THEORY AND FULL-BRIDGE MODELING OF WIND RESPONSE OF

CABLE-SUPPORTED BRIDGES
By Nicholas P. Jones,1 Member, ASCE, and Robert H. Scanlan,2 Honorary Member, ASCE

ABSTRACT: As is well known, long, suspended bridge spans require, in the design stage, careful study of their
resistance and response to site winds. This has driven, on the one hand, detailed quantitative observation of
bridge models in the wind tunnel and, on the other, a steady development and refinement of parallel theory.
Currently, both aspects have arrived at good stages of sophistication, although with continued room for improve-
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

ment. Successes in the extension of bridge spans to record-breaking lengths are mainly due to progress in wind-
resistant design, a primary component in the design of long-span bridges. Recently, multimode flutter and
buffeting analysis procedures have been developed. These procedures, which were based centrally on frequency-
domain methods, take into account the fully coupled aeroelastic and aerodynamic response of long-span bridges
to wind excitation. This paper briefly reviews the current state of the art in long-span bridge wind analysis,
emphasizing the analytical infrastructure. The focus then turns to exhibit an example of application of the theory
to the stability (flutter) and serviceability (buffeting) analyses of a new long-span bridge in North America. This
example not only demonstrates the application of the theory to a real structure but also serves to highlight some
insights into the versatility that is gained by this analytically based approach. The results demonstrate that the
analytical method with appropriate inputs and a complementary full-bridge model agree even for relatively
unusual incoming turbulence in the flow caused by the presence of structures upstream of the bridge. This paper
seeks to exhibit recent developments in the field to the interested structural/bridge engineer, outline alternative
procedures available for assessment of wind effects on cable-supported bridges, and provide an overview of the
basic steps in the process of a typical aerodynamic analysis and design.

INTRODUCTION generalized inertias. Generally, this will include at least

20 modes but, in some cases, more may be required (e.g.,
The maximum span of long-span bridges has been extended
up to 50 for very long structures).
in recent decades to where the Akashi-Kaikyo Bridge (central
• An analytical framework and computational aids for syn-
span 1,991 m) has been completed and longer bridges are
thesizing the above data—Scanlan and Jones (1990) pro-
planned [e.g., Messina Straits (3,300 m) and Gibraltar Straits
vided a comprehensive overview of the single-degree-of-
(5,000 m)]. These successes are mainly due to progress in
freedom theory, and Jain et al. (1996a,b) and Katsuchi et
wind-resistant design, a primary component in the design of
al. (1998a,b, 1999) outlined the extension of this theory
long-span bridges. Recently, multimode flutter and buffeting
to consider the interaction of multiple modes.
analysis procedures have been developed. These procedures,
which were based on frequency-domain methods, take into
account the fully coupled aeroelastic and aerodynamic re- In addition to the above components, knowledge or assump-
sponse of long-span bridges to wind excitation. tions about modal damping values and the wind environment
The current methodology for the estimation of the response are also required.
of long-span bridges to wind loads incorporates the following This paper will briefly outline the current state of the art in
components: analysis-based long-span bridge wind analysis, emphasizing
the analytical infrastructure and outlining recent developments
• Measurement of a comprehensive set of aerodynamic and in this theory, and present results of an example comparing
aeroelastic parameters for a given cross section using a the analytical predictions with full-bridge-model wind tunnel
suitably (i.e., aerodynamically) scaled section model— data for the Carquinez Bridge in California. The goal is to
These parameters include the static coefficients (lift, mo- summarize recent developments in the field to the interested
ment, and drag at a number of different angles of attack) structural/bridge engineer, describe the alternative procedures
and the flutter derivatives (generally also at several posi- available for assessment of wind effects on cable-supported
tive and negative angles of attack). It is emphasized that bridges, and provide an overview of the basic steps in the
these quantities are intended to be sectional quantities that process of a typical aerodynamic analysis and design.
will be used in the analytical model. Examples of proce-
dures can be found in Sarkar et al. (1994) and Singh et BACKGROUND THEORY
al. (1996).
• A detailed numerical (generally finite-element) dynamic The analytical method follows the state-of-the-art analysis
model for the bridge under consideration—This model procedures, as described by Scanlan and Jones (1990), Jain et
will be expected to provide a set of eigenvalues and ei- al. (1996a), and Katsuchi et al. (1999), and is based on modal
genvectors for the structure and a corresponding set of analysis in the frequency domain. The essential components
of this formulation are included below for context; interested
1
Dept. of Civ. Engrg., Johns Hopkins Univ., Baltimore, MD 21218. readers are referred to the above-referenced publications for
E-mail: [email protected] more detail.
2
Deceased May 27, 2001; formerly, Homewood Professor, Dept. of
Civ. Engrg., Johns Hopkins Univ., Baltimore, MD 21218. Basic Definitions and Problem Statement
Note. Discussion open until May 1, 2002. To extend the closing date
one month, a written request must be filed with the ASCE Manager of In a modal analysis, the vertical, lateral, and torsion deflec-
Journals. The manuscript for this paper was submitted for review and
possible publication on April 3, 2001; revised June 21, 2001. This paper
tion components of the bridge deck are represented in terms
is part of the Journal of Bridge Engineering, Vol. 6, No. 6, November/ of the generalized coordinate of the mode ␰ i (t), deck width B,
December, 2001. 䉷ASCE, ISSN 1084-0702/01/0006-0365–0375/$8.00 and dimensionless modal values of the ith mode along the
⫹ $.50 per page. Paper No. 22576. deck hi (x), pi (x), and ␣i (x) (Fig. 1)
JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001 / 365

J. Bridge Eng., 2001, 6(6): 365-375

 FIG. 1. Coordinate Definition and Carquinez Strait Bridge Cross Section
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

h(x, t) = 冘 hi (x)B␰ i (t); p(x, t) = 冘 pi (x)B␰ i (t); 1

␳U 2B 冋 h˙
⫹ KH*
B␣
˙
⫹ K 2H*␣ ⫹ K 2H*
h

冘
i i Lae = KH*
1 2 3 4
2 U U B

册
␣(x, t) = ␣i (x)␰ i (t)
(1) ṗ p
⫹ KH* ⫹ K 2H*
i
5 6
U B (6a)

冋
where x = coordinate along the deck span; and t = time. Note
that vertical, lateral, and torsion components of the deck are 1 ṗ B␣
˙ p
dealt with here because these three components are considered Dae = ␳U 2B KP*
1 ⫹ KP*
2 ⫹ K 2P*␣
3 ⫹ K 2P*
4
2 U U B

册
the primary deflection components due to the wind loading.
Following standard principles of structural dynamics, the ḣ h
governing equation of motion for generalized displacement ␰ i ⫹ KP*
5 ⫹ K 2P*
6
U B (6b)

