Numerical Prediction of Flow

S. Lu1
Graduate Student
Fields Around Circular Cylinders:
Ö. F. Turan Forced Motion and Dynamic
Associate Professor
e-mail: [email protected] Response Cases
School of the Built Environment-Mechanical
Engineering,
At Re ⫽ 2000, the predicted flow field around a circular cylinder in forced transverse
Victoria University of Technology,
oscillation is verified with experimental results. For coupled torsional and transverse
P.O. Box 14428 MC,
oscillation cases, the numerical results indicate that lock-in depends on the relative phase
Melbourne, Victoria 8001,
between torsional and translational oscillations. The dynamic response of an elastically
Australia
mounted circular cylinder in cross flow, obtained by solving the structural equations
simultaneously with the Navier-Stokes equations, is in reasonable agreement with experi-
mental data. The dynamic response results indicate that the change of wake pattern from
2S to 2P with increased frequency ratio, is not always simultaneous with the change in the
relative phase between lift force and cylinder displacement. 关S0098-2202共00兲02003-4兴

I Introduction in Section 4. The dynamic response of a spring-mounted circular

cylinder in cross-flow was also examined, along with the details
Toward predicting the wind-induced oscillation of transmission
of the flow field. These results are given in Section 5. The con-
conductors, a numerical model is developed to study the flow clusions are summarized in Section 6.
fields around moving circular cylinders. Coupled with wind tunnel
tests, a detailed understanding gained through numerical modeling
is expected to lead to modified damper designs. Among different II Numerical Scheme
types of wind-induced oscillation of structures, ‘‘galloping’’ re- Flow fields around vibrating circular cylinders have been stud-
fers to high-amplitude 共5–300 conductor diameters兲 and low- ied numerically, but mostly at Reynolds numbers of less than
frequency 共0.1–3 Hz兲 aeroelastic oscillations at wind speeds of 1000 共Chang and Sa 关5兴, Chilukuri 关6兴, Hulbert et al. 关7兴, Lecointe
6–25 m/s 共Pon et al. 关1兴兲. Several electrical transmission line and and Piquet 关8兴, to name a few兲. The reason is that the wake is fully
transmission tower failures have been caused by galloping of con- turbulent at higher Reynolds numbers, and accurate prediction
ductor cables and guy wires, especially after freezing rain storms. becomes more difficult due to problems associated with turbu-
With ice accretion, cables lose their approximately circular cross lence modeling and resolution of the numerical grid. The results
sections, and the resulting shapes can have aeroelastic instability. presented here were obtained from solutions without a turbulence
Due to very high replacement costs of such lines and towers, model at a Reynolds number of 2000 for comparison with the
conductor cable and guy wire galloping has been widely studied forced-motion experimental results of Zdero et al. 关4兴.
since the early work of Den Hartog 关2兴. In the existing analytical A FORTRAN code is developed based on the framework of the
and numerical models for galloping analysis, the assumption of a TEACH code, and a two-step time dependent solution is incorpo-
quasi-steady flow field has been widely used 共Blevins 关3兴兲. Ac- rated coupling the flow and structural equations. For the flow
cording to this assumption, the fluid force on the structure is de- solution, the continuity and Navier-Stokes equations are dis-
termined solely by the instantaneous relative velocity and angle of cretized using a finite volume method 共Patankar 关9兴兲. The compu-
attack of the flow. However, this assumption is not valid for many tational domain is divided into cells, or control volumes, by a set
observed field galloping cases 共Zdero et al. 关4兴兲. of radial lines and concentric circles in polar coordinates. For grid
In this study, the time dependent Navier-Stokes equations are independence, 100 radial cells and 120 circumferencial cells are
solved to obtain the instantaneous force coefficients, coupled with used in most cases, with the computational domain extending 40
the solution of the appropriate equations of motion for the struc- cylinder diameters radially. The size of the cells is distributed
ture, thus eliminating the need for a quasi-steady assumption. Af- nonuniformly in the radial direction, with high concentration near
ter verification with limited experimental results, the numerical the cylinder surface to resolve the boundary layer. Two layers of
results have provided more detailed information on the flow struc- cells are used radially with different linear expansion coefficients.
ture than available from experiments. The details of the code are From the cylinder surface to a distance of 0.4 radius, 20 live cells
are used, with an expansion coefficient of 1.12, and for the rest of
given in the next section. As indicated in Section 3, the results of
the domain, the expansion coefficient is 1.06. The variables are
the code have been compared first with benchmark experimental
stored on a staggered arrangement proposed by Patankar 关9兴. The
data of flow fields around stationary circular cylinders at Re
value of the velocity components at the control volume faces is
⫽100, 1000, and 2000. Subsequently, the controlled transverse
interpolated using a bounded form 共Gaskell and Lau 关10兴兲 of
oscillation cases were verified with experimental results, and the
QUICK originally devised by Leonard 关11兴, arranged in the up-
flow fields around circular cylinders in controlled translational, wind form 共Hayase et al. 关12兴兲. The discretized equations are then
torsional and combined translational and torsional oscillations solved with the SIMPLE-C algorithm 共Van Doormaal and Raithby
were studied in detail. The forced motion results are summarized 关13兴兲. Implicit time marching and tridiagonal matrix method are
1
used with alternate sweep direction.
Now with the University of California, Berkeley. Since the mesh is fixed relative to the cylinder in the numerical
Contributed by the Fluids Engineering Division for publication in the JOURNAL
OF FLUIDS ENGINEERING. Manuscript received by the Fluids Engineering Division;
solution, the translational motion of the cylinder is transformed
revised manuscript received February 8, 2000. Associate Technical Editor: G. Erle- into the time dependent boundary condition at the outer boundary
bacher. of the computational domain and no-slip condition on the cylinder

