Engineering Structures 172 (2018) 1073–1080

Contents lists available at ScienceDirect

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

Review article

Eccentric-wing flutter stabilizer for bridges – Analysis, tests, design, and T

costs
⁎
U. Starossek , T. Ferenczi, J. Priebe
Hamburg University of Technology, Denickestr. 17, 21073 Hamburg, Germany

A R T I C LE I N FO A B S T R A C T

Keywords: A device is presented that aims at preventing bridge flutter. It consists of wings positioned along the sides of, and
Passive aerodynamic damper fixed to, the bridge deck. Flutter suppression effectiveness is high provided the lateral eccentricity of the wings is
Fixed wing large. It is a passive aerodynamic device that is presumably more cost-efficient than other passive measures or
Flutter analysis devices. Moreover, it does not contain moving parts. This is an advantage over devices with moving parts, which
Wind tunnel test
meet resistance due to reliability and durability concerns. Wind-tunnel tests were performed in which the flutter
Design study
speed of a bridge deck sectional model without wings and with wings mounted in various configurations was
Cost estimate
measured. The experimental results are presented and compared with the results of flutter analyses using finite
aeroelastic beam elements. Using the analytical approach, also the effect of the distribution of the wings along
the length of a bridge was studied to optimize this distribution. Preliminary design studies for the wings and
their support structures as well as quantity and cost estimates are presented. For a representative example bridge
and wing configuration, an increase of 22% of flutter speed is reached at a cost increase of 2.5%.

1. Introduction
Flutter is a phenomenon that governs the design of long-span
bridges. Various measures have been proposed and applied to raise the
flutter resistance of bridges, that is, their critical wind speed for flutter
onset (flutter speed).
The concept of the twin suspension bridge was described by
Richardson [1] and since implemented in a few bridges. It is a passive
aerodynamic measure that takes advantage of the gap between the two
(or more) bridge decks. The flutter speed increase thus achieved comes
at the additional cost of the cross beams that are needed to connect the
individual decks.
An active aerodynamic device for raising the flutter speed was
proposed by Ostenfeld and Larsen [2]. It consists of wings, installed
along the sides of the bridge deck, whose pitch is controlled by actua-
tors. A closed-loop control is envisaged in which, based on accel-
erometer measurements, an algorithm produces the control signals for
the actuators such that the movement of the wings generate stabilizing
wind forces. With such device, the safety of the bridge depends on
energy supply and the proper functioning of control software and
hardware – a condition that meets resistance with bridge owners and
authorities due to reliability and durability concerns. A passive aero-
dynamic-mechanical device described by Starossek and Aslan [3] also
includes variable-pitch wings along the sides of the bridge deck. Instead
of being controlled by actuators, the pitch of the wings follows the
movements of tuned mass dampers inside the bridge deck to which the
wings are coupled by means of linkages or gears. With proper tuning,
the flutter suppression effectiveness can be similar to that of actively
controlled wings. Being a passive device, the safety of the bridge would
not depend on energy supply and a control system. It still includes
moving parts though, which raises the threshold of acceptance.
Diana et al. [4] examined the effect of various aerodynamic devices
rigidly attached to the deck of the envisaged Messina Strait Bridge,
including winglets positioned along the edges of the deck. The devices
are positioned close to the deck without a distinct vertical or horizontal
offset. Hence they form part of the aerodynamic contour of the deck and
influence the flow field around it. Only qualitative indications are given
in [4] concerning the impact of such devices on the flutter behavior of
the bridge and it does not become clear whether and by how much the
flutter speed is raised by the examined winglets.
Raggett [5] suggested a pair of wings rigidly mounted above, or
slightly outboard of, the two edges of the bridge deck to stabilize the
bridge against flutter. The wings are arranged with a distinct vertical
offset from the deck so that they are aerodynamically independent of
the deck. Liu et al. [6] considered a similar configuration and studied its
influence on bridge flutter both analytically and by sectional model
wind tunnel tests. When the wings are considered aerodynamically
independent of the deck, their impact on the flutter speed of the bridge

⁎
Corresponding author.
E-mail address: starossek@tuhh.de (U. Starossek).

