J . Fluid Mech. (1973), vol. 61, part 4, pp.

769-784 769
Printed in. Breat Britain

Stability of a circular cylinder oscillating in

uniform flow or in a wake
By Y . TANIDA, A. OKAJIMA? AND Y. WATANABES
Institute of Space and Aeronautical Science, University of Tokyo

(Received 13 March 1973)

The lift and drag forces were measured on both a single circular cylinder and
tandem circular cylinders in uniform flow at Reynolds numbers from 40 to 104,
60 investigate the stability of an oscillating cylinder. A cylinder (the down-
sbream one in the tandem case) was made to oscillate in either the transverse or
longitudinal direction (perpendicular or parallel to the stream). I n the case of a
single cylinder, its oscillation causes the so-called synchronization in a frequency
range around the Strouhal frequency (transverse mode) or double the Strouhal
frequency (longitudinal mode). The aerodynamic damping for transverse oscilla-
tion becomes negative in the synchronization range. I n the case of tandem
cylinders, at low Reynolds numbers in the pure Kltrmh range synchronization
was observed to occur only when the downstream cylinder oscillated inside the
vortex-formation region of the upstream one, and at high (low subcritical)
Reynolds numbers synchronization occurred irrespective of the cylinder spacing
in either oscillating mode. I n the tandem case, too, the transverse oscillation of
the downstream cylinder becomes unstable in the range of synchronization.

1. Introduction
A considerable amount of discussion is found in the literature concerning the
wind-induced oscillations of smoke-stacks, cables and obher structural forms.
Information about the fluctuating forces on an oscillating cylinder is of special
interest in aeroelasticity as well as for a basic understanding of fluid mechanics.
However, only a few studies have been made for a circular cylinder oscillating
in a uniform flow. Bishop & Hassan (1964), Jones (1968) and some others
measured the lift and drag forces acting on a, circular cylinder made to oscillate
in a direction perpendicular to the stream, and recently Okajima, Takata &
Asanuma (1971, 1972) studied experimentally and numerically the viscous
flow around a torsionally oscillating circular cylinder. I n these studies, the
existence of aeroelastic resonance was confirmed.
There is a growing interest concerning another kind of instability that occurs
in a wake from an upstream obstacle. For example, under certain flow conditions,
serious aeroelastic instabilities can be observed in the array of pipes of a heat
exchanger (Funakawa 1968) and in conductor bundles (Calvert 1969). These
t Present address : Research Institute of Applied Mechanics, Kyushu University,
Fukuoke, Japan.
$ Present address : Ishikawajima-Harima Heavy Industries, Tanashi, Tokyo, Japan.
49 F L M 61
770 Y , Tanida, A . Okujima and Y. Watalzabe

FIGURE
1. Experimental set-up.

inskabilities may be attributable not only to the forced oscillation of the down-
stream cylinder due to the non-uniformity and unsteadiness of khe wake flow,
but also to the resonance in the aeroelastic system.
The aim of the present study is to measure the fluctuating lift and drag forces
acbing on a circular cylinder which oscillates either in uniform flow or in a wake,
and to assess the possibility of aeroelastic instability, which may occur when
the aerodynamic damping force becomes negative. The results obbained are for
the following cases.
( a ) A single circular cylinder either stationary or oscillated in one direction,
perpendicular or parallel to the stream.
( b ) A tandem arrangemen%with the upstream circular cylinder a b rest and
the downstream one either stationary or oscillated in a manner similar to that
in the case (a).
The range of Reynolds numbers covered in this experiment is 40-104.

