1 s2.0 S0997754621000212 Main

European Journal of Mechanics / B Fluids 87 (2021) 135–150
Contents lists available at ScienceDirect
European Journal of Mechanics / B Fluids

journal homepage: www.elsevier.com/locate/ejmflu
Steady flow past a circular cylinder under large blockage

∗
Shravan Kumar Mishra, Subhankar Sen , Akhil Verma
Department of Mechanical Engineering, Indian Institute of Technology (Indian School of Mines) Dhanbad, Dhanbad 826 004, India
article info a b s t r a c t
Article history: The flow past a circular cylinder constrained under large blockage exhibits features even in the steady
Received 4 May 2020 regime, that strikingly contrast those corresponding to its unbounded counterpart. While several
Received in revised form 20 October 2020 studies are dedicated towards understanding this classical flow problem in unbounded medium,
Accepted 30 January 2021
surprisingly, very little efforts have been put forward to extract details of flow alterations in a highly
Available online 5 February 2021
confined flow. The absence of flow solutions and comprehensive data acts as an impetus behind the
Keywords: current work that computes steady, two-dimensional flow over the Reynolds number, Re range of
Stabilized finite-element 4–100 under a large blockage of 0.9. The imposition of large blockage stabilizes the flow; it enhances
Steady flow the Re limit of fore-aft flow symmetry to 5, delays the onset of laminar separation to Re = 27.8 and
Circular cylinder preserves the steady nature of flow even at Re = 100. The wake grows non-linearly with Re. For
Large blockage Re ≤ 5, the base pressure represents the minimum surface pressure and eliminates any development
Surface pressure of adverse pressure gradient along the cylinder. The order of magnitude rise in the drag and drop of
Drag
surface pressure are overwhelming. There exists a competition of pressure gradient on the sidewalls
and cylinder surface for 6 ≤ Re < 15.
© 2021 Elsevier Masson SAS. All rights reserved.
1. Introduction numerical treatment also differentiates the unbounded flow from

its confined counterpart [5].
For the last many decades, the flow past a circular cylin- For flow in a finite domain, a low blockage below which
der is perhaps the most explored fluid dynamics problem in the key characteristic flow quantities do not exhibit any signif-
the realm of external flow past an isolated bluff obstacle. This icant deviation marks a boundary between the unbounded and
classical flow problem has been the subject matter of several confined flows. Based on experimental findings for flow past
excellent overviews [1–3] besides completely devoted texts [4]. a circular cylinder over Reynolds number, Re between 104 and
Despite the abundant wealth of knowledge available on this 105 , West and Apelt [6] concluded that the surface pressure, Cp
flow, the observation by Sahin and Owens [5] that the vast distribution as well as drag, Cd exhibit less sensitivity to blockage
majority of investigations relate mostly to the flow in an un- while the Strouhal number, St remains invariant provided B <
bounded/unconfined medium and hardly to the confined flow, 0.06. Based on experiments in an oil tunnel for steady separated
is equally relevant even after one and half a decade. In analyt- flow past a circular cylinder at high Re, Grove et al. [7] noted
ical treatment of a flow problem, the flow domain stretches to that the surface pressure distribution at Re ≈ 175 for B =
infinite extents in all directions and hence, corresponds to an 0.05 is representative of unbounded flow. Through numerical
unbounded flow problem. However, in context of experimental or experiments on the location of lateral boundaries for flow past
computational approach of handling the flow problem, the flow a circular cylinder at Re = 100, Behr et al. [8] determined that
domain of interest is essentially finite. For such finite domains, positioning of the lateral boundaries from the cylinder centre at
the degree of confinedness is represented by blockage, B. Blockage least by eight times the cylinder diameter (or B ≤ 0.0625) renders
is the ratio of two length scales D and H, respectively signifying the vortex-shedding phenomenon insensitive to B. In conjunction
the cross-stream projection of an object and width of the flow with the onset of unsteadiness for flow past a stationary circular
domain or experimental apparatus, i.e. B = HD . Accordingly, 0 ≤ cylinder, Fig. 2 of Mittal et al. [9] suggests that the values of
B < 1 and B = 0 or H → ∞ signifies the experimental or the critical Reynolds number, Rec and critical St for HD = 40
computational parallel of unbounded flow as considered in the (or B = 0.025) are very similar to the respective values with
analytical treatment of a flow problem. Besides the transverse H
D
= 100 (or B = 0.01). For steady flow past a stationary circular
extent of the domain, the choice of boundary conditions in a cylinder, Sen et al. [10] found that the flow solutions do not alter
significantly for B ≤ 0.04 and become identical for B ≤ 0.01.
∗ Corresponding author. They concluded that the steady unbounded flow corresponds to
E-mail address: ssen@iitism.ac.in (S. Sen). B = 0.005. In light of the above observations, it therefore appears
https://doi.org/10.1016/j.euromechflu.2021.01.014
0997-7546/© 2021 Elsevier Masson SAS. All rights reserved.
S.K. Mishra, S. Sen and A. Verma European Journal of Mechanics / B Fluids 87 (2021) 135–150
that the effects of blockage on the characteristics of flow past a the confined flow, they conducted linear stability analysis and
stationary circular cylinder are less appreciable for values of B explored the effect of blockage on the critical Re corresponding
smaller than 0.05, approximately. It should, however, be noted to Hopf and Pitchfork bifurcations. The modifications of drag
that the blockage effects become stronger at lower Re and the and wake parameters of a circular cylinder due to imposition of
domain width should extend to 2000D or higher, i.e. B ≤ 0.0005 blockage was investigated by Chakraborty et al. [18]. Subject to
to achieve asymptotic solutions [11]. In the following paragraphs, the towing tank boundary condition, they simulated the steady
the prior investigations concerning the effects of lateral wall flow over Re = 0.1 − 200 and B = 0.05 − 0.65. When Re
confinement on the characteristics of flow past a circular cylinder was held constant, increasing B resulted in lesser values of Cd
are reviewed. because of additional retarding force offered by the walls on
The effects of confining walls on the stability of steady wake the cylinder. Kumar and Mittal [19] studied the dependence of
of a circular cylinder were analysed by Shair et al. [12] through critical Re and St on B. They conducted a linear stability analysis
experiments in a closed oil tunnel. For B = 0.05 − 0.2, they found of the conservation equations for 0.005 ≤ B ≤ 0.125. With
that the tunnel walls stabilize the steady wake or delay the onset increasing B, the critical Re was found to initially drop and then
of vortex-shedding. They argued that the disturbances responsi- rise. By performing extensive two-dimensional spectral-element
ble for wake instability are generated across the free-stream and computations for Re = 5 − 250, Posdziech and Grundmann [11]
get suppressed by the tunnel walls. For steady separated flow past studied the effects of domain extension and mesh resolution on
a circular cylinder at high Re, i.e. Re ≈ 30 − 300, Grove et al. the drag, base pressure, lift and shedding frequency of a circular
[7] presented limited results for surface pressure signifying the cylinder. The quantity H was varied from 40D to 8000D, yielding
wall effects on the flow. A comprehensive experimental investi- a wide range of low blockage varying from 0.000125 to 0.025.
gation on the onset of laminar separation and wake instability, Relative to the asymptotic solution at B = 0.000125, they found
dimensions of the twin eddies, location of eddy centres, contour that the difference in drag for B = 0.025 at Re = 5 is about 7%
of the wake boundary and distribution of streamwise velocity while at Re = 20 and 40, the difference drops to about 4%.
along the wake axis is due to Coutanceau and Bouard [13]. They The role of blockage (for 0.000125 ≤ B ≤ 0.8) along with
conducted flow visualization experiments for steady flow around boundary conditions in influencing the onset of steady laminar
a circular cylinder in a towing tank for Re ranging from separation separation from a circular cylinder was numerically investigated
Reynolds number, Res to Rec . Through experiments in an open by Sen et al. [10]. Two sets of boundary conditions (slip and
circuit wind tunnel for 104 < Re < 105 , West and Apelt [6] towing tank) were employed for each blockage. They also used
explored the effects of tunnel blockage and cylinder aspect ratio a third set of boundary condition involving parabolic inlet and
on the surface pressure distribution, base pressure or Cpb , wake no-slip sidewalls. With this set, they presented limited results at
frequency (measured by St) and drag. For unsteady flow past a Re = 40 for a large blockage of 0.9. By employing finite-volume
circular cylinder over Re = 60 − 180, Stansby and Slaouti [14] method on unstructured mesh, Singha and Sinhamahapatra [20]
studied the effects of blockage on St and aerodynamic forces. presented solutions for confined flow past a circular cylinder over
For unsteady flow past a circular cylinder, Behr et al. [8] Re = 45 − 250. The blockage was varied from 0.125 to 0.5. At Re =
compared the St, Cd and the maximum value of lift, Cl obtained 45 corresponding to steady flow, they found that increase in B
from five different meshes having the domain widths of 9D, results in decrease of wake length, separation angle and increase
12.5D, 16D, 24D and 32D, respectively. For domain width ex- in drag. In addition, the cylinder surface pressure was entirely
ceeding 16D, the results displayed less than 1% deviation. For a positive at the highest blockage of 0.5. For B = 0.125 and 0.167
computational domain restricted laterally by no-slip sidewalls, flow at Re = 100, they observed the phenomenon of reverse
Chen et al. [15] numerically explored the role of blockage on the Karman vortex street where the shed Karman vortices switch
formation of the closed wake behind a circular cylinder and also their positions about the wake centreline little downstream the
the onset of unsteadiness of flow. Concerning the appearance of near wake. In context of non-Newtonian fluid flow, Rao et al.
