Vous êtes sur la page 1sur 50

FLUTTER STABILITY ANALYSIS

THEORY AND EXAMPLE

Prepared by Le Thai Hoa

2004

1 | Le Thai Hoa Flutter stability analysis: Theory &Example

FLUTTER STABILITY ANALYSIS


1. INTRODUCTION
There are two typical types of bridge flutter were i) Torsional flutter that the
fundamental torsional mode dominantly involves to the flutter instability ii) Coupled flutter
that the fundamental torsional mode aerodynamically couples tendency with either of any first
symmetric or ansymmetric heaving mode at single frequency (called flutter frequency) and
also known as the so-called classical flutter (similarly to flutter of airfoil wings). Various
experiments and numerical analyses [Matsumoto et al.(1996,1997)] showed that,
moreover, the torsional flutter seems to dominate almost cases of bridges with bluff
bridge sections as low slenderness ratio (B/D) rectangular sections, H-shape sections,
stiffened truss sections, whereas streamlined boxed bridge sections are favorable for
coupled flutter. However, the Akashi-Kankyo bridge exhibited with coupled flutter
that this is never experienced before with stiffened truss sections.

Flutter generation mechanism might be more difficult, however, by uses of series of


experiments on various fundamental sections and based on flow-structure interaction
phenomena as local separation bubble, reattachment, vortex shedding on structural
surface that Matsumoto et al. (1996,2000) classified the mechanism of flutter
instability generation of 2D H-shaped and rectangular sections into detailed
branches: i) Low-speed torsional flutter, ii) High-speed torsional flutter, iii) Heaving-branch
coupled flutter, iv) Torsional-branch coupled flutter and coupled flutter, v) Heaving-torsional
coupled flutter.

Flutter problems can be approximately divided by analytical and experimental


methods and simulation. Experimental approach is thanks to free vibration tests on

2 | Le Thai Hoa Flutter stability analysis: Theory &Example

2D bridge sectional model in wind tunnel laboratory. Computational fluid dynamics


(CFD) technique has gained much development so far to become useful supplemental
tools beside analytical and experimental methods and it is also predicted broadly that
such the CFD might replace wind tunnel tests in future, however, this simulation
method still has many limitations to cope with complexity of bridge sections and
nature of 3D bridge structures.

Flutter problems

Analytical Methods

Empirical Formula

Selbergs; Kloppels

2DOF FlutterSolutions

ComplexEigenMethod
Step-by-Step Method

nDOF FlutterSolutions

Single-Mode Method
Two-Mode Method
Multi-mode Method

Experiment Method
Simulation Method

Free Vibration Method


Computational Fluid
Dynamics (CFD)

Fig. 1 Branches for flutter instability problems

Bleigh(1951) introduced empirical formula to calculate critical flutter velocity of


2DOF flutter problem for airfoil and thin-plate sections, Selberg(1961) developed
Bleichs formula by putting the shape ratio to apply for various types of bridge
sections, moreover, Kloppel(1967) exhibited under a form of empirical diagrams.
Theodorsen(1935) applied potential theory of airfoil aerodynamics by introducing socalled Theodorsens circulation functions to model self-controlled flutter forces,
meanwhile Scanlan(1971) used experimental approach to build such the selfcontrolled forces by so-called flutter derivatives. Because the potential theory
validates in certain conditions of non-separation and non-reattachment around
3 | Le Thai Hoa Flutter stability analysis: Theory &Example

structural sections, thus the Theodorsens self-controlled flutter forces are limitedly
applied only on flutter problems of airfoil and thin-plate structures, thus Scanlans
ones are widely applied so far for flutter analytical problems of 2DOF systems and 3D
bridge structures with various types of cross-sections.
For 2DOF flutter problems, there are two powerful analytical methods: the complex
eigenvalue

method

[Simui&Scanlan(1976)]

and

the

step-by-step

method

[Matsumoto(1995)]. Though the complex eigenvalue method has been applied for a
long in 2DOD flutter problems, but difficulty to investigate relationship of system
damping ratio, system frequency on wind velocity, inter-relation between flutter
derivatives as well, the step-by-step method is very favorable over such the abovementioned limitations to clarify a role of flutter derivatives on critical condition and
on flutter stabilization.

For analytical methods for bridge or nDOF systems flutter problems, there are two
approaches: i) finite differential method (FDM) in linear-time approximation and ii) finite
element method (FEM) in modal space. However, the most state-of-the-art development
of analytical methods has carried out in the later. Agar(1989) developed FDM for
flutter problem of suspension bridges. Scanlan(1987,1990) firstly introduced singmode and two-mode flutter analytical methods thanks to generalized transforms and
modal technique and based on idea that critical flutter conditions are prone to
dominant contribution of fundamental torsional mode (torsional flutter) or of
coupling between two torsional and heaving modes (coupled flutter). Many recent
studies [Pleif et al(1995), Katsuchi(1999), Ge et al.(2002)], however, pointed out that
in many cases of bridges there are not the fundamental torsional and heaving modes
involved to the critical flutter conditions, but many modes (multi-mode method)
superpose to generate more critical conditions.

4 | Le Thai Hoa Flutter stability analysis: Theory &Example

2. SINGLE TORSIONAL FLUTTER PROBLEM


The 1DOF motion equation of the torsional flutter (: torsional motion) can be written
as follow:
I C K

1
B
U 2 B 2 [ KA2*
K 2 A3* ]
2
U

(A1.1)

Transforming above equation to the ordinary form as:

2 *
Where: 2 =

1
B

U 2 B 2 KA2*
K 2 A3*
2I
U

K
C
;
I
2 K .I

(A1.2)

(A1.3)

We have the equation:

(2

1
B
1
U 2 B 2 KA2* ) (2 U 2 B 2 K 2 A3* ) 0
2I
U
2I

2 (

1
1
U B 3 KA2* ) (2 U 2 B 2 K 2 A3* ) 0
4 I
2I

(A1.4)

For simplifying, we can write:

2 2 0

(A1.5)

We easily write the solution of above equation under following form:

Ae t sin( t 0 )
Thus, the instability condition of the single torsional flutter follows:

0 or (

1
U B 3 KA2* ) 0
4 I

(A1.6)

1
U B 3 KA2*
4 I

As a result, A2*

4 I
UB 3 K

(A1.7)

Through above unequality, the significant role of the torsional-motion-related flutter


derivative A2* (aerodynamic damping force) can be clearly approved.
5 | Le Thai Hoa Flutter stability analysis: Theory &Example

3. COMPLEX EIGENVALUE FLUTTER PROBLEM FOR 2DOF HEAVINGTORSIONAL MOTION EQUATION SYSTEM
The 2DOF heaving and torsional motion equations of the flutter (h: heaving motion,
: torsional motion) can be expressed as follow:

1
h
B
mh C h h K h h U 2 B [ KH 1*
KH 2*
K 2 H 3* ]
2
U
U
I C K

(A2.1)

1
h
B
U 2 B 2 [ KA1*
KA2*
K 2 A3* ]
2
U
U

(A2.2)

Transforming above equations to the ordinary form:

1
h
B
h 2 h h h 2 h h
U 2 B KH1* KH 2*
K 2 H 3*
2m
U
U

2 *
Where: h2

Kh
m

(A2.3)

1
h
B
U 2 B 2 KA1* KA2*
K 2 A3*
2I

; 2 =

(A2.4)

K
Ch
C
; h
;
I
2 K h .m
2 K .I

Introducing time-dimensionless variable: s =

(A2.5)

Ut
B

(A2.6)

First-order, second-order differentials of t time, we have:


(.) =

d ( ) d ( ) ds
U

. ()
dt
ds dt
B

(A2.7)

(..) =

d 2 ( ) d 2 ( ) ds
U2

(
)'
'
dt 2
ds 2 dt
B2

A2.8)

Replacing eqs.(A2.7), (A2.8) into eqs.(A2.3), (A2.4), then dividing eq.(A2.3) by U 2 / B


and

eq.(A2.4)

by

U 2 / B2

we

have:

h"
B
B 2
h'
2 h h h .h' h2 2 h
[ KH *1 KA2* ' K 2 H *3* ] (A2.9)
B
U
2m
B
U
" 2

B '
B2
B 4
h'
2
2 . 2
[ KA1*
KA2* ' K A3* ]
U
2I
B
U

(A2.10)

6 | Le Thai Hoa Flutter stability analysis: Theory &Example

Putting K h

B h
B
, K
, replacing to eqs.(A2.9), (A2.10):
U
U

2 h
h"
h'
B 2
h'
2 h K h K h
[ KH 1* KH 2* ' K 2 H 3* ]
B
B
B 2m
B
2

" 2 K ' K

B 4
2I

(A2.11)

* h'

KA2* ' K 2 A3*


KA1
B

(A2.12)

Solution forms of the eqs.(A2.11), (A2.12) can be expressed under such ones as
follows:
h = h0 exp(it ) = h0 exp(

iKU sB
. ) h0 exp(iKs )
B U

(A2.13)
(A2.14)

0 exp(it ) 0 exp(iKs )
Replacing eqs. (A2.13), (A2.14) into eqs. (A2.11), (A2.12), we have:

K2
K
B 2 iK 2 *
B 2
B 2 2 *
2
*
[
2i h K h
Kh
H 1 ] h0 + [
i K H2
K H3 ] 0 0
B
B
2m B
2m
2m
4

B 4 K 2 *
B
B 2 2 *
[
i
A1 ] h0 [ K 2 2i K K K 2
i K 2 A2*
K H 3 ] 0 0
2I
B
2I
m

Conditioning that above homogenous equations have non-trivial solutions is that its
determinant must be zero:
2

B 2 2 * 1
B 2 2 * B
iK H 1 ] [
iK H 2
K 2 H 3* ]
2m
B
2m
2m
Det
=0 (A2.15)
4
4
4
2
B
1

B
[
iK 2 A1* ] [ K 2 2i K K K 2
iK 2 A2*
K 2 A3* ]
2I
B
2I
I
2

[ K 2 2i h K h K K h

Expanding the determinant (A2.15) and grouping by real part and imaginary one as
follow:
Det H = 1 i 2 =0

A2.16)

7 | Le Thai Hoa Flutter stability analysis: Theory &Example

Placing X

K h h

(A2.17)

The determinant (A2.15) is developed in such form as (A2.16). Then dividing the
2

determinant H by K h , we have:

B 2
B 2 * 2
B 2 * 2
2
*
2
(

1
)

(
2
i

iH
X
)
i

iH
X

(
H 2 X )i
h
1
3

2m
2m
2m

B 4 * 2
B 4 *
B 4 * 2 2
2
(

A
X
)
i
(

A
X

(
2

A2 X )i

2I 1

2
2I
h 2I
h

p11 p12 i
P21i

q11 q12 i

q 21 q 22 i

(A2.18)

B 2 * 2
B 2 * 2
B 2 *
Where: p11 X 1 ; p12 2 h X
H 1 X ; p 21
A1 X ; q11
H3 X ;
2m
2I
2m
2

q12

B 2 * 2
B 4 * 2 2
B 4 * 2
H 3 X ; q 21 X 2
A3 X 2 ; q 22 2
A2 X
2m
2I
h 2I
h

Expanding determinant H to the real and imaginary parts, we have:


Real part: ( p11 q 21 p12 q 22 p 21 q12 ) 0

(A2.19)

Imaginary part: ( p11 q 22 p12 q 21 p 21 q11 ) 0

(A2.20)

With the real part (A2.19), we have following equation:

B 4 * 2 2
2

( X 1) X
A3 X 2
2I
h

B 4 * 2 B 2 * 2
(
A1 X )(
H2 X ) 0
2I
2m
2


B 4 * 2
B 2 * 2
2 h X

H 1 X (2 h

A2 X )
2m
h
2I

Developing and grouping by X, we have 1:

8 | Le Thai Hoa Flutter stability analysis: Theory &Example

B 4 * B 2 B 4 * * B 2 B 4 * *
1 X [1
A3
A2 H 1
A1 H 2 ]
2I
m 4I
m 4I
4

B 2 *
2

B 4 *
B 4 *
2
+ X [2
.
H 1 2 h
A2 ] + X [ 2 4 h
1
A3 ]
h 2m
2I
h
2I
h
3

+(

2
)
h

=0

(A2.21)

Similarly developing to the imaginary part (A2.20), we have 2:

2 X 3 [

B 4 * B 2 * B 2 B 4
B 2 B 4 * *
A2
H1
.
H 1* A3*
.
A1 H 3 ]
2I
2m
m
4I
m 4I

B 4 *
B 2 * 2 B 4 *
+ X [ 2
2 h 2 h
A3 ] + X [
H1 2
A2 ] +
h
2I
2m
2I
h
2

+ [2 h

2
2 ]
2
h
h

=0

(A2.22)

As a result, the flutter motion differential equations of 2DOF heaving- torsional


system have been transformed to two polynomical equations with -variable (criticalstate circular frequency or flutter frequency). Flow chart of critical wind velocity
determination by complex eigenvalue problem for 2DOF heaving and torisional
motion system can be shown in underneath figure.

9 | Le Thai Hoa Flutter stability analysis: Theory &Example

4. STEP-BY-STEP PROBLEM FOR 2DOF HEAVING-TORSIONAL MOTION

EQUATION SYSTEM
For solving 2DOF heaving-torsional motion equations, there are two powerful
analytical methods: so-called the complex eigenvalue method [Simui&Scanlan(1976)]
and the step-by-step method [Matsumoto(1995)]. 2DOF heaving-torsional motion
system has be usually taken cases of unit structural length subjected to unit selfcontrolled forces into consideration. The 2DOF heaving-torsional motion systems,
moreover, can be known in sectional model tests in wind tunnels.
Complex Eigenvalue Method
Flutter Analytical

methods

2DOF Flutter Problems


Step-by-step Method

Fig 1. The scheme for analytical methods of 2DOF heaving-torsional flutter problems

(1) Complex eigenvalue method [Simiu&Scanlan(1976)]


The complex eigenvalue analytical method is based on some principles: i) Using some
techniques as transform of time-dimensionless variable, finding solutions under
harmonic manner; ii) Transform differential equations into linear equation system
with consistent condition its determinant must be zero; iii) Expanding determinant
and grouping by real and imaginary parts that must be simultaneously zero; iv)
Crossing point of solutions curves to determine the critical state of flutter instability.

(2) Step-by-step method [Matsumoto(1995)]


In principle, the step-by-step method is based on the serial solving technique of two
heaving-torsional motion equations, solutions of the former equation are to
determine coupled aerodynamic forces subjected to the later equation. From
10 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

transformation process, there is torsional-branch or heaving-branch step-by-step


method. Because torsional-branch instability dominates in almost cases, thus
torsional-branch step-by-step analysis will be favorable to be much more applicable
in comparision with heaving-branch one.

