Computers and Structures 84 (2006) 1584–1595

www.elsevier.com/locate/compstruc

Rain–wind induced vibrations – phenomenology, mechanical

modelling and numerical analysis
C. Seidel *, D. Dinkler
Institut für Statik, Technische Universität Braunschweig, Beethovenstr. 51, 38106 Braunschweig, Germany

Received 23 September 2005; accepted 18 January 2006

Available online 19 May 2006

Abstract

The simultaneous appearance of rain and wind at cables of cable stayed bridges and hangars of arch bridges may induce varying oscil-
lations with large amplitudes. These phenomena are identified as rain–wind induced vibrations. The paper presents a possible fluid-
mechanical interpretation of rain–wind induced vibrations. Based on this interpretation a mechanical model is deduced, in order to
enable numerical investigations. The complex system of nonlinear differential equations is analysed concerning the stability of solutions.
Rain–wind induced vibrations exist only within a certain range of wind velocity. The lower limit of the critical velocity range is deter-
mined for various cable inclinations and angles of incidence. Furthermore, a fluid mechanical interpretation is given for the existence of
the upper limit of critical velocity. The approximation of the upper limit velocities and the determination of the critical cable diameters
can be derived from established fluid-mechanical correlations. The numerical investigations show that rain–wind induced oscillations
occur on vertical cables, too. For the Faro–Falster Bridge, the numerical vibration analysis is carried out for the cables parallel and per-
pendicular to the wind direction. The analysis of the vibration frequencies is demonstrated for the Dömitz Bridge.
 2006 Civil-Comp Ltd. and Elsevier Ltd. All rights reserved.

Keywords: Rain–wind induced vibrations; Self-excited systems; Nonlinear dynamic; Stability; Aeroelastic instabilities

1. Introduction majority of bridges expensive measures were adopted to

prevent damage due to rain–wind induced vibrations.
Rain–wind induced vibrations are a fluid-structure inter-
action phenomena that occurs when rain and wind act 2. Phenomenology
simultaneously on cables, hangars and ropes. In 1986,
Hikami and Shiraishi [1] detects this kind of vibration Rain–wind induced vibrations exist for wind speeds in
behaviour of cables at the Meikonishi Bridge in Japan. the range of 5–30 m/s and for Reynolds numbers in the
The vibration phenomena may induce oscillations with subcritical flow (5 · 104 to 3 · 105). The vibrations of
large amplitudes, thus the fatigue of construction elements cables are observed for a frequency range between 0.5
is possible. This kind of vibrations are often observed at and 3 Hz and for a special range of yawing angle of wind
stay cables of cable stayed bridges and hangars of arch flow. For the phenomena of rain–wind induced vibrations
bridges. Oscillations could also occur at overhead lines, it is essential that rain and wind act simultaneously. Rain
ropes of pylons and cables of suspension bridges. During water forms rivulets on the cable surface, which follow
the last years rain–wind induced vibrations have been the gravity and flow on the lower surface of the cable.
observed at stay cable bridges all over the world. For a Above a critical wind velocity a second rivulet exists on
the upper surface of the cable. The existence of the two riv-
ulets is very significant for the vibration behaviour, as
*
Corresponding author. Tel.: +49 531 391 3684; fax: +49 531 391 8116. observations at real structures have shown. The rivulets
E-mail address: [email protected] (C. Seidel). oscillate around states of equilibrium with same frequency

0045-7949/$ - see front matter  2006 Civil-Comp Ltd. and Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2006.01.033
 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595 1585

of the cable vibrations. The reasons of rain–wind induced vortex shedding and the axial flow also without rain. This
vibrations could not yet be determined satisfyingly, demonstrates that the effects are generally an excitation
because the flow around an inclined cable with rivulets is mechanism of inclined cables with and without rain. Obvi-
very complex and the cable aerodynamics of this phenom- ously the effects increase the amplitudes of cable vibrations,
ena depend on various parameters like the inclinations and when rain is involved. Ruscheweyh [8] and Verwiebe [2]
diameters of the cable, the geometrical shape and the posi- assume that the effective cross section of cable and rivulets
tions of the rivulet as well as the velocity and the angle of are changed permanently as a result of the rivulet motion.
incidence of the flow. Furthermore the rivulet controls the separation of the wind
However, different fluid mechanical phenomena are dis- flow at the cylinder. As result of wind tunnel observations
cussed in the literature [1–7] in order to characterise rain– Verwiebe and Ruscheweyh describe three different excita-
wind induced vibrations. According to [1–4] the instability tion mechanisms for rain–wind induced vibrations parallel
does not seem to be caused by vortex induced vibrations, and perpendicular to the wind direction in dependence of
because the Strouhal number, the amplitudes and the fre- the rivulet motion at the cable surface.
quency differ from those of rain–wind induced vibrations. Another possible mechanism of excitation of rain–wind
Since the surface of the cylindrical cable is too smooth induced vibrations is derived by the authors in [9] and [10].
and a rigid edge does not exist, standard galloping effects The mechanism is based on the phenomena of the Prandtl
are not suitable, too. Wake galloping can be excluded, tripwire [11] and considers the rivulets as a movable distur-
because the distance between cables is too large in most bance. Following this approach the occurrence of the lower
cases. In comparison to the diameter of cables, the height and the upper limit of the critical velocity can be explained
and width of the rivulets do not reach a sufficient size, so and all kinds of observable vibrations may be described.
that rain–wind induced vibrations could be generated by Fig. 1 shows the effect of the Prandtl trip wire. Herein, a
the change of effective cross-sectional shape. The height wire is attached to a sphere (B 300 mm) in the critical point
of rivulets is about 1 mm and the width varies from of transition and reduces the drag coefficient significantly.
4 mm up to 10 mm, where typical cable diameters are The wire induces a transition from the subcritical to the
between 100 and 200 mm. supercritical flow at considerably lower Reynolds number
First approaches [6] address the instability problem in than for a flow around an undisturbed sphere.
terms of mechanical systems of coupled bodies with two Similar phenomena can be observed at elongated cylin-
degrees of freedom. In case of the 2-DOF instability the ders. In contrast to the fixed, immovable disturbance of the
mass of rivulet is too small for the generation of an exciting Prandtl tripwire, moving rivulets cause movable distur-
mechanism. Yamaguchi [6] determines the critical position bances regarding rain–wind induced vibrations. When the
of the rivulet at 50 with respect to the stagnation point of oscillations start, the rivulets are located near the separa-
flow by use of Den Hartog Mechanism. The results differ tion point of the subcritical flow. The Prandtl tripwire is
significantly from measurements using real water rivulets active in the same area. In contrast to the Prandtl tripwire
[1–5], where the position is observed at 80 in near of the an interaction between the motion of the cable and the riv-
separation point of flow. For the 2-DOF galloping Yama- ulets exists in the special case of rain–wind induced vibra-
guchi [6] shows that the critical position of an artificial tions. This interaction is due to the reduction of the drag
rivulet is similar to the observed position. The artificial and stiffness of the cable. For the overall mechanism it is
rivulets in [6] are described by rigid bars with a height of important that rivulets oscillate around the point of transi-
more than 10 mm which does not correspond with the size tion of the flow with the same frequency as the cable. The
of real water rivulets. result of this motion is a periodical transition of the flow
Besides of the explanation of rain–wind induced vibra- between subcritical and supercritical flow. An energy
tions as a special kind of galloping [3,6,13–16] many scien- transfer between the flow around the cable and the elastic
tists [1,2,4,5,7,8] discuss that rain–wind induced vibrations structure is induced due to the different pressure distribu-
are a new type of instability phenomena. The first interpre- tions of the subcritical and supercritical flow. The inter-
tation of the mechanism of rain–wind induced vibrations is action between cable, rivulets and flow determines the
given by Hikami and Shiraishi [1]. Hikami describes that development of a self-excitation mechanism. This leads to
the rivulet oscillation in circumferential direction of the
cable surface is due to the cable motion and leads to a peri-
odic change of cable cross section for the flow. An interest-
ing approach for the explanation of rain–wind induced
vibrations is shown by Matsumoto et al. [5] and Honda
et al. [7]. They describe an excitation mechanism of inclined
cable, which is caused by the three-dimensionality of Kár-
mán vortex shedding. The oscillations are additionally
enhanced by the axial flow in wake of an inclined cable
and by the geometrical shape of upper rivulet. Both Fig. 1. Flow around a sphere: (a) without and (b) with a Prandtl tripwire
authors observe the three-dimensional effects of Kármán [12].
1586 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595

