Sei sulla pagina 1di 25

Journal of Wind Engineering and Industrial Aerodynamics, 12 (1983) 49--73 49

Elsevier Science Publishers B.V., Amsterdam - - P r i n t e d in The Netherlands

ACROSS-WIND VIBRATIONS OF STRUCTURES OF CIRCULAR


C R O S S - S E C T I O N . P A R T I. D E V E L O P M E N T O F A M A T H E M A T I C A L
MODEL FOR TWO-DIMENSIONAL CONDITIONS

B.J. VICKERY and R.I. BASU


Faculty of Engineering Science, University of Western Ontario, London, Ontario (Canada)
(Received May 10, 1982; accepted in revised form October 26, 1982)

Summary

A model is presented for predicting the across-wind response of constant-diameter


circular cylinders vibrating in a mode of uniform amplitude and subject to uniform flow.
A key feature of the model is the representation of all motion-dependent phenomena by
a nonlinear aerodynamic damping force. This force coexists with the fluctuating force which
arises from vortex shedding on a stationary cylinder, and the two forces are assumed to be
uncorrelated.
The ability of the device used in representing the motion-induced force to model certain
aeroelastic characteristics associated with vibrating cylinders is demonstrated. The device
is shown to be capable of successfully reproducing two effects; namely, the increase of the
spanwise correlation of forces with increasing amplitude, and the phenomenon of "lock in"
where the shedding frequency is apparently dictated by the vibration frequency.
The model is developed within the framework of random-vibration theory, and a num-
ber of simplifying assumptions are necessary to incorporate the nonlinear aerodynamic
damping force and also to account for the influence of turbulence. Numerical experiments,
undertaken to examine the nature of the approximations involved in the assumptions adopt-
ed, are described. The results of the numerical experiments are very encouraging and
justify the simplifications made in the modelling process.

Notation

A p a r a m e t e r in e q n . ( 1 9 )
B dimensionless measure of bandwidth of vortex-shedding force spectrum
B0 bandwidth B for smooth flow
C p a r a m e t e r in e q n . ( 3 1 )
cL sectional r.m.s, lift coefficient
cL r.m.s, lift coefficient measured over length of cylinder
D diameter
f frequency
f0 natural frequency
f~ shedding frequency
g(s) narrow-band process representing smooth-flow vortex shedding (see
eqn. (28))

0304-3908/83/$03.00 © 1983 Elsevier Science Publishers B.V.


50

G constant (see eqn. (21))


h length of cylinder
ha component of aerodynamic force in phase with displacement
Ha dimensionless ha (see eqn. (12))
i measure of turbulence intensity (see eqn. (28))
ka component of aerodynamic force in phase with velocity
Ka aerodynamic damping parameter (= mVa/pD 2)
Ks structural damping parameter (= m~Ts/pD2)
L integral length scale of turbulence of longitudinal component
m mass per unit length
r distance separation in terms of diameter
Ro spanwise correlation of lift forces on stationary cylinder
Ry spanwise correlation of lift forces on vibrating cylinder
S Strouhal number
u(t) broad-band process representing large-scale turbulent fluctuations
(see eqn. (28))
r.m.s, longitudinal turbulence fluctuations
U mean wind speed
v~ reduced velocity (= U/foD )
Vcrit critical velocity (= roD/S)
w(t) fluctuating lift force per unit length
Y displacement
YL limiting r.m.s, displacement
Y amplitude
YL limiting amplitude
Z spanwise distance
parameter defining length scale in spectrum of turbulence (eqn. (29))
P density of air
~?a aerodynamic damping as a fraction of critical
r/s structural damping as a fraction of critical
r.m.s, quantity

1. Introduction

The dynamic response of tall slender structures of circular cross-section to


atmospheric turbulence involves a number of fluid-~tructure interaction
phenomena. It has been found convenient, and justifiable, to treat the response
in the along-wind and across-wind directions separately. The theory for the
along-wind response to the longitudinal component of atmospheric turbulence
is relatively well developed. Work undertaken in the 1960's (see for example,
refs. 1--4) has led to analytical methods which are now well est~bliahed. These
theories form the basis of calculation methods for predicting the along-wind
response in many current wind codes.
In comparison, a theory for the across-wind response has proved much more
difficult to develop. In the case of response in the along-wind direction, a single
51

mechanism, that of buffeting, is by far the dominant source of excitation,


whereas several mechanisms play significant roles in the across-wind response.
The principal source of excitation arises from the pressure fluctuations associat-
ed with the shedding of vortices. If the structure is allowed to vibrate, addi-
tional motion-dependent forces are induced. The buffeting force arising from
the lateral c o m p o n e n t of turbulence in the oncoming flow is yet another
source.
This paper, together with a companion paper [5], describe a semi-empirical
mathematical model capable of yielding estimates of the across-wind response
of structures of circular cross-section. The discussion here is, in general, con-
fined to rigid, constant-diameter cylinders vibrating in a mode of uniform am-
plitude and subject to uniform flow; the companion paper reports further
developments of the model necessary such that predictions can be made of the
response of slender structures in full-scale situations. This extension essentially
involves including the buffeting force which arises from lateral turbulence, and
accounting for variations with height of diameter, mode shape and flow
properties.
In smooth flow, measurements of the across-wind force on vibrating
cylinders of circular cross-section show the existence of, in general, two fluc-
tuating forces. The first arises from the shedding of vortices, and exists whether
the cylinder is vibrating or not, and the second is associated with the motion of
the cylinder. The response of circular cylinders to vortex-shedding forces alone
can be treated within the framework of linear random-vibration theory as long
as amplitudes are small. At greater amplitudes the contribution to the forcing
from motion-induced forces becomes significant and must be accounted for in
any realistic model of response.
A number of approaches for modelling the nonlinear behaviour of the force
have been proposed. Principal among these is the model proposed by Hartlen
and Currie [6] ; the development of their concept results in two coupled ordi-
nary differential equations, the first of which is essentially the standard, single-
degree-of-freedom equation of motion. The second equation, in which the
variable is the lift coefficient, is a second-order differential equation nonlinear
in the first derivative. The coupling arises out of the presence of cylinder velo-
city in both equations.
The second equation described in the previous paragraph is, in form, a
modified version of the van der Pol equation. Using this as a basis, Skop and
Griffin [7] developed a model applicable to flexible circular cylinders, whereas
earlier versions were restricted in application to the vibration of spring-mount-
ed rigid cylinders. This latter work has been modified by ESDU [8] such that
predictions of response can be made for structures of circular cross-section in
typical full-scale configurations. In the model developed by ESDU two distinct
regimes are identified; they are termed "broad-band" and "narrow-band", and
different response equations are developed for each regime. For a given struc-
tural and fluid-flow configuration, the applicable regime is determined by the
equation predicting the higher response. The "broad-band" regime model
52