the steady wake, they varied B from 0.1 to 0.95. For determining [21] computationally explored the effects of wall confinement
the Rec , the blockage was varied from 0.1 to 0.7. Anagnostopoulos (B = 0.167, 0.25 and 0.5) on steady and unsteady flow quantities
et al. [16] discretized the streamfunction–vorticity equations via for Re = 40 − 140. Table 1 provides a summary of the previous
finite-element method and studied the effects of blockage on the works that deal with the influence of blockage.
characteristics of flow around a cylinder. For steady and unsteady As apparent from the above discussion, the maximum value of
flows at Re = 106, three values of blockage, i.e. 0.05, 0.15 and blockage up to which results are available in the open literature
0.25 were considered. With rise in blockage, the values of the is 0.95 due to Chen et al. [15] and Chen [17]. However, detailed
separation angle, θs and drag increased while the wake length, accounts of steady flow at large blockage are unavailable in these
L shortened. The effects of blockage (for B = 0.1 − 0.95) and studies. For a representative Re of 40 and various B between 0.005
boundary conditions on the onset of steady laminar separation and 0.8, Table 2 summarizes the values of Cd obtained by Sen
were investigated by Chen [17] via finite-element computations. et al. [10] using the slip and towing tank boundary conditions.
The boundary conditions employed were: (i) fully developed or For comparison, results obtained by using the no-slip sidewalls
parabolic inlet and no-slip stationary sidewalls, (ii) free-stream are also included. It is obvious from this table, that the solutions
inlet and no-slip stationary sidewalls and (iii) towing tank bound- obtained from the slip and towing tank boundary conditions con-
ary condition (lateral boundaries are translating with free-stream verge well till B = 0.04. However, results with no-slip boundary
speed, U). For the third set of boundary conditions and B > condition differ even with B = 0.01. This deviation is even more
0.1, he noted non-linear elongation of the wake bubble with pronounced in the streamline plots presented in Fig. 7 of Sen et al.
Re. However, for the other two sets of boundary conditions, the [10]. Thus, the choice of boundary conditions assumes paramount
elongation of the wake with Re was linear. importance for flows constrained under high blockage. Following
For flow past a circular cylinder over 0 < Re ≤ 280, Sahin the lead of Sahin and Owens [5], we employ in the present work,
and Owens [5] employed a velocity-only finite-volume formula- the no-slip sidewalls and report two orders of magnitude rise
tion and studied the wall effects on the stability of flow, fluid (relative to unbounded flow) of the drag and noticeable drop of
forcing, wake dynamics and interactions between the wake and the surface pressure. We are unaware of any prior study that
sidewall boundary layers. They presented results for unconfined reports the details of steady flow under extremely large block-
(B = 0.01) as well as confined (0.1 ≤ B ≤ 0.9) flows. For age and no-slip sidewalls. While Chakraborty et al. [18] focus
136
Table 1
Flow past a stationary circular cylinder subject to varying blockage: summary of the earlier studies. The maximum blockage for
which results are available is 0.95 (highlighted in the table). Here, FEM, FVM and SEM stand for finite-element method, finite-volume
method and spectral-element method, respectively.
Study B Re Method
Shair et al. [12] 0.05–0.20 Rec Experiments
Grove et al. [7] 0.05, 0.1, 0.1875 ≈ 30−300 Experiments
Coutanceau and Bouard [13]a 0, 0.024, 0.07, 0.12 > 5, < 40 Experiments
West and Apelt [6] 0.02–0.16 > 104 , < 105 Experiments
Stansby and Slaouti [14] 0–0.5 60–180 Random vortex method
Behr et al. [8] 0.0315–0.111 100 FEM
Chen et al. [15] 0.1–0.95 Res and Rec FEM
Anagnostopoulos et al. [16] 0.05, 0.15 and 0.25 106 FEM
Chen [17] 0.1–0.95 Res FEM
Sahin and Owens [5] > 0.1, ≤ 0.9 > 0, ≤ 280 FVM
Chakraborty et al. [18] 0.05–0.65 0.1–200 FVM (ANSYS-Fluent)
Kumar and Mittal [19] 0.005–0.125 Rec FEM
Posdziech and Grundmann [11] 0.000125 − 0.025 5–250 SEM
Sen et al. [10] 0.000125–0.80 6–40 FEM
Singha and Sinhamahapatra [20] 0.125–0.5 45, 100, 150, 200, 250 FVM
Rao et al. [21] 0.167, 0.25 and 0.5 40–140 FVM (ANSYS-Fluent)
a
The definition of blockage used by Coutanceau and Bouard [13] is different from the definition used in this Section.
Table 2
Steady separated flow past a stationary circular cylinder at Re = 40: summary
of cylinder drag obtained by using slip, towing tank and no-slip conditions at
the lateral boundaries of the computational domain for different values of B.
The results for the slip and towing tank boundary conditions are taken from
Sen et al. [10].
B Slip Towing tank No-slip
0.005 1.51 1.51 1.51
0.01 1.51 1.51 1.50
0.04 1.57 1.57 1.49
0.07 1.65 1.66 1.50 Fig. 1. Problem definition for highly confined (B = 90%) steady flow around a
0.10 − − 1.54 stationary circular cylinder.
0.11 1.78 1.82 −
0.15 1.95 2.01 1.66
0.20 2.20 2.30 1.83 2. Governing equations and finite-element formulation
0.80 45.57 87.23 63.18
Let Ω ⊂ R2 be the spatial domain. The boundary of Ω ,
denoted by Γ , is assumed to be piecewise smooth. The spatial
coordinates are denoted by x (= x, y). Using vector notation, the
primarily on the drag, Singha and Sinhamahapatra [20] presents
strong form of the conservation equations governing the steady,
results for steady flow only at a single Re of 45. On the other incompressible flow of a Newtonian fluid of density, ρ , are:
hand, the results of Rao et al. [21] widely correspond to non-
Newtonian fluid flow. Thus, a fundamental and comprehensive ρ (u · ∇ u − f) − ∇ · σ = 0 on Ω , (1)
study concerning steady flow of a Newtonian fluid past a circular ∇ · u = 0 on Ω . (2)
cylinder under extremely large blockage is unavailable in the lit-
Here u (= u, v ), f and σ denote the fluid velocity, body force
erature. This research gap serves as the prime motivation behind
per unit volume and the Cauchy stress tensor, respectively. In
the current work. The research questions dealt with are: does the this study, f = 0. As the sum of inviscid (isotropic) and viscous
imposition of blockage alter the regime of fore-aft symmetry of (deviatoric) components, the constitutive relation for the stress
flow? to what extent does the value of unbounded Res alter? does tensor at a point reads as: σ = −pI+µ[(∇ u)+(∇ u)T ]. Here, p, I, µ
the wake bubble elongate linearly with Re as is generally seen? if and ∇ u are the pressure, identity tensor, dynamic viscosity of the
not, why? what is the impact of the sidewall shear layer on L and fluid and velocity gradient tensor, respectively. The flow problem
θs ? to what extent do the magnitudes of cylinder surface pressure under consideration is associated with essential (Dirichlet) as well
and drag change? how do the sidewalls influence the pressure as natural (Neumann-type) boundary conditions. These boundary
gradient around the cylinder? how does the characteristics of conditions are represented as
highly confined flow compare with those of unconfined flow? is u = g on Γg , n · σ = h on Γh , (3)
the highly confined flow at very low Re equivalent to unbounded
flow at low Re? Our efforts to analyse these follow the use of respectively, where Γg and Γh are complementary subsets of Γ , n
a stabilized finite-element solver [22] in two-dimensions. For a is its unit normal vector and h is the surface traction vector. The
blockage of 0.9, results are presented for Re = 4 − 100. boundary conditions employed in this work have been defined
in Section 3.1. The finite-element formulation is discussed in
The remaining of this article is organized in the following man-
Tezduyar et al. [22].
