Flow Measurement and Instrumentation 17 (2006) 225–235

www.elsevier.com/locate/flowmeasinst

Flow structure in the downstream of square and circular cylinders

Muammer Ozgoren ∗
Selcuk University, Faculty of Engineering and Architecture, Department of Mechanical Engineering, 42070, Konya, Turkey

Received 10 March 2005; received in revised form 18 July 2005; accepted 21 November 2005

Abstract

In this experimental work, a technique of digital particle image velocimetry (DPIV) is employed to characterize instantaneous vorticity and
time-averaged velocity, vorticity, root mean square (rms) velocities, Reynolds stress correlations and phase-averaged contours in the downstream
of circular, sharp-edged square and 45◦ orientated square cylinders in a uniform flow. Strouhal numbers for 550 ≤ Re ≤ 3400 are calculated from
wake flow patterns. Shear layers surrounding the recirculation bubble region behind the cylinder are discussed in terms of flow physics and vortex
formation lengths of large-scale Kármán vortices. Enhancement levels of Reynolds stress correlations associated with cross-stream velocity are
clarified. Finally, flow structures depending on the cylinder geometry and Reynolds number are interpreted with quantitative representations.
c 2005 Elsevier Ltd. All rights reserved.

Keywords: Square cylinder; Circular cylinder; DPIV; Vortex shedding; Wake region; Strouhal number

1. Introduction behind the bluff bodies can cause structural damage as a result
of periodic surface loading, acoustic noise and drag forces [1,2].
The motion of fluid passing circular or square cross-section Most work has been done on the flow passing a circular cylinder
cylinders is encountered in numerous industrial applications, (CC) rather than a square cylinder (SC). Even though their near-
as well as offshore and environmental settings, including tall wake flow structures are expected to be topologically similar
buildings and structures such as bridges, chimneys, trash racks, to one another, the reasons for flow separation on the cylinder
cooling towers, heat exchangers, etc. When fluid flows around surfaces are totally different. That is, flow separation occurs due
any bluff body, the vortices are alternatively shed and a well- to an adverse pressure gradient in the downstream direction
known Kármán vortex street is formed in the wake of the for the CC and the separation points on the CC surface move
cylinder. The oscillating wake rolls up into two staggered rows back and forth depending on the Reynolds number. However,
of vortices with opposite senses of rotation. The frequency the locations of the separation points are fixed at the upstream
of vorticity pairs is a function of velocity, cylinder diameter corners of the SC due to the abrupt geometrical changes.
and Reynolds number [1]. The vortex shedding frequency Relevant investigations for the SC and the CC include those
of the circular cylinder is normalized by Strouhal number, of Okajima [3], Obasaju [4], Durao et al. [5], Lyn et al. [6],
St = f s D/U∞ , where f s is the dominant vortex-shedding Saha et al. [7], Dutta et al. [8], Chen et al. [9], Sarioglu and
frequency, D is the cylinder diameter and U∞ is the freestream Yavuz [10] and Roshko [11] using measurement techniques
flow velocity. The vortex shedding frequencies and flow such as dye visualization, point-wise hot-wire/film and laser
configurations are major issues for the design of structures. The Doppler velocimetry (LDV) for various Reynolds numbers.
flow in the near-wake region behind the cylinder, incorporating Okajima [3] carried out an experimental study of flow past the
the vortex formation region, plays an important role in SC as well as the rectangular cylinder for 70 ≤ Re ≤ 20,000 to
determining the steady and unsteady forces acting on the body. determine the vortex shedding frequencies. The results showed
Periodic vortex shedding patterns and fluctuating velocity fields that there was an abrupt change in Strouhal number when
the aspect ratio of the cylinder was reduced to the range
∗ Tel.: +90 332 223 1912 or cell phone: +90 555 451 70 80; fax: +90 332 241 2–3. Strouhal numbers for the SC with an increasing angle of
06 35 or +90 332 241 06 51. incidence from 0◦ to 45◦ were examined using a hot-wire probe
E-mail addresses: mozgoren@selcuk.edu.tr, mbozgoren@yahoo.com. in a closed circuit wind-tunnel for Re = 10,000 by Obasaju [4].

0955-5986/$ - see front matter 

c 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.flowmeasinst.2005.11.005
226 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235

LDV measurements were conducted for flow past the SC in a develop reliable computational methods, numerical simulation
water tunnel for Re = 14,000 and Re = 21,400 [5,6]. Durao studies of 2-D and 3-D unsteady flows around the SC have
et al. [5] separated the periodic and random components of been addressed [19–24]. Reviews and assessments of concepts,
velocity fluctuations around the SC. The experiment results of numerical simulations, and experimental observations for bluff
Lyn et al. [6] showed a relationship between the flow topology bodies are given by Williamson [1], Ahlborn et al. [25],
and turbulence distribution. They clearly distinguished vorticity Knisely [26] and King [27]. Although many of the above
saddles and streamline saddles. Measurements were reported investigations are experimental and based on flow visualization
of two components of velocity in the wake for Reynolds and/or point-wise measurement techniques such as LDV or hot-
numbers of the SC such as 1340, 4990 and 9980 for various wire anemometry to obtain instantaneous flow characteristics,
inclinations such as 0◦ , 22.5◦, 30◦ , 45◦, and 60◦ by using smoke a few issues have remained unresolved for this class of near-
visualization and a hot-wire anemometer [7,8]. Saha et al. [7] wake flows. Specifically, none of the foregoing investigations
stated that, though the overall flow past the SC resembled that of attempted to measure quantitatively the details of the separated
the CC, there were major differences in terms of the separation shear layers and the overall properties of the flow in separated
mechanism and the related integral parameters such as Strouhal regions for the SC using DPIV.
number, lift and drag-coefficient. Vortex-shedding frequencies The purpose of this study is to investigate the effects
and surface pressures of the SC, CC and rectangular cylinders in of the cross-section of the vertical cylinder perpendicular to
a wind tunnel were investigated experimentally using hot-film the oncoming flow and the Reynolds number in terms of
measurement techniques [9,10]. They determined the dominant instantaneous and time-averaged velocity, vorticity, streamline
peak location of the vortex shedding frequency and obtained topology, and Reynolds stress correlations along with root
Strouhal numbers in the range of various Reynolds numbers mean square (rms) velocity. First of all, Strouhal numbers are
for the CC, SC and oriented square cylinder (OSC). The calculated for the sake of comparison. The physical origins of
various aspects of the quasi two-dimensional (2-D) and three- loading coefficients and flow-structure interactions have been
dimensional (3-D) structures of the shear layer vortices, as well interpreted. Phase-averaged vorticity and velocity patterns are
as the correlation of their frequency with Reynolds number for examined. Moreover, the turbulence properties of the wake,
the SC, were also given by Roshko [11]. most importantly the Reynolds stress correlations, have been
In recent decades, flow structures around CC has been addressed in relation to Reynolds numbers for the SC and CC.
investigated quantitatively by Lin et al. [12]. They employed Finally, interpretation of the effect of the Reynolds number on
the film based particle image velocimetry technique at three the vortex formation length and width of the wake size has been
values of Re = 1000, 5000 and 10,000 for the CC to obtain determined quantitatively. Informative data for both numerical
instantaneous flow patterns. They stated that representations of simulations and designers have been provided.
velocity, vorticity and streamlines showed a decrease in vortex
formation length with increasing Reynolds number. Noca et al. 2. Experimental system and technique
[13] accomplished an investigation on the concept of vortex
formation length in the wake of the CC by measuring the The experiments were performed in a closed-loop free-
location of some easily identifiable flow features for 300 < surface water channel of width W = 1000 mm, length
Re < 4000 and different aspect ratios. An alternative approach 8000 mm and free-surface height H = 750 mm. The
was to interpret the vortex formation length on the basis of a test section is preceded by a 2:1 contraction ratio and flow
characteristic velocity fluctuation in the streamwise direction conditioning, which provided a free stream turbulence intensity
given first by Bloor and Gerrard [14] and Szepessy and of less than 0.2% in the range of Reynolds numbers from
Bearman [15]. The large variation in the formation length of Re = 550 to 3400, where Re = (U∞ D) /v, based on the CC
Kármán vortices over the range of Reynolds number from diameter and width of the SC. Here, v is the kinematic viscosity.
5000 to 40,000 was observed qualitatively by Schiller and An overview of the experimental system for the SC case is
Linke [16]. Balachandar et al. [17] applied direct numerical shown in Fig. 1. Aspect ratios were 30 and 50 with respect
simulations at the lower Re number and large-eddy simulation to the submerged section and the whole length of the cylinder,
at the higher Reynolds number over several different 2- respectively. A sharp-edged end plate of 800 mm × 800 mm ×
D bluff bodies for 250 ≤ Re ≤ 140,000. They also 13 mm was located on the bottom surface of the water channel
performed experiments to support their numerical solutions. as shown in Fig. 1. The purpose of using the end plate was to
They investigated properties of the time- and phase-averaged reduce the three-dimensional flow and deflection effects. The
mean wake recirculation region in the separated flows. They SC, of 20 mm × 20 mm with sharp-edged cross-section, was
explained the distribution of Reynolds stress and pressure in made of plexiglass. On the other hand, a circular plexiglass
relation to the mean wake recirculation region (wake bubble) cylinder of 20 mm diameter had a hollow cross-section.
and vortex formation length. The response of the SC that An Nd:Yag laser was used to generate a laser sheet
was free to rotate in a uniform flow was solved numerically perpendicular to the axis of the cylinder, and a CCD camera
using the 2-D incompressible time dependent Navier–Stokes with a resolution of 1024 × 1024 pixels was used to record
equation, which was also examined experimentally on fixed the images (see Fig. 1). The thickness of the laser sheet was
and freely rotating cylinders for 1000 ≤ Re ≤ 10,000 by around 1.5 mm. The laser sheet was located horizontally at a
Zaki et al. [18]. In order to obtain flow characteristics and height of h L = 330 mm above the channel’s bottom surface,
 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235 227

factors for velocity measurement and vorticity calculation

in the DPIV method mainly comprised of the seeding
particle size, non-uniform particle distribution, particle overlap,
unmatched particle correlations, out of laser plane motion
(bias and random errors), interrogation window size, and
electronic and optical imaging noise. The details of the
effects of these factors can be found in the studies of
Adrian [28], Westerweel [29], Keane and Adrian [30],
Fouras and Soria [31], and Hart [32]. Factors contributing
to uncertainty in the velocity measurement using the DPIV
technique were critically assessed, Westerweel [29] concluding
that uncertainty estimation in the velocity measurement was
less than 2%.

3. Results and discussion

3.1. Spectral analysis and calculation of vortex shedding

frequency

The dominant vortex shedding frequency f s is determined

from power spectra of recorded DPIV data in the near-
wake region of the cylinders, particularly inside the free
shear layer region, at the various selected points. Statistical
characteristics of the instantaneous streamwise velocities in
Fig. 1. Schematic of the experimental system and definition of the parameters time series are determined in order to get information in
for the square cylinder.
the frequency domain by using spectral analysis, i.e. Fast
while the water height was h W = 600 mm in all cases. The Fourier Transformation (FFT). FFT analysis in the wake of
seeding particles in the flow were silver-coated hollow glass the body reveals significant quasi-periodic flow oscillations.
spheres with a diameter of 15 µm. The specific gravity of the As is known, the most powerful spectral peak reflects the
seeding particles, provided by Potters Industries Inc., was 1.10. vortex shedding frequency in the spectra. Input data number
The densities of the particles and water were close enough so for frequency spectra should be power of 2. It was obtained
that the distribution of particles in suspension remained uniform from padding time-averaged value of 200 DPIV instantaneous
for several hours. A Dantec DPIV system was used to measure velocities at the end of time history of the velocity in order to
velocity vectors. In image processing, a spot size of 32 × 32 extend input data number to the value of 4096 samples, for
pixels was used to allow effective interrogation. During the each spectrum analysis in Fig. 2. Depending on the locations
interrogation process, an overlap of 50% was employed in order where the spectral measurements are calculated, the magnitudes
to satisfy the Nyquist criterion. Furthermore, to satisfy the high- of the spectral peaks varied with the Reynolds number and
image-density criterion, the interrogation window typically bluff body shape. Spectral peaks do have different dominant
contained 20–30 particles per image. Finally, 200 instantaneous frequencies and magnitudes due to the different shape and
velocity vector fields were collected at the rate of 15 Hz for the separation points, as displayed in Fig. 2, even though the
each Reynolds number. The time delay between pulses ranged cylinders have the same nominal diameter and symmetrical
from 1 ms to 12 ms. In-house software was used to evaluate configuration to the oncoming flow. The results of the FFT
and remove inappropriate displacement vectors (less than 1%), calculation point out that dominant frequencies at the selected
caused by shadows, reflections, or laser sheet distortions in the points are approximately equal in the x and y directions,
flow field and replaced by using bilinear interpolation between i.e. cross-stream and streamwise velocity components. Here,
surrounding vectors in the post-processing step. The field was only the streamwise velocity component result is presented,
then smoothed by a Gaussian weighted averaging technique. To but Strouhal numbers in various positions of the wake region
minimize distortion of the velocity field, a smoothing parameter are calculated and given as mean values in Table 1 for 550 ≤
of σ = 1.3 was chosen. After producing the vector field, Re ≤ 3400. All Strouhal numbers presented for the 45◦ OSC
the vorticity patterns of the wake flow were determined by were based on the projected cross-section of the cylinder. In
using a finite difference scheme. For all the experiments, image Fig. 2, Strouhal numbers at Re = 3400 are obtained as 0.21,
magnification was 1:5.91, which yielded an effective grid size 0.13 and 0.165 for the CC, SC and OSC, respectively. The
of 2.75 mm × 2.75 mm in the physical plane of the laser sheet, peak values of the vortex shedding frequencies at the upper
corresponding to 3844 velocity vectors. The overall field of and lower wake region of the cylinder are the same, which are
view was 171 mm × 171 mm (8.5D × 8.5D) in all cylinders. 1.75, 1.12 and 1.05 Hz for the CC, SC and OSC, respectively.
Adrian [28] and Westerweel [29] provided very detailed The spectrum magnitude of the peak frequency for the SC is
information about particle image velocimetry. The uncertainty larger than that of the CC due to the higher rate fluctuations
228 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235

and 19% higher than the SC and OSC, respectively. It is worth

nothing that there are some discrepancies between the obtained
results and some of the previous studies, as seen in Table 1.
Possible sources of these discrepancies could be the difference
in Reynolds number among the various results, a different
blockage ratio, the aspect ratios of the model, the experimental
setup, surface roughness, wind tunnel or water channel, the
turbulence intensity of the free-stream flow, end conditions, and
tiny nonparallel effects [3,11,22,33,23,34]. Strouhal numbers
given in Fig. 2 and Table 1 for the present investigation are
based on evaluations of a spectral frequency interval, of about
0.004 Hz, corresponding to a resolution of about 3% in terms
of the dominant vortex shedding.

3.2. Instantaneous and time-averaged flow patterns

Representative instantaneous vorticity contours of the wake

structure for the CC, SC and OSC geometries at Re = 550 and
Re = 3400 are illustrated in Fig. 3 for Case A and Case B. The
instantaneous vorticity ω is normalized as ω∗ = ω/(U∞ /D).
The results of many qualitative and quantitative investigations
explain the well-known Kármán vortex shedding. In Case A,
there is a large-scale Kármán vortex in the upper side of the
cylinder, whereas for Case B the same flow behaviors occur
from the alternate direction, i.e. the lower side of the cylinder
in Fig. 3. Viewing the vortex street in the cross-sectional plane
gives the appearance of the upper row of negative vortices
(dashed lines) and the lower row of positive vortices (solid
lines). Flow entrainment to the rear of the cylinder coming
from the upper side of the cylinder reverses the clockwise
direction, whereas that from the lower side of the cylinder
rotates in the anticlockwise direction. These are known shear
layers, and alternatively rolls up to form the vortices in the
wake as follows. One of these shear layers coming from the
opposite side of the cylinder entrains into the growing shear
layer of the other side of the cylinder in the wake region. This
induces the detachment of the longer shear layer vortex from
the cylinder. Shorter vortex shedding located near the cylinder
has a rotating reverse flow towards the cylinder, which detaches
Fig. 2. Time history of instantaneous streamwise velocity component u (mm/s) the vortex from the cylinder surface and initiates production of
against time t (s) and spectrum analysis Su at denoted point with black donut
on the small image at Re = 3400 for CC (top), SC (middle) and OSC (bottom).
the next vortex. On the one hand, the flow separation points
in the middle column images of Fig. 3 occur at the left-side
originating from sharp-edged corners, as displayed in Fig. 2. corners of the SC. On the other hand, the separation points
The difference between the St values for the CC, SC and the are delayed up to the cylinder corners in the downstream
OSC might be explained by the different nature of their flows. side for the OSC, as seen in the right-hand column of Fig. 3.
For the SC at an angle of incidence (0◦ and 45◦ ) to the flow Instantaneous vorticity patterns in Fig. 3 clearly present the
and for the CC, the Strouhal numbers given in Table 1 were interaction between the shear layers formed on both sides of
measured by a number of authors. Table 1 clearly indicates that the cylinder. It is demonstrated that the SC has a wider wake
the present results are in close agreement with most of the region than that of the CC for all flow patterns. This effect
previous experimental and numerical results for all cylinders. contrasts with the increase in the transverse direction, which
The results of the present work for the SC agree well with the blocked the flow and thus causes an increase in drag force.
results of [3,9,22]. King [27] reported that the non-dimensional Dutta et al. [8] recorded similar observation through the smoke
wake Strouhal number for the CC was St ≈ 0.21 over a wide visualization of flow around the SC. The Reynolds number
range of Reynolds numbers, for example 102 < Re < 105 . In plays an important role in flow structure even if the bodies
the present study, Strouhal numbers obtained for the CC almost have sharp edges, as seen in Fig. 3. Instantaneous vorticity
matched the value of 0.21 generally found in the literature. The patterns quantitatively indicate the remarkable decrease in
Strouhal number at Re = 3400 for the CC is around 38% the formation length of the large-scale Kármán vortex with
 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235 229

Table 1
Comparison of Strouhal numbers

Authors Re SC OSC CC
Present study 550 to 3400 0.120 to 0.134 0.165 to 0.174 0.204 to 0.212
Okajima [3] 500 to 3400 0.121 to 0.128
Saha et al. [7] 500 0.135
Dutta et al. [8] 1340 0.142 0.191
Chen and Liu [9] 2000 to 3400 0.125 0.172
Zaki et al. [18] 1790 and 2450 0.120 and 0.122 0.156 and 0.161
Saha et al. [19] 500 0.120
Sohankar et al. [21] 500 0.122
Hwang and Sue [22] 500 and 1000 0.125 and 0.137
Davis et al. [23] 500 and 1000 0.140 and 0.150
Kim et al. [24] 500 and 3000 0.120 and 0.125
Knisely [26] 2200 0.125 0.160
Norberg [35] 550 to 3400 0.21
Nakagawa et al. [36] 3000 0.130

Fig. 3. Comparison of patterns of dimensionless instantaneous vorticity ω∗ for Re = 550 (top two rows) and Re = 3400 (bottom two rows) is presented. Both
minimum and incremental contour levels of positive (solid line) and negative (dashed line) dimensionless instantaneous vortices in all images are 0.5.

increasing Reynolds number. The vortices produced at the A and time-averaged vorticity ω∗ . Regarding the onset and
corners of the SC have a tendency to displace in the inward development of small-scale vortical structures in the separating
direction because of lower pressures prevailing within the wake. shear layer, regions of low-level vorticity concentration are
This is counter-balanced by the growing wake size, which discernible in the pattern of instantaneous vorticity for Re =
shifted the vortex centerline outwards. It is evident that the flow 550 and Re = 3400 in Fig. 3. Large-scale Kármán vortices
characteristics in the downstream of the SC and OSC are more for the SC and OSC form in the near-wake region due to the
complicated than the CC. In fact, at Re = 3400 for the OSC, the abrupt coalescence of small-scale shear-layer vortices in the
large-scale clusters of vortices induce negative vorticity layers separating shear layers that are similar to the CC work of Lin
that are immediately adjacent to the base of the cylinder as seen et al. [12]. Instantaneous vorticity patterns showed the irregular
in the right-hand column of Figs. 3 and 5, as displayed for Case vorticity shapes, and the shear layers tend to break up easily into
230 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235

Fig. 4. Comparison of dimensionless time-averaged flow patterns at Re = 550. Minimum and incremental values of the patterns are given on each image.

small concentrations of vorticity related to Kelvin–Helmholtz vrms /U∞ , and Reynolds stress correlations u v /U∞
2  for the

instabilities. They seem to combine into larger vortices during CC, SC and OSC are presented for Re = 550 and Re = 3400
the process of formation and shedding, as shown in Fig. 3. For in Figs. 4 and 5, respectively. Time-averaged vorticity patterns
Re = 3400, these concentrations become more pronounced, ω∗  clearly reflect the changes in the formation length of
and small-scale structures are clearly evident inside as well as at the large-scale vortices in the first rows of Figs. 4 and 5.
the periphery of the large-scale vortex. Concentrations of small- It is observed that the streamwise location at which the
scale vortices for the OSC are the greatest. minimum positive and negative contour lines approach each
Comparison of patterns of normalized time-averaged other near the symmetry plane of the wake is an indicator
vorticity ω∗  = ω/(U∞ /D), streamline topology ψ, of the time-averaged vortex formation length. It is evident
streamwise velocity u ∗  = u/U∞ , cross-stream velocity that the vorticity contours for all cylinder cases are deflected
v ∗  = v/U∞ , rms contours of the streamwise velocity inwards at a location immediately downstream from the base
u ∗rms  = u rms /U∞  and cross-stream velocity vrms ∗  = of the cylinder at Re = 3400. This type of inward deflection
 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235 231

Fig. 5. Comparison of dimensionless time-averaged flow patterns at Re = 3400. Minimum and incremental values of the patterns are given on each image. All
patterns are scaled with respect to the cylinder size.

occurs at a distance significantly further downstream from signed vortices interact more directly and vigorously, resulting
the base of the cylinder at Re = 550. For low Re number, in substantial cancellation of the averaged vorticity, as seen in
a large-scale vortex appears within the flow field further Figs. 4 and 5. The streamwise separation length of successive
downstream. However, at successively higher Reynolds vorticity peaks in the near-wake region for the OSC is larger
number, such as Re = 3400, it moves further upstream and longer than that of the CC and SC. This is more apparent
and is closer to the base of the cylinder. In the first row of in the first row of Fig. 3. The shape of vortex contours in the
Figs. 4 and 5, the time-averaged vorticity images reveal that near-wake is similar for both geometries of the CC and SC.
the detailed instantaneous structure of the small-scale vortices Comparison of the time-averaged vorticity patterns ω∗  and
and the far downstream part of the Kármán vorticity street are streamline topology ψ with each other and itself shows that
lost due to the unsteady flow structure and alternating direction flow structures of the wake are almost equally symmetrical
of the vortex shedding from both sides of the cylinders. In with respect to the centerline of the CC, SC and OSC for
the base region, two separated shear layers with oppositely Re = 3400. Patterns of time-averaged streamline topology
232 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235

ψ identify the major changes of the near-wake behind the sectional patterns of the wake are evident; they are consistent
cylinders. They exhibit well-defined critical points, e.g. foci with regions of abrupt wake widening observed in the sectional
(centers of vortices) designated as F1 and F2 and saddle points patterns of the flow structure for the SC and OSC, as seen in
(apparent intersections of streamlines) denoted as S, as shown Figs. 3–5.
in the second row of Fig. 4 for the OSC, and symmetrical flow Patterns of time-averaged cross-stream velocity v ∗  are
structures for the CC, SC and OSC in the second row of Fig. 5. given in the fourth row of Figs. 4 and 5. The patterns of
From the averaged streamline patterns in Fig. 4, the streamline cross-stream velocity are detectable near the interface between
shapes extending upstream to the near-base region of the CC the separating shear layers. Moreover, it is accompanied by a
and SC are elliptical and have a limited spiral cycle focused decrease in the streamwise extent of the separation bubble at
very close to the upper side of the downstream boundary of the Re = 3400 relative to the bubble length at Re = 550. The
SC and the lower side boundary of the CC due to insufficient contours of constant cross-stream velocity v ∗  close to the base
capture time and data numbers. It is worth emphasizing that of the cylinder have relatively larger values for Re = 3400.
computer controlled data acquisition systems could not allow A large amplitude of cross-stream velocity v ∗  at Re = 3400
capturing and processing DPIV data numbering more than might indicate that the developing shear layers which originally
200. Specifically for Re = 3400, whole-field representations separate from the CC, SC and OSC entrained a significant mass
of streamline patterns, along with the identification of the flow in the base region through both sides of the cylinder. A
locations of these types of critical points, could provide an possible explanation for this phenomenon is an increase in the
insight into the effective formation length of large-scale vortices level of Reynolds stress in the separating shear layer, as seen
and the degree of symmetry of either side of the near-wake. in the bottom rows of Figs. 4 and 5. It might be concluded
Streamline saddles in the near-wake similarly correspond to from pattern concentrations that Reynolds stress correlations
the time-averaged zero vorticity contour. The non-dimensional have considerably enhanced values at Re = 3400 with regard
wake lengths (L/D) measured from the center point of the to Re = 550.
cylinder to the saddle point in the averaged streamline topology The time-averaged contours of the Reynolds stress
ψ for Re = 3400 in Fig. 5 are approximately 1.65, 1.23 correlation u v /U∞
2  associated with rms streamwise velocity

and 2.20 for the CC, SC and OSC, respectively. These values ∗
u rms  and the cross-stream velocity component vrms ∗  are

become around 2.75, 1.92 and 4.0 for the CC, SC and OSC, shown in the bottom three rows of Figs. 4 and 5. The rms
respectively, for Re = 550, as seen in the second row of Fig. 4. flow patterns display similar behaviors, not only for different
Representations of the streamwise velocity component u ∗  Reynolds numbers but also for the CC, SC and OSC, whereas
over the entire field of view are also given in the third row of they have discrepancies from the points of the flow pattern
Figs. 4 and 5. Substantial regions of negative (reverse) flow concentrations and vortex formation lengths. At Re = 3400, the
are evident at Re = 550 and Re = 3400 for minimum fluctuations and irregularities are greater for the SC and OSC
and incremental contour levels. From these images, it is also than the CC in the separating shear layer. The rms streamwise
possible to define the location of a stagnation point around the velocity patterns u ∗rms  have easily detectable double peaks
wake symmetry plane, which is designated by a black dot. The at almost equal locations in the upper and lower sides of the
distances from the center point of the cylinders to the stagnation cylinder centerline for all cylinders presented in Fig. 5, which
point at Re = 550 in Fig. 4 are approximately 2.4D, 1.9D and are placed around 1.6D, 1.5D and 1.8D for Re = 3400 for
4.3D for the CC, SC and OSC, respectively. The locations of the CC, SC, and OSC, respectively. The maximum contour
the stagnation points at Re = 3400, which become around ∗  for Re = 550 in the centerline of the wake
values of vrms
1.7D, 1.2D and 2.2D for the CC, SC and OSC, respectively, occur at nearly 3.1D, 2.5D and 4.3D, whereas these distances
move upstream, as seen in the third row of Fig. 5. Viewing the for Re = 3400 decrease to approximately 2.5D, 2.2D and
patterns of Figs. 4 and 5 as a whole, the effect of orientation 4.1D for the CC, SC and OSC, respectively. Cross comparison
of the OSC substantially increases the streamwise length of of the results reveals double peaks for u ∗rms , while a single
the separated vorticity layers relative to the case for the CC peak is observed in vrms∗  with a maximum occurring along the

and the SC. Then, the growth of the separated shear layers centerline axis of the vortex near to a location coincident with
from the cylinder shoulders is associated with an increase in the stagnation point in the downstream of the cylinders.
the wake size for the OSC. Another important parameter of the Reynolds stress correlations u v /U∞ 2  depict a set of

wake is that the non-dimensional width (W/D) of the wake extrema on either side of the wake centerline for the CC, SC
can be defined as the cross-stream distance between extrema and OSC in the bottom rows of Figs. 4 and 5, which provide a
of the time-averaged vorticity layers on opposite sides of the direct comparison of patterns for two Reynolds numbers from
wake. For example, quantitative interpretations of the wake the point of momentum transfer. The patterns in the Reynolds
width from the time-averaged vorticity, corresponding to lateral stress correlations consist of both small-scale clusters located in
distances passing through maximum and minimum contour the vicinity of the base of the cylinder and large-scale clusters
values in the top row of Fig. 5, are respectively 0.95D, 1.48D located just downstream from the small-scale clusters. The
and 1.76D for the CC, SC and OSC at Re = 3400. On the other well-defined Reynolds stress patterns due to the fluctuations
hand, these values for Re = 550 are 1.21D, 1.54D and 2.17D in the shear layers produce a weaker Reynolds stress region
for the CC, SC and OSC, respectively. At these locations, as very close to the base of the cylinder which occurs as a result
well as at locations downstream, local bulges of the cross- of the flow entrainment into the wake region. The basic forms
 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235 233

of these patterns persist for Re = 550 and Re = 3400. The 200 instantaneous vortices. As seen from the time history of
extrema of the large-scale contours for Re = 3400 are displaced the streamwise velocity component u at any point in the field
downstream by distances of nearly 2.1D, 1.8D and 2.4D with of view shown in Fig. 2, there are around 14 and 25 vortex-
regard to the center of the CC, SC and OSC, respectively. shedding cycles (periods) of the SC and CC cases, respectively.
However, these distances have larger values at Re = 550 Each period includes approximately 7 and 15 instantaneous
and reach 2.8D, 2.3D and 4.1D for the CC, SC and OSC, images in sequence for the CC, SC and OSC, respectively.
respectively. The upstream extension of each large-scale cluster Second, according to the instantaneous streamwise velocity
of u v /U∞
2  migrates further upstream and the tip of the large- component u, any point in the wake region of 7 or 15 images
scale clusters is located closer to the base of the cylinders for captured during the vortex shedding period is chosen for the
Re = 3400 in Fig. 5. As clearly displayed in the bottom row of CC, SC and the OSC as a reference frame. Third, this velocity
Figs. 4 and 5, this large-scale cluster migration is the greatest component is compared with the instantaneous streamwise
for the OSC. This trend is in accord with the earlier appearance velocity component u of other periods of images in a sequence
of the confined small-scale concentrations of vorticity in the at the same location. Fourth, when they are close to each
shear layer, as indicated in the corresponding instantaneous other within the accepted experimental accuracy, the vorticity
images in Fig. 3. Same-signed Reynolds stress correlation patterns of these two instantaneous images are compared with
patterns u v /U∞
2  and cross-stream velocity components v ∗  each other by superposition. Finally, the most similar 11 images
extend one side of the centerline to the other for all cylinders. of each respective phase of the vortex shedding cycle are chosen
The patterns of Reynolds stress correlations occur over a larger to obtain phase-averaged flow patterns for the SC, OSC and
extent across the wake for the OSC. Peak values of Reynolds CC, separately. Average relative errors of the phase-averaging
stress correlations increase in the base region and then decay method between instantaneous velocity component u ∗ and
sharply in the wake region, as seen from the concentrations phase-averaged velocity component u ∗  p , which are obtained
of the flow structure. Reynolds stress correlation u v /U∞ 2  is from point to point in the field of interest shown in Fig. 6, are
zero around the cylinder axis as a result of symmetrical flow less than 4%, 8% and 9% for the CC, SC and OSC, respectively.
structure for Re = 3400 and attains its greatest magnitude These phase-averaged images, which emphasize the large-
in the upper and lower regions of the centerline, although scale repetitive vortical structures, are intended to serve as a
peak values of vrms ∗  occurred at the axis itself. Thus, velocity guide for the interpretation of instantaneous vorticity patterns
fluctuations penetrate deeper into the outer flow and finally ω∗ . Formation of the large-scale vorticity clusters of vortices
cause an increase in the drag coefficient as a result of increased and von Kármán-like vorticity is barely discernible in the
wake size. The Reynolds stress correlations in and around the patterns of instantaneous vortices ω∗ for all cylinders in Fig. 6.
mean wake recirculation region are primarily due to the time- Furthermore, large-scale concentrations of vorticity are clearly
dependent nature of the vortex shedding. In this limit, it is evident in both the upper and lower side layers of the cylinder
observed that the Reynolds stress correlation is relatively small for phase-averaged vorticity, as well as the further wake region.
in the outermost region of the flow patterns, as displayed in The upper layer is deflected substantially downwards from its
the bottom images of Figs. 4 and 5. Despite the significant shedding location, while the lower layer acts in the reverse
differences in the geometry of the bluff body, the distributions direction. Viewing the patterns of phase-referenced ω∗  p
of Reynolds stress patterns around the wake bubble for the and instantaneous vorticity ω∗ as a whole, similar vorticity
SC and OSC are quantitatively similar to the CC in the range layers are obviously obtained from phase-averaged patterns
550 ≤ Re ≤ 3400. However, the measured values of the for all cylinders. For the cases of the CC, SC and OSC, it is
peak Reynolds stresses are somewhat higher for the SC and evident that small-scale vorticity concentrations occur in the
OSC than those of the CC. Furthermore, the peak values of separated shear layer for all cases, as is clearly seen from the
normalized Reynolds stresses u v /U∞ 2  for the CC, SC and instantaneous vorticity patterns shown in Figs. 3 and 6. These
OSC are 0.08, 0.12 and 0.14 for Re = 3400, respectively. small-scale concentrations in vorticity probably come from the
3-D nature of the separated flow. However, phase-averaging
3.3. Comparison of phase-averaged and instantaneous images cancels out these vorticity layers, while preserving the identity
of large-scale Kármán vortices. In the third row of Fig. 6 for
In addition to time-averaging, phase-averaging images have all cylinders, phase-averaged streamwise velocity u ∗  p shows
been pursued. In cases where a periodic signal is recorded that rolled up flow originally starting from the upstream side
along with the DPIV images, it is possible to employ this of the SC, OSC and CC is conveyed downstream together
signal as a phase-reference, provided that the timing of with the corresponding Kármán vortex; they form a counter
DPIV images is known. Patterns of normalized phase-averaged rotating flow pair. Hence, the cylinder wake greatly enhances
vorticity ω∗  p = ω p /(U∞ /D), streamwise velocity u ∗  p = the mixing process of the flow. The wake region of the OSC for
u/U∞ and cross-stream velocity v ∗  p = v/U∞ for the phase-averaged streamwise velocity u ∗  p is the widest. The
CC, SC and OSC at Re = 3400, corresponding to representative wavelengths from one pocket center to other, placed around the
instantaneous patterns of vorticity ω∗ , are given in Fig. 6. same lateral location for phase-averaged streamwise velocity
Calculations of phase-averaged flow patterns have been u ∗  p , are approximately 3.8D, 3.2D and 2.6D for the CC, SC
accomplished as follows. First of all, a movie of instantaneous and OSC, respectively. In viewing all images in Fig. 6, phase-
vorticity patterns is created by using cinema in a sequence of averaged flow patterns reveal the alternate staggered nature
234 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235

Fig. 6. Comparison of dimensionless instantaneous vorticity and phase-averaged flow patterns at Re = 3400 for the CC, SC and OSC.

of the vortex street, thus signifying a dominant periodicity Strouhal numbers for the CC, SC and OSC are found around
in the flow, as seen in Fig. 2. In these so-called staggered 0.21, 0.13, and 0.17, respectively.
patterns, both positive and negative concentrations of phase- Quantitative comparison of the CC and SC show that length
averaged cross-stream velocity v ∗  p exist at a given spanwise scales are larger for the SC, not only in the cross-stream
location. However, the wavelengths of phase-averaged cross- direction but also in the stream-wise direction. Owing to the
stream velocity v ∗  p have two values for negative and positive fixed separation points of the SC, the flow separation and the
clusters for all cylinders. The wavelengths between the positive recirculation regions have an increasing effect in the wake
centers of v ∗  p are around 3.8D, 5.7D and 5.1D, which are region. On the other hand, the size of the wake region and
closer to the cylinder base region, shorter than that of negative the distance between oppositely signed vortices increase due
one, and have values of nearly 3.0D, 4.4D and 3.9D for the to the 45◦ orientation effect of the SC. The flow patterns have
CC, SC and OSC, respectively. The reasons for the differences considerable symmetry about the centerline of the cylinders
in the wavelengths might be the higher convective speed of the of the flow patterns, as shown in all time-averaged images
vortices and the absence of the strong interaction between the for Re = 3400. However, unsymmetrical flow structures are
oppositely signed vortices in the far-wake of the cylinder. obtained for Re = 550 because of the insufficient data numbers.
Contours of the time-averaged streamwise velocity demonstrate
4. Conclusions that the stagnation point around the symmetry plane moves
further upstream for higher Re. Time-averaged cross-stream
This investigation focused on the generation of vortical velocity contours v ∗  and Reynolds stress correlations, which
structures of the near-wake region arising from flow passing are oriented in the direction of each respective shear layer
the CC, SC and OSC for 550 ≤ Re ≤ 3400 in order to explain separation from the cylinder, are important features of the near-
the physical mechanisms of the flow structure. The present wake structures from the point of momentum transfer, as Re
paper has supported the previous works by providing detailed becomes larger. In terms of the concentration values of the flow
quantitative experimental information with DPIV in the near- patterns for three different cases, the peak value of the OSC
wake region of the CC and SC. It is found that the values of is slightly higher due to the stronger rotational fluid motion
Strouhal number, as well as the wake patterns, are functions in the wake region than that of the CC and SC. The shear
of the cross-section of the cylinders and Reynolds numbers. layer surrounding the recirculation bubble behind the cylinders
 M. Ozgoren / Flow Measurement and Instrumentation 17 (2006) 225–235 235

has a region of intense velocity fluctuation with high values of (300 < Re < 4000) using PIV. In: Proceedings of FEDSM’98. 1998, p.
Reynolds stress correlation because of the vortex interactions. 1–7.
Patterns of u ∗rms , and u v /U∞
2  have indicated double peaks, [14] Bloor MG, Gerrard JH. Measurements of turbulent vortices in a cylinder
∗ wake. Proc R Soc London, Ser A 1966;294:319–42.
whereas that of vrms  shows only a single peak. The vortices [15] Szepessy S, Bearman PW. Aspect ratio and end plate effects on vortex
shed from both sides of the cylinder are the source of these shedding from a circular cylinder. J Fluid Mech 1992;234:191–217.
two minima in the flow contours. It is physically confirmed [16] Schiller L, Linke W. Druck- und Reibungswiderstand des Zylinders bei
that the predominant features of the instantaneous and phase- Reynoldsschenzahlen 5000 bis 40,000, Z. Flugtech. und Motorluftschif-
averaged vortices are closely correlated, thereby reaffirming the fahrt 24 (7) 193–198. English translation entitled Pressure and frictional
resistance of a cylinder at Reynolds numbers 5000 to 40,000, in National
use of instantaneous vorticity patterns as representations of the Advisory Committee Aeronautics, Technical Memorandum No. 715, July
flow. 1933.
In general, vortex formation lengths decrease with increas- [17] Balachandar S, Mittal R, Najjar FM. Properties of the mean recirculation
ing Reynolds number for all the patterns presented. Present re- region in the wakes of two-dimensional bluff bodies. J Fluid Mech 1997;
351:167–99.
sults can provide sufficient information about the details of the
[18] Zaki TG, Sen M, Gad-el-Hak M. Numerical and experimental
foregoing flow features to develop a strategy for the validation investigation of flow past a freely rotable square cylinders. J Fluids Struct
of numerical models in the future. 1994;8:555–82.
[19] Saha AK, Biswas G, Muralidhar K. Three-dimensional study of flow past
a square cylinder at low Reynolds numbers. Int J Heat Fluid Flow 2003;
Acknowledgments 24:54–66.
[20] Davis RW, Moore EF, Purtell LP. A numerical study of vortex shedding
It is a pleasure to thank the Mechanical Engineering from rectangles. J Fluid Mech 1982;116:476–506.
Department of Cukurova University, Turkey for letting me carry [21] Sohankar A, Norberg C, Davidson L. Simulation of three-dimensional
out the present experimental work. This work is supported by flow around a square cylinder at moderate Reynolds numbers. Phys Fluids
the Coordinatorship of Selcuk University’s Scientific Research 1999;11(2):288–306.
[22] Hwang RR, Sue YC. Numerical simulation of shear effect on vortex
Project, No. 2004/131. shedding behind a square cylinder. Int J Numer Methods Fluids 1997;
25:1409–20.
References [23] Davis RW, Moore EF, Purtell LP. A numerical-experimental study of
confined flow around rectangular cylinders. Phys Fluids 1984;23:46–59.
[1] Williamson CHK. Vortex dynamics in a cylinder wake. Annu Rev Fluid [24] Kim Do-H, Yang KS, Senda M. Large eddy simulation of turbulent flow
Mech 1996;28:477–539. past a square cylinder confined in a channel. Computers & Fluids 2004;
[2] Naudaschser E, Rockwell D. Flow-induced vibrations: an engineering 33:81–96.
guide. Rotterdam: A. A. Balkema Press; 1994. [25] Ahlborn B, Seto ML, Noack BR. On drag, Strouhal number and vortex-
[3] Okajima A. Strouhal numbers of rectangular cylinders. J Fluid Mech street structure. Fluid Dynam Res 2002;30:379–99.
1982;123:379–98. [26] Knisely CW. Strouhal numbers of rectangular cylinders at incidence: a
[4] Obasaju ED. An investigation of the effects of incidence on the flow review and new data. J Fluids Struct 1990;4:371–93.
around a square section cylinder. Aeronaut Q 1983;34:243–59. [27] King R. A review of vortex shedding research and its application. Ocean
[5] Durao DFG, Heitor MV, Pereira JCF. Measurements of turbulent and Eng 1977;4:141–71.
periodic flows around a square cross-section cylinder. Exp Fluids 1988; [28] Adrian RJ. Particle-imaging techniques for experimental fluid mechanics.
6:298–304. Annu Rev Fluid Mech 1991;23:261–304.
[6] Lyn DA, Einav S, Rodi W, Park JH. A laser-Doppler velocimetry study [29] Westerweel J. Digital particle image velocimetry, theory and application.
of ensemble-averaged characteristics of turbulent near-wake of a square Delft University Press; 1993.
cylinder. J Fluid Mech 1995;304:285–319. [30] Keane RD, Adrian RJ. Optimization of particle image velocimeters. Opt
[7] Saha AK, Muralidhar K, Biswas G. Experimental study of flow past a Methods Flow Particle Diagnostics 1989;68:139–59. 1404 on attached.
square cylinder at high Reynolds numbers. Exp Fluids 2000;29:553–63. [31] Fouras A, Soria J. Accuracy of out-of-plane vorticity measurements
[8] Dutta S, Muralidhar K, Panigrahi PK. Influence of the orientation of a derived from in-plane velocity field data. Exp Fluids 1998;25:409–30.
square cylinder on the wake properties. Exp Fluids 2003;34:16–23. [32] Hart DP. PIV error correction. Exp Fluids 2000;29:13–22.
[9] Chen JM, Liu CH. Vortex shedding and surface pressures on a square [33] Bearman PW, Obasaju ED. An experimental study of pressure
cylinder at incidence to a uniform air stream. Int J Heat Fluid Flow 1999; fluctuations on fixed and oscillating square-section cylinders. J Fluid
20:592–7. Mech 1982;119:297–321.
[10] Sarioglu M, Yavuz T. Subcritical flow around bluff bodies. AIAA J 2002; [34] Kolar V, Lyn DA, Rodi W. Ensemble-averaged measurements in the
40(7):1257–68. turbulent near wake of two side-by-side square cylinders. J Fluid Mech
[11] Roshko AJ. Perspectives on bluff-body aerodynamics. J Wind Eng Ind 1997;346:201–37.
Aero 1993;49:79–100. [35] Norberg C. An experimental investigation of the flow around a circular
[12] Lin J-C, Towfighi J, Rockwell D. Instantaneous structure of near-wake cylinder: influence of aspect ratio. J Fluid Mech 1994;258:287–316.
of a cylinder: on the effect of Reynolds number. J Fluids Struct 1995;9: [36] Nakagawa S, Nitta K, Senda M. An experimental study on unsteady
409–18. turbulent near wake of a rectangular cylinder in channel flow. Exp Fluids
[13] Noca F, Park HG, Gharib M. Vortex formation length of a circular cylinder 1999;27:284–94.