ignores forces associated with motion, whereas the "narrow-band" regime


model includes them. Therefore it is not surprising that the transition between
the two regimes is awkward and, to some extent, arbitrary.
The essential feature of the model presented herein is the representation of
all motion-dependent phenomena by a nonlinear damping force superimposed
upon a narrow-band random force induced by vortex shedding from a station-
ary cy_linder. This simple model exhibits a smooth transition between the two
regimes and provides response estimates over all amplitude levels.
The model is developed within the framework of random-vibration theory.
In order to incorporate the nonlinear motion-dependent force in the mathe-
matical model, certain simplifying assumptions are made. Further simplifica-
tions are necessary to account for the influence of turbulence. The nature of
the approximations involved is examined by numerically solving the nonlinear
equation of motion and comparing the results with those predicted by the
mathematical model.

2. Vortex-shedding forces on stationary cylinders

As fluid flows past a bluff body such as a circular cylinder, vortices are shed
alternately from each side at most values of the Reynolds number. The shed-
ding of vortices results in a changing pressure distribution, causing an alternat-
ing side or lift force. The frequency at which this fluctuating lift force is
centred is given by the well-known relationship
fs=SV/D (t)
where fs is the shedding frequency (Hz), U the mean wind speed, and D the
diameter; S, the Strouhal number, is essentially constant for wide ranges of
Reynolds number.

urbulence

fs I:requency
Fig. 1. Form of spectrum of lift force due to vortex shedding.
53

The fluctuating lift force may be written as


~ = '/~ p U2DCL (2)
where ~ is the r.m.s, force per unit length, CL the r.m.s, lift coefficient, and
p the density of air. Measurements of the fluctuating lift force show that the
force is not strictly periodic. Even when the flow is smooth and uniform, the
spectrum of the fluctuating lift force contains a spread, though narrow, over
frequencies adjacent to fs (see Fig. 1).
The form taken by the spectrum of the lift force in turbulent flow is suggest.
ed by the nature of atmospheric turbulence. Of particular interest here is that
part of the spectrum of atmospheric turbulence containing wavelengths many
times greater than the diameter of the structure, say by a factor of ten or
more. Large-scale turbulence can be regarded as a slowly varying mean wind
speed which has the effect of altering fs, the shedding frequency, according
to the relationship given by eqn. (1). Thus if the fluctuating lift force were
actually sinusoidal in smooth flow, Gaussian turbulence would cause the
spectrum of the lift to take a Gaussian form. One such form which is found to
fit experimental data well [9] is
fSCT,(f)/C~ = (f/V"~Bfs) exp { - [(1 - f/fs)/B] 2} (3)

where f is frequency and B bandwidth. The bandwidth is a measure of the


spread of the spectrum of lift. If the lift in smooth flow were sinusoidal, then
the following relationship between bandwidth and turbulence intensity would
hold:
B = V~(ulU) (4)
where ~ is the r.m.s, longitudinal component of the wind.
Since the spectrum of lift in smooth flow has finite width, eqn. (4) requires
modification. Limited experimental data suggest the expression
B 2 =B20 + 2(~/U) 2 (5)
where B0 is the bandwidth in smooth flow, with typical values in the range
0.05- 0.1.
The discussion so far in this Section has been confined to the nature of the
fluctuating lift force at a particular section. In general, the lift forces exhibit
three-dimensional characteristics, that is, the lift forces in the spanwise direc-
tion are not perfectly correlated. The correlation is c o m m o n l y expressed in
terms of the number of diameters in the spanwise direction over which the lift
forces can be regarded as fully correlated. In most experimental work, pressure
correlations are measured which are related closely to lift-force correlation.
Considerable scatter is evident in the data: recent summaries [10, 11] show a
variation, for Reynolds number greater than ~ 10 4, from six diameters down
to one-half a diameter. The presence of turbulence will, in general, reduce the
value of the correlation length compared to the smooth-flow value.
54

Limited measurements [9] suggest that a suitable form is


R o ( z l , z : ) = cos(2r/3) exp[- (r/3) 2] (6)
where z is the spanwise distance and r = 2 Jzl - z : l / [ D ( z l ) + D(z:)]. Equation
(6) implies a correlation length of a b o u t one diameter.

3. The nature o f motion-induced forces

The dynamic response of structures and components of circular cross-section


has been the subject o f intensive research for a number of decades. This is not
surprising in view of the practical implications of such research. Structures and
structural components in a variety of industries suffer vortex-induced vibra-
tions; examples include chimneys, towers, transmission-line conductors, marine
piles~ reactor rods, distillation columns, etc.
Motion has the effect of inducing certain aeroelastic effects. The complex
nature of the f l u i d - s t r u c t u r e interactions involved precludes a direct approach
to the problem o f modelling the behaviour observed in numerous experiments.
The majority of experiments on laboratory scale have been undertaken
under two<iimensional conditions; usually, rigid cylinders elastically m o u n t e d
and subject to uniform steady flow. A recent review by Sarpkaya [12] has
summarized, in a comprehensive manner, the principal findings of such experi-
ments.
Before considering these further it is appropriate to define two dimension-
less parameters, both of which are useful in characterizing vortex-induced
vibrations. The first is the reduced velocity

