Professional Documents
Culture Documents
Elsevier
Summary
This paper reviews the literature on fluctuating lift on circular cylinders in cross-flow, both in
smooth and in turbulent uniform flows. Data of many authors on fluctuating lift and fluctuating
pressure at a section of a circular cylinder, and on the spanwise loss of correlation of fluctuating
lift, are critically analysed in terms of an equivalent Reynolds number Re (k/d)o.s. The results are
summarized in tables and graphs, and can be applied to problems of dynamic loading of structures,
and also tc relate experimental results on fluctuating lift, obtained on a cylinder slice of finite
length l, to the fluctuating lift coefficient at a section. The degree of coherence of the cylinder
wake, i.e. its more or less two-dimensional or three-dimensional character, may be evaluated from
the spanwise correlation length of fluctuating lift, ~tz.
Notation
CC circular cylinder
c~(0) sectional rms fluctuating lift coefficient
c~(/) rms fluctuating lift coefficient for a cylinder slice of length !
c~(0) rms fluctuating pressure coefficient
?
Cpb rms base pressure coefficient
?
Cpm rms maximum pressure coefficient
d cylinder diameter
k roughness height
ks coefficient proposed by Basu [3 ] to take into account the lack
of pressure correlations on a section
L cylinder total length
L.~ streamwise turbulence length scale
Re Reynolds number, R e = Ud/v
RBS cross-correlation coefficient of the fluctuating forces acting on
right and left half-sections (see Fig. 1 )
1. Introduction
right half-sectio~
U---.-)x ~ . .
left half-section
F~g. 1. Definition of coordinates.
181
results obtained with circular cylinders (CC) are presented. For the most part
these compilations encompass data that had not been gathered up tillnow and
which can be useful in the study of CC. In general the data proceed from wind
tunnel tests conducted in nearly two-dimensional conditions; both smooth flow
as well as uniform turbulent flow tests are fully analysed.
2. Discussion
In this formula the integral represents the value of C L (0) that would result
if the pressure fluctuations were to occur fully in phase at each of the half-
sections (cross-correlation coefficient--1) and fully out of phase between the
two half-sections (cross-correlation coefficientffi- 1 ). The term ½ (1 - RBS)
takes into account the lack of perfect correlation. The experimental data pre-
sented in Refs. [1,2 ] seem to confirm the validity of formula (2).
W h e n Rss is not provided, but only R9oo.27oo(the cross-correlation coefficient
of the pressures on 0 - _ 90 ° in a single cross-section), then it is possible to
adopt this value in (2) as a replacement of RBS. The substitution is reasonable,
since the sectional liftforce is dictated mainly by the pressure fluctuations in
the proximity of 0=90 ° (or 270 ° ).
Unfortunately some papers present the fluctuating pressure distribution,
but provide no measurements of sectional cross-correlation. In this case it is
still possible to estimate C~ (0), although in a less accurate way. Basu [3]
proposes the following expression:
182
f
C~(0)=kB %(0) sin 0d0.
0
(3)
bubble on
'4'-
one side strong kdrmdn
(D vortices reappear
,2_
N.-
~ ) / bubbleon /
strong kdrmdn _~ / both sides /
o
0
v°rtices st°p - ~ / ! ~ _ . _ /
Q
i , I
a , ) ultracritical
subcriticol critical supercritical
Reynolds Number
Fig. 2. Definition of flow regimes.
183
TABLE 1
Proposed values of kB for smooth CC in smooth flow of uniform turbulent flow (u'/U,~ 10%,
L~/d ~ 0.5 to 1.0 ) and corresponding values of RBs
Regime
I I , '°' i I I
-I.0 ' ' ' ' I ' ' '
!o, ~ o o ..A"AZ~
~
IO ~ -,~-- o~, --'--- Io lo \ " ' ~ 1 , Jz ULTRAC
g
o+pl "" \ I''+
+'° "il /I
;°"t
k,,
\! V c'
0¢
I- 0,0
0,2
0 _ l,O
I¢.
u.
i,u
o - 0.8
z -0.6
(2_ SUBC.~ ~ - ~ V ~ uLTRAC
1--
-I - 0.4
U,I
E
E
0 -0.2
:: /!
