of
about 1.3, or sometimes even less, can be sufficient to obtain an acceptable degree of
approximation. Frequency ratios very close to unity are fairly uncommon and usually
characterize super-long-span bridges (like the proposed Messina Strait Bridge) or non-
conventional structures, which are not expected to be analyzed with simplified methods,
requiring careful experimental campaigns. In addition, when go ! 1 in the case of realistic
deck geometries the eigenvalue approach gives often doubtful results: the problem
sometimes shows many solutions close to each other, also for very low reduced wind speed,
or in other cases no solution at all is found.
It is worth noting that the use of the approximate expression for A3 given by Eq. (35)
(approx. sol. II) instead of experimental values (approx. sol. I) does not imply significant
 ARTICLE IN PRESS
246 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

Aerodynamics: Flat plate Aerodynamics: Flat plate

Dynamics: μ = 35.7, rα = 0.249 Dynamics: μ = 178.6, rα = 0.353
300 500
450
250 400
200 350

Uc [m/s]
Uc [m/s]

300
150 250
200
100 150
50 100
50
0 0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
γω [−] γω [−]

Fig. 5. Comparison between the results given by authors’ formulas in the case of a thin flat plate and Selberg and
Rocard’s formulas.

differences in the results for the rectangular cylinder R12.5 when the mass parameters
m ¼ 35:7 and ra ¼ 0:249 are assumed. Conversely, assuming m ¼ 178:6 and ra ¼ 0:353,
reductions up to 10% of the reduced wind speed are observed. Obviously, there is no
distinction between the two approximate solutions when the aerodynamics of the flat plate
is considered.
Finally, concerning the role played by structural damping in the flutter instability onset,
it is difficult to outline clear conclusions but it seems that for streamlined cross sections [see
Figs. 4(a) and (b)] its effect is definitely marginal, especially for frequency ratios far from
unity, whereas for bluffer cross sections [see Figs. 4(c) and (d), 9(a) and (b)] it can be
quantitatively relevant depending on the mass parameters. For this reason, the fact that
in the simplified formulas damping is not neglected, is very important [see in particular
Figs. 4(d) and 9(b)]. It was observed in Chen and Kareem (2006a) and Chen (2007) that the
role of structural damping can be particularly significant for ‘‘soft-type flutter’’ in which
negative modal damping builds up slowly with increasing wind speed. This could be a
possible explanation for these results but further investigations are underway on this issue.
For a thin flat plate the results given by the proposed formulas are very similar to those
predicted by Selberg (1961) and Rocard’s (Frandsen, 1966) empirical formulas, which are
reported here with the notation adopted in the present paper:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
1
U c ¼ 2:623f a B 1  2 ra m ½Selberg, (36)
go

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  2 ffi
1 ra m
U c ¼ 6:282f a B 1 2 ½Rocard. (37)
go 1 þ 8r2a

As it can be seen in Fig. 5, all these approaches are not able to reproduce the sudden
increase in the critical wind speed appearing when the frequency ratio tends to unity,
whereas for larger values of go they all give excellent results, although authors’ formulas
also account for the contribution of structural damping. The proposed method seems to
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 247

Table 8
Case studies (coupled flutter)

Case Aerodynamics (–) m (–) ra (–) go (–) zh (%) za (%)

1 Flat plate 35.7 0.249 2.38 0.5 0.5

2 Flat plate 24.1 0.374 1.32 0.6 0.7
3 R12.5 178.6 0.353 1.96 0.8 0.8
4 R12.5 27.7 0.357 2.29 0.5 0.5
5 Akashi Kaikyo 55.6 0.422 2.34 0.5 0.3
6 Tsurumi 35.7 0.249 2.38 0.5 0.5
7 Tsurumi 55.5 0.250 2.06 0.2 0.2
8 R14.3F 47.2 0.539 1.54 0.5 0.5
9 R14.3F 27.7 0.357 2.29 0.5 0.5
10 R20 38.7 0.286 2.27 0.2 0.5

Table 9
Results for the case studies listed in Table 8 (f c is the coupling frequency, U c is the flutter wind speed)

Case eig. sol. approx. sol. I approx. sol. II

f c (Hz) U c (m/s) f c (Hz) Df c (%) U c (m/s) DU c (%) f c (Hz) Df c (%) U c (m/s) DU c (%)