vibrations with large amplitudes parallel or perpendicular

to the wind direction depending on the location of rivulets.

3. Mechanical model

Different mechanical models are presented in the litera-

ture [6,13–16] for modelling of rain–wind induced vibra-
tions. In 1990 Yamaguchi [6] formulates the first model
based on the 2-DOF galloping. The degrees of freedom
describe the translation perpendicular to the wind direction
and the rotation of the cross-section. In 1998 Geurts dem-
onstrates in [13] the standard galloping model with one Fig. 3. Spatial view of the yawed cable.
degree of freedom, where the aerodynamic coefficients are
fitted according to measured cable vibrations. Van der
Burgh [14] extends the standard galloping model with to be considered for a realistic simulation (see Fig. 3) and
respect to a power series expansion for the description of are included in the equations of motion.
the aerodynamic forces. The model of Yamaguchi is mod- The equations of motion of the cables are formulated
ified in [15] by Wang and investigated in comparison to a in an inertial system. Neglecting eigenrotations there are
simple single-degree-of-freedom model. Peil [16] develops six nonlinear coupled differential equations. These are
a 3-DOF model, which is deduced from the 2-DOF model required to compute the unknown translations of cable,
of Yamaguchi. The additional degree of freedom describes the unknown angles of rivulets and the reactive forces for
the translation parallel to the wind direction. Various anal- both rivulets.
yses of stability and parameter are demonstrated with the Eqs. (1) and (2) describe the motions of the cable in x-
3-DOF model. and y-direction in terms of the inertia, damping and stiff-
In this paper the equations of motion in two dimensions ness forces of the cable as well as the drag and lift forces
for cables and rivulets are formulated in consideration of due to wind flow. Furthermore gravity of cable, shear
the phenomenology of the Prandtl tripwire and the rivulets forces between rivulets and cable and reactive forces for
as a movable disturbance. The model is based on the both rivulets Nzi are included in the equations. br is the rel-
assumption that cables are infinitely long cylinders. It is ative angle of incidence of the quasi-stationary theory.
essential to include the influence of the rivulets and its
mS€x ¼ d x x_  k x x  F A sin br  F W cos br þ mS g sin b0
resulting interaction with the structure in the equations of
motion of the cable cross-section and in the quasi-station- þ l  ½_x sin u1 þ y_ cos u1 þ u_ 1 R sin u1  N z1 cos u1
ary theory. Furthermore, the equations of motion of the þ l  ½_x sin u2 þ y_ cos u2  u_ 2 R sin u2  N z2 cos u2 ;
rivulets are developed. Therefore the effects of the bound- ð1Þ
ary layer based on the Prandtl boundary layer equations
mS €y ¼ d y y_  k y y þ F A cos br  F W sin br  mS g cos b0
and fundamentals of the physics of drops are considered.
Fig. 2 shows the mechanical model with a flow acting  l  ½_x sin u1 þ y_ cos u1 þ u_ 1 R cos u1  N z1 sin u1
from the right to the left. The cross section of the cable is þ l  ½_x sin u2 þ y_ cos u2  u_ 2 R cos u2 þ N z2 sin u2 :
always a circle because the section is perpendicular to the ð2Þ
longitudinal axis of cable. At first rivulets and cross section
of cable are regarded as not deformable. Thus, the motions The drag force FW and the lift force FA in Eqs. (3) and
of cable and rivulets may be defined as motions of rigid (4) are formulated on the basis of the quasi-stationary the-
bodies. Spatial effects as the result of the yawed cable have ory. qL represent the density of air, u1 the flow rate, cAi the

Fig. 2. Mechanical model.

