Journal of Wind Engineering

and Industrial Aerodynamics 91 (2003) 2740

Analytical study of windrain-induced cable
vibration: SDOF model
Y.L. Xu*, L.Y. Wang
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong, Peoples Republic of China
Abstract
This paper presents an analytical study aiming to explain and predict some windrain-
induced cable vibration phenomena observed from eld measurements and simulated wind
tunnel tests. Based on the knowledge of aerodynamics and structural dynamics and some
measurement results, an analytical model is developed for the concerned problem considering
the effect of mean wind speed on the position of upper rivulet and the inuence of moving
upper rivulet on cable vibration. The analytical model is then applied to some cable models
tested in either a wind tunnel with xed articial rivulets or a windrain tunnel with moving
rivulets. It turns out that the analytical model can capture main features of windrain-induced
cable vibration.
r 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Stay cable vibration; Wind; Rain; Moving rivulet; Fixed rivulet; Analytical model; Comparison
1. Introduction
There have been many reports recently on excessive and unanticipated vibration of
stay cables in cable-stayed bridges under the simultaneous occurrence of wind and
rain. Such large amplitude cable vibrations due to wind and rain often cause fatigue
damage and corrosion to strands in surprisingly short periods. Many studies have
been thus performed to nd out the reasons behind this type of vibration and the
measures for mitigating such a vibration.
Hikami and Shiraishi [1] carried out eld measurement of stay cables in the
Meikonishi Bridge subject to wind with and without rain. They found that the
*Corresponding author. Tel.: +852-2766-6050; fax: +852-2334-6389.
E-mail address: ceylxu@polyu.edu.hk (Y.L. Xu).
0167-6105/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 6 7 - 6 1 0 5 ( 0 2 ) 0 0 3 3 3 - 1
cables, which were of small amplitude vibration under wind without rain, had
excessive vibration under wind but with rain. They then conducted a series of wind
tunnel tests with rain condition simulated and reproduced windrain-induced cable
vibration in their model tests. They concluded that the rivulet formed along the
upper surface of a cable under the action of both wind and rain changed the original
cable cross section and resulted in excessive cable vibration. Matsumoto et al. [2]
further conducted extensive windrain-tunnel tests and pointed out that windrain-
induced cable vibration was due to two major factors: one was the formation of
upper rivulet and the other was the axial ow generated in the wake of the cable.
Bosdogianni and Olivari [3] compared wind tunnel results of cables with moving
rivulet with those of the same cables but with xed rivulet. They concluded that it
was the presence of rivulet at a certain position and not the motion of rivulet that
caused cable instability. However, Ruscheweyh [4] stated that the movement of
rivulet was the trigger for windrain-induced cable vibration and if the movement
of rivulet could be stopped, rainwind-induced cable vibration would stop.
Compared with the eld measurements and windrain tunnel model tests of
cables, analytical studies are very limited. Yamaguchi [5] presented a two-degree-of-
freedom galloping model to describe windrain-induced cable vibration. In his
model, the cable was modeled as a horizontal rigid cylinder. He concluded that one-
degree-of-freedom galloping theory might not be useful to explain the mechanism of
windrain-induced cable vibration. Geurts and Staalduinen [6], however, recently
presented an engineering approach to the same problem but based on the one-
degree-of-freedom galloping theory.
It is clear that windrain-induced cable vibration is a complicated soliduid-wind
interaction problem. Its mechanism is not well understood yet. This paper thus
presents a preliminary analytical study. An analytical model is rst developed for
describing the steady-state windrain-induced vibration, using the knowledge of
aerodynamics and structural dynamics and some measurement results and taking
into account the interaction between wind, rivulet and cable. The analytical model is
then applied to some cable models tested in either a wind tunnel with xed articial
rivulets or a windrain tunnel with moving rivulets. The mechanisms of windrain-
induced cable vibration are nally explored.
2. Analytical model
Let us use a rigid and uniform inclined cylinder to represent a stay cable segment
(see Fig. 1a). The inclination of the cylinder is denoted by angle a; and the yaw angle
of the incident wind is designated by angle b: The cylinder is supposed to be
supported by springs at its ends in the plane 1-5-7. The consideration of such a
cylinder rather than a real cable is because many researchers used it in their wind
rain simulation tests [1,2] and some of the test results have to be used in this study.
As a preliminary analytical study, the upper rivulet is assumed to uniformly
distribute along the longitudinal axis of the cylinder and circumferentially vibrate
over the surface of the cylinder. Turbulent effect and axial ow effect are not
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 28
considered. The static position of the upper rivulet due to the mean wind when the
cylinder is stationary is dened by angle y
0
and the dynamic angular displacement of
the rivulet as the cylinder vibrates is designated by y with reference to y
0
(see
Fig. 1b). Since the cylinder is not perpendicular to the direction of the mean wind
speed U
0
; one needs to nd the component of mean wind speed perpendicular to the
cylinder, U; using the following equation:
U U
0