1 0.315 128.7 0.314 0.2 125.2 2.7  – – –

2 0.071 28.9 0.072 +0.8 23.6 18.3 – – – –
3 0.562 134.2 0.560 0.4 136.5 +1.7 0.553 1.6 119.4 11.0
4 0.308 66.0 0.313 +1.4 62.1 5.9 0.313 +1.5 62.3 5.5
5 0.138 79.2 0.137 0.6 79.3 +0.1 0.137 0.9 68.6 13.3
6 0.377 119.5 0.349 7.6 133.4 +11.6 0.351 6.9 130.9 +9.5
7 0.880 217.6 0.834 5.2 238.8 +9.8 0.832 5.5 241.7 +11.1
8 0.180 24.6 0.179 0.2 22.8 7.2 0.179 0.3 21.4 12.9
9 0.284 80.9 0.282 0.8 74.2 8.3 0.282 1.0 69.8 13.7
10 0.338 90.2 0.339 +0.4 87.7 2.8 0.338 0.0 92.4 +2.5

represent an extension of Selberg and Rocard’s formulas to generic bridge sections. As a

matter of fact, these formulas apply only for a flat plate and their use can be extended to
realistic bridge decks only if the critical flutter wind speed is multiplied by an empirical
coefficient accounting for the actual aerodynamics (Aerodynamic Stability Performance
Index), as already discussed in the second section of this paper.

7. Case studies

In order to further validate the proposed approach, other case studies are taken into
account. Some of those reported in Table 8 refer to existing structures, such as Case 5 for
Akashi Kaikyo Bridge (Miyata and Yamaguchi, 1993; Katsuchi et al., 1999: modified cross
section) and Case 6 for Tsurumi Fairway Bridge (Sarkar, 1992; Singh et al., 1995); the
others, as already explained, are combinations of the dynamic properties of some bridges
(see Table 2) and the flutter derivatives measured for a few cross sections prone to coupled
flutter: thin flat plate (Fung, 1993; Bisplinghoff et al., 1996), R12.5 and R20 rect-
angular cylinders (Matsumoto, 1996), rectangular cylinder with semi-circular fairings and
 ARTICLE IN PRESS
248 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

Aerodynamics: Flat plate

Dynamics: μ = 35.7, rα = 0.249 Akashi Kaikyo Bridge
8 10
eig. sol.
eig. sol. approx. sol. I
approx. sol. I
8
approx. sol. II
6
6
4 4

2 2

0
0
−2
−2
−4

−4 −6
0 2 4 6 8 10 12 14 0 5 10 15 20
URc URc

Fig. 6. Graphical solution of the approximate formula for flutter reduced wind speed [Eq. (30)] for Cases 1 and 5
(Akashi Kaikyo Bridge) of Tables 8 and 9, considering measured A3 (approx. sol. I) and approximate values given
by Eq. (35) (approx. sol. II).

chord-to-thickness ratio of 14.3, here called R14.3F (Chowdhury and Sarkar, 2004), and
Tsurumi Fairway Bridge (Singh et al., 1995). All case studies are listed in Table 8 along
with the dynamic properties assumed in the calculations. In Table 9 the flutter wind speeds
and the coupling frequencies calculated through the eigenvalue analysis and the proposed
formulas [approx. sol. I: Eqs. (19) and (30); approx. sol. II: Eqs. (19), (30) and (35)] are
compared. In Fig. 6 the graphical solution of Eq. (30) is presented for the flat plate of Case
1 and for the Akashi Kaikyo Bridge (Case 5). For the latter the proposed results have to be
interpreted merely as an idealized case study because the existing structure is unique and
the multimodal analysis cannot be waived, due to the important role played by the drag
self-excited force and in particular by the component induced by the torsional motion
(Miyata et al., 1994; Katsuchi et al., 1999).
From Table 9 it can be concluded that the simplified formulas give accurate results
in most cases, especially if experimental values of A3 are considered (approx. sol. I). In
some instances, the results are practically the same as those given by the solution of
the eigenvalue problem of stability. The formula for the coupling frequency is parti-
cularly accurate. Nevertheless, when the frequency ratio is small (e.g. Case 2), as ex-
pected, the accuracy of the approximation is poorer. The approximate formulas
retaining only two flutter derivatives (H 1 and A2 ) are less performing than the ones
adopting three experimental functions but the degree of approximation seems to be still
acceptable.
Finally, it is worth reminding that a significant scatter usually affects flutter derivatives,
especially if they are measured through the free-vibration technique (Righi, 2003; Bartoli
and Mannini, 2005b; Mannini, 2006). Moreover, non-negligible differences in the results
are just due to the chosen analytical interpolation of the experimental data. These effects
are in many cases of the same order of magnitude as the errors highlighted in Table 9, thus
confirming the validity of the present approach.
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 249