O
(n
I I / , s°'c°
v)
o 0,0
E 0 X
o
0.2
i
[2 ], u'/U= 10.5% ). For visual aid, vertical lines were drawn in Fig. 3, but they
are not to be taken as absolute limits between regimes.
For the most part the compiled results are in agreement with this approach.
The exceptions are two results of Kiya et al. [8] in turbulent flow at Re(k/
d)°"~'~ 160. They are well above the suggested values; however, it is observed
that the case yelding the smallest absolute value of cross-correlation coeffi-
cient measured by Kiya et al. (R9oo...,7oo=-0.25, see Fig. 3) in tact must be
classified on the subcritical-critical transition. The results of mean pressure
distribution and drag presented by Kiya et al. confirm this remark.
The other results (from all authors) are effectively in agreement with the
corresponding regime band. This can be verified from the analysis of mean
185
pressure distribution, drag coefficient, Strouhal number, etc. presented in each
case. The validity of the parameter R e ( k / d ) °'5 for the ordering of data, thus,
is confirmed to a certain extent.
For smooth CC k/d=O.O00035 was adopted, as proposed by Szechenyi [ 11 ].
This value seems better suited than the values in fact measured on nominally
"smooth CC". This might happen because the parameter R e ( k / d ) °'5 looses
significance for very small roughness (in the range of the -"aerodynamically
smooth").
It should be mentioned that some of the primary effects of turbulence (re-
garding CC) are iilustred by the curves in Fig. 3: turbulence triggers regime
transitions at lower Re, spreads out the critical and supercritical regimes a•d
diminishes the absolute value of the cross-correlation coefficients.
At last, it must be said that the results of Fig. 3 and Table 1 refer to turbu-
lence of moderate scale. It is suspected that turbulence of very small scale can
intensity the vortex shedding and, consequently, increase the absolute value
of the cross-correlation of the loads on one and the other side of the cylider.
2.2. Sectional lift coefficient versus lift coefficient for a cylinder of active
length l
Due to the lack of spanwise correlation, the lift coefficient measured for a
cylinder of active length l, C~. (l), is always smaller than the correspondent lift
coefficient measured for a cross-section, C~ (0). The question is how to esti-
mate C~ (0) when C~. (l) is provided or vice-versa. This will be possible when
information regarding spanwise cross-correlation of lift forces on several cross-
sections is available.
Kacker et al. [ 12 ], quoting Frenkiel, indicate the relationship between C~. (0)
and C ~ (1) as:
t [
l/t \ 0.5
C'L(l)/C'L (0)-----L(2 t (I--S)RLL (S)ds) , (4)
/
o
where s is the spanwise spacing (variable from 0 to l) and RLL (S) is the span-
wise cross-correlation coefficient of the lift forces. Unfortunatelly the results
of RLL (S) reported in the literature are very scarce, in part because measure-
ments of RLL (S) demand a relativelly complex instrumentation. The spanwise
cross-correlation coefficient of the pressures on the generator 0=90 °,
Rpp (90 o,8), however, provides a reasonable estimate of RLL (8). This happens
because from 60 ° to 120 ° Rpp (8,8) do not change very much, and this range is
the most important in the definition of RLL (8).