2. Experimental apparatus
2.1. Generul remarks
As shown in figure 1, the test cylinders ( I ) and (2), hanging vertically from the
carriage (4),are towed in a still-liquid tank (3), which is 0-7 m wide, 0.4 m
deep and about 10 m long. The carriage is pulled on tracks by towing ropes,
which are wound round a drum by an electric variable-speed motor. Hence the
uniformity of the velocity distribution in the spanwise direction of the cylinder
is sufficiently good, and a low rate of turbulence in the flow can be expected.
The effects of the free liquid surface and the clearance between the cylinder
 Stability of an oscillatilzg cylinder 771
and the tank bottom? scarcely affect the measurements as long slender cylinders
are used. This was confirmed by a preliminary test, in which the effects of end
plates were examined.$ Furthermore, the measurements are performed on the
central section of the test cylinder, so as to minimize three-dimensional flow
effects.
I n the tandem arrangement, one cylinder is held fixed upstream, and another
is placed directly downstream of it. The downstream cylinder, (2) in figure 1,
can be made to oscillate transversely with a simple harmonic motion by an
oscillating mechanism on the carriage; that is, an electric variable-speed motor
(7) drives %heoscillation of the test cylinder through a gear train (8) and Scotch-
yoke mechanism (9) suspended by parallel leaf springs (6). The amplitude of
oscillation is selected as 4-2 mm, and the direction of oscillation, either per-
pendicular or parallel to the stream, can be changed by rotating the turntable
(5). The upstream cylinder (1) is mounted at an arbitrary point on the brackeb
( l o ) , and bhe distance between the centres of the two cylinders is adjustable
in the range 2.5-20 cylinder diameters. I n the case of a single cylinder, only the
downstream cylinder is used, the upstream one being removed.
The use of water or oil as the fluid makes it possible to measure the unsteady
aerodynamic forces on the oscillating cylinder with reasonable precision. The
liquid temperature can be made as uniform as is necessary, by insulating the
tank. The available Reynolds numbers are thus about 40-150 for oil and 103-104
for water.
The time for each experimental run is more than 20 s, being sufficiently long
for the measurements to be performed with no detectable transient starting
effects.
2.2. Test cylinders
Figure 2 shows a test cylinder, which is 30 mm in diameter and 352 mm long,
being divided into three. The upper and lower sections ( a )and ( c ) are fastened
to each other by four parallel connecting rods (f).The central section ( b ) is
100 mm long and suspended from the upper dummy cylinder by two parallel
leaf springs ( d ) of phosphor bronze on which two pairs of semi-conductor strain
gauges ( e ) are patched to form a Wheatstone bridge that is sensitive to the
force acting in one direction only. The leads t o the strain gauges run out through
the centre hole of the upper cylinder. The gauges and the leads are protected
from the liquid by a soft wax coating.
The central section of the cylinder is made of hollow acrylic plastic, so that
the unsteady aerodynamic forces acting on it can be deduced with a reasonable
accuracy neglecting the inertial force even when the cylinder is oscillated. The
natural frequency of the central section is around 40Hz in water, which is

t The bottom of the tank is made flat but slightly inclined for drainage, so that the
clearance between the cylinder and the bottom of the tank varies continuously between
3 and 20 mm.
2 Two thin circular end plates 8Omm in diameter were attached to the cylinder, at
the free end of the cylinder and at a position just under the free surface of the liquid.
The effect of the end plates on the measurements could scarcely be detected, so the
present experiments were carried out without the end plates.
49-2
772 Y . Tanida, A . Olcajima and Y . Watanabe

Liquid surface

E:
c

Tank bottom
,,, ",,, , ,,'
'
i
1
::
E
c
ri
m

2. Details of test cylinder.

FIGURE
 Stability of an oscillating cylinder 773
satisfactorily high in the present study. The gaps between these three sections
are adjusted to be 0.2 mm or less; these were also checked by a preliminary test
to have no perceptible influence on the measurements.
2.3. Instrumentation
I n the present experiment, the lift and drag forces exerted on the cylinder are
measured in separate runs. As mentioned previously, these forces are detected
by the strain gauges fitted upon the beams supporting the central section of the
test cylinder. The use of semi-conductor strain gauges of the p-n type makes
the overall sensitivity of the system very high without loss of rigidiby. The
signals from the strain gauges are recorded by an ultraviolet recorder through
amplifiers and low-pass filters.
When the direction of the cylinder oscillation coincides with that of the force
to be measured, it ofken becomes difficult to deduce the fluctuating lift or drag
force at high frequency. I n this case, the inertial forces of the real and apparent
mass are subtracted electrically from the overall signals, and only the part of the
force in phase with the oscillating velocity of the cylinder is examined. This is
similar to the method adopted by Bishop & Hassan (1964), and is sufficient for
the assessment of the instability.
In operation, the driving frequency of oscillation is set in advance by reading
directly the photoelectric tachometer, which counts the revolutions of the motor
driving the oscillating mechanism. The displacement of the oscillating cylinder
is also recorded by the ultraviolet recorder. Further, for reference, the frequency
of vortex shedding is detected by a hot-wire anemometer whose probe is im-
mersed in the wake of the cylinder.