Ur = U / f o D (7)
where f0 is the natural frequency (or the frequency of vibration in experiments
where the cylinder is forced to vibrate mechanically); and the second is the
mass-damping parameter
K s = m~?s/pD 2 (8)
where ~s is the structural damping as a fraction of critical damping, and m is
the mass per unit length.
When the shedding frequency fs is in the vicinity of the natural frequency,
f0, vortex-induced vibrations are likely to occur. Maximum amplitudes are
typically attained for 5.5 < Ur < 6.5, although significant excitation is present
in the range 4.75 < Ur < 8 [12]. At sufficiently low values of Ks, and with
Ur in the range 5.5--6.5, large amplitudes of vibration are set up and a p h e n o m -
enon known as "lock in" occurs, where the shedding frequency "locks in" to
the natural frequency of the cylinder for a range of velocities, in apparen~
violation of the Strouhal relationship given b y eqn. (1).
The principal characteristics of the "lock in" condition are an increase in
magnitude of the fluctuating lift forces and an increase in their spanwise corre-
lation. These features can be incorporated in a mode] for the response by in-
troducing a negative aerodynamic damping term. In the following paragraphs
it is shown how, in a qualitative sense at least, such a device does reproduce
the features described above.
Scruton and co-workers (e.g., ref. 13) proposed expressing the aerodynamic
lift force per unit length as the sum of a displacement-dependent term and a
velocity-dependent term, as
w(t) = hay + k a y (9)
where h a and k a are coefficients for the in-phase and out-of-phase components,
respectively, of the aerodynamic force. Assuming simple harmonic motion, of
the form
y = Y sin 2nf0 t (10)
and substituting into eqn. (9) yields
w(t) = h a Y sin 2nfot + ka27rfoY cos 2~fot (11)
It is convenient to nondimensionalize h a and tea by making the following
substitutions in eqn. (11):
h a = 87r 2 f2o pD2Ha
lea = 4 7rfopD 2 Ka (12)
which yields
w( t) = 8 n2 f~ p D2 Ha Y sin 2 nfot + 8 n2f~ pD2Ka Y cos 2 rrfot (13)
This formulation of the lift force, as expressed by eqn. (9)~ makes no provision
for forces which exist when the cylinder is stationary. Thus a term is added to
allow for this additional force, giving
w(t)=1/2pU2DCL(t)+ 87r2f:opD2(HaYsin27rfot + K a Y c o s 2 7 r f o t ) (14)

where CL(t ) is the instantaneous lift coefficient. In the discussion that follows,
for the sake of simplicity the configuration assumed is that of a rigid cylinder
vibrating in a uniform-amplitude mode and, unless noted otherwise, subject
to uniform smooth flow.
It is further assumed that the forces caused by vortex shedding (the first
term on the right-hand side of eqn. (14)} are uncorrelated with the forces in-
duced by motion (the second term); this assumption is considered further in
the next Section. Before proceeding further with this development, it is
necessary to consider the nature of the parameters H a and K a.
Measurements of the value of K a and its variation with Ur have been present-
ed, sometimes implicitly, in a number of papers. The experiments from which
data for K a are available were conducted, with a few exceptions, in subcritical
smooth-flow conditions. In general K a is a function of amplitude of vibration
but can be considered nearly constant for normalized amplitudes Y]D of up to
~0.2. The nature of the variation of Ka with amplitude is considered more
properly in a later Section. K a is denoted by Ka0 at small amplitudes.
56
Re|ercllce
,:"I l:
,e.. • I~
/
20 l
t
i
Iq
i 20
KaOMAX

M;]dl N~, • 0 2~
©
\
\
1.0
\

\\ \ /
//"J-~ o
',l,n IT ~,+, U 2S
/
I~orm Interred Fnnn |'ig b \ /
of Parl 2 1 ~ I

I ,2 I ...... J
10 10n 10~ 10~' I0

REYNOLDS NUMBER

Fig. 2. Variation of aerodynamic damping with Reynolds number (after Basu and Viekery
[14]).
IO
~ ~ Rclcrcnc'e
22
23
...... 2~
...... 17
16

Kao
l! I,'\\
Ka.MAX 04
I
\\
I
0.2
I1'
I
I

I.I 12 1,3 "" ~ ~1"~ ~ ----''"~. --_ I.T


I I I ~ i J " -'~-.- i

U/UcriI

-0 2
,i
Fig. 3. Variation of KaolKao.maxwith UIUcrlt(after Viekery [21] ).

The variation of Ka~ ~ with Reynolds number is shown in Fig. 2, which is


based on data assembi~¢~'[14] from a number of experiments [ 1 5 - 2 0 ] . Figure 3
shows KaJKa, -affi as a function of us~roD, based on a compilation by ViCkery
[21] of d a t a ~ a number of studies [16, 17, 22, 23]
57

The influence of large-scale turbulence can be regarded as a slowly fluctuat-


ing mean wind speed. As a first approximation, the modulating influence of the
turbulence can be.accounted for by replacing the value of Ka0 (Fig. 3) in smooth
flow by its average value in turbulent flow. The latter can be obtained from
the integration of the product of Ka0 in smooth flow and the Gaussian proba-
bility-density function of wind speed; Fig. 4 shows the resulting Kao for various
values of turbulence intensity. It should be noted that the smooth-flow Ka0 in
Fig. 4 is a slightly modified version of the smooth-flow Ka0 shown in Fig. 3;
the modification simplifies certain computational procedures.
The parameter H a has received far less attention compared with Ka0 • H a is
representative of the aerodynamic force in phase with the motion (see eqns.
(10)--(12)) and therefore modifies the stiffness (or mass) of the structure. The
value of the stiffness force resulting from aerodynamic sources in the wind is
negligible compared with the mechanical stiffness derived from the elasticity
of the structure. However, the magnitude of H a is comparable to the magnitude
o f K a.

....... 0 2
. . . . . . 0

K~

<,+ I I . .... --..~, \

,}"
o 2
/
/ / .i // ~ ~ ~ -

.-... ,/ , , ,
ON 09 I 0 I1 1 2 I 1 1.4 I $ I 6 I ~

{I I Ci. il
Fig. 4. I n f l u e n c e o f large-scale t u r b u l e n c e o n K%/Kao,max.

