See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/270494667

Flutter analysis of long-span bridges using ANSYS

Article in Wind and Structures · February 2007

DOI: 10.12989/was.2007.10.1.061

CITATIONS READS

54 3,427

4 authors, including:

Xugang Hua Z. Q. Chen

Hunan University Hunan University
160 PUBLICATIONS 3,093 CITATIONS 165 PUBLICATIONS 2,749 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Z. Q. Chen on 25 January 2016.

The user has requested enhancement of the downloaded file.

Wind and Structures, Vol. 10, No. 1 (2007) 61-82 61

Flutter analysis of long-span bridges using ANSYS

†
X. G. Hua

Department of Civil and Structural Engineering, The Hong Kong Polytechnic University,
Hung Hom, Kowloon, Hong Kong
‡
Z. Q. Chen

College of Civil Engineering, Hunan University, Changsha, Hunan 410083, P. R. China

‡† ‡‡
Y. Q. Ni and J. M. Ko

Department of Civil and Structural Engineering, The Hong Kong Polytechnic University,
Hung Hom, Kowloon, Hong Kong
(Received February 3, 2006, Accepted November 23, 2006)

Abstract. This paper presents a novel finite element (FE) model for analyzing coupled flutter of long-
span bridges using the commercial FE package ANSYS. This model utilizes a specific user-defined
element Matrix27 in ANSYS to model the aeroelastic forces acting on the bridge, wherein the stiffness
and damping matrices are expressed in terms of the reduced wind velocity and flutter derivatives. Making
use of this FE model, damped complex eigenvalue analysis is carried out to determine the complex
eigenvalues, of which the real part is the logarithm decay rate and the imaginary part is the damped
vibration frequency. The condition for onset of flutter instability becomes that, at a certain wind velocity,
the structural system incorporating fictitious Matrix27 elements has a complex eigenvalue with zero or
near-zero real part, with the imaginary part of this eigenvalue being the flutter frequency. Case studies are
provided to validate the developed procedure as well as to demonstrate the flutter analysis of cable-
supported bridges using ANSYS. The proposed method enables the bridge designers and engineering
practitioners to analyze flutter instability by using the commercial FE package ANSYS.
Keywords: long-span bridge; coupled flutter; instability; complex eigenvalue analysis; finite element (FE)
model; ANSYS.

1. Introduction
It is well known that flexible and slender structures such as long-span cable-supported bridges,
high-rise buildings and chimneys are prone to a variety of wind-induced vibrations due to their low

†PhD Candidate, E-mail: cexghua@polyu.edu.hk

‡Professor, E-mail: zqchen@hnu.cn
‡†Associate Professor, Corresponding Author, E-mail: ceyqni@polyu.edu.hk
‡‡Chair Professor, E-mail: cejmko@inet.polyu.edu.hk
62 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

natural frequency and mechanical damping (Simiu and Scanlan 1996). The aeroelastic instabilities
have become one of important considerations during the design of long-span bridges since the
collapse of old Tocoma Narrows suspension bridge. The most dangerous one among various
aeroelastic instabilities is flutter which is a dynamic instability phenomenon, wherein at some
critical wind velocity the bridge oscillates in a divergent and destructive manner. As a result, flutter
instability is prohibitive during the design of long-span bridges, and the critical flutter wind velocity
of a bridge must exceed the design value. The objective of flutter analysis is to predict the lowest
critical wind velocity that induces flutter instability, and the corresponding flutter frequency.
During the past four decades, comprehensive studies have been carried out to develop procedures
for analyzing coupled flutter of long-span bridges by integrating analytical skills with measured
flutter derivatives. Bleich (1948) was among the first to analyze the coupled flutter problem of
suspension bridges using Thoedorsen’s formulation on unsteady aeroelastic forces. The coupled
flutter analysis for long-span bridges using measured flutter derivatives from wind tunnel tests of
section model was pioneered by Scanlan and his co-workers (Scanlan and Tomko 1971, Scanlan
1978, Scanlan and Jones 1990). At present there are two general approaches for coupled flutter
analysis of bridges: (i) the multimode flutter analysis approach where the equations of motion for
structures are represented using a modal superposition technique (Agar 1989, Namini, et al. 1992,
Tanaka, et al. 1992, Chen 1994, Katsuchi, et al. 1999, Ding, et al. 2002), and (ii) the full-order
flutter analysis approach where the aeroelastic loadings are applied directly to the physical
coordinate of structures (Miyata and Yamada 1990, Dung, et al. 1998, Ge and Tanaka 2000, Ding,
et al. 2002). A lot of research efforts have been devoted to developing efficient methods for
solution of the complex eigenvalue problem in flutter analysis (Tanaka, et al. 1992, Jain, et al.
1996, D’Asdia and Sepe 1998, Dung, et al. 1998, Katsuchi, et al. 1999) and formulating appropriate
expressions of unsteady self-excited aerodynamic forces for flutter analysis (Xie and Xiang 1985,
Boonyapinyo, et al. 1999, Chen, et al. 2000, Briseghella, et al. 2002).
Since the 1970s a variety of commercial finite element (FE) packages such as ANSYS, ABAQUS
and ADINA have emerged and received wide applications in various disciplines along with the
advancement of FE methods and computing technologies. These FE packages have friendly
graphical user interface and powerful computational capability. However, the general purpose
commercial FE packages commonly used in civil engineering community cannot be directly used
for flutter analysis of large-scale bridges due to lack of the capability of calculating motion-
dependent wind loads. Although it is possible to develop special purpose FE packages to tackle
flutter analysis of bridges such as ANSUSP (Agar 1989), NACS (Chen 1994) and NASAB (Xiao
and Cheng 2004), the incorporation of functions or modules capable of flutter analysis into general
purpose commercial FE packages provides an alternative way.
This paper presents a novel FE formulation for the analysis of coupled flutter of long-span bridges
using the commercial FE package ANSYS. In this formulation, a user-defined element in ANSYS,
namely Matrix27 (SASI 2004), is adapted to model aeroelastic forces acting on the deck of long-
span bridges. The aeroelastic stiffness and damping matrices in Matrix27 elements are derived and
expressed in terms of the flutter derivatives, using either a lumped or consistent formulation. Then
damped eigenvalue analysis is carried out with respect to the integrated system of the structure
incorporating a series of Matrix27 elements, to determine the real and imaginary parts of complex
eigenvalues. The condition for onset of flutter instability becomes that, at a certain wind velocity,
the system has a complex eigenvalue with zero or near zero real part and the imaginary part of the
eigenvalue being the flutter frequency. Case studies of three structures with typical flutter derivative
 Flutter analysis of long-span bridges using ANSYS 63