8. Torsional flutter

All the cross sections taken into account insofar are prone to classical flutter, as
confirmed by the monotonic negative trend of H 1 and A2 shown in the given references.
Nevertheless, it is well known that a wide range of bridges, whose deck cross sections are
not streamlined at all, are prone to torsional flutter. For these bridge decks the function A2
tends to reverse its sign from negative to positive at relatively low reduced wind speed,
introducing negative aerodynamic damping in the torsional modes. Therefore, the
uncoupled self-excited forces are mainly responsible for the instability.
In this case, a well-known simplified way to deal with flutter is to perform a mode-by-
mode analysis neglecting the intermodal aerodynamic coupling, i.e. just considering the
frequency shift due to A3 and the damping reduction produced by A2 .
 1=2
rB4 
fc ¼ fa 1þ A3 ðK c Þ , (38)
2I

I fa
A2 ðK c Þ ¼ 4za 
4 f ðK Þ
, (39)
rB c c

where K c ¼ U c =ðB2pf c Þ is the critical reduced frequency, U c is the critical wind speed, f c
the critical frequency. Eqs. (38) and (39) constitute a nonlinear system (1-DoF problem)
which has to be iteratively solved. The 2-DoF and 1-DoF approaches usually give close
results, unless the frequency ratio is very near to unity. Nevertheless, it was observed in
Chen and Kareem (2006a) that the coupling of heaving and pitching motions generally
reduces the critical wind speed even in the case of torsional flutter.
Although torsional flutter is just a particular case of binary coupled flutter, most of the
simplifications discussed in Sections 4 and 5 are based on the aerodynamic properties of
bridge decks prone to coupled flutter. Therefore, all the procedure has to be checked again
for torsional flutter, showing in this case the flutter derivatives a very different pattern. As
reference examples two prototypes characterized again by the dynamic properties of
Tsurumi Fairway Bridge (m ¼ 35:7, ra ¼ 0:249) and Rio Guamà Bridge (m ¼ 178:6,
ra ¼ 0:353) and by the aerodynamics of a rectangular cylinder with chord-to-thickness
ratio of 5.0 (Matsumoto, 1996), are considered. The simplifications standing behind the
approximate formula for critical frequency [Eq. (19)], discussed in Tables 3–6 for cross
sections prone to coupled flutter, are still valid if the analysis is limited to reduced wind
speeds in the neighborhood of the zero-crossing point of A2 , since it is in that range that
the critical flutter speed has to be sought. Tables are not reported here for the sake of
brevity. In Fig. 7 the weight of the terms constituting the real flutter equation [Eq. (22)] are
compared for values of X not far from unity, since torsional flutter always arises with
frequency close to the still-air torsional one. It can be understood from the presented
graphs that, beyond a certain value of X , no solution is found. In Fig. 8 the relation
between the flutter derivatives H 1 , A2 and A3 [Eq. (35)] is checked for the aforementioned
rectangular cylinder R5 and for the First Tacoma Narrows Bridge (Scanlan and Tomko,
1971). The relation assumed between the three aeroelastic coefficients is based on the
behavior of the thin flat plate and therefore is less effective in the case of a bluff cross
section. Nevertheless, Fig. 8 shows that this approximation does not seem completely
 ARTICLE IN PRESS
250 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

R5 − X = 1.00 (μ = 35.7 rα = 0.249) R5 − X = 1.75 (μ = 35.7 rα = 0.249)

80 20
60 0
40
−20
20
−40
0
−60
−20
T1 T1
T2 −80 T2
−40
T3 T3
−60 T4 −100 T4
T1+T2+T3+T4 T1+T2+T3+T4
−80 −120
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

R5 − X = 1.00 (μ = 178.6 rα = 0.353) R5 − X = 1.08 (μ = 178.6 rα = 0.353)

300 100

200 0

100 −100

0 −200

−100 −300
T1 T1
−200 T2 −400 T2
T3 T3
−300 T4 −500 T4
T1+T2+T3+T4 T1+T2+T3+T4
−400 −600
0 5 10 15 20 25 0 5 10 15 20 25
UR UR

Fig. 7. Comparison between the four terms composing the real flutter equation [Eq. (22)] for the rectangular
cylinder R5.