https://doi.org/10.1016/j.engstruct.2018.06.056

Available online 14 July 2018

0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
U. Starossek et al.
ac
wing
b b
wind
support
structure
bridge deck
Fig. 1. Bridge with eccentric-wing flutter stabilizer – cross section.
Fig. 2. Bridge with eccentric-wing flutter stabilizer around center of main span
– plan view (not to scale).
can easily be assessed analytically as will be outlined below. The au-
thors’ own parametric studies show that this impact is small for the
configurations described in [5,6], that is, for wings arranged above, or
slightly outboard of, the edges of the deck. A significant rise of flutter
speed only is produced by wings arranged with large lateral eccen-
tricity.
2. Eccentric wing flutter stabilizer
In view of the development described above, it seems promising, for
raising the flutter speed of a bridge, to pursue passive aerodynamic
measures that do not include moving parts but, at the same time, are
sufficiently effective to eliminate the need for substantial additional
structural elements or structural stiffening. The eccentric-wing flutter
stabilizer presented in the following meets these requirements.
The device consists of wings positioned along the sides of the bridge
Fig. 3. Reasons of effectiveness – angular velocity.
1074
Engineering Structures 172 (2018) 1073–1080
ac
bc bc
deck. An arrangement with wings equally positioned on both sides of
the deck is shown in Fig. 1. In certain cases, further discussed below, it
is advantageous to provide wings on only one side of the deck or to
provide wings on both sides, but to design them differently, that is, with
different widths and lateral eccentricities (dimensions 2bc and ac).
The wings are mounted on transversely orientated support struc-
tures that are laterally attached to the bridge deck. Hence they do not
move relative to the deck. It can be shown that the flutter suppression
effectiveness of the device is high provided the lateral eccentricity of
the wings is large. Wings and support structures are envisaged as light-
weight components. The wings are aerodynamically shaped such that
the lift under inclined flow is large and the resistance is small. Their
cross section can be symmetric to a horizontal plane or double sym-
metric approaching an elliptical shape. The lateral eccentricity of the
wings is on the order of the bridge deck width (ac/b ≈ 1.5 to 2.5). The
width of a wing in direction transverse to the bridge axis is on the order
of one tenth of the bridge deck width (bc/b ≈ 0.05 to 0.20). The wings
are preferably positioned above or below the bridge deck with sufficient
vertical offset to avoid aerodynamic interference between the wings
and the bridge deck including traffic.
For optimum cost efficiency, the wings are not placed over the en-
tire length of the bridge but only at regions where large vibration
amplitudes occur (Lc < L). In case flutter is governed by the first
symmetric modes of vibration, these regions lie around the center of the
main span (Fig. 2). In case flutter is governed by the first antisymmetric
modes of vibration, these regions lie around the quarter points of the
main span.
In an alternative configuration, a single wing is replaced by a certain
number of wings stacked above each other [7,8]. The flutter-suppres-
sion effectiveness of such a group of wings is approximately the same as
for a single wing provided the sum of the widths of the wings is the
same as the width of the single wing and the vertical distance between
the individual wings is not too small. A larger effective wing width can
thus be achieved at a possibly lower cost.
U. Starossek et al.
Fig. 4. Reasons of effectiveness – angular displacement.
3. Reasons of effectiveness
Because the width of a wing is comparatively small, the lift force on
a wing as a function of the vertical and angular motion of the bridge
deck can be thought of as quasi-stationary. An angular velocity of the
deck, α̇ , will result in a vertical velocity of a wing of ac α̇ (Fig. 3). This
corresponds to an apparent vertical flow of the same velocity acting on
the wing, which, when superimposed on the horizontal wind velocity, ,
leads to an apparent resulting quasi-stationary flow at an angle of attack
of ac α̇ / u and a lift force on the wing that is proportional to ac α̇ . This
force times the eccentricity, ac , produces a moment on the deck that is
proportional to ac2 α̇ and in counter-phase with α̇ .
Hence a wing produces a wind-induced moment that dampens the
angular motion of the bridge deck (aerodynamic damping) and is
proportional to the square of the eccentricity of the wing. This effect is
independent of whether the wing is positioned on the windward side or
the lee side of the deck and hence the wings on both sides produce
positive damping. By raising the damping of the angular motion of the
deck, the flutter speed is also raised.
Another effect that can be explained by quasi-stationary theory is
the aerodynamic stiffness produced by the wings. An angular dis-
placement of the deck, α , is accompanied by the same angular dis-
placement, or pitch, of the wings (Fig. 4). The horizontal wind flow
acting on a pitched wing produces a lift force on the wing that is pro-
portional to α . This force times the eccentricity results in a moment on
the deck that is proportional to ac α and in phase with + α or −α , de-
pending on whether the wing is positioned on the windward side or the
lee side of the deck.
Hence a wing produces aerodynamic stiffness related to the angular
displacement of the deck that is proportional to the eccentricity of the
wing. This stiffness is negative for a windward wing and positive for a
leeward wing. Positive aerodynamic stiffness raises the stiffness, the
natural frequency of angular vibrations, and the flutter speed – negative
aerodynamic stiffness vice versa. In a symmetrical wing layout with
wings on both sides of the bridge deck, the aerodynamic stiffness
contributions of the wings of both sides cancel and do not affect the
flutter speed.
1075
Engineering Structures 172 (2018) 1073–1080
4. Flutter analysis
The effect of the eccentric-wing flutter stabilizer on the flutter speed
has been studied analytically. It is assumed in this study that there is no
aerodynamic interference between the bridge deck and the wings nor
between the two wings. Theodorsen’s flutter derivatives [9] are used for
modelling the wind forces on the bridge deck. This is deemed suffi-
ciently accurate for determining the relative effectiveness of the various
wing configurations considered here provided the aerodynamic contour
of the deck is a streamlined closed box.
The overall system and the wind forces on the wings are modelled in
two different ways. In a first approach, the bridge system is generalized
to two degrees of freedom, that is, heave and pitch, and it is assumed
that the wind forces on the wings can be calculated by quasi-stationary
theory. The angular damping and stiffness of the two-degrees-of-
freedom system is modified by adding the aerodynamic damping and
stiffness contributions of the wings as described in the previous section.
Theodorsen’s solution procedure [9] for determining the flutter speed is
applicable, but must be supplemented by a further level of iteration,
given that the damping and stiffness contributions of the wings depend
on the wind speed.
In a second, more sophisticated approach, the bridge deck and the
wings are modelled with finite aeroelastic beam elements [10] and the
flutter speed is computed with a self-developed computer program
called DALI. This approach allows the study of discontinuous wing
configurations where the wings are not placed over the entire length of
the bridge. Moreover, the quasi-steady flow assumption for including
the wind forces on the wings is no longer required. The wind forces on
both the bridge deck and the wings are modelled by using Theodorsen’s
flutter derivatives. The wings are assumed to be connected to the deck
by rigid support structures. The degrees of freedom of the wings are
thus cinematically related to the degrees of freedom of the deck and no
additional independent degrees of freedom are introduced by including
the wings. The forces acting on bridge deck and wings can therefore be
captured in a common finite aeroelastic beam element.
5. Parametric study
Parametric computations have been performed for a long-span
U. Starossek et al. Engineering Structures 172 (2018) 1073–1080