parameters, namely a simply supported beam-like bridge with thin-airfoil cross section, a cable-
stayed bridge with bluff cross section and a suspension bridge with streamline cross section, are
conducted to validate the developed procedure and to demonstrate the flutter analysis of cable-
supported bridges using the commercial FE package ANSYS.
2. Formulation
2.1. FE for flutter analysis
The equation of motion of a bridge structure in the smooth flow can be expressed as
·· + · + =MX CX KX F se (1)
where , and are the global mass, damping and stiffness matrices, respectively; , · and ··
M C K X X X

represent the nodal displacement, velocity and acceleration vectors, respectively; se denotes the F

vector of the nodal aeroelastic forces.

The aeroelastic forces acting on unit span of bridge girder can be expressed as a linear function of
nodal displacement and nodal velocity (Scanlan 1978, Jain, et al. 1996):
· · ·
Lse = 1--- ρ U 2 ( 2 B ) KH *1 ---h- + KH *2 B
-------α + K2 H 3 α + K2 H4 --h- + KH *5 ---p- + K2 H6 --p-
* * *
(2a)
2 U U B U B
· · ·
D se = 1--- ρ U 2 ( 2 B ) KP *1 ---p- + KP*2 B
-------α + K2 P3 α + K2 P4 --p- + KP*5 ---h- + K 2 P6 --h-
* * *
(2b)
2 U U B U B
· · ·
M se = 1--- ρ U 2 ( 2 B2 ) KA *1 ---h- + KA *2 B
-------α + K2 A3 α + K2 A4 --h- + KA*5 ---p- + K2 A6 ---p-
* * *
(2c)
2 U U B U U
ω B-
where ρ is air mass density; U is wind velocity; B is the width of bridge deck; K = ------- is the
U
reduced circular frequency; h, p and α are the vertical, lateral and torsional displacements,
respectively; each dot denotes the differentiation with respect to time t; A*i , H*i and P*i (i=1, … ,
6) are non-dimensional flutter derivatives, which are functions of the reduced frequency and
determined from wind tunnel tests of section model of the bridge deck. The aeroelastic forces acting
on bridge deck are illustrated in Fig. 1.
Eqs. (2a) to (2c) represent aeroelastic forces distributed on unit deck length. In finite element

Fig. 1 Aeroelastic forces acting on bridge deck

64 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

analysis, these distributed forces are converted into equivalent nodal loadings acting at the member
ends. Thus the aeroelastic forces for element e can be expressed in terms of nodal displacement and
nodal velocity as
e
ae =
F
e
ae
e
+ eae · e
K X C X (3)
where eae and eae represent the local aeroelastic stiffness and damping matrices for element e,
K C

respectively. Similar to the general procedures in formulating element mass matrix, either a lumped
or consistent formulation can be used to derive the element aeroelastic stiffness and damping
matrices (Namini 1991). Making use of the lumped formulation, the expressions of eae and eae K C

are obtained as
e e
e
K ae = K ae 1 0 , e
C ae = C ae 1 0 (4a, b)
e e
0 K ae 1 0 C ae 1

0 0 0 0 0 0
* *
0 P6 P4 BP *3 0 0
e
K ae 1 =a0 H *6 H *4 BH *3 0 0 (4c)
* * *
0 BA6 BA4 B2 A3 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
* *
0 P5 P1 BP *2 0 0
e
C ae 1 =b0 H *5 H *1 BH *2 0 0 (4d)
* * *
0 BA 5 BA 1 B2 A2 0 0
0 0 0 0 0 0
0 0 0 0 0 0

where a = ρU2K2Le/2 and b = ρUBKLe/2; Le is the length of element e.

The user-defined element in ANSYS, Matrix27, is a versatile element with two nodes each having
six degrees of freedom, and with its local coordinate system being coincident with the global
coordinate system (SASI 2004). The element is arbitrary in geometrical configuration and the
element properties are specified by stiffness, mass, and damping coefficients. Fig. 2 illustrates the
geometry configuration of Matrix27 and its local coordinate system. Compared with other structural
elements in ANSYS, Matrix27 possesses some unique features: (i) user-specified coefficients of
mass, stiffness or damping matrices instead of physical parameters such as mass density, Young’s
modulus; (ii) accommodating both symmetric and asymmetric element matrices; and (iii)
representation of only either a mass element, or a stiffness element, or a damping element. The first
two features enable the modeling of self-excited forces using Matrix27 in ANSYS, and the last
feature implies that two elements are needed to model the self-excited force acting on each node.
 Flutter analysis of long-span bridges using ANSYS 65