冋
can be written
1 h˙ B␣
˙ h
Ii [␰¨ i ⫹ 2␨ i ␻i ␰˙ i ⫹ ␻2i ␰ i ] = Qi (t) (2) Mae = ␳U 2B 2 KA*
1 ⫹ KA*
2 ⫹ K 2A*␣
3 ⫹ K 2A*
4
2 U U B
where Ii and Qi (t) = generalized inertia and force of the ith
mode to be defined later; and ␨ i and ␻i = damping ratio to
critical and circular natural frequency of the ith mode, respec-
⫹ KA*
5
ṗ
U
⫹ K 2A*
6
p
B
册 (6c)
tively. The generalized inertia Ii is defined through where ␳ = air density; U = mean wind speed; K(=B␻/U ) =
reduced frequency; and H* i , P*
i , and A*
i = experimentally de-

Ii = 冕 structure
␩ 2i (x, y, z)dm(x, y, z) (3)
termined flutter derivatives of the deck cross section (i = 1
to 6).
Under assumed slowly varying gust action, the buffeting
where ␩ i = ith modal value for the whole bridge structure; and forces are represented through a standard pseudosteady for-
dm = infinitesimal inertia. The generalized force Qi (t) is sim- mulation
ilarly defined through 1
␳U 2B 冋 冉 冊 u
⫹ (C⬘L ⫹ CD)
w
册 1
␳U 2BLb(x, t)

冕
l
Lb = CL 2 =
2 U U 2
Qi (t) = [Lhi B ⫹ Dpi B ⫹ M␣i ] dx (4)
0 (7a)

where l = deck span length; and L, D, and M represent the

lift, drag, and pitching moment per unit span length (Fig. 1).
Db =
1
2
␳U 2B 冋 冉 冊
CD 2
u
U
⫹ C⬘D
w
U
册 =
1
2
␳U 2BDb(x, t) (7b)

冋 冉 冊 册
Representation of the complex fluid-structure interactions
1 u w 1
between the deck and the wind is a formidable modeling task. Mb = ␳U 2B 2 CM 2 ⫹ C⬘M = ␳U 2B 2Mb(x, t) (7c)
For practicality, and on the premise that the focus is on incip- 2 U U 2
ient instability and small motions of the deck under the influ- where CL, CD, and CM = static lift, drag, and pitching moment
ence of turbulent wind, an additive decomposition is adopted. coefficients (referred to deck width B) of a typical deck sec-
The lift, drag, and pitching moment per unit span are repre- tion, respectively; C⬘L = dCL /d␣, C⬘D = dCD /d␣, and C⬘M =
sented as the sum of a self-excited or aeroelastic component dCM /d␣, where ␣ = angle of incidence; and u = u(t) and w =
ae, influenced primarily by the mean speed of the incoming w(t) = along-wind and vertical velocity fluctuations of the
wind, and an external forcing or buffeting component mainly wind, respectively. Note that aerodynamic admittance is as-
due to the turbulent components in the incoming wind flow b sumed under the present circumstances to be represented using
Lift: the flutter-derivative–based form presented in Scanlan and
Jones (1999); details are not provided here.
L = Lae ⫹ Lb (5a)
System Equations of Motion
Drag:
The full multimode system of equations can be expressed
D = Dae ⫹ Db (5b) in matrix notation
Moment: I␰⬙ ⫹ C␰⬘ ⫹ D␰ = Qb(s) (8)

M = Mae ⫹ Mb (5c) where ␰ = generalized coordinate vector; ( )⬘ represents a de-

rivative with respect to dimensionless time; s = Ut/B; I = iden-
For purely sinusoidal motions of frequency ␻, the aeroelas- tity matrix; C and D = damping and stiffness matrices of the
tic forces can be expressed (Scanlan and Jones 1990; Jain et system, respectively; and Qb = generalized force vector. The
al. 1996a; Katsuchi et al. 1999) general terms of C, D, and Qb are expressed
366 / JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001

J. Bridge Eng., 2001, 6(6): 365-375

 冕 冕
l l
␳B 4lK * dx * dx
Cij (K ) = 2␨ i Ki ␦ij ⫺ [H*G
1 hihj ⫹ H*G
2 hi␣j ⫹ H*G
5 hi pj ⫹ P*G
1 pi pj G A␣5i pj = A*(x)␣
5 i (x)pj (x) ; G Hhi3␣j = H*(x)h
3 i (x)␣j (x) ;
2Ii 0 l 0 l

冕 冕
l l
⫹ P*G
2 pi␣j ⫹ P*G
5 pihj ⫹ A*G
1 ␣ihj ⫹ A*G
2 ␣i␣j ⫹ A*G
5 ␣i pj] (9a) * dx * dx
G Hhi4hj = H*(x)h
4 i (x)hj (x) ; G hHi6pj = H*(x)h
6 i (x)pj (x) ;
␳B 4lK 2 0 l 0 l

冕 冕
Dij (K ) = K 2i ␦ij ⫺ [H*G
3 hi␣j ⫹ H*G
4 hihj ⫹ H*G
6 hi pj ⫹ P*G
3 pi␣j l l
2Ii * dx * dx
G Ppi3␣j = P*(x)p
3 i (x)␣j (x) ; G Ppi4pj = P*(x)p
4 i (x)pj (x) ;
⫹ P*G
4 pi pj ⫹ P*G
6 pihj ⫹ A*G
3 ␣i␣j ⫹ A*G
4 ␣ihj ⫹ A*G
6 ␣i pj] (9b) 0 l 0 l

冕 冕 冕
l l l
␳B 4l dx * dx * dx
Qbi (s) = {Lb(x, s)hi ⫹ Db(x, s) pi ⫹ Mb(x, s)␣i} G Ppi6hj = P*(x)p
6 i (x)hj (x) ; G A␣i3␣j = A*(x)␣
3 i (x)␣j (x) ;
2Ii 0 l 0 l 0 l
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

冕 冕
(10) l l
A*4 dx A*6 dx
where ␦ij = Kronecker delta function defined G ␣i hj = A*(x)␣
4 i (x)hj (x) ; G ␣i pj = A*(x)␣
6 i (x)pj (x)
0 l 0 l
␦ij = 1, i = j; ␦ij = 0, i≠j (11) (16)

The modal integrals Gri sj are obtained by integration over The spatial integrals above can usually simply be taken as the
the deck (which is the primary aerodynamic load source) sum of integrals across the different sections; e.g.,

冕 冋 冕 冕 册
l
l* l
dx 1
Gri sj = ri (x)sj (x) (12) G H*1
= (H*) hi (x)hj (x) dx ⫹ (H*) hi (x)hj (x) dx
0 l hi hj
l
1 A 1 B
0 l*

where ri = hi, pi, or ␣i; sj = hj , pj , or ␣j ; and Ki = B␻i /U. (17)