Journal of Fluids Engineering Copyright © 2000 by ASME DECEMBER 2000, Vol. 122 Õ 703

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 1 Regions of vortex synchronization patterns given by Zdero et al. †4‡ based on a map
given by Williamson and Roshko. The flow visualization results of Zdero et al. are shown by the
dashed lines corresponding to A Õ d Ä2 and 7. The details of the synchronized wake regions are
given in the inset where vortices shed per motion cycle are enclosed by dotted lines „S, single
vortex; P, vortex pair; C, coalescence….

surface, as well as an inertial force acting on the fluid. Taking a x nuity equation is satisfied even when the boundary velocity is
and a y as the longitudinal and transverse accelerations, respec- changed. A dimensionless mass residual of 1⫻10⫺6 is used for
tively, of the cylinder in Cartesian coordinates, the inertial terms convergence.
(⫺a x sin ␪⫺ay cos ␪) and (a x cos ␪⫺ay sin ␪) are added to the mo- The present implementation allows three degrees of freedom
mentum equations. Hence, these terms contribute to the source for the cylinder: two translational and one torsional. The calcula-
terms in the discretized equations. tion of the dynamic response of an elastically mounted cylinder is
Due to the difficulty of defining an outflow condition, a time achieved through the coupling of the Navier-Stokes equations
dependent boundary condition is prescribed along the whole outer with the corresponding structural equations of motion. The total
boundary. This approach is validated by the fact that it gives the fluid force exerted on the structure including fluid damping and
same result as using an outflow condition for a stationary cylinder. inertial forces, are calculated from surface pressure and skin fric-
At each time step, the boundary condition is updated for a moving tion as a result of the flow solution. Although better accuracy
cylinder. Taking dV x and dV y as the velocity increments of the could have been achieved by using the Runge-Kutta method for
cylinder during one time step, the radial and circumferencial ve- the solution of the structural equations, a direct explicit approach
locity increments at the outer boundary are (dV x cos ␪ is used here, because the time step as dictated by the solution of
⫺dVy sin ␪) and (⫺dV x sin ␪⫺dVy cos ␪), respectively. This the flow field, is small. For most cases, 500 time steps were used
change in the boundary velocity violates the continuity condition per period of the cylinder motion to obtain a flow field that is
temporarily, and it takes a number of iterations for the boundary independent of the time step.
information to propagate to the cylinder. At each time step, the flow field is updated first, and the force
In order to reduce the number of iterations for convergence, an coefficients are computed. The fluid forces are then fed into the
inviscid flow field is also calculated from the incremental radial structural equations to update the response of the cylinder. Next,
and azimuthal cylinder velocities. This correction velocity is the results are used to update the boundary conditions, preparing
added to the velocity field of the last time step to provide an initial to move to the next time step. The details of the program as well
guess for the present time step. The resulting velocity field is as a users’ manual are given by Lu and Turan 关14兴.
already a good approximation in the inviscid region considering
the continuity equation. However, the no-slip condition is not sat- III Stationary Circular Cylinder Cases: Benchmark
isfied with this velocity field, and as a result, more sweeps are
needed close to the wall. The program is designed to perform Comparisons
more sweeps in the boundary layer region. This method was To validate the numerical implementation, detailed compari-
found to accelerate convergence considerably, because the conti- sons were made with benchmark experimental and some numeri-

704 Õ Vol. 122, DECEMBER 2000 Transactions of the ASME

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 3 Streamline and vorticity contours for f m Õ f v 0 Ä0.83, Re
Ä2000

radius and length, respectively. The Strouhal number is defined as,

St⫽2a f v 0 /U 0 , where f v 0 is the frequency of vortex shedding
from a stationary cylinder.
All other cases employ a mesh with 120 circumferencial cells
and 100 radial cells in order to fully resolve the boundary layer at
a higher Reynolds number. The evolution of near wake structure
at a Reynolds number 1000 is similar to that given by Braza et al.
关15兴. The present results give a mean drag coefficient of 1.4 and a
Strouhal number of 0.23, higher than the experimental values of
1.2 and 0.21, respectively. This difference must be due to the
two-dimensionality imposed in the simulation, and it is consistent
with the numerical results of Lecointe and Piquet 关8兴 which give a
Strouhal number of 0.24 at Re⫽875. The seemingly good agree-
Fig. 2 Cylinder displacement, force coefficient and separation ment with experimental data in earlier results 共Jordan and Fromm
point histories for f m Õ f v 0 Ä0.83, ReÄ2000 关16兴, Braza et al. 关15兴兲 is probably due to insufficient mesh reso-
lution in the boundary layer and the use of lower order 共such as
central difference兲 schemes.
cal results of the flow fields around stationary circular cylinders. From a comparison of the numerically predicted von Karman
These comparisons are summarized in this section. The first sta- vortex street formation with the experimental one 共Batchelor 关17兴兲
tionary case is vortex shedding from a circular cylinder at a Rey- at Re⫽2000, it is observed that the global vortex structure com-
nolds number of 100, where the wake is laminar. Numerical re- pares well with flow visualization, while the computed drag coef-
sults were obtained from a mesh with 90 circumferencial and 70 ficient and Strouhal number values, C D ⫽1.5 and St⫽0.24, are
radial cells. The computational domain extended to 40 cylinder higher than the experimental values of 1.2 and 0.21, respectively.
diameters. A mean drag coefficient of 1.3 and a Strouhal number From the test cases above, it is seen that the present code gives
of 0.16 were computed, in agreement with the classical experi- results comparable with experiments at Re⫽100. For Re⬎1000,
mental data. Here, the drag coefficient is defined as, C D quantitative differences occur, although the qualitative flow field
⫽D/( ␳ U 20 al), where D is the drag force on cylinder; U 0 is the is still in close agreement with the experimental results. As indi-
free-stream velocity; ␳ is the fluid density; a and l are the cylinder cated earlier, these differences are expected to be mostly due to

