Sie sind auf Seite 1von 17

Flow around a diamond-section cylinder at

low Reynolds numbers


Cite as: Phys. Fluids 33, 053611 (2021); https://doi.org/10.1063/5.0049811
Submitted: 09 March 2021 • Accepted: 03 May 2021 • Published Online: 24 May 2021

Pavan Kumar Yadav, Kumar Sourav, Deepak Kumar, et al.

ARTICLES YOU MAY BE INTERESTED IN

Flow-induced vibrations of circular cylinder in tandem arrangement with D-section cylinder at


low Reynolds number
Physics of Fluids 33, 053606 (2021); https://doi.org/10.1063/5.0048580

Large-eddy simulation of flow past a circular cylinder for Reynolds numbers 400 to 3900
Physics of Fluids 33, 034119 (2021); https://doi.org/10.1063/5.0041168

Flow separation around a square cylinder at low to moderate Reynolds numbers


Physics of Fluids 32, 044103 (2020); https://doi.org/10.1063/5.0005757

Phys. Fluids 33, 053611 (2021); https://doi.org/10.1063/5.0049811 33, 053611

© 2021 Author(s).
Physics of Fluids ARTICLE scitation.org/journal/phf

Flow around a diamond-section cylinder at low


Reynolds numbers
Cite as: Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811
Submitted: 9 March 2021 . Accepted: 3 May 2021 .
Published Online: 24 May 2021

Pavan Kumar Yadav, Kumar Sourav, Deepak Kumar, and Subhankar Sena)

AFFILIATIONS
Department of Mechanical Engineering, Indian Institute of Technology (Indian School of Mines) Dhanbad, Dhanbad 826 004, India

a)
Author to whom correspondence should be addressed: ssen@iitism.ac.in

ABSTRACT
The steady separated flow past a diamond cylinder at low Reynolds numbers, Re, is associated with diverse separation topologies not
resolved for a circular or square cylinder. The present study, conducted for Re  150, also uncovers three unique separation topologies for
the time-averaged flow. In this regard, the most striking observation is the formation of a small sub-wake around the base of the cylinder at
certain Re between 80 and 90. While two of these structures were previously captured by some recent studies, these studies did not investigate
their origin or kinematics. In the present study, conducted via stabilized finite-element computations in two-dimensions, these wake topolo-
gies are analyzed in detail. For secondary separation, the pressure at the reattachment point on the cylinder rear surface exceeds the pressure
at the corresponding separation point located upstream. In a similar manner, for primary separation, the pressure at the wake stagnation
point(s) surpasses the one at the separation points. Via direct steady and unsteady computations, the value of the critical Re indicating the
onset of vortex-shedding is found to be 41, approximately. At this Re, the flow quantities, such as wake length, drag, and pressure coefficients,
obtained from the steady and unsteady computations are found to diverge. The presentation of wake length and drag as a function of base
suction establishes that the drag and wake length share an inverse relationship.
Published under an exclusive license by AIP Publishing. https://doi.org/10.1063/5.0049811

I. INTRODUCTION unsteady flow is marked by the onset of vortex-shedding at the critical


According to Roshko (1993), a bluff body is the one that results Reynolds number, Rec. The values of Rec for circular and square cylin-
in a wide extent of separated flow and is associated with significant ders are of the order of 47 and 50, respectively. Even for a cylinder
drag force as well as vortex-shedding. The fluid dynamics concerning with its contour being composed of sharp edges alone, its varied orien-
a bluff obstacle albeit governed primarily by the Reynolds number, Re, tations alter the location of the sharp corners on it relative to the
the shape of the body is found to influence the separation of boundary incoming flow and change the flow (Jiang and Cheng, 2020). A square
layer, interactions between the separated shear layers, and conse- cylinder at 45 incidence is known as the “diamond” configuration of
quently, the fluid loading. Recent studies on flow around bluff bodies the square cylinder. A diamond cylinder permits secondary separation:
of different shapes include those due to Zhu et al. (2020), Sharma and a mode of boundary layer separation in which separation bubbles
Barman (2020), and Kim (2021). The Reynolds number is defined as instead of the classical closed wake form on the cylinder rear surface
Re ¼ UD (discussed in Kumar et al., 2018b, 2019 for Re ¼ 6  8:2).
 , where U, D, and  stand for the free-stream speed, character-
istic dimension of the body, and kinematic viscosity of the fluid, For low Reynolds numbers, the flow around a square cylinder at
respectively. Accordingly, the advent of essential fluid mechanical phe- zero incidence is well understood, thanks to the abundant wealth of
nomena associated with bluff bodies having smooth contours and available literature. Several earlier studies, such as Franke et al. (1990),
straight edges occurs at different (sometimes appreciably) Reynolds Mukhopadhyay et al. (1992), Sohankar et al. (1998), and Robichaux
numbers. For instance, for unbounded flow, the initial separation of et al. (1999), reported the features of flow past a fixed square cylinder.
laminar boundary layer from isolated circular and square-section cyl- For this configuration, Sen et al. (2011) extracted the details of the ini-
inders commences at Res ¼ 6:19 (Brons et al., 2007; Sen et al., 2009) tial separation of laminar boundary layer while Jiang and Cheng
and 1.16 (Sen et al., 2011), respectively. Here, Res signifies the separa- (2020) reported the evolution of mean streamlines with Re. They
tion Reynolds number at which steady separation of laminar boundary resolved a total of five different separation topologies; four of them
layer initiates. The transition of steady flow to two-dimensional were identified for Re ¼ 10  158 and the rest for Re ¼ 159  400.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-1


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

This can be considered as a further refinement of the three possible forces, and vortex-shedding over Re ¼ 100  200. By performing
separation topologies proposed by Robichaux et al. (1999). Prior to Lattice Boltzmann simulations over Re ¼ 50  150, Vijaybabu et al.
summarizing the overall objectives of this work, the previous studies (2018) documented results concerning flow and heat transfer around a
relating to the flow past square and diamond section cylinders are dis- permeable diamond cylinder in an unbounded fluid medium. They
cussed below. reported the formation of the twin recirculation regimes aft the cylin-
For Re  250, Zaki et al. (1994) presented numerical results for der. Alam et al. (2020) numerically explored the flow and heat transfer
flow past a fixed square cylinder at incidence. For a fixed as well as a characteristics of a square cylinder at Re ¼ 150 by altering the corner
freely rotating square cylinder, they presented experimental results over radius, r and a. As functions of the normalized corner radius (varied
Re ¼ 1000  10000. Sohankar et al. (1998) performed two- from 0 to 1) and angle of incidence (varied from 0 to 45 ), they iden-
dimensional finite-volume computations to analyze the flow around a tified four major flow patterns. For the mean flow around the diamond
fixed and an inclined square cylinder (angle of incidence, configuration with sharp corners, they resolved a small secondary
a ¼ 0  45 ) in the laminar vortex-shedding regime (45  Re  200). wake bubble in the base region.
This study explored the effects of boundary conditions, streamwise In a couple of earlier studies (Kumar et al., 2018b, 2019), the
extents of the domain, angle of incidence, and blockage, b on the flow. authors investigated the phenomenon of secondary separation from a
The quantity blockage signifies the ratio of cross-stream projection of an diamond cylinder at Reynolds numbers around 8. These studies, con-
object and the corresponding width of the flow domain considered. ducted over the regime of steady flow, unveiled a number of novel sep-
Sohankar et al. (1998) indicated that Rec attains its least value of 42 aration topologies. As apparent from the above paragraph, Vijaybabu
when the value of a is increased gradually from 0 to 45 . Via experi- et al. (2018) and Alam et al. (2020) reported the presence of novel
ments at a low Re of 410 and b < 0:07, Dutta et al. (2008) explored the wake patterns in relation to unsteady flow around a diamond cylinder.
influence of a and cylinder aspect ratio (the ratio of cylinder span/length These studies, however, did not explore the wake patterns. A prime
and characteristic dimension) on the flow properties. For Re  300 and objective of the current numerical work is, therefore, to resolve the
a high blockage of 0.25, Moussaoui et al. (2010) numerically investigated separation topologies of a diamond cylinder over the stretch of
the flow and forced convection heat transfer phenomena concerning a Re ¼ Res  150. To this end, we identify the formation of a sub-wake
diamond cylinder. They found that the value of Rec is 82. At Rec, the or a small wake within the main wake at Re  90 and a further variant
drag coefficient, Cd [defined in Eq. (4)], was found to attain its mini- of the primary/main wake at Re ¼ 150. The upper limit of Re is chosen
mum. Yoon et al. (2010) employed immersed boundary method to so as to ensure that the flow is essentially two-dimensional and lami-
extract the details of steady as well as unsteady flows past a square cylin- nar. The current investigation is driven by the following queries: Does
der at incidence. For various a, they determined the onset of primary any correspondence exist between the flow past the symmetric corner
instability/onset of unsteadiness as well as three-dimensional wake tran- of a wedge [see Fig. 4(f)] and flow around the base point of a diamond
sition. A much refined increment in a (compared to the ones used by cylinder? Can one predict the presence of a sub-wake from the distri-
Sohankar et al., 1998) resulted in a non-monotonic Rec  a variation bution of mean surface pressure? How do the magnitudes of the pres-
with Rec attaining its minimum value of 39 for a ¼ 45 . Irrespective of sure coefficient, Cp compare at the separation and attachment points?
the value of a, they noted a linear rise and non-linear fall of the wake In terms of separation topology, how many regimes of flow do exist
length, LD1 with Re. The wake length represents the shortest distance for Re  150? Does there exist a relationship between the wake length
between the base point/rear stagnation point and wake stagnation point and drag coefficient of the cylinder? The pressure coefficient is defined
of the classical closed wake [see Fig. 3(e) for example]. Feng et al. (2010) as Cp ¼ pP
1
1
qU 2
, where p is the pressure, P1 the value of p at the free-
2
computed the fluid loading and vortex-shedding frequency of diamond stream condition, and q the density of the fluid. The existence of a sec-
cylinders of aspect ratios 13, 1, and 3 at Re ¼ 100. For a diamond-section ond minimum in surface pressure near the base indicates the forma-
cylinder, aspect ratio in two-dimensions represents the axis ratio or the tion of the sub-wake. It is found that the value of surface pressure at a
ratio of the lengths of the streamwise and cross-stream diagonals of the reattachment point always surpasses the pressure at a separation point.
cylinder. The Strouhal number, St ¼ fD U , signifies the dimensionless For Re  150, the current study identifies six regimes of flow associ-
vortex-shedding frequency, where f is the dimensional vortex-shedding ated with different streamline topologies. The fifth and sixth regimes
frequency. both contain a sub-wake. The splitting of the twin eddies of the main
For b ¼ 0:25, Jahromi et al. (2011) numerically studied the effect wake into a pair of like sign vortices in regime six is a novel observa-
of incidence (for a ¼ 0  45 ) on steady flow (Re ¼ 1  40) of a tion. We establish the kinematic stability of the separation topologies
power-law fluid and associated forced convection heat transfer around corresponding to the fifth and sixth regimes of flow. To understand
a heated square cylinder. Sasmal and Chhabra (2012) investigated the the dependence between wake length and drag, these quantities are
unconfined flow of a power-law fluid and forced convection heat plotted against base suction, Cpb by treating base suction as a key ref-
transfer from a diamond cylinder. They performed two-dimensional erence parameter (instead of Re). An inverse relationship between
computations for Re ¼ 45  120, Prandtl number, Pr ¼ 0:7  80, wake length and drag is observed. We have also performed direct com-
and power-law index ¼ 0:3  1:8. For Newtonian fluids, they found putations to determine the critical Reynolds number. Data obtained
that the location of separation alters as a function of Re. For from steady as well as unsteady computations for certain characteristic
b ¼ 0:125  0:5; Re ¼ 1  40, and Pr ¼ 50, Kumar et al. (2014) pre- flow quantities, such as wake length, pressure coefficients, drag, and its
sented extensive numerical results for flow of a power-law fluid and components, have been considered toward the determination of Rec.
forced convection heat transfer relating to a diamond cylinder. For The remaining of this article in structured in the following man-
perforated square cylinders with a ¼ 0  45 , Sohankar and Najafi ner. The governing equations of incompressible fluid flow and the
(2018) numerically experimented on the control of heat transfer, fluid associated finite-element formulation are presented in Secs. II and III,