 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595 1587

lift- and cWi the drag coefficient, which depend on angles of oscillations can be excited, even when the system is not
rivulets ui and the relative angle of incidence br (Eq. (5)) moving with a velocity of cable perpendicular to the wind
1 direction in the initial state.
F A ¼ qL A  ½ðu1 þ x_ Þ2 þ y_ 2   ½cA1 ðu1  br Þ The numerical analysis of the model equations is very
2
complex, due to the high-grade nonlinearities and the cou-
þ cA2 ðu2  br Þ; ð3Þ
pled character of the differential equation system. Espe-
1 2 cially for the complete system of coupled equations the
F W ¼ qL A  ½ðu1 þ x_ Þ þ y_ 2   ½cW1 ðu1  br Þ
2 prediction of the stability of the solution appears as very
þ cW2 ðu2  br Þ; ð4Þ difficult. Mathematical transformations enable the stability
y_ of the solution of the system of nonlinear differential equa-
br ¼ arctan : ð5Þ tion. Hence, the equations of motion for the cable are for-
u1 þ x_
mulated in Cartesian coordinates, whereby the equations
For both rivulets the equations of motions are of motion for the rivulets are developed in polar
€ 1 R sin u1  u_ 21 R cos u1 Þ ¼ N 1 cos u1  T 1 sinu1 ;
mr1 ð€x  u ð6Þ coordinates.
€ 1 R cos u1  u_ 21 R sin u1 Þ ¼ N 1 sinu1 þ T 1 cos u1 ;
mr1 ð€y þ u ð7Þ mW1 ð€
u1 R  €x sin u1 þ €y cos u1 Þ ¼ T 1 ; ð13Þ
€ 2 R sin u2  u_ 22 R cos u2 Þ ¼ N 2 cos u2  T 2 sinu2 ;
mr2 ð€x  u ð8Þ mW1 ðu_ 21 R
þ €x cos u1 þ €y sin u1 Þ ¼ N 1 ; ð14Þ
€ 2 R cos u2 þ u_ 22 R sin u2 Þ ¼ N 2 sinu2  T 2 cos u2 :
mr2 ð€y  u ð9Þ mW2 ð€
u2 R þ €x sin u2 þ €y cos u2 Þ ¼ T 2 ; ð15Þ
where Ni are the normal forces and Ti are tangential forces, mW2 ðu_ 22 R  €x cos u2 þ €y sin u2 Þ ¼ N 2 : ð16Þ
respectively. The mathematical relations for both forces are Eqs. (13)–(16) describe the motions of rivulets in polar
given in Eqs. (10) and (11). The expression for the normal coordinates. Inertias are on the left hand side of these equa-
forces Ni includes effects of the reactive force Nzi, the cohe- tions and on the right hand side are the normal and tangen-
siveness between the fluid layer of the cable surface and the tial forces Ni and Ti. The six equations of motion can be
rivulets with the surface tension r0, the normal terms of reduced to four essential equations. This is not appropri-
pressure and shear forces due to flow and the normal grav- ate, since this results in numerical instabilities again. Fur-
ity, which depends on the angle of rivulets ui and spatial thermore the equations become more complex. Therefore,
angle of incidence b0. hL is the height of boundary layer it is advisable to introduce the two reactive forces as
and gL the dynamical viscosity. pseudo variables. The direct determination of both reactive
In Eq. (11) the terms describe the tangential gravity, forces is also useful under physical aspects, because they
which depends on the angle of rivulets ui and the spatial correspond to the separation condition of the rivulets.
angle of incidence b0, the viscous shear force l between riv- These differential equations describe self-excitated, param-
ulets and cable as well as tangential terms of pressure and eter-excitated or mixed forms of oscillations.
shear forces due to flow.
Z
4. Pressure distributions and aerodynamic coefficients
N i ¼ N zi  r0 Rdui  mri g cos a sinðui þ b0 Þ
Z
1 For a flow around a cylinder the normal pressure is
þ qL ½ðu1 þ x_ Þ2 þ y_ 2   cp ðui  br þ /i Þ  r cos /i d/i dominant in contrast to the shear forces of flow. The value
2
Z of shear forces are between 1% and 2% of the stagnation
R þ hL
þ 2gL  uðui  br þ /i Þ  r sin /i d/i ; ð10Þ pressure, thus the influence of shear forces can be
hL ð2R þ hL Þ
neglected. The required pressure distributions for the
T i ¼ mri g cos a cosðui þ b0 Þ þ l  ð_x sin ui  y_ cos ui þ u_ i RÞ numerical analysis are based on measurements from the lit-
Z
1 2 erature [17,18] for an infinitely long, smooth cylinder and
þ qL ½ðu1 þ x_ Þ þ y_ 2   cp ðui  br þ /i Þ  r sin /i d/i
2 the mean values of the pressure distribution cp(u).
Z
R þ hL Fig. 4(a) presents the pressure distributions around a cylin-
þ 2gL  uðui  br þ /i Þ  r cos /i d/i ; ð11Þ der for different Reynolds numbers, whereby the solid line
hL ð2R þ hL Þ
represents the supercritical flow and the dashed line the
b0 ¼ arctanðtan b  sin aÞ: ð12Þ
subcritical flow.
The spatial angle of incidence b0 is given in Eq. (12) as a The aerodynamic coefficients and their derivatives are
function of the plane angle of incidence b and the cable not investigated sufficiently for the phenomena of rain–
inclination a. By means of the aerodynamic coefficients wind induced vibrations. Measurements of pressure distri-
cAi and cWi in the Eqs. (3) and (4), the difference between butions or aerodynamic coefficients with real water rivulets
galloping and rain–wind induced vibrations is obvious. In do not exist, too. To determine aerodynamic coefficients a
contrast to aerodynamic coefficients of galloping phenom- theoretical approach is developed on the basis of the phe-
ena, the aerodynamic coefficients of rain–wind induced nomena of the Prandtl tripwire and the knowledge of the
vibrations depend on the relative angle of incidence br transition from the subcritical to the supercritical flow.
and additionally on the angles of rivulets ui. Therefore The aerodynamic coefficients are integrated analytically
1588 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595

cp 1.25
1 cp subcritical 1
cp supercritical
0.5 0.75 cw - 2 rivulets

ca , cw
ϕ 0.5 cw - 1 rivulet
0.5 1 1.5 2 2.5 3 ca
-0.5 0.25
-1 0
-1.5 -0.25
-2 -3 -2 -1 0 1 2 3
a b ϕ in rad