The diagonal terms (i = j) in (9) represent the single-degree-
of-freedom (and uncoupled) equations. The off-diagonal terms where ‘‘section A’’ extends from 0 to l* and ‘‘section B’’ ex-
introduce the aeroelastic coupling through the flutter deriva- tends from l* to l. Analysis then proceeds with the modified
tives and the mechanical coupling through the cross-modal terms.
integrals among different modes.
These equations of motion can be transformed into the re-
Flutter Instability Condition
duced frequency (K = B␻/U ) domain and written simply
¯b
E␰ = Q (13) At the incipient flutter condition, the system exhibits har-
monic motions without external excitation, so (13) reduces to
where
Eij = ⫺K 2␦ij ⫹ iKCij (K ) ⫹ Dij (K ) (14) E␰ = 0 (18)

To incorporate variations associated with aeroelastically dif- Solving this aeroelastically influenced eigenvalue problem
ferent sections, the flutter derivatives can be changed along identifies the flutter condition. The determinant of matrix E
the span (i.e., as functions of x). This may be desired to ac- must vanish to obtain a nontrivial solution for (18). Because
commodate for erection differences in the section at different the matrix E is complex, the condition of det E = 0 must be
locations or to account for proximity of the suspension cable satisfied, requiring that both the real and the imaginary parts
near the center of the structure. For long bridges, the change of the determinant be simultaneously zero.
in the pseudostatic angle of incidence due to the mean wind
can also be included. In this case, (12) must include the flutter
derivatives inside the integral and (9) and (12) are modi- Simplified Cases: Single-Mode and Two-Mode
fied to Coupled Flutter

␳B 4lK * * * * In many practical situations, the coupling that can occur in

Cij (K ) = 2␨ i Ki ␦ij ⫺ [G Hhi1hj ⫹ G Hhi2␣j ⫹ G Hhi5pj ⫹ G Ppi1pj an aeroelastically influenced flutter situation is dominated by
2Ii
* * * * *
two modes if, in fact, it is not induced by a single-mode,
⫹ G Ppi2␣j ⫹ G Ppi5hj ⫹ G A␣i1hj ⫹ G ␣Ai2␣j ⫹ G A␣i5pj] (15a) negative damping condition.
In the case of single-degree-of-freedom flutter, the single-
␳B lK 4
* * * *
mode flutter criterion (direct derivatives only) can be written
Dij (K ) = K 2i ␦ij ⫺ [G Hhi3␣j ⫹ G Hhi4hj ⫹ G Hhi6pj ⫹ G Ppi3␣j
2Ii

冋 册
1/2
* * * * * 4␨ i Ii ␳B 4l
⫹ G Ppi4pj ⫹ G pPi6hj ⫹ G A␣i3␣j ⫹G A␣i4hj ⫹ G A␣i6pj] (15b) H*Ghh ⫹ P*Gpp ⫹ A*G ␣␣ ⱖ 1⫹ A*G ␣␣
␳B 4l
1 1 2 3
2Ii
where (19a)

冕 冕
l l
* dx * dx
G Hhi1hj = H*(x)h
1 i (x)hj (x) ; G Hhi2␣j = H*(x)h
2 i (x)␣j (x) ; and, if cross-coordinate derivatives (H*
2 , H*
3 , H*
4 and A*
1 and
0 l 0 l A*)
4 are used

冕 冕
l l
dx dx
hh ⫹ P*Gpp ⫹ A*G ␣␣ ⫹ A*G h␣ ⫹ H*G
* *
G Hhi5pj = H*(x)h
5 i (x)pj (x) ; G Ppi1pj = P*(x)p
1 i (x)pj (x) ; H*G
1 1 2 1 2 h␣

冋 册
0 l 0 l 1/2

冕 冕
4␨ i Ii ␳B 4l
l
dx
l
dx ⱖ 1⫹ (A*G␣␣ ⫹ H*Ghh ⫹ H*G h␣ ⫹ A*G h␣)
␳B 4l
3 4 3 4
P*2 P*5 2Ii
G pi ␣j = P*(x)p
2 i (x)␣j (x) ; G pi hj = P*(x)p
5 i (x)hj (x) ;
0 l 0 l (19b)

冕 冕
l l
* dx * dx The modal integrals to be used in the above expressions are
G A␣1i hj = A*(x)␣
1 i (x)hj (x) ; G A␣2i ␣j = A*(x)␣
2 i (x)␣j (x) ;
0 l 0 l defined
JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001 / 367

J. Bridge Eng., 2001, 6(6): 365-375

 冕 冕
l l
dx dx where
h2i (x)

冉冊
Ghh = ; Ghp = hi (x)pi (x) ;
0 l 0 l ␻1
2
␳B 4l

冕 冕
A11 = ⫺1⫺ [H*G h1h1 ⫹ H*G h1␣1
␻
l l 4 3
dx dx 2I1
Gpp = p2i (x) ; Gh␣ = hi (x)␣i (x) ;
l l ⫹ A*G␣1h1 ⫹ A*G␣1␣1] (23a)
0 0
4 3

冕 冕 冉冊
l l 2

␣2i (x)
dx dx ␻2 ␳B 4l
G␣␣ = ; Gp␣ = pi (x)␣i (x) A22 = ⫺1⫺ [H*G h2h2 ⫹ H*G h2␣2
␻
4 3
0 l 0 l 2I2

and must be computed consistent with the generalized mass ⫹ A*G

4 ␣2h2 ⫹ A*G
3 ␣2␣2] (23b)
in (3).
␳B 4l
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

In the case of two-mode coupled flutter, the two engaged A12 = ⫺ [H*G
4 h1h2 ⫹ H*G
3 h1␣2 ⫹ A*G
4 ␣1h2 ⫹ A*G
3 ␣1␣2] (23c)
modes frequently take on similar shape: one dominant in the 2I1
vertical direction and one in the torsional direction. Even in ␳B 4l
complex cases arising in very long-span structures (e.g., the A21 = ⫺ [H*G
4 h2h1 ⫹ H*G
3 h2␣1 ⫹ A*G
4 ␣2h1 ⫹ A*G
3 ␣2␣1] (23d )
2I2
Akashi-Kaikyo), the principal coupling occurs between a pair
of modes (Katsuchi et al. 1999). This observation can be in-
ferred from the modal integrals: two modes of the structure
are unlikely to couple if their modal forms are very different
B11 = 2␨ 1 冉冊
␻1
␻
⫺
␳B 4l
2I1
[H*G
1 h1h1 ⫹ H*G
2 h1␣1

from each other, because the cross-integral terms that provide ⫹ A*G
1 ␣1h1 ⫹ A*G
2 ␣1␣1] (24a)