Stepwise procedure for torsional-branch analysis can be briefly presented hereinafter


i) The heaving motion equation will be taken into first account in which torsionalrelated coupled forces are considered as external oscillation, furthermore heaving
motion solutions are found dependant on torsional vibration parameters;

ii)

Obtained heaving motion solutions will be transformed into torsional motion


equation, then its damping ratio (or logarithmic decrement) will be determined in this
torsional-branch; iii) Checking such a damping ratio based on increment of reduced
wind velocity to obtain a critical condition for torsional-branch flutter instability.

Though the complex eigenvalue method has been applied for a long in solving 2DOF
heaving-torsional motion system to determine certain critical wind velocity, but
difficulty to investigate relationship of system damping ratio, system frequency on
wind velocity, and inter-relation between flutter derivatives as well. The step-by-step
method is favorable to deal with the complex eigenvalue methods limitation.

DOF heaving-torsional flutter equations


The flutter motion equations of 2DOF heaving-torsional system can be written as
follow:

m C K Lse
I C K M

(3.1)

se

Where: m, C , K are mass, damping coefficient and stiffness, respectively


associated with heaving motion.
11 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

I , C , K are mass inetia moment, damping coefficient and stiffness,

respectively associated with torsional motion.


Lse, Mse are self-controlled lift and moment.

The self-controlled forces Lse, Mse can be determined by either of Theodorsens


circulation function or Scanlans flutter derivatives under frequency approach. The
Scanlans self-controlled forces have been applied for the flutter motion equations for
various types of cross sections thank to experimentally-determined flutter
derivatives. Under this approach, the self-controlled forces per unit span length can
be modeled as follow:

L se (2b)U 2 kH 1* (k ) kH 2* (k )
k 2 H 3* (k ) k 2 H 4* (k )
2
U
U
b

(3.2.a)

M se (2b 2 )U 2 kA1* (k ) kA2* (k )


k 2 A3* (k ) k 2 H 4* (k )
2
U
U
b

(3.2.b)

Where: H 1* , H 4* , A2* , A3* : uncoupled derivatives


H 2* , H 3* , A1* , A4* : coupled derivatives

k is reduced frequency, k

b
U

Above equations can be rewritten under standard form as follow:

2 2

(2b)U 2 kH 1* (k ) kH 2* (k ) k 2 H 3* (k ) k 2 H 4* (k )
2m
U
U
b

2 2 (2b 2 )U 2 kA1* (k ) kA2* (k ) k 2 A3* (k ) k 2 H 4* (k )


2I
U
U
b

(3.3.a); (3.3.b)
Where:

12 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

, are free damping ratio and free circular frequency of heaving


motion, respectively

, are free damping ratio and free circular frequency of torsional


motion, respectively

; 2

m
C

2 mK

K
I

C
2 IK

The step-by-step analysis method


Step 1: Free vibration parameters of single heaving and torsional motions
Free vibration parameters will be determined by free vibration equations:

m C K 0
I C K 0

(3.4)

Writing free vibration equations above under standard form:

2 2 0

(3.5)

2 2 0
Free vibration parameters are obtained as following

; 2

m
C

2 mK

K
I

(3.6)

2 IK

Step 2: Solving the heaving motion equation in relation of coupled forces


Heaving motion equation can be written under expanded form as follow:

2 2 [

b 2
b 2 2 *
b 3
b 3 2 *
F H 1*
F H 4 ] [
F H 2*
F H 3 ]
m
m
m
m

13 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

[ 2

b 2
b 2 2 *
b 3
b 3 2 *
*
2
*

F H 1 ] [
F H 4 ] [
F H 2
F H 3 ] (3.7)
m
m
m
m

Rewriting eq(3.7) in the standard form:


2

2 ** * [

b 3
b 3 2 *
F H 2*
F H 3 ]
m
m

(3.8)

Where:

* , * are system circular frequency and system damping ratio of heaving


motion, respectively.
2

* [2

b 2 2 *
F H 4 ]
m

b 2
F H 1*
m
*
b 2 2 *
2 [2
F H 4 ]
m

(3.9)

(3.10)

Then, in order to transform the tortional-coordinate-related coupled forces in the


right-hand side to be pure external forces, technique for replacing the function can be
applied. Torsional displacement can be written under sinusoidal functional form:

sin t
cos t sin( t 900 )

(3.11)

Replacing the in to the coupled forces in left-hand side, we have:


2

2 ** *

b 3
b 3 2 *
F H 2* sin( t 90 0 )
F H 3 sin t (3.12)
m
m

Solution of eq(3.12) consists of such following components:

0 1 2

(3.13)

Where:

0 is total solution of free vibration equation:

14 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

2 ** * 0

(3.14)

1 is () solution of forced vibration equation:

b 3
F H 2* sin( t 90 0 )
m

2 ** *

(3.15)

2 is () solution of equation:
2

2 ** *

b 3 2 *
F H 3 sin t
m

(3.16)

Step 3: Finding solutions 1 , 2 of heaving forced vibration equation


2

(i) Finding 0 -solution: 1 2 **1 * 1 0


We find 0 -solution under such a form: 0 0 e

*t

sin * t

(3.17)

However, because system is motionless at initial time, thus solution of free vibration
is eliminated.
2

(ii) Finding 1 -solution: 1 2 **1 * 1

b 3
F H 2* sin( t 90 0 )
m

We find 1 -solution under such a form:

1 1 sin t 90 0

(3.18)

b 3
F H 2*
m
2

(* 2 ) 2 4 * * 2

tan (

2 **
2

* 2

*2

b 3
F H 2*
m
2 2
2
*2
(1 2 ) 4 ( 2 )
*
*

(3.19)

(3.20)

For convenience, we rewrite as follow:

15 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

1
*2

b 3
F | H 2* |
m
2
2
2
(1 2 ) 2 4 * ( 2 )
*
*

(3.21)

1 1 sin t 1 in cases of

a) 1 90 0 when H 2* 0

(3.22)

b) 1 90 0 when H 2* 0
2

(iii) Finding 2 -solution: 2 2 ** 2 * 2

b 3 2 *
F H 3 sin t
m

We also find 2 -solution under such a form:

2 2 sin t 2

(3.23)

b 3 2
F H 3*
m
2

(* 2 ) 2 4 * * 2

tan 1 (

2 **
2

* 2

*2

b 3 2
F H 3*
m
2 2
2
*2
(1 2 ) 4 ( 2 )
*
*

(3.24)

For convenience, we rewrite as follow:

2
*

b 3 2
F | H 3* |
m
2 2
2
*2
(1 2 ) 4 ( 2 )
*
*

2 2 sin t 2 in cases of

(3.25)

a) 2 when H 3* 0

(3.26)

b) 2 180 0 when H 3* 0

Thus, solution of heaving motion equation will be expressed:

1 2 1 sin( t 1 ) 2 sin( t 2 )

16 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

1 2 1 cos( t 1 ) 2 cos( t 2 )
Expanding , and noting that sin t

and cos t
, we have:

1 2 1 sin( t 1 ) 2 sin( t 2 )
1 sin t cos 1 1 sin 1 cos t 2 sin t cos 2 2 sin 2 cos t

1 cos 1 1
sin 1 2 cos 2 2
sin 2

(3.27)

1 2 1 cos( t 1 ) 2 cos( t 2 )
1 cos t cos 1 1 sin 1 sin t 2 cos t cos 2 2 sin 2 sin t

1
cos 1 1 sin 1 2
cos 2 2 sin 2

1 cos 1 1 sin 1 2 cos 2 2 sin 2

(3.28)

Step 4: Solving torsional motion equation


We have the torsional motion equation:
4
4
4
4
2 2 [ b A* b 2 A* ] [ b A* b 2 A* ] (3.29)

