Journal of Wind Engineering

and Industrial Aerodynamics 84 (2000) 273}288

Reduction of vortex-induced oscillations of

Rio}NiteroH i bridge by dynamic control devices
Ronaldo C. Battista*, Michèle S. Pfeil
COPPE - Engenharia Civil, Federal University of Rio de Janeiro, Caixa Postal 68506, CEP 21945-970,
Rio de Janeiro, Brazil

Abstract

Cross-winds of relatively low velocities have often set into vortex-induced oscillations the
lightly damped and remarkably long three continuous spans of the steel twin-box-girder
Rio}NiteroH i bridge. Whenever this happens the bridge is closed to tra$c of any vehicle, for the
sake of user's comfort and overall safety. However, because of inherent operational di$culties,
in some of these events the tra$c barrier is closed too late and users are left frightened when
crossing on the oscillating bridge. This deterrent aspect of the worlds largest steel-box-girder
bridge is explored herein to brie#y present the conceptual design of passive and active control
devices to attenuate the observed oscillation amplitudes. For this an appraisal of the actual
bridge dynamic behaviour is made "rst by using an experimentally calibrated mathemat-
ical}numerical model, including correlated aeroelastic forces along the spans. The derived
dynamic modal equations, are further combined with optimization techniques to assist in
designing feasible mechanical and robust dynamic passive and active control devices, to
upgrade the serviceability of this bridge and user's comfort. The performance of these dynamic
energy absorbers is demonstrated through comparison of numerical results obtained for time
responses of the original and the controlled structure. ( 2000 Elsevier Science Ltd. All rights
reserved.

Keywords: Dynamic control devices; Rio}NiteroH i bridge; Vortex-induced oscillations

1. Introduction

Brought into service in March 1974 the 13.3 km Rio}NiteroH i bridge spans across
the Guanabara bay in Rio de Janeiro. Most of it is a prestressed concrete structure
but its three central spans (of 200}300}200 m) are bridged by remarkably slender

* Corresponding author.
E-mail address: battista@labest.coc.ufrj.br (R.C. Battista)

0167-6105/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 9 9 ) 0 0 1 0 8 - 7
274 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

continuous steel twin box girders, as seen in Plate 1 and illustrated in Figs. 1 and 2.
The central navigation span is the largest steel girder span in the world and together
with the side and link spans weighs 13 100 t of steel and makes a total length of 848 m
[1,2].
A recurrent aeroelastic aspect of this very slender steel bridge * with average girder
height-to-span length ratio close to 1 * is its across-wind vertical bending motion.
45
Cross-winds of relatively low velocities have often set into vortex-induced oscilla-
tions the lightly damped twin-box-girders. Because of this behaviour, the bridge has
to be closed to tra$c of any vehicle, for the sake of users' comfort and overall
safety, whenever cross-winds reach velocities near 50 km/h (&14 m/s). However,
because of inherent operational di$culties, in some of these events the tra$c
barrier is closed too late and users are frightened when crossing on the oscillating
bridge.
For sustained wind velocities close to 60 km/h the blu! box section bridge experien-
ces vortex-induced oscillations in the "rst vertical bending mode. There is another

Plate 1. View of Rio-NiteroH i bridge from Santos Dumont airport, close to Guanabara bay's mouth. On the
back, well beyond the bay, the Serra do Mar skyline.

Fig. 1. Elevation of central spans of Rio}NiteroH i bridge.

 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 275

Fig. 2. Typical cross-section of the steel twin box-girders.