cos
2
b sin
2
a sin
2
b
_
U
0

sin
2
a cos
2
a cos
2
b
_
: 1
The angle of attack of the mean wind speed component U is dened as g (see
Fig. 1b), which indicates the stagnation point of incident wind on the surface of
cylinder. The position of stagnation point depends on many factors such as the cross
section of cylinder with rivulet, roughness of cylinder surface, wind turbulence, and
cable motion. No experimental results are available to the writers at this stage with
respect to the position of stagnation point on a stay cable with rivulet. Thus, the
angle of attack in this study is selected as the ideal angle of attack multiplied by an
inuence factor e:
g e sin
1
sin a sin b

cos
2
b sin
2
a sin
2
b
_
_
_
_
_
_
_ 2
Fig. 1. Modeling of windrain-induced cable vibration: (a) orientation of inclined cylinder, and
(b) relative velocity of cable and moving rivulet.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 29
when e is set as 1, g represents the ideal angle of attack for the cylinder without
rivulet [7]. When e is selected as zero, it indicates that the position of stagnation point
is the same as that on the cylinder without rivulet and yaw angle. The effects of the
mean wind speed component along the cylinder axis and wind turbulence are not
considered in this study.
In consideration of the transverse vibration of the cable of velocity yt and the
angular vibration of the upper rivulet of velocity

yt; the relative velocity of mean
wind to the cylinder with moving rivulet is therefore
U
rel

U cos g R

y cosy y
0

2
U sin g y R

y siny y
0

2
_
3
The angle between the relative velocity U
rel
and the horizontal axis in Fig. 1b is
dened as f

t; which can be expressed as

f

tan
1
U sin g y R

y siny y
0

U cos g R

y cosy y
0

; 4
where R is the radius of the cylinder and the size of the rivulet is considered small
compared with the diameter of the cylinder. The eld and laboratory observations
indicate that R

y is very small compared with the concerned mean wind speed U:

Thus, if the angles g and f

t are limited a certain range, Eq. (3) can be reduced as

f

U sin g y R

y siny y
0

U cos g

U sin g y R

yy y
0

1
6
y y
0

U cos g
: 5
The net vertical force on the cylinder per unit length in the y direction is then
F
rDU
2
rel
2
C
L
f cos f

C
d
f sin f

; 6
where r is the density of the air, D is the diameter of the cross-section of the cylinder,
C
d
is the drag coefcient, and C
L
is the lift coefcient. The drag and lift coefcients
of the cylinder with rigid rivulet measured from the wind tunnel tests are often
expressed as the function of the angle f dened in Fig. 2 [1,8]. The relationship
between the angles f

and f is given by
f f

y y
0

U sin g y R

yy y
0

1
6
y y
0

U cos g
y y
0
: 7
In consideration that the cylinder structural damping is viscous and the cylinder
mass is uniformly distributed, the equation of vertical motion of the cylinder can be
written as
. y o
2
y 2x
s
o y
F
m
0 8
in which o is the circular natural frequency of the cylinder, x
s
is the structural
damping ratio of the cylinder; and m is the mass of the cylinder per unit length.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 30
Now t the drag coefcient and lift coefcient in Eq. (6) with respect to f using
the rst three terms of the Taylors series.
C
L
f A
0
A
1
f
A
2
2
f
2

A
3
6
f
3
; 9
C
d
f B
0
B
1
f
B
2
2
f
2

B
3
6
f
3
: 10
The substitution of Eqs. (9) and (10) into Eq. (6) and then the expansion of the sine
and cosine functions in Eq. (6) with respect to f

and the use of the rst two terms of

the power series yield the vertical force acting on the cylinder per unit length of the
form.
F
rDU
2
rel
2
A
0
A
1
f
1
2
A
2
f
2