As noted before, the wings should preferably be placed only at re-

gions where large vibration amplitudes occur. Fig. 6 shows the com-
puted flutter speed increase as function of the length of the wings re-
lated to the bridge length assuming otherwise same parameters as
above and placement of the wings on both sides of the bridge at op-
timum locations. When each wing has a length of 50% of the bridge
length, the flutter speed is still raised by 22%, that is, to a value of
56.5 m/s, instead of by 28%.

Fig. 5. Computed flutter speed increase as function of width and eccentricity of

wings of eccentric-wing flutter stabilizer.

suspension bridge with a main span length of Lm = 3000 m, a single box

deck with a width of 2b = 38.0 m, and a torsional-to-vertical natural
frequency ratio of 1.76. The flutter speeds reported here were computed
by the second, more sophisticated approach described in the preceding
section. The structure was modelled with 40 finite aeroelastic beam
elements, which led to 159 degrees of freedom. The flutter speed of the
bridge without wings was computed as 46.3 m/s. Fig. 5 shows the
computed flutter speed increase for various wing configurations relative
Fig. 6. Computed flutter speed increase as function of length of eccentric-wing
to the flutter speed without wings.
flutter stabilizer.
The wings are assumed to be placed on the lee side only and along
the entire length of the bridge deck. The relative wing width, bc/b, is
varied between 0.025 and 0.170. The relative eccentricity, ac/b, is 6. Experimental determination of flutter speed
varied between 1.00 and 2.00. For a relative wing width of 0.100 and a
relative wing eccentricity of 2.00, the flutter speed is raised by 64% to a 6.1. Test setup
value of 75.8 m/s. It can also be seen in Fig. 5 that the flutter-sup-
pression effectiveness of the device increases over-proportionally with The effect on flutter stability of rigid wings mounted to a bridge
the lateral eccentricity and the width of the wings. deck was studied in sectional model wind tunnel tests [11,12]. Fig. 7
Placing wings on only one side of the bridge deck – or placing wings shows the test setup.
on both sides but designing them with different widths and lateral ec- The cross section of the bridge deck sectional model is similar to
centricities – is advantageous when the expected maximum wind speeds that of the Great Belt Bridge girder with a scale of 1:74 (Fig. 8). It has a
strongly differ for the two transverse directions. The wings on only one length of 800 mm, a width of 420 mm, and is made of acrylonitrile
side – or, when placing wings on both sides, the wings with the larger butadiene styrene (ABS). The wings are made of rigid aluminum honey-
widths and lateral eccentricities – are then placed on the lee side of the comb profiles with the intention to provide them with large stiffness
stronger wind. and high aeroelastic stability. They have a rectangular cross section
When the expected maximum wind speeds are about the same for with rounded edges and a smooth surface. The wings are connected to
both transverse directions, the wing configuration should be the same the deck model by lightweight aluminum rods, placing them with the
on both sides of the bridge deck. The flutter-suppression effectiveness of same vertical offset relative to the deck model [11]. In a modified setup
such an arrangement is reduced, as it follows from the above discussion (not shown), the vertical offset of windward and lee side wings differ to
of the reasons of effectiveness. The aerodynamic stiffness contributions
of the wings of both sides cancel and only their aerodynamic damping is
available for raising the flutter speed.
To overcome this problem, the fixed and long wings envisaged here
could be replaced by many short, appropriately formed wings sup-
ported on rotating bearings to the effect that the respective windward
wings, driven by the wind, align vertically to make them aero-
dynamically ineffective and a maximum flutter speed increase is
achieved [8]. However, the design intent of avoiding moving parts
would be abandoned.
With fixed wings on both sides of the bridge deck and under
otherwise same conditions, the computed flutter speed is still 28%
higher than the value without wings. Placing the same wings right
above the two edges of the bridge deck (bc/b = 0.100, ac/b = 1.00), as
suggested in [5,6], produces a rise of flutter speed of only 4%.
Fig. 7. Test setup.