冉冊
the coupling are unlikely to be large. Hence a reasonable
choice of two active modes is a pair that has modal resem- ␻2 ␳B 4l
B22 = 2␨ 2 ⫺ [H*G h2h2 ⫹ H*G h2␣2 ⫹ A*G ␣2h2
␻
1 2 1
blance. Further, these modes should be low in the modal hi- 2I2
erarchy according to frequency. A third criterion is that modes
chosen for coupling analysis should each have only one strong ⫹ A*G
2 ␣2␣2 ⫹ P*G
1 p2 p2] (24b)
component, vertical or torsional. ␳B l 4

Assuming, then, that only a pair of modes couple, the above B12 = ⫺ [H*G
1 h1h2 ⫹ H*G
2 h1␣2 ⫹ A*G
1 ␣1h2 ⫹ A*G
2 ␣1␣2] (24c)
equations can be simplified to the following pair, expressed in 2I1
terms of circular frequency rather than reduced frequency as ␳B 4l
originally shown by Scanlan (1981): B21 = ⫺ [H*G
4 h2h1 ⫹ H*G
3 h2␣1 ⫹ A*G
4 ␣2h1 ⫹ A*G
3 ␣2␣1] (24d )
2I2

冋冉 冊 冉 冊册
2
␻1 ␻1 ␳B 4l Finding the conditions for zero determinant are equivalent to
⫺ 1 ⫹ 2i␨ 1 ␰1 = [(i[H*G h1 h1 ⫹ H*G h1 ␣1 finding the roots of the following pair of equations:
␻ ␻
1 2
2I1

⫹ A*G Real:
1 ␣1 h1 ⫹ A*G
2 ␣1 ␣1] ⫹ [H*G
4 h1 h1 ⫹ H*G
3 h1 ␣1 ⫹ A*G
4 ␣1 h1

A11 A22 ⫺ B11 B22 ⫺ A21 A12 ⫹ B21 B12 = 0 (25a)

⫹ A*G
3 ␣1 ␣1])␰ 1 ⫹ (i[H*G
1 h1 h2 ⫹ H*G
2 h1 ␣2 ⫹ A*G
1 ␣1 h2 ⫹ A*G
2 ␣1 ␣2]
Imaginery:
⫹ [H*G
4 h1 h2 ⫹ H*G
3 h1 ␣2 ⫹ A*G
4 ␣1 h2 ⫹ A*G
3 ␣1 ␣2])␰ 2] (20a)
A11 B22 ⫹ B11 A22 ⫺ A21 B12 ⫺ B21 A12 = 0 (25b)

冋冉 冊 冉 冊册
2
␻2 ␻2 ␳B 4l This can be accomplished, for example, by fixing a value of
⫺ 1 ⫹ 2i␨ 2 ␰2 = [(i[H*G h2 h1 ⫹ H*G h2␣1
␻ ␻
1 2
2I2 U/nB and finding the roots of the above two equations. The
flutter condition—if it occurs—is when both equations have
⫹ A*G
1 ␣2h1 ⫹ A*G
2 ␣2␣1] ⫹ [H*G
4 h2h1 ⫹ H*G
3 h2␣1 ⫹ A*G
4 ␣2h1 roots for the same U/nB and ␻ in the frequency range of in-
terest.
⫹ A*G
3 ␣2␣1])␰ 1 ⫹ (i[H*G
1 h2h2 ⫹ H*G
2 h2␣2 ⫹ A*G
1 ␣2h2 ⫹ A*G
2 ␣2␣2
For a full multimode problem, the same procedure is re-
⫹ P*G quired and the lowest solution of ␻ of all solutions gives the
1 p2 p2] ⫹ [H*G
4 h2h2 ⫹ H*G
3 h2␣2 ⫹ A*G
4 ␣2h2 ⫹ A*G
3 ␣2␣2])␰ 2]
flutter-critical condition. The mode corresponding to the so-
(20b) lution of ␻ is the leading mode in the flutter condition. More-
over, the eigenvector ␰ at the flutter condition gives the flutter
Note that in this pair of equations it has been assumed for mode shape that indicates the relative participation of each
simplicity that the following flutter derivatives are of much structural mode in flutter.
lesser importance: A*5 , A*
6 , H*
5 , H* 2 ⫺ P*
6 , and P* 6 . These can
be readily included if available from the wind tunnel investi- Buffeting (Serviceability) Analysis
gation. The flutter derivative P* 1 is frequently approximated
from its pseudosteady form The buffeting analysis involves the estimation of the fluc-
tuating lift, drag, and moment forces acting on the structure
2CD
because of the incoming turbulence in the flow and the esti-
1 = ⫺
P* (21) mation of the structural response to those loads. The essential
K components of this analysis include
where CD = drag coefficient for the deck, measured in the wind • Incorporation of the aeroelastic components of the wind
tunnel. loads through the flutter derivatives per (6) (this allows
The determinant associated with the above pair of equations for the modification of modal damping and frequency val-
can be written in simplified form as follows: ues due to the influence of the self-excited forces)

冏 冏
• Representation of the sectional lift, drag, and moment
A11 ⫹ iB11 A12 ⫹ iB12 forces using a pseudosteady formulation [(7)]
=0 (22)
A21 ⫹ iB21 A22 ⫹ iB22 • Correction (as appropriate) through the use of the so-
368 / JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001

J. Bridge Eng., 2001, 6(6): 365-375

 called aerodynamic admittance functions for deviations in which the reduced frequency f = nz/U, where n = frequency
from pseudosteady behavior of the fluctuating component, z = height above ground or water
• Modeling the gust components of the incoming turbulent surface, and U = mean wind speed. The friction velocity u*
wind is a function of the surface roughness and can be estimated
• Assessing the 3D spanwise correlation of the induced from knowledge of the meterological environment (this is dis-
forces on the deck cussed more fully in the example presented below.)
Computation of the buffeting forces involves the computa-
The vector of generalized buffeting forces on the right-hand tion of ‘‘correlation integrals’’ of the form

冕冕
side of (13) is l l
dxA dxB

冕
Tmn
l H risj (K ) = ri (xA)sj (xB)Tmn(xA, xB, K ) (32)
1 dx 0 0 l l
F̄b1
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

I1 l
where ri and si = hi, pi, or ␣i; m and n = u or w; and T = C
冕
0
l
1 dx or S. Numerical computation of this integral is potentially the
F̄b2 most time-consuming component of the calculation of the buf-
␳B 4l I2 0 l
Q̄b = ⭈⭈ (26) feting response.
2 ⭈

冕
l
The power spectral density matrix for the generalized co-
1 dx ordinate ␰ is developed using standard random vibration theory
F̄b [(13)]
In 0 n l

where the integrands in the vector above are S␰␰(K ) = E⫺1SQbQb[E*]⫺1 (33)
¯ b(x, K )hi(x) ⫹ D
¯ b (x, K ) = L
F ¯ b(x, K )pi(x) ⫹ M
¯ b(x, K )␣i(x) where E* = complex conjugate transpose of matrix E.
i

