You are on page 1of 19

AEROELASTIC ANALYSIS O F CABLE-STAYED

BRIDGES
By Robert H. Scanlan, 1 Member, ASCE, and Nicholas P. Jones, 2
Associate Member, ASCE
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

ABSTRACT: The aeroelastic response of cable-stayed bridges to wind is a complex


interaction of aerodynamic loading and coupled structural motion. The present pa-
per presents an empirically based analysis method for such structures, based on a
flutter-derivative formulation. The principal perceived advantage of the approach
adopted herein is that the full three-dimensional complexities of the system may
be incorporated, while retaining insight into the physical mechanisms involved.
Two examples, covering the erection and completed stages of a twin-deck struc-
ture, are presented and discussed. Correlation of the method with three-dimen-
sional wind-tunnel tests is presented. Finally, attention is given to the effects of
turbulence in the incoming wind; the flutter-derivative approach is successfully
used to explain phenomena observed in full-scale tests and in practice.

INTRODUCTION

Developments in analysis for the aeroelastic response of long, flexible


bridges to wind owe much—as regards method—to studies of the flutter
and buffeting of aircraft (Scanlan 1966; Bisplinghoff and Ashley 1962; Dow-
ell et al. 1978; Liepmann 1952). Important early contributions to the bridge
wind-buffeting problem are indicated by Davenport (1961, 1962a, 1962b,
1966). As the load implications and mechanisms associated with experi-
mentally obtained flutter derivatives for bridge decks (Scanlan and Tomko
1971) have become more fully recognized, studies employing them (Scanlan
and Budlong 1974; Scanlan et al. 1974; Scanlan and Gade 1977; Scanlan
1978a, 1978b; Lin 1979; Lin and Arairatnam 1980; Scanlan 1981; Lin and
Yang 1983; Scanlan 1984; Huston 1986) have appeared. More recently, since
flutter derivatives are functions defined in the frequency domain (Scanlan
1988), they have been employed both in the definition for critical flutter
velocity and of the variance of structural buffeting excursions in studies based
in that domain.
The present paper examines the particular aeroelastic problems of the ca-
ble-stayed bridge. This newer type of structure is characterized by natural
vibration models that are generally more complex in form than those of the
classical suspension bridge. A number of authors (Irwin 1977, 1987; Zan
1986; Scanlan 1987; Zan and Wardlaw 1987) have addressed the problems
of the flutter and buffeting of bridges of the cable-stayed type. What distin-
guishes the approach taken with present paper is the organization of the anal-
ysis methods around experimental flutter derivatives, the recognition of the
full three-dimensional complexities of the responding bridge models, and the
focus upon frequency domain considerations throughout.
'Prof., Dept. of Civ. Engrg., The Johns Hopkins Univ., Baltimore, MD 21218-
2699.
2
Asst. Prof., Dept. of Civ. Engrg., The Johns Hopkins Univ., Baltimore, MD.
Note. Discussion open until July 1, 1990. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on August 2, 1988.
This paper is part of the Journal of Structural Engineering, Vol. 116, No. 2, Feb-
ruary, 1990. ©ASCE, ISSN 0733-9445/90/0002-0279/$1.00 + $.15 per page. Pa-
per No. 24321.

279

J. Struct. Eng. 1990.116:279-297.


Examples of some interest are then drawn from the analysis of the flutter
and buffeting of a twin-deck cable-stayed bridge in both completed and erec-
tion stages. As is now generally recognized (Irwin 1987; Zan and Wardlaw
1987), cable-stayed bridges are considerably more vulnerable during con-
struction than after completion. An item of particular interest is the manner
in which the flutter derivatives delineate the different actions of the wind-
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

ward and leeward decks of the twin-deck configuration. Finally, a note is


included demonstrating how spanwise diminution in correlation of the flutter
derivatives explains the experimentally observed increase of critical flutter
velocity when changing from smooth to turbulent flow.

BASIC ANALYSIS

The study starts conveniently by performing a standard vibration modal


analysis of the structure under consideration; it is useful to obtain the first
20 or more 3-D modes T|,(;c,y,z), with their frequencies and generalized iner-
tias

2
h ~ I f\i{x,y,z)dm{x,y,z) (1)
structure

Since loading and response concerns will focus upon the deck, where x
is the coordinate along the span, let h(x,t), a{x,i), p{x,i) represent, respec-
tively, the vertical deflection, twist, and lateral sway of the deck section at
x. In terms of the dimensionless manifestations ht(x), a,(x), pt{x) of mode
form along the deck, deflection components are
vertical
h(x,t) = X hAxMtt) i (2)

twist
a(x,t) = 2 «i(*)&(0 (3)

sway
p(x,t) = ^ P,(x)BUt) (4)
i

where £,(?) = the generalized coordinate of mode ;'; and B = deck width.
Then, if the mode has circular natural frequency w, and damping ratio-to-
critical £,•, the motion of £, is governed by
It<& + 2£,a),i + o>?&) = Q,{i) (5)
where Qt = the generalized force to be defined.
Let the lift, drag, and moment per unit span be defined by
lift
L = Lae + Lb (6)
drag
280

J. Struct. Eng. 1990.116:279-297.


D = Dae + Db (7)

moment
M = Mae + Mh. (8)
where subscripts ae and b refer to aeroelastic and buffeting, respectively.
Note that, in general, the buffeting effects include self-induced buffeting.
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