2.4. M e t M of data reduction

The lift and drag forces per unit length for a cylinder are expressed respectively
as L = ipU'dOL, D = +pU2d(C,+OD),
where p is the density of the fluid, U is bhe stream velocity, d is the diameter of
the cylinder, CL and C, are the non-dimensionalized lift and drag forces, and
the overbar and tilde denote steady and fluctuating components, respectively.
I n uniform flow, a stationary circular cylinder is subjected to, besides the
mean drag force, fluctuating lift and drag forces, with frequencies fk and 2fk
respectively, owing to the alternating vortex shedding. As is well known, this
can be characterized by a non-dimensional parameter S, = f k d / U known as the
Strouhal frequency.
When the cylinder is forced to oscillate with a frequency f, that is appreciably
smaller or larger than the Strouhal frequency, the fluctuating forces consist of
two parts: one is due to the vortex shedding (whose frequency fu = fi,) and the
other results from the cylinder oscillation. When the driving frequency of the
cylinder approaches the Strouhal frequency, the two sets of forces become
synchronized, and the cylinder and wake system oscillates a k the imposed
frequency f,, when fe = f,. Then the fluctuating lift and drag forces oscillate
regularly with fairly constant amplitudes. Outside the range of synchronization
774 Y . Tanida, A . Okajima and Y . Watanabe
as well as in the stationary case, the amplitudes of the forces are not always
constant, changing with nearly periodic beating wave forms. The records axe
then processed by averaging the amplitudes over a long period. The frequencies
of the cylinder oscillation and the vortex shedding will be reduced into non-
dimensional form as S, = f c d / U and X,, = f,d/ U respectively.
The stability of the cylinder oscillation will be examined from a purely aero-
dynamic point of view, and it can be assessed from the work done by the hydro-
dynamic forces in the direction of oscillation. For nearly sinusoidal force fluctua-
tions, or assuming that the fluctuating force c" = [b]cos (2nfct)for a cylinder
oscillation a cos (2nfct),the stability of the cylinder oscillation depends on the
force component in phase with the oscillating velocity, i.e. the instability is
probable when Im [c"] > 0, which means negative aerodynamic damping.
I n the following the results will be shown mostly for Reynolds numbers of
80 and 4000 (or 3400), being typical members of the ranges of low and high
Reynolds numbers in the present study. The Reynolds number is of course
defined 8s usual as Re = Ud/v,where v is the kinematic viscosity.

3. Results for a single circular cylinder

3.1. Preliminary remarks on a single stationary cylinder
For reference in the evaluation of the results shown later, a preliminary experi-
ment was performed on a single stationary circular cylinder in uniform flow.
As is expected from a number of previous works, the fluctuations in lift force
are not observed for R e < 40, which means that there is no vortex shedding
behind Dhe cylinder. In the pure K&rmitnrange of Reynolds numbers, a regularly
oscillating lift force with fairly constant amplitude can be seen. As the Reynolds
number increases, the lift force becomes irregular, fluctuating with a beating
wave form.
Figure 3 gives the measured lift and drag forces and the Strouhal frequency
versus the Reynolds number, compared with the results of the previous work
(Relf & Simmons 1924; Roshko 1953; Delany & Sorensen 1953; Tritton 1959;
Gerrard 1961). For the fluctuating lift, not only averaged values but also maxima
and minima are shown in the figure. The fluctuating drag force is too small to
measure accurately and is not shown here.

3.2. Case of a cylinder oscillating transversely

Figures 4 and 5 give Dhe results for a single cylinder oscillating in the transverse
(perpendicular to the stream) direction at Re = 80 and 4000, respectively,
showing the reciprocal of the vortex-shedding frequency divided by the forcing
frequency, S,lS,u,the fluctuating lift component I m [eL] in phase with the oscillat-
ing velocity, and the mean drag force cD. Similar responses are obtained in the
ranges of low and high Reynolds number although there is a slight shift of
the synchronization range between them.
When the oscillating frequency is either small or large enough, or with the
frequency detuned more than 35 !k away from the Strouhal frequency, the
vortex shedding occurs at the natural Strouhal frequency irrespective of the
 Stability of an oscillating cylinder 775

