Flow Measurement and Instrumentation 55 (2017) 23–36

Contents lists available at ScienceDirect

Flow Measurement and Instrumentation

journal homepage: [Link]/locate/flowmeasinst

Comparison of flow characteristics around an equilateral triangular cylinder MARK

via PIV and Large Eddy Simulation methods
⁎
Sercan Yagmur, Sercan Dogan, Muharrem Hilmi Aksoy, Ilker Goktepeli , Muammer Ozgoren
Selcuk University, Faculty of Engineering, Department of Mechanical Engineering, 42075 Selcuklu, Konya, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: The flow structures around an equilateral triangular cylinder, which is commonly used as a vortex shedder in the
Equilateral triangular cylinder vortex flowmeter, were investigated experimentally and numerically. Flow characteristics such as vorticity
Large Eddy Simulation (LES) contours, patterns of sectional streamlines, velocity vectors, velocity fields, Reynolds stress correlations, Strouhal
Particle Image Velocimetry (PIV) numbers and drag coefficients were examined using the Particle Image Velocimetry (PIV) technique and the
Strouhal number
Large Eddy Simulation (LES) turbulence model. Experimental studies were performed in an open water channel
Vortex flowmeter
Vortex shedding
for Re=2.9×103, Re=5.8×103 and Re=1.16×104 based on the equilateral triangle edge. A sharp-tip corner of
the cylinder with a triangle cross-section was exposed to the upstream side while the other two sharp-tip corners
were placed on the downstream side. Numerical studies were also completed at Reynolds numbers in the range
of 2.9×103≤Re≤1.16×105 to obtain the changes in the Strouhal numbers and drag coefficients. When the
results of PIV and LES are considered in the same interval of Reynolds numbers, the maximum and minimum
values of each flow pattern were nearly the same. The time-averaged patterns had considerable symmetry with
respect to the axis line passing through the sharp-tip corner of the cross-section of the triangular cylinder. The
Strouhal number was independent of the Reynolds number and was found to be approximately 0.22. The drag
coefficient decreased with increasing Reynolds numbers while increasing the Power Spectral Density (PSD) and
the vortex shedding frequency. For the same Reynolds numbers, the experimental and numerical results were in
good agreement. Therefore, the LES turbulence model is recommended for applications of flow around this type
of bluff body that is generally used in the design of vortex flowmeters to generate vortex shedding.

1. Introduction vortex shedding in a vortex flowmeter, different cross-sectional cylin-

ders and their various arrangements have been used for decades. For
Fluid flow around bluff bodies (e.g., circular, rectangular, triangular generating better vortex shedding, because of sharp corners, a prismatic
cross-sectional cylinders) has been studied by researchers for centuries cross-sectional cylinder is preferred to a circular cylinder. For instance,
because of its importance in various engineering applications such as Ozgoren [3] investigated flow characteristics of flow around circular,
flow measurements, electronic cooling, heat exchanger systems, bridge square and oriented square cylinders at a 45° orientation by using
piers, high-rise buildings, flow dividers, sensors and probes among Particle Image Velocimetry (PIV) techniques at 550≤Re≤3.4×103.
other applications [1]. One of these industrial application areas is the The author obtained Strouhal numbers such as St=0.204–0.212 for the
flow measurement by vortex flowmeters that could be used indepen- circular cylinder, St=0.12–0.134 for the square cylinder and
dently of the fluid phase. The fundamental working principle of these St=0.165–0.174 for the oriented square cylinder at
instruments is the measurement of the vortex shedding frequency, 550≤Re≤3.4×103. It was concluded that the values of the vortex
which is referred to as von Karman Vortex Street in the wake region of shedding frequency were nearly same. However, the peak frequencies
bluff bodies. As the vortex shedding frequency and free stream velocity of the Power Spectral Density (PSD) changed with the cross-sections of
change proportionally like the von Karman Vortex Street, the Strouhal the cylinder. In a water tank, Okajima [4] experimentally investigated
number remains constant [2]. the flow around the square and the rectangular cylinders for
Various methods are used for the vortex flowmeter to determine the 70≤Re≤2×104 to define the vortex shedding frequencies and the
flow rate. A well-known method includes pressure sensors near the corresponding Strouhal numbers. At values higher than Re=104, the
wake of the bluff body to obtain the vortex frequency. To generate Strouhal number was not the function of the Reynolds number.

⁎
Corresponding author.
E-mail addresses: syagmur@[Link] (S. Yagmur), sercandogan@[Link] (S. Dogan), muharremaksoy@[Link] (M.H. Aksoy),
ilkergoktepeli@[Link] (I. Goktepeli), mozgoren@[Link] (M. Ozgoren).

[Link]
Received 12 February 2016; Received in revised form 27 February 2017; Accepted 13 April 2017
Available online 14 April 2017
0955-5986/ © 2017 Elsevier Ltd. All rights reserved.
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