For purely sinusoidal motions the aeroelastic forces have taken the well-
known form (Scanlan 1978a, 1978b)

L„ = - pU2B KHf(K) - + KHt{K) — + K2Hf(K)a (9)

l ,
ae = - pU2B KPf(K) - + KPt(K) — + K2PUK)OL (10)
2

ae = - pU2B2 KAf(K) - + KA$(K) — + KzAUK)a (11)


2
under a crosswind of air density p and mean velocity U. (Note that the wind
is in general turbulent). Under assumed slowly varying gust action, the buf-
feting forces (exclusive of steady forces) due to the wind turbulence are
given by

u\ w 1
Lb = - pU2B pU2B\,b(x,t) (12)
CL\ 2-J+ (C[ + CD) -
1 , i
Db~- = - pU2B c 2 2
2 'i d) = - -PpU
U BBb(x,t) (13)

w
Mb = -1 pU2,B2,
2 HI) = - pU2B2Mb(x,t)

where CL, CD, and CM are, respectively, the lift, drag, and moment coef-
(14)

ficients (referred to deck width B) of a typical deck section; C'L = dCjdu,


C'M = dCM/da, and u = u(t), w = w(t) are wind horizontal and vertical gust
components, respectively. (These gust forces neglect self-induced buffeting
effects, which, though important in some instances, will not be introduced
here.)
With the definitions given

Qt(t) = (LhfB + DpiB + Ma,)dx (15)

where I = deck span length.

SINGLE-MODE APPROXIMATION

It has proven valuable in practical cases to investigate single-mode be-


havior, mode by mode, neglecting inter-modal and inter-component cou-
pling. Although this does not respond to the full mathematical requirements
of the problem, this will be done in what follows. It does prove useful,

281

J. Struct. Eng. 1990.116:279-297.


however, in identifing those modes susceptible of "driving" an unstable dy-
namic state. Thus, of the sinusoidal aeroelastic terms, only the aerodynamic
damping terms in H*, P*, A* will be. retained, plus the stiffness term in
A*. A discussion on this point is included later.
It is further convenient to introduce the dimensionless time s ~ Ut/B whence
Eq. 5 becomes
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

/,<£•' + l^KM + Kfc) = ^ Q,{s) (16)

with K, = B(a,/U and (;,' = dijds. Then


2
B 1 f'
—a QM = - pB4 J {K[Hf(K)ti(x) + Pf(K)p2(x) + A2*(£)a?(x)]£

+ K2AUK)<x2(x)£,i + Uh,(x) + Bbp,(x) + Mba.t(x)}dx (17)


Defining the Fourier transform of f(s) to be

f(K) = | f{s)e-'K°ds (18)


o
where K — Bw/U, we obtain an expression in the transformed domain that
is valid for general motions (;,
B2 1 f'
~2 Qi(K) = - pB4 J (K2{i[Hf(K)ti(x) + Pf(K)p2(x) + Aj(tf)a?(*)]

+ AHK)a2(x)}UK) + UiKMx) + Db(K)Pi(x) + M4(ZH(*))<& (19)


Neglecting at this point any possible variation of the flutter derivatives with
x, defining
2 dx
\ o q (x) -l = Gqi (20)
Jo
where qt = ht, pt, or a,, and taking the Fourier transform of Eq. 16 results
in
, , f>B4l f ,
(Kf -K2 + 2i£,tf,tf)i, = r— \ K2U(HfGht + PfGpl + A%Gai) + A$Ga,]&

+ | [Uhi(x)
[LMx) + DbPl(x) + M6a,(x)] y [ (21)
Jo

FLUTTER

Eq. 21 defines two conditions. When the imaginary coefficients of £,• bal-
ance to zero, a condition of no damping occurs. At this condition the ho-
mogeneous system has a natural frequency influenced by the A* term, which
is given by the relation

=1+ A G (22)
i)=U> ^ * «<
282

J. Struct. Eng. 1990.116:279-297.


Inserting this value into the zero damping condition yields
1/2
ALL I ~D4' N

H?Ghi + PtGPi + A$Gai > - ^ 1 AtGai (23)


pS I \ 21,
which may be interpreted as the negative-damping threshold or flutter con-
dition for mode i (i.e., the condition for the negative aerodynamic damping
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

effects to exceed the positive mechanical damping of the system).


Since most bridge flutter is found to be driven by negative aerodynamic
damping, usually in torsion, the aforementioned criterion, applied mode by
mode to the bridge structure, identifies the flutter-prone, or flutter-contrib-
utory, modes of the system. In practice, this method has demonstrated that
a very restricted number of modes are flutter-susceptible within the range of
realizable crosswind velocities. Further, it has been demonstrated repeatedly
that, for large civil engineering structures at relatively low wind velocities—
aerodynamically speaking—structural modes are not greatly altered by wind
forces. This supports the view that insights can be gained by the analysis of
the response of individual modes, an approach which appears to be conser-
vative. The problem is under further research by the writers.