3.1 Effect of amplitude on spanwise correlation of lift force


Earlier it was pointed out that one effect of the motion of a cylinder is to
magnify the correlation of spanwise forces compared with the correlation found
on a stationary cylinder. Some insight into the relationship between vortex-
shedding forces and forces induced by motion can be gained by considering
the spectral forms of the forces. Figure 5 (after ref. 23) shows the spectral
forms of the forces for various values of fo/fs. For values of fo/fs differing from
58

2
~ ~

~- -
.... 0.7

o °s9

10 20 30 fo 40 50 60

I- requenc~ l H /

Fig. 5. Spectral form of lift force for varying lolls (after Szechenyi and Loiseau [23] ).

unity, two distinct peaks in the spectrum of the lift force are apparent; in these
regions the assumption that the two forces are uncorrelated is justified. When
fo/fs ~- 1, a single peak is visible and it is not possible to distinguish the two
forces, tn this condition, and when amplitudes are sufficiently large, it has been
c o m m o n l y accepted that a solitary force exists, fluctuating at the frequency
of vibration. Nevertheless, the following alternative description does not
contradict the observations: two separate fluctuating forces exist at all times,
but at large amplitudes the motion-induced force dominates, and engulfs,
the vortex-shedding forces. In the present model both forces are accounted for
under all conditions and are assumed to be uncorrelated even for/Co -~ fs.
Further, since the mathematical model is intended to reproduce only the gross
features of vortex-induced motion and not the detailed physics, the precise
nature of the forces when f0 -~ fs is not of great concern.
The assumption of no correlation between the " s t a t i o n a r y " vortex-shedding
forces and the motion-induced forces allows the mean-square value of fluctuat-
ing lift per unit length to be written as
~2 = (impU2D~L)2 + 2[(2ufo)2 pD2Ha Y] 2 + 2[(2nf0): pD2Ka Y] : (15)
where CL is the r.m.s, lift coefficient. The covariance of the lift forces between
two points along the length of a cylinder is given by
w(zl)w(z2) = (%pU2CL) 2 Ro(zl, z2) + 2[(2-f0) 2 pD2HaY] 2
+ 2[(2~f0): pD2KaY] 2 (16)
where the overbar denotes averaging over time, and Ro(zl,z2) is the correlation
coefficient for vortex-shedding forces on the cylinder when stationary. Re-
arranging eqn. (16) yields the more convenient f o r m
59

2 ~2
W(Zl)W(Z2) = (lhp U2DCL) 2 {R0(zl, z2) + 8(27troD~U) 4 ( Y / D ) 2 [(H~ + Ka)/C~]
(17)
Dividing eqn. (17) b y (15) gives t h e c o r r e l a t i o n c o e f f i c i e n t f o r a v i b r a t i n g
cylinder:
Ry(ZI, z2) -- W(Zl)W(Z2)/~v 2
= [R0(zl, z2) + (a Y / D ) 2]/[1 + (aY/D) 2] (18)

w h e r e a = 2~/2(2rrfoD/U) 2 [ ( H :a + Ka)/CL] 2 ~2 vs.


A n u m b e r o f studies have b e e n c o n d u c t e d w h e r e t h e i n f l u e n c e o f a m p l i t u d e
on c o r r e l a t i o n has b e e n m e a s u r e d . T h e f o r m o f eqn. (18) m a y b e c o m p a r e d
with t h e results f r o m t w o studies [24, 25]. T a b l e I s u m m a r i z e s s o m e o f t h e
p a r a m e t e r s o f t h e e x p e r i m e n t s and also t h e results o f an exercise t o e x a m i n e
t h e f o r m o f eqn. (18); a d e s c r i p t i o n o f the p r o c e d u r e a d o p t e d follows. T h e
least-squares m e t h o d is applied to t h e d a t a in o r d e r t o establish a value f o r a
in eqn. (18). Where possible t h e required d a t a for s u b s t i t u t i o n into eqn. (18)
are e x t r a c t e d f r o m t h e p a p e r s c o n c e r n e d . T h e value o f H a is a s s u m e d to be
zero, since n e a r f0 = fs e x p e r i m e n t a l results have s h o w n t h a t H a ~ K a (see for
e x a m p l e , ref. 16). T h e r.m.s, lift c o e f f i c i e n t is a s s u m e d t o be 0 . 4 2 [8] in t h e
analysis o f T o e b e s ' results [24]. In t h e case o f t h e results o f N o v a k a n d T a n a k a
[25], CL was e s t i m a t e d b y integrating t h e f l u c t u a t i n g pressure d i s t r i b u t i o n ,
t h e n s u b t r a c t i n g t h e l o w - f r e q u e n c y c o n t r i b u t i o n s , and, finally, m u l t i p l y i n g
b y a f a c t o r to allow f o r i n c o m p l e t e c i r c u m f e r e n t i a l c o r r e l a t i o n (~ 0.8 f o r sub-
critical flow; see ref. 14). This p r o c e d u r e yields values f o r CL o f 0.42 and
0.43 f o r s m o o t h a n d t u r b u l e n t flow, respectively, w h i c h are in close a g r e e m e n t
with t h e value o f CL r e c o m m e n d e d b y E S D U [8] f o r a R e y n o l d s n u m b e r o f
1.9 X 104 .
With the value of a established by the least-squares method, the substitution
of appropriate values of ]Co and D together with CL in the expression for a
following eqn. (18) yields a value for Ka0. The values of Ka0 thus obtained are
termed "implied" and are shown in Table I. Also shown are values of Kao taken

TABLE1

Parameters of experimental measurements of correlation on vibrating cylinders and com-


parison with predicted values of K a at f0 - fs

Reference Reynolds eL a Ka
number
Implied Predicted

Toebes [24] 8.5 × 104 0.42 14.4 1.05 1.3

Novak and Tanaka [25]


Smooth flow 1.9 x 104 0.42 14.0 1.63 1.73
Turbulent flow 1.9 X 104 0.43 8.8 0.67 0.85
60

1.0

0.8
o o o o Y;D = 0.1 o

-- 0.6
t.

~9
_: 0.4
N

t~ 0.2
~ gqn. tl81 ~ 0

1' "~ 3' 4 5 6 -7 8' 9 10

Separa rio n ~Diameters )


F i g . 6. E f f e c t o f v i b r a t i o n a m p l i t u d e on correlation (experimental data from Novak and
Tanaka [25] ).

from Fig. 2; these are appropriate for smooth flow. The results o f Novak and
Tanaka [25] include measurements made in turbulent flow with a turbulence
intensity of 11%. Reference to Fig. 4 shows that this level of turbulence has the
effect of roughly halving the value of Ka0 in smooth flow. Figure 6 s h o w s a
comparison of the correlation represented by eqn. (18) with measured values
of correlation using the smooth-flow results of Novak and Tanaka [25] as an
example. The agreement between the implied values of Ka0 and those predict-
ed is encouraging, particularly in view of the uncertain nature o f m u c h of the
data associated with such experiments.