Fig. 2 Geometry configuration of Matrix27

Fig. 3 Hybrid finite element model for flutter analysis in ANSYS

The first step for flutter analysis using ANSYS is to simulate the aeroelastic forces acting on each
node by element Matrix27. To achieve this, a hybrid FE model incorporating one structural element
with four Matrix27 elements as illustrated in Fig. 3 is formulated. Because one Matrix27 element
can only model either an aeroelastic stiffness matrix or an aeroelastic damping matrix instead of
both of them simultaneously, a pair of Matirx27 elements are attached to each of the nodes in a
structural element to simulate the aeroelastic forces. For a deck element e as shown in Fig. 3,
Matrix27 elements e1 and e3 are attached after defining a fictitious node k to represent the
aeroelastic stiffness and damping at node i, while Matrix27 elements e2 and e4 are attached after
defining a fictitious node l to represent the aeroelastic stiffness and damping at node j. The pair of
Matrix27 elements attached to each structural node share the same nodes.
The aeroelastic stiffness and damping matrices for the four Matrix27 elements attached to a deck
element e can be expressed as
e1 e–1 e e2 e e+1
K = K ae + K ae , K = K ae + K ae (5a, b)
e3 e–1 e e4 e e+1
C = C ae + C ae , C = C ae + C ae (5c, d)
Assembling all element matrices into global aeroelastic stiffness and damping matrices leads to
= F ae + ·
K ae X C ae X (6)
where ae and ae denote the global aeroelastic stiffness and damping matrices, respectively.
K C

Substituting Eq. (6) into Eq. (1) results in the governing equation of motion for the structure after
incorporating Matrix27 elements, as
66 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

MX
·· + ( –
C C ae X) · +( – K ) =
K ae X 0 (7)
Eq. (7) represents an integrated system with the effect of aeroelasticity, parameterized in terms of
wind velocity and response frequency. With this equation, damped complex eigenvalue analysis can
be carried out to determine the characteristic of the parameterized system.
The system dynamic response can be approximated by a superposition of the first m conjugate
pairs of complex eigenvalues and eigenvectors, as
m
X = ∑Φe
j=1
j
λt
j
(8)

where Φj = pj ± i qj is the jth complex conjugate pair of eigenvectors; λ j = σj ± i ω j is the jth

conjugate pair of complex eigenvalues; and i = –1 .
The system is dynamically stable if the real part of all eigenvalues is negative and dynamically
unstable if the real part of one or more eigenvalues is positive. The condition for occurrence of
flutter instability is then identified as follows: at a certain wind velocity Uf the system has one
complex eigenvalue λf with zero or near zero real part, the corresponding wind velocity Uf being
the critical flutter wind velocity and the imaginary part ω f of the complex eigenvalue λf becoming
the flutter frequency.
For large-scale civil engineering structures, the FE model usually involves thousands of degrees of
freedom and it is impractical to compute all eigenvalues and eigenvectors. As flutter in real
structures always occurs with the lowest flutter wind velocity corresponding to low-order
eigenvalues, only the first several eigenvalues are required in complex eigenvalue analysis.
2.2. Incorporation of mechanical damping
The mechanical damping of a structure is generally given in terms of modal damping ratios by
assuming the Rayleigh damping matrix
C=α +β M K (9)
where α and β are proportionality coefficients. When the damping ratios for the mth and nth modes are
measured or assumed, the proportionality coefficients can be obtained by (Clough and Penzien 1993)
ωmωn
- ( ω n ξm – ω m ξn )
α = 2 ------------------ (10a)
ω 2n – ω 2m
ω m ω n ⎛ ----- ξm ξn ⎞
- + -------
β = 2 ------------------
- – (10b)
ω 2n – ω 2m ⎝ ω n ω m ⎠
where ω m and ξm are the circular frequency and damping ratio for the mth mode; ω n and ξn are
the circular frequency and damping ratio for the nth mode.
After incorporating the mechanical damping, the governing equation of motion for flutter analysis
becomes
·· + ( – ae ) · + ( – ) = 0
MX C C X aeK K X (11)
 Flutter analysis of long-span bridges using ANSYS 67

where C is the modified damping matrix and ae is the modified aeroelastic damping matrix.
C

They are expressed as

= α + β ( – ae )
C M K K (12a)
C ae = C ae +β K ae (12b)
3. Algorithm implementation in ANSYS
As shown in Eq. (4), the aeroelastic stiffness and damping matrices in Matrix27 elements are
expressed in terms of three parameters, namely wind velocity, response frequency, and reduced
frequency, but only two of them are independent. As a result, the identification of condition for
occurrence of flutter instability involves a sweep and iterative procedure. In this study the wind
velocity and response frequency are selected as independent variables in solution, and the
investigation of flutter instability involves a sweep through a range of wind velocity and iteration
with respect to response frequency.
The damped eigensolution of Eq. (7) or Eq. (11) yields m complex conjugate pairs of eigenvalues
of the form λ i = σ i ±i ω i ( i = 1 , 2 , … , m ) . Because the complex mode which corresponds to a
real flutter frequency is a priori unknown, a mode-by-mode tracing method (Ge and Tanaka 2000)
is employed to iteratively search the flutter frequency and determine the critical flutter wind
velocity. The implementation of the developed flutter analysis procedure in ANSYS is summarized
in the following steps:
1) Establish the initial structural FE model without Matrix27 elements and compute the first m
natural modes ω 0i ( i = 1 , 2 , … , m ) ;
2) Establish the FE model of the integrated system with Matrix27 elements, in which the flutter
derivatives are inputted through the command TABLE in ANSYS;
3) Set an initial guess of critical wind velocity U0 and its increment ∆ U . Let the initial
oscillation frequency ω 0 be the frequency ω 0i of each natural mode in turn. Given the
tolerance ε;
4) Determine the reduced wind velocity and the aeroelastic stiffness and damping matrices in
Matrix27 elements at the present iteration, and then carry out the damped eigenvalue analysis;
5) Compare the imaginary part of the ith computed complex eigenvalue λ i with ω0. If
) 5 Im ( λi ) – ω -0 > ε , let ω0 = Im(λi) and repeat steps 4 and 5; otherwise go to step 6;
---------------------------
Im ( λ i )
6) Repeat steps 4 and 5 over all m computed natural modes. If the real parts of all complex
eigenvalues λ i ( i = 1 , 2 , … , m ) are negative, let U = U0 + ∆ U and repeat steps 4 and 5;
otherwise terminate the iteration.
The commercial FE package ANSYS provides three tools for users to customize and expand its
existing capabilities. These are ANSYS Parametric Design Language (APDL), User Interface
Design Language (UIDL) and User Programmable Features (UPFs) (SASI 2004). The APDL is a
scripting language that enables users to automate common tasks or even build the FE model in
terms of parameters (variables). It also encompasses a wide range of other features such as
repeating a command, macros, if-then-else branching, do-loops, and scalar, vector and matrix
operations. Making use of the tool APDL, all aforementioned steps can be readily implemented in
ANSYS without difficulty.
68 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

