Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.

pdf by The University Of Western Australia user on 09 August 2021


REVIEW—Review of Flow Interference
Between Two Circular Cylinders in
Various Arrangements
There are infinite numbers of possible arrangements of two parallel cylinders posi-
tioned at right angles to the approaching flow direction. Of the infinite arrangements,
two distinct groups may be identified: in one group, the cylinders are in a tandem ar-
rangement, one behind the other at any longitudinal spacing; and in the second group,
the cylinders face the flow side by side at any transverse spacing. All other combina-
tions of longitudinal and transverse spacings represent staggered arrangements.
The tandem arrangement will be treated first. A critical survey of previous research
revealed some "odd" features which had been observed and overlooked by various authors.
The discontinuity of vortex shedding implies that a similar discontinuity should be
expected for the drag force on both cylinders. The measurements of the front (gap)
pressures of the downstream cylinder and the base pressures of both cylinders at various
spacings reveal a discontinuous "jump" at some critical spacing. The discontinuity
M. M. ZDRAVKOVICH is caused by the abrupt change from one stable flow pattern to another at the critical
Reader, University of Salford, spacing. A new interpretation is given for the existing data on the drag force for both
Salford, England cylinders. The effects of Reynolds number and surface roughness are treated in some
detail.
Following this, two cylinders arranged side by side to the approaching flow are con-
sidered. All the available data on measured forces are compiled together with additional
measurements in the range of intermittent changes of drag and lift forces. The bistable
nature of the asymmetric flow pattern around each cylinder produces two alternative
values of the drag force coupled with two alternative values of tlie lift force. The intro-
duction of the interference force coefficient exposes the physical origin of two different
forces experienced by the cylinders when arranged side by side.
Finally, the least reported arrangement of two staggered cylinders is reviewed. The
various arrangements are grouped into classes according to the sign of the lift force, or
whether the drag force is greater or less than that for a single cylinder. The measure-
ments of drag and lift forces for various arrangements reveal two different regimes for
the lift force. In one regime, the lift force directed toward the wake of the upstream
cylinder is due to the entrainment of the flow into the fully developed wake of the up-
stream cylinder. The lift force in this regime reaches a maximum value when the
downstream cylinder is near to the upstream wake boundary. In the second regime, at
very small spacings, the lift force becomes very large due to an intense gap flow which
displaces the wake of the upstream cylinder. The maximum lift force occurs with the
downstream cylinder near to the horizontal axis of the upstream cylinder. A discon-
tinuity in the lift force for some staggered arrangements is found and attributed to the
bistable nature of the gap flow.

Introduction search which has been carried out mainly to solve immediate
practical problems. Practical applications stimulated research
The history of investigations of the flow around two circular but changed with time and appeared in various areas of engi-
cylinders is an example of unsystematic and fragmentary re- neering and science. Applications started in aeronautical engi-
Contributed by the Fluids Engineering Division for publication in the neering (twin struts to support wings), moved to hydronautical
JOURNAL OF FLOIDS ENGINEERING. Manuscript received at ASME Head- engineering (periscope, snorkel and radar mast vibrations), civil
quarters, September 12, 1977. engineering (twin chimney stacks in wind and jetties and off-

618 / DECEMBER 1977 Transactions of the ASME


Copyright © 1977 by ASME
shore structures in high seas), electrical engineering (twin-con-
ductor transmission line vibration), mechanical engineering (in TWICE RESISTANCE OF SINGLE WIRE
connection with heat exchanger tube vibrations) and chemical
engineering (pipe-rack forces). DIAMETERS" OF SEPARATION
It is a common practice to assume that two cylinders should
behave in a flow in a similar, or even an identical manner, to a
single cylinder. This assumption is justified only when the two
cylinders are sufficiently apart. The interference between two
cylinders at close proximity, however, drastically changes the
flow around them and produces unexpected forces and pressure
distributions, and intensifies or suppresses vortex shedding.

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
These changes of the flow pattern are systematically described WIRES IN CONTACT
and analyzed in the present review.
All the possible arrangements of two cylinders are grouped R e = 9-72 x IO'
into three sections. Tandem arrangements are treated first,
followed by the side-by-side arrangements and, at the end,
staggered arrangements.
All the possible flows around two circular cylinders may be
classified in two categories: with and without interference.
These two categories are divided by an interference boundary. 5 IO 15 20° ANGLE OF STAGGER
The interference may be either partial, when only one cylinder Fig. 1 Combined drag of two wires [1]
is affected, or combined, when both mutually interfere. It will
be shown when the interference is weak and when it is strong.

Tandem Arrangement ferent flow patterns might be responsible for that change in the
drag increments. It seems that the authors were puzzled by the
Early Force Measurements. The first interest in flow around very large increment between the spacings of 3 and 4 dia be-
two circular cylinders arranged one behind the other relative cause they introduced, only in that range, an intermediate
to the wind, began in aeronautical engineering in the early era spacing of 3y 2 dia marked with (!) in Fig. 1.
of biplanes. The common use of cylindrical and later stream- Biermann and Herrnstein [2] reported their force measure-
lined struts to connect two wings stimulated the investigations. ments from the 7 ft X 10 ft (2.1 X 3.0 m) wind tunnel. They
Pannell, Griffiths and Coales [l] 1 reported the results of extended the range of spacings between the centers of cylinders
force measurements on two parallel "circular wires." The dis- up to 9 dia. The drag force was measured separately for each
tance between the centers of the wires varied from 1 dia (wires cylinder. They introduced an "interference drag coefficient"
in contact) to 6 dia. The combined drag force was measured defined as the difference between the drag coefficient measured
for both wires exposed to the wind in tandem and also staggered on one of the cylinders in tandem and the drag coefficient of the
to the wind at various angles up to 20 deg. The angle of stagger single cylinder at the same Reynolds number. The combined
was that between the line joining the centers of the two wires interference drag was obtained by adding the interference drag
and the wind direction. The experimental results are reproduced coefficients of both cylinders in tandem.
in Fig. 1. Pannell, Griffiths, and Coales stated "It is interesting The experimental results are reproduced in Fig. 2. The
to notice that, in the case of the circular wires, the minimum minimum interference drag coefficient of the upstream cylinder,
drag on two wires in contact is only 40 percent of the drag on curves A, coincides with the "kink" of the interference drag
one wire alone." This was due to an improved "streamlining" coefficient of the downstream one, curves B. The authors
of the flow pattern for the latter case. apparently tried to locate the position of the minimum inter-
An additional feature of the drag force is also evident from ference drag coefficient of the upstream cylinder for Re = 1.63
Fig. 1. The increments in drag force for the spacings from 1 to X 105 by adding more experimental points in the range be-
2 dia, 2 to 3 and 3 to 4, increased more and more, respectively, tween 2V2 and 4 dia, as seen in Fig. 2. However, they did
while the corresponding increments between 4 and 6 dia are very not try to determine more precisely the kink on the interfer-
much smaller. These two sets of results indicate that two dif- ence drag coefficient curve for the downstream cylinder. They
calculated, instead, the combined interference drag coefficient;
curves C are shown in Fig. 2 with dotted lines. The kink is
'Numbers in brackets designate References at end of paper. more pronounced for the combined interference drag coefficient,

"Nomenclature-

spacing
CD = drag coefficient the cylinders' centers
D = outer diameter of cylinder
CDO = drag coefficient for single T/D =
transverse spacing ratio
/ = frequency of vortex shedding
cylinder U =
local velocity
L = longitudinal spacing be-
CD — CDO = interference drag coefficient V or Um =
free stream velocity
tween cylinders' centers
CL = lift coefficient longitudinal spacing ratio X/D =
nondimensional longitudi-
L/D
Cp = pressure coefficient nal coordinate
= base pressure coefficient at R or R6 Reynolds number, I —• J Y/D = nondimensional transverse
9 = 180 deg coordinate
Cpa = gap pressure coefficient at /D a = angle of stagger
8 = Strouhal number G = circumferential angle on the
G = 0 deg for down- V
stream cylinder at small transverse spacing between cylinder

Journal of Fluids Engineering DECEMBER 1977 / 619


which is an indication that both cylinders contribute to its Hence, the decrease of the interference drag, as observed by
occurrence. the previous authors, was caused by the drop of the base pres-
The two flow patterns, inferred from Fig. 1, affect the inter- sure only.
ference drag coefficients of two cylinders in differing ways. The The pressure distribution around the downstream cylinder
flow pattern for spacings less than about 3Vs dia tends to de- showed two unusual features. Firstly, the side facing the gap
crease the interference drag of the upstream cylinder to a mini- between the cylinders had a very low negative pressure, which
mum at about 3'A dia, and to increase the interference drag was almost the same as the corresponding value of the base
of the downstream cylinder up to a maximum at about 2'A pressure of the upstream cylinder. This fact is an indication
dia spacing beyond which a small decrease follows; see Fig. 2. that the flow in the gap is almost stagnant.
The second flow pattern for spacings greater than about 3y 2 The second unusual feature was that the negative gap pres-
dia (beyond the kink) tends to increase the interference drag of sure coefficient in front of the downstream cylinder exceeded
that on the base side behind. Hence, the downstream cylinder