(Surry [13], Bruun and Davies [6], Kiya et al. [8], Sonneville [9], and
Batham [4 ] provide results of spanwise cross-correlation on generators posi-
tioned at several angles 8. From these results it is concluded that Rpp(~},s)
reaches a maximum just upstream from the separation line. Thus in the sub-
186
critical regime the pressures show better spanwise correlation on the generator
0 = 6 0 ° or 70 ° and, ii, general, this correlation diminishes for greater 0. In this
case to adopt Rpp (90 ° ,s ) as an estimate of RLL (S) is quite reasonable because
Rpp (90 °,s) ends up being approximately the mean value of Rpp (0,s) in the
range 60 ° to 120 °. In the ultracritical regime the pressures are better correlated
on the generator 0 = 9 0 ° or 100 °. In this case Rpp(0~,~q) may result in a slight
overestimation of RLL (S). In any case, in the range 60 ° to, 120 ° Rpp (0,s) do not
change considerably. )
By the other side, the spanwise cross-correlation co~f~cient of the velocities,
R,, (s), can also provide a reasonable estimate of RLL (S). These coefficients
will be similar since the velocity fluctuations are measured just outside the
boundary layer and on a line placed at 0>_-60°. Too far out frsm the separated
boundary layer or in a line 0< 60 ° the correlation of the approaching flow
predominates. Inside the separated boundary layer, i.e. in the wake, the veloc-
ity fluctuations loose spanwise correlation.
In effect, Sonneville [9 ] and El Baroudi [ 14 ] measured R,,(s) respectively
on lines positioned at 0=60 ° and 90 ° , while Kacker et al. [ 12 ] measured R,, (s)
on a line positioned 2.4d downstream of the cylinder. All measurements were
made just outside the boundary layer (see Fig. 4) and the results were quite
similar. (All tests with smooth CC, in smooth flow and in the subcritical
regime. )
Moreover, Sonneville [9 ] also carried out some measurements of spanwise
cross-correlation of the pressures on the generator 0 = 9 0 ° , and the measured
values agree with his own results of spanwise cross-correlation of velocities on
a line at 0=90 °.
Thus, supposing as discussed that Rps,(90°,s) and R,, (s) be a reasonable
estimate of RLL ( S ), and using formula ( 4 ), results of C ;. ( l ) / C ~ (0) were com-
puted from CC data of spanwise correlation provided by several authors. These
results are shown in Table 2.
The data gathered are scarce but indicate what should be expected. Exam-
TABLE 2
C~ (I)/C'L ( 0 ) values calculated for CC in smooth flow or uniform turbulent flow (u'/U,~ 10%,
L~/d=0.5 to 1.0)
ining the smooth flow data (see Table 2 ) it is observed that with the increase
of the non-dimensional spanwise spacing s/d, C'L(l)/C~ (0) diminishes slowly
in the subcritical regime, but sharply in the critical regime. This happens due
to the vortex shedding organization. While in the subcritical regime the vortex
~hedding shows a strong correlation, in the critical regime the main feature of
.he vortex shedding is its randomness.
In the ultracritical regime, as first reported by Roshko [16 ], the organized
vortex shedding reappears, in such a way that, again, CL (l)/C~ ( ) diminishes
slowly with the increase in s/d. The subcritical and ultracritical results shown
on Table 2 are similar, which would imply in a similar vortex shedding orga-
nization in these regimes. This remark, however, should be accepted with re-
servation, since the ultracriticalresults (Szechenyi [11 ] and Ribeiro [2 ] ) came
188
from tests with rough CC, and there is a suspicion that surface roughness can
improve the vortex shedding organizations.
The data obtained with CC in turbulent flow are still more scarce, and some
divergences are observed. In the subcritical regime, the values of C'L(l)/C'L (0)
computed from the results of Novak and Tanaka [ 7] are considerably larger
than the values measured directly by Surry [ 13 ].
Novak and Tanaka [ 7 ] measured spanwise cross-correlation of pressures on
the generator 0=60 °, and in the subcritical regime the pressures seem to be
better correlated just on this generator. Hence the resulting values of
C'L(I)/C'L(O) should be a little overestimated. Meanwhile the results of
C'L(l)/C'L (0) reported by Surry [ 13 ] are relatively small probabily because his
experimental set-up favored three-dimensionality: small blockage (3.9%), large
aspect ratio (l/d=38.7) and large turbulence intensity (14. 7% ). In any case,
the results from these authors can be interpreted as subcritical (minimum and
maximum) limits for CC in turbulent flow (of moderate scale and intensity
around 10% ).