However, the cross-section of the cylinder was the function of the accepted by the literature, CFD is a useful method for a design process
Strouhal number, and the cross-section varied with the width-to-height and significantly improves the efficiency of newly designed and
ratio of the cylinder. Knauss et al. [5] measured the vortex shedding manufactured industrial products. Furthermore, CFD analysis is a
frequency in the wake region of elliptical and square cylinders. They cheaper and more effective way to ensure precise quantitative output
determined that the Strouhal number depended on the geometrical into the aerodynamic and hydrodynamic flow characteristics of flow
shape of the bluff body, and the Strouhal number was almost not around a bluff body [21,22].
influenced by the Reynolds number. Yen and Yang [6] studied the flow The overall objective of the present investigation is to visualize flow
over a square cylinder in the range of 4×103≤Re≤3.6×104 with characteristics and obtain the vortex shedding frequency in the wake of
incidence angles of 0°≤θ≤45°. The minimum and maximum Strouhal a triangular cross-sectional cylinder model via PIV technique and the
numbers occurred at θ=0° and θ=15°, respectively. Another experi- CFD method using a LES turbulence model in ANSYS-Fluent. The LES
mental study on the comparison of flow characteristics of circular turbulence model has been used in the literature because the LES
cylinder and sphere was performed by Ozgoren et al. [7] at Re=5×103 turbulence model yielded results similar with respect to the experi-
and Re=10×103. They concluded that the wake flow of the sphere was mental results in some different fluid flow applications presented by
more complex than the wake flow of the circular cylinder because of the Hassanzadeh et al. [23], Gang et al. [24], Luo et al. [25] and Zhang
3-D flow. The vortex shedding frequency of the sphere was found to be et al. [26]. Although many experimental and numerical studies about
higher than the vortex shedding frequency of the circular cylinder, but the flow over bluff bodies have been performed, it is difficult to locate
the PSD of the sphere was lower than the PSD of the circular cylinder. In detailed studies of the flow characteristics and spectral analysis of a
addition to these studies, various investigations of the circular and horizontal axis triangular cylinder in the literature comparing PIV and
square cylinders were experimentally performed by Huang et al. [8], Gu the LES turbulence model. This situation has motivated us to investigate
et al. [9], Lee and Lee [10], Kalmbach and Breuer [11] and Gunes et al. the flow across an equilateral triangular cylinder in a steady flow
[12]. They considered the effect of geometry and flow control method regime. The main aim of this study was to confirm the numerical LES
on the vortex shedding frequency in their studies. turbulence model results with compared to the experimental results in
However, investigation of the vortex shedding frequency and flow terms of the vortex shedding frequency to contribute to the measure-
characteristics over bluff bodies is a common research field in fluid ment principle of the vortex flowmeters, instantaneous and time-
mechanics, but limited studies of the flow around triangular cross- averaged flow patterns by using the PIV technique and the LES
sectional cylinder-like bodies have been performed. Agrawal et al. [13], turbulence model at Re=2.9×103, Re=5.8×103 and Re=1.16×104.
El-Sherbiny [14] and Iungo and Buresti [15] determined the flow Furthermore, the concurrence between the results of PIV and LES leads
characteristics of different orientations affecting a triangular prism us to obtain the Strouhal numbers independently of the Reynolds
using various experimental flow visualization techniques. They decided number and the variation in the drag coefficient. For this reason, the
that the apex angle influenced the vortex shedding frequency and the range for the Reynolds numbers was expanded to Re=1.16×105 for the
Strouhal number. Another experimental study of the flow over a dual numerical studies.
triangulate body in a circular pipe was performed by Peng et al. [16] to
investigate the Strouhal number by using a piezoelectric sensor to 2. Methods
acquire the vortex shedding frequency. The Strouhal numbers changed
with the distance between the two bluff bodies. As a result, the 2.1. PIV measurement
Reynolds number did not have very much effect on the Strouhal
number. A study of the vortex shedding frequency of a triangular Experiments were carried out in a large-scale open water channel
cylinder was completed by Zhang et al. [17] to investigate the effect of with a rectangular cross-section and the dimensions of
the fluid type on the Strouhal number at the same Reynolds number. 770 mm×600 mm×6000 mm for width, height and length in the
The Strouhal number was found to be nearly the same for water and air Advanced Technology Research and Application Center of Selcuk
at the same Reynolds number and the Strouhal number was not affected University in Turkey. The PIV system is depicted schematically in
by the Reynolds number in the same manner as the circular and Fig. 1. To simplify laser transmission and flow visualization, test-section
different cross-sectional cylinders. However, the PSD and vortex walls were constructed from 15 mm thick transparent glass plates, and
shedding frequency were higher when air was chosen as the fluid the tank was filled with water to a level of hw=475 mm. The water was
because of the high velocity required to provide Reynolds number pumped by a centrifugal pump controlled with a frequency converter,
equality. and before reaching the test section, the water passed through a
Due to reliable results with better validation performance between honeycomb section and a two-to-one channel contraction. The uncer-
experimental and numerical studies in recent years, Computational tainty of free-stream turbulence intensity was less than 1% in the range
Fluid Dynamics (CFD) has been used in a wide range of research areas of the current Reynolds numbers, Re=(U∞ L)/v, based on the char-
as an alternative to experimental and analytical studies. For example, acteristic edge. Here, L is the side edge of the equilateral triangular
experimental and numerical studies on the comparison of a triangular cylinder, v is the kinematic viscosity and U∞ is the free-stream velocity
cylinder with trapezoidal and other cross-sectional cylinders were varied from 58 mm/s to 232 mm/s. The triangle was made from acrylic
performed by Cheng and Liu [18] and Venugopal et al. [19]. Both glass with an edge dimension of 50 mm and a 5 mm wall thickness. The
studies stated that the experimental results were in good agreement triangle is a hollow, and it is filled with water to diminish the laser
with the numerical results, and the vortex shedding frequency of a sheet deflection and highly polished to avoid surface roughness. The
trapezoidal cross-sectional cylinder was higher than the vortex shed- length of the triangle cross-section cylinder is 600 mm. The Nd: YAG
ding frequency of a triangular cylinder. Zhang and Blair [2] compared laser with a maximum of 15 Hz was used to generate a laser sheet
the flow characteristics and the Strouhal number around a triangular perpendicular to the axis of the cylinder. The thickness of the laser
cylinder via experimental and CFD results at Re=4.5×104, and they sheet was approximately 1 mm. The 10 µm-diameter suspended seeding
decided that experimental and numerical results indicated a close particles in the flow were silver-coated hollow glass spheres. The
agreement with ± 2.7% relative error. Khaledi and Andersson [20] equilateral triangular cylinder is horizontally located at the middle
performed a numerical investigation of the vortex shedding from face section of the water channel between two walls to avoid surface and
and corner orientations of a hexagonal cylinder at 100≤Re≤103. They wall effects from the channel. Moreover, to fulfill the high-image
stated that the corner orientation provided lower Strouhal numbers. density criterion, the interrogation area contains nearly 20–30 particles
Although the Strouhal number increased from Re=100 to Re=500, the per image. A Complementary Metal–Oxide–Semiconductor (CMOS)
Strouhal number remained constant from Re=500 to Re=103. As camera with a resolution of 1632×1200 pixels was used to capture

24
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

∂ ui ∂ ui uj 1 ∂p ∂τij ∂ 2ui
+ = − +v
∂t ∂xj ρ ∂xi ∂xj ∂xi ∂xj (2)
where ui is the filtered velocity component along the Cartesian
coordinates xi, ρ is the density of fluid and p is the pressure. The main
principle of the LES is low-pass filtering. When this operation is applied
to the Navier–Stokes equations, large-scale eddies are computed
directly, whereas the small scale eddies are modeled using a sub-grid
scale (SGS) model.
The influence of the small scales on the large (resolved) scales takes
place through the sub-grid scale stress as given by [30]:
τij = ρ(uiuj − uiuj ) (3)
resulting from the filtering operation, which is unknown and must be
modeled with a sub-grid scale model (SGS) occurring in several sub-
models. SGS is based on the eddy viscosity approach and is defined by
[30]:
1
τij − τkkδij = − 2μt Sij
3 (4)
where the trace of the sub-grid scale stresses τkk is incorporated into the
pressure resulting in a filtered isotropic pressure term; μt is the sub-grid-
scale turbulent viscosity, and Sij is the resolved scale of the strain rate
tensor and described by [30]:

1 ⎛ ∂ui ∂uj ⎞
Sij = ⎜⎜ + ⎟⎟
2 ⎝ ∂xj ∂xi ⎠ (5)
Fig. 1. Schematic view of the PIV system experimental setup, with laser illumination for
an equilateral triangular cylinder located in a uniform flow condition. The simplest and well-known sub-grid scale models are the
Smagorinsky and later Lilly models. In both of these models, the
2048 instantaneous image pairs at the rate of 15 Hz to calculate 1024 eddy-viscosity is characterized as [30]:
vector fields. μt = ρL s2 Sij (6)
An interrogation window of 32×32 pixels for each image was
selected and converted to approximately 1.41×1.41 mm2 grid size In Eq. (6), Ls is the length of mixing for the sub-grid scales, and
(0.028Lx0.028L) with adaptive correlations consisting of 7474 when the resolved scale of the strain rate tensor is defined by Eq. (7)
(101×74) velocity vectors. During the adaptive-correlation process, [30]:
an overlap of 50% was used for all interrogation areas to satisfy the
Sij = SijSij Ls (7)
Nyquist criterion. From instantaneous velocity vector fields, the time-
averaged flow patterns were calculated. For the post processing, Dantec The mixing length for the sub-grid scales is computed as [30]:
Dynamic Studio software was employed including proper filters to
compute the raw displacement vector field from the particle image Ls = min (κ , d , CsV 1/3) (8)
data.
In this equation, κ is the von Karman constant, d is the interval to
As explained by Ozgoren [3], the seeding particle size, particle
the closest wall, Cs is the Smagorinsky constant, and V is the computa-
overlap, non-uniform particle distribution, unmatched particle correla-
tional cell volume [30].
tions, out of laser plane motion bias and random errors, interrogation
window size, and electronic and optical imaging noise were the main
2.2.2. Computational domain and boundary conditions
uncertainty factors for velocity measurement and vorticity calculation
In the present study, the computational 3-D water flow domain was
in the PIV method. In the uncertainty evaluation of the velocity
prepared by considering the PIV measurement field. The center of the
measurement, the influence of these factor calculations was determined
flow domain was set at the sharp-tip of the triangle cross-section. The
to be in the range of ± 2–5%, and for detailed information, the studies
flow domain dimensions are −6.0L≤x≤24.0L in the streamwise
of Ozgoren [3], Adrian [27], Fouras and Soria [28], and Hart [29] can
direction, −4.75L≤y≤4.75L in the vertical direction and
be examined.
−6.0L≤z≤6.0L in the transverse direction, as shown in Fig. 2.
On the surfaces of the solid model, a no-slip boundary condition is
2.2. Large Eddy Simulation (LES) defined, and the pressure is set to zero. At the inlet boundary condition,
the velocity inlet is described with a uniform velocity and zero pressure
2.2.1. Mathematical models
For the numerical studies, the turbulent characteristics have been
computed in various ways in the literature. In this study, the LES
turbulence model, which has been associated with the ANSYS-Fluent
software package, was used to simulate the flow field around the
equilateral triangular cylinder. The governing equations for an incom-
pressible flow form of the continuity and the Reynolds Averaged
Navier-Stokes equations are specified as follows [30]:

∂ ui
=0
∂xi (1) Fig. 2. Flow domain and boundary conditions in the present study.

25
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Table 1
Grid independence study on the drag coefficient
(CD) at Re=1.16×104.

Grid size CD

2.57×106 1.582
5.68×106 1.518
9.35×106 1.443
18.28×106 1.430

Reynolds number. With respect to the results in Table 1, due to the

considerable computer capacity, analyses were performed with 9.3
million grid numbers for all Reynolds numbers.
In the LES turbulence model, the law-of-the-wall approach is
applied using the wall boundary condition. The first cell from a no-
slip wall is the symbolized dimensionless distance y+ and is character-
ized as:
τw / ρy
y+ =
v (10)
In this equation, τω and y express the shear stress and first cell
Fig. 3. Non-uniform grid structure near the body surface on the x-y plane. height, respectively. In the current study, y+ has ranged from 0.83 to
1.72. It is generally accepted that y+ should be less than 1 (y+ < 1) for
gradients. At the outlet boundary condition, the pressure outlet is most usage at low-Reynolds numbers, as given by Stringer et al. [36].
specified as zero. At the top, bottom and side boundaries, the free-slip However, a study was completed at a high Reynolds number by
conditions are implemented. The finite volume method is integrated to Jagadeesh and Murali [37], who concluded that the mesh structure of
solve the transient incompressible Navier–Stokes equation with the y+ < 2 with 5 cells in the boundary layer was sufficient to obtain
pressure and velocity coupling with the splitting of the coupled-type correct results.
algorithm. In the LES turbulence model, physical diffusion is remark-
ably affected by numerical diffusion; consequently, the central differ- 3. Results and discussion
encing schemes should be a suitable approach to conduct the spatial
discretization. Therefore, the second order implicit scheme was selected The flow patterns obtained by experimental and numerical analyses
for pressure and momentum spatial discretization with the Green-Gauss around the equilateral triangular cylinder are compared at
solver. For time integration bounded central differencing format is used Re=2.9×103, Re=5.8×103 and Re=1.16×104 in Figs. 4–12. The
[24,25,31–35]. For the transient analyses, to match PIV frequency, a dimensions of all images were normalized with the characteristic edge
constant time step Δt=0.066 s was specified during the solution with length of the triangle as x/L and y/L. The maximum and minimum
20 sub-iterations for each step. In addition, the convergence of each values are given as a legend bar for the contours in each figure. The
step was less than 10−6 for all equations to be solved. positive layers of the flow patterns are displayed as continuous line
contours with a red background color, while the negative layers are
displayed as dashed line contours with a blue background color.
2.2.3. Mesh arrangement
The flow domain is known to require enough grids to solve a
3.1. Flow characteristics for instantaneous and time-averaged vorticity
complex flow structure. The non-uniform grid structure in the vicinity
contours
of the equilateral triangular cylinder is shown in Fig. 3. Considering the
duration of the analyses and the computing capacity, the flow domain
The instantaneous and time-averaged vorticities were normalized
was divided into parts. To determine the shear layer effect and flow
with L/U∞. In Fig. 4, the instantaneous vorticity (ω*=ω L/U∞) results
characteristics in detail, the finer grids were built near the face and
clarify the von Karman vortex shedding. PIV and LES results show that
wake region of the equilateral triangular cylinder. Grid size coarser
flow separation occurs from sharp-tip corners for all Reynolds numbers.
than a critical value is not able to capture local variations of the model
Because of the shear layers, flow motion to the back of the equilateral
and flow domain on scales greater than the element size. Thus, the
triangular cylinder coming from the upper edge slants to the clockwise
time-averaged simulation results of the LES model do not coincide with
direction, while flow around the lower edge of the body rotates in the
experimental results even though convergence is occurring. For these
counter clockwise direction. After the separation of the flow from the
reasons, a grid independence study was completed by examining effects
body surface, small scale eddies roll up to form in the wake region. The
of the mesh density on the value of the mean drag force coefficients. In
Reynolds number plays an important role in the flow structure.
identifying the regular and irregular forces acting on the body, flow
Instantaneous vorticity patterns decrease in the formation length of
characteristics play an important role. These types of forces can cause
the large-scale Karman vortex with increasing Reynolds number.
structural damage. One of these forces is the drag force that is acting on
However, these patterns are not clearly observed because of the closer
the bluff body in the streamwise direction, and the drag force can be
values of the Reynolds number. Although the PIV and LES results are in
expressed by the dimensionless drag coefficient as follows:
good agreement in terms of instantaneous vorticity structure, the
2FD agreement of PIV and LES results clearly seems to be better for time-
CD = averaged vorticity patterns (< ω*=ω L/U∞ >) with maximum and
ρU∞2A (9)
minimum values in Fig. 5. Concerning the beginning and improvement
where FD is the drag force in the streamwise direction, ρ is fluid density of small-scale vorticity structures in the separating shear layer, the
and A is the projection area of the front view. The drag coefficient is small clusters of vorticity concentration are not visible in the patterns of
shown to be changing with the grid number in Table 1. With an instantaneous vorticity for all Reynolds numbers in Fig. 4. The
increasing grid number, the difference is decreasing at the same minimum positive and negative values of the vorticity contours were