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
both cylinders up to constant values (at least for the Reynolds
numbers tested). experienced a negative drag - thrust force. The decrease of the
Pressure Distributions and Velocity Profiles. Hori [3] Was the interference drag, as found by the previous authors, was caused
first who measured the pressure distribution around the two mainly by the corresponding increase of the gap pressure.
cylinders in a tandem arrangement. He carried out his measure- The two fairly flat pressure distributions in the gap and base
ments for only three spacings: 1.2, 2.0 and 3.0 dia. The pres- regions were separated by two symmetrically positioned maxi-
sure coefficient distribution is shown in Fig. 3 in polar coor- mums2 as seen in Fig. 3. It might be inferred from Fig. 3 that
dinates. The pressure distribution around the upstream cylinder they correspond to the reattachment of the flow separated from
showed that only the rear part of it was affected by the presence the upstream cylinder. The position of these reattachment
of the downstream cylinder. The base pressure coefficient in- points was symmetrical, as should be expected, and the dis-
creased as the spacing of the downstream cylinder increased and, tance between them reduced as the spacing was increased from
consequently, the drag of the upstream cylinder was reduced. 2 to 3; see Fig. 3.
Further measurements of the pressure distribution around
the downstream cylinder at higher Reynolds number have been
carried out by Zdravkovich and Stanhope [4]. The time-
average pressure coefficient distributions are given in Fig. 4
A — UPSTREAM CYLINDER for six different spacings as shown. The pressure coefficient
B — DOWNSTREAM CYLINDER distribution for the single cylinder at the same Reynolds number
C — COMBINED is given for comparison. There is a remarkable grouping of curves.
All curves for small spacings up to 3V2 dia are similar among
themselves and have the same features as those found by Hori
at lower Reynolds number.
The pressure coefficient distribution for the single cylinder
and for the tandem arrangements with spacings beyond 3V2
dia form a second group of similar curves. There is only one
maximum pressure at the stagnation point and no sign of the
reattachment peaks. The curves differed among themselves
mainly due to the reduction of the stagnation pressure coeffi-
cient from 1 (single cylinder) down to —0.2 for the 5 dia spacing.

7 8 9 L/D
2
SPACING RATIO Note that the stagnation point on the upstream cylinder represents also a
. Z Interference drag coefficient for tandem cylinders [2] maximum.

UPSTREAM CYLINDER DOWNSTREAM CYLINDER

Fig. 3 Pressure distribution around two cylinders in staggered ar-


range ment [3]

620 / DECEMBER 1977 Transactions of the ASME


(a)

Q-1 UfiO

-aura

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
Fig. 4 Pressure distribution around downstream cylinder in vari-
ous tandem arrangements [4]

A-WAKE
I 1 n
Therefore, it may be proposed that the two flow patterns have 7 O-t 0-9 ^
two types of pressure distribution around the downstream cyl-
inder:
(i) For small spacings (up to the kink) there are two sym-
metric peaks, presumably corresponding to the reattachment
points.
(ii) For greater spacings (beyond the kink) there is a single Fig. 5 Velocity profiles across the gap and wake [4]
peak corresponding to the stagnation point.
Further insight into the two flow patterns was gained by
traversing the gap between the cylinders and the wake behind
the downstream cylinder with a hot-wire. Zdravkovich and
Stanhope [4] measured three velocity profiles across the gap
and one across the wake of the downstream cylinder. The
traversing planes were situated at 1/i, l/%, and 3/i of the gap
lengths within the gap and 1/t of the gap length behind the
downstream cylinder.
The three velocity profiles within the gap were similar to each
other for all spacings up to the 3.5 dia. They showed that
a low velocity existed in the gap. Fig. 5(a) is an example for I 3 3 4 I t I ~K> 3 0 S 0 4 0 WD
L/D = 2.5. The Wake traverse, within the same range of Fig. 6 Strouhal number behind cylinders in tandem arrangement [7]
spacings, differed from the gap traverses because the magnitude
of the velocity was considerably greater in the wake behind
the downstream cylinder.
Beyond the spacing of 3.5 dia, the gap and wake profiles Oka, Kostic, and Sikmanovic [7] measured the vortex shed-
became similar and the curves were close together; see Fig. ding frequencies behind the upstream and downstream cylinders
5(b). The change between these two different flow patterns, in the intermediate subcritical Reynolds number range. There
i.e., almost no flow in the gap versus fully developed flow ap- was no distinct vortex shedding detectable behind the upstream
proaching the downstieam cylinder, should be expected to take cylinder up to the spacing L/D = 3.8, as shown in Fig. 6. Be-
place gradually as the spacing increased. The experiments, yond that spacing, the vortex shedding suddenly appeared and
however, revealed that there was a sudden change of the flow soon reached the value found behind the single cylinder.
pattern in the gap when the spacing was increased beyond Contrary to the upstream cylinder ease, the vortex shedding
3.5 dia. In the subsequent sections this feature will be discussed was detected in the whole range of spacings behind the down-
in more detail. stream cylinder. The Strouhal number (defined as the product
of the frequency of vortex shedding and diameter divided by
Vortex Shedding. The appearance of von Karman's vortex
the velocity) continuously decreased in the range 1 < L/D < 3.8
street in the wake of a single cylinder has been well known for
from values well above the Strouhal number for a single cyl-
a long time, but whether it takes place in the wakes of two cyl-
inder down to values well below it, as seen in Fig. 6. A sudden
inders in tandem was not investigated until recently. Smoke
"jump" occurred at the same spacing for which the vortex
observations of wakes of cylinders in a tandem arrangement
shedding was first detected behind the upstream cylinder. The
at low Reynolds numbers [5] revealed that a vortex street
Strouhal number approached the value found for the single
always forms behind the downstream cylinder, but only for
cylinder at first, but beyond about 8 dia it started to decrease
spacings greater than 4 dia behind the upstream cylinder.
again.3
Thomas and Kraus [6] observed (also at low Reynolds num-
ber, Re = 62) that when the spacing of cylinders was an odd
multiple of Va of the longitudinal spacing of the vortices of the
street the contraction or cancellation of the vortex street oc- 'Ishigai, et al. [91 found the same Strouhal number behind both cylinders,
curred downstream. see Fig. 8.

Journal of Fluids Engineering DECEMBER 1977 / 621


Novak [8] also measured the vortex shedding behind both
cylinders in tandem arrangements. The same trend was found
for the downstream cylinder but somehow (?) he did not detect
the jump in vortex shedding when the upstream cylinder com-
menced the vortex shedding. Ishigai, et al. [91, however, found
a distinct jump which will be described in detail in the next
section.
Novak [8] tried to establish the relationship between the
Strouhal numbers behind the upstream and downstream cyl-
inders and the Reynolds number. He found that the Strouhal
numbers for fixed spacings were constant in the range 0.45

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
X 10' S Re S 1.4 X 10'.
Novak [10] extended his measurements to cylinders of differ-
ent diameters; the ratios were 0.5 and 2, respectively. When
the small cylinder was placed upstream of the big one, the vor-
tex shedding behind the former started from the spacing of
2.25 dia (of the big one) and, for the downstream one, from
touching onwards. The reverse arrangement showed that the
small downstream cylinder could not suppress the vortex shedding
behind the big upstream one regardless of how close the former Fig. 7 Flow visualization of tandem arrangements [9J
was. The vortex shedding behind the downstream small cyl-
inder, however, was suppressed within 7 dia behind the big one.
The Two Flow Patterns The two flow patterns, first inferred
· 2 5 , . - - - -.....- - - - - - - - - ,

~2&p~pn:~::
from the force measurements, then clearly indicated from the
corresponding pressure distribution around cylinders and the
S=!Q
velocity profiles across the gap and wake, were confirmed also
by the vortex shedding measurements. V ~ xs00
The first flow pattern does not produce vortex shedding be- 5 .5
hind the upstream cylinder and distinctly influences the vortex '20 o
shedding behind the downstream one. Therefore, it should be
expected that an elastic upstream cylinder should be less prone
to vibrations induced by vortex shedding in the range of spac-
Cpb
ings 1 sLID s 3.5 than the downstream one. In fact, the
latter should be even more prone to vortex shedding induced '15 -'5
vibrations than the single cylinder due to a wide change of
Strouhal's numbers (0.15 S S s 0.26) in this range of spacings.
This has been confirmed experimentally [11] and [12] and in
field experience [13, 14].
5
The second flow pattern produces periodic vortex shedding
·10 -1'0
behind both cylinders. The commencement of vortex shedding
behind the upstream cylinder strongly affects and synchronizes 2 3 4 5 LID
the vortex shedding behind the downstream one. Hence, both Fig.8 Strouhal number and base pressure coefficient behind cylin-
cylinders should be equally prone to flow induced vibrations due ders in tandem arrangement [91 and a single cylinder with a splitter
to vortex shedding. plate after Roshko [15]

Finally, flow visualization cine-film frames by Ishigai, et al.