Journal of Fluids Engineering DECEMBER 2000, Vol. 122 Õ 705

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the two-dimensional solution, along with the lack of a turbulence
model and possibly not enough resolution of the flow field next to
the cylinder surface. With further comparison of the numerical
predictions with the experimental results for moving cylinders, it
was possible to use the same code to obtain qualitative pictures of
the flow field.

IV Controlled Transverse and Torsional Oscillation

Cases
In this section, forced-oscillation results are presented, starting
Fig. 5 Streamline and vorticity contours for f m Õ f v 0 Ä0.50, Re
with transverse oscillation cases, followed by the cases of tor- Ä2000
sional oscillation, and combined transverse and torsional oscilla-
tions. In Fig. 1, Fig. 5 of Zdero et al. 关4兴 is given, which is a
variation of the vortex synchronization map of Williamson and
Roshko 关18兴, with the dashed lines showing the experimental re-
sults of Zdero et al. In this figure, U r ⫽U 0 /2a f m is the reduced

Fig. 4 Force coefficient histories for f m Õ f v 0 Ä0.50, ReÄ2000 Fig. 6 Force coefficient histories for f m Õ f v 0 Ä0.33, ReÄ2000

706 Õ Vol. 122, DECEMBER 2000 Transactions of the ASME

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 7 Streamline and vorticity contours for f m Õ f v 0 Ä0.33, Re
Ä2000

velocity, A is the peak to peak transverse oscillation amplitude,

and d is the cylinder diameter. The transverse controlled-motion
results given in this section are compared with the experimental
results given in Fig. 1. The Reynolds number is 2000, and the
peak to peak amplitude is 2 cylinder diameters for the transverse
oscillation cases (A/d⫽2) summarized next. The experimental
results of Zdero et al. indicate that stranding does not affect the
flow field at this Reynolds number; and therefore, the numerical
results presented below also hold for stranded cases. The ratios of
the transverse motion frequency to stationary vortex shedding fre- Fig. 8 Force coefficient histories for f t Õ f v 0 Ä0.36, ReÄ2000
quency, f m / f v 0 , used in the calculations are: 0.33, 0.36, 0.5, 0.83,
0.91, and 1.07.
When a cylinder oscillates transversely at a frequency f m near
its stationary vortex shedding frequency f v 0 , the vortex shedding The lift and pitching moment coefficients and the top and bottom
process may become synchronized with the cylinder motion, and separation points fluctuate at the same frequency as the transverse
‘‘lock-in’’ occurs. It could also occur when the motion frequency oscillations; while the drag coefficient fluctuates at twice this fre-
is near a harmonic ( f m / f v 0 ⫽2,3, . . . ,n) or subharmonic quency. The corresponding Lissajous figures, which illustrate the
( f m / f v 0 ⫽1/2,1/3, . . . ,1/n) of the stationary vortex shedding fre- relative phase between the force coefficients, do not conform to a
quency. The latter is more relevant in the study of cable galloping standard form, because the force coefficients are not strictly har-
where the frequency is low. At other frequency ratios, the vortex monic. The evolution of streamlines and vorticity contours in ap-
shedding frequency may be different from the motion frequency proximately one period of cylinder motion for the case of
of the cylinder. These cases are classified as nonsynchronized, f m / f v 0 ⫽0.83 is given in Fig. 3. It can be observed that two pairs
NS, modes. of vortices are shed in one period of motion, the 2P mode. The
The cylinder displacement and force coefficient histories for the in-line spacing between pairs of vortices appears narrow as a re-
case of f m / f v 0 ⫽0.83 are given in Fig. 2. Also shown in this figure sult of over-prediction of the Strouhal frequency. Each opposite-
is the fluctuation of the top and bottom separation points, located signed pair of vortices is convected away from the centerline
at approximately 124 and 234 deg, respectively, within ⫾81 deg. while moving downstream, thus creating two rows of alternative
The time axis in this figure and in other histories to come, is vortices and a wide ‘‘dead area’’ near the centerline. The fluctua-
non-dimensionalized with a/U 0 . It can be seen that periodic mo- tion of velocity and pressure in the dead area is weak, as observed
tion was obtained after several periods of vibration, indicating that from the velocity and pressure histories at several diameters from
the vortex shedding frequency is locked onto motion frequency. the rear stagnation point of the cylinder. Similar fundamental