Comparing with the subcritical values, the critical values of C~.(l)/C'L (0)
computed from the results of Bruun and Davies [6] indicate, as expected, a
reduction. In the critical regime, besides the loss of correlation introduced by
the turbulence, there is an additional loss due to the disorganization of the
K~rm~n vortices.
The larger ultracritical C'L(l)/C'L (0) values computed from Ribeiro [2 ] are
comparable to the subcritical values and reflect the comeback of strong K~r-
m~in vortices.
Examining the results obtained with CC in smooth or turbulent flow, alto-
gether, it is observed that in all regimes the introduction of turbulence (as it
disrupts the organization of the vortex shedding) diminishes the values of
C~.(t)/C'L (O).
As discussed, the spanwise cross-correlation data reported in the literature
are scarce, so that Table 2 remains to be completed. In any case, with the aid
of Table 2, and taking into account the experimental conditions in each par-
ticular case, it is possible to get a rough estimate of C~. (0) from C~. (l) or vice-
versa.
The problem demands attention when there are no dummy cylinders around
the active length, i.e. when the active length spans the test section. In this case
C~. (l) will be smaller than CL (0) not only due to the natural lack of spanwise
cross-correlation, but also due to the action of the tunnel walls. Close to the
walls the correlation of the vortices diminishes considerably.
The last remark could be observed in tests conducted by Kacker et al. [ 12 ].
These authors, using formula (4), estimated values of C'L(I)/C'L (0) and also
measured these values with the aid of load cells and dummy cylinders of vari-
able length. The estimated and measured values matched as far as the active
length remained smaller than one half ot the total length of the CC. When the
189
TABLE 3
Ref. 10 -5 Re 103 kid Regime u' /U L~/d L/d d/b 2z/d Note
(%) (%)
Notes:
a, b: Pressure cross-correlations measured respectively at 0 = 90 ° and 60 °.
c, d: Velocity cross-correlations measured just outside the boundary layer respectively at
0 = 90 o and 2.4d downstream from the CC.
B a t h a m [4 ] used rough CC fitted with sand grains and Ribeiro [2 ] used rough CC fitted
with sand paper, wire screen, and ribs (tested in smooth flow) and also (tested in tur-
bulent flow).
190
active length approached the total length the measured values were consider-
ably smaller than the estimated values.
Table 3 presents the experimental variables as adopted in programs where
measurements of spanwise cross-correlation of pressures or velocities were made
and also presents the corresponding correlation length measured. The same
values are depicted graphically in Fig. 5.
The results of Kacker et al. [ 12 ] and El Baroudi [ 14 ], obtained from cross-
correlation of velocities, diverge to some extent from the mean. It is suspected
that, depending on the exact position of the hot-wires, measurements of cross-
correlation of velocities provide greater values than the corresponding mea-
surements of pressures. (Some comments about this subject can be found in
Ref. [12] or Ref. [14].)
,%
; ; ' I ~o - , ' ; I ' ' ; ' I ; ' ' '
/
o SMOOTH FLOW
/ o
14 0 015
N
o ,=° 6 ~, '=
18 o 8 - ~ 06
°, \o, j'
%o6 ==,~
o
5 - TURBULENT FLOW -
3- a,3
..%
I oe
0
20 I00 = 4 e I000 I0000 I00000
R e ( k / d ) °'s
Fig. 5. Spanwise correlation length for CC in smooth flow or turbulent uniform flow (u' / U~ 10%,
L~/d~0.5 to 1.0).