1
6
A
3
f
3
1
f
2
2
_ _ _
B
0
B
1
f
1
2
B
2
f
2

1
6
B
3
f
3
f

f
3
6
_ __
11
The substitution of Eqs. (1), (3), (5) and (7) into Eq. (11) and then the reservation of
the linear terms of y; y and

y only lead to
F
rD
2
G
1
RU
0

y G
2
U
0
y G
3
U
2
0
y; 12
where G
1
; G
2
; and G
2
can be found in appendix. They are not only the function of
the cable inclination, wind yaw angle, wind stagnation point, and the mean wind
speed via the static position of rivulet but also the function of cable motion and
rivulet motion which may change the values of drag coefcient and lift coefcient
as indicated by Eqs. (7), (9) and (10). The combination of Eq. (8) with Eq. (12)
-100 -75 -50 -25 0 25
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
C
L
C
d
S
t
e
a
d
y

w
i
n
d

f
o
r
c
e

c
o
e
f
f
i
c
i
e
n
t
s

C
d
,

C
L
Angle of attack (deg)
F
D
F
L
Wind

y
Fig. 2. Aerodynamic coefcients vs. wind angle of attack (Gu et al., [8]).
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 31
results in
. y o
2
y 2x
s
o
rDG
2
U
0
2m
_ _
y
rD
2m
G
1
RU
0

y G
3
U
2
0
y: 13
Based on the observations from either eld measurements or simulated windrain
tunnel tests [1], the motion of upper rivulet, y; can be assumed to be harmonic as
long as the steady-state vibration is concerned.
y a sin
u
ot: 14
The frequency of rivulet motion is almost the same as that of cable motion, as
observed from windrain tunnel tests when the vibration is in steady state [1]. The
amplitude of rivulet motion, a; is also taken from windrain-tunnel tests. Thus, the
motion of the rivulet is assumed as a known function of time in this preliminary
study. Substituting Eq. (14) into Eq. (13) produces
. y o
2
y 2
*
xo y
rDa
2m
G
1
RU
0 u
o cos
u
ot G
3
U
2
0
sin
u
ot
*
Ft 15
in which
*
x x
s

rDG
2
U
0
4mo
x
s
x
a
; 16
*
x is the total damping ratio of the cylinder, x
a
is the aerodynamic damping ratio; and
*
Ft is the normalized force due to the motion of rivulet. Since G
2
changes with the
mean wind speed U
0
through the static position of rivulet and depends on the
motions of cable and rivulet, the aerodynamic damping ratio and the total damping
ratio are the function of time.
3. Horizontal cylinder with xed rivulet
There are two approaches currently used by researchers to simulate the rivulet on
cable section model in simulation tests: one is to spray water appropriately onto the
surface of the cable model to form moving rivulet [9], and the other is to stick
articial rivulet on the cable surface [8]. To start with the simplest case, this section
investigates the dynamic behavior of horizontal cylinder with xed rivulet using the
developed model and then compares the results with the wind tunnel test results
obtained by Gu et al. [8].
When the rivulet is xed, the dynamic motion yt of the rivulet relative to the
cylinder is equal to zero and the static position of the rivulet y
0
is no longer the
function of the mean wind speed. The equation of vertical motion of the horizontal
cylinder becomes
. y o
2
y 2
*
xo y 0: 17
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 32
G
2
in Eq. (16) is also changed. For instance, G
2
can be expressed by the following
equation for the horizontal cylinder with zero wind yaw angle b:
G
2
B
0
B
1
y
0