Journal of Fluids Engineering DECEMBER 2000, Vol. 122 Õ 707

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 9 Streamline and vorticity contours for f t Õ f v 0 Ä0.36, Re
Ä2000

lock-in was also found for f m / f v 0 ⫽0.91 and 1.07. The presence
of the 2P mode for f m / f v 0 ⫽0.83, 0.91 and 1.07 is in agreement
with the experimental results given in Fig. 1.
The force coefficient histories for the case f m / f v 0 ⫽0.5 is given
in Fig. 4. An example of the corresponding streamline and vortic-
ity patterns is shown in Fig. 5. No periodicity is observed for this
case, as seen from the force coefficient histories. The instanta-
neous streamline and vorticity contour plots suggest that the vor-
tices are shed in pairs but not at fixed points within the motion
cycle 共the phase of shedding changes over time兲. This nonsyn-
chronized pattern is in agreement with the results of Williamson
and Roshko 关18兴. Zdero et al.’s results indicate that f m / f v 0 ⫽0.5
is the boundary of the 2P mode, as shown in Fig. 1. For the case
of f m / f v 0 ⫽0.36, a nonsynchronized flow field was obtained, in
agreement with Zdero et al.’s flow visualization results, although
Williamson and Roshko observed the 2P⫹2S mode in their flow
visualization at a lower Reynolds number.
As shown in Fig. 6, for the case f m / f v 0 ⫽0.33, the lift and
pitching moment histories have the motion frequency, and the
drag coefficient twice the motion frequency, as the fundamental
frequency component over which higher frequencies are imposed
due to vortex shedding. In fact, six peaks exist per period of the
lift and pitching moment histories, and per two periods of the drag
history. The streamline and vorticity plots in Fig. 7 suggest that Fig. 10 Force coefficient histories for f t Õ f v 0 Ä f m Õ f v 0 Ä0.36, in-
six alternative, equally spaced vortices are shed per motion pe- phase torsional and transverse oscillations, ReÄ2000
riod. Hence, this is virtually the same structure as what was called
3P mode by Zdero et al. as shown in Fig. 1.
Following the favorable comparison between the numerical and
experimental results of forced translational oscillation cases, sev- lock-in was present with a sudden drop in the fluctuating ampli-
eral cases were run to see the effect of torsional oscillations. The tude of the pitching moment. The wake pattern had the 2P mode,
results obtained for six cases of torsional oscillations are summa- and it was similar to the previous lock-in cases. As the cylinder
rized next. The first case is of torsional oscillations alone with a moved downward, it also rotated counter-clockwise shedding a
peak-to-peak amplitude of 60 deg and torsional frequency ratio, strong clockwise vortex and a weakened counter-clockwise vor-
f t / f v 0 , of 0.91. The second case is of combined torsional and tex, and vice versa. When the torsional motion was coupled with
in-phase transverse oscillations. The third case is of combined out-of-phase transverse motion, still keeping f t / f v 0 ⫽ f m / f v 0
torsional and out-of-phase transverse oscillations. For the second ⫽0.91, all synchronization was lost. The force coefficient histo-
and third cases, f t / f v 0 ⫽ f m / f v 0 ⫽0.91. In the remaining three ries indicated a long term of quasi-periodic evolution. The stream-
cases, the parameters are the same as in the first three cases, lines and vorticity contour plots displayed little organization.
except that the frequency ratio is set at 0.36. At f t / f v 0 ⫽0.36, the force coefficients are periodic, as seen in
The force coefficient histories of the case of torsional oscilla- Fig. 8. Similar to the histories in Fig. 6 for f m / f v 0 ⫽0.33, the lift
tions alone at f t / f v 0 ⫽0.91 had a strong component of the motion and pitching moment coefficient histories display the torsional
frequency. The streamline and vorticity contour plots displayed oscillation frequency, and the drag coefficient twice the torsional
the synchronized 2S mode. When the torsional motion was frequency, as the fundamental one. In all three histories, higher
coupled with in-phase transverse motion, f t / f v 0 ⫽ f m / f v 0 ⫽0.91, frequencies are imposed on the fundamental frequency. Both the

708 Õ Vol. 122, DECEMBER 2000 Transactions of the ASME

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 11 Streamline and vorticity contours for f t Õ f v 0 Ä f m Õ f v 0
Ä0.36, in-phase torsional and transverse oscillations, Re
Ä2000

Fig. 13 Streamline and vorticity contours for f t Õ f v 0 Ä f m Õ f v 0

Ä0.36, out-of-phase torsional and transverse oscillations, Re
Ä2000

Table 1 Summary of vortex modes for forced motion cases at

ReÄ2000

lift and drag coefficient histories have six distinct peaks per pe-
riod. As illustrated in Fig. 9, six alternate vortices are shed during
one period of torsional motion. When combined with in-phase
vertical motion, f t / f v 0 ⫽ f m / f v 0 ⫽0.36, the periodicity in the force
coefficient histories is lost, as seen in Fig. 10, and the flow pattern
depicted in Fig. 11 is not synchronized. Interestingly, when the
torsional motion is combined with out-of-phase vertical motion at
f t / f v 0 ⫽ f m / f v 0 ⫽0.36, the force coefficient histories in Fig. 12
become almost periodic, and the vortex pattern 2P⫹2S emerges,
as illustrated in Fig. 13. One vortex is shed whenever the cylinder
is near or past the maximum displacement to form a pair, and a
single vortex is shed whenever the cylinder passes the equilibrium
point.
The forced translational, torsional and combined oscillation
cases are summarized in Table 1.