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-2


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

respectively. The problem statement and computational mesh are element level integrals renders the Galerkin formulation to stabilized
depicted in Sec. IV. Section V addresses the validation as well as con- Petrov–Galerkin formulation. The Streamline-Upwind/Petrov–
vergence of the predicted results. Section VI presents and analyzes the Galerkin (SUPG) stabilization term suppresses the spurious modes of
results concerning laminar flow past a stationary diamond cylinder. velocity while the Pressure-Stabilizing/Petrov–Galerkin (PSPG) stabili-
Finally, the conclusions are drawn in Sec. VII. zation term suppresses the checkerboard modes of pressure. The sec-
ond series of element level integrals involving the stabilization
II. THE GOVERNING EQUATIONS FOR
INCOMPRESSIBLE FLOW parameter, d is added to enforce numerical stability at high Re.
Further information on the finite-element formulation can be found
The strong form of the conservation equations governing the in Tezduyar et al. (1992).
unsteady flow of an incompressible fluid is expressed using vector
notations as IV. PROBLEM DESCRIPTION AND FINITE-ELEMENT
  MESH
@u
q þ u  $u  f  $  r ¼ 0 on X  ð0; TÞ; (1) A stationary diamond cylinder with length of each diagonal D
@t resides in a rectangular computational domain EFGH of length 75D
$  u ¼ 0 on X  ð0; TÞ: (2) and width 20D [Fig. 1(a)]. The center of the cylinder is coincident
with the origin of the Cartesian coordinate system. The setup used cor-
Here, X is the spatial domain with domain boundary C and ð0; TÞ
responds to a blockage of 0.05. The upstream and downstream bound-
represents the temporal domain. The dimensional spatial and tempo-
ral coordinates are denoted by x ¼ ðx; yÞ and t, respectively. For com- aries are located at streamwise distances of 30D and 45D, respectively,
putations of steady flow, the time derivative is dropped from Eq. (1). from the origin. The cylinder wall represents a no-slip surface. Free-
In these equations, u ¼ ðu; vÞ; f, and r stand for the fluid velocity, stream condition on velocity is prescribed at the inlet while the exit
body force per unit volume, and the Cauchy stress tensor at a point, boundary is traction-free or stress-free. Along the lateral boundaries of
respectively. In this work, f ¼ 0. The constitutive relation for stress is the domain, the cross-stream component of velocity and shear stress,
read as the sum of its isotropic and deviatoric components, both disappear. A normalized time step size, 䉭tU D , of 0.005 has been
h i used for all the unsteady computations.
1 For generation of finite-element mesh, the computational
r ¼ pI þ 2leðuÞ where eðuÞ ¼ ð$uÞ þ ð$uÞT : (3)
2 domain EFGH is partitioned in five contiguous mesh blocks
Here, I, l, and e are the identity tensor, dynamic viscosity of the fluid, [Fig. 1(a)], i.e., a central block and its four neighbor rectangular blocks.
and strain rate tensor, respectively. The essential and natural boundary The central block that accommodates the cylinder is a square block of
conditions are also considered. A divergence-free velocity field is used size 4D  4D. The mesh in each block is non-uniform, structured, and
as the initial condition. The drag and lift coefficients for the cylinder composed of bilinear quadrilateral elements. Overall, the mesh is com-
are calculated as posed of 90 418 nodes and 89 700 elements. A total of 520 surface
ð nodes, Nt are used to construct the contour of the cylinder. The radial
1
Cd ; Cl ¼ ðn  rÞx;y dC: (4) thickness of the finite-elements located on the cylinder surface is
1 2 Ccyl
qU D 0:0005D. Figure 1(b) shows the mesh in the central block along with a
2 close-up of some elements near the surface.
Here, Ccyl stands for the cylinder surface and n represents the outward
local normal vector. V. TESTS FOR MESH CONVERGENCE, TIME STEP SIZE,
AND VALIDATION OF THE FORMULATION
III. THE FINITE-ELEMENT FORMULATION
A. Mesh convergence
With wh and qh denoting the weight functions for velocity and
Two meshes, labeled M1 and M2, are considered to illustrate the
pressure, respectively, the weak or variational form of Eqs. (1) and (2)
mesh convergence of results computed at Re ¼ 100 using 䉭tU D ¼ 0:005.
is represented via the following compact expression:
  The number of nodes and elements of M2 is about twice of those of
ð
@uh M1 (see Table I). The values of characteristic flow quantities obtained
wh  q þ uh  $uh  f dX
X @t from M1 to M2 are compared in Table I. In this table, Cd rms is the root
ð ð mean square or rms drag, Cl rms is the rms lift, Cp0 is the mean forward
þ eðwh Þ : rðph ; uh ÞdX þ qh $  uh dX stagnation pressure coefficient, Cpb is the mean base pressure coeffi-
X X
nel ð
    cient, and hs is the time-averaged separation angle, respectively. For
X 1 @wh
þ sSUPG q h h
þ u  $w þ sPSPG $q h each quantity, the closeness of its values obtained from M1 to M2
e q @t
e¼1 X essentially establishes the adequacy of M1 to predict mesh indepen-
   
@uh dent data. Mesh M1 with 90 418 nodes and 89 700 elements is, there-
: q þ u  $u  f þ $p dXe
h h h
fore, used for all the computations.
@t
Xnel ð þ
þ d$  wh q$  uh dXe ¼ wh  hh dC: (5) B. Test for time step size
e
e¼1 X Ch
To test the sufficiency of the time step size, the flow at Re ¼ 100
In Eq. (5), the terms without series of element level integrals con- has been computed on mesh M1 using 䉭tU D ¼ 0:005 and 0.0025. The
tribute to the Galerkin statement of the problem. Addition of series of closeness of results (see Table II) using these two time step sizes

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-3


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 1. Two-dimensional flow past a stationary, isolated diamond cylinder: (a) the problem statement with illustration of five mesh blocks and (b) finite-element mesh for the
central block.