ner. The governing equations for steady, incompressible flow are
reviewed in Section 2. Section 3 discusses the problem statement 3. Problem description
as well as finite-element mesh. The validation of the formula-
tion along with mesh insensitivity of the predicted results are 3.1. Problem set-up
addressed in Section 4. Section 5 covers the presentation and dis-
cussion of the main results. In Section 6, few concluding remarks A circular cylinder of diameter, D is placed symmetrically be-
are made. tween the upper and lower boundaries of a channel. A rectangular
137
Fig. 2. The non-uniform and three-blocked finite-element mesh used for computing the highly confined steady flow past a fixed circular cylinder. The mesh is
composed of 90460 nodes and 89472 bilinear quadrilateral elements.
domain of size 95D × 10 9

D, as depicted via Fig. 1, is used as
the computational domain. The width, H of the domain, i.e. 10 9
D
implies that the value of blockage equals 90%. The centre of
the cylinder also marks the location of the origin of the Carte-
sian coordinate system. Relative to the origin, the inlet and exit
boundaries of the domain are located at streamwise distances
of Lu = 32D and Ld = 63D, respectively. For 0.1 ≤ B ≤
0.9, Sahin and Owens [5] employed a streamwise distance of
40D for both the upstream and downstream boundaries. A fully
developed velocity profile defined as u = 1 − [2 H ]2 and v = 0
y
is prescribed as the essential boundary condition at the inlet.

The centreline or maximum streamwise velocity, Uc at the inlet
along with characteristic dimension of the cylinder, D is used to
define the Reynolds number of flow, i.e. Re = Uνc D . The essential
or Dirichlet-type no-slip, no-penetration conditions on velocity,
i.e. u = 0, v = 0 are imposed at all the rigid surfaces, i.e. the
Fig. 3. Steady separated flow past a stationary circular cylinder subject to a
channel walls as well as fluid–cylinder interface. A traction-free large blockage of 90%: comparison of predicted drag coefficient at Re = 10,
(n.σσ = h = 0) or stress-free boundary condition is specified at 20, 40 and 100 with those obtained by Sahin and Owens [5] via finite-volume
the exit boundary. This is a natural or Neumann-type boundary computations.
condition. For confined flow, the boundary conditions employed
by us are identical to those employed by Sahin and Owens [5]
except for the exit boundary. flow with those obtained by Sahin and Owens [5] via finite-
volume computations. As stated in Section 3.1, the boundary
3.2. The finite-element mesh conditions considered at the inlet, cylinder surface and sidewalls
are identical in both of these studies. For Re = 10, 20, 40 and 100,
Computation of the high blockage flow warrants the need of the comparison of Cd is illustrated in Fig. 3. A maximum deviation
a highly resolved mesh. As highlighted in Fig. 2, a non-uniform of 3.1% in predicted Cd is noted at Re = 20. Therefore, close
structured mesh composed of three mesh blocks (left, central and agreement of the drag from both sets of computations ascertains
right) is used for computations of the flow. Construction of the the accuracy of the predicted results. It may be noted that the
mesh follows the algebraic method. A total of 90460 nodes and values of Cd from the present set of computations at Re = 10, 20,
89472 bilinear quadrilateral elements are used in constructing 40 and 100 are 1574.55, 788.52 and 396.83, 167.96, respectively
the mesh. The central square mesh block is of size 10 D × 10 D
9 9 (also listed later in Table 4).
and it accommodates the cylinder. The mesh in this block is non-
Cartesian in nature, i.e. it is comprised of one family of radial
and another family of circumferential grid lines. Each radial line 4.2. Mesh convergence
contains nr + 1 = 82 + 1 = 83 nodes while a circumferential
line accommodates nt (= 96) nodes in each quadrant. Therefore, a
Mesh independence of the predicted results has been estab-
total of 4nt = 4 × 96 = 384 uniformly spaced radial grid lines and
lished by comparing at Re = 40, the values of certain flow
83 non-uniformly spaced circumferential grid lines generate this
quantities obtained from two different meshes M1 and M2. The
block. The number of nodes and number of elements in the cen-
structure of M1 has been discussed in Section 3.2. Taking M1 as a
tral block are 31872 and 31488, respectively. The shortest radial
reference mesh, M2 has been constructed by increasing the mesh
distance between the circumferential grid lines is set between the
cylinder surface (first circumferential grid line) and the second resolution in the left, central and right blocks as much uniformly
circumferential grid line. The value of this distance, denoted by as possible. With this refinement, M2 contains 206974 nodes and
hr1 , is 0.0005D. Subsequently, the radial distance increases non- 205360 quadrilateral elements, i.e. the resolution of M2 exceeds
uniformly. The left and right mesh blocks are both structured and the double of the resolution of M1. A summary of the key mesh
Cartesian. parameters in the central block along with the characteristic flow
quantities is provided in Table 3. The definition of the pressure
4. Validation and convergence of computed results coefficient is provided in Section 5.4. The wake length or eddy
length is measured from the base point or rear stagnation point.
4.1. Comparison with available results As apparent from this table, results do not exhibit any sensitivity
to mesh refinement beyond that of M1. This leads us to select
The accuracy of the computed results has been established mesh M1 for the entire set of computations of steady flow under
via comparison of the predicted drag coefficient for B = 90% high confinement.
138
Fig. 4. Results from unsteady computations of flow past a stationary circular cylinder at Re = 100 under a large blockage of 0.9: time traces of (a) Cd and (b) Cl
imply that the flow at Re = 100 is indeed steady. The closed wake (Fig. c) and vorticity contours (Fig. d) further ensure the steady nature of flow.
Table 3 For Re = 100, additional computations of the unsteady Navier–

Steady separated flow past a circular cylinder subject to a large blockage of 90% Stokes equations of motion have been performed over sufficiently
at Re = 40: effect of mesh refinement on the characteristics of flow. Also shown
are the key mesh parameters for the central block.
long time. It has been established that the flow at Re = 100 is
L indeed steady.
Mesh nr nt Nodes Elements D
Cd Cpb
M1 82 96 90460 89472 0.13 396.83 −382.89
5.1. Illustration of steady nature of flow at Re = 100
M2 110 136 206974 205360 0.13 397.04 −383.01
For flow past a circular cylinder at low blockage, the onset of

Table 4 vortex-shedding or equivalently, transition of steady flow to its
Steady flow past a circular cylinder under large blockage of 0.9: summary of unsteady counterpart, occurs at a Reynolds number of the order
characteristic flow quantities for Re = 4 − 100.
of 48 (Williamson [23] experimentally obtained the critical Re
Re Cdp Cdv Cd Cp0 Cpb Cp min θ for Cp min
value of 47.9). Kumar and Mittal [19], through a linear stability
4 3684.50 250.13 3934.63 1 −3800.09 −3800.09 180◦ analysis, determined the value of the transition Re to equal 47.336
5 2947.75 200.10 3147.85 1 −3040.05 −3040.05 180◦
6 2456.61 166.75 2623.36 1 −2533.38 −2533.39 169.69◦
for a blockage of 0.01. Sahin and Owens [5] conducted linear
10 1474.51 100.04 1574.55 1 −1520.24 −1520.56 153.75◦ stability analysis in two-dimensions and provided the values of
20 738.53 49.99 788.52 1 −761.02 −762.55 135◦ critical Reynolds as well as Strouhal numbers over a wide range
30 493.79 33.29 527.08 1 −508.58 −511.96 124.69◦ of blockage bounded by 0.1 and 0.9. For B = 0.9, they found the
40 371.91 24.92 396.83 1 −382.89 −388.70 117.19◦
flow to become unsteady for Re > 160.5. Thus, imposition of a
60 251.15 16.54 267.69 1 −258.52 −269.68 109.69◦
63 239.79 15.74 255.53 1 −246.85 −258.76 108.75◦ large blockage significantly delays the onset of vortex-shedding
64 236.25 15.50 251.75 1 −243.22 −255.37 108.75◦ or increases the value of critical Re. Based on the results of Sahin
65 232.82 15.25 248.07 1 −239.70 −252.09 108.75◦ and Owens [5] for B = 0.9, one therefore expects the flow to
66 229.50 15.02 244.52 1 −236.30 −248.91 107.81◦ be essentially steady at Re = 100, the upper limit of Re in the
67 226.29 14.79 241.08 1 −233.00 −245.85 107.81◦
68 223.17 14.57 237.74 1 −229.81 −242.89 107.81◦
present work. To test this, the B = 0.9 flow at Re = 100 is
69 220.15 14.36 234.51 1 −226.71 −240.02 107.81◦ computed using the unsteady Navier–Stokes equations of motion.