V Dynamic Response Cases

Following the prediction of forced motion cases, the dynamic
response of an elastically mounted cylinder is discussed next. By
solving the flow equations with the structural equations, onset and
development of galloping can be predicted if the shape of the
Fig. 12 Force coefficient histories for f t Õ f v 0 Ä f m Õ f v 0 Ä0.36, out- cylinder is aerodynamically unstable. In the case of a circular
of-phase torsional and transverse oscillations, ReÄ2000 cylinder, a stable cross section, vortex induced vibration results.

Journal of Fluids Engineering DECEMBER 2000, Vol. 122 Õ 709

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Unlike the controlled motion cases, the vibration amplitude may
take long to build up in the calculations, especially when the fluid
density is low.
Several experimental results from elastically mounted circular
cylinders are reported in the literature. A hysteresis was reported
by Feng 关19兴 in his experiments on a lightly damped circular
cylinder. The amplitude achieved a higher peak when the reduced
velocity was increased than when it was decreased over the same
range. Similarly, when the reduced velocity was increased, the 2S
mode continued for a longer range of reduced velocities than
when it was decreased. In other words, the 2P mode switched to
the 2S mode at a lower reduced velocity when the reduced veloc-
ity was decreased than when it was increased. Bishop and Hassan
关20兴 observed a similar hysteresis in their lift and phase measure-
ments with a cylinder in controlled motion, but it occurred over a
somewhat lower range of reduced velocities. Griffin and Ramberg
关21兴 compared the free vibration experiments on circular cylinders
in air and water. No hysteresis was reported in these experiments. Fig. 14 Response amplitudes of a spring-mounted circular
Hence, it is not clear over what range of damping and mass ratios cylinder in cross flow. Present results: 䊐, m Õ2␳ a 2 Ä1.57, ␰ y
hysteresis occurs in the dynamic response of a spring-mounted Ä0; *, m Õ2␳ a 2 Ä7.6, ␰ y Ä0.; ¿, m Õ2␳ a 2 Ä7.6, ␰ y Ä0.051…; experi-
circular cylinder. A plausible guess based on Feng’s experiments mental results „Griffin and Ramberg †21‡…: 䊏, m Õ2␳ a 2 Ä7.6, ␰ y
is that it may occur when the damping and fluid to structure mass Ä0.051.
ratios are both low. Williamson and Roshko 关18兴, on the other
hand, explained the jump in the lift and phase measurements by
the change of wake pattern, namely, between 2S and 2P modes.
As described below, the present results indicate that the change in tio of 1.25. The wake pattern was the 2P mode for all these six
the phase difference between lift and displacement at the motion cases, and it is illustrated for f v 0 / f n ⫽1.36 in Fig. 15 with the
frequency may not always be simultaneous with the change of streamline and vorticity contours.
vortex mode. Bearman 关22兴 concluded from a compilation of experimental
The mass ratios m/2␳ a 2 ⫽1.57 and 7.6 were used in the present results that the Reynolds number may not be an important param-
simulations, with the latter matching the mass ratio in Dean eter for oscillating circular cylinders away from boundary layer
et al.’s experiment in water given by Griffin and Ramberg 关21兴 transition. Therefore, the Reynolds number was fixed at 2000 in
共where m is the structural mass per unit length, ␳ is the fluid the present computations for convenience, while the natural fre-
density, and a is the radius of the cylinder兲. The lower mass ratio quency of the cylinder was varied to change the frequency ratio.
corresponds to equal mass densities of the cylinder and fluid.
Critical damping ratios of 0 and 0.051 were used with the higher
mass ratio, to represent undamped and light damping cases, re-
spectively, the light damping being the same as in Dean et al.’s
experiment. The lower mass ratio cases were run as undamped,
only. Here, the critical damping ratio is defined as, ␰ y
⫽c y /2(mk y ) 1/2 , where c y and k y are the damping and stiffness
coefficients, respectively, in the transverse direction. The corre-
sponding natural frequency is, ␻ y ⫽(k y /m) 1/2 .
In Fig. 14, the amplitude ratio calculated for the dynamic cases
is plotted against the frequency ratio f v 0 / f n . Dean et al.’s experi-
mental results in water are given in the same figure. The choice of
the frequency ratio f v 0 / f n as the horizontal coordinate, instead of
the reduced velocity, U r , is discussed first 共U r ⫽U 0 /2a f n , where
f n is the natural frequency of the cylinder in fluid兲. The computed
stationary Strouhal number, St, is 0.24 at Re⫽2000 instead of
being about 0.21. As a result, the predicted resonance range would
be different from the experimental one, if the reduced velocity
were used in Fig. 14. In order to match the two scales, computa-
tional and experimental, the frequency ratio f v 0 / f n is employed,
where f v 0 ⫽U 0 St/d is the vortex shedding frequency of the sta-
tionary circular cylinder, St⫽0.24 for the numerical cases and
St⫽0.21 for the experimental cases. f n was measured in still wa-
ter in the experiments; while for the numerical results, it is ob-
tained by adding the mass of the fluid displaced by the circular
cylinder to that of the structure, f n ⫽ f n0 关 m/(m⫹ ␲␳ a 2 ) 兴 1/2,
where f n0 is the natural frequency in air. Using the new param-
eter, f v 0 / f n , the response peak appearing near a reduced velocity
of 6 in the experiments occurs at a frequency ratio of 1.2 both
computationally and experimentally. This figure is discussed fur-
ther with respect to the three sets of computational results below.
At a mass ratio of m/2␳ a 2 ⫽1.57, the dynamic response of the
cylinder was computed at frequency ratios of f v 0 / f n ⫽1.07, 1.22,
1.36, 1.56, 1.71, and 2.14. The response amplitudes shown in Fig. Fig. 15 Streamline and vorticity contours for f v 0 Õ f n Ä1.36,
14 demonstrate wide resonance, with the maximum amplitude ra- m Õ2␳ a 2 Ä1.57, ␰ y Ä0., ReÄ2000

