Science3 PDF
Science3 PDF
Science3 PDF
Abstract
The aerostatic and aerodynamic behaviors of a long-span suspension bridge are of serious
engineering concern. As the span length increases, the nonlinear effects due to wind-structure
interactions are becoming unnegligible in determining the aerostatic and aerodynamic
behaviors of long-span suspension bridges. In this paper, an approach of three-dimensional
nonlinear aerostatic and aerodynamic analysis is presented firstly, in which the nonlinearities
of structural geometric, dynamic characteristics and aeroelastic forces due to the large
deformation under the static wind loading are considered. Numerical analyses are then
performed on the three-dimensional finite element model of a suspension bridge with a main
span length of 1490 m. The effectiveness of the proposed approach is finally demonstrated.
The results show that the nonlinear effects have a significant influence on the aerostatic and
aerodynamic behaviors of long-span suspension bridges.
r 2002 Elsevier Science Ltd. All rights reserved.
1. Introduction
*Corresponding author.
E-mail address: [email protected] (X. Zhang).
0167-6105/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 0 2 ) 0 0 2 5 1 - 9
1066 X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080
stability (mainly refers to flutter) due to the wind action is a major concern for the
design and construction of long-span suspension bridges.
Up to now, comprehensive studies on the aeroelastic problem has been performed;
also, lots of numerical analytical methods have been presented. Boonyapinyo [1]
developed a finite element approach to calculate the critical wind velocity for the
nonlinear lateral–torsional buckling instability of long-span cable-stayed bridges, the
mechanism of aerostatic instability was further described. As for flutter instability,
flutter analysis is mostly carried out in the frequency domain because of its
computational convenience. However, there are some attempts to do it in the time
domain for considering the effect of turbulence and the nonlinearities of both
structure and wind force. For three-dimensional bridge structures, three-dimensional
flutter analysis is commonly adopted at present. The calculation can be performed in
two different ways: one is to apply the unsteady aerodynamic forces directly to a
three-dimensional finite element model of the structure (called as direct method);
another is to consider the structural response separately in various vibration modes
and assemble them (called as the superposition method). Miyata and Yamada [2],
Dung [3] and Ge [4] developed the direct finite element method for flutter analysis.
The advantage of this method is that the participation of the full natural modes of
vibration can be considered, but the calculation is time-consuming. However, the
mode superposition method is commonly used due to its accuracy and efficiency.
Agar [5], Namini [6,7] and Chen [8] et al. developed the finite element method to
analyze the multi-mode flutter in frequency domain using the modal techniques.
Scanlan [9], Jones and Scanlan [10], Jain [11,12] and Tanaka et al. [13] directly
utilized the determinant search method to predict the flutter critical condition.
As observed from the above studies, their aerodynamic stability analyses were
mainly based on the linear theory. It means that the analyses were based on the
undeformed initial structures and the effects due to the nonlinear wind-structure
interactions were neglected. For long-span suspension bridges, the static wind action
will cause the large deformation because of the lower structural stiffness. The large
deformation, on the one hand, will influence structural stiffness and the dynamic
characteristics. On the other hand, the aerodynamic shape will be changed
remarkably by the large deformation, which leads to the significant variation and
nonuniform distribution of the wind force acting on the bridges. These effects will
consequently influence the aerodynamic stability of the bridge. As the span length
increases, the aerodynamic stability will be more problematic because of the
significant decrease of structural stiffness. Jones et al. [10], Dung [3], Katsuchi et al.
[14,15] and Miyata [16] also considered nonlinear effects of wind-structure
interactions in their analysis, but the mechanics and roles of these effects on the
aerodynamic stability of long-span suspension bridges are not clearly clarified.
Therefore, it is essential to investigate the nonlinear effects due to wind-structure
interactions on the aerostatic and aerodynamic behaviors of long-span suspension
bridges.