TABLE I. Comparison of characteristic flow quantities for Re ¼ 100 flow past a diamond cylinder on meshes M1 and M2.

Mesh Nodes Elements Nt Cd Cd rms Cl rms St Cp0 Cpb L1 =D hs

M1 90 418 89 700 520 1.7814 0.0296 0.4936 0.1811 1.3333 1.2502 0.9001 88.94
M2 181 264 180 000 1040 1.7889 0.0295 0.4967 0.1813 1.3981 1.2762 0.8844 89.33

TABLE II. Two-dimensional, unsteady laminar flow around a diamond cylinder at Re ¼ 100: test for the adequacy of 䉭tU
D . The computations are performed on mesh M1.

䉭tU
D Cd Cdrms Clrms St Cp0 Cpb L1 =D hs

0.005 1.7814 0.0296 0.4936 0.1811 1.3333 1.2502 0.9001 88.9386


0.0025 1.7885 0.0296 0.4902 0.1829 1.3319 1.2491 0.9062 88.9432

ascertains that 䉭tU


D ¼ 0:005 is adequate to provide solutions indepen- each study reported in Table III, it may be noted that a scatter exists in
dent of the size of time step. the value of blockage. In addition, the value of b is not mentioned in
Sohankar et al. (1998) (for diamond configuration) and Roy and
Bandyopadhyay (2006). While Moussaoui et al. (2010) considered a
C. Validation of formulation relatively high blockage of 0.25, the blockage for the remaining studies
For unsteady flow past a diamond cylinder at Re ¼ 100, Table III is restricted roughly within 0.033. In context of unsteady flow past a
provides a detailed comparison of the predicted Cd, Cl rms , St, and stationary circular cylinder at Re ¼ 100, Behr et al. (1995) numerically
L1 =D for b ¼ 0:05 with several numerical results reported earlier. For demonstrated that the vortex-shedding is insensitive to blockage for

TABLE III. Two-dimensional, unsteady laminar flow past an isolated diamond cylinder at Re ¼ 100: comparison of the predicted Cd, Cl rms , St, and L1 =D with those obtained by
previous numerical investigations.

Study b Cd Clrms St L1 =D

Sohankar et al. (1998) (data extracted from graph)  1.7141 0.4359 0.1788
Roy and Bandyopadhyay (2006) (data extracted from graph)  1.8065 0.1865
Feng et al. (2010) 0.025 1.7800 0.1656
Yoon et al. (2010) (data extracted from graph) 0.01 1.6891 0.9114
Sasmal and Chhabra (2012) 0.033 1.6943 0.4356 0.1756
Vijaybabu et al. (2018) 0 1.7545
Present 0.05 1.7814 0.4936 0.1811 0.9001

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-4


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

b  0:0625. With this result in the background, we expect the results was computed using 70 000 time steps and the instantaneous field var-
for b  0:033 and smaller to display close agreement. As revealed by iables, i.e., velocity and pressure data were recorded at each 100th time
Table III, the predicted results, in general, display satisfactory match step. The mean flow field has been computed by averaging the last 200
with other results. Since high blockage increases the magnitudes of Cd records (out of 700 records) of instantaneous velocity and pressure.
and St [Posdziech and Grundmann (2007) for a circular cylinder; For a square cylinder, Robichaux et al. (1999) showed three different
Turki et al. (2003) and Sharify et al. (2013) for a square cylinder], the mean streamline patterns for Re < 120, Re  120, and Re  150,
discrepancy in the comparison noted relative to the results of respectively; each of these patterns is composed of closed streamlines
Moussaoui et al. (2010) for b ¼ 0:25 is quite predictable. and preserves symmetry about the x axis. The flow for Re < 120 sepa-
rates only from the trailing edge and results in a symmetric mean
VI. RESULTS
wake bubble comprising of twin vortical structures similar to that of a
A. The flow circular cylinder at low Re. Unlike the circular and square cylinders,
1. The instantaneous flow field the trailing edge of the diamond cylinder reduces to a sharp corner
and presence of this sharp corner renders its mean wake to be notice-
For unsteady flow around a diamond cylinder at Re ¼ 60 and
ably distinct from those of its circular and square counterparts. The
150, respectively, the first and second rows of Fig. 2 illustrate the flow
mean wake of a diamond cylinder is shown in Fig. 3 for Re ¼ 80, 90,
field in terms of instantaneous streamlines as well as vorticity con-
100, and 150. For Re ¼ 80 [Fig. 3(a)], the classical wake comprises a
tours. The undulations or waviness in the wake streamlines (first col-
pair of counterrotating eddies and is associated with a single wake
umn of Fig. 2) are characteristic to unsteady flow. The streamwise
stagnation point. At some Re between 80 and 90, a small wake within
intervortex spacing in the Karman vortex street at Re ¼ 60 [Fig. 2(b)]
the primary or main wake, forms at the base. The formation of a small
is larger than those at Re ¼ 150 [Fig. 2(d)]. This signifies that within a
wake within the main wake appears to be a less reported phenomenon
given downstream extent, more number of shed vortices are occupied
on unsteady flow around bluff cylinders at low Re. We designate the
at high Re. Thus, intuitively, the strength of vortex-shedding or alter-
small wake as “sub-wake.” Recent studies by Vijaybabu et al. (2018)
nately the vortex-shedding frequency at Re ¼ 150 must be greater than
and Alam et al. (2020) resolved such structures. While Vijaybabu et al.
those at Re ¼ 60. This is corroborated by Fig. 9(a) depicting the St–Re
(2018) did not explore the sub-wake, Alam et al. (2020) noted a small
relationship. The vortical structures corresponding to Re ¼ 60 and 150
belong to two different regimes, namely, coherent vortex regime and hump of positive skin friction (around the base) in the distribution of
fluctuating vortex regime, respectively (Sarkar et al., 2021). In the mean skin friction coefficient around the cylinder. This hump signifies
coherent vortex regime, alternate coherent vortices are shed down- the presence of “secondary wake bubble.”
stream of the cylinder. In the fluctuating vortex regime, in contrast, For separation with a sub-wake, a total of four eddies are accom-
the initially shed vortices undergo a narrowing process downstream of modated in the mean wake of a diamond cylinder. The sense of rota-
the cylinder. High strain rate prevails in the narrow region. tion of the eddies in the sub-wake is opposite to those of the main
wake. The appearance of the additional recirculation at base is well
recognized from Fig. 3(e) depicting the variation of mean streamwise
2. The mean or time-averaged flow field
velocity, u along the wake centerline at Re ¼ 80 and 90. For Re ¼ 80, a
The mean recirculation bubble or time-averaged wake of a circu- single change of sign of u  from negative to positive at point M1 signi-
lar cylinder obtained by time-averaging the flow field over time suffi- fies the presence of a single recirculation regime with flow directed
ciently larger than the period of vortex-shedding is essentially closed toward the base. In this case, the base point denoted by the symbol A2
and symmetric (Williamson, 1996). For instance, the flow at Re ¼ 120 acts as an attachment point. In contrast, for Re ¼ 90, a pair of changes

FIG. 2. Instantaneous streamlines and vorticity at Re ¼ 60 (first row, figures a and b) and Re ¼ 150 (second row, figures c and d) for unsteady flow past a stationary diamond
cylinder. In each figure, the origin of the coordinate system is marked with  symbol.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-5


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 3. Two-dimensional unsteady laminar flow past a diamond cylinder: mean streamlines for Re ¼ (a) 80, (b) 90, (c) 100, and (d) 150. For Re ¼ 80 and 90, (e) plots the
mean u velocity along the wake centerline while (f) plots the normalized vorticity along the lower surface of the cylinder.