3.2 Influence o f amplitude on "lock i n "


Earlier in this Section the p h e n o m e n o n known as "lock in" was introduced.
It was noted that for sufficiently large amplitudes and for values of the reduced
velocity Ur in the vicinity o f 6, the shedding frequency is apparently dictated
by the frequency of vibration, rather than being that predicted by the Strouhal
relationship, for a range of wind speeds.
The mean-square value of fluctuating lift per unit length, as defined by eqn.
(15), is comprised of two forces fluctuating, in general, at different frequencies.
r h e first component, the fluctuating lift arising from vortex shedding on a
~tationary cylinder, is centred on the shedding frequency fs. The two remaining
~omponents given in eqn. (15) ar-i_sefrom the motion of the cylinder and are
;herefore centred on the frequency of vibration f0. For the task at hand it is
~'onvenient to normalize the mean-square value o f fluctuating lift per unit
ength given by eqn. (15), resulting in
~21(1/2pU2D)2 = C~ + A 2 (19)
61

where A 2 = 8[2~(Ucrit/U)S] 4 ( y / D ) 2 (H~a + K2a) and Ucrit = foD/S. Depending on


the values taken by the various parameters in eqn. {19), one or other of the
terms on the right-hand side of the equation will make a greater contribution
to the fluctuating force. For the purpose of comparing the relationship given
in eqn. (19) with experimental data it is necessary to adopt a criterion by
which to decide whether the Strouhal frequency or the vibration frequency in
eqn. (19) is dominant. A simple criterion, already implied above, is adopted,
i.e., the shedding frequency is assumed to be given by the Strouhal relationship
when C~ > A 2 and by the vibrationfrequency when A 2 > C~. It should be
noted that all discussion regarding CL pertains to values appropriate to
stationary cylinders.
Before proceeding further, a few remarks regarding the parameter Ha, in-
formation concerning which is necessary for this exercise, are appropriate.
Measurements o f H a are few for two main reasons: first, the influence of H a
on the response is negligible, and second, H a is difficult to measure, particular-
ly in air. The measurements of H a made by Nakamura et al. [16] are used in
this exercise. Figure 7 shows H a as a function of US/foD; a line has been drawn
by eye through the data presented in the aforementioned reference.
The measurements made by Feng [26] may be used to examine the nature
of the fluctuating forces in the "lock i n " region. The results given by Feng
include response and the shedding frequency as a function of wind speed for a
range of structural damping levels. These data, taken together with the
appropriate values of Ka0 and H a from Figs. 2, 3 and 7, are sufficient to cal-
culate A 2 for a given amplitude. This procedure is applied to two sets of Feng's
results, each set corresponding to a different level of structural damping. The
12

10

0 ,3
f
06

04

02

02

4J4

-0.6 t i i i i i i i i
04 0.6 08 I0 I 2 1.4 1.6 1.8 20 22

U'Ucri!

Fig. 7. Variation of aerodynamic stiffness with U/Ucrit (after Nakamura et al. [16] ).
62

results of this exercise are summarized in Fig. 8. The value of sectional CL


assumed is 0.36; this was measured on the stationary cylinder [26] at a
Reynolds number close to the "lock in" range for the vibrating cylinder. The
parameter A 2 is a rapidly varying function in the "lock in" region, and there-
fore the predicted range of "lock in" is not very sensitive to the value of CL
assumed, nor to the value of "~,~,/A2 which is used to define the predominant
frequency. In general, the range of "lock in" predicted by the described
approach compares favourably with observations.

3.3 Modelling motion-induced force at large amplitudes


The discussion so far of the effects of motion on vortex-shedding forces
has been confined to amplitudes less than ~ 20% of the diameter. A feature
(a) ~s = 0 . 0 0 1 8 1

1.0 o

0.8
o
o

o o

0.6
o
2
A
o

0.4 o

0.2
\_ ~2

..........ilk-i-
. . . . . C L = 0.13

o o

1.5

en ~ S = 0.198
1.4 • Experim •

1.3

1.2
fs/f,,
1.1

1.0 ~ o oeoo • ~Hm ~ o •

0.9

' Predicted Range of '


0.8
"Lock-ift"

I i I I
5 6 7
U.
63

{b) ~s = 0.00257
0.8

0
0
0.6
2
A 0°
0

0
0.4 0
0

0.2
~2

1.5
ooo/ \ CL= 0.13

O
O

1.4 • Experimental
S = 0.198
1.3

1.2

fs / fo 1. I

0.9 ~J

Predicted Range "~]


0.8 of "Lock-in"

,I t I m I t I
4 5 6 7
Ur
F i g . 8 . C o m p a r i s o n o f predicted range of " l o c k i n " with m e a s u r e m e n t s o f Feng [ 2 6 ] .

common to all experiments conducted on vibrating cylinders where the param-


eter Ks, as defined by eqn. (8), has approached zero and U -~ Ucrit, is that some
limiting amplitude is reached. The clear implication is that the lift force is a
nonlinear function of amplitude. The measured values of the limiting amplitude
show a fair degree of scatter; a compilation by Griffin [27] suggests an aver-
age of about one diameter. The limiting amplitudes inferred from experiments
on cantilevered models reported by Wooton [28] are much lower. This suggests
that the limiting amplitude is sensitive to aspect ratio and/or mode shape.
A number of formulations are possible in modelling the nonlinear behaviour
of the lift force as observed in experiments. Referring to eqn. (9), it is clear that
Ka, or its equivalent in terms of K a given by eqn. (12), can be regarded as an
64

aerodynamic damping coefficient, and therefore can be expressed as a fraction


of critical damping, as
a = - (4~f0 PD2 Ka/47rfo m ) = - (p D ~/m )K a ( 20 }
Simple physical arguments led Marris [29] to suggest that the aerodynamic
damping force per unit length can be represented by an expression of the form
w ( t ) = ka(y -- Gj '3) (21)
where G is a constant. The presence of the nonlinear term ensures that the re-
sponse is self-limiting.
Now consider a single-degree-of-freedom vibrating system with damping
properties of the kind described above. If the response is narrow-band in
character, it can be expressed as
y ( t ) = Y ( t ) cos 2nfot (22)
where Y ( t ) is a slowly varying function of time, and the nonlinear relationship
given by eqn. (21) be replaced by
w ( t ) = k a ) [1 - - ( Y / Y L ) 2] (23)