theoretical vortex frequency lock-in condition, for which the bridge would experience
similar amplitudes of oscillations otherwise in the second bending mode, which may
further occur if a rarer event of sustained wind velocities around 100 km/h takes place.
It is worth mentioning that the braced twin-box-girders confer enough torsional
sti!ness to preclude torsional responses for wind speeds below 200 km/h.
Theoretical-numerical and experimental results, the latter from wind-tunnel tests
on a sectional model, plus factual events corroborate the previous statements [1,3}5].
Strong vertical oscillations that left people panic-stricken and even led some to
abandon their cars, were "rst reported in relation to a storm that occurred in August
17, 1980. Frequent cross-winds of sustained velocity around 55}60 km/h have caused
the majority of the subsequent events, the two most recent in October 16, 1997 and
February 17, 1998, when large amplitudes of oscillations were experienced by users
inside their cars. In these events the sustained relatively low wind speed lasted 5}7 min
and astonishing images of the induced oscillations were captured by new video-
cameras installed on the bridge for tra$c control.
These several latest events have then, of course, caused much concern to the bridge
administration, to experts in structural dynamics and aeroelasticity and above all to
the public in general, with respect to the steel structure's integrity, serviceability and
overall safety.
Among the remedial measures to attenuate wind-induced vibrations one usually
adopted is some sort of aerodynamic appendages attached to the structure to change
its cross-sectional shape and therefore the resultant aeroelastic forces on it. One
attempt in this direction has been made when testing in the wind-tunnel the referred
sectional model [5]. This was made by covering the void between the twin-box-girders
at the lower chord, but no signi"cant di!erence in bridge model response was
observed. Some other aerodynamic appendages could be envisaged in order to
improve this very blu! section, all of them having otherwise a penalty of adding
substantial mass to the bridge. Considering that tra$c loading due to heavy vehicles
traversing the link spans' expansions joints is another important source of bridge
oscillation [1] in the same vertical bending mode excited by wind vortex-shedding, the
most appropriate approach to solve the problem is to apply control devices.
276 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

These deterrent aspects of the world largest steel box-girder bridge were explored to
develop the conceptual design of feasible passive and active control devices [1,3,4] to
attenuate the observed oscillation amplitudes * due to the action of winds as well as
tra$c loading * and, consequently, to improve bridge service life and users' comfort.
Herein an appraisal of the actual bridge dynamic behaviour is made by a math-
ematical 3D FEM model calibrated in terms of experimentally measured frequencies
and associated oscillation modes [1]. The derived modal equations, including corre-
lated aeroelastic forces along the three continuous spans, are further combined with
a multi-objective optimization technique [4,6] to assist in designing simple mechan-
ical and robust dynamic absorbers (tuned-mass-dampers or alternatively an active
control device) that may well be adopted soon to upgrade the serviceability of this
bridge, that has an average daily tra$c of 100 000 vehicles.

2. Mathematical model for vortex-induced responses

The three central steel spans of the 8.8 km long bridge structure over the
Guanabara Bay waters stand at around 65 m above sea level, closer to NiteroH i city as
shown in Fig. 3. The 848 m steel structure crosses the navigation channel coming from
the mouth of the bay, and at that point the bridge axis is in a direction whose
perpendicular lies in the south-west quadrant. Open-ocean approaching winds from
the S-W quadrant are, in this region, the strongest; and wind speeds in the range of
50}70 km/h have high probability of occurrence.

Fig. 3. Location of steel spans of Rio}NiteroH i bridge.

 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 277

Having no upstream obstacles close enough to generate oncoming turbulence, and

located well above calm waters, the steel bridge is subjected to aeroelastic forces
produced predominantly by smooth air in a laminar #ow. It is known that under this
smooth condition the bridge structure is signi"cantly more susceptible to strong
oscillations induced by vortex-shedding than if it were located in rougher terrain.
Large amplitude vortex-shedding-induced oscillations occur within a range of
shedding frequencies bracketing this bridge's "rst natural bending frequency. In this
near-resonant range of synchronized frequencies, and for laminar or low turbulence
incident #ow, the condition of alternating vortices shed in the girders' wake that is
processed in the #uid-structure interaction phenomenon is known as `lock-ina.
Vortex-induced oscillations depend on the structure's response amplitudes and, at
`lock-ina conditions are found to be self-excited and self-limiting oscillations that are
highly sensitive to the structural damping level associated with the bridge's dominant
modes of vibration.
An empirically derived simpli"ed mathematical model [7] describing the across-
#ow response of the sectional model of the bridge for wind-tunnel tests [5], takes the
general form of the one degree of freedom (1 DOF) oscillator,
m6 (yK #2m u y5 #u2 y)"FM (y, y5 , ;, t), (1)
B B B B
where, m6 is the mass per unit span length, related to a two-dimensional strip of the
B
prototype, u the circular frequency of a bending mode, m corresponding modal
B B
damping ratio, y"y(t) being the across-#ow degree of freedom, y5 and yK the corre-
sponding velocity and acceleration.
In Eq. (1) FM is a general #uid-induced forcing function [7] per unit span, de"ned at
lock-in condition (where u+u ) as
B