Based on the linear theory, an approach of aerostatic and aerodynamic analysis
considering the nonlinear wind-structure interactions is presented in this paper. The
deformation of suspension bridge under static wind loading is investigated in the
X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080 1067
beginning, and then the dynamic characteristics of such a deformed bridge structure
is analyzed. Finally, flutter analysis for the wind-deformed bridge structure is also
conducted. The nonlinear effects on structural dynamic characteristics and aerostatic
and aerodynamic behaviors due to wind-structure interactions are discussed in
detail.
The aerostatic effect is usually treated as the action of three aerostatic components
of wind force on the bridge, namely the drag force, lift force and twist moment as
shown in Fig. 1. However, the aerostatic force changes with the deformation of
bridge structure, and can be then described as the function of the effective attack
angle between the wind flow and deformed bridge deck. Fig. 1 shows the effective
attack angle a; which is the sum of initial wind attack angle y0 and the torsional angle
y caused by the static wind loading. Furthermore, the torsional deformation of deck
changes along the bridge longitudinal axis; hence the static wind force acting on the
bridge structure is a function of spatial deformation. The aerostatic force acting on
per unit length of the bridge deck can be expressed as follows:
1
Fz ¼ rU 2 D Cz ðaÞ;
2
1
Fy ¼ rU 2 B Cy ðaÞ;
2
1
Mx ¼ rU 2 B2 CM ðaÞ; ð1Þ
2
where r is the air density; U is the mean velocity at the bridge deck level; D is the
bridge deck height; B is the bridge deck width; Cz ðaÞ; Cy ðaÞ; CM ðaÞ are the aerostatic
coefficients obtained from the section model tests; a is the effective angle of attack
including the initial attack angle y0 and the additional attack angle y caused by the
torsional deformation.
Fy
θ Mx
O
α
θ0
Fz
The aerodynamic force model is generally assumed to be composed of: (1) the self-
excited force due to wind-structure interaction in the smooth oncoming wind flow
and (2) the buffeting force due to turbulence in the oncoming wind flow. The
buffeting force usually does not make the bridge aerodynamic unstable, so it is
neglected in the aerodynamic analysis. The well-known self-excited aerodynamic
force, i.e. the self-excited lift, Lh ; and drag, Dp ; as well as the moment, Ma are
generally represented by the 18 components of experimentally determined flutter
derivatives:
1 h’ B’a h p’ p
Lh ¼ rU 2 ð2BÞ KH1 þ KH2 þ K 2 H3 a þ K 2 H4 þ KH5 þ K 2 H6 ;
2 U U B U B
’
1 p’ B’a p h h
Dp ¼ rU 2 ð2BÞ KP1 þ KP2 þ K 2 P3 a þ K 2 P4 þ KP5 þ K 2 P6 ;
2 U U B U B
’
1 2 2 h B’a 2 2 h p’ 2 p
Ma ¼ rU ð2B Þ KA1 þ KA2 þ K A3 a þ K A4 þ KA5 þ K A6 ;
2 U U B U B
ð2Þ
where Hi ; Ai and Pi ði ¼ 126Þ are the experimentally measured flutter derivatives
for the deck cross section under investigation, which are the functions of the reduced
frequency K ¼ ðBo=UÞ and the effective attack angle; o is the response circular
frequency. It is known that the effect of deformation on the self-excited aerodynamic
force could be fully considered.
Eq. (2) represents the aerodynamic force acting on per unit length of the bridge
deck. To convert these uniformly distributed forces into member end effects, a simple
lumping procedure is adopted whereby one-half of the force is assumed to act at each
member end. Using a space frame element for modeling the overall bridge, the total
equivalent aerodynamic joint load, fF g; is subdivided into stiffness component ½As
and damping component ½Ad as
1 1
2
fF g ¼ rU ½As fqðx; tÞg þ ½Ad fqðx; ’ tÞg ; ð3Þ
2 U
’ tÞg are the displacement and velocity vectors; ½As ; ½Ad are the
where fqðx; tÞg; fqðx;
aerodynamic stiffness and aerodynamic damping matrices.