wrong in the range of reduced wind speeds of interest, that is around A2 zero-crossing
point.
In Fig. 9 the critical reduced wind speed is plotted against the frequency ratio for the two
reference sets of mass parameters. It is possible to remark that the critical reduced wind
speed is independent of the frequency separation, unless the ratio go is quite close to unity.
In this region the curves lose the flat trend and the 2-DoF eigenvalue approach gives results
that diverge from those obtained through the 1-DoF equations. Conversely, the 1-DoF
and 2-DoF approaches usually give close solutions for frequency ratios far enough from
unity. For a structure characterized by m ¼ 178:6 and ra ¼ 0:353 [Fig. 9(b)], the agreement
between 2-DoF and 1-DoF solutions is nearly perfect provided that go is greater than
approximately 1.5. It can be also remarked the strong dependence on structural damping,
as already experienced in Fig. 4(d). On the other hand, for a structure characterized by
m ¼ 35:7 and ra ¼ 0:249 [Fig. 9(a)], a much lower influence of damping and a larger
discrepancy between the 1-DoF and 2-DoF solutions can be observed. As previously said,
the reason for the tendency of the 1-DoF solution to overestimate the 2-DoF solution is
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 251

R5 Tacoma Bridge
9 0.7
8 A∗2 A∗2
0.6
A∗3 A∗3
7
A∗3= f(H ∗1,A∗2) 0.5 A∗3= f(H ∗1,A∗2)
6
5 0.4
4 0.3
3 0.2
2
0.1
1
0 0
−1 −0.1
0 5 10 15 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
UR UR

Fig. 8. Degree of approximation of Eq. (35) in the case of the rectangular cylinder R5 and First Tacoma Narrows
Bridge. For the latter A3 is declared as ‘‘negligible’’ (Scanlan and Tomko, 1971).

Aerodynamics: R5 Aerodynamics: R5
Dynamics: μ = 35.7, rα = 0.249 Dynamics: μ = 178.6, rα = 0.353
5.5
5 6
4.5
5
4
3.5 4
3
URc

URc

2.5 3
2
2
1.5
1 1
0.5
0 0
1 1.5 2 2.5 3 3.5 4 4.5 5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
γω γω

Fig. 9. Comparison between the solution of the 2-DoF eigenvalue flutter problem and that of the proposed
approximate formulas. ‘‘1-DoF sol.’’ refers to Eqs. (38) and (39).

the role played by the intermodal coupling with heaving motion, which becomes important
when the frequency ratio is small. Another possible explanation could be that, when the
frequency ratio is close to unity, a coupled instability arises before torsional flutter.
Nevertheless this fact, experimentally observed in Nakamura and Yoshimura (1976), is not
able to explain the discrepancy between 1-DoF and 2-DoF solutions in Fig. 9(a), which,
although decreasing, is still significant for large frequency separation. Moreover, it is
important to note in Fig. 9 and Table 11 that the results of the simplified formulas closely
follow the 2-DoF eigenvalue solution and the degree of approximation is good and
generally better than the 1-DoF approximation.
Finally, in Table 11 the approximate approach is further validated considering the case
studies presented in Table 10, where the aerodynamic behavior of the rectangular cylinder
 ARTICLE IN PRESS
252 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

Table 10
Case studies (torsional flutter)

Case Aerodynamics m (–) ra (–) go (–) zh (%) za (%)

1 Tacoma 47.2 0.539 1.54 0.5 0.5

2 R5 47.2 0.539 1.54 0.5 0.5
3 R5 35.7 0.249 2.38 0.5 0.5
4 R5 178.6 0.353 1.96 0.8 0.8
5 R5 24.1 0.374 1.32 0.6 0.7
6 R10 47.2 0.539 1.54 0.5 0.5

Table 11
Results for the case studies listed in Table 10 (differences are calculated with respect to the 2-DoF eigenvalue
solution)

Case 1-DoF sol. 2-DoF eig. sol. approx. sol. I approx. sol. II

f c (Hz) U c (m/s) f c (Hz) U c (m/s) f c (Hz) Df c (%) U c (m/s) DU c (%) f c (Hz) Df c (%) U c (m/s) DU c (%)

1 0.200 11.5 0.200 11.5 0.200 0.0 11.5 0.0 0.196 2.0 9.3 19.7
2 0.189 11.5 0.190 10.0 0.190 +0.4 10.0 0.9 0.195 +2.5 11.0 +9.6
3 0.376 66.3 0.377 58.5 0.384 +1.9 58.0 0.7 0.430 +14.0 70.8 +21.1
4 0.626 50.9 0.626 49.4 0.626 +0.0 49.5 +0.1 0.635 +1.5 50.7 +2.6
5 0.066 19.0 0.067 11.2 0.070 +3.6 12.0 +7.1 0.074 +9.8 17.4 +55.0
6 0.182 21.6 0.188 17.3 0.189 +0.1 16.9 2.2 0.189 +0.4 17.2 0.5