[9] showed convincingly the two flow patterns. They used
is also seen in Fig. 8; i.e., at the critical spacing, two different
heated cylinders and the Schlieren technique to reveal the struc-
ture of the flow. Reynolds numbers were from 2 X 10' to values of Strouhal number may intermittently exist.
4 X 10'. It is evident that there is no vortex shedding behind Ishigai, et al. compared their results with those obtained
the upstream cylinder in Fig. 7(a), (b), (c), and (d) and large with a splitter plate in the wake of a single cylinder, Roshko's
vortices can be seen in Fig. 7(e). Some kind of weak symmetric experiment [15]. There was a remarkable agreement in the
Strouhal number drop and the position of the critical spacing,
vortices can be seen in Fig. 7(d). TheY'induce in-line vibrations
of the upstream cylinder in water [12]. but also a discrepancy beyond the critical spacing. The reason
for the agreement lies in the analogous flow patterns for these
Discontinuous Change of the Two Flow Patterns. Ishigai, et al. two different geometries for spacings less than critical. In the
[9] were the first who measured simultaneously the vortex case of two circular cylinders, the flow which separates from
shedding behind the downstream cylinder and base pressures the upstream cylinder reattaches to the downstream one and
behind both cylinders, as shown in Fig. 8. They covered the this inhibits vortex shedding behind the upstream cylinder. In
range of spacings from 1 to 6 dia and reduced the base pressure the case of the circular cylinder and the splitter plate, the
data in terms of a base-pressure coefficient (defined as the dif- flow separated from the upstream cylinder re-attaches to the
ference between the pressure measured at 180 deg on the cyl- splitter plate and the vortex shedding is suppressed behind the
inders and the free-stream static pressure, divided by the free cylinder. Hence, in both cases, vortex shedding behind the
stream dynamic pressure). upstream cylinder should not be expected due to a "closed
Ishigai, et al. wrote: "The existence of a critical LID is wake" flow pattern.
evident in Fig. 8. The critical LID is equal to 3.8, below which Agreement should be expected for the Strouhal numbers meas-
the regular velocity fluctuation is observed only behind the ured behind the downstream cylinder and the splitter plate,
downstream cylinder. When LID > 3.8, the vortex streets for spacings less than critical, because of the same separation
are formed behind both cylinders and the vortex shedding fre- between the two shear layers. Abernathy and Kronauer [16]
quencies are the same." A bistable nature of the phenomenon have simulated on a computer the formation of a vortex street

622/ DECEMBER 1977 Transactions of the ASME


developed fiom two vortex layers without taking into account
the shape of the body which generated these layers.
Finally, the "disagreement" in Strouhal numbers observed
for spacings greater than critical might be due only to different
Eeynolds numbers.4 The discontinuous change of vortex shed- gap
ding strongly affected the base pressure which is an indication -CM
that a discontinuous change in flow pattern was responsible
for both. Fig. 8 shows that both base pressures decreased dis-
continuously at the critical spacing. The step-like drop of the
base pressure behind the upstream cylinder produced a similar -0-8 Cpbase
jump of the drag coefficient. This will be discussed in detail
in the next section.

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
The step-like drop of the base pressure behind the downstream m O 5-8 x IO
cylinder is smaller than that found behind the upstream cyl- V V 8'3 X IO 4
inder, as seen in Fig. 8. The fact that the vortex shedding D I I x IO"
existed behind the downstream cylinder for spacings less than
critical might suggest that the jump of the drag coefficient
should be smaller than for the upstream one. To settle this Fig. 9 Base and gap pressure coefficients of the downstream cylinder
question, additional information is necessary about changes of
the gap pressure on the front side of the downstream cylinder
at the critical spacing.
Discontinuous Change of Pressure. A simultaneous measure-
ment of the base pressure behind the downstream cylinder and [9] at lower Reynolds number is not in agreement with these
that facing the gap has been carried out in Salford [17] in the results. More research is necessary to clarify this anomaly.
upper subcritical range. Fig. 9 shows the base and gap pressure The drag coefficient is proportional to the difference of the
coefficients for three Reynolds numbers. The gap pressure coef- gap and base pressure. The difference between the two coeffi-
ficient had the lowest value when the cylinders were in contact cients, as seen in Fig. 9, rapidly decreases in the range between
and gradually increased up to the critical spacing. At the critical 1 and 2 dia spacing. Consequently, the negative drag is rapidly
spacing, the gap pressure coefficient jumped to a significantly diminished as observed in Fig. 2. In the next range of spac-
higher value and for the first time, exceeded the value of the ings, between 2 and 3 dia, the two coefficients either change
base pressure. There was a range of positions close to the critical in the same rate, as seen in Fig. 9, or the base pressure coef-
spacing in which the flow intermittently displayed low and high ficient increases faster than the gap pressure coefficient. The
gap pressures for some period of time. This confirmed the bi- consequences are that the drag coefficient does not change
stable nature of phenomenon. Beyond the critical spacing, the for the former or starts to decrease for the latter, as found in
gap pressure continued to rise because it became a "stagnation Fig. 2. Hence the kink on the drag curve is caused by the greater
point" pressure. The gap pressure followed the same trend as rate of increase of the base pressure over the gap pressure for
the base pressure of the upstream cylinder shown in Fig. 8, but spacings greater than 2.5 dia at the stated Reynolds numbers.
only up to the critical spacing, as should be expected. Ishigai, The discontinuous jumps of the gap and base pressure coeffi-
et al., [9] curve for the base pressure of the upstream cylinder cients at the critical spacing in the opposite directions cause a
showed the same trend up to the critical spacing, as seen in Fig. corresponding discontinuous jump of the drag coefficient which
8, but with an overall shift to higher Cp values. has not been shown in Fig. 2. This will be "corrected" in the
next portion of this paper.
The base pressure coefficient decreased up to the spacing of
2 dia. Above that spacing, it increased either at the same Finally, beyond the critical spacing, the gap pressure coeffi-
rate as the gap pressure coefficient or at a greater rate up to cient has a slightly higher rate of increase than the base pres-
the critical spacing. There was a discontinuous decrease of sure coefficient. The consequence is further increase of the
the base pressure coefficient at the critical spacing. Further drag coefficient at a slower rate as found in Fig. 2.
increase of spacing brought little change of the base pressure The pressure coefficients were also measured [17] on both
coefficient. By comparing the curves of the measured base sides of the downstream cylinder (90 deg from the stagnation
pressme at the three Reynolds numbers shown in Fig. 9 with point). Both side pressure coefficients also jumped from —0.8 to
the corresponding curve in Fig. 8 it is evident that they are not —1.4 at the critical spacing. It has been observed that pressures
similar. Two additional measurements have been carried out measured on opposite sides were not the same for spacings less
[17] for Re = 4.0 X IO4 and 2.5 X 104. The results for the than critical, but beyond the critical one, they became equal.
first one were similar to those in Fig. 9 and the second set just This indicated that the symmetry of the flow was improved
started to show some resemblance to the curve in Fig. 8. This for the latter.
Reynolds number effect will be discussed in detail in the next The discontinuous jump of two reattachment points into a
section. single stagnation point was also observed [17]. Fig. 10 shows
the position of the reattachment points as an angle along the
The critical spacing showed a dependence on the Reynolds
cylinder surface measured from the stagnation point. Four
number. Fig. 9 shows the observed critical spacings at the
different sets of measurements of pressure distribution around
three Reynolds numbers. The critical spacing was below L/D
the downstream cylinder were used to compile Fig. 10. The
= 3.5 for the lowest Reynolds numbei, equal to 3.5 for the
three different Reynolds numbers showed the same trend. There
Reynolds number 8.3 X 104 and above L/D = 3.5 for the
was an ultimate position of the reattachment points near to
highest Reynolds number. These results are in agreement with
the critical spacing, beyond which both points abruptly merged
the position of the kinks shown in Fig. 2 at the same Reynolds
into a single stagnation point at 0 deg.
numbers. However, the critical spacing found by Ishigai, et al.
Hence it may be summarized that at the critical spacing
the discontinuous changeover of the two flow patterns caused:

(i) Upstream cylinder:


4
Roshko's experiments were conducted in the upper subcritical range while
Ishigai, et al. were in the intermediate subcritical range. (a) commencement of vortex shedding,

Journal of Fluids Engineering DECEMBER 1977 / 623


(b) drop of the base-pressure coefficient, and
(c) jump of the drag coefficient.
O ISHIGAI ET AL
(ii) Downstream cylinder: O ZDRAVKOVICH

(a) disappearance of the reattachment points on the front X KOSTIC ET AL

side + HOR1

(6) drop of the base and side pressure coefficients


(c) jump in the vortex shedding frequency
(d) jump of the gap pressure and drag coefficients.
Drag Coefficient and Reynolds Number Effects. A variety of