EFFECT OF TURBULENCE ON FLUTTER

For a number of years it has been observed in wind tunnel model studies
that flutter under laminar flow typically occurs at a lower velocity than under
turbulent flow, and usually is entered into much more precipitously as wind
speed rises, as suggested in Fig. 1 (Scanlan and Wardlaw 1978). Various
conjectures and qualitative explanations have been advanced for this phe-
nomenon (Scanlan 1979). One view that has been documented (Huston 1987)
is that turbulence affects the sectional (two-dimensional) flutter derivatives,
although it is not certain whether these effects are always in the direction
of enhanced stability. Another view (Lin 1979) is that, since turbulence im-
plies stochastic coefficients of the equations of motion in the time domain,
these effects can alter the nature of stability, coupling deflection compo-

SECTIONAL MODE! J l

SMOOTH F L O W , FULL MODEL -

TURBULENT FLOW, FULL MODEL

60 80 100 120 140


FULL SCALE WINO SPEED, MPH

FIG. 1. Response in Smooth and Turbulent Flow

283

J. Struct. Eng. 1990.116:279-297.


nents, and postponing critical velocity to higher values.
Working in the frequency domain, as in the present study, a somewhat
simpler hypothesis can be proposed. This is to view the sectional flutter
derivatives as having spectral character dictated by the frequencies of wind
gusting and not uniquely by those of their K values associated with structural
frequency. Thus, for example, a value ofHf(Ki) measured in a laminar flow
experiment will have a unique value dictated by a structural frequency Wj,
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

while H* must be represented by a spectrum if tested at the same reduced


frequency Ki under turbulent flow conditions.
Referring to Eq. 17 it is seen that a term like

IH = I KHt(K)hf(x)dx = lKH?Ghl (24)


Jo
in which it is assumed that Hf is independent of x, must, under the afore-
mentioned interpretation, be replaced by a form acknowledging its stochastic
properties, including loss of spanwise correlation. Thus, a suggested em-
pirical alternative to IH could be
1/2
2 2 OXA dXg
I„ = (sgn Hf)lK\ SH(xA,xB,K)h, (xA)h, (xB) — — (25)

where
SH(xA,xB,K) = |Hf(£)|V**-*l/' (26)
d being an appropriate dimensionless constant. It may be noted that this form
of exponential reduction in coherence parallels that which is found experi-
mentally for wind buffeting components, as employed subsequently in this
paper. Writing IH as
tH = IKH f Ghl (27)
and replacing Gq.(qi = ht, p,, or a,) in the flutter criterion (Eq. 23), with
Gq. where
Gm^Gqt (28)
shows that loss of spanwise correlation in the flutter derivatives can, for
given wind speed, reduce the possibility of instability. The result, under
these conditions, is bridge response of the type suggested in Fig. 1.
For conservatism in the examples presented at a later point, the effect of
turbulence upon flutter stability will be neglected.

EFFECT OF WIND GUSTING BEYOND FLUTTER SPEED

In the examples of this study, the calculated flutter velocity for the com-
pleted bridge based on mode 17 (mainly torsion at 0.67 Hz) was found to
be between 130 mph and 131 mph (58.1 m/s and 58.6 m/s), interpolated
for zero effective damping to be at 130.7 mph (58.4 m/s). In this case, the
margin of bridge safety for wind velocities beyond flutter speed is small.
As an example, the following conditions are hypothetically posed: A mean
hourly wind of 125 mph (55.9 m/s) over the bridge rises abruptly to a mean
of 131 mph (58.6 m/s). The consequences of this are investigated.
284

J. Struct. Eng. 1990.116:279-297.


TABLE 1. Relationship Between CO) and T
T (sec) C(f)
(D (2)
1 3.00
10 2.32
20 2.00
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

30 1.73
50 1.35
100 1.02
200 0.70
300 0.54
600 0.36
1,000 0.16
3,600 0.00

The limited duration time T over which a given average velocity, UT,
occurs is related to the mean hourly velocity, Umh, by the relation (Simiu
and Scanlan 1986)
VT = Umh[\ + C(t)I,] (29)
where /, = gust intensity; and C(t) = a statistical function related to aver-
aging time, T, by Table 1.
In the present study, turbulence intensity at deck height z is approximated
by
l
/, = = 0.191 (30)
logz
zo
where z0 = 0.3 m, and z = 185 ft (56.4 m). Hence, by Eq. 29
131 = 125[1 + C(f)(0.191)] (31)
or
C(t) = 0.251 (32)
which yields, from Table 1
T = 818 sec = 13.6 min (33)
At 131 mph (58.6 m/s), the effective damping value of the system is
found by Eq. 23 to be
7,- = -0.00042 (34)
based on 1% structural damping.
At 125 mph (55.9 m/s), the rms single buffeting amplitude (twist of the
deck) in mode 17 is 0.233°. This value will augment exponentially, above
flutter speed, by the factor
R = e~y""T (35)
where w, = 2TTV,(V, = 0.67 Hz) and T = 818 sec. Therefore
285

J. Struct. Eng. 1990.116:279-297.


R = exp [0.00042(2ir)(0.67)(818)] = 4.25 (36)
The peak amplitude achieved, therefore, by a wind rising to just above
flutter velocity and holding for 13.6 min would be
amax = 0.233NR" = 3.67° (37)
where R = 4.25 and N = 3.71 (see Eq. 58 with T = 818 sec, v = 0.67
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

Hz).
While the aforementioned calculation suggests that there is some margin
of safety for winds above flutter speed, it also should be noted that, at least
in the present case, net negative damping grows very rapidly with wind speed
beyond flutter.