Fig. 4. Pressure distributions and aerodynamic coefficients.

by use of the well-known pressure distributions for the flow. The lift coefficient is non-zero, when one rivulet oscil-
smooth cylinder. It is important for the approach to con- lates around the transition point of flow or when two rivu-
sider the influence of the disturbance on the different range lets oscillate asymmetrically around the transition point of
of flow at the cable. The influence of the disturbance of the flow. The drag coefficient is non-zero in the subcritical and
flow is not significant in the range behind the separation also in the supercritical flow. In the critical range of flow
point of flow and in front of the transition point of flow. transition the drag collapses for two rivulets more signifi-
In this range the subcritical flow exists at the cable surface. cantly than for one rivulet, because the influence of two riv-
The lift and drag coefficients are computed by the Eqs. (17) ulets of the pressure distribution in the critical range of
and (18) with the pressure distribution of the subcritical flow transition is larger than for one rivulet. A helpful
flow assumption in the case of one rivulet is the integration of
Z the pressure distribution for the supercritical flow only in
1
cA ðuÞ ¼ cp ðuÞ  sin udu; ð17Þ this range, where the rivulet is active at the cable surface.
2
Z In the other range of the cable surface the subcritical flow
1
cW ðuÞ ¼ cp ðuÞ  cos udu: ð18Þ exists.
2 A comparison between the measurements by James [19]
In the range between the transition point and the sepa- and the computed aerodynamic coefficients demonstrates
ration point the influence of disturbance is dominant for that the theoretical approach is an useful assumption.
the flow and leads to supercritical flow in this range. Under Fig. 5(a) shows the computed drag coefficients for one riv-
consideration of the pressure distribution of the supercrit- ulet and the measured drag coefficients by James for a cyl-
ical flow the lift and the drag coefficients are computed inder with one tripwire. The size of the tripwire is 0.6% of
again by Eqs. (17) and (18). The results of the analytical the cable diameter. The computed lift coefficients for one
integration are demonstrated in Fig. 4(b). rivulet and the measured lift coefficients by James are pre-
Fig. 4(b) shows the lift coefficient cA and the drag coef- sented in Fig. 5(b). James use for this experiments a stem
ficients cW as a function of the rivulet angle for one rivulet with a size of 0.4% of the cable diameter. The unsymmetry
and for two rivulets. Both coefficients change significantly of the measured lift coefficients is caused by the special
at the critical position of rivulet close to the separation shape of stem where one side of stem possesses a sharp edge
point of the subcritical flow. This corresponds to the range and the other side is round. In comparison the measured
of the flow transition, which is calculated in the small inter- aerodynamic coefficients correspond with the computed
val between ±70 and ±80, where the supercritical flow aerodynamic coefficients in good approximation. Therefore
exists as result of the flow transition. Otherwise the coeffi- the computed aerodynamic coefficients are well suited to
cient functions conform to the conditions of subcritical describe the influence of a disturbance at a cable in wind

1.8

numerical result
1.6 measurement by James 0.5
numerical result
1.4 0.4 measurement by James
drag coefficient cw in [-]

0.3
lift coefficient ca in [-]

1.2
0.2
1
0.1
0.8
0
0.6
-0.1
0.4
-0.2

0.2 -0.3

0 -0.4
-180 -150 -120 -90 -60 -30 0 30 60 90 120 150 180 -180 -150 -120 -90 -60 -30 0 30 60 90 120 150 180
a position of rivulet ϕ in [˚] b position of rivulet ϕ in [˚]

Fig. 5. Computed and measured drag and lift coefficient for one disturbance.
 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595 1589

flow. Thus the different vibration phenomena of rain–wind and gravity is generated in the range of lower rivulets for a
induced vibration can be investigated numerically. rivulet angle of u2 = 74.06. With increasing wind veloc-
ity the state of equilibrium divides into a stable lower and
5. Numerical analysis an instable upper branch. The upper branch approaches
the separation point of the flow, whereas the lower position
The equations are investigated under consideration of converges against the minimal pressure of the flow.
different solution methods and numerical integration tech- A position of equilibrium for the upper rivulets arises
niques. Parameter studies and the analysis of sensitivities for a wind velocity of 9.48 m/s. With increasing wind veloc-
are carried out in order to determine the equilibrium posi- ities this equilibrium also divides into a stable upper and an
tion of the rivulets, as well as the upper and the lower crit- instable lower position of equilibrium. Thus, six rivulet
ical wind velocities. positions of equilibrium exist in the subcritical flow.
According to the wind tunnel experiments by Hikami
5.1. Equilibriums of rivulets for wind velocities of more than 8 m/s, a lower rivulet exists
at the position of u2 = 78.66. Further measurements of
For the prediction and analysis of rain–wind induced Hikami and Shiraishi [1] for wind velocities from 6 to
vibrations it is very important to know the correct position 9 m/s show exciting forces of cable, which are induced by
of equilibrium of the rivulets on the cable surface. There- the motion of lower rivulets. These observations are con-
fore it is essential to compute the tangential equilibrium firmed by the computed rivulet positions of equilibrium
of forces acting on the rivulets. The tangential flow forces and by numerical vibration analysis. In the experiment
can be determined by means of the solution of Prandtl an upper rivulet appears for wind velocities higher than
boundary layer equations. 9 m/s. In range of velocities from 10 to 12 m/s the mea-
Fig. 6 shows the computed equilibrium positions of the sured position of the upper rivulet corresponds to the
rivulets as a function of the wind velocity for a yawed cable numerical results. Thereby Hikami observes exciting forces
in comparison to wind tunnel measurements by Hikami of cable, which are induced as a result of the motion of
and Shiraishi [1]. The parameters of the HDPE – cable upper rivulet.
are given to d = 140 mm for the diameter, a = 45 for the
yawed angle of cable and b0 = 26.34 for the spatial angle 5.2. Lower and upper limit of critical wind velocity
of incidence of flow (with b = 35).
Different phenomena of rain–wind induced vibrations 5.2.1. Upper limit of critical velocity and critical diameter
can be identified by the equilibrium positions in Fig. 6. At Hikami and Shiraishi [1] observes that rain–wind
first, without wind flow two positions of equilibrium exist induced vibrations do not occur for wind velocities greater
for rivulets due to the gravity. Thereby the upper position than 13 m/s. At this velocity, the lower and the upper riv-
of rivulets is instable, whereas the lower position of rivulets ulet leave the equilibrium position of the subcritical flow
is stabilised due to the gravitational force. For a critical and move downstream on the cable surface. This phenom-
wind velocity of 5.07 m/s an equilibrium between flow force enon is also observed in other wind tunnel tests [3].