4. Case studies
4.1. Simply supported thin-airfoil structure
The first application illustrated here is the flutter analysis of a simply supported beam-like bridge
with thin-airfoil cross section. Since the theoretical solution of flutter frequency and flutter wind
velocity for this structure is available, this example serves as a verification of the developed
procedure. The parameters of the structure are as follows: span l = 300 m; width of the bridge deck
B = 40 m; vertical flexural rigidity EIz = 2.1×106 MPa·m4; lateral flexural rigidity EIy = 1.8×107 MPa·m4;
torsional rigidity GIt = 4.1×105 MPa·m4; mass m = 20,000 kg/m; mass moment of inertia Im = 4.5×106
MPa·m2/m; air mass density ρ = 1.248 kg/m3. The structural damping is assumed to be zero.
The unsteady aeroelastic forces acting on a thin-airfoil cross section in smooth flow were first
analytically derived by Theodorsen (1935). The relation between the flutter derivates and the
unsteady aeroelastic forces can be found in literature (e.g. Ge and Tanaka 2000). Fig. 4 shows the
variation of flutter derivatives versus the reduced wind velocity U/fB.
Natural mode analysis of the bridge FE model without Matrix27 elements is first conducted,
where the lumped mass formulation is used to construct the mass matrix for bridge distributed mass
and mass moment of inertia. The bridge is discretized by 30 deck elements as shown in Fig. 5. The
two-node beam element Beam4 in ANSYS is used to represent the deck elements; and the element
Mass21 in ANSYS is used to model the mass moments of inertia. The first ten natural modes are
extracted using ANSYS and summarized in Table 1.
After establishing the structural FE model, Matrix27 elements are incorporated into the model to
represent the aeroelastic forces acting on the bridge deck for damped complex eigenvalue analysis.

Fig. 4 Flutter derivatives for thin airfoil

Fig. 5 Finite element model for natural mode analysis

 Flutter analysis of long-span bridges using ANSYS 69

Table 1 Description of the first ten modes

Mode No. Frequency Mode shape Mode No. Frequency Mode shape
(Hz) (Hz)
1 0.1788 S-V 6 1.5030 S-T
2 0.5028 S-T 7 1.6096 S-V
3 0.5236 S-L 8 1.9976 A-T
4 0.7154 A-V 9 2.0944 A-L
5 1.0043 A-T 10 2.4867 S-T
Note: S – symmetric; A – asymmetric; V – vertical; L – lateral; T – torsional.

Fig. 6 Finite element model for flutter analysis

It is accomplished by generating the fictitious nodes and linking them with the original structural
nodes. A total of 58 Matrix27 elements are employed to formulate 29 aeroelastic stiffness matrices
and 29 aeroelastic damping matrices. The FE model of the bridge incorporating Matrix27 elements
for flutter analysis is illustrated in Fig. 6.
The initial wind velocity U0 in flutter analysis is estimated from the Van der Put’s formula (1976)

U f 0 = ⎛⎝ 1 + 0.72 ⎛⎝ --r-⎞⎠ µ ( ε – 0.05 )⎞⎠ ω h b (13)

b
where ε = ω α ⁄ ω h is the ratio of frequencies between the torsional natural mode and the vertical
bending natural mode; µ = m/(πρb2) is the ratio of bridge mass density to air mass density;
r = I m ⁄ m is the radius of gyration; and b is half of the bridge width. The initial trial frequency
for each complex mode is taken as the computed frequency of the corresponding natural mode.
Following the computational steps described in the previous section, damped complex eigenvalue
analysis is conducted for the integrated system under wind velocity ranging from 0 to 180 m/s. The
first ten conjugate pairs of complex eigenvalues and complex eigenvectors are obtained, and the
variation of these complex eigenvalues versus wind velocity is plotted in Fig. 7. It is observed that,
in the considered wind velocity range, (i) the vibration frequencies (i.e. the imaginary part of
complex eigenvalues) of vertical bending modes exhibit a slight increase with increase of wind
velocity while the real part decreases with increasing wind velocity; (ii) for lateral bending modes,
both the real and imaginary parts of complex eigenvalues remain unchanged with the increase of
wind velocity; and (iii) the imaginary part of complex eigenvalues for torsional modes decreases
with the increase of wind velocity, while the real part decreases at the outset and then increases with
increasing wind velocity. As shown in Fig. 7(a), the real part of the second complex mode becomes
zero at a wind velocity of 135.1 m/s, and the corresponding imaginary part of the complex
eigenvalue becomes 0.3940 Hz, identifying the onset of flutter instability. The input file for ANSYS
70 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

Fig. 7 Variation of complex eigenvalues versus wind velocity

Fig. 8 Variation of generalized mass versus wind velocity

to conduct flutter analysis of this structure is provided in the Appendix.