F 2
F 3
F 1
F 4
I
I
I
I

Expanding the heaving-oriented forced excitation in right-hand side:

b 3
b 3 2 *
F A1*
F A4
I
I
b 3

F A1* [ 1 cos 1 1 sin 1 2 cos 2 2 sin 2 ]


I

b 3 2 *

F A4 [ 1 cos 1 1
sin 1 2 cos 2 2
sin 2 ]
I

17 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

b 3 b 3
2
(
)(
) F

2
I
m
( A1* H 2* cos 1 A1* H 2* sin 1 F A1* H 3* cos 2
2

2
2
(1 2 ) 2 4 * ( 2 )
*
*

sin 2 )

b 3 b 3
2
(
)(
) F

I
m
( F A4* H 2* cos1 F A4* H 2*
sin 1 F2 A4* H 3* cos 2
2

2
2
(1 2 ) 2 4 * ( 2 )
*
*

A1* H 3*

F2 A4* H 3*

sin 2 )

b 3 b 3 F2
)(
)(
)
I
m * 2

(1

2
*

) 4

*2

2
*

[( A1* H 2* cos 1 F A1* H 3* cos 2 F A4* H 2* sin 1


)

F2 * *
A4 H 3 sin 2 )

(2 A1* H 2* sin 1 F A1* H 3* sin 2 F A4* H 2* cos 1 F2 A4* H 3* cos 2 ) ]

(3.30)
Replacing (3.30) in to eq(3.29), furthermore noting that in a torsional-branch
instability, the flutter frequency can be approximated to be torsional frequency, it
means that F

(3.31)
4

b
b 2 *
2 2 [
F A2*
F A3 ]
I
I
(

b 3 b 3 F2
)(
)(
)
I
m * 2

(1

2
F
*2

) 2 4 * (

2
F
*2

[( F A1* H 2* cos 1 F A1* H 3* cos 2 F A4* H 2* sin 1 F A4* H 3* sin 2 )


)

( F2 A1* H 2* sin 1 F2 A1* H 3* sin 2 2 F A4* H 2* cos 1 F2 A4* H 3* cos 2 ) ]

(3.32)

18 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Equation (3.32) can be rewritten under standard form:


2
2 * * 0

(3.33)

Where:

1 b 4
(
) F A2*
2 I
*

1 b 4 b 2 F2
(
)(
)(
) F
2 I
m * 2
2
F
*2

(1

) 2 4 * (

2
F
*2

( A1* H 2* cos 1 A1* H 3* cos 2


)

A4* H 2* sin 1 A4* H 3* sin 2 )


(3.34)

(
*2

b 4 2 *
(
) F A3
I

b 4 b 2 F2 2
)(
)(
) F
I
m * 2
2
F
*2

(1

) 2 4 * (

2
F
*2

( A1* H 2* sin 1 A1* H 3* sin 2


)

A4* H 2* cos 1 A4* H 3* cos 2 )


(3.35)
Solution of eq(3.33) can be expressed in such form: * e
2

in which: * * *

*t

sin(*t )

Step 5: Finding the critical condition of torsional instability


Flutter instability occurs if only if damping ratio

1 b 4
* (
) F A2*
2 I

1 b 4 b 2 F2
(
)(
)(
) F
2 I
m * 2
(1

2
F
*2

) 2 4 * (

2
F
*2

* 0

( A1* H 2* cos 1 A1* H 3* cos 2


)

A4* H 2* sin 1 A4* H 3* sin 2 ) 0

19 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

(3.35)
Logarithmic decrement (Log. dec) * 2 *

(
* (

b 4
) F A2*
I

b 4 b 2 F2
)(
)(
) F
I
m * 2

(1

2
F
*2

) 2 4 * (

2
F
*2

( A1* H 2* cos 1 A1* H 3* cos 2


)

A4* H 2* sin 1 A4* H 3* sin 2 ) 0


1 = (

b 4
)
I
(

2 =

(1

b 2 F2
)(
)
m * 2

F2
*

) 4 (

* 1 F A2* 1

F2
*

2 F ( A1* H 2* cos1 A1* H 3* cos 2 A4* H 2* sin 1 A4* H 3* sin 2 ) 0

We have: F , thus flutter condition can be rewritten as follow:


* 1 A2* 1

2 ( A1* H 2* cos 1 A1* H 3* cos 2 A4* H 2* sin 1 A4* H 3* sin 2 ) 0

Analytical procedure of step-by-step method


(1) Structural and dynamic parameters
- Structural parameters: b (

B
), K , K
2

- Air density:
- Dynamic parameters: m, I, C , C
(2) Flutter derivatives vs. reduced velocity U re
- Heaving derivatives:

U
fFb

H 1* , H 2* , H 3* , H 4*

20 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

- Torsional derivatives: A1* , A2* , A3* , A4*


(3) Free vibration characteristics
- Heaving motion:

- Torsional motion: 2

K
m
K
I

C
2 mK
C
2 IK

(4) Heaving-motion free vibration characteristics with uncoupled lift forces

b 2 2 *
[
F H 4 ]
m
*2

b 2
2
F H 1*
m
*
b 2 2 *
2 [2
F H 4 ]
m
(5) Initial phase angle

tan 1 (

2 **
2

* 2

case H 2* 0 then 1 90 0

else 1 90 0

case H 3* 0 then 2 180 0 else 2


(6) Torsional-branch circular frequency
2

* 2 1 2 F A3* 1 2 F2 ( A1* H 2* sin 1 A1* H 3* sin 2 A4* H 2* cos 1 A4* H 3* cos 2 )


(7) Checking of torsional-branch log. dec
* 1 A2* 1

2 ( A1* H 2* cos 1 A1* H 3* cos 2 A4* H 2* sin 1 A4* H 3* sin 2 ) 0

21 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Structural and dynamic input

b, , m, I , K , K , C , C
Free vibration parameters

, , ,
Wind velocity loop
U i (Zero first approimation)
Circular frequency loop
F , j (First

U i 1 U i U

H i* , Ai* , * , * ,

F , j 1 F , j

Frequency checking

* F , j
Log. Dec. checking

j 0
End

22 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

5.

SINGLE-MODE,

TWO-MODE

AND

MULTI-MODE

FLUTTER

PROBLEMS OF NDOF SYSTEMS


As mentioned above, the conventional complex eigen method [Simui&Scanlan(1976)]
and the Step-by-step method [Matsumoto(1995)] are very powerful to solve for 1DOF
torsional flutter equation and 2DOF tortionnal-heaving equations of 2-dimensional
structures, some analytical methods have been developed for solving nDOF flutter
equations of 3-dimensional structures.

As first, happened flutter possibilities for bridge structures will be reviewed for
explanation of numerical analytical developments of nDOF flutter problems. By
various experiments and numerical analyses from practical applications of bridge
engineering, it is shown that the fundamental torsional vibration mode dominantly
involves to the flutter instability. Moreover, with bluff cross sections like low
slenderness ratio (B/D) rectangular sections or H-shape sections or stiffened truss
sections, the flutter instability almost occur in solely fundamental torsional mode
[Matsumoto (1996)] as known the torsional flutter as the case of Tacoma Narrow
failure. Whereas the fundamental torsional mode and any first symmetric or
asymmetric heaving mode usually couple mechanically at single frequency with the
streamlined cross sections as known as the coupled flutter or the classical flutter
(studied previously on aerodynamics of airplanes airfoil wings).