B
2
2
y
2
0

B
3
6
y
3
0
A
1
A
2
y
0

A
3
2
y
2
0
: 18
For the cylinder with xed rivulet, the coefcients A
i
and B
i
i 1; 2; 3; 4 may
change with the angle f that in turn depends on the cylinder velocity y and the angle
y
0
; as seen from Eqs. (7), (9) and (10). Therefore, Eq. (17) is a nonlinear differential
equation. The RungeKutta method with the computer package MATLAB as a
platform is employed to nd the solution of the equation of motion of the cylinder
with rivulet.
Let us consider the horizontal cylinder of 120 mm diameter with a xed upper
rivulet [8]. The upper rivulet is modeled as a half ellipse with a long axis of 14.5 mm
and a short axis of 10 mm. The measured drag and lift coefcients are plotted in
Fig. 2. It is seen that when f is equal to 431, the derivative of lift coefcient with
angle of attack has a sudden change from a positive value to a negative value
whereas the derivative of the drag coefcient changes from a negative value to a
positive value. To have the best t of the measured aerodynamic coefcients using
the rst three terms of the Taylors series, the measured aerodynamic coefcients are
divided into the two ranges distinguished by the critical angle f of 431.
Fig. 3a shows the time history of displacement response of the horizontal cylinder
with xed rivulet, obtained by using Eq. (7) and Eqs. (16)(18) with a small initial
displacement. The cylinder has the following parameters: the mass per unit length of
the cylinder m is 3.48 kg/m; the natural frequency f
n
is 1.28 Hz; and the structural
damping ratio x
s
is 0.2%. The mean wind speed U
0
and yaw angle b are 10 m/s and
zero, respectively. The air density is 1.225 kg/m
3
and the position of the xed rivulet
y
0
is 321. The computed result shows that after giving a small initial displacement,
the vibration amplitude of the cylinder increases with time within the rst 320 s
because the total damping ratio
*
x is negative and constant during this period (see
Fig. 3b). After the amplitude of the cylinder is increased to a certain level, the angle
f reaches the value of 431. The aerodynamic damping ratio and the total damping
ratio then change alternately between a negative value and a positive value other
than remaining constant (Fig. 3b). As a result of the jump of aerodynamic damping,
the motion of the cylinder becomes almost periodic with nearly constant peak-to-
peak amplitude after 320 s.
To know the effect of position of the xed rivulet y
0
on cylinder vibration, the
motion of the horizontal cylinder with a wind yaw angle of 301 is computed against a
series of angles y
0
: The computed results are plotted in Fig. 4a together with the test
results [8], in which A
max
means the maximum peak to peak amplitude. A
allow
means
the allowable peak to peak amplitude that was set as 530 mm in the simulated wind
tunnel tests. The mean wind speed U
0
is 18 m/s used in the computation. A
max
=A
allow
equal to one indicates that the vibration amplitude is larger than the allowable
amplitude. It is seen that the computed and measured results both demonstrate that
within a certain range of y
0
the cylinder has large amplitude vibration, but out of this
range the cylinder has very small vibration or the cylinder stops its vibration very
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 33
fast after a small initial displacement or disturbance. The cylinder starts its large
amplitude vibration at 301, measured by the wind tunnel test and also predicted by
the analytical model. There is some difference of the upper bound of the range
between the computed and measured results.
Fig. 4b shows the variation of the normalized vibration amplitude of the cylinder
with mean wind speed for the case of 301 wind yaw angle and 351 static rivulet
position. Again, the computed results are in good agreement with the test results.
Both the wind tunnel test results and the computed results show a kind of amplitude
restricted vibration. The reason why the vibration amplitude is limited is that when
the vibration amplitude of the cylinder reaches a certain level, the aerodynamic
damping ratio and the total damping ratio change periodically between a positive
peak value and a negative peak value.
0 100 200 300 400
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
D
i
s
p
l
a
c
e
m
e
n
t

r
e
s
p
o
n
s
e

(
m
)
Time (second)
0 100 200 300 400
-0.020
-0.015
-0.010
-0.005
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.035
0.040
0.045
m=3.48 kg/m
fn=1.28 Hz

s
=0.2%

o
=32
o
U
o
=10 m/s
m=3.48 kg/m
fn=1.28 Hz

s
=0.2%

o
=32
o
U
o
=10 m/s
T
o
t
a
l

d
a
m
p
i
n
g

r
a
t
i
o

Time (second)
~
(a)
(b)
Fig. 3. Time histories of horizontal cylinder vibration with xed rivulet: (a) displacement response, and
(b) total damping ratio.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 34
4. Inclined cylinder with moving rivulet
To investigate the capability of the analytical model for predicting the motion of
inclined cable with moving rivulet, the inclined cylinder tested in a simulated wind
rain tunnel by Hikami and Shiraishi [1] is selected in this study. The diameter of the
cylinder is 140 mm; the mass per unit length is 10.2 kg/m; both the inclination and
yaw angles are 451; and the natural frequency is 1 Hz. The structural damping ratio
and the stagnation inuence factors are assumed to be 0.7% and 0.4%, respectively.
The frequency of steady-state upper rivulet motion is the same as the natural
frequency of the cylinder [1]. The drag and lift coefcient curves of the cylinder with
upper rivulet (d=D 0:1) were reported by Yamaguchi [5] and reproduced in
Fig. 5a. It is seen that there is a sudden change in the gradient of the curves at an
15 20 25 30 35 40 45 50
0.0
0.2
0.4
0.6
0.8
1.0
measured
computed
f
n
= 1.28Hz
m= 3.48kg/m