The aerostatic analysis is to predicate the equilibrium position, which is the initial
state of bridge oscillation under the dynamic wind action. The static equilibrium
position of bridge structure under a certain wind speed must be solved by the
iteration approach because of its double nonlinearity of structure and the static wind
force. The iteration equation can be expressed as
ð½Ke j
1 þ ½Kg j
1 ÞfDdj g ¼ fFj ðaj Þg
fFj
1 ðaj
1 Þg; ð4Þ
X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080 1069
where ½Ke j
1 and ½Kg j
1 are the linear-elastic stiffness matrix and geometrical
matrix, respectively; fFj
1 ðaj
1 Þg and fFj ðaj Þg are the wind load vectors correspond-
ing to the effective attack angle aj
1 in step j
1 and aj in step j under wind speed U;
respectively.
The Euclidean norm of the aerostatic coefficients of lift, drag and pitch moment is
taken as convergence criterion, which can be expressed as
(P )1=2
Na 2
1 ½CK ðaj Þ
CK ðaj
1 Þ
PNa 2
peK ðK represents L; D; MÞ; ð5Þ
1 ½CK ðaj
1 Þ
where eK is the convergence accuracy; Na is the total number of the nodes subjected
to the wind force.
When the aerostatic equilibrium position is found, the aerodynamic analysis could
be carried out on the deformed bridge structure. For a structural system with n
discreet degrees of freedom, the equation of motion due to the self-excited
aerodynamic force can be expressed as
. tÞg þ ½D fqðx;
½M fqðx; ’ tÞg þ ½K fqðx; tÞg
1 1
’ tÞg ;
¼ rU 2 ½As fqðx; tÞg þ ½Ad fqðx; ð6Þ
2 U
where ½M is the structural mass matrix; ½D is the structural damping matrix and ½K
is the structural tangent stiffness, which can be divided into two components: elastic
stiffness ½Ke and geometric stiffness ½Kg ; ½K ¼ ½Ke þ ½Kg :
Eq. (6) can be solved by the modal analysis method. For dynamic motion, the
response for all degrees can be separated into the spatial (natural modes) and time-
dependent (generalized coordinate) components as
fqðx; tÞg ¼ ½f fxðtÞg; ð7Þ
where ½f is the mode matrix, which is obtained from the dynamic characteristics
analysis on the deformed bridge structure under static wind loading; fxðtÞg is the
generalized coordinates vector, which can be assumed as a damped, harmonic form
and represented along the complex plane as
fxðtÞg ¼ fRg expðltÞ; ð8Þ
where fRg is the response amplitude; l ¼ ðd þ iÞo; d p is ffiffiffiffiffiffi
theffi response logarithmic
decrement; o is the circular frequency of response; i ¼
1:
Substituting Eqs. (7) and (8) into Eq. (6), and then pre-multiplying the transpose
of the natural mode matrix ½f ; considering the orthogonality between modes and a
nontrivial solution to exist, yielded the determinant as
2
g U U 1
½M s2 þ ½Dg s þ ½K g
12rU 2 ½Ags þ ½Agd iK
¼ 0; ð9Þ
B B B
For a given wind speed, Eq. (9) can be solved by the PK-F method [5]. The value p;
which makes the determinant equal to zero, represents the actual oscillation response;
then the logarithmic decrement and angular frequency of response can be computed as
ReðsÞ U
d¼ ; o ¼ ImðsÞ; ð10Þ
ImðsÞ B
where Re and Im are the real and imaginary parts of the complex variable,
respectively.