(27) The PSD of the physical displacements for (1) can be ob-
tained from the PSD of the respective generalized displace-
Substituting for the terms above from the appropriately ment components through (for example)
transformed (7) and following standard manipulations, the
power spectral density (PSD) matrix can be developed, a gen-
eral term of which is
Shh(xA, xB, K ) = 冘冘i j
B 2hi (xA)hj (xB)S␰i␰ j (K ) (34)

冉 冊 冕冕
2 l l
␳B 4l 1 where i and j represent the summation over the number of
SQb Q (K ) = {q˜ i (xA)q˜ j (xB)Suu (xA, xB, K ) modes being used.
i bj
2U Ii Ij 0 0
Evaluation of the spectral densities of the displacements at
⫹ r˜i(xA)r˜j(xB)Sww(xA, xB, K ) ⫹ [q˜ i(xA)r˜j (xB) combinations of discrete xA and xB will result in a matrix. The
mean-square values of these displacements can be evaluated
⫹ r˜i(xA)q˜ j (xB)]Cuw (xA, xB, K ) ⫹ i[q˜ i (xA)r˜j(xB) in terms of their respective PSD functions, for example

冕
⬁
dxA dxB
⫺ r˜i (xA)q˜ j (xB)]Quw (xA, xB, K )} ␴2h(xA, xB) = Shh(xA, xB, f ) df (35)
l l (28) 0

where where f = frequency. Covariance matrices for h, p, and ␣ are

q̃i(x) = 2[CL hi(x) ⫹ CD pi(x) ⫹ CM ␣i(x)] (29a) thus also obtained, from which statistics of the displacement
components h, p, and ␣ can be calculated.
r̃j(x) = (C⬘L ⫹ CD)hj(x) ⫹ C⬘D pj(x) ⫹ C⬘M ␣j(x) (29b) Neglecting any coupling terms among modes results in a
and in which Suu and Sww = spanwise cross-spectral densities single-mode buffeting calculation. Based on the formulations
of the u and w components, respectively; and Cuw and Quw = developed above, the PSD of the physical displacements at a
spanwise uw cospectral and quadrature-spectral densities, re- specific point x, for the single-mode buffeting, Shi (x, K ),
spectively. The latter two terms are frequently neglected in Spi (x, K ), and S␣i (x, K ) can be obtained by the same proce-
analysis because of lack of adequate models (particularly in dures as those in the multimode buffeting.
the case of Quw) or the relatively minor contributions expected Finally, the mean-square values of these displacements can
in most cases (Jones et al. 1992). be evaluated

冕
A number of different spectral forms are available in the ⬁

literature that have been used in modeling. For example, in ␴2qi (x) = Sqi (x, f ) df (36)
modeling the wind tunnel response of the Akashi-Kaikyo 0

Bridge, Katsuchi et al. (1999) used spectral forms suggested

by Roberts and Surry (1973). In other cases, standard spectra where qi = hi, pi, or ␣i.
defined in Kaimal et al. (1972) and Simiu and Scanlan (1996) For the purpose of evaluating the response in the multimode
have proven both versatile and suitable for both wind tunnel sense from single-mode responses, the square root of the sum
and prototype predictions. For atmospheric turbulence, the of square (SRSS) of single-mode responses is used; i.e.
standard Kaimal spectrum for longitudinal fluctuations takes SRSS(x) = 兹␴2q1(x) ⫹ ␴2q2(x) ⫹ ⭈ ⭈ ⭈ ⫹ ␴2qn(x) (37)
the form
where n = number of modes. Note that this form may or may
nSuu(n) 200 f
= (30) not be conservative as it neglects any cross coupling among
u2* (1 ⫹ 50 f )5/3 the modes induced by aeroelastic influences; however, for a
bridge of this length, significant coupling is not anticipated.
and the corresponding spectrum for vertical fluctuations is ex-
pressed
EXAMPLE
nSww(n) 3.36 f
= (31) The example discussed herein focuses on the Carquinez
u2* 1 ⫹ 10 f 5/3 Strait Bridge north of San Francisco Bay, built as a replace-
JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001 / 369

J. Bridge Eng., 2001, 6(6): 365-375

 ment structure to carry Interstate 80 over the Carquinez Strait. 1 ⫻ 4 m, with wind speeds continuously variable from 0 to
An artist’s rendering of the bridge is shown in Fig. 2. 6.1 m/s.
The details of the flutter analysis are omitted for brevity but A number of assumptions and adjustments are made to the
showed that the bridge was stable in both single-mode and standard analytical model to account for the unique properties
multimode flutter to a wind speed of 74 m/s (163 mi/h). of the wind tunnel and the full-bridge model. Although the
The focus of the investigation reported herein was on the model is designed to be a reasonable scaled version of the full
effects of the two existing truss bridges that are located adja- structure, a number of differences are evident and these must
cent to the new structure. The effect of these structures on the be captured in the modeling process. The comparison shows
turbulence in the incoming flow required assessment, which that the analytical model is capable of capturing the typical
necessitated some adjustments to the standard application of responses observed in the wind tunnel and can therefore be
the modeling theory presented above. The predictions were used with some confidence for the full-bridge prototype pre-
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

then compared to the results from a full-bridge model also dictions. It is also emphasized that the goal of the full-bridge
tested at the West Wind Laboratory (WWL), Carmel, Calif. model was to calibrate the analytical procedure that was used
(Fig. 3). The bridge was modeled at a scale of 1:250, resulting for the simulations; close matches of scaled dynamic proper-
in a model of width 116 mm and depth (excluding bar- ties were not rigorously sought.
riers) of 12 mm. The wind tunnel cross section has dimensions
Structural Parameters

Dimensions
For simulation purposes, full-scale dimensions were used
throughout. The final result was then scaled down appropri-
ately to facilitate comparison with the model values measured
on the full-bridge model; i.e., predicted responses were unfac-
tored full-scale estimates that were then reduced to model
scale.

Modal Quantities
Characteristics of the first 20 modes are included in Table
1, including frequencies and modal integrals. The first lateral,
vertical, and torsion modes have frequencies of 0.17, 0.19, and
0.47 Hz, respectively. Note that the relative magnitude of the
modal integrals indicates the dominant motion associated with
FIG. 2. Carquinez Strait Bridge the mode. Scaled data used in the simulations are reported in
Table 2.

Generalized Inertia
The analytical model uses dimensionless variables wherever
possible. For example, the generalized inertia (kip ⭈ s2/in. ⭈ in.2
= kip ⭈ s2 ⭈ in.) is normalized as follows:
I
Î = (38)
␳B 4ᐍ
For a perfectly scaled model, the Î will be the same as that
for the full bridge. However, it was reported that both the mass
per unit length of the deck (which dominates the generalized
inertia for the vertical and lateral modes) and the mass moment
of inertia per unit length of the deck (which dominates the
generalized inertia for the torsion modes) were high for the
model (by factors of 1.31 and 2.0, respectively).
For this reason, to represent the full-bridge model consis-
tently, the generalized inertias were increased by the same fac-
tors (the former for vertical and lateral and the latter for tor-
sion).