70 217.22 14.15 231.37 1 −223.71 −237.22 107.81◦ Figs. 4a and 4b, respectively depict the time variation of drag
80 192.17 12.35 204.52 1 −198.08 −213.60 105.94◦ and lift coefficients. At larger time, the value of drag converges
100 158.10 9.86 167.96 1 −163.36 −181.76 103.13◦
to a constant value of 167.83. No fluctuation is apparent also in
the lift signal signifying the absence of vortex-shedding. The Cl is
thus zero valued, as expected for a symmetric object. These time
5. Results traces therefore ascertain that the flow at Re = 100 belongs to the
steady regime. Fig. 4c illustrate the closed wake via streamlines.
As characteristic to a steady flow, the standing wake is symmetric
Two-dimensional stabilized finite-element computations are about the x-axis, the wake streamlines are closed and a wake
performed to analyse the steady flow past a stationary circular stagnation point exists marking the terminal point of the wake.
cylinder placed in an extremely narrow channel leading to a The vorticity generated at the solid surfaces, i.e. cylinder surface
blockage of 90%. For Re = 4 − 100, flow solutions have been as well as the channel walls, is transported via convection and
computed from the steady Navier–Stokes equations of motion. diffusion modes. For a highly confined flow, the diffusion mode of
139
Fig. 5. Steady flow past a circular cylinder under a large blockage of 0.9: depiction of flow via streamlines for Re = (a) 5, (b) 20, (c) 33, (d) 40, (e) 60, (f) 80 and
(g) 100, respectively.
transport is important even at high Re. The transport of vorticity

in the shear layers forming on the cylinder surface, is obstructed
via opposite-signed vorticity generated on the nearest sidewall
(Fig. 4d). The vorticity generated at the walls penetrates deeper
to the channel centreline along the downstream and restrict the
transport of shear layer away from the cylinder. Thus, the cylinder
shear layers do not stretch sufficiently downstream the cylinder
and consequently, the wake becomes shorter in stretch (also
see [10]). It may be noted that the vorticity contours preserve
symmetry about the x-axis.
5.2. The flow
To illustrate the development of flow with Re, streamline

contours at representative Re values of 5, 20, 33, 40, 60, 80
and 100 are shown in Fig. 5. At Re = 5 (Fig. 5a), streamlines
in each quadrant appear identical signifying fore-aft and top–
bottom symmetry. This kind of overall symmetry is characteristic
to potential flow past a circular cylinder or a highly viscous
unbounded flow in the regime of creeping flow (Re ≪ 1). For
values of Re smaller than unity, the mode of transport via diffu-
sion dominates overwhelmingly the transport due to convection
and fore-aft symmetry of flow exists away from and close to a
symmetric obstacle [24]. Therefore, the Re = 5 flow with B = 0.9
is similar to that seen in the regime of creeping flow. This further
corroborates with the observation of Section 5.1 that imposition
of blockage extends the regimes corresponding to unbounded
flow towards higher Reynolds numbers. This state of top–bottom
and fore-aft symmetry prevails at least up to Re = 5. It is
discussed later in Section 5.4 that a favourable pressure gradient
(along the cylinder surface) that is symmetric about the cylinder Fig. 6. Steady flow past a circular cylinder constrained under a large blockage
shoulder is associated with fore-aft symmetry of streamlines. As of 0.9: contours of u velocity at Re = (a) 27.7, (b) 27.8, (c) 27.9, (d) 28, (e) 29
Re is increased, the convection of vorticity continues to assume and (f) 30. Fig. g illustrates the distribution of surface vorticity for Re = 27.7,
27.8 and 27.9.
importance and the fore-aft symmetry is lost. The occurrence of
separation is indicative of stronger convective transport of vor-
ticity than transport by diffusion [17]. The fore-aft asymmetry of
streamlines is well evident for a representative Reynolds number 5d (for Re = 40), 5e (for Re = 60), 5f (for Re = 80) and
of 20 (Fig. 5b) where the boundary layer is still attached to the 5g, each ensures further loss of fore-aft symmetry but continued
cylinder surface. At Re = Res = 27.8, a tiny separation bubble existence of top–bottom or reflective symmetry of streamlines.
forms at the base of the cylinder (see Fig. 6b) marking the oc- Therefore, imposition of large blockage does not induce top–
currence of separation of laminar boundary layer. Large blockage bottom asymmetry of flow in the steady regime. This symmetry
therefore causes more than a four-fold increase of the corre- further underlines that the surface pressure distribution along
sponding value of separation Reynolds number. For unbounded the upper and lower halves of the cylinder must be balanced or
flow, both Brons et al. [25] using B = 0.02 and Sen et al. [10] using identical for each Re. Accordingly, in absence of unbalanced Cp ,
B ≤ 0.01 numerically determined that Res = 6.29. Figs. 5c (for no lift force will be generated. As characteristic to such flows,
Re = 33) through 5g (for Re = 100) illustrate the development of the closed wake consists of twin eddies wherein the streamlines
the closed steady wake with Re. The streamline plots for Figs. 5c, are essentially closed and concentric. Concerning the closed wake,
140
Fig. 8. Relationship between L −θs curves for unbounded (B = 0.005) and highly
confined (B = 0.9) flows. These two families of curves are identical up to point
Fig. 7. Steady flow past a circular cylinder under large blockage (B = 0.9): A and afterwards, they diverge. This suggests that the bubble for highly confined
variation of wake length and separation angle with Reynolds number. The flow will elongate linearly with Re up to point A and post this point, the L − Re
separation angle is measured from the base point. It may be noted that the relationship will turn out to be non-linear.
L − Re relationship is linear up to Re about 50.
increases non-linearly with Re for B = 0.125, 0.25 and 0.5. The

a noteworthy observation is that its maximum lateral extent or shift of L − Re profile from linearity marks a significant effect
width is always attained at the locations of separation and hence of imposition of blockage. A line of best fit is obtained as DL =
falls shorter than D. −0.38 + 0.01Re. For unbounded flow, Sen et al. [10] obtained
To resolve the separation bubble at inception, u velocity con- the following L − Re relationship (Equation 7.5 of their paper):
tours in the base region are illustrated in Fig. 6 for certain values D
L
= −0.4240 + 0.0668Re for Res ≤ Re ≤ 40. The slope of the line
of Re in the neighbourhood of separation Reynolds number. As with unbounded flow is much higher than for the confined flow,
obvious from the first row of Fig. 6, the boundary layer is attached signifying considerable shrinkage in the size of the wake due to
for Re up to 27.7 (Fig. 6a). The wake bubble appears first at blockage. For confined flow, the value of DL at the maximum Re of
Re = 27.8 (Fig. 6b) and continues to grow in size with increas- 100 considered is 0.96. For unbounded flow, this value of bubble
ing Re (Figs. 6c through 6f). This value of separation Reynolds length is attained at a much lower Reynolds number of 20.68 as
number determined from u velocity contours is identical to the obtained from the empirical L − Re relation of Sen et al. [10]. This
one obtained from the distribution of surface vorticity, ω. Fig. 6g comparison provides a quantitative indication of the impact of
plots the surface vorticity for Re = 27.7, 27.8 and 27.9. The large blockage.
close-up of the ω − θ curve close to the base is shown in inset. While the blockage considerably shortens the wake besides
Here, θ is the circumferential angle. The vorticity distribution for inducing non-linearity in L − Re profile, the separation angle in
each Re reveals anti-symmetry about the base point, i.e. θ = contrast, does not display any remarkable sensitivity to blockage.
180◦ . For Re = 27.7, vorticity along the entire lower surface is Alternately, blockage does not appreciably influence the position
positive and along the entire upper surface, it is negative. The of separation points. It is found that, the order of magnitude of
phenomenon of separation is associated with a change of sign of θs over the [Res , 100] range for B = 0.9 is very similar to the
surface vorticity over 0◦ < θ < 180◦ and 180◦ < θ < 360◦ . order of magnitude of θs over [Res , 40] for unbounded flow. This
Absence of a change of sign of ω at Re = 27.7 (see the inset) is implies that the value of θs at Re = 20.68 for unbounded flow
an indicator of attached flow. As the value of Re is increased to must be smaller than the value of θs at the maximum Re value
27.8, vorticity along the lower surface changes sign from positive of 100 for confined flow. The value of bubble length is however,
to negative ahead of the base point. This sign change signifies identical in both cases. The implication of this match in L and
the existence of a recirculation that forms as an outcome of mismatch in θs is that the θs − L relationship for unbounded
boundary layer separation. The corresponding zero vorticity point flow yielding linear L − Re variation does not hold for B = 0.9.