R10, characterized by a chord-to-thickness ratio of 10.0 (Matsumoto, 1996), is considered

along with the First Tacoma Narrows Bridge (Scanlan and Tomko, 1971) and the
previously mentioned rectangular cylinder R5 (Matsumoto, 1996). As reference, the
solutions of both 2-DoF [Eqs. (7) and (8)] and 1-DoF approaches [Eqs. (38) and (39)] are
reported.
It seems possible to conclude that the proposed method, albeit less useful since the
mode-by-mode analysis is a practicable approximation, applies well also in the case of
cross sections prone to torsional flutter. The formulas adopting just two flutter derivatives
(approx. sol. II) are less accurate for these bluffer cross sections but only in one case (Case
5 in Tables 10 and 11) the error is very large (nevertheless, the approximation of the 1-DoF
solution is even worse).

9. Nakamura’s formulas

Nakamura (1978) proposed a series of approximate equations for bridge deck flutter
which can be easily transformed into the simplified formula proposed herein for the critical
reduced wind speed neglecting structural damping [Eq. (31)]. For the critical frequency a
simplified formula analogous to Eq. (29) is given.
Nakamura’s results are based on the following assumptions about the unsteady
aeroelastic behavior of bridge decks:

(1) jH 2 j5jH 3 j;
(2) H 1 ¼ KH 3 , A1 ¼ KA3 , H 4 ¼ KH 2 , A4 ¼ KA2 ;
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 253

(3) the stability of the pitching branch is considered, whereas the heaving branch is
assumed to be stable. The equivalent center of rotation should not be far from the
midchord;
(4) the phase angle between heaving and pitching motions is assumed to be small.

These conditions seem quite limiting and perhaps too theoretical, but the author himself
claims that they are not seriously restrictive, being the formulas expected to be applicable
to a variety of structures ranging from a flat plate to much bluffer bridge deck sections
(Nakamura, 1978). Nakamura’s experimental work, the investigation reported in Mannini
(2006) as well as the data analyzed in this paper, seem to confirm the validity of some of the
aforementioned assumptions. In particular, Nakamura showed that the small phase angle
between heaving and pitching motions, as well as the absolute peculiarity of bridge deck
flutter with respect to aircraft wing flutter, may be attributed to the absence of inertial
coupling.
After comparison with Nakamura’s research, some interesting points of the present
work can be emphasized:

 The same result as Nakamura’s is reached in a completely independent way and without
setting any theoretical assumption, by means of a much more ‘‘empirical’’ approach,
manipulating and simplifying the flutter equations on the basis of a large number of
data. This fact seems to be a good argument supporting the validity of the proposed
formulas.
 In this paper the formula proposed for the frequency of oscillation at flutter depends on
H 1 and A2 [Eq. (20)]. Conversely, in Nakamura (1978) Eq. (29) is obtained for the same
purpose, containing only the flutter derivative A3 .
 In the proposed formulas damping is not neglected [Eqs. (19) and (30)] and its influence
on flutter onset is investigated, showing sometimes significant effects [see Figs. 4(d)
and 9(b)].
 The possibility to take into account two flutter derivatives (H 1 and A2 ) instead of three
(H 1 , A2 and A3 ) is also considered and discussed.

10. Concluding remarks

In this paper an approximate method to calculate flutter critical wind speed and
frequency is presented. This approach is only applicable to cases in which the
bending–torsional coupling is marginally affected by the structural mode shapes or by
higher modes and cannot fully replace more accurate analyses. Nevertheless, it helps to
better understand the flutter mechanism and represents the first step toward a simple
engineering tool, which, for not very large conventional structures or at preliminary bridge
design stages, allows the expeditious prediction of the critical wind speed without
performing wind-tunnel tests.
The formulas are derived from the manipulation of the flutter equations and seem to
give results close to the rigorous solution of the eigenvalue problem of stability. The
significant feature of this method is the fact that only three flutter derivatives (H 1 , A2 and
A3 ) or even two (H 1 and A2 ) are required to perform the calculation. Therefore, the
considered framework allows to take into account the uncoupled aeroelastic coefficients
 ARTICLE IN PRESS
254 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