BUFFETING

Eq. 21 may be written in the form


. pfl4/ f' . dx
[C(K) + iD(K)]i = ~ ¥(x,K) — (38)
2'i Jo '
where

C(K) = Kj - K2i\ + — AlGa) (39)

oB4l
D(K) = 2l,K,K - — K2{H?Ght + PfGPl + AfGa) (40)

where C(K) and D{K) represent aerodynamically modified frequency and


damping, respectively, and
¥{x,K) = Lb(x,K)hi(x) + t)b(x,K)Pi(x) + Mb(x,K)ai(x) (41)
From Eq. 24 there may be developed an expression for the power spectral
density S^(K) for the generalized coordinate £,; i.e.
/pB 4 A 2 , , , f' (' dx dx
Sb(K) = ( ^ I [C\K) + D\K)Yl I I S¥(xA,xB,K) y a y B (42)

where Sv(xA,xB,K) = the cross power spectral density (p.s.d.) of F(x,K)


between spanwise locations xA and xB.
Neglecting cross p.s.d.'s between u and w wind components (these would
vanish under homogeneous isotropic turbulence), SF(xA,xB,K) may be ex-
pressed as
SF(xA,xB,K) = q(xA)q(xB)S„(xA,xB,K) + f(xA)f(xB)Sw(xA,xB,K) (43)
where
q(x) = 2[CMx) + CDPi(x) + CMa,W] (44)
f(x) = (C'L + CD)ht{x) + C'Ma,{x) (45)
The cross p.s.d.'s, between spanwise points xA, xB of respective wind
components u and w may be expressed in conventional real, experimentally
286

J. Struct. Eng. 1990.116:279-297.


based form (Simiu and Scanlan 1986) as
Su{xA,xB,K) = Su{K)e~c^-xA" (46)
Sw{xA,xB,K) = S^Ve-**-"*" (47)
where
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

8«Z \6nl
(48)
U U
Use of Eq. 41 in Eq. 42 gives rise to double integrations over modal
forms. When these are so slowly varying with x as to be effectively constant,
they may be removed from the integral. In other circumstances they are
linearly or sinusoidally varying. In cases like these, the following integrals
("coherence reduction factors") are useful:
"1 /•!

*c,=
-c\i-n\
d&h\ (c - 1 + e~c) (49)
o Jo
1 /-I

Rc2 = tne-^-^dfyti]

2(1 + r 3 )
- 2 - (1 + r V c3 - 2e" c ( l + r) [l + c(l + r) + c2r] \ (50)

2TT2(1 + e"c)
Rc, = | | sin ir£ sin irne * ^dfytt) = — j + (51)
(C2 + TT2)2

The net p.s.d.'s of buffeted modes are given as a function of those modes
and S^K) by
Sh(x,K) = hHx)B%{K) (52)
Sp(x,K) = pj(x)B%(K) (53)
SJx,K) = otf(x)Sit(K) (54)
Finally, variances of h, p, and a are given by

<(*) = SJx,K)dK . (55)

where q = h, p, or a.
In the calculation of variances it is convenient to use the well-known ap-
proximation

S(n)\H(n)\2dn = | S(n)dn + — S(wi) (56)


o 41,i
o
where

\H(nf = n (57)
1 -
«i

287

J. Struct. Eng. 1990.116:279-297.


To calculate maximum excursions Mr the formula (Simiu and Scanlan 1986)
. 0.577
N = V2 log vT + , • (58)
V2 log vT
can be employed, where v = n} is the frequency (Hz) of the mode in question
and T = 1 hr = 3,600 sec.
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

WIND SPECTRA

For design it is found useful to employ the spectra (Simiu and Scanlan
1986)

200 - u%
S
M = ~. ^—575 • • • (59)

for which

Su(n)dn = 6u% (60)


Jo
and

3.36 - u\
«") = 7—^ • • (61)
1+
Kf)
for which

Sw{n)dn = \.lu\ (62)


Jo
where n = frequency (Hz); u* = friction velocity; and z = height above
ground, as employed in the boundary layer wind profile description (Simiu
and Scanlan 1986)

U = 2.5M* log - (63)


zo
with z0 = the fetch roughness length.
An alternative spectrum, useful for wind tunnel studies, is the Karman
spectrum given by

Su{n) = P ^ u%\ 1 + 7 0 . 8 ^ j (64)

where (3 = u2/u%; and Lxu = longitudinal integral scale length.


288

J. Struct. Eng. 1990.116:279-297.


USEFUL CALCULATION FORMULAS

The flutter criterion (Eq. 23) is applicable to all modes; from it a net
effective damping ratio 7,- may be stated for any given 3-D mode i

pB4l
y, = ii (HfG„, + P? GPi + AfGa) (65)
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

~4L

It is particularly to be noted that this expression takes into account aero-


dynamic damping (positive or negative) in all three coordinates, h, p, and
a. Buffeting response in mode i may then be calculated from the formula
that holds up to critical flutter velocity

pB4/ N 2

< ( * ) = q\{xW E{nt,V)Rc (66)


2IiKfU,

where q,(z) = any natural modal form (Eq. 20). E{nt,U) depends on the
particular mode in question, as reflected by Eq. 42. For example, in a "see-
saw" mode, where the dominant motion is approximately linear in h,(x),
e.g., hj(x) = hiox/l

im,Su(n,)
E(nt,U) = (2CLhi0f + 6ul
47;
r
nniSw{nt)
+ RIRKCL + cD)hl0Y + 1.7K2* (67)
4ys

where RAR = aspect ratio correction (see Fig. 2) and7? c = spanwise correlation
correction (see Eqs. 49-51). In the case of a "weather-vane" mode, where
the dominant motion is approximately linear 'mpt(x), e.g., pfa) = piax/l

TttljSJjli)
E(ni,U) = (2CDpmy + 6u% (68)
. 4-y,-

1.0- 1 1 1 1 1 1 1 1 1

0.B-
\

Jo.B-
u

a ^--^^^

0.2-

0.0- 1 1 1 1 1 1 1 1 1
0 1 2 3 4 5 8 7 8 9 10
(C^ljo/A.R.