710 Õ Vol. 122, DECEMBER 2000 Transactions of the ASME

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 16 Relative phase between lift and displacement for f v 0 Õ f n Ä„ a … 1.07, „b… 1.22, „c… 1.36, „d… 1.56, „e… 1.71 and „f…
2.14 at m Õ2␳ a 2 Ä1.57, ␰ y Ä0., ReÄ2000

In the experiments, the flow speed was adjusted, thus changing pecially in the lift coefficient, although the displacement is nearly
vortex shedding frequency of the stationary cylinder and the fre- symmetrical. The lift amplitude drops to almost zero at a fre-
quency ratio as a consequence, along with the Reynolds number. quency ratio of 1.36, with high frequency fluctuation. This case is
The force coefficient histories for f v 0 / f n ⫽1.07, 1.22, 1.36, the transitional one from the lift force being in phase with the
1.56, 1.71, and 2.14 at m/2␳ a 2 ⫽1.57 show some asymmetry, es- displacement to out of phase, as illustrated in Fig. 16. With the

Journal of Fluids Engineering DECEMBER 2000, Vol. 122 Õ 711

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 displacement plotted together with lift coefficient, the change in
the lift coefficient is clearly visible, from being in-phase to out-
of-phase with displacement. The wake pattern obtained in the
range of the frequency ratios investigated is basically the 2P
mode, while the lift is in phase with displacement for f v 0 / f n
⫽1.07 and 1.22; during transition at f v 0 / f n ⫽1.36; and while the
lift is out of phase with displacement for f v 0 / f n ⫽1.56, 1.71 and
2.14. Close inspection of the vortex formation process in all these
cases indicates that it is virtually identical to the flow pattern
during one period of the cylinder motion given in Fig. 15 for
f v 0 / f n ⫽1.36. The vorticity contours clearly show that the 2P
mode wake is also similar to those found in forced motion cases
discussed in the previous section. The only difference is that at
higher frequency ratios f v 0 / f n ⫽1.71 and 2.14, the vortex in each
pair away from the center line is weakened so that the pairs do not
form closed vortices in the streamline plots. This set of results
indicate that the change of vortex mode from 2S to 2P does not
always occur simultaneously with the change of the lift from in-
phase to out-of-phase with the displacement. The next two sets of
results confirm this conclusion further.
Next, the mass ratio is increased to m/2␳ a 2 ⫽7.6, and the fre-
quency ratios of f v 0 / f n ⫽0.89, 1.11, 1.21, and 1.33 are used. The
relative phase between the lift and displacement is illustrated in
Fig. 17 for this mass ratio. The time histories especially for the
case f v 0 / f n ⫽0.89 contain significant beat, while the flow pattern
indicated the presence of a 2S vortex pattern. The lift is in-phase
with the displacement for this frequency ratio, as seen in Fig.
17共a兲. At frequency ratios of 1.11 and 1.21, the lift patterns are in
the transition process, from being in-phase with displacement to
out-of-phase, and appear to have twice the vortex shedding fre-
quency, as shown in Fig. 17共b兲 and 17共c兲, respectively. The cor-
responding wake patterns for these two frequency ratios have al-
ready changed to the 2P mode. The lift becomes out-of-phase with
the displacement for f v 0 / f n ⫽1.33 as shown in Fig. 17共d兲. The
wake pattern for this case is also the 2P mode. Both the 2S and 2P
modes were identical to those of the forced motion cases.
The maximum amplitude ratio of 1.34 is obtained at f v 0 / f n
⫽1.21 and m/2␳ a 2 ⫽7.6, as shown in Fig. 14. This value is lower
than the limit values obtained from simplified mathematical mod-
els such as the wake oscillator and correlation models 共Blevins
关3兴兲, which give limit amplitude ratios greater than 2.0 for an
elastically mounted rigid circular cylinder. In this case, the maxi-
mum amplitude is greater than that obtained at the mass ratio of
1.57, although a fully steady amplitude was not secured in the
latter case, as shown in Fig. 16共b兲.
The cases discussed above for m/2␳ a 2 ⫽1.57 and 7.6 do not
have any structural damping. In the cases that follow, a damping Fig. 17 Relative phase between lift and displacement for
ratio of 0.051 is applied to match the value measured in Dean f v 0 Õ f n Ä„ a … 0.89, „b… 1.11, „c… 1.21, and „d… 1.33 at m Õ2␳ a 2 Ä7.6,
et al.’s experiment 共Griffin and Ramberg 关21兴兲, while the mass ␰ y Ä0., ReÄ2000
ratio was kept at 7.6. The frequency ratios used are f v 0 / f n
⫽0.89, 1.02, 1.11, 1.21, and 1.33. With light damping, the re-
sponse amplitudes drop significantly from the undamped case, as
seen in Fig. 14. The maximum amplitude compares well with the the presence of viscous damping. The 2S and 2P modes were
again identical to the forced motion patterns.
experimental value, although the resonance peak obtained is nar-
In Fig. 19, the dimensionless response frequency, f / f n , is plot-
rower than the experimental one. The discrepancy between the
ted against the dimensionless stationary vortex shedding fre-
present results and experimental data may have resulted from quency, f v 0 / f n . The results in this figure indicate that at the low
lacking a turbulence model and imposed two-dimensionality in mass ratio of 1.57, the response frequency is locked on to the
the computations. The relative phase between lift and displace- stationary vortex shedding frequency rather than the natural fre-
ment is plotted in Fig. 18. For f v 0 / f n ⫽0.89, 1.02, 1.11, and 1.21, quency of the cylinder, as it is usually obtained in experiments
the vortex shedding was in the 2S mode, while the lift remained in with high structure to fluid mass ratio. Hence, high amplitude
phase with the cylinder displacement, as shown in Fig. 18共a兲 to resonance is predicted over a wide range of reduced velocities or
18共d兲. At f v 0 / f n ⫽1.33, the vortex mode became 2P, and the lift frequency ratios for a light structure, in which the response fre-
became almost out of phase with displacement, according to Fig. quency may be controlled by stationary vortex shedding fre-
18共e兲. The result from an FFT analysis of the lift and displacement quency 共inverse lock-in兲. When the mass ratio is increased mod-
histories indicated that the lift now actually lead the displacement erately to 7.6, the response frequency remains between the natural
by 145 deg. The departure of phase difference from 180 deg pro- frequency of the structure and the stationary vortex shedding fre-
vided energy to the cylinder to maintain a constant amplitude in quency. It is interesting to note that adding light viscous damping