s
= 0.2%
U
o
= 18m/s
= 0
o
A
m
a
x
/
A
a
l
l
o
w
A
m
a
x
/
A
a
l
l
o
w
Position of rivulet
o
(deg)
0 2 4 6 8 10 12 14 16 18 20
0.0
0.2
0.4
0.6
0.8
1.0
measured
computed
f
n
= 1.28Hz
m= 3.48kg/m

s
= 0.2%
o
= 35
o

= 0
o
Wind speed U
o
(m/s)
(a)
(b)
Fig. 4. Variations of cylinder vibration amplitude: (a) with rivulet position, and (b) with mean wind speed.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 35
angle of 551. The curves are subsequently tted to the Taylors series of the rst
three terms distinguished by the critical angle of 551. For the moving rivulet, its
static position y
0
is the function of mean wind speed. The wind tunnel test results
related to this position [1] are plotted in Fig. 5b and tted by a quadratic function.
The results obtained from the analytical model are showed in Fig. 6 together with
the wind tunnel test results. In this gure, the x-coordinate is the mean wind speed
U
0
and the y-coordinate is the maximum displacement response amplitude of the
cylinder. Both analytical and measured results show that the windrain-induced
cylinder vibration only occurs within a certain range of mean wind speed. The wind
rain-induced cylinder vibration is also a kind of vibration of restricted amplitude
rather than galloping. The predicted maximum vibration amplitude is moderately
-90 -60 -30 0 30
-0.5
0.0
0.5
1.0
1.5
C
L
C
d
S
t
e
a
d
y

w
i
n
d

f
o
r
c
e

c
o
e
f
f
i
c
i
e
n
t
s

C
d
,

C
L
Angle of attack (degree)
8 10 12 14 16 18
20
40
60
80
fitted
measured
S
t
a
t
i
c

p
o
s
i
t
i
o
n

o
f

r
i
v
u
l
e
t

(
o
)
Mean wind speed (m/s)
F
D
F
L
Wind

y
(a)
(b)
Fig. 5. Aerodynamic properties of windrain-induced cylinder vibration: (a) aerodynamic coefcients
vs. wind angle of attack [5], and (b) static position of upper rivulet vs. mean wind speed (Hikami and
Shiraishi [1]).
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 36
larger than that from the windrain-tunnel tests but the on-set mean wind speed at
which the large cylinder vibration starts is slightly lower in the computation than the
windrain-tunnel test. Such differences may be due to uncertainties in the selection
of stagnation inuence factor and structural damping ratio.
One particular mean wind speed corresponding to the global maximum
displacement response, that is, 9.8 m/s is selected to examine cable vibration
features. It is found that at the mean wind speed of 9.8 m/s, the angle ft changes
with time and its negative amplitude is less than 551. Thus, the aerodynamic
damping ratio and the total damping ratio alternate with time. Because the structural
damping ratio is relatively large, the total damping ratio is kept within the positive
value range. However, the upper and lower bounds of the total damping ratio are
quite small but the peak-to-peak amplitude of excitation force due to the motion of
rivulet is relatively larger. As a result, the cylinder exhibits a large amplitude
vibration. Thus, one may conclude that the restricted large amplitude vibration of
the cylinder at 9.8 m/s wind speed is because of very small total damping ratio and
relatively large amplitude force due to the motion of rivulet.
The time-history of the displacement response of the cylinder is shown in Fig. 7.
During the rst 50 s, the cylinder vibration is in a transient vibration period and the
smallest amplitude of the angle ft is above 551. This transient vibration period
does not reect the real situation because of the steady-state rivulet motion
assumption used in this study. However, when the vibration amplitude of cylinder
increases to a certain level, the smallest amplitude of the angle ft becomes less than
551. The aerodynamic damping ratio and the total damping ratio no longer remain
constant. The normalized force also changes with time because of the interaction
between incident wind, moving rivulet and cylinder motion. All these features of
windrain-induced cable vibration cause the cylinder to vibrate with almost constant
amplitude.
0 2 4 6 8 10 12 14 16 18 20
0.00
0.05
0.10
0.15
0.20
0.25
measured(Hikami
and Shiraishi,1988)
computed
M
a
x
i
m
u
m