FIG. 2. Correction for Finite Aspect Ratio

289

J. Struct. Eng. 1990.116:279-297.


EXAMPLES

Completed Bridge
A twin-deck cable-stayed bridge with side-by-side roadways' is taken as
an example. The main span length is 1,250 ft (381 m), with symmetrical
side spans of 485 ft (147.828 m). Each deck has a width of B = 78.167 ft
(23.825 m); the parallel disposition of the decks is illustrated in Fig. 3. Note
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

that the twin towers are integrally joined structurally at the "knee" at deck
level. To illustrate the relative participation of the h, p, and a components
of the bridge modes, the values of lGqi are listed for the first 17 modes in
Table 2. (The scale to which these are presented depends on arbitrary nor-
malization of the computational units, in which /, = 1.) This table allows a
judgment as to the predominant deflectional action (h, p, or a) of any given
mode. It is noted, for example, that the lowest mode, mode 1, has pro-
nounced lateral sway (IGP. s= 0.35 X 10"6), and that mode 4 is strong in
vertical bending (lGh. = 0.33 X 10~s), while the first mode that is strong in
torsion is mode 17 (lGa. = 0.32 X 10"5). (The display of Table 2 also is
useful in selecting principal modes with strong characteristics of interest to
wind tunnel section model studies. This aspect will not be pursued here.) It
will be observed that the first mode that is strong in torsion (and thus sus-
ceptible to flutter) is mode 17, which occurs at the relatively high frequency
of 0.67 Hz.
In flutter, wind is most critical when normal to the span. Experimental
flutter derivatives Hf,A*, and A* for windward and leeward decks are given
in Fig. 4. The striking difference between windward and leeward results is
evident. A mechanical damping value of £, = 0.01 was assumed for all
modes. Application of the flutter criterion (Eq. 23) led to a calculated flutter
velocity of 135 mph (60.4 m/s), which was later confirmed as conservative
by wind tunnel tests. The integrally connected twin-deck configuration in

• > . . '

FIG. 3. Parallel Twin-Deck Construction of Baytown Bridge

290

J. Struct. Eng. 1990.116:279-297.


TABLE 2. Deck Generalized Modal Factors—Completed Bridge
Mode number Frequency (Hz) Vertical lGh. Lateral IGP. Torsional /G„
(D (2) (3) (4) (5)
1 0.178 0.9010834E-13 0.3468579E-06 0.17272641E-07
2 0.179 0.1443272E-08 0.1416995E-12 0.11422142E-11
3 0.190 0.8095872E-10 0.3457035E-06 0.12867202E-07
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

4 0.273 0.3324330E-06 0.7890348E-13 0.31182583E-11


5 0.279 0.3348605E-06 0.1158093E-09 0.76769890E-09
6 0.350 0.3275767E-06 0.8875293E-14 0.41010849E-11
7 0.361 0.3318449E-06 0.5906351E-09 0.12462385E-08
8 0.388 0.4930751E-09 0.3230640E-06 0.52642148E-07
9 0.439 0.1900286E-12 0.3422200E-06 0.94378642E-07
10 0.537 0.3275868E-06 0.1395707E-13 0.17552783E-11
11 0.549 0.3300172E-06 0.2544302E-08 0.88823286E-08
12 0.582 0.6363877E-08 0.2333250E-06 0.18926808E-07
13 0.611 0.1121834E-06 0.2254676E-08 0.81263565E-07
14 0.612 0.2646134E-07 0.1223409E-07 0.46492072E-06
15 0.626 0.3371242E-06 0.1636895E-12 0.34783224E-11
16 0.642 0.2342440E-06 0.8722898E-09 0.49586891E-07
17 0.670 0.8540877E-09 0.1051735E-07 0.31795792E-05

1.0- • \—— 1 • 1 1

O.B-

0.6-
/
0.4-
/
0.2-
/

0.2-

(1.4- 1 H -_ 1 1 1 .
> 2 4 6 E 2 4 6 E
Reduced Velocity, U/NB Reduced Velocity, U/NB

1.0- • 1 1 1 1 1 .

0.8-

0.B- •

0.4-

-> 0.2-
A

'w
0.2-

0.4- — , — i — , — i — , — i — , —
0 2 4 6 E
Reduced V e l o c i t y , U/NB

FIG. 4. Flutter Derivatives—Completed Structure: Windward Deck, Lee-


ward Deck

291

J. Struct. Eng. 1990.116:279-297.


i
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

Stage B, Mode 1
N = 0.189 Hz.