To illustrate the characteristic of coupled flutter, Fig. 8 provides the variation of the generalized
mass in vertical and torsional directions for the second complex eigenvalue versus wind velocity. As
expected, the second complex mode is a purely torsional mode with the generalized mass in
torsional direction being unity when the wind velocity is zero, and then it becomes a vertically and
torsionally coupled mode with the increase of wind velocity.
For this simple structure, a time-domain analysis of the integrated system can be carried out to verify the
frequency-domain computational results. For a given wind velocity, the time-domain response of the
system subjected to an initial excitation can be readily computed using a time integration scheme. Figs. 9
to 11 show the response time histories of mid-span under different wind velocities. It is seen that the
system is neutrally stable with a response frequency of 0.3940 Hz at the critical wind velocity of 134.1 m/s
(Fig. 9). The system is dynamically stable when the wind velocity is lower than the critical value (Fig. 10)
 Flutter analysis of long-span bridges using ANSYS 71

Fig. 9 Time-domain response of mid-span at wind velocity of 134.1 m/s

Fig. 10 Time-domain response of mid-span at wind velocity of 120.0 m/s

Fig. 11 Time-domain response of mid-span at wind velocity of 140.0 m/s

and becomes dynamically unstable when the wind velocity is larger than the critical value (Fig. 11).
Table 2 gives a comparison of the results obtained by different methods. The exact solution of the
flutter frequency and critical wind velocity is obtained by using the two-mode classical flutter
theory (Theodorsen 1935, Bleich 1948). The result from the M-S method (Chen 1994) taking into
account multimode participation is also listed in the table for comparison. It is seen that the present
procedure gives rise to agreeable results with the multimode method as well as the exact solution.
72 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

Table 2 Comparison of flutter analysis results

Method Flutter velocity (m/s) Flutter frequency (Hz)
Present procedure (frequency domain) 135.1 0.3940
Present procedure (time domain) 134.1 0.3940
M-S method 134.3 0.3936
Exact solution 136.3 0.3914

Fig. 12 Illustration of Dongting Lake cable-stayed bridge

Fig. 13 Typical cross section of bridge deck

4.2. Dongting Lake cable-stayed bridge

The Dongting Lake Bridge, as shown in Fig. 12, is a multi-span cable-stayed bridge with three
towers supporting a center spans of 310 m each and two identical side spans of 130 m each. The
bridge towers are mede of reinforced concrete, and the heights of the central tower and the two side
towers above pylon base are 123.48 m and 98.3 m, respectively. The prestressed concrete bridge
deck has a shallow π-type cross section in a width of 23.4 m as shown in Fig. 13. It is supported by
two inclined cable planes emanating from deck anchorages to tower tops, each plane comprising
112 cables. The horizontal distance between two adjacent cables is 8 m.
The flutter derivatives H*i and A*i ( i = 1 , 2 , 3 , 4 ) have been determined by wind tunnel tests of a
scaled section model of the bridge deck (Chen and Yu 2002). Fig. 14 illustrates the 1:75 scale
section model used for wind tunnel tests. The flutter derivatives under different wind incidences
have been measured and those under a wind incidence of zero degree are shown in Fig. 15.
A triple-girder FE model considering the effect of warping stiffness has been established for the

Fig. 14 Scaled section model for wind tunnel testing (unit: mm)
 Flutter analysis of long-span bridges using ANSYS 73

Fig. 15 Flutter derivatives of scale section model

bridge for natural mode analysis. The structural model, as shown in Fig. 16, is composed of 925
elements. The first twenty natural modes are extracted using the Lanczos method in ANSYS, and
the natural frequencies and mode shapes of the first ten modes are summarized in Table 3.

Fig. 16 Finite element model of Dongting Lake Bridge for natural mode analysis

Table 3 Description of the first ten modes for Dongting Lake Bridge
Mode No. Frequency Mode shape Mode No. Frequency Mode shape
(Hz) (Hz)
1 0.1369 A-LS 6 0.6118 S-L
2 0.2332 A-V 7 0.6813 S-V
3 0.4329 S-V 8 0.8258 A-V
4 0.4884 A-V 9 0.8574 S-V
5 0.5283 A-L 10 0.9339 S-T
Note: S – symmetric; A – asymmetric; V – vertical; L – lateral; T – torsional; LS – longitudinal sway.
74 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

Fig. 17 Finite element model of Dongting Lake Bridge for flutter analysis

Fig. 18 Variation of complex eigenvalues versus wind velocity for Dongting Lake Bridge

A series of Matrix27 elements are then attached to the nodes at the central girder to model
aerodynamic coupling effects between the bridge and air flow. A total of 226 Matrix27 elements are
used, half of them modeling the aerodynamic stiffness while the remaining modeling the
aerodynamic damping. Fig. 17 illustrates the FE model of the bridge incorporating Matrix27
elements for flutter analysis, where the boundary conditions for the constrained nodes are not
displayed for clarity.
Damped complex eigenvalue analysis is first carried out without considering mechanical damping
under wind velocity ranging from 0 to 100 m/s. The first ten conjugate pairs of complex
eigenvalues and complex eigenvectors are obtained, and the variation of these complex eigenvalues
versus wind velocity is plotted in Fig. 18. It is observed that the imaginary part of complex
eigenvalues for all modes remains almost unchanged in the considered wind velocity range, while
the real part of complex eigenvalues for bending and torsional modes exhibits a significant
alteration with the increase of wind velocity. As shown in Fig. 18(a), the real part of the 10th
complex mode crosses over zero at a wind velocity of 63.1 m/s, identifying the occurrence of flutter
instability. The flutter frequency is predicted as 0.9315 Hz.
 Flutter analysis of long-span bridges using ANSYS 75