r
e
s
p
o
n
s
e

a
m
p
l
i
t
u
d
e

(
m
)
Mean wind speed U
o
(m/s)
Fig. 6. Maximum cylinder vibration amplitude vs. mean wind speed.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 37
5. Concluding remarks
An analytical model for describing steady-state windrain-induced cable vibration
has been established using the knowledge of aerodynamics and structural dynamics
and some measurement results. The effect of mean wind speed on the position of
upper rivulet and the inuence of moving upper rivulet on cable motion were taken
into account in the model. The analytical model was applied to some test cable
models with either moving rivulet or xed articial rivulet. It was found that the
analytical model is able to capture main vibration features of inclined cylinders with
moving rivulet, such as velocity-restricted vibration and amplitude-restricted
vibration. The occurrence of velocity- and amplitude-restricted vibration is mainly
because of alternating aerodynamic damping ratio and/or alternating excitation
force due to the interaction between rivulet motion, cable motion and wind. For the
vibration of horizontal cylinder with xed articial rivulet, the vibration of large
amplitude is due to negative aerodynamic damping ratio. However, its vibration
amplitude will be eventually restricted because the oscillation of cylinder leads
aerodynamic damping ratio change within the range bounded by a negative value
and a positive value.
It should be pointed out that in the derivation of windrain-induced aero-
dynamic force on the cylinder, the effects of axial ow and turbulence are
not considered in this investigation because there are no enough information
and knowledge for the writers to quantify their effects. The amplitude of upper
rivulet is also taken from windrain tunnel tests and assumed to be independent
of the cylinder motion, indicating that only steady-state oscillation is con-
cerned. Further investigation is needed to see if the latter assumption can be
removed.
0 100 200 300 400
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
m = 10.2 kg/m f
n
= 1 Hz = 0.007 =45 =45
D
i
s
p
l
a
c
e
m
e
n
t

r
e
s
p
o
n
s
e

(
m
)
Time (second)
Fig. 7. Time history of cylinder vibration with moving rivulet.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 38
Acknowledgements
The writers are grateful for the nancial support from the Research Grants
Council of Hong Kong through a RGC research grant (CERG 5066/97E).
Appendix A. Aerodynamic parameters C
1
C
3
c
0
cos
2
b sin
2
a sin
2
b; 19
c
1
2 g y
0

c
0
_
; 20
c
2
2 sin g

c
0
_
; 21
c
3

1

c
0
_
cos g
; 22
c
4

y
0
y
3
0
=6

c
0
_
cos g
; 23
c
5
tan g; 24
c
6
c
5
y
0
; 25
G
1
c
0
c
4
1
c
2
5
2
_ _
c
1
c
5
1
c
2
5
6
_ _ _ _
B
0
c
0
c
4
c
5
1
c
2
5
6
_ _
c
0
c
4
c
6
1
c
2
5
2
_ _
c
1
c
5
c
6
1
c
2
5
6
_ _ _ _
B
1
c
0
c
4
c
5
c
6
1
c
2
5
6
_ _

c
0
c
4
c
2
6
2
1
c
2
5
2
_ _

c
1
c
5
c
2
6
2
1
c
2
5
6
_ _ _ _
B
2

c
0
c
4
c
5
c
2
6
2
1
c
2
5
6
_ _

c
0
c
4
c
3
6
6
1
c
2
5
2
_ _

c
1
c
5
c
3
6
6
1
c
2
5
6
_ _ _ _
B
3
c
1
1
c
2
5
2
_ _
c
0
c
4
c
5
_ _
A
0
c
0
c
4
1
c
2
5
2
_ _
c
1
c
6
1
c
2
5
2
_ _ _
c
0
c
4
c
5
c
6
A
1
c
0
c
4
c
6
1
c
2
5
2
_ _