1076
U. Starossek et al. Engineering Structures 172 (2018) 1073–1080

Table 2
Measured and computed flutter speeds.
Setup Eccentr. (ac) Wing arrangement Wind tunnel [m/s] Analysis [m/s]

M01 2.0 b – 5.95 6.97

M02 2.0 b Windward side 5.3 6.61
M03 2.0 b Both sides 8.0 10.1
M04 2.0 b Lee side 11.1 12.6
M05 2.0 b – 5.85 5.72
M06 2.0 b Windward side 5.15 5.61
M07 2.0 b Both sides 9.25 28.6
Fig. 8. Sectional model of deck with windward and lee wings.
M08 2.0 b Lee side > 11.7* ∞

M09 2.5 b – 5.5 5.62

avoid aerodynamic interference between both wings [12,13]. M10 2.5 b Windward side > 8.5* 5.93
The tests are performed for 24 different setups (Table 1), that is, for M11 2.5 b Both sides > 15.5* 59.1
three different wing eccentricities, two different wing widths, and four M12 2.5 b Lee side > 13.5* ∞
different wing arrangements (no wings/windward wing only/windward M13 2.5 b – 7.6 3.10
M14 2.5 b Windward side > 13.0* 23.6
and lee wings/lee wing only). The wing eccentricities are 1.5, 2.0, or M15 2.5 b Both sides > 14.0* ∞
2.5 times the half-width of the deck model, that is, 315 mm, 420 mm, or M16 2.5 b Lee side > 13.1* ∞
525 mm. The wing widths are 0.10 or 0.15 times the deck model width, M17 1.5 b – 8.0 7.90
that is, 42 mm or 63 mm. The focus is on the aerodynamic effect of the M18 1.5 b Windward side 6.7 7.24
wings. Therefore, a wing that is absent in a particular arrangement is M19 1.5 b Both sides 7.65 9.10
substituted by ballast at the same eccentric position. To produce flutter M20 1.5 b Lee side 8.75 10.82
M21 1.5 b – 7.7 7.35
at low wind speeds and thus improve the observability, the test setup is
M22 1.5 b Windward side 6.15 6.57
tuned to a low angular-to-vertical natural frequency ratio of 1.11. M23 1.5 b Both sides 7.15 9.28
M24 1.5 b Lee side 8.95 12.7

6.2. Test procedure and results

* No upper limit could be found experimentally.