Fig. 19 Least square fit of proportionality coefficients

Fig. 20 Illustration of Humen suspension bridge

Then the analysis is repeated by assuming that the damping ratios ξj for all the first ten modes are
0.5%. In this case the proportionality coefficients α and β of Rayleigh damping matrix are obtained
by least square fit of the function
2 ω j ξj = α + βω 2j (14)
Fig. 19 shows the function values obtained from the assumed modal damping ratios and those
reconstructed by use of the fitted proportionality coefficients. After introducing the mechanical
damping matrix in the complex eigenvalue analysis, the flutter wind velocity is predicted as 183.5
m/s and the corresponding flutter frequency is 0.9246 Hz.
4.3. Humen suspension bridge
The Humen Bridge, as illustrated in Fig. 20, is a suspension bridge with a main span of 888 m
(Zheng and Yang 1998). The bridge deck is a stiffening steel box girder of 36.1 m wide and 3.0 m
high. The two main cables are 33 m apart and the bridge deck is suspended by hangers at intervals
of 12 m. The two bridge towers are 150.5 m high reinforced concrete structures. Wind tunnel tests
have been conducted on an aeroelastic model of the bridge (SLDRCE 1995) and the flutter
derivative parameters A*2 , A*3 and H*1 have been measured under the wind attack angle of 0 degree
and 3 degrees, respectively. Fig. 21 shows the measured flutter derivatives.
In recognizing that only partial flutter derivative parameters were measured, complex eigenvalue
analysis is first carried out using theoretical flutter derivatives of the airfoil-like cross section. By
76 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

Fig. 21 Flutter derivatives of bridge deck

Fig. 22 Variation of complex eigenvalues versus wind velocity obtained using theoretical flutter derivatives

disregarding mechanical damping and sweeping wind velocity from 0 to 120 m/s, the first ten
conjugate pairs of complex eigenvalues and complex eigenvectors are extracted and the variation of
the complex eigenvalues versus wind velocity is shown in Fig. 22. The closely spaced modes make
it quite difficult to correctly trace a specified mode during the sweep of wind velocity. To ensure a
correct mode trace, the following correlation coefficient of mode shapes is examined for each
increment step of wind velocity
(φ ϕ) T 2
C = ---------------------------- (15)
(φ φ)(ϕ ϕ)
T T

where φ is the mode shape obtained at the previous wind velocity step Ui−1; and ϕ is the mode
shape obtained at the current wind velocity step Ui. More the value of C is close to 1, more likely
the two mode shapes are similar. From Fig. 22, the critical flutter wind velocity and flutter
 Flutter analysis of long-span bridges using ANSYS 77

Fig. 23 Variation of complex eigenvalues versus wind velocity obtained using measured flutter derivatives at
0o attack angle

Fig. 24 Variation of complex eigenvalues versus wind velocity obtained using measured flutter derivatives at
3o attack angle

frequency are estimated to be 85.90 m/s and 0.2781 Hz, respectively, when having no mechanical
damping. Then the complex eigenvalue analysis is conducted again by assuming that the damping
ratios for all modes are equal to 0.5%. In this case the critical flutter wind velocity and flutter
frequency are computed to be 92.23 m/s and 0.2752 Hz.
As mentioned before, only three flutter derivative parameters were measured for the Humen
Bridge. The remaining flutter derivative parameters are calculated using Theodorsen’s formulation
(Theodorsen 1935). Then the measured flutter derivative parameters together with the calculated
flutter derivative parameters are used to predict the critical flutter wind velocity and flutter
frequency under the wind attack angle of 0 degree and +3 degrees, respectively, where the
mechanical damping ratios for all concerned modes are set as 0.5% which was also adopted in the
aeroelastic model test of the bridge (SLDRCE 1995). Figs. 23 and 24 illustrate the variation of the
predicted complex eigenvalues versus wind velocity under the wind attack angle of 0 degree and +3
degrees, respectively. Table 4 shows a comparison of the flutter wind velocity and flutter frequency
78 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

Table 4 Comparison of flutter analysis results

Flutter velocity (m/s) Flutter frequency (Hz)
Method
0 attack angle
o
3o attack angle 0 attack angle
o
3o attack angle
Present procedure 89.36 75.20 0.2973 0.3050
M-S method 89.55 71.68 0.3009 0.3089
Wind tunnel tests 88 72 --- ---