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
measurements of drag forces on two circular cylinders in tandem
arrangement have been carried out by various authors at differ-
ent Reynolds numbers. Practical engineering problems, like
the twin chimney stacks and twin conductor lines in wind,
prompted research only for the spacing relevant to the par-
ticular application and for the staggered arrangements adjacent
to it. Some examples of these measurements will be given L/D
through three groups of papers. Counihan [18] carried out Fig. 10 Position of the reattachment point on the downstream cylin-
research on flow about smooth cylinders and stranded cables der [17]
but confined himself to only two spacings, 10.7 and 16 dia,
for two cylinders in tandem. Maekawa [19] tested only three
types of "conductors" at L/D = 8, but the Reynolds number
range covered subcritical and supercritical flow regimes. Price
[20] presented data for smooth cylinders at L/D = 5.0, 9.0
and 19.5 and stranded cables at L/D = 8 and 25.4 dia. He
also included the effect of turbulence on the drag force, but
confined himself to the downstream cylinder only. All these
results can be reproduced as single points in the plot of the drag
coefficient, CD, versus the spacing L/D.
Other authors covered a wide range of spacings in a narrow
range of Reynolds numbers. Cooper and Wardlaw [21] ex-
tended the range of spacings up to 50 dia for the Reynolds
number 5 X 104 only. Suzuki, et al. [22] calculated the drag
coefficients for both cylinders up to L/D = 11 from the pressure
distribution measurements at Reynolds number 2.3 X 10s only.
Tanida, et al. [23] measured the drag forces on both cylinders
towed in water up to L/D = 20 for the Reynolds numbers
equal to 80 and 2300. It is interesting to note that none of these
authors found the kink on the curve of the drag coefficient for
the downstream cylinder.
Nagai and Kurata [24] reproduced Biermann and Herrn-
stein's [2] curves from Hoerner's [25] book, but presented their
results without the kink for the downstream cylinder, and with- WMMANN ET AL
out the minimum for the upstream cylinder (!). Kostic and 0 WARDLAW AND COOPER
Oka [26] found the kink for the downstream cylinder but O SUZUKI I T AL
missed the minimum on the curve of the upstream cylinder TANIDA ET AL
at Re = 1.3 X 104, 2.6 X 104 and 4 X 104. The existence of Z IBRAVKOVICH
the kink on the curve of the downstream cylinder drag was H BOB I
e COUNIHAN
carefully verified at Salford [27] and found not only for strictly p PRICE
two dimensional flow, but also for finite cylinders with a free (FULL SYMSOLS- UPSTREAM
end (length to diameter ratio was 8), and even when a third CYLINDER)

cylinder was placed behind the second (monitored) one.


Fig. 11 Drag coefficients of cylinders
Finally, there was a third group of authors who tried to cover
a whole range of spacings within a Reynolds number range as
wide as possible. First systematic measurements of the drag
forces were reported by Wardlaw and Cooper [28] on smooth belong to the so-called first group of authors), some with single
and stranded twin conductors at 6 Reynolds numbers, from curves of different lengths, (the second group of authors), and
2 X 104 up to 9 X 104, for spacings from 1.2 up to 35 dia for some with sets of curves (the third group of authors). The
the downstream cylinder, and from 1.5 to 5.0 for the upstream present author took the liberty of introducing a break in the
one. Further extension of similar measurements for the smooth curves between the last experimental point preceding the critical
cylinders only at spacings up to 50 dia was reported by Cooper spacing and the first beyond it. The existence of the ciitical
[29] in the same range of Reynolds numbers. The present author spacing was not known at the time when the forces were meas-
resented [17] in a single plot the force measurements for spac- ured, and, unfortunately, it is now impossible to show its actual
ings from 1 to 10 dia in the Reynolds number range from 3.1 position with certainty. The exceptions are: the early measure-
X 104 to 1.2 X 105 for the downstream cylinder. The kink and ments shown in Fig. 2 where it seems clear that for Re = 6.5
the discontinuous jump were shown only in the last paper. X 104 and 1.6 X 105 the critical spacings were less than and
After this lengthy historical introduction it will be clear why greater than the 3.5 dia spacing, respectively, and Ishigai, et al.
some results in Fig. 11 are shown only with single points (they [9] who observed the critical spacing at 3.8 dia.

624 / DECEMBER 1977 Transactions of the ASME


The discontinuous jump of the drag coefficient of the up- was still 20 percent below the value of a single cylinder at the
stream cylinder is self-evident in Fig. 11 despite the fact that Reynolds number 5 X 104 [21].
only three full curves were available. The jump at the lower The curves compiled in Fig. 11 might have been affected by
Reynolds number of 3.4 X 103 seems to be reduced. The main different blockage effects in closed and open test sections and
feature for all Reynolds numbers is that the single cylinder by different measuring techniques. It is desirable to prove
drag force is reached soon beyond the critical spacing. It means the effect of the Reynolds number by tests in one wind tunnel
that thereafter the downstream cylinder had no effect on the only. Cooper's [29] experimental data are compiled in a single
upstream one. graph and reproduced in Fig. 12. The agreement with Fig. 11
The drag coefficient of the downstream cylinder showed a is obvious and the measurements at Salford [17] showed the
very strong dependence on the Reynolds number. There was same effects. Cooper wrote that the data at the lower test
considerable scatter of data in the range of spacings less than speeds were scattered due to low signal levels, and so the in-

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
critical. This scatter probably was caused by different blockage flexion shown on the top two curves is questionable.
effects (ratio of the height of the test section to the cylinder
diameter). These differed considerably in various measurements Effects of Surface Roughness. Most of the practical applicar
as shown in Table 1. tions (in fact, almost all in civil and electrical engineering),
The range of spacings above the critical showed two different produce a flow regime in which the boundary layers become
trends. At low Reynolds number the drag coefficient of the fully turbulent well before the separation points on a circular
downstream cylinder approached the value of the upstream cylinder. This flow regime will be called "postcritical."5 The
cylinder. At high subcritical Reynolds numbers, the wake fully turbulent boundary layers postpone the separation up to
turbulence from the upstream cylinder induced a supercritical about 100 deg, which results in a narrow wake. Nevertheless,
flow around the downstream one and hence the drag remained periodic vortex shedding was found in a regime, which, for the
small even at large spacing. In between these two extremes smooth cylinder, begins for Reynolds numbers greater than
there was a gradual change of the drag coefficient with a con- 6 X 106. Fage and Warsap [30] found that surface roughness
tinuous slope upwards with increasing spacing, and decreasing can promote transition to turbulence in the boundary layers
Reynolds number. It might be argued that the drag coefficient around the circular cylinder and consequently reduce the actual
of the downstream cylinder was affected in a similar way as a Reynolds number at which fully turbulent separation occurs.
single cylinder exposed to a free stream with different levels of
turbulence. The reader is referred to a paper by Fage and
Warsap [30] for full details of the latter. It is remarkable that B
Some authors called it, after Roshko, the transcritical and others super-
even at 50 dia spacing the downstream cylinder drag coefficient critical.

Table 1 Test rigs and models

Author(s) Year L/D B/D Tu% Test rig


1. Biermann, et al. 1933 48 33.5 (?) Open jet
2. Hori 1959 120 120 (?) Closed wind tunnel
3. Suzuki, et al. 1971 17.5 26 (?) Open jet
4. Cooper 1972 25 25 Low Closed wind tunnel
5. Tanida, et al. 1973 10.5 23 0% Water tank
6. Wardlaw, et al. 1973 46 69 Low Closed wind tunnel

L/D - Length to diameter ratio


B/D - Width of test section to diameter ratio
Tu% - Level of turbulence

io

oa

_L_i_J
SO 4 0 BO L/D

Fig. 12 Drag coefficient of the downstream cylinder for various


Reynolds numbers [29]

Journal of Fluids Engineering DECEMBER 1977 / 625


Systematic measurements of the change of the drag force for increased as the spacing decreased but only down to 2 7 4 dia.
vaiious stranded cables [18, 19, 20, 28] showed t h a t the end For smaller spacings, very odd changes occurred as seen in Fig.
of the subcritical regime (laminar boundary layer separation) 14. The authors observed a remarkable feature and wrote:
occurred at a Reynolds number of 3 X 104. T h e critical regime "Apparently the type of flow (refers to positive and negative
was followed by fully turbulent separation which occurred be- interference drag) changes rapidly with a change in spacing; it
yond Re = 7 X 104. This feature of stranded cables gave an may even change while t h e spacing is held constant." This was
insight into a possible mechanism of interference of two smooth the first indication of the bistable nature of two different flow
circular cylinders at much higher Reynolds numbers. patterns a t these spacings.
Wardlaw and Cooper [28] systematically measured drag forces Landweber [31] undertook a photographic study of the wake
of stranded cables in tandem. Only three sets of data are re- behind two cylinders towed side by side in water. T h e spacings
produced: A, the end of the subcritical regime, B, the middle chosen were: 1 (cylinders in contact), l1/*, lVs, 2, 2'A, 3, and 4
dia. H e observed a single vortex street for the first two spac-