Structural Damping
Damping values used in the modeling were obtained from
free-vibration, ring-down tests on the full-bridge model. These
values—which may not be representative of full scale—are
used directly for the prediction of the full-bridge model re-
sponse under the various turbulence spectra. Values used are
summarized in Table 2.

Meteorological Characteristics

Wind Speed Profile

The basic log-law is assumed to describe the wind speed
FIG. 3. Full Bridge Wind Tunnel Model (Courtesy WWL) profile
370 / JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001

J. Bridge Eng., 2001, 6(6): 365-375

 TABLE 1. Modal Properties (from OPAC Finite-Element Analysis)

Frequency Period
Mode number (Hz) (s) Ghh Gpp Gaa Gph Gpa Gha
1 0.146 6.87 6.47E⫺09 1.55E⫺14 5.07E⫺14 7.62E⫺12 ⫺2.26E⫺14 ⫺1.76E⫺11
2 0.166 6.02 8.09E⫺15 6.95E⫺09 2.82E⫺11 1.88E⫺12 ⫺4.41E⫺10 ⫺1.51E⫺13
3 0.194 5.15 7.37E⫺09 7.17E⫺16 2.05E⫺16 ⫺2.02E⫺12 1.35E⫺16 ⫺4.62E⫺13
4 0.261 3.84 7.36E⫺09 5.47E⫺16 1.20E⫺14 2.02E⫺13 8.71E⫺16 ⫺3.31E⫺12
5 0.307 3.26 6.29E⫺14 1.35E⫺09 1.01E⫺11 ⫺2.40E⫺12 ⫺1.23E⫺11 ⫺2.18E⫺14
6 0.349 2.87 2.62E⫺13 6.05E⫺10 1.59E⫺11 ⫺6.07E⫺13 ⫺8.02E⫺11 ⫺1.55E⫺13
7 0.357 2.80 7.36E⫺09 2.21E⫺14 2.20E⫺13 1.75E⫺12 ⫺3.39E⫺14 1.44E⫺12
8 0.380 2.63 3.37E⫺11 4.74E⫺14 1.95E⫺16 4.19E⫺13 1.03E⫺15 ⫺1.52E⫺14
9 0.391 2.56 5.37E⫺09 1.29E⫺14 1.75E⫺13 2.64E⫺12 ⫺4.39E⫺14 ⫺3.26E⫺12
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

10 0.440 2.27 4.67E⫺14 2.68E⫺12 2.52E⫺10 ⫺8.70E⫺14 2.39E⫺11 ⫺2.32E⫺13

11 0.442 2.26 1.09E⫺13 9.90E⫺13 7.26E⫺11 ⫺9.73E⫺14 7.60E⫺12 ⫺2.37E⫺13
12 0.449 2.23 3.95E⫺15 4.54E⫺10 8.17E⫺09 2.13E⫺13 1.50E⫺09 1.36E⫺12
13 0.455 2.20 3.57E⫺14 7.11E⫺10 8.94E⫺09 2.52E⫺13 1.37E⫺09 2.23E⫺12
14 0.472 2.12 1.55E⫺14 1.27E⫺09 3.29E⫺08 ⫺1.94E⫺12 1.64E⫺10 6.13E⫺12
15 0.478 2.09 5.63E⫺09 5.29E⫺14 1.59E⫺13 2.51E⫺12 ⫺4.91E⫺14 ⫺5.84E⫺13
16 0.489 2.04 7.06E⫺14 2.72E⫺09 6.50E⫺09 ⫺2.28E⫺12 ⫺7.29E⫺10 1.05E⫺12
17 0.493 2.03 5.67E⫺10 6.63E⫺14 2.23E⫺12 ⫺5.04E⫺13 ⫺9.69E⫺14 ⫺2.52E⫺12
18 0.537 1.86 1.66E⫺13 4.03E⫺13 7.74E⫺14 ⫺6.63E⫺15 5.92E⫺15 1.65E⫺14
19 0.542 1.84 4.74E⫺09 2.33E⫺15 4.33E⫺13 1.10E⫺12 2.83E⫺15 ⫺1.06E⫺11
20 0.567 1.76 3.07E⫺13 4.21E⫺10 1.61E⫺09 ⫺1.50E⫺13 ⫺8.28E⫺11 1.48E⫺12

TABLE 2. Assumed Structural Damping from Model Tests (from TABLE 4. Spectral Model Parameters Fitted
WWL)
Condition A B C
Scaled
Coastal fetch 200 50 1.5
Mode Dampinga Frequency, frequency, Coastal fetch ⫹ 1958 bridge 160 50 1.15
number Mode type (%) modeledb target Coastal fetch ⫹ 1927, 1958 bridges 105 50 0.75
1 First vertical, 3.0 2.3 2.3
asymmetric
2 First lateral, 0.6 2.5 2.6 To simulate the presence of the upstream bridges, an arti-
symmetric ficial, ‘‘equivalent’’ boundary layer flow description was de-
3 First vertical, 3.5 2.4 3.0 veloped as follows:
symmetric
4 Second vertical, 0.3 3.6 4.1 • The relationship between ␤ and z0 tabulated in Simiu and
asymmetric
14 First torsion, 0.8 3.7 7.5 Scanlan (1996) is used.
symmetric • Using the measured turbulence in the tunnel due to the
a upstream bridges and (39)–(41) above, a second relation-
Note that actual full-bridge model damping was reported as >5%.
b
These are frequencies measured on WWL model and simulated in ship between ␤ and z0 can be determined (for known z)
prediction.
兹␤ = 2.5I ln 冉冊
z
z0
(42)
TABLE 3. Turbulence Intensity, ␤, and Surface Roughness Estimates
• The intersection of these two ‘‘curves’’ provides an esti-
I mate of both the ␤ and the z0 for the particular turbulence
Condition (%) ␤ z0 intensity measured I.
Coastal fetch 11.1 6.5 0.005
Coastal fetch ⫹ 1958 bridge 16.8 5.65 0.175 For the cases considered herein, the parameters estimated
Coastal fetch ⫹ 1927, 1958 bridges 23.7 4.69 1.285 by the above procedure are shown in Table 3. With these val-
ues, a ‘‘pseudoatmospheric boundary layer’’ is defined for the