A B
1 y2 y5
FM " o;2(2D)> (K) 1!e , (2)
2 1 D2 ;
where D is the average across-#ow dimension (e.g. height) of the twin-box-girders,
o the air density, K"uD/; " reduced frequency of vortex-shedding, u"vortex-
shedding frequency that satis"es the Strouhal relation (out side lock-in): uB/;"2pS,
S the Strouhal number, U the laminar wind-#ow velocity, and > and e are,
1
respectively, linear and non-linear parameters representing the components of the
aerodynamic force damping.
The parameters > and e are to be determined from wind-tunnel measurements of
1
two free resonant harmonic motion amplitudes related to two distinct structural
damping values. This is accomplished through the quadratic equation in y,
4pm6 Sm !oD2> #o> ey2/4"0, (3)
B B 1 1
resulting from the condition of zero average energy dissipation at steady amplitudes
(self-limiting motion), in which S is the Strouhal number.
Fig. 4 [5] shows the variation with the prototype wind-speed of the vortex-induced
amplitudes augmented according to geometric scale factor applied to the sectional
model results obtained from wind-tunnel tests for three di!erent damping factors. The
278 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

Fig. 4. Variation with prototype wind speed of the vortex-induced amplitudes augmented according to
geometric scale factor applied to sectional model results from wind tunnel tests [5].

Fig. 5. Lumping of the lock-in forces in FEM model.

sectional model represents the central portion of the mid-span with section height
equal to 7.42 m. It is worth noting that the amplitudes shown in Fig. 4 cannot be taken
directly as an estimate for the prototype amplitudes since the mode shape involved
and the spanwise correlation of the aeroelastic forces could not be represented in
a sectional model. From these experimental results the parameters > and e were
1
calculated for m "1% by using the amplitudes obtained for damping factors equal to
B
0.5% and 1.5%, yielding > "12.0 and e"1724.
1
For a FEM model of the twin-box-girders discretized adequately with space frame
elements [1] the nodal force vector is obtained by lumping the force per unit span,
expressed by Eq. (2), at each node, as shown in Fig. 5. Herein, assumption is made that
the values of > and e obtained from the sectional model results can be extended to
1
the entire span. For a modal analysis the displacement vector is transformed to
 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 279

a generalized coordinate vector yielding a set of modal equations of motion written in

the following form for one bending mode:
m [yK #2m u y5 #u2 y ]"f , (4)
B B B B B B B B
where apart from the parameters already de"ned in relation to Eq. (1) m is the modal
B
mass, y is the generalized displacement and f is the generalized wind force related to
B B
the considered vertical bending mode. Performing the matrix operation f "/t F,
B B
where / is the normalized eigenvector and F is the nodal force vector, f is written as
B B

C D
1 2D e
f " o;2 > W ! W y2 y5 , (5)
B 2 ; 1 2 D2 4 B B
in which
N
W " + /n l , n"2 or 4, (6a)
n Bi i
i/1
l being the in#uence length of the elements adjacent to each node i (see Fig. 5), and
i
/ the eigenvector amplitude at node i corresponding to vertical direction.
Bi
In Eq. (5) the modal summations W and W account for the mode shape e!ect on
2 4
the resultant generalized force f , considering full spanwise correlation of the aeroelas-
B
tic force. The partial spanwise correlation e!ect [8], due to three-dimensionalities of
the vortices can be accounted for by modi"ed modal summations WH
n
N
WH" + /M /n l , n"2 or 4, (6b)
n Bi Bi i
i/1
in which /M is the component associated to node i of /M , which is a shape vector
Bi B
constructed with the absolute values of the components of the normalized eigenvector.
This shape vector allows for the spanwise correlation dependence on the amplitudes
of oscillation due to the organizing e!ect of the oscillating structure on the vortices
shed in the wake, leading to high correlated forces where amplitudes are large and low
correlated ones at locations where amplitudes are small.