It is very

interesting, however, by both analyses and experiments to mark that coupled flutter
has occurred in case of the Akashi-Kaikyo bridge, that has been never seen before
with such kinds of the stiffen truss-girder cable-supported bridges [Katsuchi(1998)]. It
is questionable from case of the Akashi-Kaikyo bridge, thus, that coupled flutter also
possibly happens to very flexible long-span cable-supported bridges with bluffsections.

23 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Some recent analytical studies [Scanlan(1990), Pleif(1995), Jain(1996), Katsuchi(1998)],


furthermore, pointed out that in many cases of coupled flutter there are not the
fundamental torsional and heaving modes involved to the critical condition, but
many modes (multi-modes) superpose to gain more critical condition at a lower onset
velocity.
Some analytical methods for flutter problems have been developed from above
happened

possibilities

as

single-mode

[Simui&Scanlan(1976)],

two-mode

[Scanlan(1981)], multi-mode methods [Scanlan(1990), Jain(1996), Katsuchi(1998),


Ge(2000)]. In principle, above-mentioned analytical methods are carried out on the
modal space thank to generalized coordinate transform and modal superposition
technique.
Bluff cross-sections
Low B/D or H/- sections
Stiffen truss sections

Torsional Flutter
Heaving Flutter

Single-mode Method

nDOF Flutter problems


Two-mode Method
Coupled Flutter
Multi-mode Method
Airfoil
Thin plate sections
Streamlined sections

Diagram for analytical analysis methods of nDOF flutter problems

It is suggested that the single-mode method can be applied for cases of torsional
flutter possibly happens, whereas, the two-mode method for simplified approach and

24 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

the multi-mode method for more accuracy should be applied for tendency cases of
coupled flutter.

NDOF flutter motion equations


The motion equations of the nDOF structural systems in the steady flow can be
expressed under the Finite Element Method (FEM) as follow:
MU CU KU P (t )

(A3.1)

Where: P(t) is the self-controlled flutter forces subjected to structure (However,


noting that the self-controlled flutter forces above are only valid in cases of
steady wind flows, whereas unsteady buffeting forces must be associated with
in self-controlled forces in the unsteady wind flows).
The self-controlled aerodynamic forces per unit length of bridge deck can are
popularly formed thank to the Scanlans experimentally-determined flutter
derivatives:

L se BU 2 KH 1* ( K ) KH 2* ( K )
K 2 H 3* ( K ) K 2 H 4* ( K )
2
U
U
B

1
p
B
p
D se BU 2 KP1* ( K ) KP2* ( K )
K 2 P3* ( K ) K 2 P4* ( K )
2
U
U
B

(A3.2)

M se B 2U 2 KA1* ( K ) KA2* ( K )
K 2 A3* ( K ) K 2 H 4* ( K )
2
U
U
B

Where: H 1* , H 4* , A2* , A3* , P1* , P4* : uncoupled derivatives


H 2* , H 3* , A1* , A4* , P2* , P3* : coupled derivatives

25 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Above self-controlled aerodynamic forces can be explicitly divided by the


displacement-dependant aerodynamic elastic force-component and first-order
derivative-dependant aerodynamic damping one, we have:
P (t ) P1 (t ) P2 (t ) P1U P2U

(A3.3)

Thus, the nDOF flutter motion equations are written hereafter:


MU CU KU P1U P2U
MU [C P1 ]U [ K P2 ]U 0
MU C *U K *U 0
Where:

(A3.4)

C * C P1 ; K * K P2

C*, K* are the system damping-force and elastic-force matrices, respectively.


Because above matrices have no longer symmetrical, thus eigenvalues of
frequency equation of eq.(A3.4) must be conjugate complex pairs.

Transforming to the modal space and generalized coordinates


The motion equations can be transformed from the ordinary coordinates into the
principle ones (generalized coordinates) with the generalized coordinates are defined
as follow: U

(A3.5)

Where: is the generalized coordinates


is the mass-normalized eigenmodal matrix

By using the mass-normalized technique, we transform

M C * K * 0
*
*
I C K 0
*

(A3.6)
*

T
*
T
*
Where: C C ; K K

26 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

t
Solution of eq.(A3.) found under such form: e

(A3.7)

Expending and transforming eq.(A3.6) in the frequency equation (the


characteristic equation):

I C
2

Det

(A3.8)

2n eigenvalues and eigenvectors can be determined through above equation.


Because the system damping and elastic matrices have no longer to be symmetrical,
thus eigenvalues from eq.(A3.8) will be exhibited under the conjugate complex
eigenvalue pairs:

i i j i

(A3.9)

Global response of bridge in the generalized coordinates can be obtained by


superposing of combined modal responses as form:
2n

i e i t

(A3.10)

i 1

Where:

i is the modal scaling factor

n is the number of modes combined in global response


i is the eigenvector of ith mode, also presented under

conjugate

complex eigenvector as i pi jqi

Global response of bridge in the generalized coordinates can be rewritten hereby:


2n

j i ) p i jq j e ( i

j i ) t

i j i p i jq i e ( i j i ) t

i 1

27 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

2n

e i t 2 i p i i q i sin i t 2 i q i i p i cos i t

i 1

(A3.11)
Global response amplitude of bridge in the conventional coordinates can be expressed
as follow:
2N

{U } {i }e i t [2({ i }{ pi } { i }{qi }) sin it 2({ i }{qi } { i }{ pi }) cos i t ]


i 1

From eq.(A3.11), it can be seen clearly the role of the real part i of complex
eigenvalues in the system stability and instability problem, when real part of complex
eigenvalue become positive, system response amplitude is to be divergent and flutter
instability occurs (known as the Liapunovs Theorem).

Linearly-discretized technique of the self-controlled aerodynamic forces


Uniform aerodynamic forces are linearly lumped at deck nodes. Six nodal
displacements and their first derivatives can be expressed in to element coordinates
as follows: {U } {0 h p 0 0}T and {U } {0 h p 0 0}T .

Element coordinates and nodal displacements

28 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

The self-controlled aerodynamic forces along bridge deck can be linearly discretized
at any bridge deck node:

Diagram for nodal linear-discretization of self-controlled forces

From this linearly-discretized technique, finite-element damping and elastic


aerodynamic force matrices P1, P2 (12x12) can be easily obtained:
0
0
0
H 1*

0
1
K 0
P 1 U 2 B L
4
U 0 BA1*
0
0

0
0
0
0

0
1
P 2 U 2 BK 2 L
4
0
0

0
0
0
0
0
0

0
0
P1*
0
0
0

0
0
0 H 3*
0 P3*
0 BA3*
0
0
0
0

0
BH 2*
BP2*
B 2 A2*
0
0
0
0
0
0
0
0

0
0
0

0
0

0
0
0
0
0
0

0
0
0

0
0

(A3.12)

29 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Noting that the matrices P 1 , P 2 (6x6) above are only presented at single node of
element, and element force matrices P1, P2 (12x12) will be built symmetrically from
above P 1 , P 2 (6x6).

Multi-mode flutter analysis


The quadratic eigenvalue problem will be difficult, in order to transform the
frequency equation (A3.8) into the standardized eigenvalue problem, motion
equation (A3.6) will be written in the state-space as follow:
I 0 0

*
C * 0 K 0

0
I

(A3.13)

We replace hereby:
0 I
A I C * ;

I 0
B 0 K *

(A3.14)

t
t
Y
e ; Y
e


We will have:

A Y B Y


A
B

B Z A Z
Here

(A3.15)

Expanding from eq.(A3.15), we have:

A1 B Z Z

30 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

C * K *


I
0

Z Z

C * K *
Replacing: D

I
0

As

result,

the

standardized

(A3.16)

eigenvalue

problem

D Z Z

has

been

achieved:
(A3.17)

The standardized eigenvalue problem above can be solved by the many solving
techniques such as Jacobi diagonalization, QL or QR transformation, subspace
iteration and another.