冉冊
bridge location, which simulates the turbulence intensity and
z mean speeds measured. In particular, the friction velocity u*
U = 2.5u* ln (39)
z0 can be estimated as required for spectral calculations (see
below).
where U = mean wind speed; u* = friction velocity; z = ele-
vation; and z0 = roughness length. Wind Spectra
For the coastal fetch, the roughness length was assumed to
Spectra of longitudinal velocity fluctuations were measured
be z0 = 0.005 m and the structure was modeled at 38.7 m/s
by the WWL for the full-bridge model under various condi-
(85 mi/h; 7.9 ft/s; 1,500 in./s). Knowledge of the deck ele-
tions. Of relevance here are the coastal fetch, coastal fetch with
vation then enables the computation of the u*. The turbulence
the 1958 bridge, and coastal fetch with the 1958 and 1927
intensity
bridges. No measurement of the vertical velocity fluctuations
兹 u2 was made.
I= (40) For atmospheric turbulence, the standard Kaimal spectrum
U
(Simiu and Scanlan 1996) for longitudinal fluctuations takes
can be estimated from the form in (30) above. For simulation purposes, the following
procedure was adopted:
u2 = ␤u*
2
(41)
• Measured longitudinal spectra were fitted over the range
where u2 = mean square velocity fluctuation; and ␤ is defined of (at model scale) 0.8–10 Hz with a perturbed form of
below. the Kaimal spectrum:
JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001 / 371

J. Bridge Eng., 2001, 6(6): 365-375

 nS(n) Af
= (43)
2
u* (1 ⫹ Bf )C

• The spectrum of vertical velocity fluctuations was then

modeled (using the same exponent C and a scaled ampli-
tude A ):
nS(n) 3.36 ⭈ (A/200) f
= (44)
u2* (1 ⫹ 10 f )C

• For frequencies >25 Hz, the spectral forms in (30) and

FIG. 6. Measured and Fitted Spectra for Coastal Fetch ⫹ 1958 and
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

(31) were assumed.

• For frequencies 10 Hz < n < 25 Hz, a linear extrapolation 1927 Bridges
between the 10 and 25 Hz values was assumed. (Note that
all the model frequencies of interest were between 2 and
4 Hz, so the fits outside this range are of limited interest.)

Fits to the recorded data (0.8–10 Hz range) are summarized

in Table 4 and Figs. 4–6, and a comparison of all spectra used
for simulation in Fig. 7. In each case, the ‘‘Sw estimate’’ is
the spectrum of vertical fluctuations estimated from (44)
above.

Spanwise Coherence
Loss of coherence along the span is modeled analytically
using the following functional form:

Srr(xA, xB, n) = Srr(xA, n)exp 冉 冉

⫺k
xB
ᐍ
⫺
xA
ᐍ
冊冊 (45)

where k = cnᐍ/U ; and rr denotes the velocity or force com-

ponent of interest. Past experience has suggested that with a
FIG. 7. Comparison of Spectra Used in Simulation
coastal or close-to-coastal fetch, a value of c = 5 is appropriate
for the aerodynamic loads. This was also assumed for the
coastal fetch herein. the presence of the upstream bridge(s) reduces the spanwise
Based on the WWL data on correlation, it is evident that coherence in the flow considerably. Using exponential fits to
the WWL correlation data, the factor c = 5 for the coastal fetch
TABLE 5. Loss of Correlation Parameters Assumed was increased (implying reduced coherence) to the following
for simulation purposes. Values used are shown in Table 5.
Condition c
Coastal fetch 5.00 Aerodynamic/Aeroelastic Parameters
Coastal fetch ⫹ 1958 bridge 6.67
Coastal fetch ⫹ 1927, 1958 bridges 8.33
Flutter Derivatives
Dimensionless flutter derivatives used for the simulations
were determined from the WWL section model tests per-
formed at a scale of 1:50 (model dimensions 1,620 mm long,
580 mm wide, and 60 mm deep) and are shown in Fig. 8.
Specifically, set CSB19 was used for the west wind conditions
and set CSB26 was used for the east wind conditions. It was
assumed that the flutter derivatives measured on the section
model were wholly applicable to the full-bridge model.

Static Coefficients
FIG. 4. Measured and Fitted Spectra for Coastal Fetch Static coefficients used for the simulations were determined
from the WWL section model tests (Fig. 9). It was assumed
again that the static coefficients measured on the section model
were wholly applicable to the full-bridge model.

Results of Comparison
Comparisons between measured and predicted responses are
outlined in detail below and summarized in Figs. 10–14. In
these figures ‘‘East⫹58’’ refers to an incident east wind with
the 1958 bridge present, ‘‘East⫹27⫹58’’ refers to an incident
east wind with both the 1927 and the 1958 bridges present,
and ‘‘East’’ is an east wind without either bridge disturbing
FIG. 5. Measured and Fitted Spectra for Coastal Fetch ⫹ 1958 Bridge the flow.
372 / JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001

J. Bridge Eng., 2001, 6(6): 365-375

Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

FIG. 8. Flutter Derivatives for CSB19 (West Wind) and CSB 26 (East Wind) (Courtesy WWL)

The predicted responses conservatively estimate the mea- • Prediction of lateral response differs from that measured
sured responses and may therefore be considered very good, by a few 10ths of a millimeter in the case of the coastal
given the uncertainties outlined below. Specifically fetch.
• Prediction of torsional response differs from that mea-
• Prediction of vertical response differs from that measured sured by 5 ⫺ 8 ⫻ 10⫺3 rad and the edge displacement
by a few 10ths of a millimeter in the case of the coastal associated with this rotation by a few 10ths of a milli-
fetch. meter in the case of the coastal fetch. As expected, this is
JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001 / 373

J. Bridge Eng., 2001, 6(6): 365-375

Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

FIG. 9. Static Coefficient Data Measured in Smooth Flow (Courtesy WWL)

FIG. 10. Comparison of Vertical Responses FIG. 13. Comparison of Torsional Responses

FIG. 11. Comparison of Vertical Responses—Modal Breakdown

FIG. 14. Comparison of Vertical Edge Responses (Torsional Rotation)

• The full-bridge model is a dynamic and aerodynamic ap-

proximation to the prototype. Although corrections were
applied as deemed appropriate, there are still inevitable
similitude differences between the model and the proto-
type it represents.
• Unfortunately, no information was available in the present
circumstances on the vertical turbulence spectra devel-
oped in the wind tunnel. The form assumed for calcula-
tion—although consistent with standard practice for at-
mospheric turbulence—must be considered approximate
FIG. 12. Comparison of Lateral Responses
in this application and could easily differ from the actual
vertical turbulence in the tunnel.
the weakest of the comparisons to the model because the • The amplitudes measured are found to be very small. It
actual full-bridge model damping was reported as consid- is difficult to predict the response of a model with the
erably higher (5%) than that used in the simulation precision tacitly implied, given the uncertainties present.
(0.8%); the trend evidenced is consistent with that obser- • In general, the inevitably small size of the model made it
vation (i.e., the analysis overpredicts the response). inherently less accurate in its ability to realize all pertinent
• In the cases of turbulence initiated by bridges upstream, features.
the analytical predictions tend to diverge a little more • Damping was measured from free-vibration tests. The
from those measured in the model. The calculation almost modal damping present in the model at the very low am-
always overpredicts the measured response. plitudes experienced during the buffeting testing may in-
deed be higher, disproportionately reducing the response.
In evaluating these differences, the following points are worth Although this effect should be evaluated in the future,
remembering: data for such an evaluation are not available to the writers.
374 / JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001