Fig. 6. Equilibrium of the rivulets for a yawed cable.

1590 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595

The Reynolds number, computed with a wind velocity Re  mL

d crit P : ð20Þ
of 13 m/s and the given cable diameter, yields the value u1
of Re = 1.2 · 105. A plausible explanation for the observed describes an upper threshold for the appearance of rain–
phenomenon is possible for this Reynolds number regard- wind induced vibrations. In general the wind velocity has
ing the fluid mechanical relations of the excitation. to be at least 2–3 m/s since a certain energy transfer from
The Reynolds number of 1.2 · 105 is in the range of the wind flow is necessary to induce vibrations of the
1 · 105–3.5 · 105, where the natural transition from the structure.
subcritical to the supercritical flow takes place. This transi-
tion occurs also without the influence of a disturbance on
the cylindrical surface. Hereby, the turbulence of the wind 5.2.2. Lower limit of critical velocity
flow and the roughness of the structure affect significantly Eq. (13) enables the numerical determination of the
the critical Reynolds number. upper limit velocities. The lower limit velocities result from
Regarding rain–wind induced vibrations, the flow tran- the analysis of the equilibrium state of the tangential forces
sition means that the fluid mechanical reasons disappear acting on the rivulets. Various parameters have to be con-
and the self excitation mechanism due to oscillating rivulets sidered for the calculation of the upper limit velocity.
is inactive. In the transition point the rivulets move from These are the inclination of cable, the spatial angle of inci-
the equilibrium state of the subcritical flow to the equilib- dence, the cable diameter as well as shape and size of the
rium position of the supercritical flow (see Fig. 6). Again, rivulet. Fig. 6 shows the lower limit of the critical velocity
the numerical results agree with the measured values by range for various cable inclinations a from 0 to 75 and
Hikami for wind velocities above of 13 m/s. It is important angles of incidence b0 from 90 to +90. The diameter
to note, that rain–wind induced vibrations do not exist in of cable is 150 mm and the rivulet has an unsymmetric
the supercritical flow, which is described by the given cross section.
mechanical model. According to Fig. 7 the equilibrium of the rivulets are
Considering Reynolds numbers within the interval of stable for negative spatial angle of incidence and instable
transition for the flow around a cylinder, an upper range for positive spatial angle of incidence. This is also observed
for the wind velocity can be determined, where rain–wind in nature and in wind tunnel tests [1,3,4]. The values for the
induced vibrations do not occur above. By use of the rela- critical velocity arise from the computation as very small,
tion for the Reynolds number the upper range follows to when the spatial angles of incidence are negative. In addi-
tion, with increasing inclination of the cable the critical
Re  mL
uupper
crit P : ð19Þ wind velocities decline, which is caused by the augmenta-
d tion of flow forces and the reduction of gravitational forces
where Re denotes the Reynolds number, mL the kinematic from the calculation of equilibrium positions. For a cable
viscosity of air and d the diameter of the cable. When real inclination of a = 90 (vertical cable) only flow forces exist
structures are regarded, the range of Reynolds numbers and the curve of the critical velocities is identical to a hor-
truncates to 1 · 105–2 · 105, since turbulence-free flow does izontal line at a critical wind velocity of u1 = 0 m/s. Thus
not occur in nature. Analogously, the critical diameter rain–wind induced vibrations in principle are possible on

Fig. 7. Critical velocities of rain–wind induced vibrations for different inclinations a of cable.
 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595 1591

α = 25˚, β0 = -19.525º (β = - 40º), D = 160 mm

measurement by Flamand
numerical results
0.25
vibrations caused
by lower and upper rivulet
0.2

amplitude in [m]
supercritical
subcritical flow
flow
0.15

vibrations caused
0.1
by lower rivulet

0.05

0
4 5 6 7 8 9 10 11 12 13
wind velocity u∞ in [m/s]

Fig. 8. Vibration amplitudes and wind velocity.

