Flow Measurement and Instrumentation 15 (2004) 187–197

www.elsevier.com/locate/flowmeasinst

Flow characteristics of circular cylinders arranged side-by-side

in shallow water
Huseyin Akilli
, Atakan Akar, Cuma Karakus
Çukurova University, Faculty of Engineering and Architecture, Department of Mechanical Engineering, 01330 Adana, Turkey

Received 6 January 2004; received in revised form 2 April 2004; accepted 7 April 2004

Abstract

The flow around two and three side-by-side circular cylinders of equal diameter in shallow water was investigated experimen-
tally using the particle image velocimetry (PIV) technique, over a transverse gap ratio in the range of G=D ¼ 1:0 3:0 with an
increment of 0.25. The experiments were conducted in a closed-loop water channel with Reynolds number 5000 based on the cyl-
inder diameter. In addition to the PIV measurements, flow visualization studies were also carried out. The PIV technique pro-
vided instantaneous and averaged patterns of velocity, vorticity and Reynolds stress contours and streamline topology. For the
two side-by-side cylinder case, it was found that the flow structure behind the cylinders is asymmetrical at small gap ratios as a
result of jet-like flow between the cylinders. The jet-like flow tends to deflect toward the narrow wake region that has a higher
vortex shedding frequency. In the case of three cylinders, both an asymmetrical flow structure at small gap ratio (G=D ¼ 1:25)
and a symmetrical flow structure at intermediate gap ratios (1:5 G=D  2:0) were observed. Bistable wake regions were obtained
for the asymmetrical cases. The Reynolds stress (u0 v0 ) downstream of the upper cylinder was significantly attenuated for G=D ¼
1:25 where the jet-like flow is more effective for both two and three circular cylinder arrangements.
# 2004 Elsevier Ltd. All rights reserved.

Keywords: Side-by-side cylinders; Shallow water; PIV; Wake flow; Vortex shedding

1. Introduction wake of islands, and understanding of island wake flow

is very important in terms of water quality. Complex
A shallow water flow is defined as the situation eddy-like motion in the wake region of islands creates
where the horizontal length scale of a typical eddy is poor water quality on the sheltered side of islands and
significantly larger than the vertical extent of the flow, can trap sediments and pollutants. Understanding of
which means that the dominating processes take place this kind of flow plays an important role in the
in the horizontal plane. Many wake flows can be char- location of outfall discharges, mud disposals, cooling
acterized as shallow turbulent flows in nature, for intakes, marine parks and reserves [4].
example, in wide rivers, lakes, estuaries, shallow coastal Studies of Ingram and Chu [1] and Chen and Jirka
waters, mountains or the stratified atmosphere. Differ- [5] show that the flow structure behind bluff bodies in
ent flow patterns around islands were investigated by shallow water is different from that in deep water. First
Ingram and Chu [1], Wolanski et al. [2] and Lloyd et al. of all, in certain cases, the limited water depth prevents
[3]. Aerial photographs or satellite images can reveal the development of three-dimensional instabilities. Bot-
striking occurrences of such flows, often associated tom friction, which acts as a mechanism for sup-
with large-scale well ordered and coherent flow pat- pression of the transverse growth of disturbance, plays
terns. Due to the complex flow pattern of water in the an important role in the development of the turbulent
wake in shallow water. Ingram and Chu [1] studied dif-

ferent instability mechanisms in shallow water flow and
Corresponding author. Tel.: +90-322-338-6084x2722; fax: +90-
322-338-6126. defined a dimensionless wake stability parameter S to
E-mail address: [email protected] (H. Akilli). account for the stabilizing effect of the transverse shear.
0955-5986/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.flowmeasinst.2004.04.003
188 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197

The parameter S which is based on the ratio of the [16]. The work of Spivack [17] addresses the unsteadi-
small-scale kinetic energy loss to the production of ness of gap flow observed in different gap ratios (G=D,
large-scale turbulent kinetic energy is defined as S ¼ ratio of distance between the center of cylinders to the
Cf D=h for the bluff body, where Cf is the bottom fric- diameter of cylinders). He pointed out that the gap
tion coefficient, D is the width or diameter of a bluff ratio could lead to the formation of single or multiple
body, and h is the water depth. Chen and Jirka [5] wakes. Moreover, the vortex shedding frequency of the
described the critical values of S in order to classify the wake downstream of two circular cylinders is the same
different instability mechanisms. For S < 0:2, a Kár- as that of the wake of a single circular cylinder for the
mán vortex street forms. For S > 0:5, a steady-bubble gap ratio larger than 2.0. When the gap ratio is adjus-
wake occurs as a result of the suppression of large- ted between 1.4 and 2.0, the vortex shedding fre-
scale instabilities by the bottom friction. For quencies on both sides of the wakes are different. For
0:2 < S < 0:5, an unsteady-bubble wake occurs. Lloyd the gap ratio smaller than 1.4, single vortex shedding
and Stansby [6] investigated both experimentally and occurs.
numerically the flow patterns around conical island Bearman and Wadcock [9] measured pressure distri-
models in shallow water. They defined regimes of bution around two cylinders in a side-by-side arrange-
possible wake pattern. ment. The flow becomes asymmetric with two cylinders
The wake flow behavior behind bluff bodies in shal- experiencing different drags and base pressure and the
low water flow helps us to understand the transport drag of the cylinders in combination was less than the
characteristics of the flow. Akilli and Rockwell [7] sum of the drags of the isolated cylinders. Kiya et al.
investigated the near wake of a circular cylinder in ver- [11] investigated the wake behavior downstream of the
tical position in shallow water using a combination of circular cylinders and observed that the characteristics
visualization marker and particle image velocimetry of the wake were significantly related to the arrange-
technique (PIV). They located a laser sheet parallel to ment patterns of the cylinders. Williamson [12] con-
the bed of the water channel at different heights to
cluded that the low-frequency vortex shedding is
investigate the flow characteristics in close regions of
accompanied by a broad wake flow but the high-fre-
the bed and at the various locations above it. At the
quency vortex shedding occurs due to the narrow
bed, the time-averaged streamline topology down-
wake. He also concluded that when two parallel vortex
stream of the base of the cylinder takes on a form
streets are in anti-phase; two different vortex streets are
known as an owl face of the first kind, which was orig-
formed. When there is in-phase oscillation of two vor-
inally defined for a completely different exterior flow.
tex streets, a large-scale vortex street occurs. Wei and
Immediately adjacent to the base of the cylinder, an
Chang [18] found that the characteristics of the wake
additional system of saddle points (Sp) is located at
which occurred behind side-by-side arranged cylinders
either end of a nodal line. At locations above the bed,
one of the two principal saddle points of the owl face are significantly related to the cross-sectional geome-
of the first kind disappears and the principal foci (F) tries of the bluff-body. The gap flow tends to deflect
are transformed from a stable to an unstable state. toward the narrow wake side downstream of the circu-
It is observed that when more than one cylinders are lar cylinders. The biasing characteristic of the gap flow
located in a fluid flow, the vortex shedding pattern is is relatively bistable. The deflected gap flow may
completely different than that of a single cylinder. The change over intermittently from one side to another.
applications of side-by-side arrangements of two and Zhou et al. [19] proposed that the frequency of the nar-
more cylinders can be found in many branches of row wake is approximately three times larger than that
engineering such as heat exchangers, cooling tower of lower frequency of the wide wake.
arrays, fuel and cooling systems for nuclear power The unsteady structure of the flow behind side-by-
plants, offshore structures, bridges, chimneys, groups of side arrangement of circular cylinders in close proxim-
tall buildings and power transmission lines. ity has not been investigated in detail in shallow water
When more than one cylinder is placed in a uniform flows. The aim of the present work is to obtain quali-
flow, the flow characteristics are completely different tative and quantitative information on the downstream
from the case of a single cylinder depending on the flow characteristics of side-by-side arrangement of
arrangement or spacing of the cylinders. There are cylinders with equal diameter in shallow water flow
numerous studies in the literature concerning the flow using the high-image-density PIV technique. This tech-
around side-by-side circular cylinders in deep-water nique provides patterns of average and instantaneous
flows. Flow around side-by-side circular cylinders in representations of the flow structure such as velocity,
deep flow has been investigated by Ishigai et al. [8], vorticity, streamline topology, Reynolds stress and
Bearman and Wadcock [9], Kamemoto [10], Kiya et al. root-mean-square fluctuations. The experiments were
[11], Williamson [12], Jendrzejczyk and Chen [13], Kim carried out in the range 1:0  G=D  3:0 with an
and Durbin [14], Le Gal et al. [15] and Sumner et al. increment of 0.25. The laser sheet was located parallel
 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197 189

to the bottom of the water channel at the mid-section propagate through them. Cylinders were located at a
of the water height. distance of 5500 mm from the leading edge of the test
section in order to maintain fully developed turbulent
2. Experimental system and technique boundary layer. The depth of the water was adjusted to
hw ¼ 25 mm for all experiments using a platform with
Experiments were undertaken in a low turbulence a height of hp ¼ 600 mm located in the water channel.
closed-loop open-surface water channel test section The solid blockage ratio of each cylinder was 2.5%,
having the following dimensions: a length of 8000 mm, having a maximum blockage of 7.5% for three-cylinder
width of 1000 mm, and depth of 750 mm. This facility case. Configurations of just a single cylinder, two cylin-
was built for PIV experiments for steady and unsteady ders and three cylinders were tested. The gap distances
flows. Longitudinal free-stream turbulence intensity is between the cylinder centers for these configurations
less than 0.5%. A pump driven by an electric motor were changed from G=D ¼ 1:0 3:0 with an increment
having a variable speed controller was used to create a of 0.25. Experiments were carried out at a free-stream
desired mean flow in the test section. Water was velocity U1 of 0.162 m/s, yielding Reynolds number
pumped into a settling chamber through a honeycomb based on the cylinder diameter of Re ¼ 5000. Stability
section. The water was then conveyed through a two- parameter S was calculated as 0.0082. The friction
dimensional two-to-one contraction to reach the test coefficient of the water table Cf was estimated from the
section which was constructed of 15 mm thick clear standard smooth-wall relation as Cf ¼ 0:0068.
Plexiglas sheet. The overall length of the water channel Using the PIV technique, instantaneous velocity vec-
is approximately 14000 mm. An overview of the shal- tors can be measured in a region illuminated by a two-
low water experimental system is shown in Fig. 1. The dimensional laser sheet. Velocity vector measurements
cylinder models having a diameter of D ¼ 30 mm were have been carried out using Dantec PIV system. The
constructed from Plexiglas so that the laser light could flow field illumination was provided by two Nd:Yag

Fig. 1. Schematic of the experimental system and definition of the parameters for side-by-side arrangement of cylinders.
190 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197

pulsed laser sources of a wavelength of 532 nm, each 3. Results and discussions
with a maximum energy output of 120 mJ. Dantec
FlowMap Processor which controlled the timing of the The quantitative dye images of the near-wake region
data acquisition was used to synchronize the image of two side-by-side circular cylinders are shown in
taking and laser units. The thickness of the laser sheet Fig. 2 for different gap ratios. As seen in Fig. 2a, when
illuminating the flow field was approximately 1.5 mm. two side-by-side circular cylinders are in contact with
PIV measurements were taken in the mid-plane elev- each other, Kármán vortex shedding was observed as a
ation. The water was seeded with the neutrally buoyant whole, rather than seen in the case of the individual
spherical particles of 12 lm in diameter. The move- cylinders, and vortex shedding frequency in the wake is
ments of the particles were recorded using a CCD cam- about half of the average of the single cylinder case.
era with a resolution of 1024  1024 pixels. The camera When the gap ratio is varied between 1 and 1.15, the
was equipped with a 60 mm focal-length lens. Dantec effect of the jet-like flow between cylinders is negligibly
weak. Therefore, a single large-scale Kármán vortex
flow grabber digital PIV software employing frame-to-
street is still observed up to G=D ¼ 1:15. As the gap
frame cross-correlation technique was used to calculate
ratio increases, the jet-like flow between the two cylin-
the raw displacement vector field from the particle
ders becomes effective. This jet-like flow interacts with
image velocity data. In the image processing, 32  32
the vortex structure shed from the two cylinders and
pixels rectangular effective interrogation windows were
causes bistable behavior of the near-wake flow. Figs. 2b
used. During the interrogation process, an overlap of
and c show the asymmetric flow pattern for
50% was employed in order to satisfy the Nyquist cri-
G=D ¼ 1:25. The gap flow is deflected to one side
terion. Erroneous vectors were removed (less than 1%)
(toward the lower cylinder as seen in Fig. 2b), which is
and replaced by using bilinear interpolation between
in agreement with the observations of Kim and Durbin
surrounding vectors in the post-processing step. Velo-
[14], Bearman and Wadcock [9], and Guillaume and La
city vectors of the flow were calculated from this dis-
Rue [21] in deep-water flows. However, this deflected
placement vector field. The vorticity patterns of the
flow was not observed on the same side (toward the
wake flow were determined from the velocity field upper cylinder as seen in Fig. 2c) all the time. The jet-
using a finite difference scheme. Streamlines and circu- like flow sometimes moves towards the upper cylinder
lation were obtained by post-processing of the velocity and sometimes moves towards the lower cylinder.
data. For all of the experiments, the image magnifi- However, this bistable nature of the wake flow is not
cation was 1:5.5 which yields an effective grid size of periodic, it is completely arbitrary. Bistable flow
2:91  2:91 mm in physical plane of the laser sheet, characteristic is not caused by the misalignment of the
corresponding to 3844 velocity vectors. The overall cylinders or other external influences but is an intrinsic
field of view was 178  178 mm (5:93  5:93 cylinder property of the flow [22]. Any perturbation on the flow
diameters). Patterns of instantaneous particle images field or change in flow conditions might cause this
(total of 200 images for a continuous series) were taken switching behavior of the flow. The flow visualization
at the rate of 15 Hz, thereby spanning 13.4 s. Averaged shown in Fig. 2b indicates a wider wake flow region
patterns of the flow structure were calculated from all downstream of the upper cylinder, which corresponds
of the instantaneous images. The time delay between to smaller vortex shedding frequency and Fig. 2c shows
pulses ranged from 1.5 to 2 ms. The velocity vector a wider wake flow region downstream of the lower cyl-
field, vorticity contours, streamline topology and other inder which corresponds to a higher vortex shedding
turbulent statistics were computed. In terms of uncer- frequency. Approximately, two cylinder diameters
tainty of velocity measurements using the PIV tech- downstream from the base of the cylinders, both wake
nique, the digital PIV approach carried out during the regions merge together to form a wider but weaker
experiments is similar to that employed by Westerweel wake flow region. A similar flow behavior was
[20] who concluded that an uncertainty in velocity is observed for G=D ¼ 1:5 but, having smaller deflection
less than 2%. angle, which is the angle between the directions of free-
Flow visualization experiments using a dye were car- stream and jet flow, of the jet-like flow.
ried out in order to reveal the flow characteristics of This kind of deflected flow is observed until G=D ¼
two and three side-by-side cylinder arrangements. The 2:0 after which the wake regions of these cylinders are
fluid motion was recorded with a digital video camera not affected from each other and from the jet-like flow
(Sony DCR TRV 355E). The images were processed between them. The flow visualization around side-by-
with a frame grabber and then these images were side cylinders having a gap ratio of 2.5 is shown in
enhanced for analysis using Adobe Photoshop soft- Fig. 2d. Two distinct coherent vortex streets are
ware. Each cylinder had three dye injection holes each observed downstream of the each cylinders which are
of which having 1 mm diameter located at the mid- not virtually different from that of the flow of single
span and near the points of boundary layer separation. cylinder. Therefore, shedding frequency of these two
 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197 191

Fig. 2. Flow visualization of two side-by-side arrangement of circular cylinders.

vortex structures are found to be almost the same and downstream of the outer cylinders as seen in Fig. 3d.
equal to the shedding frequency of single cylinder. The effect of wake regions on one another is not
In case of three side-by-side cylinders, different vor- observed at this gap ratio and also at higher gap ratios.
tex shedding patterns are observed. When three cylin- The frequency of vortex shedding of each cylinder is
ders are touching one another, a coupled vortex street equal to the frequency of single cylinder.
is observed as seen in Fig. 3a. Vortex shedding fre- The variation of non-dimensional vortex shedding
quency of the wake is approximately equal to 0.3 frequency (Strouhal number) for two side-by-side cylin-
which is one-fourth of a single cylinder case. Flow der arrangements is shown in Fig. 4 as a function of
visualizations for G=D ¼ 1:25 indicate that there are the gap ratio. The Strouhal number data in Fig. 4
two different flow patterns as shown in Figs. 3b and c. show that two dominant vortex shedding frequencies
Fig. 3b indicates that a wider wake flow region with are detected at the same gap ratio which corresponds
a lower shedding frequency occurs downstream of to two different Strouhal numbers. The smaller fre-
the central cylinder and two narrow wake regions quency corresponds to the vortex shedding frequency
downstream of the outer cylinders accompany this of wider wake flow region and the bigger frequency
wider wake which is symmetrical about the centerline depicts the smaller wake’s frequency. For G=D ¼ 1:25,
of the middle cylinder. The gap flows on both sides of the vortex shedding frequency of smaller wake is one-
the central cylinder are directed towards the outer third of the wider wake. The ratio of the vortex shed-
cylinders. Deflection angles of the biased flow are ding frequencies of narrow and wider wake decreases
nearly identical. For the second flow pattern, two wider when the gap ratio increases up to a value of 2.25.
wake regions downstream of the outer cylinders domi- Since the jet flow does not affect the vortex shedding
nate the flow. However, the wake region of one of the patterns of both cylinders for the gap ratio of 2.25, the
cylinders is sometimes larger than that of the other shedding frequency of the cylinders is the same.
one. Similar flow patterns were also observed for the In order to determine detailed information on the
gap ratio of G=D ¼ 1:5 with smaller deflection angles flow by obtaining instantaneous flow patterns of velo-
of the biased flow around the central cylinder. Both city, vorticity and streamlines, the PIV technique was
PIV measurements and dye visualization experiments applied to the flow field. By averaging instantaneous
show that deflection angle decreases with an increase in images of velocity, the time-averaged velocity field and
the gap ratio. When the gap ratio is adjusted to related time-averaged vorticity and streamline topology
G=D ¼ 2:5, Kármán type vortex shedding is observed were obtained.
192 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197

Fig. 3. Flow visualization of three side-by-side arrangement of circular cylinders.

Comparisons of the patterns of the averaged velocity by-side cylinder case. Significant reverse velocities
and corresponding patterns of the averaged stream- occur in the wake region and the magnitude of the
lines, vorticity and Reynolds stress for one, two and reverse velocities increases with the number of cylin-
three side-by-side circular cylinders in contact with one ders. Both time-averaged velocity and streamline top-
another are presented in Fig. 5. The development of ology indicate that two foci (F1 and F2) and one
the flow fields for two and three cylinders exhibit simi- saddle point (Sp) are evident for all the three cases.
larity to that of a single cylinder. The averaged velocity These two foci and the saddle points are indicated in
patterns shown in Fig. 5a indicate that single wake Fig. 5b. Streamline topology also shows the size of the
region occurs downstream of each cylinder. Vortex wake flow region downstream of the cylinder arrange-
shedding frequency of single cylinder is approximately ments. Averaged vorticity contours are shown in
two times greater than that of two cylinder arrange- Fig. 5c. Positive vorticity (anticlockwise) and negative
ment, and three times greater than that of three side- vorticity (clockwise) layers are represented by solid
lines and dashed lines, respectively. For single cylinder,
the transverse spatial extent of wake region remains
relatively narrow. On the other hand, for the cases of
two and three cylinders, each vorticity layer is broadly
distributed in the transverse direction and moreover,
the central region of each layer is deflected towards the
plane of symmetry. The peak value of vorticity concen-
tration is approximately the same for all the three
cases. Fig. 5d presents the contours of dimensionless
2
Reynolds stress (normalized with U1 in which U1 is
the free-stream velocity) at the mid-plane level of the
shallow water layer. The separated shear layers, which
also correspond to vortex centerlines, from the cylin-
ders show a very high level of Reynolds stress. The
peak value of Reynolds stress is located in the reversed
Fig. 4. Variation of Strouhal number with the gap ratio for two flow region downstream of the cylinders and the
side-by-side arrangement of cylinders. location of extrema of Reynolds stress is displaced on
 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197 193

Fig. 5. (a) Time-averaged velocity vector field downstream of the cylinders for G=D ¼ 1:0 in the plan view, (b) corresponding time-averaged
streamline topology, (c) patterns of averaged vorticity contours. Dashed and solid lines correspond to negative and positive vorticity, respectively.
Minimum and incremental values of vorticity are xmin ¼ 4 and Dx ¼ 2 s1 , (d) contours of normalized Reynolds stress. Dashed and solid lines
correspond to negative and positive Reynolds stress, respectively. Minimum and incremental values of Reynolds stress are ½hu0 v0 i=U1 2
min ¼ 0:01
and D½hu0 v0 i=U1
2
¼ 0:01.

the same streamwise distance with saddle point associa- side-by-side circular cylinder configurations with
ted with the generation of intensive fluctuations in this G=D ¼ 1:25, 1.5, 2.0 and 2.5 are shown in Fig. 6. Aver-
location for all cases. In terms of peak concentration aged velocity vector image presented in Fig. 6a shows
values for different cases, due to the stronger rotational that the flow structure of the wake is dominated by the
fluid motion in the separated layer, the peak value of strong jet-like flow between two cylinders for the cases
Reynolds stress for three-cylinder case is slightly higher of G=D ¼ 1:25 and 1.5. For G=D ¼ 1:25, separate vor-
than that of two-cylinder case and approximately four tex streets are observed downstream of each cylinder.
times higher than that of single cylinder case. Beside Actually, this kind of flow behavior starts to occur for
the well-defined patterns of Reynolds stress for two the gap ratio higher than 1.15. Despite the geometrical
and three-cylinder cases, additional but weaker con- symmetry, the jet-like flow for G=D ¼ 1:25 and 1.5
tours are observed closer to the base of the cylinders cases is deflected towards the lower cylinder side which
due to sweeping of the vorticity layers across this has a narrow wake region and higher vortex shedding
region of the flow during formation of the large-scale frequency. However, upper cylinder has a wider wake
vortical structures, which are evident in the corre- region and lower vortex shedding frequency. The wake
sponding patterns of averaged velocity [23]. As a result region of upper cylinder is much longer than that of
of wider wake region of the three-cylinder case, con- the lower cylinder and that of two-cylinder case with
tours of Reynolds stress are broadly distributed in G=D ¼ 1 arrangement. The deflection angle and the
transverse direction. streamwise extent of the wake region reduce as the gap
Averaged velocity vectors, streamline, vorticity and ratio enlarges. As mentioned previously, the bistable
Reynolds stress contours corresponding to the two nature of the flow is observed during the PIV measure-
194 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197

Fig. 6. (a) Time-averaged velocity vector field at the mid-plane for two side-by-side circular cylinders for different gap ratios in the plan view,
(b) corresponding time-averaged streamline topology, (c) patterns of averaged vorticity contours. Dashed and solid lines correspond to negative
and positive vorticity, respectively. Minimum and incremental values of vorticity are xmin ¼ 4 and Dx ¼ 2 s1 , (d) contours of normalized
Reynolds stress. Dashed and solid lines correspond to negative and positive Reynolds stress, respectively. Minimum and incremental values of
Reynolds stress are ½hu0 v0 i=U1
2 0 0 2
min ¼ 0:005 and D½hu v i=U1 ¼ 0:005.

ments and flow visualization observations. The flow also indicate that as the gap ratio increases, the size of
does not remain deflected towards the same cylinder the wake regions downstream of the cylinders decreases
side at all times. Sometimes, deflection occurs towards to the size of the wake region of single cylinder case.
the upper cylinder side without changing the alignment Averaged streamline and vorticity contours for differ-
of the cylinders or the water flow speed or any other ent gap ratios are depicted in Figs. 6b and c. The jet-
parameter that could affect the flow structure. For like flow significantly affects the vortex shedding beha-
greater gap ratios (G=D > 2:0), the effect of jet-like vior downstream of the cylinder arrangements for
flow is not observed and the cylinders behave as inde- ratios smaller than 2.0. Elongated negative (dashed
pendent and isolated cylinders. However, some interac- lines) vorticity layers are evident downstream of the
tions are observed between the two cylinder wake upper cylinder for both G=D ¼ 1:25 and 1.5 cases. Due
regions for the gap ratio of 2.0. This interaction is to the jet-like flow and interactions of wake regions
clearly seen from the Reynolds stress contours in with each other, the streamwise extend of both positive
Fig. 6c for G=D ¼ 2:0. Averaged velocity vector fields vorticity layers of upper cylinder and negative vorticity
 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197 195

layers of lower cylinder are shorter and both of them tance between separated vorticity layers) is at least two
are deflected toward the lower cylinder side. The peak times larger than the width of the wake region of single
vorticity levels of these inner layers and the outer lay- cylinder. The separating shear layers from the outer
ers are approximately the same. On the other hand, cylinders are remarkably deflected towards the wake of
vorticity contours of cylinder arrangements of G=D ¼ the central cylinder causing significantly narrow wake
2:0 and 2.5 are remarkably similar to the vorticity con- flow region downstream of the central cylinder. Deflec-
tours downstream of the single cylinder. Well-defined tion angle of the jet-like flow between the upper cylin-
Kármán vortex streets are obtained and the vortex der and the central cylinder is higher than the jet-like
shedding frequencies of both wakes are equal to the flow between the lower cylinder and the central cylin-
shedding frequency of single cylinder. Both in-phase der. When the gap ratio is increased to a value of
and anti-phase vortex sheddings are observed at these G=D ¼ 1:5, symmetrical flow patterns, wide wake
gap ratios. Images shown in Fig. 6d provide contours regions of approximately equal size behind the upper
of constant Reynolds stress as a function of gap ratio and the lower cylinders and narrow wake region down-
between two cylinders. Consider, first of all, the pat- stream of the central cylinder, are obtained. The size of
terns of Reynolds stress for G=D ¼ 2:5. Well-defined the wake regions which appear downstream of the cen-
concentrations of Reynolds stress occur downstream of tral cylinders of G=D ¼ 1:25 and 1.5 are nearly equal
both cylinders, which show no changes in the flow field to each other. Bistable flow structure was also observed
at this gap ratio. The observation of Reynolds stress from PIV measurements and flow visualizations at this
pattern for G=D ¼ 2:0 suggests that there is a small gap ratio (G=D ¼ 1:5). A large wake flow region down-
modification of the flow structure. Well-defined Rey- stream of the central cylinder and narrow wake regions
nolds stress concentrations occur downstream of the behind outer cylinders are obtained. The jet-like flow
upper cylinder. However, the peak concentration of the between cylinders is biased towards the outer cylinders.
Reynolds stress contours (especially negative contours Symmetric biased flow is also obtained for the gap
represented with dashed lines) of the lower cylinder is ratio of G=D ¼ 2:0. A similar wake structure occurs
slightly smaller compared with the peak concentrations downstream of the cylinders compared to the case of
of upper cylinder. Decreasing values of gap ratio from G=D ¼ 1:5. However, the effect of jet-like flow on the
G=D ¼ 2:0 to G=D ¼ 1:25, negative concentrations of wake region downstream of the cylinders is less pro-
Reynolds stress downstream of both cylinders nounced relative to the cases of smaller gap ratios. The
especially for the concentrations at the upper cylinder reattachment length of the wake behind the central cyl-
gradually disappear and they are barely detectable for inder is longer than that of G=D ¼ 1:5 case. It can be
G=D ¼ 1:25 arrangement as a result of the direction of concluded for G=D ¼ 2:0 arrangement that jet-like flow
jet-like flow in that region. The peak concentration of has an important effect on the wake region for three
positive Reynolds stress occurs downstream of the side-by-side arrangements compared to the two cylinder
lower cylinder. The patterns of Reynolds stress close to arrangement. At sufficiently large G=D (G=D ¼ 2:5),
the base of the cylinders are not evident. The location well-defined vortex sheddings downstream of the three
of the peak values of positive Reynolds stress of upper cylinders occur with the same vortex shedding
cylinder is approximately at the same level for all gap frequency.
ratios. Corresponding patterns of average vorticity pre-
Averaged velocity vectors, streamline patterns, vor- sented in Fig. 7c show further features of the wake
ticity contours and Reynolds stresses are shown in regions. An asymmetrical vorticity pattern is evident
Fig. 7 for the case of three side-by-side cylinders hav- for the smallest gap ratio (G=D ¼ 1:25). Because of
ing the gap ratios as G=D ¼ 1:25, 1.5, 2.0 and 2.5. decreased gap distance between the upper and the cen-
Despite the geometrical symmetry, images presented in tral cylinders, an elongated negative concentrations
Figs. 7a and b showing the average velocity vector field shed from the upper surface of the upper cylinder
and streamline topology indicate asymmetrical flow occur in the case of G=D ¼ 1:25. Positive vorticity con-
pattern downstream of the cylinders for the case of tours of the outer cylinders are relatively small and are
G=D ¼ 1:25. A wider wake flow region downstream of deflected toward the wake region of the central cylin-
the upper cylinder and an intermediate wake down- der. Both positive and negative vorticity contours shed
stream of the lower cylinder occur and a significantly from the central cylinder are deflected toward the plane
narrow wake flow region occurs downstream of the of symmetry. When the gap ratio is increased to a
central cylinder. However, both PIV measurements and value of G=D ¼ 1:5, deflection angle decreases and
flow visualizations show that the region of the wider negative and positive vorticity of outer cylinders move
wake is so expanded in the streamwise direction that toward the center of symmetry of each cylinder. A
the reattachment point of the wake region is out of similar flow structure is obtained for G=D ¼ 2:0. How-
field of view of the PIV measurements. Moreover, the ever, for this case, the streamwise extent of the vor-
width of the wider wake flow region (transverse dis- ticity contours of the central cylinder is longer
196 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197

Fig. 7. (a) Time-averaged velocity vector field at the mid-plane for three side-by-side circular cylinders for different gap ratios in the plan view,
(b) corresponding time-averaged streamline topology, (c) patterns of averaged vorticity contours. Dashed and solid lines correspond to negative
and positive vorticity, respectively. Minimum and incremental values of vorticity are xmin ¼ 4 and Dx ¼ 2 s1 , (d) contours of normalized
Reynolds stress. Dashed and solid lines correspond to negative and positive Reynolds stress, respectively. Minimum and incremental values of
Reynolds stress are ½hu0 v0 i=U1
2 0 0 2
min ¼ 0:005 and D½hu v i=U1 ¼ 0:005.

compared to the smaller gap ratio cases. Finally, wake flow region of the upper cylinder which has a
Kármán vortex shedding is observed behind the cylin- wider wake flow region, less concentrated Reynolds
ders for G=D ¼ 2:5 case. stress contours which combine with the negative con-
Patterns of Reynolds stress for different gap ratios tours of the central cylinder are observed. That is why,
are given in Fig. 7d. Concerning G=D ¼ 1:25 case, the negative Reynolds stress contours of the central
negative Reynolds stress contours of the upper cylinder cylinder appear to be broadly distributed and elon-
are not detectable because of the direction of the jet- gated both in streamwise and transverse direction. As
like flow between the upper and the central cylinder. the effect of jet-like flow decreases with increasing gap
As the saddle point of the upper cylinder moves far ratio, negative Reynolds stress contours of the upper
away from the cylinder, the location of the peak cylinder are detectable. Fig. 7d also indicates that
concentration of the positive Reynolds stress moves Reynolds stress contours of the central cylinder are
further away from the base of the cylinder. Since the elongated in streamwise direction when the gap ratio is
jet-like flow does not have a remarkable effect on the increased.
 H. Akilli et al. / Flow Measurement and Instrumentation 15 (2004) 187–197 197

4. Conclusions [2] E. Wolanski, J. Imberger, M.L. Heron, Island wakes in shallow

coastal water, Journal of Geophysical Research 89 (1984)
The effect of gap ratio between side-by-side arranged 10553–10569.
[3] P.M. Lloyd, P.K. Stansby, D. Chen, Wake formation around
circular cylinders on the wake flows downstream of
islands in oscillatory laminar shallow-water flows. Part 1.
two and three cylinders of equal diameter in shallow Experimental investigation, Journal of Fluid Mechanics 429
water was characterized using PIV technique and flow (2001) 217–238.
visualizations. Time-average patterns of velocity vec- [4] A.A. Kolyshkin, M.S. Ghidaoui, Stability analysis of shallow
tors, vorticity and Reynolds stress contours and wake flows, Journal of Fluid Mechanics 494 (2003) 355–377.
[5] D. Chen, H. Jirka, Experimental study of plane turbulent wakes
streamline topology, obtained from PIV measurements, in a shallow water layer, Fluid Dynamics Research 16 (1995)
were used to explain the physical mechanisms of the 11–41.
flow downstream of the two and three side-by-side cir- [6] P.M. Lloyd, P.K. Stansby, Shallow-water flow around model
cular cylinder arrangements. Jet-like flow between conical island of small side slope. I: surface piercing, Journal of
Hydraulic Engineering 123 (1997) 1057–1067.
cylinders was remarkably effective for small gap ratios
[7] H. Akilli, D. Rockwell, Vortex formation from a cylinder in
(G=D  1:5) especially for the gap ratio of G=D ¼ 1:25. shallow water, Physics of Fluids 14 (2002) 2957–2967.
It had a minor influence at moderate gap ratios [8] S. Ishigai, E. Nishikawa, K. Nishimura, K. Cho, Experimental
(1:5 <G=D  2:0). For larger gap ratios (G=D 2:5), study of structures of gas flow in tube banks with tube axes nor-
the effect of jet-like flow on the flow characteristics mal to flow (Part I, Kármán vortex flow from two tubes at vari-
ous spacing), Bulletin of the JSME 15 (1972) 949–956.
behind the cylinders was negligible. Despite the sym-
[9] P.W. Bearman, A.J. Wadcock, The interaction between a pair of
metrical geometry, asymmetrical flow structures were circular cylinders normal to a stream, Journal of Fluid Mech-
obtained for different arrangement of the cylinders anics 61 (1973) 499–511.
especially at smaller gap ratios. At G=D ¼ 1:25 2:0, the [10] K. Kamemoto, Formation and interaction of two parallel vortex
biased flow created a narrower wake flow region with a streets, Bulletin of the JSME 19 (1976) 283–290.
[11] M. Kiya, M. Arie, H. Tamura, H. Mori, Vortex shedding from
higher vortex shedding frequency and a wider wake two circular cylinders in staggered arrangement, Transactions of
flow region with a lower vortex shedding frequency for ASME, Journal of Fluids Engineering 102 (1980) 166–173.
two-cylinder case. In the case of three side-by-side [12] C.H.K. Williamson, Evolution of a single wake behind a pair of
cylinders, asymmetric flow structure was observed only bluff bodies, Journal of Fluid Mechanics 159 (1985) 1–18.
[13] J.A. Jendrzejczyk, S.S. Chen, Fluid forces on two circular cylin-
at the gap ratio of G=D ¼ 1:25. Symmetrical flow
ders in cross flow, in: S.S. Chen, J.C. Simonis, Y.S. Shin (Eds.)-
occurred for higher gap ratios (G=D 1:5). Flow-Induced Vibration PVP 104 Chicago, IL, 1986, pp. 1–13.
Bistable nature of the flow was observed especially [14] H.J. Kim, P.A. Durbin, Investigation of the flow between a pair
for the gap ratio of G=D ¼ 1:25, both for two and of circular cylinders in the flopping regime, Journal of Fluid
three side-by-side cylinders arrangements. Without Mechanics 196 (1988) 431–448.
[15] P. Le Gal, M.P. Chauve, R. Lima, J. Rezende, Coupled waked
changing any experimental conditions, the jet-like flow behind two circular cylinders, Physical Review A 41 (1990)
changed direction and switched from one side to the 4566–4569.
other. For example, wider wake region occurred down- [16] D. Sumner, S. Wong, S.J. Price, M.P. Paidoussis, Two and three
stream of upper cylinder most of the time for two-cyl- side-by-side circular cylinders in steady cross-flow, Proceedings
inder case, but, sometimes wider wake occurred of the 16th Canadian Congress of Applied Mechanics, vol. 1,
1997, pp. 273–274.
downstream of the lower cylinder. Similar phenomenon [17] H.M. Spivack, Vortex frequency and flow pattern in the wake of
was observed for three-cylinder case. However, this bis- two parallel cylinders at varied spacing normal to an air stream,
table nature of the flow was not observed for the gap Journal of the Aeronautical Science 13 (1946) 289–297.
ratios bigger than 1.5. [18] C.Y. Wei, J.R. Chang, Wake and base-bleed flow downstream of
bluff bodies with different geometry, Experimental Thermal and
Finally, negative Reynolds stress concentrations were
Fluid Science 26 (2002) 39–52.
barely detectable downstream of the cylinder which [19] Y. Zhou, Z.J. Wang, R.M.C. So, S.J. Xu, W. Jin, Free vibra-
had a wider wake flow region at the gap ratio of G=D tions of two side-by-side cylinders in a cross-flow, Journal of
¼ 1:25 for both two and three cylinders arrangements. Fluid Mechanics 443 (2001) 197–229.
Another important point is that the peak concentra- [20] J. Westerweel, Digital Particle Image Velocimetry, Theory and
Application, Delft University Press, 1993.
tions of Reynolds stress occurred very close to the sad- [21] D.W. Guillaume, J.C. La Rue, Investigation of the flopping
dle points. regime with two-, three- and four-cylinder arrays, Experiments
in Fluids 27 (1999) 145–156.
[22] D. Sumner, S.S.T. Wang, S.J. Price, M.P. Paidoussis, Fluid
References Behavior of side-by-side circular cylinders in steady cross-flow,
Journal of Fluids and Structures 13 (1999) 309–338.
[1] R.G. Ingram, V.H. Chu, Flow around island in Rupert Bay: an [23] A. Kahraman, B. Sahin, D. Rockwell, Control of vortex forma-
investigation of the bottom friction effect, Journal of Geophysi- tion from a vertical cylinder in shallow water: effect of localized
cal Research 92 (1987) 14521–14533. roughness elements, Experiments in Fluids 33 (2002) 54–65.