J. Bridge Eng., 2001, 6(6): 365-375

 CONCLUDING REMARKS Singh, L., Jones, N. P., Scanlan, R. H., and Lorendeaux, O. (1996). ‘‘Iden-
tification of lateral flutter derivatives of long-span bridges.’’ J. Wind
The comparison described above demonstrates acceptable Engrg. and Industrial Aerodynamics, 60, 81–89.
performance of the analytical model in its ability to predict
the model response. The analytical model (with suitable choice NOTATION
of parameters) may therefore be considered suitable for the
prediction of the estimated prototype response. Overall, the The following symbols are used in this paper:
complete methodology employed in this paper represents the
general versatility of the analytical and experimental tech- A*
k = flutter derivatives for aeroelastic moment;
niques for the aeroelastic design of long-span bridges. B = deck width;
CD = static drag coefficient;
ACKNOWLEDGMENTS Cij = general term of aeroelastically modified damping ma-
Downloaded from ascelibrary.org by "Indian Institute of Technology, Kanpur" on 01/02/22. Copyright ASCE. For personal use only; all rights reserved.

trix C;
The writers would like to thank Dr. Jon Raggett of WWL for perform- D = aerodynamic drag force/deck height;
ing the aerodynamic and aeroelastic tests reported in this paper: both the D = normalized aerodynamic drag force;
section model tests to determine static coefficients and flutter derivatives D̄ = transformed normalized aerodynamic drag force;
and the full-bridge model tests that were used in the model comparison.
The writers would also like to thank Mark Ketchum of OPAC Consulting
Dij = general term of aeroelastically modified stiffness matrix
Engineers, San Francisco, Calif., who was responsible for the finite-ele- D;
ment analysis of the structures analyzed herein that provided the structural E = impedance matrix;
dynamic data. The support of both writers by the National Science Foun- f = frequency (Hz);
*
dation, Washington, D.C., is also gratefully acknowledged. G Trimsj = modal integral, T* m = H* m , P* m ; ri = hi, pi, or ␣i;
m , or A*
and sj = hj, pj, or ␣j;
REFERENCES H* k = flutter derivatives for aeroelastic lift;
h(x, t) = total vertical deflection;
Jain, A., Jones, N. P., and Scanlan, R. H. (1996a). ‘‘Coupled flutter and hi (x) = vertical component of ith modal deflection;
buffeting analysis of long-span bridges.’’ J. Struct. Engrg., ASCE,
Ii = generalized inertia in ith mode;
122(7), 716–725.
Jain, A., Jones, N. P., and Scanlan, R. H. (1996b). ‘‘Coupled aeroelastic K = reduced frequency, B␻/U;
and aerodynamic analysis of long-span bridges.’’ J. Wind Engrg. and L = aerodynamic lift force;
Industrial Aerodynamics, 60, 69–80. L = normalized aerodynamic lift force;
Jones, N. P., Jain, A., and Scanlan, R. H. (1992). ‘‘Comparison of wind L̄ = transformed normalized aerodynamic lift force;
cross-spectral data with models.’’ Proc., Probabilistic Methods Spec. l = deck length;
Conf., ASCE, New York. M = aerodynamic pitching moment;
Kaimal, J. C., Wyngaard, J. C., Izumi, Y., and Cote, O. R. (1972). ‘‘Spec- M = normalized aerodynamic pitching moment;
tral characteristics of surface-layer turbulence.’’ Quarterly J. Royal Me- M̄ = transformed normalized aerodynamic pitching moment;
teorological Soc., Bracknell, U.K., 98, 563–589.
P* k = flutter derivatives for aeroelastic drag;
Katsuchi, H., Jones, N. P., Scanlan, R. H., and Akiyama, H. (1998a). ‘‘A
study of mode coupling in flutter and buffeting of the Akashi-Kaikyo p(x, t) = total lateral deflection;
Bridge.’’ J. Struct. Mech. and Earthquake Engrg., Tokyo, 15(2), 175s– pi (x) = lateral component of ith modal deflection;
190s. qi (t) = generalized force in ith mode;
Katsuchi, H., Jones, N. P., Scanlan, R. H., and Akiyama, H. (1998b). s = dimensionless time, Ut/B;
‘‘Multi-mode flutter and buffeting analysis of the Akashi-Kaikyo t = time;
Bridge.’’ J. Wind Engrg. and Industrial Aerodynamics, 77 and 78, 431– U = mean wind velocity;
441. x = along-span coordinate;
Katsuchi, H., Jones, N. P., Scanlan, R. H., and Akiyama, H. (1999). ‘‘Mul- y = across-span coordinate;
timode coupled buffeting and flutter analysis of the Akashi-Kaikyo
z = vertical coordinate/elevation above sea level;
Bridge.’’ J. Mech. Engrg., ASCE, 125(1), 60–70.
Roberts, J. B., and Surry, D. (1973). ‘‘Coherence of grid-generated tur- ␣(x, t) = total torsional deflection;
bulence.’’ J. Hydr. Div., ASCE, 99(6), 1227–1245. ␣i (x) = torsional component of ith modal deflection;
Sarkar, P. P., Jones, N. P., and Scanlan, R. H. (1994). ‘‘Identification of ␦ij = Kronecker delta function;
aeroelastic parameters of flexible bridges.’’ J. Engrg. Mech., ASCE, ␨i = damping ratio, ith mode;
120(8), 1718–1742. ␰ i (t) = generalized coordinate of mode i;
Scanlan, R. H. (1981). ‘‘State-of-the-art methods for calculating flutter, ␳ = air density; and
vortex-induced, and buffeting response of bridge structures.’’ Rep. No. ␻i = ith mode circular natural frequency.
FHWA/RD-80/50, Federal Highway Administration, Washington, D.C.
Scanlan, R. H., and Jones, N. P. (1990). ‘‘Aeroelastic analysis of cable-
stayed bridges.’’ J. Struct. Engrg., ASCE, 116(2), 279–297. Subscripts
Scanlan, R. H., and Jones, N. P. (1999). ‘‘On aerodynamic admittance in
bridge aeroelastic analysis.’’ J. Fluids and Struct., 13, 1017–1027. ae = aeroelastic;
Simiu, E., and Scanlan, R. H. (1996). Wind effects on structures, 3rd Ed., b = buffeting; and
Wiley, New York. i = mode index.

JOURNAL OF BRIDGE ENGINEERING / NOVEMBER/DECEMBER 2001 / 375

J. Bridge Eng., 2001, 6(6): 365-375