Journal of Fluids and Structures (1987) 1, 239-261

THE EFFECTS OF INTERFERENCE BETWEEN

CIRCULAR CYLINDERS IN CROSS FLOWt

M. M. ZDRAVKOVICH
Department of Aeronautical and Mechanical Engineering,
University of Salford, Salford M5 4WT, U.K.

(Received 29 June 1986 and in revised form 20 September 1986)

One is dismayed if not surprised by the disorderly state of a topic of such practical
importance. An attempt is first made to categorize various cluster arrangements in terms of
the number of pipes (cylinders), their mutual spacing and orientation to the oncoming
current. The second step classifies the basic interference flow regimes in the subcritical,
critical and postcritical states, for two pipes only, in all possible arrangements. The third step
goes further and considers various clusters and distinguishes between the basic interference
flow regimes, however modified they may be, and the additional ones for particular
orientations. This approach allows some qualitative interpretation and even the prediction of
the forces exerted on any of the pipes in the cluster, for various orientations. The following
clusters are discussed in some detail: aligned, triangular, square, staggered, all with pipes of
the same size, and clusters with pipes of different diameters, i.e. small pipes arranged
circumferentially around a big central one. This field is an example of ad hoc testing, leading
to a proliferation of undigested and uncorrelated data.

"There is no part of hydrodynamics more perplexing to the student than that which treats
the resistance of fluids" Lord Rayleigh (1876)

1. I N T R O D U C T I O N
WHEN MORE THAN ONE bluff body is placed in a fluid flow, the resulting forces and vortex
shedding pattern may be completely different from those found on a single body at the
same Reynolds number. For example, the drag force exerted on a downstream pipe
(cylinder) arranged in tandem closely behind an upstream pipe may be negative, i.e.
acting as a "thrust" force against the free stream. This appears "more perplexing" than
D'Alembert's paradox because it stems from experiments in a real fluid.
Another perplexing paradox occurs at high Reynolds numbers, when the boundary
layers along the pipe become turbulent before separation. The drag force on the
"shielded" downstream pipe in the tandem arrangement is greater than that on the
unshielded upstream pipe.
The paradoxes are not restricted to tandem arrangements. When two identical pipes
are arranged a small distance apart and side by side to the oncoming flow, the forces
exerted on them, and the nearwakes formed behind them, are always different. An
intrinsically asymmetric flow is produced by a geometrically symmetric side-by-side
arrangement as shown in Figure l[a, (ii)].
Yet another example is a staggered arrangement of two pipes. The force exerted on the
downstream pipe has a component in the direction normal to the freestream, i.e. a mean
t An earlier version was originally presented as an invited paper, entitled "Forces on pipe clusters", at the
International Symposium on Separated Flow around Marine Structures, Norwegian Institute of Technology,
Trondheim, Norway, 26-28 June 1985.

0889-9746/87/020239 + 23 $03.00/0 © 1987 AcademicPress Inc. (London) Limited

240 M . M . ZDRAVKOVICH

lift force. The paradox is that the low pressure in the wake of the upstream pipe cannot
account for more than 1/4 of the measured lift force.
All these paradoxes are caused by different kinds of interference produced by two pipes.
The corresponding flows around two pipes in various arrangements will be called the
basic interference flow regimes and discussed in some detail in the next section.

2. C A T E G O R I Z A T I O N O F C L U S T E R A R R A N G E M E N T S
The flow around pipes in a cluster is likely to be more complicated than the flow
around two pipes. The interference phenomena are highly non-linear and at present
beyond a reliable theoretical or computer analysis. However, for any particular
orientation of the pipe cluster relative to the free stream, one of the basic two-pipe
interference flow regimes may be applicable to flow around some of the pipes. These
particular orientations may serve as useful guides in interpreting the measured forces on
the pipes and in predicting the forces on them, albeit in a qualitative way.

a=O a-90 O< a < 90

( I ) Tandem (11) Side by side (111) Staggered

(a)

- O~<a ~<90 ° 0 ~<a ~<60 ° 0~a~<180 °

( t ) Ahgned (H) Regular (In) Irregular
trtangular trlangu[ar
(b)

~(~f~_O 0 0 0 oooo
~>,-0 0 0 ~7 °0 0°0 0° °
0 0 00.0 oooo
(i)2x2 (n) 5 x 5 (m)4x4
(c)

o o .e'
O<a < 6 0 °
( i ) Triangular (11) Irregular (ill) Regular
stagger stagger ctrcular
(d)

Figure 1. Classification of pipe clusters: (a) for two pipes; (b) for three-pipe clusters; (c) for regular square multi-
pipe clusters; (d) for irregular and triangular multi-pipe clusters.
 CYLINDERS IN CROSSFLOW 241
All kinds of interference depend strongly on the orientation, spacing and number of
pipes in the cluster. There are, of course, the usual influencing parameters such as the
Reynolds number, surface roughness, free stream turbulence and all others known to
affect the flow around a single pipe. Hence, one possible way to categorize the pipe
clusters is according to the number of pipes and their mutual arrangement.
Figure 1 shows such an attempt. It is assumed that the smallest number of pipes in the
cluster is three. Two different arrangements are often found in engineering applications:
aligned and triangular. The latter is usually a regular arrangement forming an equilateral
triangle and rarely an irregular arrangement, as shown in Figure l(b).
Additional pipes can be arranged in equidistant square clusters such as 2 x 2, 3 x 3,
4 x 4, etc., as shown in Figure l(c). The square clusters appear to the oncoming flow as
in-line when ~ 0 and as regularly staggered when ~ = 45 °.
=

There are other arrangements where the basic matrix is triangular, either regular or
irregular, as depicted in Figure l(d). A rather special type of pipe cluster is frequently
used for marine risers when a big-diameter pipe is surrounded by smaller pipes in a
circumferential arrangement, Figure lid, (iii)]. The interference between the large and
small pipes depends strongly on the ratio of the pipe diameters. The spacing may be
related to either the big or small pipe diameter.