Stage D, Mode 4
N = 0.330 Hz.
I

FIG. 5. Baytown Bridge—Construction Stages

this instance provided an important flutter advantage, as the wake-protected


leeward deck lent positive daming to the otherwise much more vulnerable
windward deck, which, by itself, had a flutter speed of only 105 mph (47
m/s).
Buffeting of the twin full-span structure was studied in mode 4 (strong
vertical deflection) and mode 17 (strong twist), using the theory outlined
and the same basic parameters, at a wind speed of 125 mph (55.9 m/s).
Static force coefficient data were CL = -0.27, CD = 0.18, CM = -0.0076,
C'L = 4.22, and C'M = 0.36. In mode 4 (0.273 Hz) the single-amplitude rms
vertical deflection of the windward deck was found to be uh = 0.7 ft (0.213
m), with an expected maximum (3.87a-,,) of 2.71 ft (0.826 m). A coherence
reduction factor of Rc =0.0815 was used. In mode 17 the single-amplitude
rms deflection of the same deck was o-a = 0.23°, with a peak value (4.09o-a
= 0.95°). A coherence reduction factor RCl = 0.034 served in this mode.

Partially Constructed Bridge


During construction, a number of potentially critical stages were identi-
fied. These were the following.
292

J. Struct. Eng. 1990.116:279-297.


TABLE 3. Summary of Buffeting Responses: Deck Outer-Tip Deflection: Bay-
town Bridge-Construction Stages

Mode Frequency Rocking peak to peak, Lateral peak to peak,


number (Hz) in. (cm) (cm)
(1) (2) (3) (4)
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

(a) Stage B
1 0.189 36.95 (93.85) —
2 0.289 18.42 (46.79) 1.48 (3.76)
3 0.463 5.94 (15.09) 1.66 (4.22)
4 0.515 — 1.91 (4.85)
(b) Stage C
1 0.143 7.72 (19.61) 18.36 (46.63)
2 0.169 — 13.40 (34.04)
3 0.224 49.25 (125.10) —
4 0.307 41.17 (104.57) —
5 0.521 9.32 (23.67) —
6 0.580 2.20 (5.59) 0.36 (0.91)
(c) Stage D
1 0.281 — 4.07 (10.34)
2 0.307 33.51 (85.12) —
3 0.322 — 3.17 (8.05)
4 0.330 40.33 (102.44) —

1. Just prior to the installation of temporary tie-downs (stage B).


2. Just prior to the addition of the side-pier (stage C).
3. Just before completion of the main span (stage D).

The total deck length in each case was 389.7 ft (118.8 m), 877.9 ft (267.6
m), and 1,098.7 ft (334.9 m), respectively; the configurations are sketched
in Fig. 5. A similar approach for determining flutter proclivity and esti-
mating buffeting response was used as in the preceding example. At each
of the aforementioned stages, a dynamic analysis of the structure was per-
formed and the first 20 modes computed. Flutter was not observed to be a
critical factor at any of the stages. Several modes were identified as being
potentially susceptible, as in the preceding example, but it was found that
the significant coupling between the modal degrees of freedom lent sufficient
positive aerodynamic damping to the structure in each case so as to stabilize
the deck. It should be emphasized that the leeward deck again contributed
significantly to this stability, as in the case of the completed bridge. In a
number of cases, the windward deck was indeed prone to instability, but
transfer of energy to, and subsequent dissipation by, the leeward deck af-
forded overall stability.
Buffeting was studied in the lowest, most flexible modes of each of the
three stages. Buffeting response was checked in what can be described as
"weather vane" (where the decks rotate essentially as rigid bodies about a
vertical axis through the center of the towers) and "seesaw" (where the decks
rotate essentially as rigid bodies about a horizontal axis through the towers)
modes. In all cases, it was found that the most critical mode was the "see-
293

J. Struct. Eng. 1990.116:279-297.


saw" or rocking mode, and the most extreme was an estimated peak-to-peak
deflection at the tip of the deck of 49.3 in. (125.2 cm) in mode C-3. The
results are summarized in Table 3. It should be noted that a correction has
been applied to the stage B results to account for the finite aspect ratio of
the relatively short deck and the resulting non-two-dimensionality of the flow
field around the structure. An estimate of this correction may be determined
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

from Fig. 2 as 0.7 for the dimensions associated with stage C.


The computed results were compared to those obtained from a series of
wind tunnel tests on 1:125 scale models. The computed results compared
favorably with the wind tunnel measurements.

CONCLUSIONS

An analysis method for computing the response of cable-stayed bridge


structures to aeroelastic excitation has been presented. The method is for-
mulated around the experimentally determined flutter derivatives and a full
three-dimensional modal analysis of the structure. The technique allows re-
alistic incorporation of a number of three-dimensional effects in the analysis,
including finite aspect ratio, differing aerodynamic properties of windward
and leeward decks in a twin-deck configuration, and the lack of spanwise
coherence in the incoming flow. A generalization of the method can be
used to explain the less-precipitous transition to flutter instability observed
in practice.
Analysis of a twin-deck structure during erection and when completed ver-
ified the susceptibility of cable-stayed bridges to wind during construction.
While the peak buffeting-induced deflection expected in the completed struc-
ture for a design wind speed of 125 mph (55.9 m/s) was 2.71 ft (0.83 m),
the peak response during erection [design wind speed 120 mph (53.6 m/s)]
in stage C was estimated as 4.11 ft (1.25 m). In both cases, the flutter
susceptibility was hindered by the favorable aerodynamic properties of the
leeward deck, which offset the influence of the inherently unstable windward
deck. In the completed structure, the critical flutter velocity was determined
to be 135 mph (60.3 m/s), while the erection stages showed no proclivity
to flutter due to the aforementioned factors.