3. Passive control of wind-induced oscillations

The basic apparent di$culties of applying distributed passive devices to control the
wind-induced motion of large bridge structure were avoided by using a specially
developed general automatic procedure to assist designing a system of tuned-mass-
dampers (TMDs) to structures displaying either well spaced or compound modes,
subjected to frequency varied forcing functions.
This procedure, intended to deal with multiple degree of freedom (MDOF) struc-
tural systems (i.e., resulting from the discretization in "nite elements), is based on the
conception of substructures for dynamic analysis combined with modal superposition
and goal programming optimization techniques. The former conception leads to
a reduced degree-of-freedom system having a complex eigenvalue problem resulting
280 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

from non-proportional damping. The latter technique leads to optimized values of

frequency, damping factor and mass for each of the absorbers at di!erent locations on
the structure [6]. Due to lack of space this optimization technique is not discussed
herein; the interested reader should also see previous fundamental works [9].
The dynamic behaviour of a structure with TMDs is described by the following
di!erential equation:
(T TM T#M )X$ #(T TC T#C )XQ #(T TK T )X "T TF (7a)
0 A 0 0 A 0 0 0
or
MHX$ #CHXQ #KHX "T TF, (7b)
0 0 0
where F is the vector of dynamic external applied forces, T is a transformation matrix
used for reduction of DOF

C D
1 0
T" (8)
/ /
45 4
and
U "TX , (9a)
0 0
X "(X , X )T, (9b)
0 . 4
U "(U , U )T, (9c)
0 . 4
where the subscripts ( ) and ( ) stand, respectively, for the structural and absorbers
0 A
systems, being / and / , respectively the matrices of eigenvectors and static modes of
4 45
the original structure. X and U stand for the transformed and original substructure
0 0
displacement vectors; the latter written in terms of generalized coordinates, where the
subscripts &s' and &m' denote, respectively, the slave (or ordinary) and master (or
connection) nodes where the TMD's are linked, as indicated in Fig. 6.
Assuming harmonic form of solution of the homogeneous part of Eq. (7b) yields the
complex eigenvalue problem (10),
CH MH KH
AC D C DB
0
k# U"0 (10)
MH 0 0 !M H
where from k, U are used to derive modal equations, on which an optimization
technique is applied to seek for the most e$cient ratios between frequencies (u /u ),
A 0
damping (c /c ), and mass (M /M ), of TMDs and structure [6].
A 0 A 0
For the simpler case of a bridge structure with well-spaced oscillations frequencies
on which is installed one TMD (simulating a group of smaller ones close distributed
around the cross-section of maximum amplitude) to control response amplitudes in
the dominant bending mode, the dynamics of the resulting 2DOF system (bridge
vibrating in one generalized DOF#TMD) is governed by the following coupled
second-order di!erential equations:
m (yK #2m u y5 #u2 y )!c v5 !k v"f , (11a)
B B B B B B B A A B
m vK #c v5 #k v"!m yK , (11b)
A A A A B
 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 281

Fig. 6. Passive control. (a) Typical section with tuned-mass-dampers, (b) substructure with TMDs.

where, apart from the variables and parameters already de"ned in relation to Eq. (4),
m , c , k are, respectively, the mass, damping and sti!ness of the subsidiary dynamic
A A A
system (TMD); and v(t) is the vertical displacement of the passive absorber's mass
relative to the structure, as indicated in Fig. 6.
Eqs. (11a) and (11b) or (4), related to the controlled and uncontrolled systems are
then solved for y (t) through a fourth-order Runge}Kutta numerical integration
B
scheme to investigate the behaviour of the structure under vortex-induced vertical
bending motion and the performance of the TMDs in reducing oscillation amplitudes.

4. Active control

The basic conception of a closed-loop feed-back active control of the motion in the
"rst bending mode by using servo-hydraulic activated masses is illustrated schemati-
cally in Fig. 7, in which F(t) is the external excitation force and u(t) is the resulting
control force added to the system. For motion control of the modal mass m optimum
B
control principles are used to determine u(t).
The dynamics of the resulting 1DOF system (bridge vibrating in one generalized
DOF related to one bending mode#active control force) is governed by the follow-
ing second-order di!erential equation:
m (yK #2m u y5 #u2 y )"f #u, (12)
B B B B B B B B
where apart from the variables and parameters already de"ned in relation to Eq. (4),
u"/Tu is the modal control force.
In state}space notation the dynamic behaviour of the controlled system is described
by the following "rst order di!erential equations:
x5 (t)"Ax(t)#Bu(t)#f (t), (13)
where x is the state vector de"ned by
xT"[x , x ]"[y , y5 ] (14)
1 2 B B
282 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