3. BASIC I N T E R F E R E N C E F L O W REGIMES
There is an infinite number of kinds of flow interference between two parallel pipes of
circular cross section. However, four particular kinds can be distinguished:
(i) Proximity intereference, P, which takes place when the pipes are close to each
other, but none of them is submerged in the wake of another. For example,
interference for pipes in side-by-side arrangements belongs to this category.
(ii) Wake interference, W, which takes place when one pipe is near to or submerged
into the wake of the other. This kind of interference is related only to the
downstream pipe in tandem and staggered arrangements inside the interference
boundary. Figure 2 shows a sketch of the extent of the proximity and wake
interference regions.
(iii) There is a third region, P + W, which represents a combination of proximity and

/ \
/ ,-,o \

P-SS p_sT\" ~ No mterference

Proximity ~ Interference ""~i~ -

.. -;:w' -. . . . . w-u
t I ~°e~',\ ..... Wake interference

I I II
-2 -I

W-TI W-T2 W-T (14-2)

Figure 2. Definition of regions of flow interference for two-pipe arrangements. IN, bistable flow regions.
242 M.M. ZDRAVKOVICH

wake interference. In that region an extreme variation of forces both in magnitude

and direction was found to occur and will be discussed in the next section.
(iv) Finally, there is a substantial region depicted in Figure 2 where the interference is
negligible. The flow around each pipe in that region is effectively identical to that
around a single pipe.
The boundary of wake interference extends always beyond the width of the wake of the
upstream pipe. The width of the wake is defined as a line along which the wake velocity
reaches 999/0 of the free stream velocity. The wake interference boundary is defined as a
line along which the measured lift coefficient on the downstream pipe becomes less than
19/o. Similar definitions are applied for the two proximity interference boundaries shown
in Figure 2. The reference pipe shown in Figure 2 represents a downstream one for
L/D > 0 and an upstream one for LID < O.
The proximity interference region may conveniently be subdivided into three ranges:
P-SS: side by side arrangements of the two pipes;
P-SI: upstream pipe in staggered arrangements;
P-S2: downstream pipe in staggered arrangements.
The wake interference region can be subdivided into two ranges:
W - T : tandem arrangements;
W-S: staggered arrangements.
The proximity-and-wake region, P + W, precedes the wake interference region. It will be
shown that entirely different flow regimes are produced by proximity-and-wake
interference.
4. PROXIMITY INTERFERENCE
4.1. P-SS: SIDE-BY-SIDE ARRANGEMENTS
Three different flow regimes have been observed by Williamson [1] in the laminar state
and by Zdravkovich [2] in the subcritical state of flow around two stationary pipes, as
follows:
P-SSA: 1 < T/D < l'l to 1"2. A single vortex street is formed behind both pipes with a
base bleed through the gap, as depicted in Figure 3.

W-TI
J ",..z.,.~ l L,OUplea Ol (el W-S
5 "1 ~,vortex (
I 0 /streets /L2_._LL -

P-ssB--_1 ,2e g
f ¢f)_f,b,stob,e, }
- "'-"~1 X ~ . , D2 (d)
'[~\( ~"~ 1Single W-T(I+2)~ I'~,~ ~.~, (e)
J \~ I'~vortex k.J --~ ~
.,"'~"-. ( 1 ~.~m/street (fl
."~-"~- - ~ .
/ [ X , " - ~ (b) (c) (d) (e) [f)
\ v ] I 2 5 Bistable 4 5 6 LID
~- J ,-" ~. Two vortex streets
One vortex street
Figure 3. Interference flow regimes for side-by-side, tandem and staggered arrangements of two pipes. For
tandem regimes: (a)single slender body; (b) alternate reattachment; (c)quasi-steady reattachment; (d)inter-
mittent shedding; (e) discontinuous jump (d)~ (f); (f) binary vortex street.
 CYLINDERS IN CROSS FLOW 243
P-SSB: 1.2 < T/D < 2 to 2"2. Narrow and wide wakes are formed behind two identical
pipes, respectively, and the gap flow forms a jet biased towards the narrow wake. The
flow is bistable, i.e. the biased jet can switch in the opposite direction at irregular time
intervals, and the narrow and wide wakes interchange behind the tubes.
P-SSC: 2.7 < T/D < 4 or 5. Both nearwakes are equal in size but the two vortex strt cts
are coupled and " m i r r o r " each other along the gap axis. The vortex shedding is
synchronized, both in phase and frequency, as sketched in Figure 3.
Figure 4 shows the variation of lift and drag coefficients, and recently corrected
Strouhal numbers by Williamson [1], in the subcritical state of flow. The biased jet-flow

(e) ]
1'6

Re
"l" 1.6 ×
<~ 0 ' 4 -'~% [] • 8×10 3
,~ o , 6xlO 4
~@*'~'~ L ~..~ .,% A 2.5x}0 4

"T =i" .... ~' , i ,

2 3 4 5 6

! ~ / ° o

3x I

~ 0-2 ~.s'~ I
2X
', |.
zt
p •
../
•

IX •
I L L
I'0 1.5 2.0 2.5
T/D

Figure 4. Proximity interference for side-by-side arrangements: (a) lift and drag coefficients, after
Zdravkovich [2]; (b) Strouhal number for subcritical and laminar states, after Williamson [1]; (1 x, 2 x, 3 x
are multiples of measured frequencies).
244 M.M. ZDRAVKOVICH

regime produces always larger drag, lift and Strouhal number behind the pipe with the
narrow near-wake. This is in stark contrast with the observation of the same parameters
behind a single pipe, where a wide nearwake is always associated with an increase in drag
and a decrease in Strouhal number. The small vortices produced in the narrow wake
dissipate rapidly and coalesce with the vortices formed in the wide wake. Only a single
lower Strouhal number can be found further downstream. This flow regime is bistable and
the switch of the asymmetric flow pattern can excite large amplitude oscillations, as
reviewed in some detail elsewhere [3, 4].
Figure 5 shows the variation of lift, drag and Strouhal numbers on both pipes in the
critical Reynolds number regime. The formation of separation bubbles occurs only on the
outward sides of the pipes at small spacings because the stagnation points are shifted
towards the gap side. This leads to a significant increase in lift coefficient in comparison
with that in the subcritical state. The appearance of the first bubble on one of the pipes on
the gap side leads to a sudden collapse of lift coefficient. Both drag coefficients appear to
be almost equal as shown in Figure 5. The Strouhal number is also plotted, although the
peaks in the frequency spectra were small and wide-banded (except at small spacings
when a single vortex street was formed behind the two pipes).
The postcritical state, simulated by surface roughness, inhibited the formation of
separation bubbles and the lift coefficient was small, while the drag coefficient was in
excess of that found on smooth pipes (Zdravkovich [5]).

4.2. P-S1 AND P-S2: STAGGEREDARRANGEMENTS

Figure 6 shows the lift, drag and Strouhal number variation typical for the subcritical
state of flow. The upstream pipe experienced positive lift directed outward for c~> 40 °
coupled with a relatively small decrease in drag and a significant increase in Strouhal
number. There is a region of negative lift directed towards the gap in the P-S1 region
with a sharp, almost discontinuous, peak at LID > 1.25 and T/D = 0.25. Note that the
vortex shedding is suppressed in that region and no Strouhal number is given.
The bistable flow found in the side-by-side arrangements extends for the tubes
staggered up to Ix/D[ = 0.15. The narrow and wide wakes can intermittently change over

0"5

o, _ , ,
A

0'2
i•yl
I bubble I /
",. --,.5yI.A
' I ~'-I
IB I I t~_~CylA
I I^ IB I Ol
"~, I

JX!'~ 2A "~ bubbles Ai

0 I I I I
2 5 2 3
T/D

Figure 5. Proximity interferencefor side-by-sidearrangements in the critical state. - - - , Re = 3.4 x 105; - -