of sign of u indicates the co-existence of two recirculation regimes. structure degenerates to a new arrangement of vortices with each pri-
Along the wake centerline, u is positive (flow away from base point S2) mary vortex splitting into a pair of like sign vortices. The vortex split-
for the first regime while for the second regime, it is negative. The ting results into two additional saddle points M3 and M4 within the
symbols A2 and S2 are discussed in Subsection VI A 3. The outward primary wake and not on the wake centerline [Fig. 3(d)]. The resultant
and inward moving flows result in an additional saddle point M2 on vortices are analogous to the “cat’s eye” vortices found in several ear-
the wake centerline. The mean wake therefore consists of the first and lier investigations (Kuhlmann et al., 1997; Luzzatto-Fegiz and
second wake stagnation points M1 and M2, respectively [Fig. 3(d)]. Williamson, 2012). The time-averaged flow, therefore, undergoes two
The singular points on the cylinder corresponding to the vortex pair of additional bifurcations, i.e., formation of sub-wake and cat’s eye vorti-
the sub-wake are therefore attachment points rather being separation ces as Re is progressively increased. For Re ¼ 80  150, the streamline
points. The main wake while forms essentially due to separation of the pattern [Figs. 3(a)–3(d)] along the front half of the cylinder is very
laminar boundary layer, the sub-wake does not. The length of the sub- similar and signifies attached flow; the streamlines differ appreciably
wake L2 represents the shortest distance between the sub-wake saddle aft the cylinder. Accordingly, for each Re, the distribution of Cp
point M2 and base point. The presence of a single wake at Re ¼ 80 and becomes very similar along the front lateral edges and noticeably dif-
two wakes at Re ¼ 90 can be also confirmed from the distribution of ferent along the rear lateral edges.
surface vorticity, x shown in Fig. 3(f). Here, h stands for the circum-
ferential angle measured counterclockwise from the forward stagna- 3. The separation topology
tion point. For h ¼ 90  180 , two zero-crossings of x at Re ¼ 80
signify the presence of a single recirculation zone. In contrast, for The separation topology is composed of zero streamlines
Re ¼ 90, three zero-crossings of x ascertain the existence of two con- (streamfunction, w ¼ 0) and singular points where the zero stream-
tiguous regimes of recirculation. lines generally intersect. The cylinder surface is represented by a zero
Owing to extremely small dimension at inception, the pair of streamline and streamlines emanating from or terminating at the cyl-
small eddies forming at the base is not visible at Re ¼ 90 [Fig. 3(b)]. inder surface are also zero streamlines. For steady flow past a diamond
These structures are discernible at Re ¼ 100 as evident from the close- cylinder, Kumar et al. (2018b) and (2019) identified three possible sep-
up in Fig. 3(c). As Re is increased, the streamwise extent of the wake aration topologies prior to the formation of the classical closed wake.
shortens but the sub-wake vortices enlarge. The vortical structures in These topologies respectively correspond to the attached flow
the main wake undergo further alterations at Re ¼ 150. The twin eddy [Fig. 4(a)], distinct separation bubbles on rear lateral edges [Fig. 4(b)],

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-6


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 4. Flow past a stationary diamond cylinder for Re  150: schematic presentation of separation topology for (a) attached flow and (b)–(e) and (g) separated flow. (f)
inspired from Moffatt (1964) shows the vortical structures for flow of an electrically conducting magnetic fluid past a corner.

and joining of the separation bubbles at the base point [Fig. 4(c)]. The fluid-solid interface and correspond to three-way saddles or half-
latter two topologies arise due to secondary separation. Each of these saddles. At A, two streamlines diverge (move away) and one stream-
topologies was shown to satisfy the kinematic constraint proposed by line converges (approaches) while at S, two streamlines converge and
Hunt et al. (1978) for stability of the flow structures, expressed as one streamline diverges. The wake stagnation points or confluence
    points M1 and M2 are four-way saddle points and located in the fluid.
X 1X 0 X 1X 0
Nþ N  Sþ S ¼ 1  n: (6) Two additional four-way saddle points in the fluid are M3 and M4.
2 2 The separation topology for the classical closed wake [Fig. 4(d)] corre-
In Eq. (6), N and S, respectively, denote the four-way nodes and sponds to steady primary separation as well as the mean flow in the
four-way saddles while N 0 and S0 stand for nodes and saddles, respec- unsteady regime for Re <90, approximately.
tively, that are three-way in nature. The connectivity of the section of For the
P twin-wake structure [Fig. 4(e)], points M1 and M2 con-
the flow under consideration is denoted by n. For flow around an iso- tribute to S ¼ 2. The centers P P N1, N2, N3, and N4,
of the eddies, i.e.,
lated obstacle, n ¼ 2. A four-way saddle point is denoted by symbol M. represent four-way nodes, thus, N ¼ 4 and N 0 ¼ 0. The singu-
These points are characterized by two pairs of streamlines—a converg- lar points at the fluid-solid
P interface, i.e., points A1, A2, A3, S1, S2, and
ing pair and a diverging pair. Figure 4 schematically illustrates all the S3, contribute to S0 ¼ 6. Thus, the left-hand side of Eq. (6) simpli-
separation topologies of a diamond cylinder identified for Re  150. fies to 4 þ 0  2  62 ¼ 1. The same value is also reached from the
In this figure, the attachment and separation points are denoted by right-hand side. This verification of the kinematic constraint of Hunt
the symbols A and S, respectively; these points are located at the et al. (1978) ascertains that the twin-wake structure is stable. For

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-7


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

Stokes flow of an electrically conducting magnetic fluid past a sharp influencing the surface pressure, the Cp  h curves for Re ¼ 60, 80, 90,
corner formed by the intersection of two solid planes at high Hartman 100, and 150 are shown together in Fig. 5(f). Irrespective of Re, the
number, Moffatt (1964) analytically investigated the role played by a pressure is symmetric about the base point indicating the vanishing of
line current passing along the intersection of planes. As apparent from the mean lift force. In consistency with the general trend for flow past
the schematic presented in Fig. 8 of Moffatt (1964) and shown here as a bluff body, the value of Cpmax attained at the leading edge or forward
Fig. 4(f), the corner flow is characterized by two pairs of counter- stagnation point exceeds unity. In the unsteady regime, the difference
rotating eddies that preserve symmetry about the plane of symmetry. between the maximum and minimum Cp appears to be smaller than
The wake vortices resolved in the current study for Re  90 and above those in the steady regime. The pressure is positive (toward the cylin-
are quite similar to those predicted by Moffatt (1964) ahead of a sharp der) over a narrow stretch of h about the leading edge; for rest of the
corner despite different contexts of flow. The time-averaged wake of a surface, Cp is negative (away from the cylinder). The pressure under-
diamond cylinder is therefore structurally equivalent to that obtained goes an abrupt drop at the sharp corners or shoulders (h ¼ 90 and
for flow past a corner. The flow near a corner is composed of Moffatt 270 ) and the minimum value of Cp is attained at these points. For a
vortices, i.e., counter-rotating eddies of gradually diminishing size and circular cylinder in contrast, Cp attains the minimum value ahead of
strength. The wake vortices appear analogous to Moffatt vortices. For h ¼ 90 or 270 . The minimum points in Cp denote the limits of h
@C
the final separation topology comprising ofPtwin-wakes andPcat’s eye beyond which the circumferential pressure gradient, @hp changes sign.
P P
vortices [Fig. 4(g)], S ¼ 4; N ¼ 6; N 0 ¼ 0, and S0 ¼ 6. Thus, starting from the leading edge, the pressure gradient for each Re
These parameters along with n ¼ 2 satisfy the constraint of Hunt et al. remains favorable till the shoulders are reached. Owing to this, the
(1978). incoming streamlines closely follow the body contour over the left half
of the cylinder and any possibility of boundary layer separation in this
4. The correspondence between Cp and mean flow region is ruled out. For Re ¼ 60 [Fig. 5(f)], the monotonic rise of Cp
from the shoulders up to the base point renders this region subject
Figure 5 illustrates the distribution of surface pressure for only to adverse pressure gradient. This explains the formation of the
Re ¼ 10  150. As apparent from Figs. 5(a) (for Re ¼ 10) and 5(b) main wake at Re ¼ 60. For Re  80, there appears a second minimum
(for Re ¼ 40) for steady flow, the range of variation of Cp (the differ- in Cp in the base region indicating favorable pressure gradient and
ence between the maximum pressure, Cpmax and minimum pressure, hence the existence of a pair of reattachment points in this region
Cpmin ) decreases with increasing Re. The mean or time-averaged pres- [points A2 and A3 in Fig. 4(g) for Re ¼ 150]. The presence of the sub-
sure in the unsteady regime is shown at Re ¼ 80 [Fig. 5(c)], 90 wake is therefore well consistent with the appearance of the second
[Fig. 5(d)], and 150 [Fig. 5(e)]. To understand the role of Re in minimum in Cp that is otherwise absent for flow past a circular

FIG. 5. Distribution of Cp for flow past a diamond cylinder at Re ¼ (a) 10, (b) 40, (c) 80, (d) 90, and (e) 150. Cp at Re ¼ 60, 80, 90, 100, and 150 are plotted together in (f).