The energy dissipated by the force in eqn. (23) per " c y c l e " will be the same as
that for eqn. (21), provided that
YL = 1/~rfos/~G (24)
The amplitude YL is the equilibrium or limiting amplitude that would be
attained in the absence of external forcing and structural damping. The
presence of the nonlinearity in the damping severelycomplicatesthe com-
putation of the response to random excitation, and as a simplificationit is
proposed that the damping force be expressed as
F ( t ) = kay [1 - ( 7 / Y L ) 2] (25)
where YL = YL/X/2. This presumes linear damping with a coefficient that is
dependent upon the mean-square displacement ~ 2 averaged over a long period.
The negative aerodynamic damping, as a fraction of critical, is then expressed
as
~?a = - ( P D 2 / m ) g a 0 [1 - ( ~ / y L ) 2] (26)
where Ka0 is simply the value of Ka when y -+ 0.
For small amplitudes of vibration, w h e r e no account is taken of motion-
induced forces, and for cases where B, the bandwidth: of the v o r t e x ~ e d d i n g
force spectrum, is much greater than ~s, the structural d a m p S , Vickery and
Clark [9] developed a simple expression for the tesponae. I n c o r p o ~ n g the
development described above, this expression b e c o m e s
~ /D = (~,/8~ =) (U/foD) 2 (pD=fm) {(~/,r/4B) (foil'=) exp { - [(1 "fo/f,)lB] ~} /
(~s + ha)} ~ (27)
65

where ~'2 rh J0
CL = ~0 rh~2
"~L Ro(zlz2) dzl dz2,. ~C~ is the sectional r.m.s, lift coeffi-
cient, and Ro(z~, z2) the spanwise correlation of lift force for a stationary
cylinder.
The validity of the simplifications made in replacing eqn. (23) by (25) to-
gether with earlier assumptions concerning the influence of large-scale turbu-
lence on the bandwidth B and on the aerodynamic damping r~a or K a can be
explored by numerical simulation. This is the subject of the next Section.

4. Numerical simulation of mathematical model

Numerical experiments were undertaken to examine the soundness of the


various simplifications made in the development described in the previous
Section. The response of a cylinder to numerically simulated vortex-shedding
forces was determined by direct integration of the nonlinear equation of
motion. The fluctuating lift force due to vortex shedding was generated from
an equation of the form
EL(t) = CL 1/2pU~Dh [1 + iu(t)] 2g(s) (28)
where s = ftoU[1 + iu(t)] dt. The function u(t) was generated as a broad-band
Gaussian random variable with zero mean, unit variance and a spectrum of
the form
Su (f) = constant/[1 + (~f/fo) 2 ] (29)

The function u(t) represents the normalized fluctuating longitudinal com-


ponent of turbulence; multiplication of u(t) by i in eqn. (28) yields the
instantaneous wind speed with the mean wind speed subtracted. The function
g(s) in eqn. (28) was generated as a narrow-band Gaussian random variable
with zero mean, unit variance and a relative bandwidth of 0.1; the centre
wavenumber was set equal to S/D. The relative bandwidth of 0.1 is representa-
tive of smooth flow.
The aerodynamic damping force was computed from the equation
FD(t ) = 4~pD2fogao (y - Vy3)h (30)
with G selected to yield a specified limiting amplitude YL, and where
Ka0 = ¢ U[1 + iu(t)]/foD as defined by Fig. 4 for ~ / U = O.
From the foregoing discussion it can be seen that the parameter i in eqn.
(28) can be interpreted as the intensity of longitudinal turbulence, and the
parameter 2~S/[3 as the ratio of the integral scale of turbulence to the cylinder
diameter. Equation (28) is based on the assumption that, in large-scale turbu-
lence, the forces due to vortex shedding have a magnitude and frequency
determined by the instantaneous wind speed.
Numerical integration ~of r
the equation was performed for a range of values
of U/foD, {3 and ~s, with CL = 0.2, S = 0.2, Ka0,m~x = 1.0, m / p D 2 = 100 and
"YL/D = 0 . 5 / X / - 2 . For each set of values of U/foD, [3 and Vs, values were com-
66

puted for the r.m.s, response "~/D and the mean and r.m.s, values of the re-
sponse envelope ~/y2 + ~2/(2rrfo)2/D. All averages were c o m p u t e d over a
sample length of 1000/f0, selected after the passage of sufficient time to
ensure decay of the initial conditions. It is useful, prior to analyzing the results
of the numerical experiments, to consider limiting forms, in terms of the net
damping, of the response equation (27), In its simplest form, eqn. (27) can be
written as
"~/D = C / [ K s - Ka0 (1 - ~ 2 / ~ ) ] v2 (31)
where Ks is defined in eqn. (8). The factor C embodies all parameters not
associated with damping; the denominator is proportional to the square root

* I:ixperimenlal 128]
Re No ~ 600000
Height/Diameter= 11,5
0.10
0.0043

023D

"'Lock-in'"
Regime ]

(Y)
D max

0.01

"'Transition"
Regime

i > "Forced Vibr~


[ Regime

0,001 t I t i l t I
0.1 0.2 0.4 0.6 0.8 1.0 2.0 4.0
Ks

Fig. 9, Measured and fitted responses in smooth flow (after Viekery [30] ).
67

of the net damping. Equation (31) has approximate solutions given by


~/D ~- el(Ks - ga0) ~' (Ks >~ ga0) (32a)
~]~L ~ (1 -- gs/gao) V2 (Ks < Ka0) (325)
Ks is defined by eqn. (8) and Ka0 by eqn. (26). Equation (32a) represents the
low-amplitude case of a linear system with reduced damping (Hs + Ha0) subject
to random forcing, where Ha0 is defined by eqn. (26) with the condition ~ -~ 0.
Equation (32b) represents the nonlinear large-amplitude or "lock in" situation,
in which the amplitude is determined by the nonlinear damping and is virtually
independent of external forcing. Linking these two regimes is a transition zone
in the vicinity of Ks = Kao in which the amplitudes are dependent on both the
external forcing and the nonlinear damping. The three zones are illustrated in
Fig. 9, where measurements made on a vibrating cantilever are presented; an
equation of the form of eqn. (31) was fitted to the data and is also shown.