Fig. 7. Schematic diagram of active control system in the bridge.

and with and

c "2m u m (15)
B B B B

C D
0 1
A" , (16)
!u2 !c /m
B B B
BT"[0, 1/m ], (17)
B
f T"[0, f /m ]. (18)
B B
The resulting control force u(t) in this closed-loop control problem, as indicated in
Fig. 7, is regulated only by the feedback response state vector x(t), which is measured
by means of sensors located on the structure. It follows from modern control theory
[10] that, in this 1 DOF linear regulator problem, the optimal feedback control law
for u(t) is the best linear function of x(t) for minimizing the quadratic objective
function J(t)

G P H
1 t&
Min J" [xT(t)Qx[t]#ru2(t)] dt . (19)
2
0
For a su$ciently large controlling time interval (i.e., t PR) it also follows from this
&
theory that the control law for u(t) is given by
u(t)"!r~1BTPx(t)"Gx(t), (20)
in which G is a constant gain vector and the symmetric matrix P satis"es the Riccati's
nonlinear algebraic matrix equation
Q#P A#ATP!P B r~1BTP"0, (21)
where Q"v Q , Q y is a positive semi-de"nite weighing diagonal matrix associated
11 22
with the controlled components of the state vector x(t), and r is a weighing positive
parameter related to the required level of u(t).
 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 283

Taking into account the symmetry of P and the diagonal form of Q (in which
Q and Q are immaterial constants), the Riccati equation (21) may be rewritten in
11 22
the explicitly form of the following three nonlinear algebraic equations:
Q !2u2 P !r~1m~2P2 "0, (22a)
11 B 12 B 12
P !2m u P !u2 P !r~1m~2P P "0, (22b)
11 B B 12 B 22 B 12 22
Q #2P !4m u P !r~1m~2P2 "0, (22c)
22 12 B B 22 B 22
which may be solved for P by using any appropriate iterative technique, as the
Newton-Raphson approach, resulting after appropriate substitutions in the following
expression for the control force:
u(t)"(rm )~1[!P , !P ][y , y5 ]T. (23)
B 21 22 B B
Alternatively, e$cient algorithms such as the instantaneous optimal control ap-
proach [11] may be used in this closed-loop control problem. In this approach the
objective function is a time-dependent quadratic function J(t)
J(t)"xT(t)Qx(t)#ru2(t), (24)
that is used as a concurrent performance index in which Q, x, r and u have already
been de"ned in relation to Eq. (19).
J(t) being minimized under the constraint given by the state equation (13) leads to
the following closed-loop control law:

C D
*t *t Q
u(t)"! r~1BTQx(t)" 0,! 22 [y , y5 ]T, (25)
2 2r m B B
B
in which *t is the numerical integration step for the solution of the state equation of
controlled motion (13) by means of a standard fourth-order Runge}Kutta scheme.

5. Uncontrolled and controlled dynamic responses

Plates 2a and b show, on a TV-screen, two frozen video images of the oscillating
bridge at two instants of time (&1.5 s apart) corresponding to the maxima upward
and downward displacements of the central spans. They are related to strong vertical
oscillations induced by low velocities cross-winds occurred in October 1997. They
were taken by a video camera intended for tra$c control, installed on a pole located
close to the middle section of Rio's side span. By using a horizontal reference line
drawn on the TV screen, and the central barrier height as a reference value, one can
estimate a total relative displacement of 600 mm between middle section locations at
side and central spans. By knowing, from the numerical eigenvector associated to the
"rst bending mode, the ratio between displacements in those two locations, the
observed amplitude at the middle of central span is approximately $250 mm.
The performance of added passive control is now examined numerically using
base-line parameters extracted from the `as-builta drawings of Rio}NiteroH i bridge
284 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

Plate 2. Maximum wind-induced oscillation amplitude. (a) Upward, (b) downward.