26
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 4. Comparison of the normalized experimental and numerical instantaneous vorticity (ω*) around the equilateral triangular cylinder.

observed near the symmetry plane, which is definition of vortex triangular cylinder for all Reynolds numbers. However, the longer size
formation length, at Reynolds numbers of 2.9×103, 5.8×103 and of the velocity vectors can be observed along the shear layer. From
1.16×104. The form of the vorticity contours in the near-wake is Re=2.9×103 to Re=1.16×104, relatively longer velocity vectors were
similar for all Reynolds numbers. A large-scale vortex occurs in the formed in the downstream of the body.
downstream for Re=2.9×103. While increasing the Reynolds number, As emphasized in Fig. 5, from time-averaged vorticity patterns, the
it moves to the upstream and approaches the back side of the triangular streamline patterns (< ψ >) of the triangular cylinder demonstrate that
cylinder. In addition, in the part of the downstream region of the the flow structures in the wake region are almost equally symmetrical
triangle, the time-averaged vorticity patterns are nearly equal to zero with respect to the centerline in Fig. 7. There are two symmetrical focus
because the direction of the vortex shedding changes continuously. points F1 and F2 with a similar size and cycling in opposite directions,
Therefore, the time-averaged vorticity patterns show that the von and a saddle point S (evident for intersections of streamlines) is formed
Karman Vortex Streets observed in the instantaneous vorticity images clearly in the downstream of the wake. Saddle points in the wake region
forming small-scale vortices have disappeared. agree with the time-averaged vorticity contours in the PIV and LES
results for all Reynolds numbers. The saddle point could also ensure a
3.2. Comparison of streamline topology and the corresponding velocity demonstration of the main formation length of the large-scale vortices.
contours Saddle point distances normalized as x/L were approximately measured
from the back side of the equilateral triangular cylinder at
A reversed flow has occurred in the wake region as displayed by the Re=2.9×103, Re=5.8×103 and Re=1.16×104 as 1.70/1.75, 1.65/
velocity vectors (< V >) in Fig. 6. Velocity vectors with very small 1.70 and 1.60/1.65 by PIV/LES, respectively.
magnitude have been established in the wake region of the equilateral The distributions of the dimensionless time-averaged streamwise

Fig. 5. Comparison of the normalized experimental and numerical time-averaged vorticity (< ω* >) around the equilateral triangular cylinder.

27
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 6. Comparison of the experimental and numerical velocity field (< V >) around the equilateral triangular cylinder.

velocity components (< u*= u/U∞ >) are illustrated in Fig. 8. Time- tively. Stagnation points can be attributed to the saddle (S) point from
averaged streamwise velocity was normalized with free stream velocity streamline topology in Fig. 7.
(U∞). Both PIV and LES results indicate the same minimum and The time-averaged cross-stream velocity contours (< v*= v/U∞ >)
maximum levels of the dimensionless values of streamwise velocity are presented in Fig. 9. The time-averaged maximum positive and
contours at Re=2.9×103, Re=5.8×103 and Re=1.16×104. The flow negative values of cross-stream velocity were observed on the upper
clearly accelerates after passing the sharp-tip corners and reaches the and down sides of the equilateral triangular cylinder. The peaks of
maximum value. The flow acceleration was observed in the region near positive and negative values for PIV and LES increased with increasing
the sharp-tip corners, and this region shrank due to the increase in the Reynolds number. After the flow passing the equilateral triangular
Reynolds number. It is also possible to describe the location of the cylinder, two different pairs of clusters were developed in the down-
stagnation point along the wake symmetry plane. From the PIV and the stream of the wake. The intersection of negative and positive cluster
LES results, the distance from the rear surface of the triangle to the cross-stream velocity (< v* >) in the wake of the equilateral triangular
stagnation point at Re=2.9×103 is approximately 1.7L. With increas- cylinder close to rear surface of the triangle have relatively larger
ing Reynolds number, the growth of the separated shear layers from the values with increasing Reynolds numbers. From the PIV and LES
tip corner of the triangular cylinder shoulders is associated with a results, the small size of the positive and large size of the negative
decrease in the wake size. Therefore, the location of the stagnation cross-stream velocity contours are formed in the half-upper side of the
point approached the rear surface of the equilateral triangular cylinder wake region. In addition, just the opposite of this case is also formed in
and for PIV and LES results, indicated that the stagnation points are the half-lower side of the wake region for both Reynolds numbers along
nearly 1.65L and 1.60L at Re=5.8×103 and Re=1.16×104, respec- the axis line where there is a ‘‘switching’’ of the orientation from

Fig. 7. Comparison of the experimental and numerical streamline topology (< ψ >) around the equilateral triangular cylinder.

28
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 8. Comparison of the normalized experimental and numerical time-averaged streamwise velocity (< u* >) around the equilateral triangular cylinder.

positive to negative values. The arrangement of the cross-stream the wake region for all cases that are placed nearly 1.15L at
velocity, (< v* >), indicates two negative and positive peak points Re=2.9×103, 1.275L at Re=5.8×103 and 1.4L at Re=1.16×104 for
downstream of the equilateral triangular cylinder with the same both experimental and numerical analyses. A possible explanation for
intervals of PIV and LES results: ± 0.29 at Re=2.9×103, ± 0.36 at these peak points can be related to the foci F1 and F2 (centers of
Re=5.8×103 and ± 0.38 at Re=1.16×104 demonstrate symmetrical vortices) regions in Fig. 7, where the streamwise velocity continuously
flow structures in the wake region. fluctuates between the minimum and maximum values. The maximum
values of rms cross-stream velocity are nearly same from the results of
PIV and LES for each Reynolds number. Although double peaks occur
3.3. Variations of time-averaged fluctuating velocity patterns for the rms streamwise velocity along the axis line, only one peak is
demonstrated in the wake of the rms cross-stream velocity (< v*rms >)
The time-averaged normalized rms streamwise (< u*rms= urms/ on the axis line in Fig. 11. The locations of peak points can be associated
U∞ >) and cross-stream wise velocity (< v*rms=vrms/U∞ >) are de- with the stagnation point in the wake of the equilateral triangular
monstrated in Figs. 10 and 11, respectively. They are not affected by cylinder.
Reynolds number and display nearly the same attitude for both PIV and
LES results. In Fig. 10, rms streamwise velocity (< u*rms >) demon-
strates that symmetrical double peaks are formed along the axis line in

Fig. 9. Comparison of the normalized experimental and numerical time-averaged cross-stream velocity (< v* >) around the equilateral triangular cylinder.

29
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 10. Comparison of the normalized experimental and numerical rms streamwise velocity fluctuations (< u*rms >) around the equilateral triangular cylinder.