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
of critical regime (minimum drag on a single cable) and C, the
postoritical regime. The data are compiled in Fig. 13. The curve ings, "confused" 4 rows of vortices for the third, and two distinct
A for both cables was in agreement with the corresponding curve vortex streets for the rest. He neglected the "confused 4 rows"
for smooth cylinders at the end of the subcritical regime. The and devoted the rest of his paper to coupled vortex streets.
kink and both jumps are at the right spacings and the shape Spivack [32] was the first to carry out a systematic meas-
of the curve beyond the critical spacing is identical to those urement of the vortex shedding frequency behind two cylinders
shown in Figs. 11 and 12. in the side-by-side arrangement. He found for all transverse
The critical regime showed a disappearance of the kink and spacings, T, greater than 2 dia a single frequency in both wakes
the subsequent jump is questionable. I t is known t h a t there is which, when reduced to Strouhal number, was the same as
no periodic vortex shedding in this regime so the jump should for t h e single cylinder. For the transverse spacing less than
not be expected. The force on the upstream cable was not 2 dia, two different frequencies were recorded in both wakes as
measured; more research is needed for this regime. shown in Fig. 15. T h e upper frequency disappeared for small
The postoritical regime brought back the kink and the jump, spacings (about 1.4 dia) and only the lower one continued down
as seen in Fig. 13. Both features are expected because vortex to the—cylinders in contact. Spivack suggested an empirical
shedding resumes in this regime. The upstream cable followed formula for the Strouhal number: S = 0.09 + | (T/D - 1/2)/
the expected trend with a marked reduction of the jump. (T/D + 1) |2. The possible explanation for the existence of
Similar data for smooth cylinders are not yet available. two different frequencies was due to two different wakes be-
hind two cylinders, but why they were found in both wakes was
left unexplained. A very low frequency was detected along the
Side by Side Arrangement gap centerline (between the cylinders only) for spacings less
Historical Survey of Early Research. Biermann and Herrnstein than 1.4.
[2] carried out measurements of the drag force when two cir- Hori [3] measured the surface pressure distribution around
cular cylinders were arranged to face the wind side by side. one of the two cylinders for various wind speeds, gaps between
They expressed the results through the interference drag coeffi- them, and angles between the wind direction and the plane of
cient which was defined in the same way as for the tandem
arrangement. Fig. 14 shows that the interference drag was
zero for all spacings greater than 5 dia. The interference drag

+ 1-63 x IO !

L/D-»« A

6 T/D
Fig. 14 Interference drag coefficient for the side by side arrange-
ment [2]

SOL/D 9 WITHIN WAKES


a ALONG CAP CENTRE LINE

_ _ A J-i « 1 0 '
. 14 « lO 4
C 1-3 « lO 5

I 2 3 T/D 4
Fig. 13 Drag coefficient of stranded cables [28] Fig. 15 Strouhal number for the side by side arrangement [32]

626 / DECEMBER 1977 Transactions of the ASME


the cylinders' axes. Fig. 16 shows the base pressure coefficient
at Re = 8 X 103 for three spacings 1.2, 2, and 3 dia. () = 0
deg gives the base pressure of the upstream cylinder in the tan-
dem arrangement, () = 90 deg on one cylinder in the side-by-side
arrangement, and () = 180 deg on the downstream one in the
tandem arrangement. There is a discontinuous change of the
base pressure coefficient for the side-by-side arrangement when
the spacings were 1.2 and 2.0 dia. Hori carefully verified this
by adding additional experimental points in front of and behind
the location of the jump, and stated: "When () = 90 deg and
the gap is less than 2 dia, there is a difference in Cpb of the two

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
cylinders, although the geometrical arrangement is symmetrical
with respect to the stream." It was a third indication that the
wakes behind the two cylinders were different.
Bistable Nature of Biased Flow Pattern. Ishigai, et al. [9] used
Fig.16 Base pressure coefficient for various arrangements [3]
the Schlieren optical method and photographed the flow pattern
around the two cylinders, Fig. 17 shows a selection of frames
from a cine film for five different side-by-side arrangements. For
the transverse spacing ratios 3 and 2.5, the vortex shedding is
"coupled," as seen in Figs. 17(a) and (b). The vortex formation
and shedding is markedly symmetric about the axis of the gap.
At TID = 2.0, Fig. 17(c), the vortex shedding is still coupled,
but the gap flow is slightly deflected upwards. The vortex
fhedding becomes uncoupled and weak at TID = 1.5 in Fig.
17(d), while the gap flow is biased to one side. Consequently,
wide and narrow wakes are formed behind the cylinders. The
same trend continues at TID = 1.25 as seen in Figs. 17(e)
and (f). The biased flow in the gap is bistable and intermitt-
ently changes over either downwards, Fig. 17(e) or upwards,
17(f).
Ishigai, et al. [9] suggested that the biased flow in the gap
was due to the Coanda effect, i.e., flow attaching itself more ( ) ..
TJIJ 3
(b)
lIfJ-2.$
«)
TW-2
to one cylinder smface than the other due to unsymmetric R. :CXi) R~ • .c-SO /0',-3 )
separation. This was incorrect because the biased flow was Fig.17 Flow pattern for side-by-side arrangements [9]
observed [33] behind two flat plates placed side-by-side (fixed
separation points).
Bealman and Wadcock [33] were the first who measured the
Cpb
base pressure simultaneously on both cylinders, and found that
-I"
the cylinders always experienced different base pressures in the
-1·4
range of spacings between 1.1 and 2.3 dia, Fig. 18 (Hori's [3]
data are also included). They stated, "The base pressure AUTHOR II.
changed from one steady value to another, or simply fluctuated o HAllMAN 2'5. 104
between the two extremes. Stopping and starting the tunnel X KORI • • 10'
again could cause the pressure to change over." The bistable
nature of the biased flow pattern was confirmed. Some physical
insight into the origin of the biased flow will be given in the
section dealing with the staggered arrangements.
-.
-..

-'2
0

0'-1--'----!2~--'---!:---
.......-~4 T/O
Lift and Drag Forces on Two Cylinders. The bistable nature of
Fig. 18 Base pressure coefficient for the side-by-side arrangement
the flow around the two cylinders means that in the range of [33] and [3]
spacings from 1.1 to 2.2 dia there are two bistable forces ex-
perienced by the cylinders. Two different forces act on the
two cylinders only for some time and change over of these
forces should always be taken into account.
There is an additional feature of the biased flow. The gap
flow biased to one side will produce a resultant force on the

---+--+--
cylinder which is deflected relative to the free stream direction.
Hence, there will be a component of the force acting perpen-
dicular to the free stream direction which may be called a lift
force. Repulsive lift forces between cylinders will be taken as CD C L Rc IIEI':
1••• 101
positive.
All available lift and drag coefficients were compiled in a
, + (2)
o • • • 10' (.)
Cl It 0 8 • • ,04 (17)

~\,
graph shown in Fig. 19. All the experimental points for the
6 6 2'5.104 (n)
drag coefficients measured by Biermann and Herrnstein [2],
Fig. 14, at Re = 1.63 X 105, were retained, but two curves are
drawn through them in Fig. 19. The measurements at Salford '~\.~-i==~_ _...J..._ _..J...._...J...=
0L._---l:;=:~::::L

[17] at Re = 6 X 10 4 were also included. They lay within I 2 I 4 5 • T/O


Bielmann's and Herrnstein's curves. Hori's [3] calculated drag Fig. 19 Lift and drag coefficients for the side-by-side arrangement
coefficients for three spacings were very close to the Salford [17]

Journal of Fluids Engineering DECEMBER 1977 / 627


measurements. Bearman and Wadcock [33] stated values only bistable range of the side by side arrangement yields two vec-
for the higher drag coefficient and these values were included tors acting on each of the two cylinders. One of the cylinders
in Fig. 19. The overall trend of all curves is the same despite experiences the force corresponding to upstream stagger while
the differences in the Reynolds number. The remarkable feature the other encounters, at the same time, the force corresponding
of the interference between the two cylinders is that the sum to downstream stagger. The vectors on the two cylinders are
of the bistable high and low drag is always less than twice the almost identical in the lift direction but markedly different in
drag of the single cylinder. the drag direction. Hence, the bistable side-by-side arrange-
Lift coefficients were not measured by Biermann and Herrn- ments represent a transition from the upstream to the down-
stein. Hori calculated lift for three spacings, and the values stream stagger.
are shown in Fig. 19. The two lift coefficients are very close A direction of the interference force coefficient from right to
to one another. left indicates that the drag at that position is less than for the
Bearman and Wadcock [33] stated only the lift coefficient single cylinder, while the opposite direction means that the drag