marks the lower separation point. Anti-symmetry of ω implies This is consistent with L − Re being non-linear for B = 0.9. For
another sign change along the upper surface and existence of the B = 0.005 (unbounded flow) and 0.9, Fig. 8 plots the relationship
upper separation point. For Re = 27.9, the change of sign of ω between bubble length and separation angle for Res < Re ≤ 40
occurs well ahead of the base point, implying increased extent and Res < Re ≤ 100, respectively. Up to point A (B = 0.9, Re = 50
of the recirculation and hence, a bigger separation angle. The or B = 0.005, Re = 10), the L − θs curves are roughly identical.
identification of separation and attachment points from surface Since the L − Re relationship for B = 0.005 is linear for all Re, we
vorticity was proposed by Lighthill [26]. conclude that the L − Re relationship for B = 0.9 will be linear up
to Re ≈ 50. Since these two families of curves are not similar
5.3. Dimensional aspects of the wake beyond point A and location of separation points do not alter
significantly due to blockage, the bubble length beyond Re ≈ 50
The eddy length, separation angle and maximum width, W for confined flow must increase non-linearly with Re. Therefore,
characterize the symmetric, closed wake. Fig. 7 illustrates the in relation to bubble dimensions, the unbounded and highly
dependence of DL and θs on Re. While the wake bubble elongates confined flows behave similarly up to point A. This, however, does
with increasing Re, the L − Re relationship is not truly linear. not necessarily suggest that these two cases of flow are identical
In context of steady confined flow past an equilateral triangular up to point A. It may be noted that for B = 0.005 and Re = 10, the
cylinder placed symmetrically in the flow such that a corner faces value of Cd is 2.80 whereas Cd = 319.11 for B = 0.9 and Re = 50.
the flow, Dhiman and Kumar [27] found that the bubble length Considering the two families of L − θs curves, for comparable
141
dominant as the Re is increased (see Fig. 10d for Re = 60, 10e

for Re = 80 and 10f for Re = 100). Thus, the circumferen-
tial extent of wake increases with increasing Re. For the cases
of separated flow, the presence of a saddle point close to the
cylinder base is noted. This point has been shown in Fig. 10e
for Re = 80. Appearance of the vorticity saddle point close to
the base point seems to be an indicator of separated flow. An
interesting observation from Fig. 10 is that the thickness of the
cylinder shear layers just above or below the cylinder shoulders is
roughly an invariant (independent of Re). The implication is that
the wall shear layers do not significantly influence the vorticity
dynamics along the cylinder circumference. The wall shear layers
rather interact more with the downstream convection of vorticity
from the shear layers on the cylinder surface. As the value of Re
exceeds 40, the gap between the wall shear layers near the flow
centreline increases. This allows the vorticity of cylinder shear
Fig. 9. Variation of normalized streamwise velocity component with y for Re = layers to make further inroad in the downstream direction. This
5 − 100. The velocity component is considered in the gap along the y-axis and
is associated with elongation of the wake. In summary, the shear
between the upper shoulder of the cylinder and upper sidewall. It may be noted
that the boundary layer on the cylinder is attached at Re = 5, 10 and 20. layers forming on the sidewalls are responsible for shortening of
the eddy length compared to those of unbounded flow while the
shear layers generated on the cylinder surface are responsible for
maintaining high value of separation angle.
maximum values of θs , the maximum value of bubble length for
The effect of large blockage on the wake profile is remarkable
unbounded flow is more than two times the maximum value for
in that, along the ψ = 0 streamline emanating from the sepa-
confined flow.
ration points, the width of the wake continues to decay as one
The increase in the value of θs with Re can be explained by
travels towards the wake stagnation point (see Fig. 11b and also
analysing the streamwise or u velocity in the gap along the y-
the sketch in Fig. 11c). Thus, the maximum width of the wake for
axis and between the cylinder shoulder and channel sidewalls
B = 0.9 is attained at the location of separation. For unbounded
and also from the vorticity field. The distribution of normalized
flow, such wake profiles form at Reynolds numbers little above
streamwise velocity along part of the y-axis, i.e. the vertical gap
Res . For instance, the wake profile presented by Sen et al. [10]
between the upper shoulder of the cylinder and upper sidewall
at Re = 8 and B = 0.01 (see Fig. 6c of their paper) is similar
of the channel is illustrated in Fig. 9 for Re = 5 − 100. This flow
to the wake profile shown in Fig. 11b. The gradual narrowing of
section being the narrowest, the flow is highly accelerated. The
wake width under large blockage is a highly contrasting feature
normalized centreline velocity at inlet undergoes an approximate
relative to the nature of wake profile for unbounded flow at
ten-fold rise in this region. For each Re, the velocity profile is
Re ≫ Res where the maximum width of wake is attained at
parabolic. The velocity profiles for various Re differ in the neigh-
some distance downstream the separation points. This is evident
bourhood of centreline velocity such that the maximum velocity
from Fig. 11a for B = 0.01, Re = 40 and also from Fig. 19 of
drops as Re is progressively increased (shown in close-up). Thus,
Coutanceau and Bouard [13]. For B = 0.9, the wall shear layers
a decelerated flow with increasing Re separates early resulting in
push inwards the shear layers from the cylinders and towards
higher value of θs .
the wake centreline. As discussed above, the effect of wall shear
For attached as well as separated flows, Fig. 10 shows the
layers increase in the downstream direction and the shear layers
vorticity field at various representative Reynolds numbers. At
cause gradual shrinkage of the wake across the flow direction.
Re = 5 pertaining to the creeping flow-like regime, thick shear
Thus the wake profile never develops to its standard shape as in
layers of opposite sign form on the channel walls and spread
Fig. 11a. The dependence of W on Re is illustrated via Fig. 11c.
to the centre of the channel (Fig. 10a). The shear layers formed
Since W denotes the vertical distance between the points of
on the upper and lower surfaces of the cylinder are sandwiched
separation, D2 sin θs = W2 or W D
= sin θs . Since θs increases non-
by the sidewall shear layers. Overall, a fore-aft symmetry of the
linearly with Re (Fig. 7) and sin θ is a monotonically increasing
vorticity contours, as expected in a diffusion dominated flow, is
function of θ for 0◦ ≤ θ ≤ 90◦ , the wake width W D
or sin θs must
identified. As seen from Fig. 10b, the convection of vorticity at
exhibit a non-linear rise with Re.
Re = 20 becomes appreciable and streamwise stretching of the
shear layers on the cylinder becomes quite apparent. The fore-aft
5.4. Distribution of surface pressure
symmetry no longer exists. In both of these cases corresponding
to attached flow, the circumferential distance between the shear
Irrespective of the value of Re, the pressure coefficient is
layers along the cylinder surface in the base region is minimum. p−p
defined as Cp = 0.5ρref
U2
where pref signifies the reference pressure.
At Re = 40 (Fig. 10c) post Res , downstream convection of vorticity
in the cylinder shear layers becomes more prominent. In addition, In general, the free-stream pressure, p∞ , i.e. pressure at the
fraction of vorticity from either of the shear layers forming on inlet boundary is used as the reference pressure. As discussed
the cylinder surface penetrate the opposite-signed shear layer on by Posdziech and Grundmann [11], difficulty arises to properly
define the reference pressure. In context of flow past a circular
the cylinder. A close-up near the base region depicts this effect.
cylinder, the pressure at the inlet was used as pref by Henderson
The divergence of the shear layers away from the cylinder surface
[28] while Braza et al. [29] considered the forward stagnation
due to opposite-signed penetrated vorticity indicates the lack of
pressure, p0 as the reference pressure. Braza et al. [29] defined
supply of vorticity near the base region and hence, presence of +0.5ρ U 2
p−p
ref 0 p−p +0.5ρ U 2
a separation bubble. It may be noted that such penetration of the pressure coefficient as Cp = 0.5ρ U 2
= 0.5ρ U 2
(see
opposite signed vorticity is absent for attached flow. This pen- Section 4.1, page 90 of their paper). Posdziech and Grundmann
etrated vorticity in effect deflects the opposite-signed adjacent [11] pointed out that for either of these choices, the distance Lu
shear layer and increases the circumferential distance between of the inlet boundary assumes importance to reduce deviations
the shear layers along the cylinder surface. This effect turns more in the value of (base) pressure coefficient. This is because the
142
Fig. 10. Steady flow past a circular cylinder under extremely large blockage of 0.9: vorticity contours for Re = (a) 5, (b) 20, (c) 40, (d) 60, (e) 80 and (f) 100,
respectively. The vorticity saddle point is highlighted in Fig. e.