The flutter speed is determined in the following way. In each trial, Most test setups with wings show an increase of flutter speed
the wind speed is set to a fixed value, the model is released from a compared to the respective test setup without wings. Furthermore, the
certain initial deflection, and the vibration behavior (decaying or flutter speed mostly increases with greater eccentricity and width of the
growing) is established. This is repeated seven times. For the next and wings. As for the wing arrangement, a lee side wing mostly leads to a
further rounds of trials, the wind speed is increased by a small incre- higher flutter speed than a both-sides wing arrangement, which in turn
ment. A lower limit (highest wind speed with decaying vibration) and leads to a higher flutter speed than a windward wing. These findings are
an upper limit (lowest wind speed with growing vibration) of the flutter compatible with the reasons of effectiveness outlined above.
speed are thus identified. The flutter speed lies in between both limit When narrow wings are mounted on both sides of the deck with an
values and is taken as the average of both. The flutter speeds de- eccentricity of 2.0 b (M03), a flutter speed increase of 34% (relative to
termined in this manner are shown in Table 2. They were obtained for M01) is observed. When further increasing the eccentricity to 2.5 b
the modified setup (see above) with windward and lee side wings (M11), the flutter speed is more than twice the value of the respective
placed at different vertical offsets relative to the deck model [12]. arrangement without wings (M09).
In the latter and some other cases, only a lower limit of the flutter
Table 1 speed can be determined. This is due to local flutter of the windward wing-
Overview of test setups. rod sub-assemblage or to excessive angular displacements produced by the
Setup Eccentr. (ac)* Windward side Lee side static wind force acting on the asymmetric sectional bridge deck model.
In some cases, placing wings on the windward side only or on both
M01 2.0 b Ballast (180 g) Ballast (180 g) sides leads to a reduction of flutter speed. For the both-sides wing ar-
M02 2.0 b Wing (bc = 0.10 b) Ballast (180 g)
rangement, this only occurs for the minimum eccentricity of 1.5 b. Even
M03 2.0 b Wing (bc = 0.10 b) Wing (bc = 0.10 b)
M04 2.0 b Ballast (180 g) Wing (bc = 0.10 b) if mounting a wing on the lee side only mostly produces the highest
M05 2.0 b Ballast (288 g) Ballast (288 g) flutter speed increase, it is anticipated that for applications in the ma-
M06 2.0 b Wing (bc = 0.15 b) Ballast (288 g) jority of real projects a symmetric wing arrangement with wings on
M07 2.0 b Wing (bc = 0.15 b) Wing (bc = 0.15 b) both sides of the deck is most appropriate.
M08 2.0 b Ballast (288 g) Wing (bc = 0.15 b)
The overall result is that wind tunnel tests confirm the former
M09 2.5 b Ballast (180 g) Ballast (180 g) finding based on flutter analysis: fixed wings attached to a bridge deck
M10 2.5 b Wing (bc = 0.10 b) Ballast (180 g)
can produce a substantial increase in flutter speed provided they are
M11 2.5 b Wing (bc = 0.10 b) Wing (bc = 0.10 b)
M12 2.5b Ballast (180 g) Wing (bc = 0.10 b) positioned with a large lateral eccentricity.
M13 2.5 b Ballast (288 g) Ballast (288 g)
M14 2.5 b Wing (bc = 0.15 b) Ballast (288 g)
6.3. Comparison with computed flutter speed
M15 2.5 b Wing (bc = 0.15 b) Wing (bc = 0.15 b)
M16 2.5 b Ballast (288 g) Wing (bc = 0.15 b)
The results of the wind tunnel tests are compared to the results of
M17 1.5 b Ballast (180 g) Ballast (180 g)
M18 1.5 b Wing (bc = 0.10 b) Ballast (180 g)
flutter analyses in which the test setup is modelled by finite aeroelastic
M19 1.5 b Wing (bc = 0.10 b) Wing (bc = 0.10 b) beam elements as described in Section 4. Theodorsen’s flutter derivatives
M20 1.5 b Ballast (180 g) Wing (bc = 0.10 b) [9] are used for modelling the wind forces on deck and wings. The system
M21 1.5 b Ballast (288 g) Ballast (288 g) parameters are determined from the measured natural frequencies of
M22 1.5 b Wing (bc = 0.15 b) Ballast (288 g)
vertical and angular vibrations of the test setup without wind and from
M23 1.5 b Wing (bc = 0.15 b) Wing (bc = 0.15 b)
M24 1.5 b Ballast (288 g) Wing (bc = 0.15 b) the geometry and mass of the sectional models of deck and wings. The
structural damping of the test setup is small and was neglected.
* Lateral eccentricity of a wing, where b is the half-width of the deck. The results of the flutter analysis are included in Table 2. A value of

1077
U. Starossek et al. Engineering Structures 172 (2018) 1073–1080

Table 3
Flutter speeds of setups M01 to M04 and flutter speed increases.
Setup Windward side Lee side Wind incr. rel. Analysis incr. rel.
tunnel to M01 [m/s] to M01
[m/s]

M01 Ballast Ballast 5.95 – 6.97 –

M02 Wing Ballast 5.3 −11% 6.61 −5%
M03 Wing Wing 8.0 +34% 10.1 +45%
M04 Ballast Wing 11.1 +87% 12.6 +81%