In general, the multi-mode method has been still based on prior selection of concrete
modes in combination. Recently, it can be automatically combined total modes from
free vibration analysis for flutter analysis, so-called full-mode method [Ge(2002)],
however, this full-mode method dont pay much more accuracy out of control than
multi-mode method but time-consuming.

Single-mode flutter analysis in the modal space


The nDOF flutter motion equations can be written in the modal space and
generalized coordinate under such mass-normalized form:
I C K T P1 T P2

(A3.18)
T

T
T
Where: I M ; C C ; K K

I is the mass-normalized unit matrix

C is the diagonal normalized damping matrix

31 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

(containing modal damping coefficients)

K is the diagonal normalized elastic matrix


(containing modal eigenvalues)

Flutter motion equation of ith mode in the generalized coordinates can be written after
the normalized technique: i 2 i i i i2 i pi (t ) (A3.19)
Where: pi(t) is the self-controlled aerodynamic force of ith mode (or called as
the normalized generalized aerodynamic force) determined as follow:
pi (t ) i P1 i i P2 i
T

i (x) or j (x)

and i = p i (x) or p j (x)

i (x) or j (x)

(A3.20)

(A3.21)

x is an deck-alongside coordinate
i, j are an index for combination between two modes

Expanding (A3.20) with noting that aerodynamic matrices P1, P2 determined by


eq.(A3.12) and grouping in generalized coordinates ( ) and their first-order
derivatives

( ), the normalized generalized aerodynamic force can be obtained

below:
1
BK *
pi (t ) U 2
[ H1 Ghi h j BH 2* Ghi j P1* G pi p j BP2* G pi h j BA1*G i h j B 2 A2* G i j ]i
2
U

1
U 2 BK 2 [ H 3* Ghi j P3* G pi j BA3* G i j ] i
2

(A3.22)

Where: Grmsn is the modal integral sums determined by such formula in


discretized manner of mode shapes.

32 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Grmsn =

(r,k)m (r,k)n

(A3.23)

k 1

Omitting cross-modal integral sums Grmsn (rs) due to their small, remaining automodal integral ones Grmsn (r=s), we easily obtain:
1
BK
1
pi (t ) U 2
[ H 1* Ghi h j P1* G pi p j B 2 A2* G i j ]i U 2 BK 2 [ BA3* G i j ] i (A3.24)
2
U
2

Putting eq.(A3.24) in eq.(A3.19), transforming into an ordinary 1DOF differential


equation:
1
1
i [2 i i U 2 ( H 1*Ghi h j P1*G pi p j B 2 A2*G i j )]i [ i2 U 2 BK 2 ( BA3*G i j )] i 0
2
2

i 2 i ii i i 0

(A3.25)

Where:

i2

B 4 *
1
A3 ( K i )G i j
2

(A3.26)

i =

i i B 4 *

[H 1 (K i ) G hihj P1* ( K i )G pipj B 2 A *2 ( K i )G ij (A3.27)


4
i

Ki

B i
U

(A3.28)

Two-mode flutter analysis in the modal space


In cases of two-mode coupled flutter, two modes as common sense are engaged: one is
dominant heaving mode and other is dominant torsional mode. It might be postulated
that two modes can couple if they have similarly modal shapes and closely natural
frequencies, and cross-modal integral sums in these cases play more important role.

33 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Generally, the two-mode flutter method has developed from problems of nDOF systems
single-mode analysis and of 2DOF systems complex eigen analysis. The two motion
equations of ith and jth modes with the coupled normalized generalized aerodynamic
forces can be expressed following:
i)

ith modal motion equation

i 2 ii i 2ii
1
BK *
U 2
[ H1 Ghihi BH 2*Ghi i P1*G pi pi BP2*G pihi BA1*Gi hi B 2 A2*G ii ]i
2
U
1
U 2 K 2 [ H 4*Ghihi BH 3*Ghii P4*G pi pi BP3*G pii BA4*G ihi B 2 A3*Gi i ]i
2
+

1
BK *
U 2
[ H1 Ghi h j BH 2*Ghi j P1*G pi p j BP2*G pi h j BA1*G i h j B 2 A2*G i j ] j
2
U

1
U 2 K 2[ H 4*Ghi h j BH 3*Ghi j P4*G pi p j BP3*G pi j BA4*G i h j B 2 A3*G i j ] j
2

ii)

(A 3.29a)

jth modal motion equation

j 2 j j j 2 j j
1
BK *
U 2
[ H1 Gh j h j BH 2*Gh j j P1*G p j p j BP2*G p j h j BA1*G j h j B 2 A2*G j j ]j
2
U
1
U 2 K 2[ H 4*Gh j h j BH 3*Gh j j P4*G p j p j BP3*G p j j BA4*G j h j B 2 A3*G j j ] j
2
1
BK *
U 2
[ H1 Gh j hi BH 2*Gh j i P1*G p j pi BP2*G p j hi BA1*G j hi B 2 A2*G ji ]i
2
U
1
U 2 K 2 [ H 4*Gh jhi BH 3*Gh ji P4*G p j pi BP3*G p j i BA4*G jhi B 2 A3*G ji ] i
2

(A 3.29b)

Solution for two modal motion equations under coupled forces can be carried out by
following steps:

34 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Step 1: Solutions assumed under such forms as i i 0 ei F t , j j 0 ei F t . Then expanding


and grouping into equation system under terms of i 0 , j 0 . We can obtain 2 equations:

[(

i 2

) 1 2i i ( i )] i
F
T

1 2
B {i[ H1*Ghihi BH 2*Ghi i P1*G pi pi BP2*G pihi BA1*G ihi B 2 A2*G ii ]
2
[ H 4*Ghi hi BH 3*Ghi i P4*G pi pi BP3*G pi i BA4*Gihi B 2 A3*Gi i ]} i 0

1 2
B {i[ H1*Ghi h j BH 2*Ghi j P1*G pi p j BP2*G pi h j BA1*G i h j B 2 A2*G i j ]
2

(A 3.30a)

[ H 4*Ghih j BH 3*Ghi j P4*G pi p j BP3*G pi j BA4*G ih j B 2 A3*G i j ]} j 0


and

[(

j 2

) 1 2i j ( j )] j
F
T

1 2
B {i[ H1*Gh j h j BH 2*Gh j j P1*G p j p j BP2*G p j h j BA1*G j h j B 2 A2*G j j ]
2

[ H 4*Gh j h j BH 3*Gh j j P4*G p j p j BP3*G p j j BA4*G j h j B 2 A3*G j i ]} j 0


1
B 2{i[ H1*Gh j hj B2*Gh j j P1*G p j pi BP2*G p j hi BA1*G jhi B 2 A2*G ji ]
2

(A 3.30b)

[ H 4*Gh j hi BH 3*Gh j j P4*G p j pi BP3*G p j i BA4*G j hi B 2 A3*G j i ]}i 0

Step 2: Grouping by i 0 , j 0 , obtaining two-equation system, conditioning nontrivial


solutions that their determinant must be zero. We can write the determinant under such a
form:

Det

A11 iB11

A12 iB12

A21 iB21

A22 iB22

(A 3.31)

35 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Step 3: Expanding the determinant, two equations of real and imaginary parts can be
obtained and must be simultaneously zero, we have:
Real part:

Aii A jj Bii B jj A ji Aij B ji Bij 0

Imaginary part: Aii B jj Bii A jj A ji Bij B ji Aij 0

(A 3.32a)
(A 3.32b)

Where:

Aii (i / F )2 1 1 / 2( B 2 )[H 4*Ghi hi BH 3*Ghi i P4*G pi p i BP3*G p i i BA4*G i hi B 2 A3*G i i ]

Aij 1 / 2( B 2 )[ H 4*Ghi h j BH 3*Ghi j P4*G pi p j BP3*G pi j BA4*G i h j B 2 A3*G i j ]


A jj ( j / F ) 2 1 1 / 2( B 2 )[ H 4*Gh j h j BH 3*Gh j j P4*G p j p j BP3*G p j j BA4*G j h j B 2 A3*G j j ]
A ji 1 / 2( B 2 )[ H 4*Gh jhi BH 3*Gh j j P4*G p j pi BP3*G p j i BA4*G jhi B 2 A3*G j i ]
Bii 2 i (i / F ) 1 / 2( B 2 )[ H1*Ghihi BH 2*Ghii P1*G pi pi BP2*G pihi BA1*G ihi B 2 A2*G ii ]

Bij 1 / 2( B 2 )[ H1*Ghi h j BH 2*Ghi j P1*G pi p j BP2*G pi h j BA1*G i h j B 2 A2*G i j ]


B jj 2 j ( j / F ) 1 / 2( B 2 )[ H1*Gh jh j BH 2*Gh j j P1*G p j p j BP2*G p jh j BA1*G jh j B 2 A2*G j j ]
B ji 1 / 2( B 2 )[ H1*Gh j hi BH 2*Gh j i P1*G p j pi BP2*G p j hi BA1*G j hi B 2 A2*G j i ]

Step 4: Solutions of Eq.(A 3.32a), Eq.(A 3.32b) are found simultaneously, intersected point
of solution curves determine the critical flutter condition.

36 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

6. NUMERICAL EXAMPLE OF A CABLE-STAYED BRIDGE


Numerical example of a cable-stayed bridge for the flutter analysis will be presented
under three approaches:
i.

Complex-eigen analysis for a 2DOF torsional-heaving system (first


torsional and heaving modes selected for the analysis)

ii.

Step-by-step analysis for a 2DOF torsional-heaving system ( also first


torsional and heaving modes selected for the analysis)

iii.

Single-mode and multi-mode analysis for the cable-stayed bridge

Fig A4.1. Cable-stayed bridge for numerical analysis example

37 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Structural characteristics and flutter derivatives


Table A4.1. Sectional characteristics of example cable-stayed bridge
Gider
Material parameters
E =3600000 T/m2
G =1384600 T/m2
=0.3 Poison ratio
Geometrical parameter
A =6.525 m2
I33 =0.11 m4
I22 =114.32 m4
J =0.44m4

Tower
Material parameters
E =3600000 T/m2
G =1384600 T/m2
=0.3 Poison ratio
Geometrical parameter
A =1.14 m2; I33=0.257 m4
I22 =0.118 m4;J=0.223m4
A =1.14 m2; I33=0.257 m4
I22 =0.118 m4;J =0.223m4

Stayed cables
Material parameters
E = 19500000 T/m2
Geometrical parameter
A =26.355 cm2 Type 19K15
A =16.69 cm2 Type 12K15

Flutter derivatives
20
15

Hi
(i=1,2,3)

10
5
0
0

10

11

12

-5
-1 0
-1 5
-2 0

Reduced velocity K

Flutter derivatives

A*i
(i=1,2,3)

3.5
3
2.5
2
1.5
1
0.5
0
-0.5 0

9 10 11 12

Reduced frequency K

Fig A4.2. Diagrams of the flutter derivatives H*i, A*i (i=1-3) given by
quasi-steady formula [Scanlan(1989), Pleif(1995)]
38 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Free vibration analysis


Mode 1
f=0.6099Hz

Mode 2
f=0.8016Hz

Mode 3
f=0.8522Hz

Mode 4
f= 1.1949Hz

Mode 5
f=1.2931Hz

Mode 7
f=1.5819Hz

Mode 6
f=1.4495Hz

Mode 8
f=1.6304Hz

39 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Mode 7
f=1.5819Hz

Mode 8
f=1.6304Hz

Fig A4.3. Fundamental 10 natural mode shapes

Free vibration characteristics and modal integrals


Table A4.1. Characteristics of the first 10 natural mode shapes
Mode Eigenvalue Frequency
shape

(Hz)

Period

Modal Character

(s)

1.47E+01

0.609913

1.639579

S-V-1

2.54E+01

0.801663

1.247406

A-V-2

2.87E+01

0.852593

1.172893

S-T-1

5.64E+01

1.194920

0.836876

A-T-2

6.60E+01

1.293130

0.773318

S-V-3

8.30E+01

1.449593

0.689849

A-V-4

9.88E+01

1.581915

0.632145

S-T-P-3

1.05E+02

1.630459

0.613324

S-V-5

1.12E+02

1.683362

0.594049

A-V-6

10

1.36E+02

1.857597

0.53830

S-V-7

Note :

S: Symmetric mode

V: Heaving mode shape

A: Asymmetric mode

T: Torsional mode shape


P: Horizontal mode shape

40 | L e T h a i H o a F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Dng dao ng th 2
( Dng un th 2)

Dng dao ng th nht


(Dng un th nht )
0,1

0,04
0,02

-0,06
-0,08

29

27

25

23

21

19

17

15

13

11

0
1

Gi tr dng

29

27

25

23

21

19

17

15

13

11

0
-0,02
-0,04

0,05

Gi tr dng

0,06

-0,05
-0,1

-0,1
-0,12

-0,15

Dng dao ng th 3
(Dng xon th nht )

Dng dao ng th 4
( dng xon th 2 )

0,02

0,015
0,01

28

25

22

19

16

-0,005

13

10

-0,01

28

25

22

19

16

13

10

0
7

0,005

0,005

Gi tr dng

0,01

Gi tr dng

0,015

-0,005
-0,01

-0,015

-0,015

-0,02

-0,02

Dng dao dng th 6


(dng un th 4)

Dng dao ng th 5
(dng un th 3)
0,06

0,1

0,04
0,05

25

27

29

27

29

21

19

17

15

13

11

25

- 0,08

23

- 0,06

23

- 0,04

- 0,02

Gi tr dng

29

27

25

23

21

19

17

15

13

11

0
1

Gi tr dng

0,02

-0,05

- 0,1

-0,1
- 0,12

-0,15

Dng dao ng th 8
(dng un th 4)

Dng dao ng th 7
(dng xon th 3)
1,00E-02

0,12

5,00E-03

0,08

-1,50E-02

0,04
0,02
21

19

17

15

13

11

0
-0,02

28

25

22

19

16

13

10

0,06

-1,00E-02

Gi tr dng

-5,00E-03

0,00E+00
1

Gi tr dng

0,1

-0,04
-0,06

-2,00E-02

-0,08

Fig. Modal amplitude of normalized mode shapes

41 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Table A4.2. Modal integral sums of first 10 natural mode shapes


Mode Frequency Modal
shape

(Hz)