and from a 3D "nite element model (FEM) analysis of the 3 spans continuous steel
twin-box-girders [1]. A feasible mechanical arrangement for this sort of passive
control device can be thought by using distributed small masses along a short strip of
the center span. Table 1 presents the modal parameters related to the "rst vertical
bending mode, together with the corresponding TMD's dynamic characteristics.
Experimental measurements that were carried on the actual bridge [1] showed that
many oscillation modes are excited by the tra$c of heavy vehicles, with main
frequencies ranging from 0.32}0.64 Hz (see Table 2). The calibrated 3D-FEM model
of the bridge structure * including its piers and mechanical bearings (see Fig. 1),
foundations and interaction between piles and soil strata * displays this same
multi-mode behaviour under tra$c loading. But, for wind loading it displays a re-
sponse in the dominant "rst bending mode, as visually observed in the actual
behaviour of the bridge under near-resonant wind forces.
Then, in what follows, numerical results for vortex-induced responses in the "rst
vertical bending mode of the structure, with damping ratio of 1%, are shown for
 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 285

Table 1
Structure and passive TMDs characteristics

Characteristics Structure ( ) 8 TMDs ( ) Ratios ( ) /( )

B A A B
Frequency (Hz) 0.32 0.29 &0.9
Mass, m (t) 5]103 50.0 0.01
Damping, n (%) 1.0 7.5 7.5

Table 2
Frequencies and modes shapes for the prototype

Frequency (Hz) Mode shape from 3D FEM model

Experimental Theoretical

0.32 0.32 First vertical bending

0.48 0.45 First lateral bending
0.55 0.55 Second vertical bending
0.64 0.61 Second lateral bending

lock-in condition at a wind velocity of 62 km/h (&17 m/s). The controlled and
uncontrolled responses are shown herein in terms of vertical displacement vs. time at
the middle section of the centre span; they are shown over a period of time of 240 s,
shortly beyond which steady state is reached for the original bridge.
Figs. 8a and 9a show, respectively, the vortex-induced responses of the bridge under
active and passive control. From the obtained controlled responses, as compared to
the uncontrolled response of the original structure, it seems that the active control has
a better performance than the passive counterpart in reducing and controlling the
amplitudes of the vortex-induced oscillations. Fig. 8b shows the variation of the total
control force u(t) envisaged as being produced by two hydraulic actuators, within
a feasible servo-mechanical system, accelerating masses of smaller values than those
needed for the TMDs.
The main results for the uncontrolled and controlled simulations are summarized in
Table 3, which contains theoretical values of the vertical displacement y at the
B
middle of the center span at about 240 s, together with an estimation made from the
video images.
It should be noted however that the performance of the TMD passive control (see
Figs. 9a and b) could be taken as wholly satisfactory as it enables a reduction of about
85}90% of the uncontrolled displacement amplitudes, keeping low the relative dis-
placement (see Fig. 9b) between the absorber's mass and structure, which is an
important design parameter. Also, passive control with TMDs is simple to design,
construct and install, and its mechanical robustness demands low maintenance.
The mass of each of these eight TMDs is only 6.25 t and corresponds to 0.125% of
the modal mass associated with the "rst bending mode of the bridge (i.e., steel
286 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

Fig. 8. (a) Actively controlled and uncontrolled responses in terms of vertical displacement at middle
section of central span. (b) Total active control force variation.

structure, asphaltic surfacing, concrete barriers, etc.) or to only 0.05% of the total
mass of the steel structure. By any of these "gures the proposed TMDs weigh less than
an ordinary two-axle loaded truck. Dampers of the TMD also resemble those of
a truck. In practice, however, mounting soft elastic springs to accommodate low-
frequency large amplitudes of vertical motion of a considerable mass is not an easy
task. Although amplitudes of the relative dynamic displacement between an absorb-
er's mass and bridge structure are kept well within practical values as shown in Fig. 9b
by using adequate damping, the amplitude of the static displacement of the TMDs
mass (y "m g/k ) reaches a value close to 2.5 m.
45 A A
Even so, in the case of the Rio-NiteroH i bridge the overall dimensions of the
cross-section of its box girders may well accommodate both static plus dynamic
vertical displacement amplitudes of the auxiliary masses. The springs could be
envisaged as several long helical springs hanging from the main #oor beams and
sustaining the TMDs masses (6.25 t) somewhere between top and bottom of the box
girder as illustrated in Fig. 6a.
 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288 287

Fig. 9. (a) Passively controlled and uncontrolled responses in terms of vertical displacements of
middle section of central span. (b) Relative displacement v between TMDs and structure and total spring
force f .
4

On the other hand, although of a more complex design and demanding a better
maintenance, the active control device does not su!er this drawback as in it actuators
accelerate small masses with closed-loop controlled short strokes.