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-8


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

TABLE IV. Laminar flow past a diamond cylinder over Re ¼ 10  150: summary of The mean wake of a circular cylinder [shown in Fig. 6(a) for
the maximum and minimum surface pressure, base pressure, and location of the Re ¼ 100] is known not to have a sub-wake while the mean wake of a
minimum surface pressure.
diamond cylinder does. To understand the differences in mean wake
@C
of these cylinders, the associated Cp and @hp are compared at Re ¼ 100
Re Cpmax or Cp0 Cpmin h for Cpmin Cpb
(Fig. 6). For the circular cylinder, the Cp  h variation involves a
10 2.4432 4.2994 66.9079 0.7080 monotonic fall of Cp followed by its monotonic rise. This underscores
40 1.5357 2.2671 87.9798 0.4915 the existence of a favorable pressure gradient followed by an adverse
pressure gradient [Fig. 6(b)]. For a diamond cylinder in contrast, the
60 1.4265 2.2794 90.6923 0.7567
construction of the Cp  h curve includes an additional segment
80 1.3681 2.2800 90.6923 1.0027
where Cp decreases with h. Overall, Cp initially drops till h  90 , then
90 1.3486 2.2860 90.6923 1.1273 rises, and eventually drops. The minimum pressure of a circular cylin-
100 1.3333 2.2924 90.6923 1.2503 der is reached well ahead of the shoulder or h ¼ 90 [Fig. 6(a)] and
150 1.2961 2.2932 90.6923 1.7692 hence, the adverse pressure gradient sets in relatively early [inset
6(b-i)]. Till the base point, Cp gradually rises and hence the pressure
gradient between the minimum pressure point and base point is essen-
cylinder over identical range of Re. The enlargement of the sub-wake tially adverse. Absence of a favorable pressure gradient nullifies the
with Re [Figs. 3(c) and 3(d)] is associated with stronger favorable pres- possibility of presence of additional reattachment points and hence
sure gradient at the base (the second peak in Cp becomes more nega- that of a bubble at the base. In contrast, for the diamond cylinder, the
tive) that pushes the attachment points away from the base point. The additional Cp  h segment with decaying Cp and a consequent favor-
variation of Cp for various Re [Fig. 5(f)] is quite similar along the front able pressure gradient near the base [Fig. 6(b-ii)] ascertains the forma-
lateral edges while it differs appreciably along the rear lateral edges. As tion of a recirculation. Cpmin of the diamond cylinder is significantly
Re is increased, Cp continues to drop (becomes more negative) along smaller than that of a circular cylinder. The severity of adverse pres-
the rear half surface of the cylinder. Thus, the base pressure decreases sure gradient of the former therefore significantly exceeds the one for
with Re. For Re ¼ 10–150, Table IV lists the values of the base pres- the latter [evident from Fig. 6(b) for h slightly above 90 , also shown
sure, and maximum and minimum surface pressure along with the in the inset 6(b-i)]. This plays a crucial role in rendering the dimen-
location of the minimum pressure. The minimum pressure is attained sions of the wake of the diamond cylinder to appreciably exceed those
invariably at the shoulder signaling a fixed h location for commence- of a circular cylinder. Thus, the magnitude of adverse pressure gradient
ment of the adverse pressure gradient and hence, an almost constant aids the separation of laminar boundary layer more than the location of
value of the separation angle. The maximum change in Cpb over commencement of adverse pressure gradient does. Figure 6(a) reveals
Re ¼ 60–150 is one order of magnitude higher than the change of that Cp0 and Cpb of a diamond cylinder are, respectively, higher and
Cpmax (this quantity also represents Cp0 ) and two orders of magnitude smaller than that of a circular cylinder. Thus, a higher ðCp0  Cpb Þ dif-
higher than the change of Cpmin . Thus, Cpmin exhibits the least sensitivity ference for the diamond cylinder causes lesser recovery of forward pres-
to changes in Re. By analyzing surface pressure data, it is found that the sure at the base. Based on higher loss or lesser recovery of pressure, one
value of Cp at a reattachment point (whether on the solid surface or can predict a higher value of pressure drag, Cdp as well as total drag for a
within the fluid) always exceeds the pressure at the separation points. diamond cylinder as compared to a circular cylinder.

@C
FIG. 6. Comparison of (a) Cp and (b) @hp along the lower half surface of circular and diamond cylinders at Re ¼ 100 and b ¼ 0:05. (a) also shows the wake of a circular cylin-
@C
der at Re ¼ 100. In (b), insets (i) and (ii) show @hp near the shoulder and base, respectively.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-9


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 7. Flow around a stationary diamond cylinder for Re  150: classification of the regimes of flow based on streamline topology.

5. Summary of the flow regimes for Re £ 150 point exhibits dual behavior, it acts as a separation point for regimes I,
II, V, and VI whereas for regimes III and IV, it acts as an attachment
Kumar et al. (2018b) and (2019) identified three regimes of flow point.
past a diamond cylinder for Re  8:18. The first regime or regime I
bounded by Re <7.3 corresponds to attached flow. At Re ¼ 7.3, a pair
of distinct separation bubbles form on the lateral edges of the cylinder B. Relationship between separation angle and
via secondary separation. These bubbles do not merge as long as Re is Reynolds number
smaller than 8.13. The regime II therefore stretches over For the regimes of primary and secondary separations, Fig.
7:3  Re < 8:13. In regime III stretching over Re ¼ 8:13  8:18, the 8(a) plots the variation of separation as well as attachment angles
elongated separation bubbles meet at the base point but do not form a with Re. At the onset (Re ¼ Rss ¼ 7:3) and closure (Re ¼ 8.18) of
wake. The classical closed wake forms first at Re ¼ 8.19 and marks the secondary separation, the values of separation angle are 25.96 and
initiation of regime IV. This regime due to the main/primary separa- 50:72 , respectively. At the onset of primary separation at
tion extends for Re up to 90 and is characterized by the presence of Re ¼ 8.19, hs ¼ 0 . For Re ¼ 8:19  8:2, the separation angle
two counter-rotating eddies within the wake. The state of twin-wakes remains approximately equal to 51 [see Fig. 8(b) which shows the
(Re  90  140) defines the fifth regime (regime V) of flow compris- separation and attachment angles for secondary separation in
ing of two pairs of counter-rotating eddies. An additional regime of close-up]. Therefore, over the regime of secondary separation, hs
flow (regime VI) where each of the primary wake vortices splits into a rises sharply with Re. At the upper limit of steady flow, i.e.,
pair of smaller vortices of same sense of rotation is resolved at Re ¼ 40, the separation angle is as high as 87:94 . From here
Re ¼ 150. The flow regimes are illustrated in Fig. 7. For each regime, onwards, the separation points remain virtually fixed at the should-
the forward stagnation point acts as an attachment point. The base ers of the cylinder implying that hs  90 .

FIG. 8. Flow past a diamond cylinder: (a) variation of separation and attachment angles with Re. (b) shows the close-up of the highlighted box in (a).

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-10


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

Figure 8 also depicts the dependence of the location of reattach- edges slanted at 45 with the horizontal, the transverse spacing
ment points on Re. The reattachment angles for secondary separation between the separated shear layers of a diamond cylinder appears to
and sub-wake are denoted by ha and hA, respectively. At Res, be smaller than those of a circular cylinder where the rear surface is
ha ¼ 17:17 and ha decreases with increasing Re. At the upper extrem- convex and separation point movable. Vigorous interactions between
ity of secondary separation, the attachment point coincides with the the shear layers owing to lesser lateral gap might be responsible for
base point implying that the least value of ha of 0 is attained [Fig. higher vortex-shedding frequency of the diamond cylinder.
8(b)]. The sub-wake reattachment angle increases with Re and the For a stationary symmetric bluff body, the mean lift is zero-
reattachment point A approaches the separation point S [Fig. 8(a)]. valued. Hence, the features of lift force are represented in terms of its
amplitude, either by the rms lift or by the maximum lift. The rms lift,
C. The vortex-shedding frequency and fluctuating computed using the fluctuations of lift, follows an increasing trend
fluid forces with Re [Fig. 9(b)] similar to that of St. A least squares curve fit yields
the following empirical relationship between Cl rms and Re:
The gradual narrowing of the diamond cylinder next to its
shoulders renders the opposite-signed shear layers separated from it to 11:8051
Cl rms ¼ 9:7650  for b ¼ 0:05 and Re ¼ 41  150:
be in close proximity. The vortex-shedding phenomenon and associ- Re0:0525
ated lift force rely on the extent of interactions between these shear (9)
layers. With increasing Re, enhanced interactions between these shear 2
layers result in vigorous vortex-shedding. This is reflected by the The values of standard error and R for Cl rms are 0.006 421 and
monotonic rise of St with Re [Fig. 9(a)]. The St–Re relationship is 0.99, respectively. The rms of drag exhibits a slow rise with Re and is
expressed by the following two-term empirical formula obtained by an order lower in magnitude than the rms of lift. For flow-induced
least squares curve fitting: vibrations of an elliptic cylinder at a ¼ 0 , Kumar et al. (2018a)
explained the order of magnitude differences in Cd rms and Cl rms . They
1:0170 showed that the pressure at the forward and rear stagnation points
St ¼ 0:2668  for b ¼ 0:05 and Re ¼ 41  150:
Re0:5351 oscillates with the frequency of drag while those at the shoulders (Cpt
(7) and Cpd at the top and bottom shoulders, respectively) oscillate with
2 the vortex-shedding frequency or frequency of lift. The pressure at the
The values of standard error and R for the above formula are stagnation points, therefore, represents the features of drag and that at
0.001 34 and 0.99, respectively. The quantity R2 bounded by 0 and 1 is the shoulders accounts for the lift. Kumar et al. (2018a) noted that the
a measure of the goodness of curve fit. For Re ¼ 50  150 flow past a fluctuations in Cp0 and Cpb are lower than those in Cpt and Cpd by an
circular cylinder, Roshko (1954) proposed the following empirical order of magnitude. Thus, the dominance of Cl rms over Cd rms was
equation: explained. For the diamond cylinder, Fig. 10(a) plots the time traces of
  Cp0 and Cpb at Re ¼ 100. The corresponding power spectra overlap
21:2
St ¼ 0:212 1  : (8) with each other [Fig. 10(b)] and exhibit an excellent convergence with
Re
the frequency of Cd. The amplitude of both Cpt and Cpd [Fig. 10(c)]
Equations (7) and (8) suggest that St of diamond and circular cyl- appreciably surpasses those of Cp0 and Cpb. An out of phase relation-
inders vary as Re0:5351 and Re1 , respectively. Thus, the rate of ship between Cpt and Cpd, similar to those reported by Kumar et al.
growth of St is much higher for a diamond cylinder than its circular (2018a) and consistent with alternate shedding of vortices, is apparent
counterpart. For the same characteristic dimension, the area of a dia- from Fig. 10(c). The power spectra of Cp at the shoulders shown in
mond cylinder is smaller than its circular counterpart. For the rear Fig. 10(d) reveal that both Cpt and Cpd oscillate with the vortex-