Modal integral sums Grmsn

Character

Ghihi

Gpipi

Gii

0.609913

S-V-1

5.20E-01

7.50E-11

0.00E+00

0.801663 A-V-2

4.95E-01

7.43E-09

1.35E-09

0.852593

S-T-1

3.79E-09

5.23E-05

1.14E-02

1.194920 A-T-2

1.78E-07

1.82E-05

1.07E-02

1.293130

S-V-3

5.07E-01

1.36E-07

23.62E-09

1.449593 A-V-4

4.99E-01

2.10E-09

9.42E-09

1.581915 S-T-P-3

2.67E-07

1.10E-03

1.10E-02

1.630459

S-V-5

5.03E-01

1.43E-07

1.27E-08

1.683362 A-V-6

1.64E-06

1.77E-04

1.09E-02

10

1.857597

4.16E-06

2.78E-03

1.11E-02

S-V-7

Note: Modal integral sums: Grmsn =

(r,k)m (r,k)n

k 1

lk:
Discretized deck lengths
(r,k)n : Discretized modal values

42 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

2
X33
1.5
X44
1

X43

0.5

0
X32
-0.5
Intersection
X41
X31

-1

X 42
-1.5

4
5 5.3
Reduced Frequency K

FigA4.4.Diagram of complex eigen solutions of 2DOF torsional-heaving system (first


heaving mode + first torsional mode)

43 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

1.2
Mode
Mode
Mode
Mode
Mode

1
2
3
4
5

(Heaving)
(Heaving)
(Torsional)
(Torsional)
(Heaving)

System damping ratio

0.8

0.6

Mode 1

Mode 2

0.4

Mode 5

0.2
Mode 4

Mode 3
-0.2
10

20

30

40
50
60 64.5 70
Wind velocity (m/s)

80 88.5 90

Fig.A4.5.Diagram of wind velocity vs. system damping ratio (V-)

1.3

Aerodynamic interaction

Mode 3 (Torsional)
Mode 4 (Torsional)

1.2
Mode 3

Frequency (Hz)

1.1

1
Aerodynamic interaction
0.9

0.8
Mode 4
0.7

0.6
10

20

30

40
50
60
Wind velocity (m/s)

70

80

90

Fig.A4.6.Diagram of wind velocity vs. frequency (V-f)


44 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Time-history modal amplitude of first 5 modes at certain wind velocities


1

Mode 1

0
-1
10

10

20

30

40
50
Mode 2

60

70

80

90

-1
10

100

10

20

30

40
50
Mode 2

60

70

80

90

100

10

20

30

40
50
60
70
Mode 3 (Divergence)

80

90

100

10

20

30

40
50
Mode 4

60

70

80

90

100

10

20

30

40
50
Mode 5

60

70

80

90

100

10

20

30

40

50
60
Time (s)

70

80

90

100

-1
10

10

20

30

40
50
Mode 3

60

70

80

90

100

0
-1
10

10

20

30

40
50
Mode 4

60

70

80

90

100

M o d a l A m p lit u d e

M odal A m plitude

-1
20
0
-2
10
0

-1
10

10

20

30

40
50
Mode 5

60

70

80

90

-1
10

100

0
-1

Mode 1

0
0

10

20

30

40

50
Time (s)

60

70

80

90

-1

100

Modal amplitude of first 5 modes at U=50m/s Modal amplitude of first 5 modes at U=65m/s
1

0
-1
10

10

20

30

40
50
Mode 2

60

70

80

90

-1
10

100

-1
50

10

20

30

40
50
Mode 3

60

70

80

90

100

0
-5
10

10

20

30

40
50
Mode 4

60

70

80

90

100

M odal A m plitude

M odal Am plitude

10

20

30

40
50
Mode 2

60

70

80

90

100

-1 x 105
10
10

20

30
40
50
60
Mode 3 (Divergence)

70

80

90

100

10

20

30
40
50
60
Mode 4 (Divergence)

70

80

90

100

10

20

30

40
50
Mode 5

60

70

80

90

100

10

20

30

40

50
60
Time (s)

70

80

90

100

0
-1
20
0

0
-1
10

10

20

30

40
50
Mode 5

60

70

80

90

100

-2
10
0

0
-1

Mode 1

Mode 1

10

20

30

40

50
60
Time (s)

70

80

90

Modal amplitude of 5 modes at U=70m/s

100

-1

Modal amplitude of 5 modes at U=90m/s

Fig.A4.7.Diagrams of modal amplitudes of first 5 modes at some certain wind


velocity values
45 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Investigation on change of modal amplitude of some major modes following wind


velocities and time intervals

0.06
0.04

Modal amplitude

0.02
0
-0.02

11 13 15 17 19 21 23 25 27 29

-0.04

Initial

-0.06

50m/s
65m/s

-0.08

70m/s
90m/s

Decay

-0.1
-0.12

Deck nodes

Diagram of 1st heaving modal amplitude vs. wind velocity after 2 seconds

0.1

Modal amplitude

0.05

0
1

11

13 15

17 19

21 23

25

27 29

-0.05
Initial
50m/s

-0.1

65m/s

Decay
-0.15

70m/s
90m/s

Deck nodes

Diagram of 2nd heaving modal amplitude vs. wind velocity after 2 seconds

46 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

0.01

Modal amplitude

0.005
0
1

11 13 15 17 19 21 23 25 27 29

-0.005
Initial

-0.01

50m/s
65m/s

-0.015

70m/s

Divergence

90m/s

-0.02
-0.025
Deck nodes

Diagram of 1st torsional modal amplitude vs. wind velocity after 2 seconds

0.015

Modal amplitude

0.01
0.005
0
1

Divergence
11 13 15 17
19 21 23 25 27 29

-0.005
Initial

-0.01

50m/s
65m/s

-0.015

70m/s
90m/s

-0.02
Deck nodes

Diagram of 2nd torsional modal amplitude vs. wind velocity after 2 seconds

47 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

0.06

Initial

0.04

50m/s
65m/s

Modal amplitude

0.02

70m/s
90m/s

-0.02

11 13

15 17 19

21 23 25 27

29

-0.04
-0.06
-0.08

Decay

-0.1
-0.12

Deck nodes

Diagram of 3nd heaving modal amplitude vs. wind velocity after 2 seconds

0.06

Modal amplitude (at 50m/s)

0.04
0.02
0

-0.02

11

13

15

17

19

21

23

25

27

29

Initial

-0.04

1seco nd

-0.06

2seco nds

-0.08

3seco nds
5seco nds

-0.1

10seco nds

-0.12

Deck nodes

Diagram of 1st heaving modal amplitude at wind velocity 50m/s vs. time intervals

48 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

0.06
Modal amplitude (at 70m/s)

0.04
0.02
0

-0.02

11

13 15

17

19 21

23

25 27

29

Initial

-0.04

1second

-0.06

2seco nds

-0.08

3seco nds
5seco nds

-0.1

10seconds

-0.12
Deck nodes

Diagram of 1st heaving modal amplitude at wind velocity 70m/s vs. time intervals

Modal amplitude (at 50m/s)

0.01
0.005
0
1

11 13 15 17 19 21 23 25 27 29

-0.005
Initial

-0.01

1seco nd
2seco nds
3seco nds

-0.015

5seco nds
10seco nds

-0.02

Deck nodes

Diagram of 1st torsional modal amplitude at wind velocity 50m/s vs. time intervals

49 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

0.01

Modal amplitude (at 70m/s)

0.005
0
1

11

13

15

17

19

21

23

25

27

29

-0.005
-0.01
Initial
1seco nd
2seco nds

-0.015

3seco nds

-0.02
-0.025

5seco nds
10seco nds

Deck nodes

Diagram of 1st torsional modal amplitude at wind velocity 50m/s vs. time intervals

50 | L e T h a i H o a - F l u t t e r s t a b i l i t y a n a l y s i s : T h e o r y & E x a m p l e

Vous aimerez peut-être aussi