,s 20--h- ,
15

1.0

0.5 -
Trittoii

(4
Relf &2 Siininons

0 0 0 0 0

D c h y & Sorenscii

0 - I I 1 I 1 t I 1 I

oscillation, and the two sets of fluctuating forces are superposed upon each
other. As the driving frequency approaches the Strouhal frequency, however,
the system is synchronized to oscillate at the forcing frequency, and vortices
are shed with frequencyfc or SJS, = 1. The synchronization occurs over a finite
frequency range, in which ( a )the lift force oscillates regularly with fairly constant;
amplitude and takes a maximum value at around the centre of the synchroniza-
bion range, being accompanied by a sudden phase lag, and ( b ) the mean drag
force also takes a maximum value at around the centre of the range.
Hence, in the lower half of the synchronization range, the fluctuating lift
component Im [CL]in phase with the oscillating velocity becomes positive, as
shown in figures 4 ( b ) and 5 ( b ) , and then it is possible for the single cylinder to
oscillate autonomously owing to the hydrodynamic excitation. These results
agree with those of the previous work, e.g. by Tanaka & Takahara (1970).
 -3
-3
Q,

0 0.1 0.2 0.3 0.4

S' SC
0.6 I I
1 (h)

0.4 0.4 -
t
0.2 -
r-l
c
1
t, I I
Y I ,
0 b
0.1 0.3 0.4
Stable SC
- 0
'-
0.2 1 Stable -0.3
\ I
Tanaka & Takahara A\
-0.4 1 - 0.4

1.4 1

0.8 I I I 1 I
05 0 0.1 0.2 0.3 0.4
0 0.1 0.2 0.3
8,
FIGUBE4. Case of a single cylinder oscillating FIGURE 5. Case of a single cylinder oscillating transversely.
transversely. Re = 80. (a)Oscillation/vortex- Re = 4000. (a)Osoillation/vortex-sheddingfrequency ratio ;
shedding frequency ratio ; ( b )in-phase lift com- ( b )in-phase lift component,; ( c )mean drag.
ponent; ( c ) mean drag.
 Stability of a n oscillating cylinder 777

3.3. Case of a cylinder oscillating longitudinally

I n figures 6 and 7 the results for a circular cylinder oscillating in the longitudinal
(parallel t o the stream) direction in uniform flow are given. Synchronization
can be observed in a range around double the Strouhal frequency, where vortices
are shed with a frequency half the imposed one. Synchronization of this type
has been already found by Tatsuno (1972) in his flow visualization experiment.
Figures 6 and 7 show that the fluctuating lift and mean drag forces take
maxima in the middle of the synchronization range, as in the case of transverse
oscillation. In figure 7 ( b ) for the higher Reynolds number, the vertical lines in
the synchronization range denote that periods in which the lift fluctuation
vanishes arise alternatively. Then it is probable that the oscillation of this mode
acts not only to excite the vortex shedding but also to suppress it occasionally.
The fluctuating drag component Im [OD]in phase with the oscillating velocity
is always negative in this case, so that the longitudinal oscillation is stable
from the autonomous point of view as far as the present experiment is con-
cerned. However, the instability is likely to occur at much higher Reynolds
number, because Im [OD]tends to become positive in the synchronization range
as the Reynolds number increases, deviating from the quasi-steady value as
shown in figure 7 (c).

4. Results for tandem circular cylinders

4.1. Case of both cylinders at rest
Figures 8 and 9 give the measured mean drag and fluctuating lift forces acting
on two stationary cylinders in tandem a t Re = 80 and 3400 respectively, to- I

gether with the Strouhal frequency, for separations up to 20 diameters, i.e. for
s 5 20d. At the higher Reynolds number, the fluctuating lift on both cylinders
changes with a, beating wave form, the maxima and minima of which are shown
by vertical lines in figure 9 (b).
For separations smaller than 5 diameters at Re = 80 and 3 diameters at
Re = 3400, no fluctuations in lift can be observed. This implies that the down-
stream cylinder suppresses the vortex shedding from the upstream one, playing a
role like that of a splitter plate behind a cylinder. This is supported by Thomas &
Kraus’s (1964) results of flow visualization for S/d = 3-6 at Re = 62. It is noted
that the critical separation for the vortex shedding, sld = 5 at Re = 80 and
sld + 3 at Re = 3400 in the present case, is consistent with the lengbh of the
so-called vortex-formation region given in the previous work (Schaefer &
Eskinazi 1959; Bloor & Gerrard 1966).
For separations larger than the critical value, vortices are shed periodically
from the upstream cylinder, and both cylinders undergo fluctuations in lift and
drag forces with common frequencies. As the separation increases, the amplitude
]cLll of the fluctuating lift acting on the upstream cylinder soon settles at the
value for a single stationary cylinder (indicated by an arrow in the figures).
On the downstream cylinder, however, the lift amplitude 10Lzlsharply increases
with increasing separation and reaches a maximum that is 8 to 10 times as large
 I 4
I I
4
00