Re = 4 × 105; after Okajima and Sugitani [46].
 CYLINDERS IN CROSS FLOW

-5 -4 -3 -2 -I 0 I 2 3 4 5
L/D
Figure 6. Force coefficients for various two-tube arrangements ttle.dsured on one tube in subcrltical state: (a) lift
coeffictent; (b) drag coefficient, after Zdravkovich [Z]; (c) Strouhal number, after Kiya et a[. [47].

behind the staggered pipes. The flow in the staggered arrangement is bistable because
apparently the pipe with a narrow wake takes the role of the “upstream” pipe and the
other with a wide wake appears as the “downstream one”. For (x/D1 > O-15 the
downstream pipe produces always a wide wake and experiences lower lift and drag, and
the Strouhal number decreases as the wake becomes wider [see Figure 6(c)]. Severe
oscillations were found in P-S2 arrangements (Zdravkovich 141).
246 M. M. ZDRAVKOVICH

The critical state of flow was simulated for S/D = 1.325 and g =45 ° and 135°, by
Zdravkovich [5]. As seen in Figure 7 the measured forces on the downstream cylinder are
almost identical to those in the subcritical state of flow. The drag and lift coefficient plots
are compiled from the measurements of Wardlaw and Cooper [6] on stranded cables in
the artificially induced postcritical state at Re = 1.5 x 105.

5. WAKE INTERFERENCE
3.1. W-T: TANDEMARRANGEMENTS
Three different flow regimes have been observed at low Reynolds numbers in the laminar
state (Zdravkovich [7]) and in the suhcritical state of flow around two stationary pipes as
described in more detail by Zdravkovich [2]. These regimes are depicted in Figure 3 and
they are as follows:
W-TI: 1 < LID < 1.2 to 1.8, depending on Reynolds number. The free shear layers
separated from the upstream pipe 1 do not reattach on the downstream pipe 2. The
vortex street behind the latter is actually formed by the free shear layers detached from
the former (Igarashi [8]).
W-T2:1.2 to 1.8 < LID < 3.4 to 3.8 depending on Reynolds number (Zdravkovich and
Pridden [9]). The free shear layers separated from pipe 1 reattach on the upstream
side of pipe 2. A vortex street is formed only behind the downstream pipe 2.
W-T(1 +2): LID > 3.4 to 3.8. The separated shear layers roll up alternately and form
vortices behind pipe 1 in front of the downstream pipe 2. Two vortex streets are
formed behind the pipes. The vortex street behind pipe 2 is called binary
(Zdravkovich [10]), because each vortex really consists of two vortices formed behind
the upstream and downstream pipe. Figure 6 shows the variation of drag and Strouhal
number. The binary vortices cause an excessive fluctuating lift on the downstream pipe
(Arie et al. [11]).
All three flow regimes are bistable in the transitional regions D1 and D2 where sudden
changes of flow regime occur over a range of spacings as depicted in Figure 3. In both D1
and D2 regions a large hysteretic effect is exhibited which depends strongly on whether
either velocity or spacing is increased or decreased.
Figure 7 shows a similar variation in the postcritical state simulated by surface
roughness. Stranded cables were used, with 30 strands wound helically around a core,
with strand-to-core diameter ratio d/D = 0.009.
Figure 8(a) shows the variation of drag coefficient for LID = 3 and 5 over a wide range
of Reynolds numbers. There is a big difference between measured CD1 and Co2 on the two
pipes in the subcritical state, when a wide wake is formed behind the upstream pipe. In
the critical range, the formation of separation bubbles on the upstream pipe reduces
drastically the width of the upstream wake, which leads to an increase in CD2. The
disturbed supercritical downstream wake is wider than the upstream one and produces
the paradox that Co2 is greater than Col. In the postcritical state, the separation bubbles
are completely obliterated on the upstream pipe and Co2 is less than Cot. The Strouhal
number jumps from 0.2 in the subcritical state to 0.3 when the first bubble is formed and
to 0.4 when the second bubble appears. Regular vortex shedding ceases in the
supercritical regime when the separation bubbles are fragmented along the span, and the
separation lines become irregular. However, when the pipes are rough, the bubbles are
not formed, the separation line remains straight and the vortex shedding is prominent.
Figure 8(b) shows the variation of drag and Strouhal number for L/D = 3 and 5, which is
in stark contrast with the measured values for smooth pipes.
 CYLINDERS IN CROSS FLOW 247

(o)
TIO

O9

-5 -2 -I 0 I 2 ~u-~ 3 ,4 5
L/O

(b)

3 cL:o/~X
/

-5 -2 -I 0 I 2 3 4 5
L/D

Figure 7. Force coefficients for two-stranded cables measured in post-critical state: (a) drag coefficient; (b) lift
coefficient, after Wardlaw and Cooper 1-25].

U ,- L _,_ 4D _, L/D
1,5
.~ . . . . . Col 5
L©
t~ stz Co2 5
1'5 coi- --, ~"
__ -~.- 5- Coz 1.0
CoL %~ .. - - ' - - 3
I.O
-" --N~"'~ U ~_~Or--'--5-
, E-E-1?;?uco~
L, --.o~ co~ @ 0.5
Q 0.5 Coz }"'- - -

St ...... I St - • --...__--:.-

I I I I I~'//l 4 I I 1 , I I I -""7"" x~..

4 6 8105 ,... 6 810e 2 4 6 4.,,," 8 105 2 4- 6 8 106
CD~" . . . . . Re Co"~ Re
(a)

-0.5
(b) J
-0.5
Figure 8. Measured variation of drag coefficient and Strouhal number for (a) smooth pipes, after Pearcey
e t a l . [24]; (b) rough pipes, after Okajima [48].
248 M . M . ZDRAVKOVICH

5.2. W-S: STAGGEREDARRANGEMENTS

Two different flow regimes have been observed in the subcritical state of flow around two
staggered stationary pipes:
(i) Gap Flow, W-SG: 1-1 < LID < 3.5 and T/D = 0.2. The strong gap flow between the
pipes induces a significant "inner" lift force (Zdravkovich [2]), which suddenly
disappears for T/D < 0.2 when the gap flow ceases.
(ii) Wake Displacement, W-SD: LID > 2-8 and T/D > 0-4. The "outer" lift force is
produced by the displacement of the fully formed wake of the upstream tube by the
flow around the downstream tube (Zdravkovich [2]). The outer lift reaches a
maximum value near the edge of the wake boundary when T/D >0-4 and
diminishes gradually to zero when the tandem arrangement W-T2 is reached.
Figure 6 shows the variation of the lift, drag, and Strouhal number. The "inner" and
"outer" lift forces are always directed towards the axis of the upstream wake. The
deflected nearwake behind the downstream cylinder in the W-SG region produces a
minimum drag force at the same staggered arrangements where the "inner" lift force is
maximum. In contrast, the "outer" lift maximum has no effect on the drag force or
Strouhal number as seen in Figure 6(a) and (b), see also Bokaian and Geoola [12, 13].
Figure 7 shows the variation of lift and drag in the wake of stranded cables in the
postcritical state of flow. The main difference is a significant reduction in the size of both
the W SD region and the maximum values of lift coefficient. The gap flow in the W-SG
regime is weakened significantly by surface roughness.
Figure 9 shows the variation of lift and drag coefficients for one particular staggered
arrangement (LID = 5 and T/D = 1.54). The increased velocity on the inward side of the
downstream pipe is induced by the wake displacement of the upstream pipe. This triggers
the formation of the first separation bubble with a discontinuous jump in lift coefficient
and a fall in drag coefficient. However, the separation bubble is fragmented due to the
disturbances carried in the upstream wake and the second bubble is formed on the
outward side at a higher Reynolds number. These bubbles produce a net lift force directed
outwards, away from the wake axis, with a very small change in drag coefficient. The