Depending on the sign of the logarithmic decrement, the response can be defined
as do0; stable; d ¼ 0; neutrally stable; d > 0; unstable. The wind speed that produces
the neutrally stable condition is termed as flutter speed, Uf ; with the corresponding
flutter frequency, of :
4. Computational procedure
5. Numerical analyses
The linear and nonlinear aerostatic analyses under the initial wind attack angle of
01 are performed to show the nonlinear effects on the aerostatic behaviors, where the
linear analysis refers to the case that the effect of deformation on the aerostatic force
Table 1
The mechanical properties of the bridge
CZ
CY
1.0 CM
0.9
0.8
Aerostatic force coefficients
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
-6 -4 -2 0 2 4 6
is neglected. For the bridge deck, three aerostatic components are acted on the deck,
but only the drag component is considered for the cables, hangers and towers. Fig. 5
shows the variation of lateral, vertical and torsional displacements of the deck at
mid-span with growth of wind speed. The additional wind attack angles due to the
torsional deformation distributing along the bridge deck under several wind speeds
are shown in Fig. 6.
As observed from Fig. 5, it is found out that the vertical and torsional
displacements obtained from nonlinear analyses are greater than those from linear
analyses. Furthermore, the difference between the two cases increases rapidly with
the growth of wind speed. So the nonlinear effects caused by the deformation have
significant influence on the vertical and torsional displacements. On the contrary, the
lateral displacements under the two cases are just identical. The lateral displacement
X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080 1073
is hardly influenced by the nonlinear effects. As seen in Fig. 6, the additional wind
attack angle caused by the torsional deformation distributes nonuniformly along the
bridge deck. Moreover, the distributing nonuniformity increases significantly as
wind speed increases. It will finally affect the aerodynamic behaviors as mentioned
below. Therefore, the nonlinear effects should be considered in the aerostatic
analysis of long-span suspension bridges.
As mentioned above, the bridge configuration and internal force, particularly the
cable tension force, are changed under the action of aerostatic force. Structural
stiffness will be consequently changed, and leads to the variation of the dynamic
characteristics. To investigate the nonlinear effects due to the wind-structure
interactions on the dynamic characteristics of the bridge, numerical analyses of
structural dynamic characteristics on the deformed structure with increase of wind
speed are performed under wind attack angles of
31, 01 and +31. Fig. 7 shows the
modal frequency of the first, the second symmetric vertical bending and the first
torsional modes varying with wind speed.
- 5˚ 0˚
- 3˚ +3˚
0 0.02
0˚ -3˚
0.00
-1 +5˚
+ 3˚ -0.02
+ 5˚ -0.04 -5˚
-2
-0.06
H2*
-3 -0.08
H1*
-0.10
-4
-0.12
-5 -0.14
-0.16
-6 -0.18
-0.20
-7
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12
Reduced velocity Reduced Velocity(U/fB)
o 0˚
0
o +3˚
+3 0.4
0.1 o -3˚
-3 0.2
0.0 0.0
o +5˚
-0.1 +5 -0.2
-0.2 -5
o -0.4 -5˚
-0.6
-0.3 -0.8
-0.4 -1.0
-0.5 -1.2
H4*
H3*
-0.6 -1.4
-1.6
-0.7 -1.8
-0.8 -2.0
-0.9 -2.2
-1.0 -2.4
-2.6
-1.1
0 2 4 6 8 10 12 0 2 4 6 8 10 12 14 16 18
Reduced Velocity (U/fB) Reduced Velocity (U/fB)
Fig. 4. Flutter derivatives at varying angles of wind incidence versus the reduced velocity.
1074 X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080
7 0.02
0.00
0°
6 -0.02
+3° -0.04
5 -0.06
-3°
-0.08
4 +5° -0.10
-0.12 0°
A2*
A1*
3 -5°
-0.14
-0.16 +3°
2 -0.18 -3°
-0.20
1 -0.22 +5°
-0.24
0 -5°
-0.26
-0.28
-1
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12
Reduced Velocity (U/fB)
Reduced Velocity (U/fB)
3.0
2.5
0° 2.0
1.2 1.5
+3° 1.0
1.0 -3° 0.5
0.0
+5° -0.5
0.8 -1.0
A4*
-5° -1.5
-2.0 0°
A3*
0.6
-2.5
-3.0 +3°
0.5 -3.5 -3°
-4.0
-4.5 +5°
0.4
-5.0
-5.5 -5°
0.2 -6.0
-6.5
0 2 4 6 8 10 12 0 2 4 6 8 10 12 14 16 18
Reduced Velocity (U/fB) Reduced Velocity (U/fB)
Fig. 4 (continued).