vertical cables and correspond with observations at real Hikami and Shiraishi [1] determines the influence of the
bridge structures by Verwiebe [2]. natural frequency of the rain–wind induced vibrations for a
cable. The measurements by Hikami are conducted for a
5.3. Vibration amplitudes and the influence of frequency cable model with natural frequencies of 1, 2 and 3 Hz
under the same Scruton number and rain condition. The
Fig. 8 demonstrates the measurements by Flamand [4] parameters of the HDPE – cable are given to d =
and the numerical results of the amplitudes of rain–wind 140 mm for the diameter, a = 45 for the yawed angle of
induced vibrations as a function of the wind velocity. The cable and b0 = 35.26 for the spatial angle of incidence
wind-tunnel tests by Flamand are based on a HDPE – cable of flow (with b = 45). Regarding these parameters the
model with a diameter of 160 mm, 0.16 kg/m for the mass, a computed critical onset velocity follows to 9.5 m/s, which
logarithmic decrement of 0.1 and eigenfrequency of 1.0 Hz. is independent on the natural frequency of the cable.
The other parameters of the HDPE – cable are given to Hikami observes the same phenomena in the wind tunnel
a = 25 for the yawed angle of cable and b0 = 19.52 for test. At this wind velocity rain–wind induced vibrations
the spatial angle of incidence of flow (with b = 40). occur and increase with the square of the wind velocity.
For the given cable parameters the critical wind velocity Fig. 9 presents the measurements by Hikami and the
for the lower rivulet follows to 5.5 m/s. The lower rivulet computed vibration amplitudes as a function of wind
causes a cable vibration, which starts with an amplitude velocity and for the three different natural frequencies of
of 0.03 m for a wind velocity of 6 m/s. The amplitudes 1, 2 and 3 Hz. The measured values are in good correspon-
are increasing for higher wind velocities whereby the dence with the numerical results, which are obtained by the
increase of the computed amplitudes develops with the proposed mechanical model on basis of the phenomena
square of wind velocity. The increase of the amplitudes of the Prandtl tripwire and the rivulets as movable
with the square of wind velocity is confirmed by wind tun- disturbance.
nel measurements of Schwarzkopf [20]. For a critical wind In accordance to the onset velocity the upper critical
velocity of 8.46 m/s the upper and the lower rivulet are wind velocity does not depend on the natural frequency
active. Both rivulets cause a rapid increase of vibration of the cable. Here, the upper critical wind velocity follows
amplitudes as demonstrated in Fig. 8. Again the vibration to 13 m/s and rain–wind induced vibrations terminates.
amplitudes increase with the square of wind velocity. The This effect is due to the natural transition of the subcritical
measurements by Flamand correspond to the results of to the supercritical flow for the Reynolds number of
computation for vibrations caused by the lower rivulet 1.2 · 105. The measurements by Hikami and the numerical
and also for vibrations caused by the upper and the lower results demonstrate that the natural frequencies of cable
rivulet (see Fig. 8). The vibration amplitudes are reduced do not influence the lower and the upper critical wind
significantly at the measurement by Flamand for a wind velocity of cable vibrations. However the vibration
velocity of 11.5 m/s and a Reynolds number of 1.2 · 105. amplitudes are influenced by the natural frequency. Low
For this Reynolds number the natural transition from the natural frequencies enable vibrations with large ampli-
subcritical to the supercritical flow takes place and the tudes, high natural frequencies lead to vibrations with
rain–wind induced vibrations terminate. small amplitudes.
1592 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595

α = 45º, β0 = -35.264º (β = - 45º), D = 140 mm

0.2
1Hz (numerical result)
1Hz (measurement by Hikami)
0.18
2 Hz (numerical result)
2 Hz (measurement by Hikami)
0.16
3 Hz (numerical result)
3 Hz (measurement by Hikami)
0.14
amplitude in [m]
0.12

0.1

0.08

0.06

0.04

0.02

0
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
wind velocity u∞ in [m/s]

Fig. 9. Vibration amplitudes and influence of frequency.

5.4. Numerical vibration analysis of Faro–Falster Bridge of rivulets at the cable surface. For inclinations of cable
from 80.2 to 26 the critical wind velocities follow to
A method for calculation of the critical velocity is shown 4.3–5.5 m/s for short cables and to 5.5–11.9 m/s for long
in Section 5.2. At the critical velocity the rivulet is posi- cables. The upper limit of the critical velocity yields 19–
tioned at the critical point of transition and initiates the 29 m/s (short cables) and 15–25 m/s (long cables). Above
self-excitation mechanism of rain–wind induced vibrations. this limit rain–wind induced vibrations do not exist.
It is important to analyse whether this mechanism enforces In the second step of calculation the vibration behaviour
or decays the oscillations. To solve this question coupled of the cables is investigated by means of the coupled non-
nonlinear equations of motion (see Section 3) are analysed linear equations of motion from Section 3. Forced oscilla-
and then applied to the example of Faro–Falster Bridge in tions parallel to the wind direction arise for a wind velocity
Denmark. At these bridge oscillations parallel to the wind of 10.5 m/s. Fig. 10(a) shows the amplitudes of vibrations
direction can be measured in rain [21] with large ampli- for the cable and the rivulets as a function of time. Thus
tudes up to 1.5 m for wind velocities from 10 to 15 m/s, a the amplitudes of rivulets are significantly smaller than
plane angle of incidence from 20 to 45 with respect the amplitudes of cable. Due to the inertia both dynamic
to the longitudinal axis of bridge. The oscillations perpen- responses possess a phase shift of a quarter period, which
dicular to the wind direction display small amplitudes induces a maximum of energy transfer into the structure.
of 0.1 m in the measurements and are of minor importance. In the steady state the amplitudes of cable are about
Parameters of cables are given to 60.4 kg/m or 115.3 kg/m 1.3 m in wind direction (see Fig. 10(b)). The amplitudes
for the mass, 160 mm or 200 mm for the diameter, 0.7– are increasing for higher wind velocities, so that observed
1.9 Hz for the eigenfrequency, and 0.03 for the logarithmic amplitudes of 1.5 m are possible.
decrement. Fig. 11 illustrated the nonlinear behaviour of the oscilla-
In the first step of calculation the critical velocities are tions, which converges to a limit cycle finally. Here, the
analysed under consideration of the tangential equilibrium response frequencies appear in the interval from 0.3 to

0.4
cable 1
0.2 rivulet
x in m,ϕ in rad

x in m,ϕ in rad

0.5
0 0

-0.2 -0.5

-0.4 -1
0 20 40 60 80 100
0 200 400 600 800 1000
a time ins b time in s

Fig. 10. Vibration of the cable and rivulets parallel to wind direction: u1 = 10,5 m/s.
 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595 1593

x in m, ϕ in rad -0.02

-0.04

-0.06

-0.08

-0.1
0 200 400 600 800 1000
time in s

Fig. 11. Vibration of the cable and rivulets parallel to wind direction:
u1 = 10 m/s.