712 Õ Vol. 122, DECEMBER 2000 Transactions of the ASME

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 19 Response frequency, f , of a spring-mounted circular
cylinder. 䊐, m Õ2␳ a 2 Ä1.57, ␰ y Ä0; *, m Õ2␳ a 2 Ä7.6, ␰ y Ä0.; ¿,
m Õ2␳ a 2 Ä7.6, ␰ y Ä0.051.

Table 2 Summary of vortex modes and the relative phase be-

tween lift force and cylinder displacement for dynamic re-
sponse cases at ReÄ2000

VI Conclusions
The numerical predictions of the flow fields around stationary
and transversely oscillating circular cylinders, as well as of the
dynamic response of a spring-mounted circular cylinder, have
been found to be in reasonable agreement with the available ex-
perimental data. The predictions were obtained with a FORTRAN
code developed using the SIMPLE-C algorithm with the QUICK
scheme.
From the results of the controlled motion cases, it can be seen
that when torsional motion is coupled with the vertical oscillation,
the lock-in of the vortex shedding is possible, but it depends
largely on the relative phase between the torsional and transla-
tional oscillations. With torsional motion alone, the flow field may
not be modified significantly.
The numerical simulation of the dynamic cross flow response
of a spring-mounted circular cylinder confirm the phase change,
Fig. 18 Relative phase between lift and displacement for
f v 0 Õ f n Ä„ a … 0.89, „b… 1.02, „c… 1.11, „d… 1.21, and „e… 1.33 at
namely, from the lift force being in-phase with the cylinder dis-
m Õ2␳ a 2 Ä7.6, ␰ y Ä0.051, ReÄ2000 placement to out-of-phase, with increased frequency ratio, as well
as the change of wake pattern from 2S to 2P. However, the change
of wake pattern does not always occur simultaneously with
the change of the relative phase between the lift force and
displacement.
affects the response frequency as a result of nonlinear interaction,
while viscous damping has little effect on the natural frequency in
a linear spring-mass system at similar damping levels. Acknowledgment
A summary of the vortex modes and lift phases is given in Financial support for this project was provided by NSERC
Table 2 for the three sets of dynamic cases. 共Natural Sciences and Engineering Research Council兲 Canada,