only. The formulas apply well also in the case of cross sections prone to torsional flutter,
this fact confirming the wide generality of the method, which provides an improvement
with respect to the mode-by-mode approximation.
This work, underlying the key role played by the flutter derivatives H 1 and A2 , confirms
the results found by Matsumoto for a wide range of rectangular cylinders. Nevertheless, it
also suggests the necessity to consider in many instances a third aeroelastic coefficient (A3 )
in order to obtain accurate estimations of flutter wind speed and frequency. Moreover, the
proposed formulas are not directly obtainable from the flutter equations either by setting
to zero five out of eight flutter derivatives or only applying the inter-dependence
relationships suggested by other authors (e.g. Matsumoto, 1996).
Finally, the role played in the flutter instability mechanism by several structural
parameters is investigated. In particular, it is shown that, despite what is usually believed,
the contribution of structural damping to flutter onset is not always negligible, depending
both on the dynamic and aerodynamic properties of the bridge. Consequently, in some
particular cases the flutter critical wind speed can be significantly over- or under-predicted
if structural damping is not correctly accounted for. For this reason, the fact that the
proposed simplified formulas do not neglect this parameter should be emphasized.

Acknowledgment

This work has partially been supported by MIUR (Italian Ministry of University and
Research) as a part of a National Research Program and within the International Doctoral
Course on ‘‘Risk Management on the Built Environment’’ (University of Florence,
Italy–TU Braunschweig, Germany).

References

AFGC, 2002. Comportement au Vent des Ponts, first ed. Presses de l’Ecole Nationale des Ponts et Chaussées,
Paris (Association Franc- aise de Génie Civil).
Augusti, G., Chiarugi, A., Vignoli, A., 1980. Experimental analysis of a cable-stayed bridge. Inserto Costruzioni
Metalliche 3, 2–16.
Bartoli, G., Mannini, C., 2005a. From multimodal to bimodal approach to flutter. In: Soize, C., Schuëller, G.I.
(Eds.), Proceedings of the 6th European Conference on Structural Dynamics. Millpress, Rotterdam, Paris,
pp. 349–354.
Bartoli, G., Mannini, C., 2005b. Reliability of bridge deck flutter derivative measurements in wind tunnel tests. In:
Augusti, G., Schuëller, G.I., Ciampoli, M. (Eds.), Proceedings of the 9th International Conference on
Structural Safety and Reliability. Millpress, Rotterdam, Rome, pp. 1193–1200.
Bartoli, G., Procino, L., 2004. Prove in galleria del vento su modello sezione di una passerella pedonale in loc.
Ruffolo (Siena). Internal Report, CRIACIV.
Bartoli, G., Righi, M., 2006. Flutter mechanism for rectangular prisms in smooth and turbulent flow. J. Wind
Eng. Ind. Aerodyn. 94, 275–291.
Bartoli, G., Borri, C., Gusella, V., 1995. Aeroelastic behaviour of bridge decks: a sensitivity analysis of the
turbulent nonlinear response. In: Proceedings of the 9th International Conference on Wind Engineering. Wiley
Eastern Ltd, New Delhi, pp. 851–862.
Bisplinghoff, R.L., Ashley, H., Halfman, R.L., 1996. Aeroelasticity, first ed. Dover Publications Inc., New York.
Borri, C., Costa, C., Salvatori, L., 2005. Reliability of indicial functions in bridge deck aeroelasticity. In: Augusti,
G., Schuëller, G.I., Ciampoli, M. (Eds.), Proceedings of the 9th International Conference on Structural Safety
and Reliability. Millpress, Rotterdam, Rome, pp. 1217–1224.
Bucher, C.G., Lin, Y.K., 1988a. Effect of spanwise correlation of turbulence field on the motion stability of long-
span bridges. J. Fluids Struct. 2, 437–451.
 ARTICLE IN PRESS
G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256 255