infinite appears for the computed flutter speed of some test setups, in Fig. 9. Wing overall geometry.
particular those with large wing eccentricities and/or wing on the lee
side only. In such cases, there exists no wind speed at which one of the
solutions of the eigenvalue problem to be solved here corresponds to a
vibration with constant or increasing amplitude. Analysis thus predicts
that flutter does not occur at any wind speed. In all these cases, no
upper limit of the flutter speed could be found experimentally either.
When comparing the results from wind tunnel tests and flutter
analysis, good agreement is mostly found for setups without wings or
with windward wing only. These setups also exhibit the smallest flutter
speeds. The difference between measured and computed flutter speeds
becomes greater for setups with wings on both sides or with leeward
wing only. Nevertheless, a tendency to larger flutter speeds, compared
to the no-wing and windward-wing setups, is generally indicated by
both approaches. In most cases, the flutter speed determined by analysis
is larger than the value determined from the wind tunnel tests. The
relatively large differences between measured and computed flutter Fig. 10. Wing ribs and spars.
speeds for certain setups is attributed to the higher flutter speeds of
these setups and hence the higher influence of the wind forces on the
dynamic force equilibrium. Approximating the motion-induced wind
forces by means of Theodorsen’s flutter derivatives, instead of by using
the flutter derivatives of the actual cross section, will then produce less
accurate analytical results for the flutter speed of the test setup. This
interpretation and the finding that the computed values are mostly
higher than the measured ones is consistent with the empirical finding
[14] that bridge deck cross sections generally lead to smaller flutter
speeds than the thin flat plate assumed in Theodorsen’s theory.
Setup M03 represents a feasible and recommended wing configura-
tion. The measured and computed flutter speeds for setups M01 to M04,
which only differ in wing arrangement, and the respective increases re-
lative to M01 (no wings) are shown in Table 3. The relative increases of
measured and computed flutter speeds agree comparatively well.
Fig. 11. Wing ribs.
7. Wing design
Table 4
7.1. Wing structure Material and cross-sectional properties of a wing.
Property Symbol Value Unit
An elliptical cross section is chosen for the wings with a height-to-
width ratio of 0.16 and a width of 2bc = 3.80 m for a relative wing Material of skin Al EN-AW 5454 H14
width of bc/b = 0.100 at an assumed deck width of 2b = 38.0 m. The Thickness of skin t 6 mm
Material of ribs and spars St S355
design is inspired by aircraft wing design. Accordingly, the wings
Thickness of ribs tr 4 mm
consist of three main structural components: skin, ribs and spars (Figs. 9 Thickness of spars ts 6 mm
to 11). Various design alternatives have been developed. In the design Cross-sectional area A 0.1022 m2
alternative presented in the following, the skin is made of aluminum Area moment of inertia Iy 4.036∙10−3 m4
and the ribs and spars are made of steel. Area moment of inertia Iz 113.5∙10−3 m4
Torsional moment of inertia It 10.92∙10−3 m4
The skin together with the spars are the main structural components
Mass (incl. ribs) m 329 kg/m
of a wing and transmit the loads to the supports. The ribs are stiffening Mass moment of inertia (incl. ribs) Im 380 kgm2/m
and forming components that incorporate the elliptical shape of the
wing. The skin adopts this shape when attached to the ribs.
The wings are envisaged as simply supported beams spanning be- 7.2. Design verification
tween the support structures that are attached to the bridge deck at a
spacing of 20 m. The plate thicknesses are established on the basis of The design is verified for static loads and dynamic wind loads. Two
finite shell element analyses. The chosen materials and thicknesses as kinds of static loads are considered: self-weight of wing and static wind
well as the resulting composite cross-sectional properties, based on the load on wing. A wind speed of 60 m/s and a maximum angle of attack of
modulus of elasticity of aluminium, are shown in Table 4. 5° are the assumed design conditions. The static uplift force and the

1078
U. Starossek et al. Engineering Structures 172 (2018) 1073–1080

aerodynamic moment are computed based on potential flow theory

assuming a flat-plate cross section [15]. The resulting maximum uplift
force along the wing axis is 4.9 kN/m and the aerodynamic moment is
4.7 kNm/m.
Dynamic wind loads are determined for vortex shedding off the wings.
The flexibility of the support structures is taken into account. With a
Strouhal number of 0.16 for an elliptical cross section [17] and assuming
a vertical support stiffness of 1800 kN/m (alternative C support structures
described in Section 8.3), the first and second natural frequencies of
vertical vibrations are 2.7 Hz and 5.0 Hz and the respective critical wind
speeds are 10.5 m/s and 18.9 m/s. With these input data, the forces due to
vortex shedding are determined applying the provisions of Eurocode EN
1991-1-4, Appendix E [16]. A more refined determination of these forces
and the inclusion of buffeting loads are desirable and under way. Fig. 13. Support structure – design alternative B.
The stress analysis and ultimate limit state verification is performed
with the commercial analysis program ANSYS using a finite shell ele- require additional bracing in the direction of the bridge axis.
ment model [18,19]. Disadvantages are a relatively high mass of 13 t per support structure, a
Furthermore, it is verified that local flutter of the flutter stabilizer large number of connections of truss members, and possibly main-
itself, that is, of the sub-system consisting of wings and support struc- tenance difficulties due to corrosion and poor accessibility.
tures, will not occur. Once more, the self-developed program DALI
based on finite aeroelastic beam elements [10] and Theodorsen’s flutter 8.2. Design alternative B
derivatives are used. The accuracy of the analysis can be expected to be
good given that the elliptical wing has a streamlined slender cross Design alternative B intends to overcome the disadvantages just
section that is aerodynamically close to the flat plate assumed in outlined. Aluminum instead of steel and a planar truss instead of a
Theodorsen’s theory [14]. The program allows for the modelling of space truss are used (Fig. 13) [21]. The truss members are quadratic or
flexible supports. Flutter speeds well above the design wind speed of circular hollow sections made of aluminum EN-AW 5454 H14. Analysis
60 m/s are obtained for all three support structure design alternatives and verification are performed according to EN 1999-1-1 [22]. The
described in the next section. mass of one support structure is 5.3 t. It can be prefabricated entirely
with all shop joints being welded. The planar trusses are stiffened in the
8. Design of support structures direction of the bridge axis by tension rod bracing with the rods being
anchored at the bridge deck.
The support structures are cantilevers that are attached to the Advantages of this alternative are a reduced weight, a smaller
bridge deck at a spacing along the bridge axis of 20.0 m. They support number of connections, and reduced corrosion and maintenance pro-
the wings that are located 19.0 m outboard of the bridge deck for a blems. On the other hand, the welded connections between the alu-
relative wing eccentricity of ac/b = 2.00 at an assumed deck width of minum elements and between the aluminum trusses and the steel deck
2b = 38.0 m. Self-weight and the static and dynamic loads from the must be carefully designed and executed to avoid corrosion and loss of
wings are taken into account in the design. In the following, three truss structural resistance. Furthermore, the tension rod bracing must be
design alternatives are presented. In all alternatives, the truss is at- designed such that the failure of an individual rod does not trigger a
tached to the upper and lower chord of the bridge deck and is designed failure progression along the bridge.
for ease of transportation and erection.
8.3. Design alternative C
8.1. Design alternative A
A third alternative is a three-chord aluminum truss that combines
the advantages of design alternatives A and B (Fig. 14) [23]. The
Design alternative A is a space truss cantilever made of structural
stiffness in the direction of the bridge axis is greater than in alternative
steel S235 (Fig. 12) [19]. European double symmetric steel profiles are
B so that no bracing is required. The truss members are circular hollow
chosen for all truss members. Analysis and verification are performed
according to EN 1993-1-1 [20].
The truss can be prefabricated entirely with all shop joints being
welded or it is prefabricated in two segments with welded or bolted site
connections. Advantages of this alternative is that the support struc-
tures are made of the same material as probably the bridge deck, fa-
cilitating connections, and that the space truss is stiff and does not