Fig. 11. Steady separated flow past a circular cylinder at low Re: wake contours for (a) B = 0.01 flow at Re = 40 and (b) B = 0.9 flow at Re = 100. Fig. c depicts
the dependence of maximum width of the wake on Re.
value of pressure at the inlet or forward stagnation point vary be defined only by subtracting the same quantity from p. Thus,
with Lu . As pointed out by Posdziech and Grundmann [11], the the inlet pressure cannot be used as the reference pressure for
differences in (base) pressure coefficient using both the choices defining the pressure coefficient. To retain consistency in the
for pref become more apparent for short Lu . values of Cp , we employ the definition used by Braza et al. [29],
In the current study, we employ a coupled velocity–pressure p−p +0.5ρ U 2
0
i.e. Cp = 0.5ρ U 2
. This definition of Cp renders a unit value of
finite-element formulation. This formulation does not use an the forward stagnation pressure coefficient for each Re.
explicit equation for pressure and hence, no initial or boundary The most impactful effect of large blockage occurs perhaps on
condition is required to compute the pressure field. Thus, the the cylinder surface pressure distribution. The surface pressure
pressure is an unknown quantity at all the nodes of the com- profile undergoes fundamental modifications, the order of magni-
putational domain. With respect to the fully developed parabolic tude of Cp sees a huge drop. The Cp along the lower half surface of
velocity profile at the inlet, the pressure at the inlet does no the cylinder (0◦ ≤ θ ≤ 180◦ ) for B = 0.01 and 0.9 are compared
longer represent the free-stream pressure. This is apparent from in Fig. 12 for Re = 5 and 40, respectively. For B = 0.01 (Fig. 12a),
Figs. 19d, e, f for values of Cp at the lower left corner of the the surface pressure at Re = 5 is maximum at the forward
computational domain, corresponding to Dx = −32. As revealed stagnation point. As one travels away from the forward stagnation
by Figs. 19d, e, f, the inlet pressure is found to vary with the point (θ = 0◦ ) along the cylinder surface or as θ increases, Cp
Reynolds number. This implies that use of the inlet pressure as turns negative close to θ = 40◦ and remains negative thereafter.
pref amounts to subtraction of a variable pressure from p. Pressure The minimum value of Cp is attained at θ ≈ 117◦ next to the
being a relative variable (page 130 of [30]), a consistent Cp can cylinder shoulder followed by a recovery/rise of pressure. The
143
Fig. 12. Steady flow past a circular cylinder under low (B = 0.01, first column, Figs. a and d) and large (B = 0.9, second and third columns) blockages at Re = 5
and 40: comparison of cylinder surface pressure profile. In this figure, the third column (Fig. c for Re = 5 and Fig. f for Re = 40) shows the pressure gradient.
existence of a minimum in Cp − θ plot signifies a change of sign is entirely positive at Re = 45. They, however, defined the
∂C p−p
of the pressure gradient, ∂θp , i.e. the pressure gradient along the pressure coefficient as Cp = 0.5ρ U∞2 . By using the same definition
cylinder surface is favourable till the point of minimum pressure, as used by Singha and Sinhamahapatra [20], it is found in the
Cp min is reached. The gradient turns adverse next to the minimum present work that the base pressure is positive for the entire
pressure point. A large blockage of 0.9 induces drastic alterations range of Re considered. In addition, for Re up to 63, the surface
of the pressure distribution at the same Re (see Fig. 12b). The pressure is positive throughout. As already stated, we follow the
surface pressure turns negative at an angle as small as 15◦ and as definition of Cp due to Braza et al. [29]. For Re = 4 − 100, Fig. 13
θ approaches 180◦ , its value decreases by several hundred times. illustrates the development of surface pressure distribution. The
The maximum (= 1) and minimum (= −3040.05) values of Cp are Cp − θ curve at Re = 4 is similar in construction to the one
attained at the forward and rear stagnation points, respectively. for Re = 5 discussed above. At Re = 6, the point of minimum
For steady unbounded flow around a circular cylinder at Re = 1, pressure shifts upstream from the base point (see Table 4) sig-
Fig. 16 of Dennis and Shimshoni [31] for Cp −θ variation illustrates nalling the first occurrence of the adverse pressure gradient. The
monotonic decay of Cp with θ and attainment of minimum Cp at adverse pressure gradient though initiated at Re = 6, is weak
the base point; the base pressure being a negative quantity. As and continues to strengthen with rise in Re. It becomes strong
evident from Fig. 12b, Cp is virtually constant over θ ≈ 120◦ to enough at Re = 27.8 to cause the separation of boundary layer.
180◦ . Besides, Cp maintains its least value. A sharp drop in surface With increasing Re, the value of the minimum pressure increases
pressure takes place ahead of this range of θ . The absence of a further (becomes less negative) and the location of the minimum
minimum point prior to the base point or absence of pressure pressure travels upstream (Fig. 13b for Re = 30 and 13c for Re =
recovery implies that the confined flow at Re = 5 is strictly 60). As apparent from Figs. 13d, 13e and 13f for Re = 64, 70 and
under favourable pressure gradient (Fig. 12c) and hence, must be 100, respectively, the regime of negative pressure downstream
attached. It may be noted that the favourable pressure gradient is of the shoulder weakens (turns less negative) with rising Re. The
symmetric about the lower shoulder of the cylinder. This further minimum pressure point does however, never advance ahead of
corroborates with the fact that the flow at Re = 5 displays fore-aft the cylinder shoulder. This is a contrasting feature relative to
symmetry. unbounded flow at Re ≫ Res , such as 40 (Fig. 12d). Overall, the
The surface pressure for B = 0.01 does not exhibit any quali- value of Cp at a given θ increases with increasing Re.
tative change when the value of Re is increased to 40 (Fig. 12d). Fig. 8 indicates that the value of separation angle for B = 0.9
The values of the minimum pressure and base pressure increase, at Re = 100 surpasses the one for B = 0.005 at Re = 40.
i.e. become less negative. In addition, the location of the mini- The higher value of separation angle in the former case is due
mum pressure travels upstream, signifying a wider region under to the penetrated opposite-signed vorticity and divergence of
adverse pressure gradient and a consequent larger wake. In con- shear layer on the cylinder away from the wall. The magnitude of
trast, for B = 0.9, the Cp − θ profile at Re = 40 undergoes large pressure gradient along the cylinder wall also appears to be a key
quantitative alterations as evident from Fig. 12e. The magnitude factor governing the location of separation points. Figs. 14a and
of Cp drops noticeably relative to Re = 5. The base point no longer 14b, respectively suggest that the extent of the region of adverse
represents the location of minimum pressure; the minimum pres- pressure gradient for B = 0.01 at Re = 40 is bigger compared
sure is attained at θ = 117.19◦ . Therefore, a regime of pressure to those with B = 0.9 at Re = 100. However, the magnitude of
∂ Cp
recovery exists between this location and the base point. The non- ∂θ
for the confined flow overwhelmingly overshadows its coun-
monotonic variation of Cp (decrease and subsequent increase) terpart for the unbounded flow. The dominance of the adverse
with θ signifies the existence of a regime of adverse pressure pressure gradient for confined flow appears to contribute to the
gradient (Fig. 12f). The existence of the adverse pressure gradient earlier separation and higher θs .
is consistent with the fact that at Re = 40, the flow is separated. For steady separated flow at high Reynolds numbers, the nu-
For steady flow around a circular cylinder using B = 0.5, merically computed surface pressure of Fornberg [32] shows pos-
Singha and Sinhamahapatra [20] noted that the surface pressure itive base pressure at Re = 500 and 600 (Fig. 14a of their paper).
144
Fig. 13. Steady flow past a circular cylinder over Re = 4 − 100 for B = 0.9: the distribution of surface pressure for Re = (a) 4, (b) 30, (c) 60, (d) 64, (e) 70 and (f)
100.
Fig. 14. Comparison of pressure gradient along the lower half surface of the cylinder for (a) B = 0.01, Re = 40 and (b) B = 0.9, Re = 100.
For flow in an unbounded medium, Cpb is a negative quantity. In shear stress, τw ∗ Re illustrated in the second row of the figure. The
such cases, the base suction or −Cpb decays with Re for steady vorticity and shear stress curves for fixed Re are identical. Thus,
flow and increases with Re for two-dimensional unsteady flow the viscous drag component, Cdv of total drag can be computed
(Fig. 3 of [3]). For steady unbounded flow, such a Cpb − Re either from surface vorticity or from shear stress. Dennis and
relationship signifies a rise in the value of Cpb with increasing Chang [33] employed streamfunction–vorticity formulation for
Re (see Fig. 5 of [33]). For steady confined flow, the Cpb − Re solving the steady flow past a circular cylinder and they used
relationship exhibits an identical trend. As evident from Fig. 15, surface vorticity for computing Cdv . In the present work, shear
Cpb continues to rise as Re is increased. stress along the cylinder surface has been used to compute the
viscous drag. The values of vorticity as well as shear stress shoot
5.5. Distribution of cylinder surface vorticity and shear stress up (or down) at the shoulder region where velocity gradient is
very high owing to flow through the constriction. The symme-
The distribution of normalized surface vorticity and normal- try of surface pressure and anti-symmetry of surface vorticity
ized shear stress along the cylinder surface is shown in Fig. 16 at or shear stress about the base point ensures that the lift force
representative Reynolds numbers of 4, 40 and 100. on the cylinder must vanish. The symmetric or anti-symmetric
The surface vorticity curves (first row) display anti-symmetry distribution is a function of whether the bluff obstacle is itself
about the base point. Same holds true for the normalized wall geometrically symmetric or not.