2
Fig. 13. Limit cycle of a Vibration parallel to the wind direction: u1 = 10,
1 5 m/s.
y in m, ϕ in rad

-1 tion behaviour of the hangars in detailed wind-tunnel

experiments, because rain–wind induced vibrations with
-2 large amplitudes have been caused significant damage of
0 200 400 600 800 1000
time in s structural components. The hangars are built with an
inclination of 79, a length up to 25.76 m and a diameter
Fig. 12. Vibration of the cable and rivulets perpendicular to wind
direction: u1 = 20, 83 m/s. up to 130 mm. The wind-tunnel tests are based on cable
models with a diameter of 110 mm, a reduced mass
between 3.6 and 9.4 kg/m, logarithmic decrements of
0.9 Hz. Oscillations of cable decay for wind velocities smal- 0.056 and 0.061 and eigenfrequencies of 3.36 and
ler than 10.5 m/s (Fig. 11). Consequently the excitation 4.0 Hz. Based on these parameters the hangar vibrations
mechanism causes damping in the system below the critical are investigated using the mechanical model from Section
wind velocity of 10.5 m/s. 3. Experimental and numerical analysis show that domi-
Forced oscillations perpendicular to the wind direction nant vibrations parallel to the wind direction exists for
exist above a critical velocity of 20.83 m/s and can lead low wind speeds. In the range of high wind speeds the
to amplitudes of 2.0 m (see Fig. 12). Below the critical cable vibrations show large amplitudes perpendicular to
velocity the excitation mechanism acts as a damper for the wind direction.
the cable vibrations. The monitoring of Faro–Falster Fig. 14 presents the amplitudes of the cable and the riv-
Bridge yields similar results, where significantly enforced ulets for dominant vibrations parallel and perpendicular to
oscillations perpendicular to the wind direction cannot be the wind direction for 15 and 25 m/s. Here, the amplitudes
observed for wind velocities from 10 to 15 m/s. of the nonlinear vibration behaviour converge to a limit
The example of Faro–Falster Bridge indicates that rain– cycle after approximately 200 s. Oscillations of cable decay
wind induced vibrations can lead to oscillations with large for wind velocities smaller then 10 m/s, because the excita-
amplitudes. Other structures may show oscillations with tion mechanism causes damping in the vibration system
smaller amplitudes and larger frequencies, in dependence consisting of cable and rivulets.
of the specific structural properties. Fig. 15 shows frequency response functions for different
The kind of instability can be also identified by means of wind velocities. Fig. 15(a) depicts the frequencies for vibra-
the numerical vibration analysis. For rain–wind induced tions parallel to the wind direction for a wind velocity of
vibrations self-excitation phenomena occur only, if a min- 15 m/s, whereby Fig. 15(b) illustrates the frequencies for
imum of energy transfer from the wind flow is guaranteed vibrations perpendicular to the wind direction for a wind
(strong self-excitation). Here, an inner instable and an velocity of 25 m/s. Both frequency response functions are
outer stable limit cycle exist, as illustrated in Fig. 13 for governed by a dominant fundamental frequency and over-
vibrations parallel to the wind direction. tones from the first to the third order.
The existence of overtones is typical for the nonlinear
5.5. Analysis of vibration frequencies of Dömitz Bridge characteristics of rain–wind induced vibrations. The first,
the second and the third overtone appear in the frequency
Finally the analysis of vibration frequencies is demon- analysis of vibrations parallel to the wind direction. For
strated for the Dömitz Bridge across the river Elbe in Ger- amplitudes perpendicular to the wind direction a first and
many. The Dömitz Bridge is an arch bridge with inclined a second overtone exist, whereby the concentration of
arches and solid steel hangars. The structural properties energy for the second overtone is higher than for the first
are given in [22]. Lüsse et al. [22] investigated the vibra- one.
1594 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595

0.6
0.2 0.4

x/d in [–], ϕ in [rad]

y/d in [–], ϕ in [rad]

0.2
0
0

-0.2 -0.2
-0.4
-0.4 -0.6
0 200 400 600 800 1000 0 200 400 600 800 1000
time in [s] time in [s]

Fig. 14. Vibrations parallel to wind direction (u1 = 15 m/s) and perpendicular to wind direction (u1 = 25 m/s).

0.5
0.5
0.4
amplitude in m

amplitude in m
0.4
0.3
0.3
0.2 0.2
0.1 0.1

0.25 0.5 0.75 1. 1.25 1.5 1.75 2. 2.25 2.5 0.25 0.5 0.75 1. 1.25 1.5 1.75 2. 2.25 2.5
a frequency in Hz b frequency in Hz

Fig. 15. Vibration frequencies of cable parallel (u1 = 15 m/s) and perpendicular to wind direction (u1 = 25 m/s).

Fig. 16 shows the response frequency as a function of Another phenomenon is the continuous change of the
the wind velocity. In the range of low wind speed the figure frequency for vibrations parallel to the wind direction to
demonstrates the frequency for the vibration parallel to the vibrations perpendicular to the wind direction. The reason
wind direction. In the range of higher wind speed the figure of this phenomena seems to be the excitation mechanism,
presents the frequency for the vibration perpendicular to which is similar for rain–wind induced vibrations parallel
the wind direction. The fundamental frequency correlates and perpendicular to the wind direction. The overtones
to a linear function of the wind velocity for the vibration show the nonlinear characteristics of the frequency as a
parallel to the wind direction and also for the vibration per- function of the wind velocity. In case of vibration parallel
pendicular to the wind direction. This property is similar to to the wind direction the frequency of the first overtone
vortex induced vibrations. In contrast, the gradient for vor- seems to be a linear function of wind velocity and a nonlin-
tex induced vibrations is steeper than for rain–wind ear function for the higher overtones. The fundamental fre-
induced vibrations. quencies of cable vibration appear in the interval from 0.5

α = 79˚, β = -90˚, D = 110 mm

3
vibration parallel Uα vibration perpendicular Uα

to wind direction to wind direction

2.5

3rd overtone 2nd overtone

2
frequency f in [Hz]

2nd overtone
1.5
1st overtone

1 1st overtone

fundamental frequency
0.5
fundamental frequency

0
0 10 20 30 40
wind velocity u∞ in [m/s]

Fig. 16. Frequency and wind velocity.