0 0.1 0.2 0.3 0.4 0.5

JSC
S,
1-0 I 1

s,

-1
Quasl-stcady

FIGURE 6.Case of a single cylinder oscillating FIGURE 7. Case of a single cylinder oscillating
longitudinally. Re = 80. (a)Oscillation/vortex- longitudinally. Re = 4000. (a)Oscillation/vortex-
shedding frequency ratio ; ( b )lift; (c) mean drag shedding frequency ratio ; ( b ) lift; ( 6 ) mean drag
and in-phase drag component. end in-phase drag component..
 ri
0

‘0
8
3

m
0,
....
5
0
.-..
3
*
‘“
/6-
Stability of a n oscillating cylinder

v
3
t

FIGURE 8. Characteristicsof tandem stationary FIGURE 9. Characteristicsof tandem stationary

cylinders. R e = 80. ( a ) Mean. drag; ( b ) lift; cylinders. Re = 3400. ( a ) Mean drag; ( b ) lift j
778

(c) Strouhal frequency. (c) Strouhal frequency.

t*
780 Y . Tanida, A . Okajima and Y . Watanabe

0 0.1 02 0.3
SC
0.6 I I

FIUURE 10. Case of downstream cylinder oscillating transversely. Re = 80. ( a )Oscillation/

vortex-shedding frequency ratio ; ( b )in-phase lift components on the downstream cylinder.
0 , sld = 2.5; 0, s1d = 3; A,81d = 4; 0, s/d = 5 ; 0 , sld = 15.

as that for a single cylinder, and remains large even at sld = 20, although it
gradually decreases until it reaches the value for a single cylinder at infinite
separation.
The Strouhal frequency S, in this case is slightly lower for small separations
than in the case of single cylinder, but it soon recovers with increasing separation.
A t Re = 3400, it is seen that S, increases again for very small separations near
the critical point.
As shown in figures 8 (a)and 9 (a),in both the range of low and of high Reynolds
numbers, the mean drag force cDlacting on the upstream cylinder is slightly
subject to the influence of the downstream one only when the separation be-
tween them is very small. On the other hand, the front surface of the downstream
cylinder is strongly affected by the negative pressure of the upstream wake, so
that its mean drag force CDzdecreases remarkably at small spacings. At much
higher Reynolds numbers, the remarkable decrease in the mean drag may
occur on the downstream cylinder over a wide range of separations (Biermann &
Herrnstein 1933).
 Stability of an oscillating cylinder 781

1 I I

Unstable

FIQURE11. Case of downstream cylinder oscillating transversely. Re = 4000. ( a ) UEcilla-

tion/vortex-shedding frequency ratios ; ( b ) in-phase lift components of the downstream
s/d = 3 ; A,s/d = 4; 0 , B/d = 10.
cylinder. 0 , s/d = 2 . 5 ; 0,

4.2. Case of downstream cylinder oscillating transversely

Figure 10 shows that, when the downstream cylinder oscillates t ransversely at
Re = 80, synchronization does not occur for separations s/d > 5; that is, irres-
pective of the oscillation of the downstream cylinder, vortices are s hed from tl;e
upstream cylinder with Che natural Strouhal frequency as in the case of both
cylinders at rest, being washed downstream without being much i nfluenced by
the downstream cylinder. For separations s/d c; 5, when no vort ex shedding is
observed in the stationary case, the oscillation of the downstr earn cylinder
forces vortex shedding, just as an aerofoil with an oscillating trail ing edge does,
and brings about synchronization. In the range of synchronizatio n ,the ampli-
tudes of the fluctuating lifts on the two cylinders increase simulta n eously, and
the in-phase lift component Im [CL2] on the downstream oscilla ti ng cylinder
becomes positive, which leads to the possibility of instability as i n the case of
a single cylinder.
Figure I1 shows that at R e = 4000 synchronization can be de tected for all
782 Y . Tanida, A . Okajima and Y . Watanabe

FIGURE 12. Case of downstream cylinder oscillating longitudinally. Re = 4000. (a)Oscil-

lation/vort,ex-shedding frequency ratios ; ( b ) mean drag on the downstream cylinder ; ( c )
in-phase drag components on the downst,ream cylinder. 0,s/d = 3 ; 0, s/d = 5 ; 0 ,
s/d = 7 .