3.4. Variations of Reynolds stress correlations nearly 1.2L in the wake region at Re=2.9×103, Re=5.8×103 and at
Re=1.16×104. The large-scaled clusters moved further upstream and
Reynolds stress correlations (< u'v'/U∞2 >) indicate the compar- are positioned near to the base of the equilateral triangular cylinder
ison of momentum transfer at three Reynolds numbers for both analyses with increasing Reynolds number. In addition to this movement, the
in Fig. 12. Two clusters with small and large scales are formed at positive and negative peak values decrease from ± 0.15 at
Reynolds stress correlation images that are symmetrical along the axis Re=2.9×103 to ± 0.125 at Re=5.8×103. From Re=5.8×103 to
line. The Reynolds stress correlations in and around the mean wake Re=1.16×104, the peaks of negative and positive Reynolds stress have
recirculation region are mainly because of the time-dependent behavior increased from ± 0.125 to ± 0.13.
of the vortex shedding. In this context, as shown in time-averaged
vorticity counters in Fig. 5, the Reynolds stress correlation is nearly
zero in the outermost region of the flow fields. The small scaled clusters 3.5. Variations of time-averaged normalized streamwise velocity and
are positioned at the rear surface of the equilateral triangular cylinder Reynolds stress distributions along the vertical lines downstream of the
and the large-scaled clusters occur on the back side of the small ones. equilateral triangular cylinder
Because of the fluctuations in the shear layers, the small-scale Reynolds
stress region is close to the back side of the equilateral triangular Variations of the time-averaged streamwise velocity component
cylinder which occurs because of the flow entrainment into the wake (< u* >) in Fig. 8 and Reynolds stress correlations (< u'v'/U∞2 >) in
region. The similarity of the PIV and LES results shows that the center Fig. 12 for all studied Reynolds numbers are described along the
of the maximum negative and positive large-scaled clusters is placed vertical stations as ‘‘K, L, M, N and O’’ which are 0.25L, 0.5L, 1.0L, 1.5L
and 2.0L, respectively, away from the rear surface of the triangle along

Fig. 11. Comparison of the normalized experimental and numerical rms cross-stream velocity fluctuations (< v*rms >) around the equilateral triangular cylinder.

30
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 12. Comparison of the normalized experimental and numerical Reynolds stress correlations (< u'v'/U∞2 >) around the equilateral triangular cylinder.

the +x axis. The distributions of the time-averaged streamwise velocity symmetrical flow structure and achieve its maximum positive and
component (< u* >) in Fig. 13 and the Reynolds stress correlations negative values in the upper and lower regions of the axis line. For that
(< u'v'/U∞2 >) in Fig. 14 are given downstream of the equilateral reason, we can conclude that the Reynolds stress correlations are nearly
triangular cylinder for different cross-sections in the interval of 0.25L to zero in the free stream flow region. However, the Reynolds stress
2.0L, the first row at Re=2.9×103, the second row at Re=5.8×103 correlations obtain their maximum positive and negative values
and the bottom row at Re=1.16×104. The red and blue lines represent between the stations N and O associated with the rms of the velocity
the PIV and the LES results, respectively. distributions in Figs. 10 and 11.
It is apparent that the streamwise velocities (< u* >) are nearly
zero in the shear layer between the back side of the equilateral
triangular cylinder and the vertical station K for all Reynolds numbers. 3.6. Spectral analysis and calculation of vortex shedding frequency
Because of the low-pressure region in the wake of the equilateral
triangular cylinder, negative velocity components occur from station K Mass flow rate measurement plays an important role in terms of
to O, and the maximum negative values are obtained nearly at the scientific studies and industrial applications due to its stability under
vertical station of M for all Reynolds numbers. The experimental and different temperature and pressure conditions compared to volume.
numerical results demonstrate that the stagnation point in the wake of Mass flow rate is also obtained from the multiplication of volumetric
the triangle occurs between the stations of N and O for all Reynolds flow rate and fluid density. There are two options to attain the density.
numbers, but it approaches the station N with increasing Reynolds Fluid density is obtained with respect to measured medium conditions
numbers. The maximum values are reached at y/L= ± 0.6 for vertical such as temperature, pressure etc. as a first option and a densitometer is
stations of K, L and M at all Reynolds numbers. However, values of considered for the second option. Nevertheless, the desired accuracy
streamwise velocities approach the uniform flow conditions at the may not be provided as a result of the measurements. Although Coriolis
vertical stations N and O at y/L= ± 0.6 in Fig. 13. The minimum and and thermal flowmeters are utilized to measure the mass flow rate
maximum values of the time-averaged streamwise velocity and the directly, these devices remain incapable for some practical applications.
distribution of its contour fields through the vertical lines for PIV are in For this reason, the vortex shedding principle is accepted as an
close agreement with LES. Both methods clearly show the symmetrical alternative method to measure the mass flow rate by using a vortex
flow structure in the wake region. Moreover, the difference between flowmeter. In this content, the vortex flowmeter benefits from a bluff
PIV and LES was found to be less than ± 5% for the maximum values of body to generate the von Karman vortex shedding. In this study, an
streamwise velocity contours in Fig. 13. equilateral triangular cylinder model has been utilized as a bluff body
As shown in Fig. 14, the Reynolds stress correlations (< u'v'/U∞2 >) or a vortex shedder, and flow around this geometry led to fluctuations
fluctuate due to small and large scaled clusters in the rear surface of the in the wake region. Moreover, these fluctuations have certain frequency
triangular cylinder. In Fig. 14, it is more visible that the small scaled values with respect to flow conditions and these frequency values have
clusters disappear after the station of M. Furthermore, the diminishing been obtained both experimentally and numerically. Thus, the mass
of the small scaled clusters occurs earlier along the vertical location of flow rate is defined by evaluating the frequency values.
M at Re=1.16×104 than at Re=5.8×103. In the vertical direction In the present study, experimental and numerical results have been
after the location of approximately y/L= ± 0.7, the values of the obtained and then compared. The vortex shedding frequency of the
Reynolds stress correlations become zero for the examined flow field equilateral triangular cylinder has been attained with respect to the
and Reynolds numbers. The Reynolds stress correlations downstream of experimental and numerical results. The occurrence of the von Karman
the mean wake recirculation region occur mainly because of the time- Vortex Street can be observed clearly the from instantaneous vorticity
dependent nature of the vortex shedding [3]. Therefore, the Reynolds contours in Fig. 4. The dominant vortex shedding frequency f with the
stress correlations are zero along the symmetry axis because of the help of PSD is calculated from the time history of the streamwise
velocity of the recorded PIV and LES data in the wake region of the

31
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

(K)0 Re=2900 (L)0 (M)0 (N)0 (O)0

y/L y/L y/L y/L y/L
1.25 1.25 1.25 1.25 1.25

<u*> <u*> <u*> <u*> <u*>

-0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5

-1.25 -1.25 -1.25 -1.25 -1.25

y/L Re=5800 y/L y/L y/L y/L

1.25 1.25 1.25 1.25 1.25

<u*> <u*> <u*> <u*> <u*>

-0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5

-1.25 -1.25 -1.25 -1.25 -1.25

y/L Re=11600 y/L y/L y/L y/L

1.25 1.25 1.25 1.25 1.25

<u*> <u*> <u*> <u*> <u*>

-0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5

-1.25 -1.25 -1.25 -1.25 -1.25

Fig. 13. Variations of the time-averaged streamwise velocity component (< u* >) described along the vertical lines as ‘‘K, L, M, N and O’’, which are 0.25L, 0.5L, 1.0L, 1.5L and 2.0L,
respectively, away from the rear surface of the equilateral triangular cylinder along the +x axis.