4.1 Results of numerical modelling

4.1.1 Behaviour at low amplitudes


Selected numerical results for the low-amplitude region, for a range of
turbulence intensities and turbulence scale, are presented in Fig. 10. The varia-
tion of y/D with reduced velocity Ur is shown in Fig. 10(a); the broadening of
the response curves and the reduction of the maximum response owing to the
increase of bandwidth with turbulence is clearly evident. The reduction of the
maximum response is compared with the predictions of eqn. (32a); the results
of the numerical integration are ~ 10% lower than the theoretical prediction
but this difference is barely significant, since the limited sample length from
which the r.m.s, value was evaluated (T = 1000/f0) results in a coefficient of
variation of the estimate of ~ 15%. The influence of the integral scale of
turbulence at Ur = 5.5 is illustrated in Fig. 10(c). Over the range 10/n < LID
< 500/~, the variation is of the order of 10%, which again is barely significant.

4.1.2 Behaviour at large amplitudes


The influence of intensity of turbulence on the response at large amplitudes
is shown in Fig. 11, together with the predictions from eqns. (32a) and (32b)
with Kao defined in accordance with Fig. 4 and Ka0,m~x = 1.0. The effect of
turbulence scale in this regime is illustrated in Fig. 12. It is apparent from both
figures that the numerical results are in good agreement with the predictions
of the equations concerned.

4.1.3 Probability distribution of response


While it has been demonstrated that the comparatively simple limiting forms
of eqn. (31), i.e., eqns. (32a) and (32b), yield r.m.s, displacement estimates that
compare very favourably with a detailed nonlinear treatment, the probability
distribution is indeterminate. The general trend can be argued from the form
68

(a) Response vs. Ur


0.01
o.0o9
0.008
0.007 / ~ ~ L
0.006 = 30

0.005 ~ D

0.003:32 ,:
f • i 3 l 0;; • ~ O .

0.002 l a I I l , 1 n , , I
4 5 6 7 8
Ur
Ib~ Response as a Function of Turbulence In[ensiU.'
0.01
0.009
Ur=5
0.008
Equation (32a)
0.007 ~ ~ LA) = 30

0.006

0.005

0.004 1 l I
10% 2if, i 30';

(c} Response as a Function of Turbulence Scale


0.01
0.009 • i : 20 ~
0.008 • i = 10~'; U r = 5.5

0.007

0.006
~qr~ - I ° ~-----'~'°~ ~-~-o~
0.005

0.004
2, 4t 6, 8, ,
10 ,
20 '
40 ' 80' 1~00
60 2O0
L/D

Fig. 10. Numerical model results in low-amplitude regime: Ka0,max= 1.0. C~, = 0.2, S = 0.2,
m/oD 2 = 100, ns = 0.02.

of eqn. (31) and, in particular, its approximations, eqns. (32a) and (82b). Away
from Ks = Ka0 and for Ks > Kao, the linear solution described by eqn. (32a) is
applicable and the response will be Gaussian with an envelope well approximated
by a Rayleigh distribution. Away from Ks = Kao and with Ks < Kao, eqn. (32b)
indicates independence from the random forcing, and hence the amplitude
fluctuations will be due to variations of the instantaneous aerodynamic damp-
hag with wind speed. The nonlinear dependence of clamping on amplitude then
suggests that the distribution witl tend to sinusoidal as Ks tends to zero.
69

0.4

• i = I0';
- - Equation (32b)
• i = l l ';
A i=12 ',
0.2

0. t i = 30!#

0.08

~ / - - Equation 132a)

0.06

0.04

0.02
Equation i31 ~ \\ w ~ •

0.01 0 0 iI O.'2 0.3~ 0. '4 '


0.5 0.6' L
0.7 0.8

Ks

Fig. 11. Numerical model results in high-amplitude r e g i m e : Ka0,max = 1 . 0 , C L = 0 . 2 , S = 0 . 2 ,


m/pD ~ = 100, L/D = 30.

The general behaviour described above was in fact observed in the numerical
solution, and the peak factors given in Fig. 13 show a rapid transition from
- 1 . 5 for Ks/Kao < 0.5 to ~ 4 . 0 for Ks/Kao > 2. The peak factors expected for
sinusoidal and Gaussian distributions are 1.414 and 3.87, respectively. The
nature of the variation of displacement with time is demonstrated in Fig. 14,
which shows computed response histories for the three regimes, i.e. (i} Ks < Ka0,
the "lock in" regime; (ii) Ks = Ka0, the "transition" regime; and (iii) K s >> Ka0 ,
the "forced-random-vibration" regime. The computed displacement histories
shown in the figure bear very close resemblance to experimental observations
at corresponding levels of the ratio Ks/Kao.
71

Ks K~o ~ 1

0.50 i

4) 5
Ks Kao ~ I
0.1
0 - tl'~JilUliililUu'~' "lllllllll~t*iu~m*it~i~iU~U~U~i~i~U~i~i~W~...........'~'................................................,~,.,,,,,,,,.,,,.,,ll
~t m u . . v Tnm . b. . . . . . . . .u u umu
Illlllllll~lllVlr" ~'~'v~m~vImI~m~M~~M~m1u~v ............. ,.,,,m,,mv, ,~
4).1

Ks Kao~>I
0,03

0 ........... ........... " ' . . . . . . . . . . . . . . i '


D ......... '
'""qF'"'" ' ' v.,,, ...........~,r.T.,mm,,,"',,,~nv~,,,,' ~qWw'[l"~l~w'qfllv~r,'

-0.03 L
Fig. 14. Computed displacement histories for high, moderate and low structural damping.

5. Concluding remarks

A semi-empirical mathematical model for estimating the across-wind re-


sponse of circular cylinders under two-dimensional conditions has been
presented. A key feature of the model is the representation of the motion-
induced force, which in reality involves complicated fluid--structure inter-
action phenomena, by a relatively simple device. This representation is achiev-
ed through the inclusion of negative aerodynamic damping which is depen-
dent on Reynolds number, on amplitude, and on the ratio of vibration fre-
quency to shedding frequency. The principal characteristics of the "lock in"
condition are reproduced by using this approach. The increase in the spanwise
correlation of lift force with amplitude, and the "locking in" of the shedding
frequency onto the vibration frequency, are demonstrated.
The motion-induced force is nonlinear in character and is essentially a
deterministic process under ideal conditions. Since the model is developed
within the framework of linear random-vibration theory, a number of simplifi-
cations are required to adapt the representation of motion-induced force to
the model. In order to make an assessment of the approximations involved,
the true nonlinear equation of motion was examined numerically. Results
were obtained for a wide range of amplitudes under a variety of turbulence
conditions. A comparison of the numerical results and those predicted by
the simplified theory showed that the errors introduced by the approximations
are of an acceptable level.
72

Acknowledgement

The work described herein was supported by a grant from the National
Science and Engineering Research Council of Canada.