FIG. 9. Two-dimensional, laminar flow past a diamond cylinder for Re ¼ 41  150: variation of the (a) normalized vortex-shedding frequency and (b) rms of drag and lift forces
with Re.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-11


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 10. Flow past a fixed diamond cylinder at Re ¼ 100: (a) time series of Cp0 and Cpb, (b) power spectra of Cp0 and Cpb, (c) time series of Cpt and Cpd, and (d) power spec-
tra of Cpt and Cpd.

shedding frequency. The order of magnitude difference between Cd rms Rec is determined by solving the linearized Stuart–Landau equation
and Cl rms can therefore be explained with respect to the higher amplifi- (Provansal et al., 1987; Schumm et al., 1994; Sohankar et al., 1998), lin-
cation of pressure (due to vortex-shedding) at the shoulders than at ear stability analysis (Yang and Zebib, 1989; Kumar and Mittal, 2006),
the stagnation points. or direct computation (Henderson, 1995; Rajani et al., 2009). In the
current work, Rec is determined by direct steady and unsteady compu-
tations of the flow around Re ¼ 40. The generation of lift marks the
D. The critical Re for the onset of unsteadiness/ onset of unsteadiness. This fact is utilized to determine Rec from
vortex-shedding the time trace of Cl (see Rajani et al., 2009 for a circular cylinder).
The transition of steady flow to two-dimensional unsteady flow Figures 11(a) and 11(b) illustrate the time series of Cl at Re ¼ 40 and
occurs through a Hopf bifurcation (Jackson, 1987; Provansal et al., 41, respectively. At Re ¼ 40, after the initial transient, the lift stagnates
1987; Noack and Eckelmann, 1994; Kumar and Mittal, 2006) at Rec. to a zero value and signifies the steady nature of flow for Re at least up

FIG. 11. Two-dimensional, laminar flow past a stationary diamond cylinder for b ¼ 0:05: time series of Cl at Re ¼ (a) 40 and (b) 41.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-12


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

to 40. At Re ¼ 41, development of a periodic wave form of Cl confirms the steady regime followed by a non-linear decay in the unsteady
the presence of vortex-shedding. For b ¼ 0:05, Rec of a diamond cyl- regime. Such an L1  Re relationship can be found in Park et al.
inder, therefore, lies between 40 and 41. While the first occurrence of (1998) and Silva et al. (2003), among others for flow past a circular cyl-
Cl signals the onset of unsteadiness, discontinuous changes of certain inder. The elongation of the sub-wake also follows a linear variation
characteristic quantities, such as L1, Cd, Cdp, viscous drag coefficient, with Re as opposed to the non-linear decay of the classical wake
Cdv, and Cp0 , also signify the same. Within the steady regime, the val- length. The inset of Fig. 12(a) illustrates the Re regimes of attached
ues of these quantities obtained from the steady and unsteady compu- flow, secondary separated flow, and primary separated flow.
tations converge. For Re  Rec , the values of these quantities obtained Roshko (1993) argues that the traditional Cd  Re plots fail to
from the steady and unsteady computations essentially diverge. Thus, adequately reveal the sensitivity of flow to changes in Re and hence
the determination of Rec via direct computations boils down to the Cpb justifiably replaces Cd to locate the transitions in a flow.
identification of Re at which the values of the above quantities from However, besides base suction, the drag and its components can also
steady computations mark the first departure with those obtained via be used to precisely determine Rec (Henderson, 1995). The base suc-
unsteady computations. This can be clearly understood from Cd and tion coefficient is a fundamental parameter that exhibits sensitivity to
Cpb plots of Henderson (1995) based on direct numerical simulation various instabilities occurring in a flow over different values of Re
data. (Roshko, 1993; Williamson, 1996). In the neighborhood of the onset
For b ¼ 0:05 and 0.01, respectively, Sohankar et al. (1998) and of laminar vortex-shedding, the variation of Cpb with Re experiences
Yoon et al. (2010) determined the value of Rec of a diamond cylinder a trend change as compared to the variation in the steady regime
to be 42 and 39. Rec ð¼ 41Þ predicted by us using b ¼ 0:05 is in close (Roshko, 1993; Henderson, 1995; Williamson, 1996). For a circular
agreement with these data. The value of Rec for a circular cylinder is cylinder, the Cpb –Re curve (Roshko, 1993; Henderson, 1995;
approximately equal to 47 (Provansal et al., 1987; Norberg, 1994; Williamson, 1996) indicates that Cpb attains its minimum value at
Morzy nski et al., 1999; Kumar and Mittal, 2006). Using a low blockage the upper Re extremity of the steady separated flow. Discontinuity in
of 0.01, Chopra and Mittal (2019) recently determined the value of Rec the slope of Cpb represents a transition from steady to time-
of a circular cylinder as 46.985. For a square cylinder, Jackson (1987) dependent flow and the Cpb - Re curve must contain a kink at Rec
and Schumm et al. (1994) predicted the value of Rec to be approxi- (Rajani et al., 2009). The pressure coefficients Cp0 and Cpb are plot-
mately 50. For the same geometry, using b ¼ 0:025, Sohankar et al. ted against Re in Fig. 12(b). As is apparent from the inset ii of
(1997) obtained Rec ¼ 47 6 2 and for b ¼ 0:05, Sohankar et al. Fig. 12(b), Cpb becomes the minimum (with steady and unsteady
(1998) later determined Rec ¼ 51:2. Among the diamond, circular, solutions converging) at Re ¼ 40. Subsequently, the slope of the
and square cylinders, the value of Rec is, therefore, the minimum for Cpb  Re profile undergoes a change at Re ¼ Rec ¼ 41 so that the
the diamond cylinder. value of Cpb obtained from the unsteady computations surpasses
those obtained from steady computations. The divergence of Cp0 from
1. Rec from the wake length and surface pressure the steady and unsteady computations at Re ¼ 41 is relatively less
prominent (see inset i) as compared to Cpb , yet discernible. Relative
The relationships of the lengths of the classical wake as well as to Re, the pressure coefficients display contrasting behavior; while Cp0
the sub-wake with Re are depicted via Fig. 12(a). The inception of the drops with increasing Re, the base suction decreases in the steady
classical closed wake of a diamond cylinder occurs at Re ¼ 8.19 regime followed by an almost linear rise in the unsteady regime.
(Kumar et al., 2018b, 2019). The L1  Re curve exhibits a discontinu- Concerning flow past a circular cylinder, Roshko (1993) suggests that
ity at Re ¼ 41 signifying the onset of vortex-shedding. Prior to the the increase in Cpb with Re over Re ¼ 50  180 results from
onset of vortex-shedding (Re ¼ 8:19  40), the growth of L1 with Re enhanced wake instability and Reynolds stress in regions near the cyl-
is approximately linear. L1 attains its peak value at the upper limit of inder. For flow past a flat plate at a ¼ 90 and having a splitter plate

FIG. 12. Flow past a diamond cylinder: variation of the (a) length of the main and sub-wake and (b) pressure coefficients with Re. The inset in (a) demarcates the regimes of
attached flow, secondary separated flow, and primary separated flow.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-13