Fig. 14. Variations of Reynolds stress correlations (< u'v'/U∞2 >) described along the vertical lines as ‘‘K, L, M, N and O’’, which are 0.25L, 0.5L, 1.0L, 1.5L and 2.0L, respectively, away
from the rear surface of the equilateral triangular cylinder along the +x axis.

32
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 15. Time history of instantaneous streamwise velocity component u (m/s) against t (s) and Power Spectral Density (PSD) (m2/s) analysis at the denoted point with black on the small
image for all studied Reynolds numbers.

equilateral triangular cylinder, near the shear layer region denoted with i.e., Fast Fourier Transformation (FFT). For FFT analyses, the input data
a small black dot in Fig. 4. number should be a power of 2. Instantaneous cross-stream velocities
In Fig. 15, the 1024 instantaneous streamwise velocity components were also used to perform the FFT analyses, but the result was nearly
were taken to obtain the dominant frequency by using spectral analysis, the same as in the streamwise velocities. The streamwise velocity

33
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

Fig. 16. Variation of Strouhal numbers with changing Reynolds numbers.

results are the only results given in the present study. FFT analyses Table 2
reveal the vortex shedding frequency in the wake of the equilateral Comparison of the drag coefficients and Strouhal numbers with the literature.
triangular cylinder, as shown in Fig. 15. The dominant vortex shedding
Authors Re CD St
frequency can be defined as the peak value of PSD, determined to be
0.245/0.259 Hz, 0.511/0.524 Hz and 1.016/1.044 Hz for PIV/LES at Rasool et al. [39] 100 and 150 1.7 and 2.0 0.204 and 0.225
Re=2.9×103, Re=5.8×103 and at Re=1.16×104, respectively. This Srikanth et al. [40] 150 1.93 0.221
vortex shedding frequency, f, is a function of the Strouhal number, St= De and Dalal [41] 150 1.87 0.225
Chatterjee and Mondal 150 1.904 0.203
(f L)/U∞, and it is calculated as 0.211/0.214, 0.22/0.225 and 0.225/
[38]
0.219 for PIV/LES results at Re=2.9×103, Re=5.8×103 and Dhiman and Shyam [1] 150 1.894 0.204
Re=1.16×104, respectively. Because the results obtained from PIV Ganga Prasath et al. [42] 200 1.43 0.201
and LES were in agreement in terms of flow patterns and Strouhal Bovand et al. [43] 200 1.91 0.193
Agrawal et al. [44] 520 and 0.65 and 2.47 0.20 and 0.197
numbers, the numerical analyses were also completed at the Reynolds
1.04×103
numbers in the range of 2.9×103≤Re≤1.16×105 to investigate the Twigge-Molecey and 2.6×103 and – 0.21
effect on the Strouhal number. The magnitudes of the spectral peak Baines [45] 3.6×104
variation with the Reynolds number are illustrated in Fig. 15. With an Present paper 2.9×103 to – 0.211–0.225
increase in the Reynolds number, the PSD and vortex shedding (experimental) 1.16×104
Present paper (numerical) 2.9×103 to 1.43–1.33 0.214–0.23
frequency were also increased. However, the Strouhal number was
1.16×105
not affected significantly by the change in the Reynolds number in the Okamoto et al. [46] 9×103 1.4 0.22
range examined. Comparison of the present study with the literature in Zheng et al. [47] 14440 – 0.21
terms of the Strouhal numbers against Reynolds number is given in Protos et al. [48] 3.9×104 – 0.22
Obara and Matsudaira 7.5×104 – 0.26
Fig. 16. The Strouhal numbers in the related studies change approxi-
[49]
mately between 0.2≤St≤0.23 in the wide range of the Reynolds Nakagawa [50] 105 – 0.22
numbers. When the Reynolds number increases, there is also an Venugopal et al. [19] 1.3×105 and – 0.20
increment in the vortex shedding frequency to remain an almost 2.2×105
constant value, as aforementioned. In addition, it is possible to associate
the Strouhal number values with the Reynolds numbers investigated.
Correlatively, Strouhal numbers obtained from both the experimental
and numerical results remain almost constant due to the simultaneous
rise in the velocity and the vortex shedding frequency. With this
perspective, constancy in the Strouhal numbers presents an advantage
as the vortex flowmeter and thus is utilized as a measurement device. In
addition, the values of the Strouhal numbers and drag coefficients
coincide with the values from literature summarized in Table 2.

Fig. 17. The time history of drag coefficient at Re=5.8×103.

3.7. Drag history of the equilateral triangular cylinder

due to the vortex shedding in the shear layer. The time-averaged drag
One of the most important issues in studying flow over bluff bodies
coefficient value was determined as CD=1.47 at Re=5.8×103. Altera-
is related to the loading effect and the drag coefficient (CD). In the
tion of the mean drag coefficient with the Reynolds number is shown in
present study, the flow characteristics and spectral analyses of flow
Fig. 18. The drag coefficient tends to increase from Re=2.9×103 to
around the equilateral triangular cylinder are given, in detail, by the
Re=5.8×103 as determined by the study. Accordingly, the same trend
PIV and LES results. The present PIV results are similar to the results of
in increasing drag coefficients has also been observed in the study
the LES analyses in terms of time-averaged flow field images and
prepared by Chatterjee and Mondal [38] at lower Reynolds numbers.
spectral analyses. Therefore, the drag coefficient is numerically calcu-
They have also studied the equilateral triangular cylinder in cross flow
lated, due to the matching of flow characteristics. After forming the
at low Reynolds numbers and obtained the following results:
regular vortex shedding, the time-dependent drag coefficient at
CD=1.5334, 1.7546 and 1.9037 at Re=50, 100 and 150, respectively.
Re=5.8×103 was derived in Fig. 17. The drag coefficient fluctuates

34
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

1.5 equilateral triangular cylinder between the PIV and the LES yields a
relative error less than ± 5%, which is quite acceptable. In the design
1.45 and the manufacturing processes of different vortex flowmeters for
industrial applications, using the PIV technique or the LES method
saves time and cost.
CD 1.4
Acknowledgement
1.35
This work is partially supported by Selcuk University's Scientific
Research Project Office Contract no's. 11401058 and 11401059 and
1.3 Turkish Republic Government Planning Organization (DPT) via the
0 2 4 6 8 10 12 2009K12180 coded project of Advanced Technology Research and
Re x 104 Application Centre.
Fig. 18. Alteration of drag coefficient obtained by CFD for the different Reynolds
numbers. References