References

1 A.G. Davenport, The application of statistical concepts to the wind loading of struc-
tures, Proc. Inst. Civ. Eng. (London), 19 (1961) 449--472.
2 A.G. Davenport, The response of slender line-like structures to gusty winds, Proc. Inst.
Civ. Eng. (London), 23 (1962) 389--408.
3 B.J. Vickery, On the reliability of gust loading factors, Natl. Bur. Stand. (U,S.), Build.
Sci. Ser. No. 30, U.S. Department of Commerce, Washington, DC, 1969.
4 J. Vellozi and E. Cohen, Gust response factors, Am. Soc. Civ. Eng., J. Struct. Div., 94
(1968) 1295--1313.
5 R.I. Basu and B~I. Vickery, Across-wind vibrations of structures of circular cross-
section. Part II. Development of a mathematical model for full-scale application. J.
Wind Eng. Ind. Aerodyn., 12 (1983) 75--97.
6 R.T. Hartlen and I.G. Cuttle, Lift--oscillator model of vortex-induced vibration, Am.
Soc. Civ. Eng., J. Eng. Mech. Div., 96 (1970) 577--591.
7 R.A. Skop and O.M. Griffin, On a theory for the vortex-excited oscillations of ~lexible
cylindrical structures, J. Sound Vib.. 41 (1975) 263--274.
8 ESDU, Across-flow response due to vortex-shedding: isolated circular cylindrical struc-
tures in wind or gas flows, Item 78006, Eng. Sci. Data Unit, London, 1978.
9 B.J. Vickery and A.W. Clarke, Lift or across-wind response of tapered stacks, Am. Soc.
Civ. Eng., J. Struct. Div., 98 (1972) 1--20.
10 H.H. Bruun and P.O.A.L. Davies, An experimental investigation of the unsteady
pressure forces on a circular cylinder in turbulent cross flow, J. Sound Vib., 40 (1975)
535--559.
11 R. King, A review of vortex shedding research and its application, Ocean Eng., 4
(1977) 141--171.
12 T. Sarpkaya, Vortex-induced oscillations -- a selective review, J. Appl. Mech., Trans.
Am, Soc. Mech. Eng., 46 (1979) 241--258.
13 C. Scruton and A.R. Flint, Wind-excited oscillations of structures, Proc. Inst. Cir. Eng.
(London), 27 (1964) 673--702.
14 R.I. Basu and B.J. Vickery, The across-wind response of tall slender structures of
circular cross-section to atmospheric turbulence. Part 2. I n p u t parameters, Rep.
BLWT-2-81, Faculty of Eng. Sci., University of Western Ontario, London, Ontario,
Canada, 1981.
15 A. Protos, V. Goldschmidt and G. Toebes, Hydroelastic forces on bluff cylinders, J,
Basic Eng., Trans. Am. Soc. Mech. Eng., 90 (1968) 378--386.
16 Y. Nakamura, S. Kaku and T. Mizota, Effect of mass ratio on the vortex excitation of
a circular cylinder, Proc. 3rd Int. Conf. on Wind Effects on Buildings and Structures,
Tokyo, 1971, pp. 727--736.
17 T. Yano and S. Takahara, Study on unsteady aerodynamic forces acting on an oscillat-
ing cylinder, Proc. 3rd Int. Conf. on Wind Effects o n Buildings and Structures, Tokyo,
1971, pp. 737--746.
18 G.W. Jones, J.J. Cincotta and R.W. Walker, Aerodynamic forces on a stationary and
oscillating circular cylinder at high Reynolds numbers, NASA Tech. Rep. R-30O, Natl.
Aeronaut. Space Admin., Washington, DC, 1969.
73

19 J.F. Howell and M. Novak, Vortex shedding from circular cylinders in turbulent flow,
in J.E. Cermak (Ed.), Proc. 5th Int. Conf. on Wind Engineering, Fort Collins, CO,
1979, Pergamon, Oxford, 1980, Vol. 1, pp. 619--629.
20 A. Daly and B.J. Vickery, Across-wind forces on a vibrating circular cylinder in the
Reynolds number range 3 × 103--2 × 104, Boundary-Layer Wind Tunnel Rep., Uni-
versity of Western Ontario, London, Ontario, Canada, 1983 (in preparation).
21 B.J. Vickery, A model for the prediction of the response of chimneys to vortex shed-
ding, Proc. 3rd Int. Chimney Design Syrup., Munich, 1978.
22 L.V. Schmidt, Measurements of fluctuating air loads on a circular cylinder, J. Aircraft,
2 (1965) 49--55.
23 E. Szechenyi and H. Loiseau, Portance instationnaires sur un cylindre vibrant dans un
~coulement supercritique, Rec. Aerospace, 1 (1975) 45--57.
24 G.H. Toebes, Fluidelastic features of flow around cylinders, Int. Res. Semin. on Wind
Effects, Ottawa, Canada, 1967, University of Toronto Press, 1968, Vol. 2, pp. 323--
341.
25 M. Novak and H. Tanaka, Pressure correlations on a vibrating cylinder, in K.J. Eaton
(Ed.), Proc. 4th Int. Conf. on Wind Effects on Buildings and Structures, Heathrow,
1975, Cambridge University Press, London, 1977, pp. 227--232.
26 C.C. Feng, The measurement of vortex-induced effects in flow past stationary and
oscillating circular and D-section cylinders, M.A. Sc. Thesis, University of British
Columbia, Vancouver, 1968.
27 O.M. Griffin, OTEC cold water pipe design for problems caused by vortex-excited
oscillations, Ocean Eng., 8 (1981) 129--209.
28 L.R. Wooton, The oscillations of large circular stacks in wind, Proc. Inst. Cir. Eng.
(London), 43 (1969) 573--598.
29 A.W. Marris, A review of vortex streets, periodic wakes and induced vibration phenom.
ena, J. Basic Eng., Trans. Am. Soc. Mech. Eng., 86 (1964) 185--193.
30 B.J. Vickery, Across-wind buffeting in a group of four in-line model chimneys, J.
Wind Eng. Ind. Aerodyn., 8 (1981) 177--193.

Potrebbero piacerti anche