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
for the high drag regime as shown. The lift coefficients measured is greater than for the single cylinder. The upward direction
at Salford [17] were well below Bearman's and Hori's values. in the upper half and downward direction in the lower half of
The difference between the lift coefficients of the two flow re- the figure indicate a repulsive, positive lift force while the
gimes was greater than found by Hori, particularly at the spac- opposite directions in the corresponding halves indicate a neg-
ing of 1.25 dia. The cause for these differences is at present ative lift force.
unknown and more research for very small gaps is necessary. All possible arrangements of two cylinders were classified into
Very little is known at present about the existence of the biased regions by taking into account whether the drag force is greater
flow and its bistable nature in the post-critical flow regime. or less than for the single cylinder and whether the lift force
The research at Salford, currently in progress, revealed that is positive, negative or negligible. These regions are shown in
both do exist even when the cylinders were finite (H/D = 5). Fig. 20. The upstream cylinder can be situated in three regions.
(1) Negligible lift force and reduced drag force.
Staggered Arrangement (2) Small repulsive lift force and reduced drag force.
(3) Repulsive lift force and increased drag force.
Classification of Staggered Arrangements. The staggered ar-
rangement is likely to occur most often in engineering appli- The downstream cylinder can be situated, in addition to the
cations, but has attracted the least research interest. The first above three, in the following regions:
systematic measurements of surface pressure distribution around
(4) Negligible lift force and increased drag force. This is a
one of the two parallel cylinders in various staggered arrange-
small region and beyond it there is no interference.
ments were carried out by Hori [3]. He rotated the cylinder
(5) Negative lift force and decreased drag force. This is a
pairs and exposed them in staggered arrangements to the wind
for three spacings only (L/D = 1.2, 2.0 and 3.0). The measured dominant region for the downstream cylinder.
base pressure was shown in Fig. 16. It is interesting to note that the upstream and downstream
Hori calculated drag and lift coefficients from the surface cylinder in any staggered arrangement usually belong to differ-
pressure measurements and presented them in a table. From ent regions. The cylinders in tandem arrangements for spacings
these values, the resultant interference force coefficients were less than the ciitical, and side-by-side arrangements beyond the
plotted in Fig. 20. The resultant interference force coefficient bistable range, are both in the same region. When the cylinders
was obtained by vectorially subtracting the drag coefficient for are in the side-by-side arrangement within the bistable range,
the single cylinder from the given force coefficients. The inter- then one belongs to region 3 and the other to region 2, or vice
ference effects are proportional to the vectors shown. The versa.

CB-W T/B I C t ° 0
/ §-^ © . - . • " C o * C„

v
- ^ © / ® 4.'\

© IO 3

FLOW
W I K
-AU*- 3 L/O
i

/
(?) r ^ „A X

- /f\A®K © \

/ \
© /
©
c8-cBO\
Fig. 20 Interference force coefficient for all arrangements [3]

628 / DECEMBER 1977


Transactions of the ASME
The wide variety of changes of the interference force reflect not coincide with the angle of stagger. This fact prompted the
the very complicated nature of the corresponding flow patterns. present author to abandon the angle of stagger in all subsequent
Suzuki, et al. [22] measured the distribution of time average sections and to describe all the phenomena in terms of L/D
pressure around the downstream cylinder in region 5, and their and T/D instead.
results are shown in Fig. 21. Some insight into the flow patterns
can be inferred from Fig. 21, bearing in mind that the Eeynolds Forces on the Downstream Cylinder. Systematic measurements
number is higher than in Fig. 20. The most prominent feature of the lift and drag force at Salford [17] covered tandem arrange-
is that the axis of "symmetry" of the pressure distribution does ments from 1 to 5 dia spacing, side-by-side arrangements from
1 to 3 dia, and a series of staggered arrangements within these
two limits. The results are presented in the L/D - T/D plane
and then the curves of constant coefficients are drawn by in-
fegS-003 fr«0 terpolation between the measured values of lift and drag coef-

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
ficients, as shown in Figs. 22 and 23 at Re = 6 X 104.
The plot of constant lift coefficient curves in Fig. 22 shows a
positive (repulsive) force in the regions 2 and 3. The remainder
is region 5 with a negative lift force directed towards the axis
of the upstream cylinder wake. It is surprising that there are
two separate curves representing the maximum lift coefficient
(chain dot lines). One lies well inside the upstream cylinder
wake for L/D up to 3, and the other is near to the wake bound-
ary for spacings greater than 2.7 dia. The overlap is an indica-
tion that there should be two entirely different origins of the
outer and inner maximum negative lift force as will be shown in
the next section.
The corresponding drag coefficient plot for the same Reynolds
Fig. 21 Pressure distributions around the downstream cylinder in number is shown in Fig. 23. The most striking feature is that
staggered arrangement [22] for spacings less than 3 dia the minimum drag coefficient of

eL roa R0 ™a » lo*

Fig. 22 Lift force coefficient for a downstream cylinder at Re = 6.X


X10< [17]

Fig. 23 Drag force coefficient for a downstream cylinder at Re = 6.1


X 10* [17]

Journal of Fluids Engineering DECEMBER 1977 / 629


the downstream cylinder occurs in the staggered arrange- only around the upper part of the downstream cylinder.
ment, and not in the tandem arrangement, as might be ex- It should be pointed out that "wake displacement" hypothesis
pected. This region of the minimum drag coefficient almost is a novel one. All the previous authors assumed that the down-
completely coincides with the inner maximum negative lift co- stream cylinder is submerged in the wake of the upstream one.
efficient (shown as a chain dot line). This coincidence will be Maekawa [19] was the first who put forward that assumption
used to explain the origin of the inner lift force in the subse- after measuring the gradient of the static pressure across the
quent section. single cylinder wake. He hypothesized that the same wake
The second feature seen in Fig. 23 is that the drag coefficient exists behind the upstream cylinder and induces a "buoyancy"
increases regularly and gradually with the transvet"se spacing. force on the downstream cylinder directed towards the center
This feature is not affected in the region where the outer lift line of the upstream cylinder wake. Taylor [36], however,
coefficient maximum occurs. found that when the circular cylinder was subjected to a shear
The two values of the drag coefficient for the bistable side-by- flow similar to that behind the upstream cylinder, the lift

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
side spacing are shown on two sides of the TID axis, Fig. 23. force should be in the positive direction. This was verified by
The "spread" of the bistable values int() the nearby staggered Snyder [37]. Cooper [29] showed by pressure integration that
arrangements is schematically shown with two parallel lines. less than half of the observed lift can be ascribed to the pressure
This should be taken only as a qualitative indication because gradient across the wake. Despite that qualitative and quan-
the actual spread of the bistable values is not known at present. titative contradiction, all subsequent researchers followed the
For example, Wardlaw and Cooper [28] observed that the inter- same path.
mittency of the drag force at TID = 1.5 persisted within the Price [38] found that the wake characteristics of the upstream
range of stagger from -0.16 ~ LID ~ 0.16. Roberts [34] cylinder had a negligible effect on the lift of the downstream
called this intermittency "the jet switch" and found severe cylinder. He also found that the lift force increased when the
vibrations when every alternate cylinder in a cascade was upstream cylinder was smaller than the downstream one, and
flexible. it substantially decreased when the downstream cylinder was
smaller than the upstream one. If the lift force was due to the
The Origin of the Outer Lift Force. A plausible explanation of
upstream cylinder wake displacement, then it should not de-
the origin of a negative lift force has been offered by Mail' and
pend on the wake characteristics. If the upstream cylinder is
Maull [35]. They proposed that the negative lift is due to the
smaller than the downstream one, the latter will be able to
entrainment of the flow into the wake boundary of the up- "displace" the "smaller" wake to a greater extent which re-
stream cylinder. This explanation is applicable as long as the
sults in the greater lift. Contrary to that, when the smaller
downstream cylinder is not in the wake of the upstream one.
downstream cylinder is in the wake of a bigger upstream cyl-
Sketch A in Fig. 24 shows that the flow rate around the down-
inder, the former is less effective in displacing a "bigger wake"
stream cylinder is unaffected on the outer side, but is increased
and the result is a decrease of the lift force. When the down-
due to the entrainment between the inner side and the fully
stream cylinder is much smaller than the upstream one, it might
formed wake of the upstream cylinder. The measurements of
be unable to displace the upstream cylinder wake and it will
the mean pressure distribution around the downstream cylinder
be submerged into the shear flow. Price found, in that case, a
by Suzuki, et aI. [22] and Zdravkovich [11] showed that a shift
positive lift force as should be expected from Taylor's theory
in the stagnation point was always in the direction away from
[36] and Snyder's tests [37]. It seems to the present author
the side of the upstream wake. The pressure distribution was
almost symmetric about the axis passing through the shifted
stagnation point and the center of the cylinder as shown in
Fig. 21 (for large spacings only). Hence, the resulting pressure
force was inclined to the freestream velocity and had a com-
ponent in the negative lift direction.
The negative lift force increases as the downstream cylinder
approaches the upstream cylinder wake. The maximum is
reached when the downstream cylinder "displaces" the upstream
cylinder wake and in doing so "squeezes" the streamlines be-
tween its inner side and the displaced wake, sketch B in Fig. 24.
Flow visualisation pictures by Ishigai, et aI. [9] and Cooper
[39] clearly show that. The latter are reproduced in Fig. 25.
It is evident that, even at TID = 1, all the wake flow passes

=:::-=========------------
~~:~.
@~ u V
l -=--
.~4'6
i C..·.02.
II
~~":1

Fig.24 Sketch of flow patterns in staggered arrangement Fig. 25 Flow visualisation of staggered arrangements [39)