predicted by the present procedure with those obtained using the M-S method and measured from
wind tunnel tests. A good agreement between the results obtained by the three approaches is
observed.
5. Conclusions
An ANSYS-based FE model for coupled flutter analysis of long-span bridges is developed in this
paper. The FE model utilizes a user-defined element, Matrix27, to model the aeroelastic forces
acting on the bridge, where the entries of aeroelastic stiffness and damping matrices in Matrix27
elements are parameterized in terms of wind velocity and response frequency. The critical flutter
wind velocity and flutter frequency are determined through complex eigenvalue analysis of an
integrated system of the structural FE model incorporating Matrix27 elements. Three case studies
are provided to verify the proposed method and demonstrate its capability for analyzing coupled
flutter of long-span bridges. In these examples, the flutter analysis results obtained by the proposed
method are compared with those obtained by an analytical solution, a multi-mode analysis
procedure, or wind tunnel tests, and a good agreement is observed. The proposed method enables
the full-order flutter analysis of long-span bridges to be conducted using the commercial FE
package ANSYS.
Acknowledgements
The work described in this paper was supported in part by National Natural Science Foundation
of China under Grant Nos. 50178103 and 50478051 and partially by The Hong Kong Polytechnic
University under Grant No. G-T770. These supports are gratefully acknowledged.
References
Agar, T. J. A. (1989), “Aerodynamic flutter analysis of suspension bridges by a modal technique”, Eng. Struct.,
11, 75-82.
Bleich, F. (1948), “Dynamic instability of truss-stiffened suspension bridges under wind action”, Transactions of
the American Society of Civil Engineers, Paper No. 2385, 1177-1222.
Boonyapinyo, V., Miyata, T. and Yamada, H. (1999), “Advanced aerodynamic analysis of suspension bridges by
state-space approach”, J. Struct. Eng., ASCE, 125, 1357-1366.
Briseghella, L., Franchetti, P. and Secchi, C. (2002), “Time domain flutter analysis of the Great Belt East
Bridge”, Wind and Struct., 5, 479-492.
Chen, X., Matsumoto, M. and Kareem, A. (2000), “Aerodynamic coupling effects on flutter and buffeting of
bridges”, J. Eng. Mech., ASCE, 126, 17-26.
 Flutter analysis of long-span bridges using ANSYS 79

Chen, Z. Q. (1994), “The three dimensional analysis and behaviors investigation on the critical flutter state of
bridges”, Proceedings of the International Symposium on Cable-Stayed Bridges, Shanghai, China, 10-13.
Chen, Z. Q. and Yu, X. D. (2002), “A new method for measuring flutter self-excited forces of long-span
bridges”, China Civil Eng. J., 35, 34-41 (in Chinese).
Clough, R. W. and Penzien, J. (1993), Dynamics of Structures, 2nd edition, McGraw-Hill, New York.
D’Asdia, P. and Sepe, V. (1998), “Aeroelastic instability of long-span suspended bridges: A multi-mode
approach”, J. Wind Eng. Ind. Aerodyn., 74, 849-857.
Ding, Q., Chen, A. and Xiang, H. (2002), “Coupled flutter analysis of long-span bridges by multimode and full-
order approaches”, J. Wind Eng. Ind. Aerodyn., 90, 1981-1993.
Dung, N. N., Miyata, T., Yamada, H. and Minh, N. N. (1998), “Flutter responses in long span bridges with wind
induced displacement by the mode tracing method”, J. Wind Eng. Ind. Aerodyn., 77-78, 367-379.
Ge, Y. J. and Tanaka, H. (2000), “Aerodynamic analysis of cable-supported bridge by multi-mode and full-mode
approaches”, J. Wind Eng. Ind. Aerodyn., 86, 123-153.
Jain, A., Jones, N. P. and Scanlan, R. H. (1996), “Coupled flutter and buffeting analysis of long-span bridges”, J.
Struct. Eng., ASCE, 122, 716-725.
Katsuchi, H., Jones, N. P. and Scanlan, R. H. (1999), “Multimode coupled flutter and buffeting analysis of the
Akashi-Kaikyo Bridge”, J. Struct. Eng., ASCE, 125, 60-70.
Miyata, T. and Yamada, H. (1990), “Coupled flutter estimate of a suspension bridge”, J. Wind Eng. Ind.
Aerodyn., 33, 341-348.
Namini, A., Albrecht, P. and Bosch, H. (1992), “Finite element-based flutter analysis of cable-suspended
bridges”, J. Struct. Eng., ASCE, 118, 1509-1526.
Namini, A. H. (1991), “Analytical modeling of flutter derivatives as finite elements”, Com. Struct., 41, 1055-
1064.
Scanlan, R. H. (1978), “Action of flexible bridges under wind, 1: flutter theory”, J. Sound Vib., 60, 187-199.
Scanlan, R. H. and Jones, N. P. (1990), “Aeroelastic analysis of cable-stayed bridges”, J. Struct. Eng., ASCE,
116, 270-297.
Scanlan, R. H. and Tomko, J. J. (1971), “Airfoil and bridge deck flutter derivatives”, J. Eng. Mech. Div., ASCE,
91, 1117-1137.
Simiu, E. and Scanlan, R. H. (1996), Wind Effects on Structures: Fundamentals and Applications to Design, 3rd
edition, John Wiley, New York.
State-key Laboratory for Disaster Reduction in Civil Engineering (SLDRCE) (1995), “Investigations on wind-
resistant behavior of Humen suspension bridge”, Research Report, Bulletin of Laboratory of Wind Tunnel,
Tongji University, Shanghai, China (in Chinese).
Swanson Analysis Systems Inc. (SASI) (2004), ANSYS User’s Manual, Version 8.0, Houston, Pennsylvania.
Tanaka, H., Yamamura, N. and Tatsumi, M. (1992), “Coupled mode flutter analysis using flutter derivatives”, J.
Wind Eng. Ind. Aerodyn., 42, 1279-1290.
Theodorsen, T. (1935), “General theory of aerodynamic instability and the mechanism of flutter”, NACA Report
No. 496, 1935.
Van der Put, M. (1976), “Rigidity of structures against aerodynamic forces”, Proceedings of the International
Association for Bridge and Structural Engineering, IABSE, Zurich, Switzerland.
Xiao, R. C. and Cheng, J. (2004), “Advanced aerostatic stability analysis of suspension bridges”, Wind and
Struct., 7, 55-70.
Xie, J. and Xiang, H. F. (1985), “State-space method for 3-D flutter analysis of bridge structures”, Proceedings
of the 1st Asia Pacific Symposium on Wind Engineering, India, 269-276.
Zheng, M. Z. and Yang, G. Z. (1998), “The Humen Pearl River Bridge”, Struct. Eng. Int., 2, 93-94.
Appendix: Macro for flutter analysis of the structure in case 1
/UIS,MSGPOP,3
/CLEAR
/PREP7
ET,1,beam4
80 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