However, with the increase in the Reynolds number, the mean drag [1] A. Dhiman, R. Shyam, Unsteady heat transfer from an equilateral triangular
cylinder in the unconfined flow regime, Int. Sch. Res. Netw. Mech. Eng. 2011
coefficient drops after the critical Reynolds number of Re=5.8×103.
(2011) 776–789, [Link]
The drag coefficient as a term consists of the pressure (form) drag [2] X. Zhang, P. Blair, Turbulent vortex shedding from triangular cylinder using the
coefficient and the friction (viscous) drag coefficient. The friction drag turbulent body force potential model, in: Proceedings of the ASME 2000 Fluids
coefficient is under the influence of the shear caused by the fluid, and Engineering Division Summer Meeting, Boston, MA, USA FEDSM2000-11172,
2000.
the friction drag coefficient is more important for laminar flow [3] M. Ozgoren, Flow structure in the downstream of square and circular cylinders,
conditions while the pressure drag coefficient is related to the pressure Flow Meas. Instrum. 17 (4) (2006) 225–235, [Link]
difference between the upstream and downstream area of the body, and flowmeasinst.2005.11.005.
[4] A. Okajima, Strouhal numbers of rectangular cylinders, J. Fluid Mech. 123 (1982)
it is more significant in turbulent flow conditions. For this reason, when 379–398, [Link]
the Reynolds number rises, the fluid flow approaches the turbulent flow [5] D.T. Knauss, J.E.A. John, C.H. Marks, The vortex frequencies of bluff cylinders at
or exceeds the critical value and becomes completely turbulent. In this low Reynolds numbers, J. Hydronautics 10 (4) (1976) 121–126, [Link]
10.2514/3.48149.
way, the pressure drag coefficient becomes more important than the [6] S.C. Yen, C.W. Yang, Flow patterns and vortex shedding behavior behind a square
friction drag coefficient in terms of flow characteristics. With the cylinder, J. Wind Eng. Ind. Aerodyn. 99 (2011) 868–878, [Link]
increasing turbulence level of the flow, the wake region of the body 1016/[Link].2011.06.006.
[7] M. Ozgoren, E. Pinar, B. Sahin, H. Akilli, Comparison of flow structures in the
shrinks and then becomes closer to the body. The smaller size of the
downstream region of a cylinder and sphere, Int. J. Heat Fluid Flow 32 (6) (2011)
wake region for the body means that there is less pressure difference 1138–1146, [Link]
between the upstream and downstream areas of the body. Therefore, [8] R.F. Huang, B.H. Lin, S.C. Yen, Time-averaged topological flow patterns and their
influence on vortex shedding of a square cylinder in cross-flow at incidence, J.
less pressure difference constitutes less drag for the body since the
Fluids Struct. 26 (2010) 406–429, [Link]
pressure drag coefficient is more dominant under these conditions. The 01.003.
drag coefficient seems to be less affected at high Reynolds numbers and [9] F. Gu, J.S. Wang, X.Q. Qiao, Z. Huang, Pressure distribution, fluctuating forces and
is going to be an asymptotic trend. vortex shedding behaviour of circular cylinder with rotatable splitter plates, J.
Fluids Struct. 28 (2012) 263–278, [Link]
11.005.
[10] S.J. Lee, J.Y. Lee, PIV measurements of the wake behind a rotationally oscillating
4. Conclusions circular cylinder, J. Fluids Struct. 24 (2008) 2–17, [Link]
jfluidstructs.2007.06.001.
[11] A. Kalmbach, M. Breuer, Experimental PIV/V3V measurements of vortex-induced
Numerical and experimental flow characteristics around the equi- fluid structure interaction in turbulent flow—A new benchmark FSI-PfS-2a, J.
lateral triangular cylinder have been obtained and compared by using Fluids Struct. 42 (2013) 369–387, [Link]
07.004.
the PIV and the LES methods. The present paper has supported
[12] S. Gunes, V. Ozceyhan, O. Buyukalaca, Heat transfer enhancement in a tube with
quantitative experimental and numerical information in the near wake equilateral triangle cross sectioned coiled wire inserts, Exp. Therm. Fluid Sci. 34 (6)
region of the equilateral triangular cylinder. In general, the compara- (2010) 684–691, [Link]
tive results of PIV and LES are quite similar in terms of flow [13] N. Agrawal, S. Dutta, B.K. Gandhi, Experimental investigation of flow field behind
triangular prisms at intermediate Reynolds number with different apex angles, Exp.
characteristics and spectral analyses. Therm. Fluid Sci. 72 (2016) 97–111, [Link]
The time-averaged flow fields show symmetrical flow patterns 2015.10.032.
occurring along the axis line of the equilateral triangular cylinder. [14] S. El-Sherbiny, Flow separation and reattachment over the sides of a 90° triangular
prism, J. Wind Eng. Ind. Aerodyn. 11 (1–3) (1983) 393–403, [Link]
With the increase in the Reynolds number, the wake of the equilateral 1016/0167-6105(83)90116-2.
triangular cylinder shrinks, and the stagnation points approach the back [15] G.V. Iungo, G. Buresti, Experimental investigation on the aerodynamic loads and
side of the triangular cylinder for both methods. The Strouhal numbers wake flow features of low aspect-ratio triangular prisms at different wind direc-
tions, J. Fluids Struct. 25 (7) (2009) 1119–1135, [Link]
are nearly St=0.22 for PIV and LES at all Reynolds numbers, and it can jfluidstructs.2009.06.004.
be inferred that these values are independent of the Reynolds number. [16] J. Peng, X. Fu, Y. Chen, Experimental investigations of Strouhal number for flows
However, the PSD and the vortex shedding are functions of the past dual triangulate bluff bodies, Flow Meas. Instrum. 19 (6) (2008) 350–357,
[Link]
Reynolds number, and they increase with increasing Reynolds number.
[17] H. Zhang, Y. Huang, Z. Sun, A study of mass flow rate measurement based on the
The drag coefficient is also obtained numerically and after vortex shedding principle, Flow Meas. Instrum. 17 (1) (2006) 29–38, [Link]
Re=5.8×103, the drag coefficient displays a decreasing trend. In org/10.1016/j.flowmeasinst.2005.08.002.
[18] M. Cheng, G.R. Liu, Effects of after body shape on flow around prismatic cylinders,
addition, it should be emphasized that the cell volume size is an
J. Wind Eng. Ind. Aerodyn. 84 (2) (2000) 181–196, [Link]
important function to decide filtering operations for small scale eddies S0167-6105(99)00050-1.
in the LES method. Therefore, to obtain accurate results using LES, the [19] A. Venugopal, A. Agrawal, S.V. Prabhu, Influence of blockage and shape of a bluff
fine grid system must be created with a small grid that can result in body on the performance of vortex flowmeter with wall pressure measurement,
Measurement 44 (5) (2011) 954–964, [Link]
increasing the solution time and thus requiring a powerful computing 2011.02.015.
capacity. Comparison of the streamwise velocity and the Reynolds [20] H.A. Khaledi, H.I. Andersson, On vortex shedding from a hexagonal cylinder, Phys.
Stress correlation along the vertical lines in the wake region of the Lett. A 375 (2011) 4007–4021, [Link]