CLI
--~'-~, I I f~" I I I I[I I I f I I I II
Cot "" %. Cz,Z ~ Inward AJr-CL2
-I'0 %. 1 ; bubble f T ~ _ ,~ -
x J I 5p.j~7 ~ ~02
"-PI ~ "'.~'~ J8° coz

_o+
c~2 k i r'--"~ ~"
(~ C D ~ ~ ~ CL2

o
I

C L 2 ~ Outward (?)
05 I I t I I 1 ~ I I bubble I I I I I I I
2 4 6 8 IO e 2 4 6 8 107
Re

Figure 9. Measured variation of drag and lift coefficient for wake interference m tandem arrangements: S/D = 5.
= 18 °, adapted from Pearcey et al. [24].
 CYLINDERS IN CROSSFLOW 249
local disruption of that bubble occurs at Re = 8 x 105 which reduces discontinuously the
lift and increases the drag. Beyond that is a typical supercritical regime and the
posteritical state is established finally beyond (Re = 6 x 106). Note that CDz > Cm because
pipe 2 is almost outside the wake of pipe 1.

6. CLUSTER OF THREE PIPES

6.1. INTRODUCTION
Three pipes can be arranged in various ways which have been conveniently grouped in
Figure l(b) into three categories:
(a) all three pipes aligned and oriented from in-line c~= 0 to side-by-side arrangements
= 90°;
(b) regular triangular arrangement where all pipes are spaced equidistantly and the
flow pattern repeats itself every c~+re/3;
(c) irregular triangular arrangements, composed of all other possible arrangements,
except (a) and (b).

6.2. ALIGNED PIPES

The addition of a third pipe to a tandem arrangement of two, produces a new kind of flow
which differs from both the flows behind the first two pipes. Early pressure distribution
measurements by Pierce [14], for LID = 2, showed that the flow around the first two
pipes was almost unaffected by the addition of the third and fourth pipe. The pressure
distributions around the last two pipes (3 and 4) were similar (Igarashi [8]), and might be
taken as typical for any additional number of pipes aligned in a column. This was
confirmed by the similarity of aeroelastic responses of pipes arranged in the third and
fourth row by Southworth and Zdravkovich [15], and also by heat transfer measurements
by Aiba et al. [16] around four aligned pipes.
A systematic investigation of the flow around three aligned pipes at c~= 0 ° has been
carried out by Igarashi [8] in the subcritical state of flow. Figure 10 shows the pressure
drag coefficients for all three pipes and Strouhal numbers measured behind the third one.

1.2 I I I I 0.25 I I 1 I T

~ (a) ~W-TI (b)

0.8
CD3 t"' -- 0.20
W+T (1+2+3)
0.4-
@ I 0'15
w-T~,.. W-TC2+~/
0

0-10
-0-4
Re = 2-2 x I04

-0.~ l l l i l 0.05 I I ] I ;
2 5 4 2 3
L/D
Figure 10 Effect of spacing on flow around three aligned pipes: (a)drag coefficient (Re = 2.2x 104);
(b) Strouhal number (Re = 2 7 x 10% after Igarashi and Suzuki [17].
250 M.M. ZDRAVKOVICH
The three flow regimes which have been observed for the two pipes in tandem, were again
identified for a third pipe but within a different range of spacings and in a modified form.
The flow regime of a vortex street formed by the free shear layers separated from the
first pipe has distinct jumps in CD2 and CD3 and a fall in the Strouhal number at the lower
LID end of the range. The bistable nature of occasional reattachments is evident in
hysteresis in Figure 10(b). The second flow regime W-T3 was characterized by the
reattachment of the free shear layers first to the third pipe, and subsequently to the second
pipe, W-T(2 + 3). The angle of reattachment on the third pipe was significantly less than
that on the second pipe. This promoted the formation of vortices in the gap between the
second and third pipe as early as LID = 2 to 2.2. The subsequent attainment of the
W - T(1 + 2 + 3) flow regime behind the first pipe at LID = 3.5 decreased slightly CD3 and
increased considerably the Strouhal number in a discontinuous and bistable manner as
depicted in Figure 10(b). Similar flow regimes were found behind four aligned tubes by
Igarashi [17].
Dalton et al. [18] measured only the drag component of force exerted on each of three
aligned cylinders in the range 0°< e < 90 °. Figure 11 shows a summary of data for
R e = 6 . 7 x 104. It is evident that the variation of the drag coefficient on the third
(downstream) pipe is considerably different from that behind the second pipe. There was
an increase in CD3 in comparison to CD2, but Coa was always less than Col.

(o)

4
O9

-Fg-OB-o4-o.2 o
0 I 2 3 4 5
L/D

r/D (b)

-4 -3 -2 -I 0 r 2 3 4 5
L/D
Figure 11. Measured drag coefficient on three aligned pipes in tandem, staggered and side-by-side arrangements
at Re = 6.7 x 10 4, adapted from Dalton and Szabo [18]. (a) Second of three cylinders; (b) third cylinder.
 CYLINDERS IN CROSS FLOW 251

A simulation of the postcritical state, Re~,m = 1.5 x 107, was attempted by Gerhardt and
K r a m e r [19] by using rough surfaces (k/D=6x 10 -3) on pipes with a free end
(H/D = 28.6). The middle pipe experienced less drag than the side ones for small spacings
T/D = 1.5 and 2 but the reverse was found for T/D = 3. This indicated that some sort of
biased gap flow produced unequal nearwakes even at T/D = 3. The different drag on each
tube was also found when the fourth side pipe was added at the same spacing to the
other three.

6.3. EQUIDISTANT TRIANGULAR CLUSTER

One of the main reasons for introducing the basic interference flow regimes for two pipes
was the expectation that the fluid forces on three pipes in a triangular cluster would bear
some similarities and have the same physical origin. As an example, Figure 12 shows the
variation of measured lift and drag forces on four equidistant triangular arrangements in
the subcritical state of flow (Re = 2 x 105, H/D = 10, free ends).

1.0
(a)

T
A

@ 05
_

~-~~
Model on T~/ ~
L 5 ~

/ "b~/.....
"%.~_.._S
"""" '.