Bucher, C.G., Lin, Y.K., 1988b. Stochastic stability of bridges considering coupled modes. J. Eng. Mech. ASCE
114 (12), 2055–2071.
Bucher, C.G., Lin, Y.K., 1989. Stochastic stability of bridges considering coupled modes II. J. Eng. Mech. ASCE
115 (2), 384–400.
Caracoglia, L., Jones, N.P., 2003. Time domain vs. frequency domain characterization of aeroelastic forces for
bridge deck sections. J. Wind Eng. Ind. Aerodyn. 91, 371–402.
Chen, X., 2007. Improved understanding of bimodal coupled bridge flutter based on closed-form solutions.
J. Struct. Eng. ASCE 133 (1), 22–31.
Chen, X., Kareem, A., 2000. Nonlinear response analysis of long-span bridges under turbulent winds. In:
Proceedings of the 4th International Colloquium on Bluff Body Aerodynamics and Applications. Bochum.
Chen, X., Kareem, A., 2006a. Revisiting multimode coupled bridge flutter: some new insights. J. Eng. Mech.
ASCE 132 (10), 1115–1123.
Chen, X., Kareem, A., 2006b. Understanding the underlying physics of multimode coupled bridge flutter based on
closed-form solutions. In: Proceedings of the 4th International Symposium on Computational Wind
Engineering. Yokohama, pp. 145–148.
Chowdhury, A.G., Sarkar, P.P., 2004. Identification of eighteen flutter derivatives of an airfoil and a bridge deck.
Wind Struct. 7 (3), 187–202.
D’Asdia, P., Sepe, V., 1998. Aeroelastic instability of long-span suspended bridges: a multi-mode approach.
J. Wind Eng. Ind. Aerodyn. 74–76, 849–857.
Diana, G., Bruni, S., Cigada, A., Collina, A., 1993. Turbulence effect on flutter velocity in long span suspended
bridges. J. Wind Eng. Ind. Aerodyn. 48, 329–342.
Diana, G., Cheli, F., Zasso, A., Bocciolone, M., 1999. Suspension bridge response to turbulent wind: comparison
of a new numerical simulation method results with full scale data. In: Larsen, A., Larose, G.L., Livesey, F.M.
(Eds.), Wind Engineering into the 21st Century. Proceedings of the 10th International Conference on Wind
Engineering. Balkema, Rotterdam, Copenhagen, pp. 871–878.
Diana, G., Resta, F., Zasso, A., Belloli, M., Rocchi, D., 2004. Forced motion and free motion aeroelastic tests on
a new concept dynamometric section model of the Messina suspension bridge. J. Wind Eng. Ind. Aerodyn. 92
(6), 441–462.
Diana, G., Bruni, S., Rocchi, D., 2005. A numerical and experimental investigation on aerodynamic nonlinearities
in bridge response to turbulent wind. In: Náprstek, J., Fischer, C. (Eds.), Proceedings of the 4th European and
African Conference on Wind Engineering. Prague.
Dyrbye, C., Hansen, S.O., 1997. Wind Loads on Structures, first ed. Wiley, New York.
ECCS, 1987. Recommandations pour le calcul des effets du vent sur les constructions, second ed. European
Convention for Constructional Steelwork (ECCS), Comité Technique 12 ‘‘Vent’’.
ENV1991-2-4, 1995. Eurocode 1: Basis of Design and Actions on Structures. Part 2–4: Wind Actions.
Farquharson, F.B., 1949. Aerodynamic Stability of Suspension Bridges with Special Reference to the Tacoma
Narrows Bridge. Bulletin of the University of Washington, Seattle, Washington.
Flamand, O., 2003. Validation de la stabilité aérodynamique d’une passerelle à Turin. EN-CAPE 03.210 C-V0,
CSTB, Nantes.
Frandsen, A.G., 1966. Wind stability of suspension bridges. Application of the theory of ‘‘thin airfoils’’. In:
Proceedings of the International Symposium on Suspension Bridges. Lisbon, pp. 609–627.
Fung, Y.C., 1993. An Introduction to the Theory of Aeroelasticity. Dover Publications Inc., New York.
Gimsing, N.J., 1997. Cable Supported Bridges—Concepts and Design, second ed. Wiley, New York.
Jain, A., Jones, N.P., Scanlan, R.H., 1996. Coupled flutter and buffeting analysis of long-span bridges. J. Struct.
Eng. ASCE 122 (7), 716–725.
Katsuchi, H., Jones, N.P., Scanlan, R.H., 1999. Multimode coupled flutter and buffeting analysis of the Akashi
Kaikyo Bridge. J. Struct. Eng. ASCE 125 (1), 60–70.
Larsen, A., 1998. Advances in aeroelastic analyses of suspension and cable-stayed bridges. J. Wind Eng. Ind.
Aerodyn. 74–76, 73–90.
Mannini, C., 2006. Flutter vulnerability assessment of flexible bridges. Ph.D. Thesis, University of Florence,
Italy–TU Braunschweig, Germany.
Martini, S., 2004. Entwurf der Gründung einer Hängebrücke unter Berücksichtigung von Erdbebenbean-
spruchungen gemäb EC8-Teil2. Master’s Thesis, Bergische Universität, Wuppertal.
Matsumoto, M., 1996. Aerodynamic damping of prisms. J. Wind Eng. Ind. Aerodyn. 59, 159–175.
Matsumoto, M., 2005. Flutter instability of structures. In: Náprstek, J., Fischer, C. (Eds.), Proceedings of the 4th
European and African Conference on Wind Engineering. Prague.
 ARTICLE IN PRESS
256 G. Bartoli, C. Mannini / J. Wind Eng. Ind. Aerodyn. 96 (2008) 229–256