191
The other results match fairly well. In smooth flow the spanwise cross-cor-
relation length (~z) reaches 3d or 4d in the subcritical regime (when the K~r-
m~n vortices are strong), with a pronounced decrease to ~ ld in the critical
regime (when the K~rmdn vortices are weak) and again reaches 3d or 4d in
the ultracritical regime (when strong K~rm~n vortices reappear).
Qualitatively the same behavior is observed for the case of CC in turbulent
flow. Quantitatively, however, the spanwise cross-correlation length dimin-
ishes in turbulent flow because in this condition the organization of the K~r-
m~n vortices is not as good as in smooth flow. Mean values of ~z regarding CC
in turbulent flow would be around 1.5d to 2.5d in the subcritical and ultracrit-
ical regimes and around 0.5d in the critical regime.
1C~.(6.8d) is the value of C~. measured (with load cells) from an active cylinder of length l-6.8d.
192
OmB I - -
I
0,7 - SMOOTH
FLOW
0.6 -
o
0.5 - 7
0.4 " A
29 2 b
30
b" o.3 / 28A
o
24
0.2 ' 26 ] b30
0,1 -
4 ,,i
0,0 ~ -- "6 :
'!, I I
•
I
, I
, i t t
23
;:
~2(
13
0.5 " TURBULENT
B FLOW
08 D
D
O,4 I?
0 v
_"', 0.3 2e
0
0.2 -
%
0.1
4
0.0 -.--
20 I00 s 4 s elO00 I0000 I00 000
Re(kid) °'s
Fig. 6. Sectional r,,s lift coefficient for CC in smooth flow or turbulent flow. Strips: / / / [ 2 1 ], i i I
[111, \ \ \ [201.
section, where the blockage ratio was 16%. Results presented in the literature
from tests where the blockage ratio was smaller than 10%, however, seem not
to diverge noticeably and indicate a maximum subcritical value of C~ (0) around
0.45 or 0.50.
From these remarks it is concluded that, regarding CC, blockage ratios smaller
than 10% do not exert an important effect on C~. Blockage ratios greater than
10%, as far as they contribute to improve the vortex shedding organization,
should exert an effect gradually greater on C~. The results from tests with
blockage ratios greater than 15% should be accepted with reservation.
The results presented by Gerrard [ 10 ] also diverge from the mean. The max-
imum subcritical value measured by Gerrard was 0.85, and he attributes this
large value to the small turbulence intensity of the wind tunnel (u' / U < 0.3% ).
Another reason might have been the dynamic flexibility of the setup used by
193
TABLE 4
C~. or Cp (0) measurements on CC. Principal experimental variables and information about the approach
followed to obtain C~. (0)
Notes:
a: Measurements of C~, (I) with load cells.CL (0) estimate with the aid of Table 2.
b: Measurements of c p (0), C L (0) computed using (3).
c: Measurements of c p (e) and partial measurements of cross-correlation.
C~ (0) computed using (2).
Measurements of Cp (0) and complete measurements of circumferentialcross-correlation.CL (0) pro-
vided by the authors.
C ~ (0) provided by the author from c p~t,,) (8) and supposing perfectcorrelation (out of phase ) between
half-sections.Cp(ru,~(0) isthe rms value of the liftcoeff,in a 1/3 of octave band (centred on the Strouhal
frequency).
Ci, (0) provided by the author from the realtime sum Of Cp (0).
g: Cp or C' provided by the author was corrected for blockage.
h: Test tunnel with gaps to compensate blockage, so that the effectiveblockage ratio is smaller than the
nominal value.
i: The tests conducted by James et al. [26] included C C with d=152, 316 and 452 ram. The C C with
d = 152 m m was provided with end platesof diameter 3d and 4d.
j: The mean fluctuatingvalue from crest to crest of the liftcoefficientwas provided by the author, The
rms value was estimated dividing the former by x/~.
k: Results in brackets indicate some uncertainty about the data.
h Measurements of peak values, rms values estimated from rms =peak/peak factor. C~ (0) computed
using (3).