630 IDE CE M B E R 1977 Transactions of the ASME


that this decisive evidence in favor of the wake displacement rected toward the gap between the cylinders. T h e result of such
hypothesis proved it to be the correct one. a pressure distribution was a very high lift force.
There are, in the literature, misleading hypotheses concerning T h e second pressure distribution shows t h e disappearance of
a wake boundary and an interference boundary. T h e wake the low pressure between 270 and 360 deg, and appearance of a
boundary is a line along which the velocity becomes the same new one at 90 deg. T h e distribution resembles those shown in
as the free stream one. T h e interference boundary is the line Fig. 3 for the tandem arrangement at spacings less than critical.
along which the drag force is identical to the single cylinder T h e main difference is that the present one is markedly asym-
drag force and the lift force becomes zero. Cooper and Wardlaw metric. Hence, it m a y be inferred t h a t the flow pattern resem-
[21] proposed t h a t both boundaries coincide for all longitudinal bles those in the tandem arrangement with no-flow and without
spacings. This is not so, as seen in Fig. 25, and it is not even vortex shedding in the gap. T h e result is a very small lift.
necessary for a theoretical explanation once the shear flow con- For spacings less than 3 dia, it was shown in Fig. 23 t h a t the
cept is abandoned. The wake displacement concept coupled

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
minimum drag coefficient occurred for the same transverse
with Mair and Maull's wake entrainment concept indicate t h a t spacing for which a maximum lift coefficient was found. T h e
the interference boundary must always be outside the wake pressure distribution for high lift revealed t h a t it was due to the
boundary. very low pressure around 360 deg in comparison to that at 180
Finally, some of the literature does not employ the proper deg. Hence, the minimum negative drag coefficient in t h e
definition of the interference boundary. Many authors observed staggered arrangement is a by-product of the gap flow while
that the zero lift coefficient was reached at a greater transverse the negative drag in t h e tandem arrangement is due to "no-flow"
spacing than the spacing at which the drag became equal to in the gap.
the single cylinder value. This is in agreement with the wake The most striking feature of the two pressure distributions
entrainment concept. shown in Fig. 26 was that they changed over intermittently at
The slight asymmetry of the flow around the downstream irregular time intervals. This indicated a bistable nature of t h e
cylinder induced by the entrainment of the flow into the up- fully developed maximum flow through the gap between the
stream wake causes the lift force, while the drag force remains cylinder and no-flow pattern similar to t h a t found in the tandem
unchanged. The first noticeable decrease of the drag force can arrangements at these spacings. T h e alternating changes of
be expected only after a sufficient lift force is produced, i.e., pressure distribution resulted in a discontinuous change of the
when the resulting force is sufficiently tilted. Hence, the correct lift coefficient whilst the drag coefficient was less affected and
and precise definition of the interference boundary is that it is showed little difference in the measured values. T h e same
a line along which the lift force becomes zero or negligible. behavior was found not only at the Reynolds number of 6 X 104
but also at 8.5 X 104 and 1.1 X 106. Wardlaw and Cooper
The Origin of the Inner Lift Force. The maximum negative lift [28] experimental d a t a showed a similar discontinuity in the
force for all longitudinal spacings L/D less than 3 dia cor- lift coefficient at R e = 2.5 X 104. I t is not known a t present
responds to a position of the downstream cylinder well inside whether the gap flow exists in the whole range of subcritical
the wake of the upstream one, as seen in Fig. 22. T h e mag- Reynolds numbers. Severe vibrations may be excited by the
nitude of the lift coefficient is higher than those on the outer bistable gap flow [11].
Ci m ax curve. It is evident that the side of the downstream
cylinder crossed the axis of the wake of the upstream one. Reynolds Number Effects. Our present knowledge of the post-
Immediately after the Ci m „x was reached, a step-like fall brings critical flow regime around a single cylinder indicates some
the lift coefficient down to a very low value. The position of features which should be expected in the case of two cylinders
the downstream cylinder suggests t h a t an intense gap flow in a staggered arrangement. The first well known feature is
between the cylinders might be responsible for such a high t h a t the wake behind a single cylinder narrows compared with
Ctmax. Further reduction of the transverse spacing might the subcritical regime. This will cause a similar contraction of
"block" t h e gap flow with the corresponding sharp fall of t h e the interference boundary in the case of two staggered cylinders.
lift coefficient. This trend has been found by various authors [19, 20, 21, and
The time-average pressure distribution round the downstream 28]. I t should be pointed out t h a t all these authors observed
cylinder gave full support to the above description. Two differ- this contraction behind stranded cables and not behind smooth
ent pressure distributions were found [17] as shown in Fig. 26. cylinders at high Reynolds numbers. The so-called "wake-
One of the distributions has a very low pressure between 270 galloping" instability boundary moved closer to the upstream
and 360 deg which is due to a high velocity around t h a t side of cable and its wake axis [21]. The outer CLI™* curve was found
the downstream cylinder. The stagnation point, seen as the posi- [28] to be almost half-way between the Chm**. of the subcritical
tive pressure peak in Fig. 26, was at 45 deg, and the negative regime and the axis of the upstream cable. I t seems, however,
pressure did not appear at 90 deg. These facts indicate t h a t the t h a t the inner Ct m a x remained at the same transverse spacings
bulk of the flow approaching the downstream cylinder was di- and extended over the same range of longitudinal spacings,
L/D. More research is necessary to confirm this finding for
smooth cylinders and to investigate the effect of surface rough-
ness on the gap flow.
Some preliminary research at Salford showed t h a t a discon-
tinuous suppression of the gap flow also exists in the fully tur-
bulent regime.
The most complicated range of spacings is that where the
outer Ci m l i X and the inner one overlap. Some qualitative in-
sight may be obtained from the two Figs. 27 and 28. T h e experi-
mental data given by Wardlaw and Cooper [28] were compiled
into two curves. T h e lift coefficient measured on t h e down-
stream cylinder (both were smooth) is shown versus the trans-
verse spacing. Only the data for the lowest, Re = 2.42 X 10 4
and highest Re = 8.46 X 104 are shown in Fig. 27. Two peaks
are evident; one indicating the position of the inner Ctmax and
the second indicating the position of the outer Cimax. Both
Fig. 26 Pressure distribution around the downstream cylinder in
staggered arrangement [17] curves are similar and they differ only at very small trans-

Journal of Fluids Engineering DECEMBER 1977 / 631


L/0 I n
O e ©
o-s -®- •9-
1 s 9
2 0 9
4
2-4 K I d " • t
SPIVACK'S RESULTS (32)

—2 -

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
Fig. 27 Lift coefficient of the downstream cylinder in staggered ar-
rangements at L/D = 3 [28]

L/O *3
T/D
Fig. 29 Strouhal number for cylinders in staggered arrangement [9]

(FULLY TURBULENT)
cylinder. Contrarily, t h e Strouhal numbers for both cylinders
should be different at small spacings.
T h e only available data were measured in t h e intermediate
subcritical range by Ishigai, et al. [9]. Their results are re-
produced in Fig. 29. I t is evident t h a t the upstream cylinder
with a narrow wake has a higher Strouhal number. T h e peak
is reached at L/D = 1. The downstream cylinder is less affected
and the Strouhal number is always less than for the single
cylinder.
Fig. 28 Lift coefficient of a downstream stranded cable in staggered
arrangement [28]
Closing Comments
The complexity of the flow interference between two circular
verse spacings. At the higher Reynolds number the lift coeffi- cylinders may be due to differing flow patterns. T h e flow pat-
cient is positive for very small transverse spacings and this terns provide phenomenological explanation, b u t t h e intrinsic
was also found by [5] and [22]. The discontinuous jump due to nature of the flow patterns remains a mystery. Hence, future
the gap flow is greater for the higher Reynolds number than research should be directed in t h a t direction.
for the lower one. The existence of bistable flow patterns m a y excite flow-in-
The lift coefficients measured on two stranded cables at the duced vibrations of large amplitude. These vibrations m a y
same Reynolds number are shown in Fig. 28. The lower Reyn- occur at any natural frequency of the elastic cylinders. They
olds number corresponds to a subcritical flow regime and the will be the subject of another paper.
higher one to the fully turbulent flow regime. There is a marked
difference between these two curves. Neither the position of Acknowledgment
the two peaks nor their magnitudes are the same.
The low Reynolds number curve as seen in Fig. 28, which I would like to express m y gratitude to the Referees whose
corresponds to the subcritical regime, is not similar to those comments greatly improved the presentation of the paper.
curves shown in Fig. 27. The position of the outer peak is the
same but the magnitude is much higher. The inner peak is References
not seen and it may be argued t h a t the surface roughness affected
the gap flow in an unknown manner. 1 Pannell, J. R., Griffiths, E. A., and Coales, J. D., "Experi-
ments on the Interference between Pairs of Aeroplane Wires of
The higher Reynolds number curve which corresponds to the Circular and Lenticular Cross Section," (British) Advisory Com-
fully turbulent regime, see Fig. 28, is similar to the higher mittee for Aeronautics, Reports and Memoranda No. 208, 1915,
Reynolds number curve in Fig. 27. Both peaks are distinct, Annual Reports for 1915-1916, Vol. 7, pp. 219-221.
the outer one markedly shifted to smaller transverse spacings 2 Biermann, D., and Herrnstein, W. H., Jr., "The Inter-
ference between Struts in Various Combinations," National Ad-
as expected. The magnitude of the lift coefficient is much visory Committee for Aeronautics, Tech. Rep. 468, 1933.
lower than in the subcritical regime except at small transverse 3 Hori, E., "Experiments on Flow around a Pair of Parallel
spacings. The latter is an opposite trend to that seen in Fig. 27. Circular Cylinders," Proc. 9th Japan National Congress for Ap-
Finally, it may be suggested that future research should take plied, Mech., Tokyo, 1959, pp. 231-234.
4 Zdravkovich, M . M., and Stanhope, D. J., "Flow P a t t e r n
into account separately all four parameters: Re, L/D, T/D in the Gap Between Two Cylinders in Tandem," University of
and the surface roughness with an emphasis on L/D less than Salford Internal Report F M 5/72, 1972.
three. 5 Zdravkovich, M. M., "Smoke Observations of Wakes of
Tandem Cylinders at Low Reynolds Number," The Aeronautical
Vortex Shedding. T h e most neglected feature of the flow Journal, Vol. 76, Feb. 1972, pp. 108-114.
around two staggered cylinders was, and still is, the measure- 6 Thomas, D. G., and Kraus, K. A., "Interaction of Vortex
Streets," Journal of Applied Physics, Vol. 35, No. 12, Dec. 1964,
ments of vortex shedding. I t is expected that at large spacing, the pp. 3458-3459.
Strouhal number should be equal to t h a t found for a single 7 Oka, S., Kostic, Z. G., and Sikmanovic, S., "Investigation