145
Fig. 15. Variation of the base pressure coefficient with Re for steady flow past Fig. 17. Cd − Re relationship for a highly confined (B = 0.9) steady flow past a
a circular cylinder subject to a large blockage of 0.9. circular cylinder over Re = 4 − 100. The inset illustrates the variation of viscous
and pressure drag coefficients with Re.
5.6. The large drag in confined flow

14710.20
Cdp = 4.33 + . (5)
The drastic alteration of surface pressure manifests in enor- Re
mous increase in the magnitude of drag coefficient. The value of These equations are valid for B = 0.9 and Re = 4 − 100. As
drag coefficient at Re = 4 approximately reaches a value as high evident, the drag and its pressure component decay as Re−1 . For
as 4000 (see Table 4). At Re = 100, the value of drag coefficient steady unbounded flow around a circular cylinder, Cd varies as
1
lowers down yet to a considerably high value of 167.96. This Re− 2 [10,35]. Therefore, the rate of decay of drag with Re is much
magnitude of drag is surprisingly large in view of the corre- higher for flows under large blockage.
sponding drag data for unbounded flow at the same Re. For The loss of pressure due to presence of the cylinder (implying
steady separated flow past a circular cylinder at Re = 100 in an that Cp0 > Cpb ) is represented by the difference of pressure
unbounded medium, Dennis and Chang [33] obtained Cd = 1.056 between the stagnation points, i.e., the quantity (Cp0 − Cpb ). A
while Fornberg [34] obtained Cd = 1.058. Fig. 17 plots the Cd − Re positive value of this difference signifies drag force on the body
variation while the inset shows the dependence of pressure drag, imposed by the flow while a negative value implies negative drag
C
Cdp and viscous drag on Re. At Re = 4, the Cdp ratio attains the or thrust on the cylinder. The higher the (Cp0 − Cpb ) difference,
dv
smallest value of 14.73 and this ratio increases monotonically lower is the recovery of pressure at the base and higher is the
C
with increasing Re. The maximum value of Cdp ratio is 16.04, pressure drag and hence, total drag. The recovery of pressure is
dv
attained at Re = 100. Empirical relations obtained by fitting the closely linked to the magnitude of Reynolds number. The pres-
computed Cd and Cdp data are expressed as: sure drag decays with Re since the (Cp0 − Cpb ) difference drops
15711.30 monotonically with increasing Re (Fig. 18a). The pressure drag
Cd = 4.23 + , (4) is precisely in a linear relationship with the quantity (Cp0 − Cpb )
Re
Fig. 16. Distribution of normalized surface vorticity (first row) and normalized shear stress (second row) along the cylinder surface for highly confined (B = 0.9)
steady flow past a circular cylinder. The upper row illustrates the normalized surface vorticity for Re = (a) 4, (b) 40 and (c) 100 while the lower row illustrates the
normalized shear stress for Re = (d) 4, (e) 40 and (f) 100. The normalized wall shear stress, τw ∗ is defined as τw ∗ = ρτUw2 .
146
Fig. 18. Highly confined (B = 0.9) steady flow past a circular cylinder over Re = 4 − 100: (a) illustration of the decaying nature of Cdp and (Cp0 − Cpb ) with increasing
Re and (b) linear relationship between Cdp and (Cp0 − Cpb ).
Table 5
Steady unbounded as well as highly confined flow past a circular cylinder at Re = 40 and 100: comparison of characteristic flow
quantities obtained from the present study and those reported by Dennis and Chang [33] and Fornberg [34].
Study L/D θs Cd
Dennis and Chang [33], unbounded, Re = 40 2.35 53.8◦ 1.52
Fornberg [34], unbounded, Re = 40 2.24 54.97◦ (from graph) 1.50
Present, B = 0.9, Re = 40 0.13 22.40◦ 396.83
Dennis and Chang [33], unbounded, Re = 100 6.56 66.2◦ 1.06
Fornberg [34], unbounded, Re = 100 6.35 68.02◦ (from graph) 1.06
Present, B = 0.9, Re = 100 0.96 53.93◦ 167.96
(Fig. 18b). The line of best fit is obtained as Re = 40 and 100, the characteristic flow quantities for unbounded
and highly confined flows. The last column of Table 4 suggests
Cdp = −0.52 + 0.97(Cp0 − Cpb )
(6) that the minimum pressure for B = 0.9 is attained well beyond
for B = 0.9 and Re = 4 − 100. cylinder shoulder. Thus, inception of adverse pressure gradient
Since the slope of the Cdp line (= 44.12◦ ) is very close to 45◦ , occurs behind the shoulder. For unbounded flow, Fig. 12d of the
Eq. (6) can be further simplified to Cdp ≈ Cp0 − Cpb . Thus, one can present paper and Fig. 5 of Dennis and Chang [33] suggests that
have a reasonably close estimate of pressure drag (within 4% of the minimum pressure at Re = 40 is attained just ahead of the
computed data) based on surface pressure data at the stagnation shoulder. Fig. 5 of Dennis and Chang [33] further indicates that as
points. the value of Re exceeds 40, the location of the minimum pressure
For a diffusion dominated flow at low Re, the fluid travels advances upstream of the shoulder. Thus, the separation angle for
with low values of approach velocity towards the cylinder and unbounded flow at the same Re always exceeds the separation
also around the close vicinity of the cylinder. However, for ap- angle for confined flow. In general, earlier separation is associated
proach flow, the fluid travels towards the stagnation zone and with wider wake and larger pressure drag. Despite a delayed
flow gets decelerated. In contrast, near the shoulders, flow is separation and smaller θs , the pressure drag for B = 0.9 is too
through a converging passage and hence, accelerated (Fig. 9). In large owing to the large range of surface pressure.
the shoulder region, the fluid flows like a jet. The fluid jets from
the top and bottom shoulders interact near the base and transfer 5.8. Surface vorticity and pressure along the channel walls
momentum to the flow. Owing to this, unlike the leading edge,
flow in the base region is not decelerated. According to Bernoulli’s The distribution of surface vorticity along the bottom wall
principle (even though Bernoulli’s principle is not strictly appli- of the channel is plotted in the upper row of Fig. 19. Results
cable for viscous fluid flow, Jordan and Fromm [36] noted that have been presented for Re = 4 (Fig. 19a) belonging to the
high pressure zones for unsteady flow past a circular cylinder are attached flow regime, 40 (Fig. 19b) and 100 (Fig. 19c) in the
associated with regions of low velocity), this disparity of velocity regime of separated flow. Except for the region of wall close to the
field causes large pressure at and near the forward stagnation cylinder shoulder, the wall vorticity for each Re, is fairly constant.
point, low pressure near the shoulders and minimum pressure The vorticity on the lower wall is negative. The sign of vorticity
at the base. Thus, at Re within the regime of fore-aft symmetry, is consistent noting that the vorticity on the lower half of the
i.e. Re ≤ 5, the (Cp0 −Cpb ) difference is high yielding large pressure cylinder surface is positive. A vorticity spike forms on the wall
drag. As Re is increased beyond this regime, the convection mode at the x = 0 location. The value of the maximum vorticity of this
starts to dominate over diffusion and velocity of flow increases. spike is the same as the maximum vorticity on the lower surface
This causes Cpb to reduce with increasing Re. of the cylinder (Fig. 16a). The high vorticity signifies the presence
of large velocity gradient as well as shear stress on the cylinder
5.7. Summary of the key characteristic flow quantities shoulder and x = 0 region of the wall.
The pressure along the bottom wall is shown for Re = 4
The values of the key characteristic flow quantities for Re = (Fig. 19d), 40 (Fig. 19e) and 100 (Fig. 19f). Overall, the pressure
4 − 100 are summarized in Table 4. The Cp data clearly establishes generally decreases along the wall. Based on the magnitude of
that the steady laminar flow around a bluff obstacle subject to pressure, a pair of pressure zones can be identified along the
extremely high blockage involves a wide range of Cp where the wall, i.e. the regime of high pressure ahead of the cylinder and
values of surface pressure decrease by several orders of mag- a low pressure zone downstream of the cylinder. A sharp drop
nitude as compared to those for unbounded flow. The drag, in of pressure signifying highly accelerated flow at x = 0 bridges
contrast, undergoes significant magnification. This order of mag- these pressure levels. The location of pressure drop coincides
nitude rise of Cd is further evident from Table 5 that compares at with that of vorticity spike and hence, maximum shear stress.