c
0
c
4
c
5
c
2
6
2

c
1
c
2
6
2
1
c
2
5
2
_ _ _ _
A
2

c
0
c
4
c
2
6
2
1
c
2
5
2
_ _

c
0
c
4
c
5
c
3
6
6

c
1
c
3
6
6
1
c
2
5
2
_ _ _ _
A
3
26
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 39
G
2
c
0
c
3
1
c
2
5
2
_ _
c
2
c
5
1
c
2
5
6
_ _ _ _
B
0
c
0
c
3
c
5
1
c
2
5
6
_ _
c
0
c
3
c
6
1
c
2
5
2
_ _
c
2
c
5
c
6
1
c
2
5
6
_ _ _ _
B
1
c
0
c
3
c
5
c
6
1
c
2
5
6
_ _

c
0
c
3
c
2
6
2
1
c
2
5
2
_ _

c
2
c
5
c
2
6
2
1
c
2
5
6
_ _ _ _
B
2

c
0
c
3
c
5
c
2
6
2
1
c
2
5
6
_ _

c
0
c
3
c
3
6
6
1
c
2
5
2
_ _

c
2
c
5
c
3
6
6
1
c
2
5
6
_ _ _ _
B
3
c
2
1
c
2
5
2
_ _
c
0
c
3
c
5
_ _
A
0
c
0
c
3
1
c
2
5
2
_ _
c
2
c
6
1
c
2
5
2
_ _ _
c
0
c
3
c
5
c
6
A
1
c
0
c
3
c
6
1
c
2
5
2
_ _

c
0
c
3
c
5
c
2
6
2

c
2
c
2
6
2
1
c
2
5
2
_ _ _ _
A
2

c
0
c
3
c
2
6
2
1
c
2
5
2
_ _

c
0
c
3
c
5
c
3
6
6

c
2
c
3
6
6
1
c
2
5
2
_ _ _ _
A
3
27
G
3
c
0
c
5
1
c
2
5
6
_ _
B
1
B
2
c
6

1
2
B
3
c
2
6
_ _
c
0
1
c
2
5
2
_ _
A
1
A
2
c
6

1
2
A
3
c
2
6
_ _
: 28
References
[1] Y. Hikami, N. Shiraishi, Rainwind induced vibrations of cables in cable stayed bridges, J. Wind Eng.
Ind. Aerodyn. 29 (1988) 409418.
[2] M. Matsumoto, N. Shirashi, H. Shirato, Rainwind induced vibration of cables of cable-stayed
bridges, J. Wind Eng. Ind. Aerodyn. 4144 (1992) 20112022.
[3] A. Bosdogianni, D. Oliver, Wind and rain induced oscillations of cables of stayed bridges, J. Wind
Eng. Ind. Aerodyn. 64 (1996) 171185.
[4] H.P. Ruscheweyh, The mechanism of rainwind induced vibration, Proceedings of the Tenth
International Conference on Wind Engineering, Balkema, Rotterdam, Vol. 2, 1999, pp. 10411047.
[5] H. Yamaguchi, Analytical study on growth mechanism of rain vibration of cables, J. Wind Eng. Ind.
Aerodyn. 33 (1990) 7380.
[6] C.P.W. Geurts, P.C. van Staalduinen, Estimation of the effects of rainwind-induced vibration in the
design stage of inclined stay cables, Proceedings of the Tenth International Conference on Wind
Engineering, Balkema, Rotterdam, Vol. 2, 1999, pp. 885892.
[7] M. Matsumoto, M. Yamagishi, J. Aoki, N. Shiraishi, Various mechanism of inclined cable
aerodynamics, Proceedings of the 9th International Conference on Wind Engineering, Wiley Eastern,
New Delhi, Vol. 2, 1995, pp. 759769.
[8] M. Gu, Y.L. Xu, C.J. Liu, H.F. Xiang, Wind tunnel study of response characteristics of cables with
articial rivulet, Proceedings of the International Conference on Advances in Structural Dynamics,
Elsevier Science, Vol. 2, 2000, pp. 14791486.
[9] O. Flamand, Rainwind induced vibration of cables, J. Wind Eng. Ind. Aerodyn. 57 (1995) 353362.
Y.L. Xu, L.Y. Wang / J. Wind Eng. Ind. Aerodyn. 91 (2003) 2740 40