ACKNOWLEDGMENTS

The writers are particulary grateful to Ben Christopher and Steven L. Stroh
of Greiner, Inc., Tampa, Fla., and to the Texas State Department of High-
ways for information on the Baytown Bridge as well as Figs. 3 and 5, which
were used in the preparation of this paper. Thanks are also due J. D. Raggett
of West Wind Laboratory, Carmel, Calif., who obtained the flutter deriv-
atives, and to Timothy A. Reinhold of Applied Research Associates, Ra-
leigh, N.C., who performed both erection stage and full-bridge wind tunnel
model studies.

APPENDIX I. REFERENCES

Bisplinghoff, R. L., and Ashley, H. (1962). Principles of aeroelasticity. John Wiley


and Sons, New York, N.Y.
Davenport, A. G. (1961). "The application of statistical concepts to the wind loading
294

J. Struct. Eng. 1990.116:279-297.


of structures." Proc. Inst. Civ. Engrg., 19, Aug., 449^472.
Davenport, A. G. (1962a). "Buffeting of a suspension bridge by storm winds." J.
Struct. Div., ASCE, 88(3), 233-268. „
Davenport, A. G. (1962b). "The response of slender, line-like structures to a gusty
wind." Proc. Inst. Civ. Engrg., 23, Nov., 389-407.
Davenport, A. G. (1966). "The action of wind on suspension bridges." Proc, Int.
Symp. on Suspension Bridges, Laboratorio Nacional De Engenharia Civil, Lisbon,
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

Portugal, 79-100.
Dowell, E. H., et al. (1978). A modern course in aeroelasticity. Sijthoff and Noord-
hoff, Alphen aan den Rijn, The Netherlands.
Huston, D. R. (1986). "The effect of upstream gusting on the aeroelastic behavior
of long suspended-span bridges." Thesis presented to Princeton Univ., at Prince-
ton, N.J., in partial fulfillment of the requirements for the degree of Doctor of
Philosophy.
Huston, D. R. (1987). "Flutter derivatives extracted from fourteen generic deck sec-
tions." Bridges and Transmission Line Structures, ASCE, L. Tall, ed., 281-291.
Irwin, P. A. (1977). "Wind tunnel and analytical investigations of the response of
Lions' Gate Bridge to a turbulent wind." Report NAE-LTR-LA-210, National Re-
search Council, Ottawa, Canada.
Irwin, P. A. (1987). "Wind buffeting of cable-stayed bridges during construction."
Bridges and Transmission Line Structures, ASCE, L. Tall, ed., 164-177.
Liepmann, H. W. (1952). "On the application of statistical concepts to the buffeting
problem." J. Aeronaut. Sci., 19(12), 793-800, 822.
Lin, Y. K. (1979). "Motion of suspension bridges in turbulent winds." J. Engrg.
Mech., ASCE, 105(6), 921-923.
Lin, Y. K., and Ariaratnam, S. T. (1980). "Stability of bridge motion in turbulent
winds." / . Struct. Mech., 8(1), 1-15.
Lin, Y. K., and Yang, J. N. (1983). "Multimode bridge response to wind excita-
tion." J. Engrg. Mech., ASCE, 109(2), 586-603.
Scanlan, R. H. (1978a). "The action of flexible bridges under wind: I. Flutter the-
ory." / . Sound and Vibration, 60(2), 187-199.
Scanlan, R. H. (1978b). "The action of flexible bridges under wind: II. Buffeting
theory." / . Sound and Vibration, 60(2), 201-211.
Scanlan, R. H. (1979). "On the state of stability considerations for suspended-span
bridges under wind." Practical Experiences with Flow-Induced Vibrations, E.
Naudascher and D. Rockwell, eds., Springer-Verlag, Berlin, W. Germany, 595-
618.
Scanlan, R. H. (1981). "State-of-the-art methods for calculating flutter, vortex-in-
duced, and buffeting response of bridge structures." Report No. FHWA/RD-80-
050, Federal Highway Administration, Washington, D.C., Apr.
Scanlan, R. H. (1984). "Role of indicial functions in buffeting analysis of bridges."
J. Struct. Engrg., ASCE, 110(7), 1433-1446.
Scanlan, R. H. (1988). "On flutter and buffeting mechanisms in long-span bridges."
Prob. Engrg. Mech., 3(1), 22-27.
Scanlan, R. H. (1987). "Aspects of wind and earthquake dynamics of cable-stayed
bridges." Bridges and Power Line Structs., ASCE, L. Tall, ed., 329-340.
Scanlan, R. H., Beliveau, J. G., and Budlong, K. S. (1974). "Indicial aerodynamic
functions for bridge decks." J. Engrg. Mech., ASCE, 100(4), 657-672.
Scanlan, R. H., and Budlong, K. S. (1974). "Flutter and aerodynamic response con-
siderations for bluff objects in a smooth flow." Flow-Induced Structural Vibra-
tions, E. Naudascher, ed., Springer-Verlag, New York, N.Y., 339-354.
Scanlan, R. H., and Gade, R. H. (1977). "Motion of suspended bridge spans under
gusty wind." J. Struct. Div., ASCE, 103(9), 1867-1883.
Scanlan, R. H., and Rosenbaum, R. (1951). Aircraft vibration and flutter, Mac-
millan, Inc., New York, N.Y.
Scanlan, R. H., and Tomko, J. I. (1971). "Airfoil and bridge deck flutter deriva-
tives." J. Engrg. Mech., ASCE, 97(6), 1717-1737.
Scanlan, R. H., and Wardlaw, R. L. (1978). Cable-stayed bridges. Bridge Div.,

295

J. Struct. Eng. 1990.116:279-297.


r ;
Office of Engineering, Federal Highway Administration, Washington, D.C., 169-
202.
Simiu, E., and Scanlan, R. H. (1986). Wind effects on structures, 2nd Ed., John
Wiley and Sons, New York, N.Y.
Zan, S. J. (1986). "Analytical prediction of the buffeting response of the ALRT
Fraser River crossing during erection and upon completion." Report NAE-LTR-LA-
280, National Research Council, Ottawa, Canada.
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

Zan, S. J., and Wardlaw, R. L. (1987). "Wind buffeting of long span bridges with
reference to erection base behaviour." Bridges and Transmission Line Structures,
ASCE, L. Tall, ed., 432-448.

APPENDIX II. NOTATION

The following symbols are used in this paper:

Af = flutter derivatives for aeroelastic moment;


B = deck width;
cD = static drag coefficient;
cL = static lift coefficient;
cL = derivative of static lift coefficient;
CM = static moment coefficient;
t~" = derivative of static moment coefficient;
C(K) = aerodynamically modified frequency parameter (Eq. 39);
C(t) = statistical relationship between averaging time and wind
velocity (Table 1);
c = exponential decay factor for spatial correlation (Eq. 48);
D = aerodynamic drag force;
D = normalized aerodynamic drag fore;
D = transformed normalized aerodynamic drag force;
D(K) = aerodynamically modified damping parameter (Eq. 40);
E(n,,U) = integral of aerodynamic force spectra;
¥(x,K) = transformed aerodynamic buffeting force (Eq. 41);
G9I = modal integral, defined in Eq. 20; qt = h:, pt, or a,-;
Gqt = modal integral, modified for wind turbulence;
m = flutter derivatives for aeroelastic lift;
h(x, t) = total vertical deflection;
h,(x) = vertical component of ith modal deflection;
IH = modal integral, defined in Eq. 24;
It = generalized inertia in the *'th mode;
I, = gust intensity;
K = reduced frequency, Bu/U\
L = aerodynamic lift force;
L = normalized aerodynamic lift force;
L = transformed normalized aerodynamic lift force;
Tx = longitudinal integral length scale in Karman spectrum;
I = deck length;
M = aerodynamic moment;
M = normalized aerodynamic moment;
M = transformed normalized aerodynamic moment;
N = peak response factor, from Eq. 51;
n = frequency (Hz);
= flutter derivatives for aeroelastic drag;
296

J. Struct. Eng. 1990.116:279-297.


p(x,t) = total lateral deflection;
Pi(x) = lateral component of ith modal deflection;
=
Qi(t) generalized for in the ith mode;
q(x) = term defined by Eq. 44;
Qi(K) = transformed generalized for in the /th mode;
7? = factor b y which amplitude increases for post-flutter v e -
Downloaded from ascelibrary.org by New York University on 05/04/15. Copyright ASCE. For personal use only; all rights reserved.

locity gust;
RAK = aspect ratio correction (Fig. 2);
RCt = coherence reduction factors (Eqs. 49-51);
f(x) = term defined in Eq. 44;
SF(xA,xB,K) = cross spectrum of aerodynamic buffeting forces;
SH = spectrum of H* flutter derivative, for turbulent wind;
Sh(x,K) = power spectrum of ^-response;
Sp(x,K) = power spectrum of /(-pressure;
Su(xA,xB,K) = cross spectrum of w-velocity;
Sw(xA,xB,K) = cross spectrum of w-velocity;
Sa(x,K) = power spectrum of a-response;
5f. = spectrum of generalized coordinate response;
s = dimensionless time, Ut/B;
t = time;
U = mean wind velocity;
UT = mean wind velocity for duration T;
Umh = mean hourly wind velocity;
u = longitudinal velocity fluctuation;
u* = friction velocity;
w = vertical velocity fluctuation;
x = along-span coordinate;
y = across-span coordinate;
z = vertical coordinate, deck height;
z0 = effective roughness height;
a(x,t) = total twisting deflection;
a,(x) = twisting component of rth modal deflection;
(3 = ratio u2/u%;
y, = effective system damping in (th mode (mechanical plus
aerodynamic);
£,• = dampinig ratio, ith mode;
T|, = full-bridge ith mode shape;
p = air density;
v = frequency parameter (Hz) in Eq. 58;
ii(t) = generalized coordinate of mode i;
£,,(K) - transformed generalized coordinate of m o d e i;
CT^ = variance of ^-response; and
to,- = ith mode circular natural frequency.

Subscripts
ae = aeroelastic;
b = buffeting; and
i = m o d e index.

297

J. Struct. Eng. 1990.116:279-297.

You might also like