194
.5 i ' i ' 1= I I ~ I I I I
o'1
26
.4 m O F SMOOTH FLOW
~7~ e 09
0 %
17
o A ""
0 --~29 0 26
04 2
.a,,
E A26
0 26 0
.2
62 27 27 o21q~026
44
.J -- ,.~ _
0 o n n
~7~8~ 6 17
% 0 13 o
.Q.
E % "4 -""---.-'- " = ' '
~ I k e -(Lila,~ 17
.2 m "q~u 6 ~ n ,?
% v:
°j -.-
0zO * t , I
IOC)
-
=
i ,
* *
, ,
I000
I i i , ,
I0000
i , , , ,
I00000
R e ( k / d ) °'s
Fig. 7. Maximum rms fluctuating pressure coefficient for CC in smooth flow or uniform turbulent
flow.
,5 ' " ' I ' ' ' ' I ' ' ' ' I I , , ,-
o4 i o SMOOTH FLOW
28
,3 -- 0 T,d~,--- ,,,
IIJ 0 0
17 31
o--------D ---.---~40
.2 -- o
z6 Z
jle
) !lw" .. z z z o ze
4 17 "4 val
°~1 02t
~
.4 -- O "& TURBULENT FLOW -
013
.3 -- 0
0
.2 -- 13 17 T4 Wait -.
Re(kid) °'~
Fig. 8. Wake rms fluctuating pressure coefficient for CC in smooth flow or uniform turbulent flow.
3. Conclusions
References
1 H. Loiseau and E. Szechenyi, Analyse exp~rimentale des portances sur un cylindre immobile
soumis ~ un ~coulement perpendiculaire ~ son axes ~ des nombres de Reynolds ~lev~s, La
Recherche A~rospatiale, 5 (1.972) 279-291.
2 J.L.D. Ribeiro, Effects of surface roughness on the two-dimensional flow past circular cyl-
inders, part 2: fluctuating forces and pressures, J. Wind Eng. Ind. Aerodyn., 37 (1991) 311-
326.
197
25 Y.C. Fung, Fluctuating lift and drag acting on a cylinder in a flow at supercritical Reynolds
numbers, J. Aerosp. Sci., 27 ( 11 ) ( 1960 ) 801-814.
26 W.D. James, S.W. Paris and G.N. Malcolm, A study of viscous cross-flow effects on circular
cylinders at high Reynolds numbers, Paper presented at AIAA 12th. Fluid and Plasma Dy-
namics Conference, Williamsburg, Virginia, July 23-25, 1979, 14 pp.
27 B.M. Franck, Roughness effect on the flow past circular cylinders at high Reynolds numbers,
Mast. thesis, Faculty of Graduate School of the University of Minnesota, Minneapolis, 1981,
76 pp.
28 M.J. Tunstall, Fluctuating pressures on circular cylinders in uniform and turbulent flows,
Lab. Note RD/L/N45-70, Central Electricity Research Laboratories, Leatherhead, Surrey,
UK, 1970.
29 D.M. Mc Gregor, An experimental investigation on the oscillating pressures on a circular
cylinder in a fluid stream, Toronto, University of Toronto UTIA Tech. note 14 (1957).
30 D.T. Hove, W.C.L. Shih and R.J. Curtis, Roughness effectson cylinder flows, Paper pre-
sented at the 8th. U.S. Na lal Congress of Applied Mechanics, University of California,
Los Angeles, June 26-30, 1978, 13 pp.
31 B. Cantwell and D. Coles, An experimental study of entrainment and transport in the tur-
bulent near wake of a circularcylinder,J. Fluid Mech. 136 (1983) 321-374.
32 H. Ruscheweyh, Wind loadingson the televisiontower, Hamburg, Germany, J. Ind. Aerodyn.,
1(4) (1976) 315-333.