' 'i,'.\
~\
E
M
2
~ It'
17/
'
f.

...

-- f h~e b a l o n~c e l ~ x .
(f,,
/..¢/d--IO ~ ' "
"% y;
...j,
S/d
.......... 2.0
2.5
..... 50
5.0
I

o Ts % ~ ~ " ,"h/b/
~/'~--~ ~_ D @/-. " !,,~
o.~o,-- - .~r~-L_ / ....

Modelon ~ " \ ,.r"%r- . !i~:!

theboiance/\\~ ~ / .._._~\ ~! , ~

d~ o ....... , -..,,, ,6/j.

-o.25 T,,..i ' ,I m

N s/u i~ I! N
• ~ ......... 2.0 . i\,~,~ It
,i DE
- 0.50 - - M 25 -- "~! M
..... 30
I - - 5 0 2
-0.75 I , I
0° 30 ° 60 ° 90 ° 120 ° 150 ° 180 °
(2

Figure 12. Forces measured on one pipe in three-pipe cluster at R e = 2 x 105: ( a ) d r a g coefficient; (b)lift
coefficient, after Bardowicks [-49]. For • = 30 ° the monitored tube is upstream in T A N D E M 1 arrangement,
and for c~= 150 ° it is the downstream one in TANDEM 2 arrangement.
252 M. M. ZDRAVKOVICH

The upstream pipe for a =40“, experienced the lift directed inwards only for the
smallest spacing. Considerable lift was exerted on the downstream pipe for a = 130” at all
spacings, as it was in the category W-SD. However, an outward lift force was found for
the smallest spacing governed by the P-SSB flow regime. The third set of peaks
correspond to the category W-SC, except the smallest peak which belongs to the W-SD
category. The measurements of drag force of the cluster of roughened pipes
(k/D = 1.8 x 10e3, in order to simulate Re, = 8 x lo6 around smooth pipes) showed a
similar variation of drag coefficient (see Gerhardt and Kramer [19]). There is a complex
interaction of the three wakes downstream, as observed at low Reynolds number by
Zdravkovich [20,21].

6.4. DISTORTED TRIANGULAR CLUSTER

An additional insight into the variation of fluid forces on one of the three pipes in the
cluster can be obtained by displacing that pipe from the initially equidistant triangular
arrangement [22]. Figure 13 shows the variation of drag and lift coefficients produced by
the gradual displacement of the upstream (L/D = - 1.5) or downstream (L/D = l-5) pipe
in the transverse direction. Note that now it is not possible to explain results in terms of
those for a two-pipe group.
When two of the pipes are in the tandem arrangement maximum lift was measured on
the third one despite the drag being a minimum. The third pipe induced presumably a
strong gap flow between the other two and produced a repulsive (outward) lift force. The
expected inward lift force did appear when the downstream pipe was staggered relative to
the upstream one. The drag was reduced in comparison with that found in the W-SC
range in Figure 6. When the traversing pipe was placed upstream, the same type of gap
flow produced an inward lift force. Price and Paidoussis [22] also measured the forces for

(a)
0.80

040

-0.80
-0.80 I
-5 -4 -3 -2 -

Figure 13. Force coefficients on the upstream pipe and downstream pipe in a cluster, after Price et al. 1221:
(a) drag coefficient (--, L./O= 1.5; ---, L/D= -1.5); (b) lift coefficient (--, L/D= 1.5; ---,
L/D = - 1.5).
 C Y L I N D E R S IN C R O S S F L O W 253

the same transverse spacing at x / D = 5. The large lift coefficients were absent which
proved that a W SD interaction was occurring and induced the small "external" lift.

7. S Q U A R E C L U S T E R S
The main feature of square clusters, as seen in Figure l(c), is that at c~= 0 all pipes in
successive rows are arranged in-line relative to the free stream and at c~= 45 ° they are all
regularly staggered. The overall resultant force coefficient on square clusters is found to
be almost the same, irrespective of the n u m b e r of pipes in the cluster (Ball and Hall [23],
Pearcey et al. [-24]).
Table 1 shows the drag coefficient as measured in a simulated high Reynolds n u m b e r
flow and reported by Pearcey et al. [24]. The expected values of CD were found for the
upstream and downstream pipes in 2 x 2 clusters. However, the 3 x 3 cluster p r o d u c e d a
drag coefficient in the third row which was less than in the second row. This was at
variance with CD measured on three aligned tubes in the subcritical state. The same trend
can be seen in Table 1 at c~= 45 ° on the third pipe in the middle column. A considerable
increase of drag coefficient occurred at e = 18 ° owing to wake displacement, not only for
the pipes in the second row but also, and even m o r e so, for the pipes situated in the third
row. The latter is in stark contrast to that found on the third aligned pipe. The m o s t
important conclusion to be drawn from Table 1 is that the m e a n drag coefficient for the
cluster is not representative of any individual pipe in it. F o r example, the pipes in the
upstream or last row m a y experience up to 30~o higher drag than the mean value,
depending on the orientation of the cluster. There is also a lift c o m p o n e n t in all staggered
arrangements which has not been measured by Pearcey et al. [24].
Figure 14(a) shows the variation of b o t h drag and lift coefficients for a stranded cable
in the postcritical and subcritical state of flow respectively. The latter reduced the drag
coefficient when the downstream pipe was staggered but not when it was in tandem. This
was in agreement with Figure 8, where the s m o o t h d o w n s t r e a m pipe in the critical state
had higher drag when shielded. In the post critical state, however, the reverse was true

TABLE 1
Drag coefficient for pipes in square clusters (S/D = 5)

Angle of stagger

Cluster 0° 18° 45 °

2x2 0.61 0.61 0-60 0"63 0-62

Individual C D 0.44 0-43 0.69 0.71 0-64 0-64
0-44

Mean C o 0.52 0.64 0.58

0.58
0.61 0.62 0.62 0-58 0.60 0.66 0.61 0.63
3×3 0.43 0-41 0.43 0-64 0-69 0.76 0-69 0-48 0.68
Individual C D 0-37 0.39 0.34 0.73 0.75 0.61 0.47 0.47
0.40

Mean. Co 0.47 0.67 0.56

4 × 4 Mean C D 0.43 0-62 0.59

254 M. M. ZDRAVKOVICH
1,2
(b}
1.0 f---x it---

O8

@o6 k 0
,
~ ,
\
~
/
/

o
o4
0
o 15d
o2

o-o O0
<.Y
-o 2 0 ?10 4'0 6tO
0 4O ....

0 o

Figure 14. Measured force coefficients of four-tube clusters of stranded cables: (a) S/D = 10, Re = 2 × 10s;
(b) S/D = 13, Re = 4 × 104 ( - - - , smooth flow; - - , grid turbulence 10%); after Ko [26] and Wardlaw
et al. [6].

and this is in variance with the results obtained with stranded cables. Hence, the
simulation of the postcritical state by surface roughness does not produce the same flow
pattern as do the s m o o t h pipes and the interference flow regimes are evidently different.
The effect of free stream turbulence was also studied by Wardlaw et al. [25] and Figure
14(b) shows that 10% grid turbulence decreases both the lift and drag coefficients in the
subcritical state of flow a r o u n d stranded cables ( S / D = 13). A further increase in the
n u m b e r of stranded cables to 6 and 8, arranged circumferentiaUy, produced the same
variation of lift and drag coefficients for the same interference flow regimes (see K o [-26]).

TABLE 2
Lift and drag coefficient on the pipe in the second row of a three-row cluster
(T/D = 1.66 and Re = 4 × 104; Zdravkovich et al. [27])
(a) Regular triangular stagger

S/D t'46 2"16 2"63

CL 0'03 0'21 0'26

Co 0.43 0"59 0'40

(b) Rectangular in-line arrangement

L/D 1'2 2.0 2-4

CL 0'03 0'0 0'01

Co -0"012 -0'07 0"09

(c) Irregular triangular stagger

(S/D)1 1-32 2-07 2.56 1.23 2.02 2.51

(S/D)2 1.63 2.28 2.73 1.83 2.43 2.85

CL 0.52 0.48 0.30 0.70 0.81 0.77

CD 0"03 0'21 0'16 0"50 0'48 0'34
 CYLINDERS IN CROSS FLOW 255
c;

0 I~ '~i 90 180 270 I #"1560°

'

• dl
/ K\ $/
.._
,_. ~
Row

> ,"~s
"~-'-": ~
I
I
/,/ /

'4 .,.+- •
M M
Figure 15. Measured pressure distribution around pipe situated in first three rows of a six-row tube bank
depicted in Figure 16, after Zdravkovich and Namork [50].

8. STAGGERED CLUSTERS
Square clusters are only one of the possible regular arrangements. Another arrangement,
which is often used in heat exchangers, is an equidistant triangular stagger, as shown in
Figure l(d). A peculiar feature of this arrangement is the "free alley" running diagonally
between the pipes, see Figure 16, where fluid can flow along straight paths. This peculiar
feature may enable a continuous biased flow to take place along the free alleys. Table 2
demonstrates the existence of a significant lift coefficient, evaluated from a mean pressure
distribution around the pipe situated in the second row. The lift coefficient appears also
for isosceles triangles.
The structure of interstitial flow between the tubes for the equidistant triangular stagger
S/D = 1.375 may be inferred from the pressure distribution measured around tubes in the
first three rows in a six-row deep bank. Figure 15 shows a significant variation of pressure
distribution, not only in the separated region but also in the region of favourable pressure
gradient. The tube in the first row is subjected to an accelerated flow due to the vicinity of
the adjacent tubes on both sides, as sketched in Figure 16. The tube in the second row
experiences an additional contraction of flow due to the wide nearwakes formed by the
tubes in the first row. This produces a drastic increase in favourable pressure gradient
leading to a first Cpmi, at about 50°, where the minimum "flow cross section" occurs. A
small adverse pressure gradient follows, before another acceleration of flow takes place
due to the vicinity of the tubes in the second row. The second Cpmi, occurs at 90 ° where
the gap is a minimum, as seen in Figure 16. Separation occurs around 120°-130 °, leading

\
(
;>

S
Figure 16. Interstitial flow within a six-row deep tube bank (schematic), after Zdravkovich and Namork [50].
256 M.M. ZDRAVKOVICH

o (2h +"zj+ 67~

£:0; ,¢',
-i e+e+e t#

-IOf 7
-12 i | I I I I I I I i t
0 90 180 270 560
8 (degrees)
Figure 17. Pressure distribution around tube in second row, staggered by (2/3)T/D, measured for three
streamwise displacements [L/D = 1-2 (~), 2.0 (O), 2.5 (E])], after Zdravkovich et at. [27].

to a narrow and cusped wake, which has been "sized" by a hot-wire traversed along the
chain-dotted line in Figure 16. The tube in the third row is subjected to a less steep
favourable pressure gradient than the tube in the second row because the flow
constriction by the narrow wake is smaller. The gap at 90 ° represents the minimum
contraction for the flow and results in a Cpmin in the third row. The cusped nearwake is
very similar to that found behind the tubes in the second row. There was very little
variation of measured pressure distribution in subsequent rows from that found in the
third row.

® -'®* ®
=p I
-2
®@® //
-4
@
-6

-8

-I0
I I I I I I l I I I I I
0 90 180 270 :560
8 (degrees)
Figure 18. Measured pressure distribution around tube located m second row, staggered by (I/3)T/D, for three
streamwise displacements [LID = 1.2 (V), 2.0 (O), 2-5 (D)], after Zdravkovich et al. 1-27].
 CYLINDERS IN CROSS FLOW 257
A further insight into the structure of interstitial flow can be gained by varying the gap
size on opposite sides of the monitored tube. This has been achieved by keeping the
transverse pitch constant (T/D= 1.66) and varying the location of the second row
upstream and downstream and in the transverse direction, as depicted with crosses in
Figures 17 and 18. Two unsymmetrical pressure distributions are shown for the (2/3)T/D
and (1/3)T/D stagger, that is when the second row as a whole was displaced towards the
in-line arrangement. The most interesting feature in Figure 17 is a discontinuous drop in
pressure following the adverse pressure gradient at 0 = 60 ° for LID -= 1.2. Some kind of
"separation bubble" was formed on the upstream side of the cylinder. This conjecture was
recently corroborated by the flow visualization study carried out by Abd-Rabbo and
Weaver [28], and Fujita et al. [29] who triggered separation bubbles with trip wires.
Table 2 specifies the spacings of three irregular triangular arrangements and presents
the evaluated pressure lift and drag coefficients. A significant variation of both lift and
drag coefficient is found, with the former exceeding the latter for some arrangements.
There is another even more "disturbing" feature of the interstitial flow which deserves
attention. The fluctuating pressure around all tubes increases significantly, despite the fact
that a distinct narrow-band peak, corresponding to vortex shedding, cannot be found in
the closely packed banks (Weaver and Graver 1-30]). The magnification of the fluctuating
pressure is particularly strong in the second row, as seen in Figure 19. This leads to the
excitation of oscillations of large amplitude, which usually commences in the second row
and can lead to fatigue failure and fretting of clashing pipes. This phenomenon was
modelled mathematically by Blevins [31], and described in several review papers by
Paidoussis [32, 33]. It will be treated in detail by Chen [34] in a forthcoming book.

Scale of
3~0 Prrm

5O

°f
100
150
N/m z

Figure 19. Measured rms fluctuating pressure and time-averaged pressure distribution around a tube located in
rows 1-6, after Zdravkovich and N a m o r k 151].
258 M. M. ZDRAVKOVICH

TABLE 3
Critical spacing in terms of diameter ratio (D1 = diameter of upstream pipe)

D1/D2 0"132 0"263 0"395 0'526 1'000 1"47

LID 1 8'8 5'5 4'0 3"2 3"8 2'0

9. PIPES OF DIFFERENT DIAMETERS

In some practical applications, such as marine risers and axial rod shrouds (Zdravkovich
[35]), two different sizes of pipes are combined in a cluster. A typical configuration
consisting of a regular circular arrangement of small pipes around a large diameter pipe is
depicted in Figure lld, (iii)]. The interference flow is affected by an additional parameter
which is the ratio of pipe diameters.
When the small pipe is laid upstream of the big one it promotes transition in the
boundary layers of the latter by its turbulent wake, with high-frequency vortex shedding.
The "jump" phenomenon, a characteristic feature of two identical pipes in tandem, was
also observed in the gap between two cylinders of different diameters by Hiwada et al.
[36]. However, the critical spacing where the jump occurred depended not only on the
Reynolds number but also on the diameter ratio, as seen in Table 3.
Igarashi [37] studied the case when the large pipe was upstream in a tandem
arrangement, D1/D2= 1.47. However, the "jump" in the base pressure and drag
coefficient on the upstream cylinder did not induce a discontinuous jump in Strouhal
number, as was the case for two pipes of the same diameter. Further research was recently
reported by Bokaian and Geoola [12, 38] and Barnes et al. [39]. The cluster of five
roughened pipes arranged as shown in Figure l[d(iii)] was studied by Gerhardt and
Kramer [19]. The overall force coefficient of the cluster is shown in Figure 20 for various
orientations ~. It is regrettable that hydrodynamic forces measured on 15 multitube
production risers were based on the central tube (Demirbilek and Halvorsen [40]). This
made it impossible to single out the optimal geometry.
An axial-rod shroud placed around a large cylinder, where diD = 0.025, was studied by
Zdravkovich [35] and Zdravkovich and Southworth [41].

Resim= 7 x 106
, 55d
k/d= 12 x I 0 -2
k/D = 71 x I 0 -3
1.0

@ ~3d~

0.5

0 I I
0o 45 ° 90 °
a

Figure 20. Measured overall drag force coefficient C e on circular cluster of five tubes, after Gerhardt and
Kramer [191 (D/d = 2.33). k = size of roughness.
 CYLINDERS IN CROSS FLOW 259

10. C L O S I N G C O M M E N T S
The attempted categorization of the flow through pipe clusters has revealed that the
available experimental data diminishes rapidly as the number of pipes increases. The most
complete physical picture of the flow patterns was developed for only two pipes in various
arrangements. These basic interference flow regimes are not sufficient to describe the
additional interactions of pipes in larger clusters (n >i 3).
A particularly limited data base (at least unclassified) is available for the supercritical
and postcritical states of flow. The main thrust of future research should be directed to
these regimes.
Another important aspect which has not attracted sufficient attention is the effect of
pipe motion on vortex shedding. The modification of vortex shedding in the
synchronization range for a single oscillating pipe was shown to be excessive, see the
reviews by Ericsson [42], Zdravkovich [43] and Bearman [44]. Even more dramatic
modifications and the emergence of m a n y more interference-induced excitations were
discovered for two pipes in various arrangements and were classified recently by
Zdravkovich [45]. The closely packed pipe clusters and their intrinsic instability is still a
mystery.

REFERENCES
1. C. H. K. WILLIAMSON1985 Evolution of a single wake behind a pair of bluff bodies. Journal of
Fluid Mechanics 159, 1-18.
2. M. M. ZDRAVKOVICH1977 Review of flow interference between two circular cylinders in
various arrangements. A SME Journal of Fluids Engineering 99, 618-631.
3. M. M. ZDRAVKOVICH1984 Classification of flow-induced oscillations of two parallel circular
cylinders in various arrangements. A S M E Flow Induced Vibration Symposium (ed. M.P.
Paidoussis et al.), 2, pp. 1-18. New York: ASME.
4. M. M. ZDRAVKOVICH1985 Flow induced oscillation of two interfering circular cylinders.
Journal Sound and Vibration 101, 511-521.
5. M. M. ZDRAVKOVICH1980 Aerodynamics of two parallel circular cylinders of finite height at
simulated high Reynolds numbers. Journal of Wind Engineering and Industrial Aerodynamics
6, 59-71.
6. R. L. WARDLAW,K. R. COOPER, R. G. Ko and J. A. WATTS 1974 Wind tunnel and analytical
investigations into the aeroelastic behaviour of bundled conductors. I.E.E.E. Summer Meeting
and Energy Resources Conference, Anaheim, California. Paper T74, 368-7.
7. M. M. ZDRAVKOVICH1972 Smoke observations of wakes of tandem cylinders at low Reynolds
numbers. The Aeronautical Journal 76, 108-114.
8. T. IGARASHI1981 and 1984 Characteristics of the flow around two circular cylinders arranged
in tandem. Bulletin of the Japan Society of Mechanical Engineers, Report 1, 24, 323-331;
Report 2, 27, 2380-2387.
9. M. M. ZDRAVKOVICHand D. J. PRIDDEN 1977 Interference between two circular cylinders;
series of unexpected discontinuities. Journal of Industrial Aerodynamics 2, 255-270.
10. M. M. ZDRAVKOWCH1986 Discussion on effect of vibrating upstream cylinder of two circular
cylinders in tandem arrangement by M. Moriya and H. Sakamoto. ASME Journal of Fluids
Engineering 108, 180-185 (in press).
11. M. Am~, M. KIYA, H. TAMURA and H. MOR~ 1983 Pressure fluctuations on two circular
cylinders in tandem arrangement. A S M E Journal of Fluids Engineering 105, 161-167.
12. A. BOKAIANand F. GEOOLA 1984 Vortex shedding from two circular cylinders. Proceedings
A SCE, Journal of Engineering Mechanics Division 110, 673-678.
13. A. BOKA~ANand F. GEOOLA 1985 Wake displacement as cause of lift force on cylinder pair.
Proceedings ASCE, Journal of Engineering Mechanics Division 111, 92-95.
14. H. R. PIERCE 1973 Noise and vibration in heat exchangers. Ph.D. Thesis, Oxford University.
15. P. J. SOUTHWORTHand M. M. ZDRAVKOWCH1975 Cross flow induced vibrations of finite tube
banks in in-line arrangements. Journal of Mechanical Engineering Sciences 17, 190-198.
260 M.M. ZDRAVKOVICH

16. S. AII3A,H. TSUCH1DAand T. OTA 1981 Heat transfer around a tube bank. Bulletin of the Japan
Society of Mechanical Engineers 24, 380-387.
17. T. IGARASHIand K. Suzuki 1984 Characteristics of the flow around three circular cylinders
arranged in-line; and 1986 Characteristics of flow around four circular cylinders. Bulletin of the
Japan Society of" Mechanical Engineers 27, 2397-2404, and 29, 751-757.
18. C. DALTON and J. M. SZABO 1977 Drag on group of cylinders. ASME Journal of Pressure
Vessel Technology 99, 152-157.
19. J. J. GERHARDTand C. KRAMER 1981 Interference effects for groups of stacks. Journal of Wind
Engineering and Industrial Aerodynamics 8, 195-202.
20. M. M. ZDRAVI(OVICH1968 Smoke observations of the wake of a group of three cylinders at low
Reynolds number. Journal of Fluid Mechanics 32, 339-351.
21. M. M. ZDRAVKOVICH 1973 Smoke visualisation of three dimensional flow patterns in a
nominally two dimensional wake. Journal de M4canique 12, 225-233.
22. S. J. PRICE and M, P. PAIDOUSSIS1984 The aerodynamic forces acting on groups of two and
three circular cylinders when subject to a cross-flow. Journal of Wind Engineering and Industrial
Aerodynamics 17, 329-347.
23. D. J. BALL and C. D. HALL 1980 Drag of yawed pile groups at low Reynolds numbers. Journal
of the Waterway, Port, Coastal and Ocean Division, Proceedings ASCE 106, 229-238.
24. H. H. PEARCEV,R. F. CASH, I. J. SALTERand A. BORmOND 1982 Interference effects on the drag
loading for groups of cylinders in uni-directional flow. National Maritime Institute Report
NMI R 130.
25. R. L. WARDLAW and K. R. COOVER 1973 A wind tunnel investigation of the steady
aerodynamic forces on smooth and stranded twin bundled power conductors for the Aluminum
Company of America. National Research Council of Canada, LTR-LA-117.
26. R. G. Ko 1973 A wind tunnel investigation into the aerodynamic stability of bundled
conductors for Hydro-Qu6bec, Part III. National Research Council of Canada, Report
LTR-LA-110.
27. M. M. ZDRAVKOVICH,S. SINGH, J. A. NUTTALL and D. M. CAUSON 1976 Flow induced
vibrations in staggered tube banks. Proceedings 6th Thermodynamics and Fluid Mechanics
Convention, Institution of Mechanical Engineers, London, pp. 237-243.
28. A. ABD-RABBOand D. S. WEAVER1984 A flow visualisation study of a square array of tubes in
water cross flow. ASME Symposium on Flow Induced Vibrations (eds M. P. Paidoussis et al.) 2,
pp. 165 178. New York: ASME.
29. H. FUJITA, H. TAKAHAMAand T. KAWAI 1985 Effects of tripping wires on heat transfer from a
circular cylinder in cross flow. Bulletin of the Japan Society of Mechanical Engineers 28, 80 88.
30. D. S. WEAVERand L. K. GROVER 1978 Cross flow induced vibrations in tube bank. Journal of
Sound and Vibration 59, 263-294.
31. R. D. BLEVINS 1977 Flow Induced Vibrations. New York: Van Nostrand, Reinhold.
32. M. P. PAIDOUSSIS1980 Flow induced vibrations in nuclear reactors and heat exchangers. In
Practical Experiences with Flow Induced Vibrations (eds E. Naudascher and D. Rockwell),
pp. 1-81. Springer Verlag, Berlin.
33. M. P. PAIDOUSSIS1981 Fluid-elastic vibration of cylinder arrays in axial and cross flow. Journal
of Sound Vibration 76, 329-360, and 1983 Nuclear Engineering and Design 74, 31-60.
34. S. S. CHEN 1987 Flow Induced Vibration of Circular Cylindrical Structures. Hemisphere Publ.
Co, Washington DC (in press).
35. M. M. ZDRAVKOVICH1971 Circular cylinder enclosed in various shrouds. ASME Vibrations
Conference, Toronto, Canada, Paper 71-VIBR-28.
36. M. HIWADA, T. TAGUCHI, I. MABUCHI and M. KUMADA 1979 Fluid flow and heat transfer
around two circular cylinders of different diameters in cross flow. Bulletin of the Japan Society
of Mechanical Engineers 22, 715-723.
37. T. IGARASHI1982 Characteristics of flow around two circular cylinders of different diameters
arranged in tandem. Bulletin of the Japan Society of Mechanical Engineers 25, 349-357.
38. A. BOKAIAN and F. GEOOLA 1985 Hydrodynamic forces on a pair of cylinders. Offshore
Technology Conference, Houston, Paper OTC 5007.
39. F. H. BARNES,A. J. BAXENDALEand I. GRANT 1985 The flow past two cylinders having different
diameters. The Aeronautical Journal 89, 125-134; also 1986 A lock-in effect in the flow over two
cylinders. The Aeronautical Journal 90, 128-138.
40. Z. DEMIRBILEKand Z. HALVORSEN 1985 Hydrodynamic forces on multitube production risers
exposed to currents and waves. ASME Journal of Energy Resources Technology 107, 226-234.
 CYLINDERS 1N CROSS FLOW 261
41. M. M. ZDRAVKOVICHand D. J. SOUTHWORTH 1973 Effect of shroud eccentricity on suppression
of flow induced vibrations. Journal of Sound and Vibration 27, 77-84.
42. L. E. ERICSSON 1980 Karman vortex shedding and the effect of body motion. AIAA Journal 18,
935-944.
43. M. M. ZDRAVKOVICH 1982 Modification of vortex shedding in the synchronisation range.
A SME Journal of Fluids Engineering 104, 513 517.
44. P. W. BEARMAN 1984 Vortex shedding from oscillating bluff bodies. Annual Review of Fluid
Mechanics 16, 195-222.
45. M. M. ZDRAVKOVICH 1984 Reduction of effectiveness of means for suppressing wind induced
oscillation. Engineering Structures 6, 344-349.
46. A. OKAJIMA and K. SUGITANI 1981 Aeroelastic oscillation of a cylinder in tandem
arrangements. Bulletin of the Research Institute of Applied Mechanics, Kyushu University 57,
247-265 (in Japanese).
47. M. KIYA, M. ARIE, H. TAMURA and H. MORI 1980 Vortex shedding from two circulars in
staggered arrangement. A SME Journal of Fluids Engineering 102, 166-173.
48. A. OKAJIMA1977 Aerodynamic characteristics of stationary tandem cylinders at high Reynolds
numbers. Bulletin of the Research Institute of Applied Mechanics, Kyushu University 46, 111-127
(in Japanese). See also 1979 Flows around two tandem cylinders at very high Reynolds
numbers. Bulletin of the Japan Society of Mechanical Engineers 22, 504-511.
49. H. BARDOWICKS1980 A new six-component balance and applications on wind tunnel models of
slender structures. Private communication.
50. M. M. ZDRAVKOVICH and J. E. NAMORK 1979 Structure of interstitial flow between closely
spaced tubes in staggered array. Flow Induced Vibration (eds S. S. Chen et al.), pp. 41-46. New
York: ASME.
51. M. M. ZDRAVKOVICHand J. E. NAMORK 1980 Excitation amplification and suppression of flow
induced vibrations in heat exchangers. Practical Experiences with Flow Induced Vibration (eds
E. Naudascher and D. Rockwell), pp. 109-117. Berlin: Springer Verlag.