Fig. 12. Support structure – design alternative A. Fig. 14. Support structure – design alternative C.

1079
U. Starossek et al.
sections made of aluminum EN-AW 5454 H14. This design alternative
has been particularly optimized. The mass of one support structure is
3.0 tons.
9. Quantities and cost estimate
On the basis of the design of wings and support structures, quan-
tities are determined and costs are estimated. Concerning the support
structures, design alternative A uses steel and has a mass per support
structure of 13 t. Design alternative B uses aluminum and has a mass
per support structure of 5.3 t. Design alternative C also uses aluminum
and has a mass per support structure of 3.0 tons. The wings are made of
aluminum and steel components and have an aluminum mass of
127 kg/m/side and a steel mass of 202 kg/m/side.
The design of wings and support structures provides for a relative
wing width of 0.100 and a relative wing eccentricity of 2.00 at an as-
sumed deck width of 38.0 m. It was shown in Section 5 that a flutter
speed increase of 22% is achieved in such a configuration when wings
are placed on both sides of the bridge deck over 50% of the bridge
length. A long-span suspension bridge with a main span length of
Lm = 3000 m with a deck width of 2b = 38.0 m was assumed in that
analysis. In the following, the relative cost increase for achieving this
flutter speed increase is estimated.
For comparison, the Great Belt East Bridge is considered. The total
mass of steel deck, hangers, and cables is 22.7 t/m [24]. The deck width
is 31.0 m. The corresponding mass of a 38.0 m wide bridge is estimated
at (38.0 m/31.0 m) × 22.7 t/m = 27.8 t/m. The relative mass increase
– due to wings and alternative C support structures – is
½ × 2 × (0.127 + 3.0/20.0)/27.8 = 1.0% of aluminum plus
½ × 2 × 0.202/27.8 = 0.73% of steel. When the unit price for alu-
minum construction is estimated at 1.8 times the unit price for steel
construction, the relative cost increase is 1.8 × 1.0% + 0.73% = 2.5%.
Note that the cost increase relative to the total construction cost – in-
cluding the cost of pylons and substructure and considering the com-
paratively higher costs of the cables – is actually smaller. Furthermore,
the thus obtained flutter speed increase of 22% is a conservative esti-
mate because the mass considered in the flutter analysis is larger than
27.8 t/m.
10. Conclusions
Flutter analyses substantiated by wind tunnel tests indicate that the
eccentric-wing flutter stabilizer is a promising device for raising the
flutter speed of bridges. The flutter speed increases over-proportionally
with the lateral eccentricity of the wings, an effect due to aerodynamic
damping of the angular motion of the bridge deck.
The device is cost-efficient given that a substantial increase of flutter
speed is reached at a small additional cost. The cost estimate is based on
a design of the wings and their support structures. For a representative
bridge and wing configuration, the flutter speed is increased by at least
22% at a cost increase of 2.5%. Higher flutter speed increases can be
achieved with corresponding designs, for instance, with groups of wings
stacked above each other, without increasing the cost in the same ratio.
The above statement on cost-efficiency is based on flutter analyses
for a streamlined box girder super long-span bridge. Note that the ec-
centric-wing flutter stabilizer can cost-efficiently be applied also to
1080
Engineering Structures 172 (2018) 1073–1080
bridges of smaller spans, say, on the order of 200 m and greater. Such
bridges are nowadays mostly built with streamlined box girder decks
for reducing wind loads and achieving flutter stability. When using the
flutter stabilizer, which also dampens other kinds of wind-induced vi-
brations in addition to flutter, such cross sections could be replaced by
more economical ones such as, say, open plate-girder sections.
References
[1] Richardson JR. The development of the concept of the twin suspension bridge, NMI