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

attached to it, Roshko (1993) used the “free streamline” or “cavity mod- flow quantities considered by Roshko (1993) include the wake length
el” of Riabouchinsky (1921) and found that increase in Cpb results in (taken from Riabouchinsky, 1921) and drag coefficient. Following the
shortened wake length. Comparing the behaviors of wake length and lead of this figure, we plot the variation of DL with Cpb in Fig. 14(a)
base suction in the steady and unsteady regimes, we conclude that the for Re ¼ 8:19  150. Owing to the inverse relationship of L and Cpb
wake length varies inversely with base suction. Concerning flow past an with Re, the L-Cpb curves corresponding to the steady and unsteady
axisymmetric body at higher Reynolds numbers, such as Re ¼ 300, 800, regimes converge at a sharp peak; the sharp peak marks the upper
1500, and 9:6  104 , Mariotti et al. (2015) had similar observations. extremity of the steady regime of flow. For Re ¼ 38  40, the curve is
nearly vertical implying that the steady regime terminates very close to
2. Rec from drag and its components Re ¼ 40. This figure also indicates that the steady to unsteady transi-
tion occurs close to the least value of base suction and accordingly, the
The magnitude of Cdp varies directly with pressure recovery at maximum value of wake length. The value of Cpb at Re ¼ 40 is
the cylinder base. In the steady regime, Cp0 drops rapidly with Re while 0.4915 and it is 0.5311 at Re ¼ 41. The value of base suction at Rec is
Cpb decays at a much slower rate [Fig. 12(b)]. Accordingly, therefore bracketed by 0.4915 (approximately) and 0.5311.
(Cp0  Cpb ) decays rapidly or pressure recovery increases [Fig. 13(a)] In a manner similar to Fig. 14(a), the total drag and its compo-
causing sharp drops in Cd and Cdp [Fig. 13(b)]. For unsteady flow, Cp0 nents are plotted against base suction in Fig. 14(b). In the steady
decays slowly while Cpb increases rapidly with Re. The value of the regime, the decrease in base suction (or increase in Re) lowers the pres-
ðCp0  Cpb Þ difference becomes the minimum at the upper limit of sure drag but increases the recirculation length. Thus, elongation of
the steady flow, i.e., Re ¼ 40 [see the inset of Fig. 13(a)]. At this Re, the the wake corresponds to decrease in Cdp (or Cd) and vice versa. Such
pressure drag and total drag attain their respective minimum values as inverse relationship between the length of mean recirculation region
[see Fig. 13(b)]. For a circular cylinder, a well-defined transition kink and base drag has been reported by Mariotti et al. (2015) in conjunc-
at Rec characterizes the Cd and Cdp curves while the kink is relatively tion with high Re flow past an axisymmetric bluff body. The pressure
less prominent in Cdv (see Henderson, 1995). For the diamond cylin- drag in the unsteady regime increases linearly with base suction. The
der, an obvious trend change exists in the ðCp0  Cpb Þ vs Re curve at viscous drag decays in the steady as well as unsteady regimes. This
Rec ¼ 41. Beyond the critical Reynolds number, ðCp0  Cpb Þ increases decay appears to be associated with an increasing Re and not with
almost linearly with Re. The widening gap between Cp0 and Cpb with changes in base suction.
increasing Re in the unsteady regime is reflected by the increasing VII. CONCLUSIONS
nature of Cd and Cdp. In the unsteady regime, Cp0 registers very slow
decay and hence, increasing base suction amounts to increase in the The two-dimensional, laminar, incompressible flow around a sta-
pressure (or total) drag. The divergence of the steady and unsteady tionary diamond cylinder is investigated for Re  150 using a stabi-
results over Re  Rec can be clearly observed in the total drag, its pres- lized finite-element solver. The flow is found to be steady for Re up to
sure component [inset (i) of Fig. 13(b)], and viscous component [inset 40, approximately and vortex-shedding commences at some Re
(ii) of Fig. 13(b)]. With increasing Re, the dominant convection mode between 40 and 41. The time-averaged flow undergoes a pair of topo-
logical bifurcations as the value of Re exceeds 80, approximately. The
of transport suggests weakening of viscous effects and a consequent
first topological bifurcation at Re  90 involves the inception of a sub-
decaying nature of Cdv.
wake attached to the base. The presence of a sub-wake is indicated by
the presence of a second minimum in the surface pressure distribution
E. The relationship of base suction with wake length close to the base region. This local minimum signifies the existence of
and drag a region of favorable pressure gradient, thus presence of attachment
Figure 4 of Roshko (1993) plots several key characteristic flow points and hence, presence of a sub-wake. The formation of sub-wake
quantities in the unsteady regime as functions of the base suction. The appears to be a strong function of body geometry. The absence of a

FIG. 13. Two-dimensional, laminar flow past a diamond cylinder: the variation of (a) ðCp0  Cpb Þ and (b) Cd, Cdp, Cdv with Re for Re ¼ 7:3  150.

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-14


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 14. Two-dimensional, laminar flow past a diamond cylinder: the relationship of (a) the length of the classical closed wake and (b) drag and its components with base
suction.

sub-wake for a circular cylinder and its presence for a diamond cylin- unsteady flows. The L1  Cpb and Cd  Cpb plots indicate an inverse
der appear to imply that a sub-wake generally forms for cylinders with relationship between wake length and drag.
straight edges and having gradually narrowing cross section toward
the base. The advent of sub-wake underlines the existence of two sad- DATA AVAILABILITY
dle points of opposite nature and located on the wake centerline. The The data that support the findings of this study are available
saddle point M1 of the primary wake signifies attachment while the from the corresponding author upon reasonable request.
saddle point M2 of the sub-wake indicates separation. As an outcome
of the second bifurcation, each of the standing vortices of the primary REFERENCES
wake splits into twin like-sign vortices forming the cat’s eye vortices.
Alam, M. M., Abdelhamid, T., and Sohankar, A., “Effect of cylinder corner radius
The separation topologies resulting from the bifurcations are shown to and attack angle on heat transfer and flow topology,” Int. J. Mech. Sci. 175,
be stable. As the value of Reynolds number is progressively increased 105566 (2020).
to 150, a total of six different regimes of separation topology are identi- Behr, M., Hastreiter, D., Mittal, S., and Tezduyar, T. E., “Incompressible flow past
fied. The base point exhibits a dual nature; it acts as a separation point a circular cylinder: Dependence of the computed flow field on the location of
for regimes I, II, V, and VI whereas for regimes III and IV, it behaves the lateral boundaries,” Comput. Methods Appl. Mech. Eng. 123(1-4), 309–316
as an attachment point. (1995).
Brons, M., Jakobsen, B., Niss, K., Bisgaard, A. V., and Voigt, L. K., “Streamline
For a separation bubble on the cylinder surface or the classical
topology in the near wake of a circular cylinder at moderate Reynolds
wake, the value of surface pressure at a reattachment point always numbers,” J. Fluid Mech. 584, 23–43 (2007).
exceeds the corresponding value at a separation point. This holds valid Chopra, G., and Mittal, S., “Drag coefficient and formation length at the onset of
for attachment points located at the fluid-solid interface and also vortex shedding,” Phys. Fluids 31(1), 013601 (2019).
within the fluid. The separation of laminar boundary layer appears to Dutta, S., Panigrahi, P. K., and Muralidhar, K., “Experimental investigation of
be more a function of the magnitude of adverse pressure gradient than flow past a square cylinder at an angle of incidence,” J. Eng. Mech. 134(9),
788–803 (2008).
the location where the adverse pressure gradient commences. In the
Feng, D. K., Zhang, Z. G., Wang, L., and Mudaliar, A. V., “Numerical analysis of
steady and unsteady regimes, respectively, both Cp0 and Cpb vary fluid flow around diamond cylinders,” Chin. J. Comput. Mech. 27(4), 618–623
non-linearly and linearly with Re. This appears to result in a non- (2010).
linear variation of drag (and its components) in the steady regime fol- Franke, R., Rodi, W., and Schonung, B., “Numerical calculation of laminar
lowed by linear variation in the unsteady regime. Stronger fluctuation vortex-shedding flow past cylinders,” J. Wind Eng. Ind. Aerodyn. 35, 237–257
in surface pressure at the shoulders than at the stagnation points (1990).
results in Cl rms significantly exceeding Cd rms . The critical Reynolds Henderson, R. D., “Details of the drag curve near the onset of vortex shedding,”
Phys. Fluids 7(9), 2102–2104 (1995).
number signifying the onset of vortex-shedding is determined by Hunt, J. C. R., Abell, C. J., Peterka, J. A., and Woo, H., “Kinematical studies of
direct computations. The values of the characteristic flow quantities, the flows around free or surface-mounted obstacles; applying topology to flow
such as wake length, pressure coefficients, and drag coefficient, visualization,” J. Fluid Mech. 86(1), 179–200 (1978).
obtained using steady and unsteady computations are found to con- Jackson, C. P., “A finite-element study of the onset of vortex shedding in flow
verge. The Re at which the computed values start diverging signifies past variously shaped bodies,” J. Fluid Mech. 182, 23–45 (1987).
Rec ð¼ 41Þ. Similar to Re, the base suction also reflects the transition Jahromi, J. A., Nezhad, A. H., and Behzadmehr, A., “Effects of inclination angle
on the steady flow and heat transfer of power-law fluids around a heated
of steady flow to its unsteady counterpart. The (steady to unsteady) inclined square cylinder in a plane channel,” J. Non-Newtonian Fluid Mech.
transition kink in the CdCpb plot is much more prominent that the 166(23-24), 1406–1414 (2011).
one in the Cd  Re plot. The wake length and cylinder drag plotted Jiang, H., and Cheng, L., “Flow separation around a square cylinder at low to
against base suction sharply demarcate the regimes of steady and moderate Reynolds numbers,” Phys. Fluids 32(4), 044103 (2020).