632 / DECEMBER 1977 Transactions of the ASME


of the Heat Transfer Processes in Tube Banks in Cross Flow," ders in Open Channel Flow," Transactions, Japan Society Civil
International Seminar on Recent Developments in Heat Ex- Engineers, Vol. 3, pp. 57-64 (in Japanese).
changers, Trogir, 1972, Yugoslavia (preprint). 25 Hoerner, S. F., Fluid Dynamics of Drag (published by the
8 Novak, J., "Strouhal Number of a Square Prism, Angle author), 1958.
Iron and Two Circular Cylinders Arranged in Tandem," (in 26 Kostic, Z. G., and Oka, S. N., "Fluid Flow and H e a t
English) Acta Technica, Czechoslovak Akaolemy of Sciences, No. 3, Transfer with Two Cylinders in Cross Flow," International
1974, p p . 361-373. Journal of Heat and Mass Transfer, Vol. 15, 1972, pp. 279-299
9 Ishigai, S., Nishikawa, E., Nishimura, K., and Cho, K., 27 Zdravkovich, M. M., and Pridden, D. J., "Flow Around
"Experimental Study on Structure of Gas Flow in Tube Banks Two Circular Cylinders; Research Report," Proceedings 2nd
with Tube Axes Normal to Flow (Part 1, Karman Vortex Flow U. S. National Conference on Wind Enqineerinq Research, Fort
around two Tubes at Various Spacings)," Bulletin of the Japan Collins, 1975, Paper IV-18.
Society Mechanical Engineers, Vol. 15, No. 86, 1972, pp. 949-956. 28 Wardlaw, R. L., and Cooper, K. R., "A Wind Tunnel In-
10 Novak, J., "Strouhal Number for Two Cylinders of Dif- vestigation of the Steady Aerodynamic Forces on Smooth and
ferent Diameters Arranged in Tandem," Acta Technica, Czecho- Stranded Twin Bundled Power Conductors for the Aluminium

Downloaded from http://asmedigitalcollection.asme.org/fluidsengineering/article-pdf/99/4/618/5899556/618_1.pdf by The University Of Western Australia user on 09 August 2021
slovak Academy of Sciences, No. 3,-1975, p p . 366-374. Company of America," National Aeronautical Establishment,
11 Zdravkovich, M. M., "Flow Induced Vibrations of Two Canada, 1973, LTR-LA-117.
Cylinders in Tandem, and their Suppression," Proceedings Inter- 29 Cooper, K. R., "Wind Tunnel Measurements of the Steady
national Symposium, Flow Induced Structural Vibrations, Karls- Aerodynamic Forces on a Smooth Circular Cylinder Immersed
ruhe 1972; Springer Verlag 1974, pp. 631-639. in the Wake of an Identical Cylinder," National Aerodynamic
12 King, R., and Johns, D. J., "Wake Interaction Experi- Establishment, Canada, 1974, LTR-LA-119.
ments with Two Flexible Circular Cylinders in Flowing Water," 30 Fage, A., and Warsap, J. H., " T h e Effects of Turbulence
Journal of Sound and Vibration, Vol. 45, No. 24, Mar. 1976, p p . and Surface Roughness on the Drag of a Circular Cylinder,"
259-283. Aeronautical Research Council, Reports and Memoranda No.
13 Dickey, W. L , and Woodruff, G. B., "The Vibration of 1283, 1929.
Steel Stacks," Proceedings ASCE, Vol. 80, Paper No. 540, 1954. 31 Landweber, L., "Flow About a Pair of Adjacent, Parallel
14 Ozker, M. S., and Smith, J. O., "Factors Influencing the Cylinders Normal to a Stream," Navy Dept. D . W. Taylor Model
Dynamic Behaviour of Tall Stacks under the Action of the Basin, Report 485, 1942, Washington, D.C.
Wind," Journal of Power, TKANS. ASME, Vol. 78, 1956, pp. 32 Spivak, H. M., "Vortex Frequency and Flow Pattern in
1381-1391. the Wake of Two Parallel Cylinders at Varied Spacings Normal
15 Roshko, A., "On the Wake and Drag of Bluff Bodies," to an Air Stream," Journal Aeronatuical Sciences, Vol. 13, 1946,
Journal Aeronautical Sciences, Vol. 22, 1955, pp. 124-132. pp. 289-297.
16 Abernathy, F . H., and Kronauer, R. E., " T h e Formation 33 Bearman, P. W., and Wadcock, A. J., " T h e Interaction
of Vortex Streets," Journal of Fluid Mechanics, Vol. 13, 1962, Between a Pair of Circular Cylinders Normal to a Stream,"
pp. 1-20. Journal Fluid Mechanics, Vol. 61, pp. 499-511.
17 Zdravkovich, M. M., "Interference Between Two Circular 34 Roberts, B. W., "Low Frequency Aeroelastic Vibrations in
Cylinders; Series of Unexpected Discontinuities," Journal of a Cascade of Circular Cylinders," Mechanical Engineering
Industrial Aerodynamics, Vol. 2, 1977, pp. 255-270. Science Monograph, No. 4, 1966.
18 Counihan, J., "Lift and Drag Measurements on Stranded 35 Mair, W. A., and Maull, D . J., "Aerodynamic Behaviour
Cables," Imperial College, London, Aero Report No. 117, 1963. of Bodies in the Wake of Other Bodies," Transactions Royal
19 Maekawa, T., "Study on Wind Pressure against ACSR Society, A, Vol. 269, 1971, pp. 425-437.
Double Conductor," Electrical Engineering in Japan, Vol. 84, 36 Taylor, G. I., "Motion of Solids in Fluids when the Flow
1964, pp. 21-28. is. Irrotational," Proceedings Royal Society, A, Vol. 93, 1917, p p .
20 Price, S. J., "Wake Induced Flutter of Power Transmission 99-113.
Conductors," Journal of Sound and Vibration, Vol. 38, No. 1, 37 Snyder, M. A., "Testing of Cylinders in Shear Flow,"
1975, pp. 125-147. J. Aircraft, Vol. 8, 1971, pp. 593-596.
21 Cooper, K. R., and Wardlaw, R. L., "Aeroelastic Instabili- 38 Price, S. J., "The Origin and Nature of the Lift Force on
ties in Wakes," Proceedings International Symposium on Wind Two Bluff Bodies," Aeronatuical Quarterly, Vol. X X V I , 1976, p p .
Effects on Buildings and Structures, Tokyo, 1971, Paper IV.1. 154-168.
22 Suzuki, N., Sato, H., Iuchi, M., and Yammamoto, Sh., 39 Cooper, K. R., "Wind Tunnel and Theoretical Investiga-
"Aerodynamic Forces Acting on Circular Cylinders Arranged in tions into the Aerodynamic Stability of Smooth and Stranded
Longitudinal Row," Proceedings International Symposium, Wind Twin Bundled Power Conductors," National Aeronautical Es-
Effects on Buildings and Structures, Tokyo, 1971, pp. 377-386. tablishment, Canada, 1973. LTR-LA-115.
23 Tanida, A., Okajima, A., and Watanabe, Y., "Stability 40 Quadflieg, H., "Vortex Induced Load on the Cylinders
of Circular Cylinder Oscillating in Uniform Flow or in a Wake," Pair at High Reynolds Numbers," (in German), Forsch. Ing.-
Journal of Fluid Mechanics, Vol. 61, 1973, pp. 769-784. Wes., Vol. 43, 1977, pp. 9-18. This paper was published after
24 Nagai, S., and Kurata, K., "Interference between Cylin- the present review was written.

Journal of Fluids Engineering DECEMBER 1977 / 633

You might also like