35
S. Yagmur et al. Flow Measurement and Instrumentation 55 (2017) 23–36

[21] J. Wang, J. Piechna, N. Müller, A novel design of composite water turbine using [36] R.M. Stringer, J. Zang, A.J. Hillis, Unsteady RANS computations of flow around a
CFD, J. Hydrodyn. Ser. B 24 (1) (2012) 11–16, [Link] circular cylinder for a wide range of Reynolds numbers, Ocean Eng. 87 (1) (2014)
6058(11)60213-8. 1–9, [Link]
[22] B. Saavedra-Moreno, S. Salcedo-Sanz, A. Paniagua-Tineo, L. Prieto, A. Portilla [37] P. Jagadeesh, K. Murali, Application of low-Re turbulence models for flow
Figueras, Seeding evolutionary algorithms with heuristics for optimal wind turbines simulations past underwater vehicle hull forms, J. Nav. Archit. Mar. Eng. 2 (1)
positioning in wind farms, Renew. Energy 36 (11) (2011) 2838–2844, [Link] (2005) 41–54, [Link]
[Link]/10.1016/[Link].2011.04.018. [38] D. Chatterjee, B. Mondal, Mixed convection heat transfer from an equilateral
[23] R. Hassanzadeh, B. Sahin, M. Ozgoren, Large eddy simulation of free-surface effects triangular cylinder in cross flow at low Reynolds numbers, Heat Transf. Eng. 36 (1)
on the wake structures downstream of a spherical body, Ocean Eng. 54 (1) (2012) (2015) 123–133, [Link]
213–222, [Link] [39] T. Rasool, A. Dhiman, M. Parveez, Cross-buoyancy mixed convection around a
[24] G. Hu, K.T. Tse, K.C.S. Kwok, Y. Zhang, Large eddy simulation of flow around an confined triangular bluff body, Numer. Heat Transf. Part A: Appl. 67 (4) (2015)
inclined finite square cylinder, J. Wind Eng. Ind. Aerodyn. 146 (2015) 172–184, 454–475, [Link]
[Link] [40] S. Srikanth, A.K. Dhiman, S. Bijjam, Confined flow and heat transfer across a
[25] K. Luo, S. Zhang, Z. Gao, J. Wang, L. Zhang, R. Yuan, J. Fan, K. Cen, Large-eddy triangular cylinder in a channel, Int. J. Therm. Sci. 49 (2010) 2191–2200, http://
simulation and wind-tunnel measurement of aerodynamics and aeroacoustics of a [Link]/10.1016/[Link].2010.06.010.
horizontal-axis wind turbine, Renew. Energy 77 (2015) 351–362, [Link] [41] A.K. De, A. Dalal, Numerical study of laminar forced convection fluid flow and heat
org/10.1016/[Link].2014.12.024. transfer from a triangular cylinder placed in a channel, J. Heat Transf. 129 (5)
[26] H. Zhang, J. Yang, L. Xiao, H. Lu, Large-eddy simulation of the flow past both finite (2006) 646–656, [Link]
and infinite circular cylinders at Re=3900, J. Hydrodyn. Ser. B 27 (2) (2015) [42] S. Ganga Prasath, M. Sudharsan, V. Vinodh Kumar, S.V. Diwakar, T. Sundararajan,
195–203, [Link] Shaligram Tiwari, Effects of aspect ratio and orientation on the wake characteristics
[27] R.J. Adrian, Particle-imaging techniques for experimental fluid mechanics, Annu. of low Reynolds number flow over a triangular prism, J. Fluids Struct. 46 (2014)
Rev. Fluid Mech. 23 (1991) 261–304, [Link] 59–76, [Link]
010191.001401. [43] M. Bovand, S. Rashidi, J.A. Esfahani, Enhancement of heat transfer by nano fluids
[28] A. Fouras, J. Soria, Accuracy of out-of-plane vorticity measurements derived from and orientations of the equilateral triangular obstacle, Energy Convers. Manag. 97
in-plane velocity field data, Exp. Fluids 25 (5) (1998) 409–430, [Link] (2015) 212–223, [Link]
10.1007/s003480050248. [44] N. Agrawal, S. Dutta, B.K. Gandhi, An experimental study on flow past an
[29] D.P. Hart, PIV error correction, Exp. Fluids 29 (1) (2000) 13–22, [Link] equilateral prism at intermediate Reynolds number and the effect of its orientation,
10.1007/s003480050421. in: Proceedings of the ASME Fluids Engineering Division Summer Meeting, Incline
[30] Anonymous, Fluent 14.0 User Guide, Fluent Inc., 2013. Village, NV, USA (2013) FDSM2013-16474, 2013.
[31] J. Taghinia, M.M. Rahman, T. Siikonen, Large eddy simulation of flow past a [45] C.F.M. Twigge-Molecey, W.D. Baines, Aerodynamic forces on a triangular cylinder,
circular cylinder with a novel sub-grid scale model, Eur. J. Mech. – B/Fluids 52 J. Eng. Mech. Div. 99 (4) (1973) 803–818.
(2015) 11–18, [Link] [46] T. Okamoto, M. Yagita, K. Ohtsuka, Experimental investigation of the wake of a
[32] C. Liang, G. Papadakis, Large eddy simulation of pulsating flow over a circular wedge, Bull. JSME 20 (141) (1977) 323–328, [Link]
cylinder at subcritical Reynolds number, Comput. Fluids 36 (2) (2007) 299–312, jsme1958.20.323.
[Link] [47] Y. Zheng, A. Rinoshika, S. Fujimoto, Triangle cylinder wake analysis based on
[33] A.R. Vasel-Be-Hagh, R. Carriveau, D.S.K. Ting, Numerical simulation of flow past an wavelet and POD techniques, Procedia Eng. 126 (2015) 108–112, [Link]
underwater energy storage balloon, Comput. Fluids 88 (15) (2013) 272–286, org/10.1016/[Link].2015.11.188.
[Link] [48] A. Protos, V.W. Goldschmidt, G.H. Toebes, Hydroelastic forces on bluff cylinders, J.
[34] J. Tu, D. Zhou, Y. Bao, Z. Han, R. Li, Flow characteristics and flow-induced forces of Basic Eng. 90 (3) (1968) 378–386, [Link]
a stationary and rotating triangular cylinder with different incidence angles at low [49] H. Obara, Y. Matsudaira, Large vortex formation-mechanism behind wedge under
Reynolds numbers, J. Fluids Struct. 45 (2014) 107–123, [Link] several separation conditions, JSME Int. J. Ser. B Fluids Therm. Eng. 41 (4) (1998)
1016/j.jfluidstructs.2013.11.004. 788–795, [Link]
[35] S. Yagmur, S. Dogan, M.H. Aksoy, E. Canli, M. Ozgoren, Experimental and [50] T. Nakagawa, Vortex shedding mechanism from a triangular prism in a subsonic
numerical investigation of flow structures around cylindrical bluff bodies, Eur. flow, Fluid Dyn. Res. 5 (2) (1989) 69–81, [Link]
Phys. J. Conf. 92 (2015) 02113, [Link] 5983(89)90012-9.

36