6. Concluding remarks

Subsidiary active or tuned vibration absorbers may play a great role in reducing
dynamic responses of existing large civil structures to excitation forces which were not
envisaged or accounted for in their original design. This is the case of the vortex-
induced oscillations of the world largest span steel box girders bridge: the Rio}NiteroH i
bridge.
With the intent to forward a proposal to upgrade the serviceability of this bridge,
the feasibility of applying passive control devices to attenuate its vortex-induced
288 R.C. Battista, M.S. Pfeil / J. Wind Eng. Ind. Aerodyn. 84 (2000) 273}288

Table 3
Summary of main results for uncontrolled and controlled response amplitudes

Source of estimation > (cm) v (cm)

B A
Original bridge * Full correlation $34
theoretical Partial correlation $26 *
Estimated from video images $25($5)
Bridge with eight TMDs-theoretical $2.1 &$8.0
Bridge with active control &0

oscillations have been investigated. Results from numerical simulation of this aero-
elastic problem combined with some few practical aspects are used to demonstrate
that simple mechanical and robust control devices may come to ful"ll this intention.
Near future monitoring of the bridge [1] will allow for better comparisons between
measured and calculated amplitudes of vortex-induced oscillations, and consequent
re"nement of mathematical models and control devices.

References

[1] R.C. Battista et al., Global analysis of the structural behaviour of the central spans of Rio-NiteroH i
bridge, PONTE SA Contract Report, Vol. 3, ET-150747 COPPETEC, Rio de Janeiro, November
1997 (in portuguese).
[2] J. Upstone, D. Reily, Construction of the navigation spans of the Rio}NiteroH i bridge, Brazil,
Proceedings of the Institution of Civil Engineers, Part 1, Vol. 66, May 1979, pp. 227}246.
[3] R.C. Battista, M.S. Pfeil, Passive damping of vortex-induced oscillations of Rio-Niteroi bridge,
Passive Damping Vol. 2445, Proceedings of SPIE's Smart Structures & Materials Conference, San
Diego, CA, USA, March 1995, pp. 252}263.
[4] R.C. Battista, M.S. Pfeil, Active}passive control of vortex-induced oscillations of Rio}NiteroH i bridge,
EURODYN'96, Proceedings of the Third European Conference on Structural Dynamics, Structural
Dynamics, Florence, Italy, June 1996, Vol. 1, Balkema, Rotterdam, 1996, pp. 561}567.
[5] R. Robinson, M.G. Savage, Wind tunnel investigation of the President Costa e Silva bridge, Rio
de Janeiro, Brazil, Laboratory Technical Report LTR - LA - 311, National Aeronautical Establish-
ment, National Research Council Canada, Ottawa, March, 1989.
[6] R. Battista, C. Magluta, Optmized vibration absorbers for structures with non-proportional damping,
Paper no. 19, Vol. 2193, Passive damping, Proceedings of SPIE's Smart Structures & Materials
Conference, Orlando, FL, February 1994, pp. 202}212.
[7] E. Simiu, R. Scanlan, Wind E!ects on Structures, 3rd Edition, Wiley, New York, 1996.
[8] F. Ehsan, R.H. Scanlan, Vortex-induced vibrations of #exible bridges, J. Eng. Mech. ASCE 116 (6)
(1990) 1392}1411.
[9] J.P. Ignizio, Linear Programming in Single and Multiple Objective System, Prentice-Hall, Englewood
Cli!s, NJ, 1982.
[10] Meirovitch, Dynamics and Control of Structures, Wiley, New York, 1990.
[11] N. Yang, A. Akbarpour, P. Ghaemmaghami, New optimal control algorithms for structural control,
J. Eng. Mech. ASCE 113 (1987) 1369}1386.