Matsumoto, M., Kobayashi, K., Niihara, Y., Shiarto, H., Hamasaki, H., 1995. Flutter mechanism and its
stabilisation of bluff bodies. In: Proceedings of the 9th International Conference on Wind Engineering. Wiley
Eastern Ltd, New Delhi, pp. 827–838.
Matsumoto, M., Kobayashi, Y., Shirato, H., 1996. The influence of aerodynamic derivatives on flutter. J. Wind
Eng. Ind. Aerodyn. 60, 227–239.
Matsumoto, M., Daito, Y., Yoshizumi, F., Ichikawa, Y., Yabutani, T., 1997. Torsional flutter of bluff bodies.
J. Wind Eng. Ind. Aerodyn. 69–71, 871–882.
Minh, N.N., Miyata, T., Yamada, H., Sanada, Y., 1999. Numerical simulation of wind turbulence and buffeting
analysis of long-span bridges. J. Wind Eng. Ind. Aerodyn. 83 (1–3), 301–315.
Miyata, T., Yamaguchi, K., 1993. Aerodynamics of wind effects on the Akashi Kaikyo Bridge. J. Wind Eng. Ind.
Aerodyn. 48, 287–315.
Miyata, T., Tada, K., Sato, H., Katsuchi, H., Hikami, Y., 1994. New findings of coupled flutter in full model wind
tunnel tests on the Akashi Kaikyo Bridge. In: Proceedings of the International Symposium on Cable-Stayed
and Suspension Bridges. Deauville, pp. 163–170.
Miyata, T., Yamada, H., Boonyapinyo, V., Santos, J.C., 1995. Analytical investigation on the response of a very
long suspension bridge under gusty wind. In: Proceedings of the 9th International Conference on Wind
Engineering, New Delhi, pp. 1006–1017.
Nakamura, Y., 1978. An analysis of binary flutter of bridge deck sections. J. Sound Vib. 57 (4), 471–482.
Nakamura, Y., 1979. On the aerodynamic mechanism of torsional flutter of bluff structures. J. Sound Vib. 67 (2),
163–177.
Nakamura, Y., Mizota, T., 1975. Torsional flutter of rectangular prisms. J. Eng. Mech. Div. 101 (EM2), 125–142.
Nakamura, Y., Yoshimura, T., 1976. Binary flutter of suspension bridge deck sections. J. Eng. Mech. Div. 102
(EM4), 685–700.
ONERA, 1989. Etude du comportement dans le vent du tablier définitif du pont de Normandie. Rapport
Technique 15/3588 RY 091R-391G.
Righi, M., 2003. Aeroelastic stability of long span suspended bridges: flutter mechanism on rectangular cylinders
in smooth and turbulent flow. Ph.D. Thesis, University of Florence.
Rocha, M.M., Loredo-Souza, A.M., Paluch, M.J., Núñez, G.J.Z., 2003. Aspectos aerodinâmicos a serem
considerados no projeto de pontes estaiadas—Estudo de caso: ponte sobre o Rio Guamá. In: Proceedings of
the 5th Simpósio EPUSP sobre Estruturas de Concreto, São Paulo.
Sarkar, P.P., 1992. New identification methods applied to the response of flexible bridges to wind. Ph.D. Thesis,
John Hopkins University, Baltimore.
Scanlan, R.H., Tomko, J.J., 1971. Airfoil and bridge deck flutter derivatives. J. Eng. Mech. Div. Proc. ASCE 97,
1717–1737.
Scanlan, R.H., Jones, N.P., Singh, L., 1997. Inter-relations among flutter derivatives. J. Wind Eng. Ind. Aerodyn.
69–71, 829–837.
Selberg, A., 1961. Oscillation and aerodynamic stability of suspension bridges. In: Acta Polytechnica
Scandinavica, Civil Engineering and Building Construction Series 13, Oslo.
Simiu, E., Scanlan, R.H., 1996. Wind Effects on Structures, third ed. Wiley, New York.
Singh, L., Jones, N.P., Scanlan, R.H., Lorendeaux, O., 1995. Simultaneous identification of 3-dof aeroelastic
parameters. In: Proceedings of the 9th International Conference on Wind Engineering. Wiley Eastern Ltd,
New Delhi, pp. 972–981.
Theodorsen, T., 1934. General theory of aerodynamic instability and the mechanism of flutter. NACA Technical
Report 496, Aeronautics, 20th Annual Report of the National Advisory Committee for Aeronautics, NACA.
Zasso, A., 1996. Flutter derivatives: advantages of a new representation convention. J. Wind Eng. Ind. Aerodyn.
60, 35–47.