ET,2,mass21
N,1,0,0,
N,31,300,0,0
FILL,1,31
R,1,10,10,85.714,40,0.25,,,
RMORE,0,5.076,,,,,,
R,2,,,,4500,,,,
MP,ex,1,2.1e7
MP,dens,1,0.2
MP,gxy,1,8.077e6
TYPE,1
REAL,1
*DO,i,1,30
E,i,i+1
*ENDDO
TYPE,2
REAL,2
*DO,i,2,30
E,i
*ENDDO
/SOLU
D,1,ux,0,0,,,uy,uz,rotx
D,31,uy,0,0,,,uz,rotx
ANTYPE,modal
MODOPT,lanb,10
MXPAND,10
LUMPM,on
SOLVE ! undamped eigenvalue analysis
/POST1
*DIM, freq0,,10,
*DO,i,1,10
*GET,freq0(I),mode,i,freq,,,,
*ENDDO
FINI
/PREP7
ET,3,matrix27,,1,4,,,,,
ET,4,matrix27,,1,5,,,,
R,3,,,,,, ! initialize real constant
R,4,,,,,,
R,5,,,,,,
R,6,,,,,,
NGEN,2,40,2,30,1,0,-10,0 ! generate the fictitious nodes
TYPE,3
REAL,3
*DO,i,3,29
E,i,i+40
*ENDDO
TYPE,4
REAL,4
*DO,i,3,29
E,i,i+40
*ENDDO
TYPE,3
REAL,5
E,2,42
 Flutter analysis of long-span bridges using ANSYS 81

TYPE,4
REAL,6
E,2,42
TYPE,3
REAL,5
E,30,70
TYPE,4
REAL,6
E,30,70
NSEL,s,,,42,70
D,all,all
! A1 = DRV(1,1)
! A2 = DRV(1,2)
! A3 = DRV(1,3)
! A4 = DRV(1,4)
! H1 = DRV(1,5)
! H2 = DRV(1,6)
! H3 = DRV(1,7)
! H4 = DRV(1,8)
FINISH
*DIM,STIF,,4,
*DIM,DMP,,4,
*DIM,ll,,2
*dim,drv,table,11,8,
*tread,drv,drv,txt,,, !input flutter derivatives from drv.txt
*DIM,freq1r,,10,
*DIM,freq1i,,10,
*cfopen,result,txt
b =40.0
p = 1.248e-4
ll(1)=10
ll(2)=15
! Notation:
! ii -- cycle over wind velocity
! jj -- cycle over number of comlex modes
! kk -- frequency iteration
*DO,ii,0,180 !wind velocity range
U=ii
*DO,jj,1,10 ! number of complex modes
/PREP7
omiga = freq0(JJ)*2*3.1415926
f0 = omiga/(2*3.1415926)
*DO,kk,1,5 ! iteration steps
/PREP7
ru = u/(f0*b)
rku =2*3.1415926*f0*b !RKU=U*RK
*DO,i,1,2
KKK= ll(i)*p*(rku**2) ! stiffness coef
 82 X. G. Hua, Z. Q. Chen, Y. Q. Ni and J. M. Ko

CCC = ll(i)*p*b*rku ! damping coef

STIF(1) = -KKK*DRV(RU,8)
STIF(2) = -KKK*B*DRV(RU,7)
STIF(3) = -KKK*B*DRV(RU,4)
STIF(4) = -KKK*B*B*DRV(RU,3)
DMP(1) = -CCC*DRV(RU,5)
DMP(2) = -CCC*B*DRV(RU,6)
DMP(3) = -CCC*B*DRV(RU,1)
DMP(4) = -CCC*B*B*DRV(RU,2)
RMODIF,2*(i+0.5),13,STIF(1),,STIF(2)
RMODIF,2*(i+0.5),34,STIF(4)
RMODIF,2*(i+0.5),64,STIF(1),,STIF(2)
RMODIF,2*(i+0.5),73,STIF(4)
RMODIF,2*(i+0.5),83,STIF(3)
RMODIF,2*(i+0.5),122,STIF(3)
RMODIF,2*(i+1),13,DMP(1),,DMP(2)
RMODIF,2*(i+1),34,DMP(4)
RMODIF,2*(i+1),64,DMP(1),,DMP(2)
RMODIF,2*(i+1),73,DMP(4)
RMODIF,2*(i+1),83,DMP(3)
RMODIF,2*(i+1),122,DMP(3)
*ENDDO
FINISH
/SOLU
ANTY,modal
MODOPT,damp,20
MXPAND,20
LUMPM,on
ALLSEL,ALL
SOLV
FINISH
*DO,I,1,10
K=2*(I-1)+1
!extracting the real part (damping) of complex modes
*GET,FREQ1R(I),MODE,K,FREQ,,,,
!extracting the imaginary part (frequency) of complex modes
*GET,FREQ1I(I),MODE,K,FREQ,IMAG,,,
*ENDDO
F0=FREQ1I(JJ)
*ENDDO
/post1
temp1=FREQ1R(jj)
temp2=FREQ1I(jj)
*vwrite,temp1,temp2
(4(f10.5,5x))
*ENDDO
*ENDDO
CC

View publication stats