 C. Seidel, D. Dinkler / Computers and Structures 84 (2006) 1584–1595 1595

to 0.9 Hz and the first overtone in the interval between 1.0 Bridge. Thereby the fundamental frequency correlates to
and 1.8 Hz. In contrast to positive gradients of the funda- a linear function of the wind velocity for vibrations parallel
mental frequency and the first overtone, the second over- and perpendicular to the wind direction. The overtones
tone decreases for velocities higher than 32 m/s. show the nonlinear characteristics of the frequency as a
function of the wind velocity.
6. Conclusions
References
The paper presents a possible fluid-mechanical interpre-
[1] Hikami Y, Shiraishi N. Rain–wind induced vibrations of cables in
tation of rain–wind induced vibrations, based on the phe- cable stayed bridges. J Wind Eng Ind Aerodyn 1988;29:409–18.
nomena of the Prandtl tripwire under consideration of [2] Verwiebe C. Erregermechanismen von Regen-Wind-induzierten Sch-
the rivulets as a movable disturbance. Based on this inter- wingungen. In: Peil U, editor. Baukonstruktionen unter Wind-
pretation a mechanical model is deduced, which can einwirkung (TU Braunschweig 1997). Aachen: Windtechnologische
Gesellschaft e.V.; 1998.
describe different phenomena of rain–wind induced vibra-
[3] Bosdogianni A, Olivari D. Wind- and rain-induced oscillations of
tions. The numerical results are computed by this mechan- cables of stayed bridges. J Wind Eng Ind Aerody 1996;64:171–85.
ical model and demonstrate good correspondence to the [4] Flamand O. Rain–wind induced vibration of cables. J Wind Eng Ind
wind tunnel tests and the measurements at real structures. Aerody 1995;57:353–62.
To determine aerodynamic coefficients a theoretical [5] Matsumoto M, Shiraishi N, Shirato H. Rain–wind induced vibration
of cables of cable-stayed bridges. J Wind Eng Ind Aerody 1992;41–
approach is developed on the basis of the phenomena of
44:2011–22.
the Prandtl tripwire. The computed aerodynamic coeffi- [6] Yamaguchi H. Analytical study on growth mechanism of rain
cients approximately correspond to the measured aerody- vibration of cables. J Wind Eng Ind Aerody 1990;31:73–80.
namic coefficients. Therefore the computed aerodynamic [7] Honda A, et al. Wind tunnel test on rain-induced vibration of the
coefficients are well suited to describe the influence of a dis- stay-cable. In: Proceedings of the international symposium on cable
dynamics. Liege; 1995.
turbance at a cable in wind flow.
[8] Ruscheweyh HP. The mechanism of rain–wind induced vibration. In:
The lower limit of the critical velocity range is deter- Proceedings of wind engineering into the 21st century. Rotterdam;
mined by the mechanical model for various cable incli- 1999.
nations and angles of incidence. Furthermore, a fluid [9] Seidel C, Dinkler D. Regen-Wind induzierte Schwingungen – Ursache
mechanical interpretation is given for the existence of the und Beschreibung. In: Proceedings of GAMM. Augsburg; 2002.
[10] Seidel C, Dinkler D. Phänomenologie und Modellierung Regen-Wind
upper limit, which bases on the natural transition from
induzierter Schwingungen. Bauingenieur 2004;79:145–54.
subcritical to supercritical flow. The approximation of the [11] Prandtl L. Der Luftwiderstand von Kugeln. Nachr. Der Gesellschaft
upper limit velocities and the determination of the critical der Wiss. Zu Göttingen, Math.-physikalische Klasse, 1914; 177–90.
cable diameters can be derived from established fluid- [12] Zierep J. Ludwig Prandtl, Leben und Wirken. In: Meier GEA, editor.
mechanical correlations. Numerical investigations show Ludwig Prandtl, ein Führer in der Strömungslehre. Braunschweig:
Vieweg; 2000.
that rain–wind induced oscillations also occur on vertical
[13] Geurts C, Vrouwenvelder T, van Staalduinen P, Reusink J. Numer-
cables. ical modelling of rain–wind-induced vibration: erasmus bridge,
The computation of the vibration amplitudes as a Rotterdam. Str Eng Int 1998;2:129–35.
function of the wind velocity shows that the computed [14] Van der Burgh AHP, Hartono J. Rain–wind-induced vibrations of a
amplitudes increase with the square of wind velocity. Fur- simple oscillator. Int J Non-Linear Mech 2004;39:93–100.
[15] Wang LY, Xu L. Analytical study of wind–rain-induced cable
thermore, the numerical investigations indicate the influ-
vibration: 2DOF model. Wind Struct 2003;6(4):291–306.
ence of the rivulets at the vibration amplitudes, e.g. for a [16] Peil U, Nahrath N. Modeling of rain–wind induced vibrations. Wind
wind tunnel test by Flamand [4]. Measurements by Hikami Struct 2003;6(1):41–52.
and Shiraishi [1] and numerical results demonstrate that [17] Sockel H. Aerodynamik der Bauwerke. Braunschweig: Vieweg;
the natural frequencies of cable do not influence the lower 1984.
[18] Ruscheweyh H. Dynamische Windwirkung an Bauwerken. Wiesba-
and the upper critical wind velocity of cable vibrations. For
den: Bauverlag; 1982.
wind tunnel tests both by Hikami and by Flamand, the [19] James F, Truong Q. Wind load on cylinder with spanwise protrusion.
rain–wind induced vibrations terminate for the Reynolds Eng Mech Div 1972;98:1573–89.
number where the natural transition from subcritical to [20] Schwarzkopf D. Regen-Wind induzierte Schwingungen – Grundlagen
supercritical flow takes place. und ein Berechnungsmodell. Dissertation. RWTH Aachen, 2004.
[21] Langsoe HE, Larsen OD. Generating mechanisms for cable stay
For the Faro–Falster Bridge, the numerical vibration
oscillations at the faro bridges. In: Proceedings of the international
analysis is carried out for the cable vibrations parallel conference on cable-stayed bridges. Bangkok; 1987.
and perpendicular to the wind direction. The analysis of [22] Lüsse G et al. Regen-Wind-induzierte Schwingungserscheinungen an
the vibration frequencies is demonstrated for the Dömitz der Elbebrücke in Dömitz. Stahlbau 1996;65(3):105–14.