Journal of Fluids Engineering DECEMBER 2000, Vol. 122 Õ 713

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Ontario Hydro, Canada, and Victoria University of Technology, dure Based on Quadratic Upstream Interpolation,’’ Comput. Methods Appl.
Mech. Eng., 19, pp. 59–98.
Australia. The authors are grateful to D. G. Havard of Ontario
关12兴 Hayase, T., Humphrey, J. A., and Greif, R., 1992, ‘‘A Consistently Formulated
Hydro for many helpful discussions. A shorter version of this QUICK Scheme for Fast and Stable Convergence Using Finite-Volume Itera-
paper was presented at the 1996 ASME Fluids Engineering Divi- tive Calculation Procedures,’’ J. Comput. Phys., 98, pp. 108–118.
sion Conference 共Turan and Lu, 1996兲 in San Diego, California. 关13兴 Van Doormaal, J. P., and Raithby, G. D., 1984, ‘‘Enhancement of the SIMPLE
Method for Predicting Incompressible Fluid Flows,’’ Numer. Heat Transfer, 7,
pp. 147–163.
References 关14兴 Lu, S., and Turan, Ö. F., 1996, ‘‘A FORTRAN Code to Simulate Flow Fields
关1兴 Pon, C. J., Havard, D. G., Currie, I. G., and MacDonald, R., 1989, Aeolian around Oscillating Conductors,’’ Departmental Report, MET9608, Mech. Eng.
Vibration of Bundle Conductors, CEA REPORT 117-T-510, March. Dept., Victoria University of Technology, Melbourne.
关2兴 Den Hartog, J. P., 1932, ‘‘Transmission Line Vibration due to Sleet,’’ AIEE 关15兴 Braza, M., Chassaing, P., and Ha Minh, H., 1986, ‘‘Numerical Study and
Trans., 51, No. 4, pp. 1074–1086. Physical Analysis of the Pressure and Velocity Fields in the Near Wake of a
关3兴 Blevins, R. D., 1977, Flow-Induced Vibration, Van Nostrand Reinhold, New Circular Cylinder,’’ J. Fluid Mech., 165, pp. 79–130.
York. 关16兴 Jordan, S. K., and Fromm, J. E., 1972, ‘‘Oscillatory Drag, Lift, and Torque on
关4兴 Zdero, R., Turan, Ö. F., and Havard, D. G., 1995, ‘‘Galloping: Near Wake a Circular Cylinder in a Uniform Flow,’’ Phys. Fluids, 15, No. 3.
Study of Oscillating Smooth and Stranded Circular Cylinders in Forced Mo- 关17兴 Batchelor, G. K. 1967, An Introduction to Fluid Dynamics, Cambridge Uni-
tion,’’ Exp. Therm. Fluid Sci., 10, pp. 28–43. versity Press.
关5兴 Chang, K. S., and Sa, J. Y., 1992, ‘‘Patterns of Vortex Shedding from an 关18兴 Williamson, C. H. K., and Roshko, A., 1988, ‘‘Vortex Formation in the Wake
Oscillating Circular Cylinder,’’ AIAA J., 30, No. 5, pp. 1331–1336.
of an Oscillating Cylinder,’’ J. Fluids Struct., 2, pp. 355–381.
关6兴 Chilukuri, R., 1987, ‘‘Incompressible Laminar Flow Past a Transversely Vi-
关19兴 Feng, C. C., 1988, ‘‘The Measurement of Vortex Induced Effects in Flow past
brating Cylinder,’’ ASME J. Fluids Eng., 109, pp. 166–171.
Stationary and Oscillating Circular and D-section Cylinders,’’ M.Sc. thesis,
关7兴 Hulbert, S. E., Spaulding, M. L., and White, F. M., 1982, ‘‘Numerical Solution
of the Time Dependent Navier-Stokes Equations in the Presence of an Oscil- University of British Columbia, Vancouver.
lating Cylinder,’’ ASME J. Fluids Eng., 104, pp. 201–206. 关20兴 Bishop, R. E. D., and Hassan, A. Y., 1964, ‘‘The Lift and Drag Forces on a
关8兴 Lecointe, Y., and Piquet, J., 1990, Computation of Unsteady, Laminar, Incom- Circular Cylinder Oscillating in a Flowing Fluid,’’ Proc. R. Soc. London, Ser.
pressible, Viscous Flows Using the Vorticity Streamfunction Formulation, A, 277, pp. 51–75.
Computational Methods in Viscous Aerodynamics, Murthy, T. K. S., and 关21兴 Griffin, O. M., and Ramberg, S. E., 1982, ‘‘Some Recent Studies of Vortex
Brebbia, C. A., eds., Elsevier and Computational Mechanics, pp. 77–116. Shedding with Application to Marine Tubulars and Risers,’’ ASME J. Energy
关9兴 Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, McGraw- Resour. Technol., 104, pp. 2–13.
Hill, New York. 关22兴 Bearman, P. W., 1984, ‘‘Vortex Shedding from Oscillating Bluff Bodies,’’
关10兴 Gaskell, P. H., and Lau, A. K. C., 1988, ‘‘Curvature-Compensated Convective Annu. Rev. Fluid Mech., 16, pp. 195–222.
Transport: Smart, A New Boundedness Preserving Transport Algorithm,’’ Int. 关23兴 Turan, Ö. F., and Lu, S., 1996, ‘‘Numerical prediction of Flow Fields around
J. Numer. Methods Fluids, 8, No. 6, pp. 617–641. Circular Cylinders: Forced Motion and Dynamic Response Cases,’’ ASME
关11兴 Leonard, B. P., 1979, ‘‘A Stable and Accurate Convective Modelling Proce- Fluids Engineering Division Conference, FED-Vol. 237, pp. 647–652.

714 Õ Vol. 122, DECEMBER 2000 Transactions of the ASME

Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm