Experiments in Fluids 34 (2003) 16–23

DOI 10.1007/s00348-002-0484-x

Influence of the orientation of a square cylinder

on the wake properties
S. Dutta, K. Muralidhar, P.K. Panigrahi

16
Abstract An experimental investigation of flow around a fu frequency of vortex shedding based on the u-trace
square cylinder placed at various angles with respect to the fv frequency of vortex shedding based on the v-trace
approach fluid velocity is reported. The focus of the study Re Reynolds number, qUB/l
is toward examining the sensitivity of the wake properties Su Strouhal number based on the u-signal, fuB/U
to the cylinder orientation and Reynolds number. Angles Sv Strouhal number based on the v-signal, fvB/U
of incidence in the range of 0–60
and Reynolds numbers u x-component velocity
of 1340, 4990, and 9980 have been considered. Velocity v y-component velocity
measurements have been carried out using an X-wire U upstream velocity
hotwire anemometer. The Strouhal number and the drag u¢ x-component velocity fluctuation
coefficient of the cylinder have been computed from the v¢ y-component velocity fluctuation
wake measurements. Utilizing the velocity traces at dis- u0 v0 dimensionless
 turbulent
 shear stress,
tinct probe locations in the near and the far wake, statis- 1
P 0 0 1
¼ 100
N ui vi U 2
tical properties such as the RMS velocities and the spectra i
have been obtained. Results obtained in the present work x, y dimensionless coordinates from the cylinder center
revealed that for a cylinder with zero inclination, flow scaled by B
separates from the corners on the face exposed to the in- X, Y dimensional coordinates from the cylinder center
coming flow. For inclinations greater than zero, the points
of separation on the cylinder move downstream and the Greek symbols
wake size increases, but the separated shear layer rolls up q fluid density
over a shorter distance. These factors lead to a reduced l dynamic viscosity
drag coefficient and a higher Strouhal number. The center- h orientation of the cylinder with respect to the in-
line recovery of the time-averaged velocity and the decay coming flow
rates of velocity fluctuations depend on the Reynolds x in-plane component of vorticity scaled by U/B,
number. A marginal effect of the cylinder orientation is =(¶v/¶x–¶u/¶y)
also seen.
1
Nomenclature Introduction
Ac excess projected area of an inclined cylinder with Fluid flow past objects of nonaerodynamic cross sections
respect to the straight, =(D–B)/B% is often encountered in engineering applications. The
B edge of the square cylinder forces acting on these objects are closely correlated to the
D projected dimension normal to the flow direc- wake properties. To understand the interdependence be-
tion,=B(cosh+sinh) tween the geometry, the wake, and the forces experienced
CD drag coefficient based on
the upstream velocity and by a bluff object, an experimental study of flow past a
D, ¼ drag per unit length 12 qU 2 D square cylinder has been undertaken in the present work.
Eu dimensionless power spectrum of the u-component Unlike a circular cylinder, the separation points for square
of velocity cylinder are fixed. For a cylinder at zero angle of incidence,
Ev dimensionless power spectrum of the v-component these points are commonly at the leading edge. Thus the
of velocity flow characteristics are relatively insensitive to Reynolds
f frequency (Hz) number, but will depend on the orientation of the cylinder
with respect to the approach flow. The configuration
considered in the present work is depicted in Fig. 1. A
Received: 19 November 2001 / Accepted: 20 April 2002 fixed square cylinder with side B is positioned at an angle h
Published online: 12 November 2002 with respect to the uniform upstream velocity U. The
Springer-Verlag 2002 cylinder is long in the axial direction, in relation to its edge
B. The aspect ratio in the present experiments was close to
S. Dutta, K. Muralidhar (&), P.K. Panigrahi 15. For the range of velocities considered, incompressible
Department of Mechanical Engineering,
Indian Institute of Technology Kanpur, flow conditions can be assumed to prevail. Experiments
Kanpur 208016 UP, India for the present work were carried out in a low-speed wind
E-mail: [email protected] tunnel.
 the experiments. The wind tunnel motor was electronically
controlled to produce stable velocities over the duration of
the individual experiments. Free stream turbulence was
found to be less than the background noise present in the
anemometer (<0.05%), while the flow parallelism was
better than 98% over 95% of the tunnel height. The square
cylinder used in the experiments was made of Perspex,
25 mm·25 mm in cross section. The square rod was ma-
chined to ensure that the corners were sharp. The com-
mercially available hotwire anemometer (Dantec) and an
X-wire probe were employed in the present work. The
17
probe was calibrated in the tunnel itself against a pitot-
static tube and a 19.99-mm-H2O digital manometer
(Furnes Control). The DC values of the anemometer
Fig. 1. Schematic drawing of a square cylinder placed at an angle h to
voltage output were determined by using signals that were
the incoming flow 10-s long. All other quantities were determined by
numerically integrating the computer-acquired signals.
The x- and y-components of the instantaneous velocity
Okajima (1982) carried out an experimental study of were acquired simultaneously. Instantaneous voltage
flow past square and rectangular cylinders in the Reynolds signals were recorded using an A/D card (National
number range of 70–2·104. The dependence of the Instruments) placed in a computer (HCL P-II). LabVIEW
Strouhal number on the aspect ratio of the cross-section as software (National Instruments) was used to collect the
well as Reynolds number was brought out through the anemometer data. The wake patterns as a function of the
study. Durao et al. (1988) conducted laser-Doppler veloc- cylinder inclination were visualized in a smoke tunnel that
imetry (LDV) measurements for flow past a square cylin- used kerosene. Flow patterns were subsequently imaged
der at a Reynolds number of 14,000 in a water tunnel. Lyn with a digital camera (Sony).
et al. (1995) carried out LDV measurements in the wake of The time-averaged drag coefficient, the Strouhal num-
a square cylinder in a water tunnel at a Reynolds number ber, the time-averaged profiles of velocity, and the velocity
of 21,400. The experiments brought out the differences fluctuations were determined from the experiments. The
between the base region and the near wake, on one hand, time-averaged drag acting on the cylinder was determined
and the relationship between the flow topology and the by the wake-survey method. The momentum loss calcu-
turbulence distribution, on the other. Strouhal numbers lation was carried out at the midplane of the cylinder at a
for rectangular cross sections as a function of their ori- dimensionless distance of x=18. A considerable variation
entation have been reviewed by Knisely (1990). A more in the static pressure was seen within the wake and so the
detailed study relating the forces and shedding frequencies measurements were directed individually at the total and
to the flow details has been reported by Norberg (1993). static pressures. The experimental facility employed in the
The roles of confinement effects and the intensity of free present work is quite similar to the one reported by Saha
stream turbulence on the forces were included in the et al. (2000a).
study. Sohankar et al. (1998) have carried out numerical
calculations at low Reynolds numbers (Re<200) to study 3
the influence of the cylinder orientation on the macro- Data reduction
scopic flow properties such as forces, moments, and the In experiments with low as well as high velocities,
Strouhal number. Literature pertaining to the experiments streamwise velocities close to the cylinder are negative and
on flows past cylinders of square cross section at zero the velocity fluctuations are expected to be large. In the
angle of incidence has been reviewed by Saha et al. present experiments, the first measurement location was at
(2000a). x=3.0, and a definite positive value of the time-averaged
The objective of the present work is to examine the velocity was recorded. The conversion of anemometer
properties of the wake of a square cylinder as a function ofvoltages to velocities was accomplished through an explicit
its orientation to the incoming flow. Three Reynolds non-real time method of Chew and Simpson (1988). The
numbers, namely 1340, 4990, and 9980, and orientations of method does not have a restriction with respect to the
0, 22.5, 30, 45, and 60 degrees have been considered in the magnitude of the velocity fluctuation.
experiments. The wake has been characterized in terms of Earlier studies show the size of the base recirculation
the drag coefficient, the Strouhal number, the velocity region at high Reynolds numbers to be in the range
profiles, and the velocity fluctuations. 1.3–1.4 (Lyn et al. 1995; Durao et al. 1988). A discussion
on the measurements of large fluctuations in low velocity
2 zones has been provided by Swaminathan et al. (1986).
Apparatus and instrumentation These authors have shown that the turbulence correction
The experiments for the present study were carried out in factor in the time-averaged quantities is less than 10% for
an open-circuit wind tunnel with nominal cross section of turbulence intensities as high as 45%. In the present
40 cm·40 cm and a 2-m long test section. Free stream work, the largest recorded velocity fluctuation was 30% at
velocities in the range 0.8–6 m/s have been utilized in all x=3.0. Thus the choice of the first measurement station
 for the time-averaged velocity as 3.0 can be taken as The Strouhal numbers for the square cylinder wake at
acceptable. zero angle of incidence have been reported by several
Power spectra were determined by first converting the authors (Kelkar and Patankar 1993; Norberg 1993; Franke
anemometer voltage signals to velocity and subsequently et al. 1990; Okajima 1982; Okajima et al. 1990; Davis and
applying the FFT algorithm. A sampling frequency of Moore 1982; Durao et al. 1988; Lyn et al. 1995). For a
5 kHz was employed during A/D conversion. The total Reynolds number range of 1000 to 21,400, the Strouhal
numbers have been found to lie between 0.122 and 0.142.
signal length was 13.2 s, yielding a total of 66,000 points in
In the present study, the values of the Strouhal number
the time series. The data set was split into eight parts for
the calculation of the spectrum, which was later averaged. were in the range 0.141–0.143. The near-constancy of the
This procedure yielded a useful frequency range 0.1– Strouhal number is indicative of a nearly constant wake
1 kHz. For Strouhal number calculations, the vortex- width and is again related to the fixed position of the
18
shedding frequency was measured at the dimensionless separation points. The percentage variation in the Strouhal
location of x=5, y=1.2. The offset location produced clear number is, however, marginally greater from one experi-
peaks in the power spectra at the shedding frequency. ment to another when compared to the drag coefficient.
To estimate the degree of repeatability of the experi- This can be explained by the fact that the shedding fre-
mental data, the following approach was adopted. Exper- quency in the near wake, and, in turn, the wake size is
iments were performed at a nearby Reynolds number influenced by the upstream turbulence level, blockage, and
approximately 10% away from the original. The two sets of confinement effects of the individual experiments, as well
data were then compared in terms of the dimensionless as simulation.
velocity and the velocity fluctuations. The agreement be- A comparison of the time-averaged drag coefficients
with the experiments of Norberg (1993) for cylinders at
tween the two sets of data was very good. In this sense, the
results reported in the present work have a high degree of various angles of incidence is presented in Fig. 2. For a
repeatability. The difference in the time-averaged quanti- cylinder placed at an inclination to the main flow, the drag
ties between Re=4990 and 9980 is also quite small, indi- coefficient is based on the projected dimension D, rather
cating low levels of uncertainty in the present set of than on the cylinder size B. All Reynolds numbers in the
experiments. Based on this criterion, uncertainty levels ofpresent work are based on the cylinder size B. At the three
± 5% and ± 10% can be assigned to the local values of the Reynolds numbers considered in Fig. 2, the drag coeffi-
time-mean and root-mean square (RMS) velocities. cient shows a reduction with respect to the zero-angle
configuration and is a minimum at 45
. The reduction is
4 around 40% with respect to the zero-angle configuration.
Results and discussion It is practically equal to the increase in the projected area,
Wake properties of a square cylinder at various angles of that is, 41%. Thus, in dimensional terms, there is no
incidence over a range of Reynolds numbers are discussed change in the drag force with the angle of incidence at a
in the following sections. The measured data has been
compared, where appropriate, with those published in the
literature in Sect. 4.1. The time-averaged flow properties
and the spectra are reported in Sects. 4.2 and 4.3, re-
spectively.

4.1
Comparison with published data
Time-averaged drag coefficients for a square cylinder at
zero angle of incidence have been reported by various
authors (Norberg 1993; Knisely 1990; Igarashi 1987;
Nakaguchi et al. 1968; Bearman and Trueman 1972;
Courchesne and Laneville 1979). The Reynolds number in
the referred studies varies over 500–40,000. The corre-
sponding drag coefficients fall in the range 2.0 to 2.4, with
no definite trend being discernable. The drag coefficient
obtained at the highest Reynolds number (Re=9980) of
the present work was 2.21 and increased to 2.27 at the
lowest Reynolds number of 1340. The agreement among
the drag coefficients reported in the literature with the
present work can hence be considered satisfactory. The
near constancy of the drag coefficient with the Reynolds
number has been discussed by Saha et al. (2000a). It has
been traced to the fact that the dimensionless flow fields in
the wakes of square cylinders are practically independent
of the Reynolds number. This is primarily because the Fig. 2. Comparison of the present experiments with Norberg (1993)
points of separation are fixed by the sharp corners facing in terms of the drag coefficient and Strouhal number as a function of
the incoming flow. the angle of incidence
given upstream velocity. This result was seen for all indi- the transverse dimension D decreases with increasing h,
vidual Reynolds numbers and angles of orientation. It can and is a minimum at an angle of 45
. The shift in the point
thus be concluded that the leading order correction to of separation is, however, responsible for an abrupt change
cylinder drag is contained in the change in the projected in the drag coefficient as the inclination increases from 0

area. Factors such as the Reynolds number and the wake to 22.5
.
properties are secondary and contribute to the higher The increase in the Strouhal number with the angle h
order. can be interpreted with reference to Fig. 3 along the fol-
The Strouhal numbers (based on the cylinder dimen- lowing lines. The movement of the separation points in-
sion B) as a function of angle and Reynolds number are creases the wake size. On the other hand, the separated
also presented in Fig. 2. The data of Norberg (1993) are boundary layer from an inclined cylinder rolls up over a
shown by lines, while data from the present experiments shorter distance in comparison to the wake size. The
19
are shown by symbols. The variation of the Strouhal vortex-shedding frequency correlates with the smaller of
number with the angle h is not monotonic. A change in the the length scales, namely the size of the rolled up shear
angle from 0 to 22.5
leads to a sharp increase in the layer, as against the wake width. For h=0 the wake width is
Strouhal number. Smaller changes are seen for other val- the smaller length dimension, and the vortex-shedding
ues of h. For h=45
the Strouhal number is high, but not frequency scales with this dimension. Thus, for angles of
necessarily a maximum. There are visible changes in the inclination greater than zero, the effective length scale of
trends with respect to the Reynolds number. the shed vortex decreases, correspondingly increasing the
To explain the trends in the drag coefficient and the shedding frequency and the Strouhal number.
Strouhal number with the incidence angle, flow visualiza- The variation of the RMS lift coefficient with the cyl-
tion studies were carried out in a smoke tunnel facility at a inder orientation was reported by Sohankar et al. (1998)
Reynolds number of 3920. The recorded images are shown based on a numerical calculation at Re=200. It was seen
in Fig. 3. We can conclude from Fig. 3 that for h=0 the that the RMS lift coefficient increased initially with h, and
point of separation is at the left corners of the cylinder, leveled off in the range 30–45
. The wake may thus be
namely, those exposed to the approach flow. At higher considered to become increasingly vigorous for h>0,
angles, flow separation is delayed up to the cylinder cor- reaching a maximum in the range 30–45
. These results
ners on the downstream side. Hence for all angles greater can be compared with the present work by examining the
than zero, there is a greater pressure recovery. This effect variation of the RMS velocity fluctuations in the near wake
is contrasted against the increase in the transverse di- with the angle h.The comparison is based on the expec-
mension blocking the flow. An examination of the data in tation that the pressure fluctuations that lead to the fluc-
Fig. 2 shows the second factor to be more important than tuating lift component are well-correlated with one or both
the first. Hence the drag coefficient defined with respect to of the velocity fluctuations. The comparison has been

Fig. 3. Smoke visualization of

flow around a square cylinder at
various angles of incidence with
Re=3920
 carried out with velocity measurements at x=5, y=1.2, probe was located at the offset location y=1.2 in both the
where the vortex-shedding frequency was determined. The experiments. The Strouhal numbers at the two x-locations
x- and y-components of the velocity fluctuations as a based on the time traces of the two velocity components
function of the incidence angle are shown in Fig. 4 for the are presented in Tables 1, 2 and 3. At Re=1340, the wake is
three Reynolds numbers. At the lowest Reynolds number, in the transitional zone and the dependence on the probe
u¢ increases with h up to 30
and subsequently decreases. location is tangible. During experiments, it was found that
On the other hand, v’¢ remains relatively constant. At the vortex-shedding frequency at the lowest Reynolds
Re=4990, both fluctuations increase with h up to 45
and number exhibited time dependence as well, which in turn
decrease for higher angles. However, at Re=9980, the ve- was a source of scatter. The flow complexity in the lower
locity fluctuations show a monotonic reduction with the range of Reynolds numbers has been discussed by Saha
incidence angle. Thus a degree of similarity is seen be- et al. (2000b) and the references therein. For higher Rey-
20
tween the numerical predictions of Sohankar et al. (1998) nolds numbers, Tables 1, 2 and 3 show that the difference
and the present experiments for Reynolds numbers up to in the Strouhal numbers between the two x-locations is
5000 (Fig. 4). The intensification of the wake for angles quite small. Though the power spectra of the velocity
between 0 and 45
can also be inferred from the flow vi- fluctuations depended on the probe location, the position
sualization in Fig. 3. Here the size of the rolled up shear of the spectral peak representing the vortex-shedding
layer decreases for angles greater than zero, in comparison frequency barely changed with the x- and y-coordinates.
to h=0. The deviation from this trend at the highest Rey- Tables 1, 2 and 3 record the Strouhal numbers Su and
nolds number can be related to a faster decay rate of the Sv when the vortex-shedding frequency is obtained by
velocity fluctuations. The decay referred to here is partly locating the peak in the power spectrum of the u and v
due to viscous dissipation. More importantly, the trans- velocities, respectively. For an object symmetric about the
port of kinetic energy from the core of the wake to the
outer flow is Reynolds number dependent (Saha et al. Table 1. Variation of the Strouhal number based on the x- and
2000b). y-components of velocity with orientation of the square cylinder with
Re=1340
4.2 h X=5B X=10B
Time-averaged wake properties
To examine the sensitivity of the measurement of vortex- Su Sv Su Sv
shedding frequency to the probe location, experiments
were carried out at stations x=5 and 10, respectively. The 0 0.141 0.141 0.143 0.143
22.5 0.194 0.191 0.195 0.185
30 0.194 0.186 0.186 0.189
45 0.192 0.192 0.199 0.192
60 0.191 0.195 0.195 0.186

Table 2. Variation of the Strouhal number based on the x- and

y-components of velocity with orientation of the square cylinder with
Re=4490

h X=5B X=10B

Su Sv Su Sv

0 0.143 0.143 0.140 0.141

22.5 0.198 0.198 0.197 0.198
30 0.194 0.186 0.186 0.189
45 0.193 0.192 0.192 0.191
60 0.193 0.194 0.193 0.193

Table 3. Variation of the Strouhal number based on the x- and

y-components of velocity with orientation of the square cylinder with
Re=9980

h X=5B X=10B

Su Sv Su Sv

0 0.143 0.143 0.142 0.142

22.5 0.193 0.192 0.192 0.192
Fig. 4. Variation of the streamwise and the transverse velocity 30 0.190 0.191 0.190 0.190
fluctuations as a function of angle. Measurements carried out at X/ 45 0.190 0.190 0.188 0.189
B=5, Y/B=1.2 for three Reynolds numbers. The error bars indicate the 60 0.189 0.189 0.188 0.188
extent of variation possible in the vicinity of the measurement point
x-axis, it is expected that the two sets of Strouhal numbers The recovery curves are more definite at higher Reynolds
are equal. When the cylinder is oriented at an angle, a loss numbers. The overall trend seen in Fig. 5 is that the
of symmetry is introduced and the two values may be x-velocity decreases initially beyond the base region of the
unequal. Tables 1, 2 and 3 show that the extent of variation wake, while the trend of recovery is initiated in the near
between Su and Sv is also quite small at angles other than wake, beyond x=10. This trend conforms both with ex-
zero. Measurable differences are seen only at the lowest periments as well as computations over a wide range of
Reynolds number studied. The orientations of 0
and 45
Reynolds numbers for an angle h=0 (Lyn et al. 1995). The
are both symmetric configurations, and the two Strouhal distance at which the recovery of the x-velocity is initiated
numbers are nearly equal. The greatest difference is seen at is dependent on the incidence angle of the cylinder. The
Re=1340 for an angle of 30
. The magnitude of the dif- recovery rates in turn are also dependent on the angle,
ference is, however, uniformly small. though in a nonmonotonic manner. The slowest recovery
21
The centerline recovery of the time-averaged velocity is seen for h=45
, while the recovery rates are similar for
and the respective RMS fluctuations are discussed below. the other angles.
Data have been presented beyond the base region of the The recovery of the time-averaged velocity components
cylinder (x>3) for a dimensionless downstream distance of along the centerline is related quite closely to the wake
up to x=20. dynamics. The loss of momentum of the fluid, which
The recovery of the time-averaged x-component of the equivalently appears as forces acting on the cylinder, re-
fluid velocity, the RMS velocity fluctuations, and the cross sults in reduced velocities within the wake. For small
correlation are shown in Fig. 5. Four angles, namely 0, 30, values of x, the flow is dominated by transport of kinetic
45 and 60
have been considered at Reynolds number energy from the core of the wake to the outer flow
Re=9980. Practically identical trends were seen at a Rey- (Townsend 1949). While this factor leads to a decay in the
nolds number of 4990 and have not been repeated. A velocity fluctuations, it can also result in a lowered time-
Reynolds number effect was seen in the data of Re=1340. averaged velocity. Farther away from the cylinder, the
Major observations from Fig. 5 can be summarized as wake size and the centerline velocity increase, indicating
follows. The x-velocity evolves along the centerline and that the wake entrains fluid from the outer flow. The
approaches an asymptotic value of unity, the approach transport of kinetic energy from within to the outer edges
velocity. The far wake tends to be symmetric, irrespective of the wake in the transverse direction generates nonzero
of the object shape. This was found in the wake velocity vertical velocities, even for the symmetric configuration of
profiles used in the determination of the drag coefficient. h=0.For other angles, the transverse velocity is greater, but
Hence the y-velocity approaches zero in the downstream uniformly decreases with increasing distance from the
direction. The data in Fig. 5 show scatter, which was much cylinder.
higher at the lowest Reynolds number of 1340. This trend The decay rates of the RMS velocity fluctuations u¢, v¢,
could possibly be explained by the intrinsic three-dimen- and the time-average u0 v0 are shown in Fig. 5 at Reynolds
sionality and low-frequency unsteadiness of the low Rey- number Re=9980. The velocity scale used for nondimen-
nolds number wake as well as probe interference effects. sionalization is the approach velocity U. The decay rate of
u¢ is marginally higher for the 45
angle, but the overall
patterns in Fig. 5 look similar for all the angles. The decay
rate of v¢ is greater, but is the slowest for the 45
config-
uration. The cross correlation u0 v0 was practically a con-
stant at the lowest Reynolds number. A strong dependence
on the cylinder orientation is seen in Fig. 5 at the highest
Reynolds number, particularly for X/B<10. The magnitude
of the cross correlation, however, decreases in the down-
stream direction in all the experiments. The transverse
profiles of the velocity fluctuations were flat at all mea-
surement stations considered, except the first, at
x=3.These profiles have not been reported in the present
study.
The decay of fluctuations in the wake of the cylinder is
partly due to viscous dissipation, but arises primarily from
the transport of the kinetic energy of the fluctuations in
the transverse direction. The transport is by the large-scale
vortices shed by the cylinder. This mechanism is also re-
sponsible for making the transverse profiles practically
uniform. Marginal differences in the decay rates of the
streamwise fluctuation with orientation (Fig. 5) are related
to (a) the shift in the point of separation, and (b) the
change in the circulation strength of the shed vortices. As
seen in Fig. 3, the separation point moves downstream for
Fig. 5. Centerline recovery of u, u¢ (rms), v¢ (rms), and u0 v0 with all angles greater than zero. The wake size increases
distance at Re=9980 with the angle h, but the delay in the point of separation
 indicates a reduced drag (Fig. 2), a lowered momentum
loss, and hence a smaller velocity deficit in the wake. Since
the strength of the circulation is roughly proportional to
the flow blockage, factors (a) and (b) contribute to the
decay rates of the fluctuations in conflicting terms. As a
result, a definite influence of the cylinder orientation is not
felt in the evolution of the streamwise velocity fluctuations.
The transverse fluctuation essentially responds to changes
taking place in the streamwise component. At h=45
, the
wake deficit is a minimum, which is consistent with that in
the drag coefficient. Hence, the time-averaged velocity
22
undergoes smaller changes with distance and leads to
lower decay rates in v¢.
The time-averaged vorticity profiles at three streamwise
distances are presented in Fig. 6. Data has been presented
for four cylinder orientations. Vorticity values have been
derived by differentiating the time-averaged velocity data.
Reynolds numbers of 1340 and 9980 have been considered.
For all the cylinder positions, the vorticity profiles in Fig. 6
show a maximum and a minimum, on each side of the
centerline (y=0). The magnitude of vorticity at these
points diminishes with the x-distance, though gradually.
The positions of the vorticity extrema are seen to shift with
the streamwise distance for h=0, but for other cylinder
Fig. 6. Time-averaged vorticity profiles as a function of the orienta- angles the positions are practically fixed. This is an
tion of the cylinder for Re=1340 (dashed) and Re=9980 (solid)
evidence of two oppositely oriented vortices being

Fig. 7. Power spectra of the u- and v-compo-

nents of velocity at x=5, 10 for Re=1340 (dashed)
and Re=9980 (solid)
transported in the wake. The vortices are produced at the component shows a similar behaviour. The cross-cor-
corners of the cylinder and have a tendency to get dis- relation function u0 v0 strongly depends on the cylinder
placed inwards because of the lower pressures prevailing orientation before diminishing to zero along the wake
within the wake. This is counter-balanced by the growing centerline.
wake size, which shifts the vortex centers outwards. Thus 3. At x=5, the spectra of both components of velocity are
the vortices move practically along a straight path in the dominated by the contribution of the vortex-shedding
wake. For h=0, the vortices emerge from the forward frequency. The peaks are sharper in the v-component
corners of the cylinder, as opposed to the rear corners for and persist even at x=10.
all other angles. The earlier separation for the noninclined
cylinder leads to a larger pressure deficit between the inner
and the outer regions of the wake, and the vortices get References 23
visibly displaced toward the axis. Once the shed vortices Bearman PW, Trueman DM (1972) An investigation of the flow
come close to one another, they can be expected to merge around rectangular cylinders. Aeron Quart 23:229–237
Chew YT, Simpson RL (1988) An explicit non-real time data reduction
and move as a single entity. method of triple sensors hot-wire anemometer in three-dimen-
sional flow. ASME J Fluids Eng 110:110–119
4.3 Courchense J, Laneville A (1979) A comparison of correction methods
Power spectra used in the evaluation of drag coefficient measurements for two
dimensional rectangular cylinders. ASME J Fluids Eng 101:506–
The dimensionless power spectra as a function of fre- 510
quency are shown in Fig. 7 for two Reynolds numbers of Davis RW, Moore EF (1982) A numerical study of vortex shedding
the present study. At the lowest Reynolds number, a few from rectangles. J Fluid Mech 116:475–506
harmonics away from the peak were seen to be excited. No Durao DFG, Heitor MV, Pereira JCF (1988) Measurements of turbu-
significant differences were otherwise seen in the shape of lent and periodic flows around a square cross-section cylinder.
Exp Fluids 6:298–304
the spectra with a change in the Reynolds number. Four Franke R, Rodi W, Schonugh B (1990) Numerical calculation of
angles of inclination have been included in Fig. 7 for laminar vortex shedding flow past cylinders. J Wind Eng Ind
comparison. The measured power spectra have been Aerodyn 35:237–257
nondimensionalized by the total area under the curve. Igarashi T (1987) Fluid flow and heat transfer around rectangular
cylinders (the case of a width/height ratio of a section of 033-15).
Figure 7 shows that the peaks of vortex shedding di- Int J Heat Mass Trans 30:893–901
minish rapidly with distance in the spectrum of the u-ve- Kelkar KM, Patankar SV (1993) Numerical prediction of vortex
locity, while they persist in the v-component of velocity. shedding behind a square cylinder. Int J Num Meth Fluids 14:327–
Spectra in the v-component of velocity recorded at x=18 341
continued to show peaks corresponding to vortex shedding. Knisely CW (1990) Strouhal numbers of rectangular cylinders at in-
cidence: a review and new data. J Fluids Struct 4:371–393
A survey of the spectra in Fig. 7 shows that the vortex- Lyn DA, Einav S, Rodi W, Park JH (1995) A laser-Doppler velocimetry
shedding frequency is only a weak function of the angle h. study of ensemble-averaged characteristics of turbulent near wake
The inertial range of the spectra representing a fully de- of a square cylinder. J Fluid Mech 304:285–319
veloped turbulent profile was seen for x>10 in the u-velocity Nakaguchi H, Hashimoto K, Muto S (1968) An experimental study of
aerodynamic drag of rectangular cylinders. J Jpn Soc Aerospace
and for x>20 in the v-velocity, with the signature of vortex Sci 16:1–5
shedding fully suppressed. The influence of the cylinder Norberg C (1993) Flow around rectangular cylinders: pressure forces
orientation was found to be of secondary importance. and wake frequencies. J Wind Eng Ind Aerodyn 49:187–196
Okajima A (1982) Strouhal number of rectangular cylinders. J Fluid
Mech 123:379–398
5 Okajima A, Nagahisa T, Rokugoh A (1990) A numerical analysis of
Conclusions flow around rectangular cylinders. JSME Int J Series II 33:702–711
An experimental study of flow past of a square cylinder at Saha AK, Muralidhar K, Biswas G (2000a) Experimental study of flow
intermediate Reynolds numbers and various inclinations past a square cylinder at high Reynolds numbers. Exp Fluids
has been reported. The following conclusions have been Saha29:553–563
AK, Muralidhar K, Biswas G (2000b) Vortex structures and ki-
arrived at in the present work: netic energy budget in two-dimensional flow past a square cyl-
inder. Comput Fluids 29:669–694
1. For inclinations greater than 0
, the points of separa- Sohankar A, Norberg C, Davidson L (1998) Low Reynolds number
tion move downstream, the wake size increases, but the flow around a square cylinder at incidence: study of blockage,
separated shear layer rolls up over a shorter distance. onset of vortex shedding and outlet boundary condition. Int J
These factors lead to a reduced drag coefficient and a Numer Meth Fluids 26:39–56
Swaminathan MK, Rankin GW, Sridhar K (1986) Evaluation of the
higher Strouhal number. basic systems of equations for turbulence measurements using a
2. The velocity fluctuations decay in the downstream di- Monte Carlo technique. J Fluid Mech 170:1–9
rection. The decay rates are the slowest for h=45
. The Townsend AA (1949) The fully developed turbulent wake of a circular
centerline recovery of the time-averaged x-velocity cylinder. Aust J Sci Res 2:451–468