cylinder separations with which the present experiments is concerned, although

it is more or less random and accompanied by temporary pauses. When the
downstream cylinder is located inside the vortex-formation region of the up-
stream one, or for s/d < 3, the fluctuating lifts on both cylinders synchronize
with the imposed oscillation. However, when the downstream cylinder is located
far downstream, or for sld > 3, the fluctuating lift on the downstream cylinder
synchronizes with the oscillation, whereas the upstream cylinder is scarcely
affected by the oscillation of the downstream one. Figure 11 ( b ) gives the in-
phase lift component Im [cL2]
acting on the downstream cylinder, showing the
negative aerodynamic damping effect in the range of synchronization as in the
case of the single cylinder.
 Stability of a n oscillating cylinder 783

4.3. Case of downstream cylinder oscillating longitudinally

At low Reynolds numbers such as Re = 80, synchronization cannot be induced
by longitudinal oscillation of the downstream cylinder for any separation in the
experiment. For smaller separations such as sld < 5, no vortex shedding is
observed in spite of the oscillation of the downstream cylinder, and for sld > 5,
vortices are shed almost regardless of the oscillation, and the fluctuating lifts
on both cylinders change at a common frequency. There is no remarkable feature
in the fluctxating drag component in phase with the oscillating velocity, and it
is almost proportional to the oscillating frequency, yielding a positive damping
effect.
At a high Reynolds number Re = 4000, figure 12 shows that synchronization
can be detected for any separation over 3 diameters, when the downstream
cylinder may be outside the vortex-formation region of the upstream one. How-
ever, the effect of synchronization is so limited thab vortices are shed from the
upstream cylinder as in the case of single cylinder, while the vortex shedding
from the downstream cylinder is performed at half the driving frequency.
Hence, the synchronization does not greatly affect the mean drag forces on both
cylinders, and the damping force Im [OD,] acting on the downstream cylinder
is always negative, although it has a tendency to become positive owing to
synchronization.

5. Concluding remarks
The stability of a circular cylinder oscillating in uniform flow or in a wake
has been studied by measuring the lift and drag forces in the case of a single
cylinder and tandem cylinders. It was found that, in either case, the aerodynamic
damping force becomes negative when a cylinder (the downstream one in the
tandem case) oscillabes transversely in the synchronization range; that is, the
transverse oscillation of the cylinder may become unstable when the cylinder
motion and the vortex shedding are synchronized.

The authors are indebted to Prof. T. Asanuma of the Institute of Space and
Aeronautical Science, University of Tokyo, for his invaluable advice.

REFERENCES
BIERMANN, D. & HERRNSTEIN, W. H. 1933 N.A.C.A. Tech. Rep. no. 468.
BISHOP,R. E. D. & HASSAN, A. Y. 1964 Proc. Roy. SOC.A 277, 51.
BLOOR,M. S. & GERRARD, J. H. 1966 Proc. Roy. Soc. A 294, 319.
CALVERT, J. R. 1969 Central Electricity Generating Board Rep. RD/M/N377.
DELANY, N. K. t SORENSEN, N. E. 1953 N.A.C.A. Tech. Note, no. 3038.
FUNAKAWA, M. 1968 J . Japan SOC. Mech. Engrs, 71, 1477 (in Japanese).
GERRARD, J. H. 1961 J . Fluid Mech. 11, 244.
JONES,G. W. 1968 A.S.M.E. Paper, 68-FE-36.
OKAJIMA, A,, TAKATA, H. & ASANUMA, T. 1971 Trans. Japan Soo. Mech. En.grs, 37,2309
(in Japanese).
784 Y . Tanida, A . Okajima and Y . Watanabe
OKAJIMA, A., TAEATA, H. & ASANUMA, T. 1972 Trans. Japan SOC.Mech. Engrs, 38,2044
(in Japanese),
RELF,E. F. & SIMMONS, L. F. G. 1924 Aero. Res. Counc. R . &I M . no. 917.
ROSHKO, A. 1953 N.A.C.A. Tech. Note, no. 2913.
SCHAEFER, J. W. & ESKINAZI,S. 1959 J . Fluid Mech. 6, 241.
TANAKA,H. & TAKAHARA, S . 1970 Ilfitmbishi Heavy I d .Tech. Rev. 7 , 123 (in Japanese).
TATSUNO, M. 1972 Bull. Res. Inst. Appl. Mech., Kyushu Univemity, 36, 25 (in Japanese).
THOMAS, D. 0. & KRAUS,K. A. 1964 J. Appl. Phys. 35, 3458.
TRITTON, D . J . 1959 J . Fluid Mech. 6 , 547.