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-15


Published under an exclusive license by AIP Publishing
Physics of Fluids ARTICLE scitation.org/journal/phf

Kim, I., “On the morphology of two-dimensional laminar vortex streets behind Roshko, A., “Perspectives on bluff body aerodynamics,” J. Wind Eng. Ind.
triangles,” Phys. Fluids 33(3), 033601 (2021). Aerodyn. 49(1-3), 79–100 (1993).
Kuhlmann, H. C., Wanschura, M., and Rath, H. J., “Flow in two-sided lid-driven Roy, A., and Bandyopadhyay, G., “A finite volume method for viscous incom-
cavities: Non-uniqueness, instabilities, and cellular structures,” J. Fluid Mech. pressible flows using a consistent flux reconstruction scheme,” Int. J. Numer.
336, 267–299 (1997). Methods Fluids 52(3), 297–319 (2006).
Kumar, A., Dhiman, A. K., and Bharti, R. P., “Power-law flow and heat transfer Sarkar, S., Mondal, C., Manna, N. K., and Saha, S. C., “Forced convection past a semi-
over an inclined square bluff body: Effect of blockage ratio,” Heat Transfer- circular cylinder at incidence with a downstream circular cylinder: Thermofluidic
Asian Res. 43(2), 167–196 (2014). transport and stability analysis,” Phys. Fluids 33(2), 023603 (2021).
Kumar, B., and Mittal, S., “Prediction of the critical Reynolds number for flow Sasmal, C., and Chhabra, R. P., “Momentum and heat transfer characteristics of a
past a circular cylinder,” Comput. Methods Appl. Mech. Eng. 195(44-47), long parallelepiped submerged in power-law fluids in the laminar vortex shed-
6046–6058 (2006). ding regime,” Int. J. Heat Mass Transfer 55(9-10), 2285–2314 (2012).
Kumar, D., Mittal, M., and Sen, S., “Modification of response and suppression of Schumm, M., Berger, E., and Monkewitz, P. A., “Self-excited oscillations in the
vortex-shedding in vortex-induced vibrations of an elliptic cylinder,” Int. J. wake of two-dimensional bluff bodies and their control,” J. Fluid Mech. 271,
Heat Fluid Flow 71, 406–419 (2018a). 17–53 (1994).
Kumar, D., Sourav, K., Sen, S., and Yadav, P. K., “Steady separation of flow from Sen, S., Mittal, S., and Biswas, G., “Steady separated flow past a circular cylinder
an inclined square cylinder with sharp and rounded base,” Comput. Fluids 171, at low Reynolds numbers,” J. Fluid Mech. 620, 89–119 (2009).
29–40 (2018b). Sen, S., Mittal, S., and Biswas, G., “Flow past a square cylinder at low Reynolds
Kumar, D., Sourav, K., Yadav, P. K., and Sen, S., “Understanding the secondary numbers,” Int. J. Numer. Methods Fluids 67(9), 1160–1174 (2011).
separation from an inclined square cylinder with sharp and rounded trailing Sharify, E. M., Saito, H., Harasawa, H., Takahasi, S., and Arai N., “Experimental
edges,” Phys. Fluids 31(7), 073607 (2019). and numerical study of blockage effect on flow characteristics around a square-
Luzzatto-Fegiz, P., and Williamson, C. H. K., “Structure and stability of the section cylinder,” J. Jpn. Soc. Exp. Mech. 13, s7–s12 (2013).
finite-area von Karman street,” Phys. Fluids 24(6), 066602 (2012). Sharma, B., and Barman, R. N., “Steady laminar flow past a slotted circular cylin-
Mariotti, A., Buresti, G., and Salvetti, M. V., “Connection between base drag, sep- der,” Phys. Fluids 32(7), 073605 (2020).
arating boundary layer characteristics and wake mean recirculation length of Silva, A. L. F., Lima, E., Silveira-Neto, A., and Damasceno, J. J. R., “Numerical
an axisymmetric blunt-based body,” J. Fluids Struct. 55, 191–203 (2015). simulation of two-dimensional flows over a circular cylinder using the
Moffatt, H. K., “Viscous and resistive eddies near a sharp corner,” J. Fluid Mech. immersed boundary method,” J. Comput. Phys. 189(2), 351–370 (2003).
18(1), 1–18 (1964). Sohankar, A., and Najafi, M., “Control of vortex shedding, forces and heat trans-
Morzy nski, M., Afanasiev, K., and Thiele, F., “Solution of the eigenvalue prob- fer from a square cylinder at incidence by suction and blowing,” Int. J. Therm.
lems resulting from global non-parallel flow stability analysis,” Comput. Sci. 129, 266–279 (2018).
Methods Appl. Mech. Eng. 169(1–2), 161–176 (1999). Sohankar, A., Norberg, C., and Davidson, L., “Numerical simulation of unsteady
Moussaoui, M. A., Jami, M., Mezrhab, A., and Naji, H., “MRT-lattice Boltzmann flow around a rectangular two-dimensional cylinder at incidence,” J. Wind
simulation of forced convection in a plane channel with an inclined square cyl- Eng. Ind. Aerodyn. 69-71, 189–201 (1997).
inder,” Int. J. Therm. Sci. 49(1), 131–142 (2010). Sohankar, A., Norberg, C., and Davidson, L., “Low-Reynolds-number flow
Mukhopadhyay, A., Biswas, G., and Sundararajan, T., “Numerical investigation around a square cylinder at incidence: Study of blockage, onset of vortex shed-
of confined wakes behind a square cylinder in a channel,” Int. J. Numer. ding and outlet boundary condition,” Int. J. Numer. Methods Fluids 26(1),
Methods Fluids 14(12), 1473–1484 (1992). 39–56 (1998).
Noack, B. R., and Eckelmann, H., “A global stability analysis of the steady and Tezduyar, T. E., Mittal, S., Ray, S. E., and Shih, R., “Incompressible flow computa-
periodic cylinder wake,” J. Fluid Mech. 270, 297–330 (1994). tions with stabilized bilinear and linear equal-order-interpolation velocity-
Norberg, C., “An experimental investigation of the flow around a circular cylin- pressure elements,” Comput. Methods Appl. Mech. Eng. 95(2), 221–242
der: Influence of aspect ratio,” J. Fluid Mech. 258, 287–316 (1994). (1992).
Park, J., Kwon, K., and Choi, H., “Numerical solutions of flow past a circular cyl- Turki, S., Abbassi, H., and Nasrallah, S. B., “Effect of the blockage ratio on the
inder at Reynolds numbers up to 160,” Korean Soc. Mech. Eng. Int. J. 12(6), flow in a channel with a built-in square cylinder,” Comput. Mech. 33(1), 22–29
1200–1205 (1998). (2003).
Provansal, M., Mathis, C., and Boyer, L., “Benard-von Karman instability: Vijaybabu, T. R., Anirudh, K., and Dhinakaran, S., “LBM simulation of unsteady
Transient and forced regimes,” J. Fluid Mech. 182, 1–22 (1987). flow and heat transfer from a diamond-shaped porous cylinder,” Int. J. Heat
Posdziech, O., and Grundmann, R., “A systematic approach to the numerical cal- Mass Transfer 120, 267–283 (2018).
culation of fundamental quantities of the two-dimensional flow over a circular Williamson, C. H. K., “Vortex dynamics in the cylinder wake,” Annu. Rev. Fluid
cylinder,” J. Fluids Struct. 23(3), 479–499 (2007). Mech. 28, 477–539 (1996).
Rajani, B. N., Kandasamy, A., and Majumdar, S., “Numerical simulation of lami- Yang, X., and Zebib, A., “Absolute and convective instability of a cylinder wake,”
nar flow past a circular cylinder,” Appl. Math. Modell. 33(3), 1228–1247 Phys. Fluids A 1(4), 689–696 (1989).
(2009). Yoon, D. H., Yang, K. S., and Choi, C. B., “Flow past a square cylinder with an
Riabouchinsky, D., “On steady fluid motions with free surfaces,” Proc. London angle of incidence,” Phys. Fluids 22(4), 043603 (2010).
Math. Soc. 2(1), 206–215 (1921). Zaki, T. G., Sen, M., and Gad-El-Hak, M., “Numerical and experimental investi-
Robichaux, J., Balachandar, S., and Vanka, S. P., “Three-dimensional gation of flow past a freely rotatable square cylinder,” J. Fluids Struct. 8(7),
Floquet instability of the wake of square cylinder,” Phys. Fluids 11(3), 560–578 555–582 (1994).
(1999). Zhu, H., Tang, T., Zhou, T., Liu, H., and Zhong, J., “Flow structures around trap-
Roshko, A., “On the drag and shedding frequency of two-dimensional bluff bod- ezoidal cylinders and their hydrodynamic characteristics: Effects of the base
ies,” Technical Report No. NACA TN 3169 (1954). length ratio and attack angle,” Phys. Fluids 32(10), 103606 (2020).

Phys. Fluids 33, 053611 (2021); doi: 10.1063/5.0049811 33, 053611-16


Published under an exclusive license by AIP Publishing

Das könnte Ihnen auch gefallen