The modal frequencies change very little under 01 and +31 wind attack angles, but
they increase under
31 wind attack angle. In the cases of 01 and +31 wind attack
angles, the deformation and the variation of structural internal force are small, so the
modal frequencies change slightly. However, in the case of
31 wind attack angle,
the bridge is deformed significantly, and the cable tension force is greatly increased
due to the downward action of the aerostatic lift force as wind speed increases. The
gravity stiffness is increased, and further the modal frequency is increased
consequently. Although the deformation has a slight influence on the modal
frequencies, the modal shape, particularly the torsional mode, is changed greatly.
The vertical and lateral components in the torsional mode will be enhanced as wind
speed increases. The above effects on flutter critical wind speed will be further
discussed later.
As known from the above analyses, the increase in wind speed emphasizes the
importance of structural deformation caused by the static wind loading on the
dynamic characteristics. A large deformation will also change the attack angle along
the bridge deck greatly, which leads to the variation of the aerodynamic force acting
on the bridge deck. The flutter responses will be influenced consequently. In order to
X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080 1075
linear
-7
nonlinear
lateral displacement(m)
-6
-5
-4
-3
-2
-1
0
0 20 40 60 80
wind speed(m/s)
-1.4
vertical displacement(m)
linear
-1.3 nonlinear
-1.2
-1.1
-1.0
-0.9
-0.8
-0.7
-0.6
10 20 30 40 50 60 70 80 90
wind speed(m/s)
Additional wind attack angle(deg)
linear
-0.4
nonlinear
-0.3
-0.2
-0.1
0.0
0 20 40 60 80
wind speed(m/s)
investigate the above nonlinear effects on the flutter stability, numerical flutter
analyses are carried out under the four cases A, B, C and D by the computational
procedure BSNAA. Case A refers to the linear flutter analysis, in which all the
nonlinearities are not included in the analysis. As for Case B, only the variation of
structural dynamic characteristics with the growth of wind speed is considered,
1076 X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080
0.0
-0.1
-0.2
U=20m/s
-0.3 U=40m/s
U=60m/s
-0.4 U=80m/s
0.24
0.24
1st vertical bending mode 0.23
0.22 nd
2 vertical bending mode 0.22
Frequency(Hz)
0.24
0.23
0.22
0.21 1st vertical bending mode
Frequecy(Hz)
whereas in Case C, only the nonlinear effect of aeroelastic force varying with the
deformation is considered. Case D refers to the nonlinear flutter analysis described in
this paper. It is found out from the above analyses that the first and second
symmetrical vertical bending modes and the first symmetrical torsional mode (modes
X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080 1077
4, 5, 11, respectively) are the dominant modes participating in flutter response. Fig. 8
shows the modal damping and response frequencies of modes 4, 5 and 11 varying
with wind speed. The critical flutter speeds at different angles of wind incidence are
given in Table 2.
It can be seen that the flutter speeds obtained from the nonlinear analysis are
generally close to those obtained from the sectional model tests. It is worthwhile to
Mode 4
0.00 Mode 5
Logarithmic decrement
0.24 Mode 11
-0.02
Mode 4
Frequency (Hz)
-0.04 0.22
Mode 5
-0.06 Mode 11
0.20
-0.08
0.18
-0.10
0.16
-0.12
-0.14 0.14
-0.16 0.12
10 20 30 40 50 60 70 10 20 30 40 50 60 70
Wind speed (m/s) Wind speed (m/s)
(a)
0.00 Mode 4
Mode 5
-0.02
Logarithmic decrement
Mode 11
Mode 4 0.24
-0.04 Mode 5 0.23
0.22
Frequency (Hz)
-0.06 Mode 11
0.22
-0.08 0.20
-0.10 0.19
0.18
-0.12 0.17
-0.14 0.16
0.15
-0.16 0.14
-0.18 0.13
10 20 30 40 50 60 70 0.12
10 20 30 40 50 60 70
Wind speed (m/s) Wind speed (m/s)
(b)
Mode 4
0.00 Mode 5
0.25 Mode 11
Logarithmic decrement
-0.01 0.24
-0.02 Mode 4 0.23
-0.03 0.22
Frequency (Hz)
Mode 5
0.21
-0.04 Mode 11 0.20
-0.05 0.19
-0.06 0.18
0.17
-0.07 0.16
-0.08 0.15
-0.09 0.14
0.13
-0.10 0.12
-0.11 0.11
10 15 20 25 30 35 40 45 50 10 15 20 25 30 35 40 45 50
Wind speed (m/s) Wind speed (m/s)
(c)
Fig. 8. Modal frequencies and damping ratio varying with the wind speed: (a)
31 wind attack angle; (b)
01 wind attack angle and (c) +31 wind attack angle.
1078 X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080
Table 2
The flutter speeds under different cases (m/s)
note that the section model test could not really represent the aerodynamic behavior
of the actual three-dimensional bridge structure. The difference between the two
cases may be caused by this reason to a certain extent. On the other hand, it may be
caused by the fact that structural damping of the model is really different from the
values simply assumed in the numerical analyses. On the whole, the proposed
nonlinear method is effective to predict the critical flutter speed. It is to be noted that
the flutter response of this bridge is multi-mode coupled flutter, which is dominated
by the first vertical bending, second vertical bending and first torsional mode.
By comparing Case B with Case A, it is found that the flutter speeds under 01 and
31 wind attack angles are greatly increased by the effect of the dynamic
characteristics varying with wind speed under the static wind loading. The reason,
on the one hand, could be attributed to the increase of torsional frequency as shown
in Fig. 7. In these cases, the axial forces of main cables increase due to the downward
action of aerostatic lift force, and consequently structural stiffness and also the
response frequency are increased. But the flutter speeds increase slightly due to this
effect. On the other hand, it could be attributed to the coupling between the vertical
and torsional response. As wind speed increases, the mode shape, particularly the
torsional mode, changes greatly. The lateral and vertical displacement components
are enhanced, and the affinity between the bending and torsional modes is then
decreased. In the case of +31 wind attack angle, the response frequencies decrease
with wind speed, but the flutter speed increases due to the coupling effect.
In Case A, the flutter speeds decrease by 8.1 and 10.8 m/s, respectively at the wind
attack angles of +31 and –31 when the nonlinear effect of aerodynamic force varying
with the deformation and its three-dimensional effect are considered. The decrease of
flutter speed is caused mainly by the additional wind attack angle of deformed bridge
structure under static wind loading. Fig. 9 shows the additional wind attack angles
distributing along the bridge deck under the flutter critical condition. The large
additional wind attack angle makes the aerodynamic shape of the bridge deck more
bluff, which leads to the great decrease in flutter speed. But under 01 wind attack
angle, the aerodynamic stability is improved due to the negative additional attack
angle because the aerodynamic stability at
31 attack angle is higher than that at 01
attack angle in this case.
It can be concluded from the above analyses that the nonlinear effects of structural
dynamic characteristics and aeroelastic force due to the deformation have an
X. Zhang et al. / J. Wind Eng. Ind. Aerodyn. 90 (2002) 1065–1080 1079
0.5
-0.5
-1.0
-3°
-1.5 0°
-2.0
+3°
-2.5
-800 -600 -400 -200 0 200 400 600 800
Span length(m)
Fig. 9. The additional wind attack angle along the bridge deck at flutter critical condition.
6. Concluding remarks
Acknowledgements
The research is partly supported by the National Science Foundation under grant
No. 59895410, which is gratefully acknowledged.
References