R 125. Feltham, UK: National Maritime Institute; 1981.
[2] Ostenfeld KH, Larsen A. Bridge engineering and aerodynamics. In: Larsen A, editor.
Aerodynamics of large bridges. Proc., First international symposium on aero-
dynamics of large bridges. Copenhagen, Denmark: Balkema; 1992.
[3] Starossek U, Aslan H. Passive control of bridge deck flutter using tuned mass
dampers and control surfaces. In: Proc., Seventh European conference on structural
dynamics – Eurodyn 2008, Southampton, UK; 2008.
[4] Diana G, Fiammenghi G, Belloli M, Rocchi D, Resta F, Zasso A. Sensitivity analysis
on the effects of different aerodynamic devices on the behavior of a bridge deck. In:
12th international conference on wind engineering, Cairns, Australia; 2007.
[5] Raggett JD. Stabilizing winglet pair for slender bridge decks. Bridges and trans-
mission line structures. Orlando, Florida: ASCE Structures Congress; 1987.
[6] Gao Liu, Fanchao Meng, Xiuwei Wang. Mechanism of flutter control of suspension
bridge by winglets. In: Fourth international symposium on computational wind
engineering, Yokohama, Japan; 2006.
[7] Starossek U. Eccentric-wing flutter stabilizer for bridges. In: IABSE conference,
Guangzhou, China; 2016.
[8] Starossek U. Eccentric-wing flutter stabilizer for long-span bridges. Bridge Struct
2016;12:3–9. http://dx.doi.org/10.3233/BRS-160098.
[9] Theodorsen T. General theory of aerodynamic instability and the mechanism of
flutter. NACA, Technical Report No. 496; 1934.
[10] Starossek U. Prediction of bridge flutter through use of finite elements. Struct Eng
Rev 1993;5(4):301–7.
[11] Ziems H. Experimentelle Überprüfung der Flatterstabilisierung weitgespannter
Brücken mittels exzentrischer Flügel [Master Thesis]. Structural Analysis Institute,
Hamburg University of Technology; 2015.
[12] Erismis H. Aerodynamische Untersuchung der Flatterstabilisierung eines
Brückenmodells mittels exzentrischer Flügel und numerische Validierung [Master
Thesis]. Structural Analysis Institute, Hamburg Institute of Technology; 2016.
[13] Flaga A. Wind tunnel tests of aerodynamic interference between two high-rise
buildings, EACWE 5. Italy: Florence; 2009.
[14] Klöppel K, Thiele F. Modellversuche im Windkanal zur Bemessung von Brücken
gegen die Gefahr winderregter Schwingungen. Der Stahlbau 1967;36(12):353–65.
[15] Försching HW. Grundlagen der Aeroelastik. Berlin: Springer Verlag; 1974.
[16] Comité Européen de Normalisation. EN 1991-1-4:2005+A1 Eurocode 1: Actions on
structures: Part 1-4: general actions – wind actions, incorporating corrigendum
2010; 2010.
[17] Trivellato F, Castelli MR. Appraisal of Strouhal number in wind turbine en-
gineering. Renew Sustain Energy Rev 2015;49:795–804.
[18] Karlidag I. Flutter stabilization of long-span bridges by means of wings and im-
proving these wings considering conventional wing design [Master Thesis].
Structural Analysis Institute, Hamburg University of Technology; 2016.
[19] Schwarzkopf C, Wiegand A. Entwurf und Berechnung eines Flügels und des
zugehörigen Auslegers zur Flatterdämpfung eines Brückendecks [Project Work].
Structural Analysis Institute, Hamburg University of Technology; 2017.
[20] Comité Européen de Normalisation. EN 1993-1-1:2005 Eurocode 3: Design of steel
structures: Part 1-1: general rules and rules for buildings, incorporating corrigenda
2006 and 2009; 2009.
[21] Vukovic S. Konstruktion und Bemessung des Auslegers eines starren Flügels zur
Flatterstabilisierung von Brücken [Master Thesis]. Structural Analysis Institute,
Hamburg University of Technology; 2017.
[22] Comité Européen de Normalisation. EN 1999-1-1:2007+A1 Eurocode 9: design of
aluminium structures: Part 1-1: general structural rules; 2009.
[23] Goos F, Jaacks F. Bemessung und Optimierung eines Auslegers für eine
Flügelkonstruktion zur Flatterstabilisierung von Brücken [Project Work]. Structural
Analysis Institute, Hamburg University of Technology; 2017.
[24] Larsen A. Aerodynamic aspects of the final design of the 1624 m suspension bridge
across the Great Belt. Wind Eng Indus Aerodyn 1993;48(2–3):261–85.