147
Fig. 19. Highly confined (B = 0.9) steady flow past a circular cylinder: the distribution of surface vorticity along the lower wall of the channel at Re = (a) 4, (b) 40
and (c) 100. Figs. d, e and f, respectively depict the pressure along the lower wall of the channel at Re = 4, 40 and 100.
At Re = 4, the wall pressure continues to decay with x. Thus, flow and unbounded flow, both at low Re. Using an extremely low
the pressure gradient is essentially favourable along the wall. The blockage of 0.00004 and Re ≤ 0.1, Pantokratoras [37] listed the
drop of cylinder surface pressure from its maximum or Cp0 to values of total drag for circular, square cylinders and a flat plate
minimum or Cpb at Re = 4 occurs monotonically with θ (Fig. 13a). oriented normal to the flow. Flow of both Newtonian and non-
This trend ascertains that the surface pressure on the cylinder Newtonian fluids was considered. For flow of a Newtonian fluid
is absolutely favourable. The favourable pressure gradient on past a circular cylinder, his computed Cd values for Re = 0.001,
the wall aids the favourable pressure gradient on the cylinder 0.01 and 0.1 are 2846, 373.67 and 55.711, respectively. Using a
surface and the flow remains strongly attached to the cylinder. very low blockage of 0.00083 for flow around a circular cylinder,
At Re = 6, the location of the minimum pressure on the cylinder Sivakumar et al. [38] obtained Cd = 417.3587 for Re = 0.01
surface shifts from the base point towards upstream (see Table 4). and Cd = 56.4324 for Re = 0.1. The value of Cd obtained by
Therefore, the pressure gradient on the cylinder surface turns Pantokratoras [37] for unbounded flow at Re = 0.001 is close
locally adverse for the first time at Re = 6. In contrast, on to the Cd value for highly confined flow in the Re range of 5 − 6
the bottom wall, the adverse pressure gradient is noted first at (apparent from Table 4). Similarly, the Cd corresponding to Re =
Re = 15. Therefore, for 6 ≤ Re < 15, the pressure gradient at the 0.01 for unbounded flow appears to be attained for confined
bottom wall of the channel aids the flow whereas on the bottom flow at an Re close to 40. While Pantokratoras [37] presented
half surface of the cylinder, it opposes the flow. This competition results for no other flow variable than drag, Sivakumar et al.
of pressure gradients continues unless the pressure gradient on [38] presented detailed results including the surface pressure
the bottom wall also turns adverse close to the Dx = 0 location. distribution. The surface pressure presented by Sivakumar et al.
At Re = 20, the pressure gradient at the bottom wall becomes [38] (first row of Fig. 11 of their paper) for low blockage differs
locally adverse and hence, opposes the flow. The combined effect from the Cp − θ profile attained for B = 0.9 (Fig. 12e of the
of adverse pressure gradient drives the separation of boundary present paper). At Re = 0.01 and B = 0.00083 [38], the base point
layer to commence at Re = 27.8. As seen from the inset of represents the minimum pressure point whereas for B = 0.9
Fig. 19e, the wall pressure for Re = 40 first drops near Dx = 0.3 flow at Re = 40 (present work), the minimum pressure point is
and then rises for a short interval of Dx . Subsequent to this, the located well upstream of the base point. In addition, the fore-aft
pressure registers a steady drop with x. Therefore, the pressure symmetry of streamlines for Re = 0.01 and B = 0.00083 (see Fig.
gradient is favourable for Dx < 0.3 and turns adverse for 0.3 < 5b of [38] for power-law index, n = 1) signifies creeping flow.
D
x
< 1.2. For Dx > 1.2, the pressure gradient turns favourable In contrast, the Re = 40 flow for B = 0.9 is separated. Thus, the
again. The implication of the adverse pressure gradient on the unbounded flow at very low Re and highly confined flow at low
wall is that it aids the adverse pressure gradient on the cylinder Re are not analogous.
surface and hence advances the location of separation on the
cylinder or increases the magnitude of θs . The adverse pressure 6. Conclusions
gradient on the bottom wall becomes even stronger as Re is
increased to 100 (Fig. 19f). The effective adverse pressure gradient A stabilized finite-element formulation permitting the use of
on the cylinder therefore receives further enhancement leading to the same order of interpolation for velocity as well as pressure
a larger separation angle. is employed to investigate the features of steady, laminar flow
around a circular cylinder over Re = 4 − 100. The cylinder is
5.9. Correspondence between highly confined flow at very low Re placed symmetrically in an extremely narrow channel leading
and unconfined flow at low Re to a blockage of 0.9. The direct consequence of imposing such
a large blockage is the drastic fall of cylinder surface pressure
In view of the large drag for B = 0.9, one might anticipate that and much delayed inception of adverse pressure gradient along
there exists a close correspondence between the highly confined the cylinder surface. For unbounded flow, presence of a local
148
minimum in surface pressure plot (Fig. 6 of [39]) quite ahead Declaration of competing interest
of the base point at Re = 1.6 indicates the advent of adverse
pressure gradient at a much lower Re. At B = 0.9, the adverse The authors declare that they have no known competing finan-
pressure gradient develops first at a fairly higher value of Re = 6 cial interests or personal relationships that could have appeared
(compared to unbounded flow) indicating delayed onset of steady to influence the work reported in this paper.
separation of laminar boundary layer from the cylinder surface.
Overall, the role of the sidewalls is thus, to stabilize the flow or Acknowledgement
delay the essential sequence of transition. The fore-aft symmetry
of streamlines about the cylinder prevails up to Re about 5, Partial support for this research came from the FRS grant
(FRS(124)/2018-19/MECH. ENGG.) provided by Indian Institute of
signifying the dominance of diffusion of vorticity than convection
Technology (Indian School of Mines) Dhanbad, Dhanbad - 826004,
of vorticity. The separation of boundary layer on cylinder surface
Jharkhand, India.
commences at Re = 27.8 as opposed to about 6.19 for unbounded
flow. The elongation of the standing wake of the cylinder is no
References
longer purely linear over the entire range of Re and the bubble
length at the same Re is much shorter relative to its unbounded [1] M.V. Morkovin, Flow around a circular cylinder–a kaleidoscope of chal-
counterpart. The DL − θs curves for B = 0.005 and 0.9 converge up lenging fluid phenomena. Symposium on Fully Separated Flows (ASME)
to the location of B = 0.9, Re = 50 signalling that the bubble New York, 1964, pp. 102–118.
[2] J.H. Lienhard, Synopsis of lift, drag and vortex frequency data for rigid
elongates linearly with Re for Re ≤ 50 and subsequently, the circular cylinders, in: College of Engineering Research Division Bulletin,
elongation is non-linear. Imposition of a large blockage does not vol. 300, Technical Extension Service, Washington State University, 1966.
introduce top–bottom asymmetry of flow in the steady regime. [3] C.H.K. Williamson, Vortex dynamics in the cylinder wake, Annu. Rev. Fluid
The effect of blockage on the separation angle is less appreciable. Mech. 28 (1996) 477–539.
[4] M.M. Zdravkovich, Flow Around Circular Cylinders, Volume 1, Oxford
The shortening of the wake is due to thickening of sidewall University Press, Oxford, 1997.
shear layers whereas relatively high value of separation angle is [5] M. Sahin, R.G. Owens, A numerical investigation of wall effects up to high
maintained via interactions between the cylinder shear layers and blockage ratios on two-dimensional flow past a confined circular cylinder,
Phys. Fluids 16 (5) (2004) 1305–1320.
penetrated vorticity of opposite-sign. The presence/absence of a
[6] G.S. West, C.J. Apelt, The effects of tunnel blockage and aspect ratio on
vorticity saddle point close to the base point appears to indicate the mean flow past a circular cylinder with Reynolds number between
a separated/attached flow. The sidewall shear layers hinder the 104 and 105 , J. Fluid Mech. 114 (1982) 361–377.
natural development of the wake and the bubble width in the [7] A.S. Grove, F.H. Shair, E.E. Petersen, A. Acrivos, An experimental investiga-
tion of the steady separated flow past a circular cylinder, J. Fluid Mech.
afterbody region never exceeds the width at separation points. In 19 (1) (1964) 60–80.
particular, WD
= sin θs . [8] M. Behr, D. Hastreiter, S. Mittal, T.E. Tezduyar, Incompressible flow past a
For the entire range of Re, the base pressure is strictly negative. circular cylinder: Dependence of the computed flow field on the location
For Re ≤ 5, the base pressure also represents the minimum of the lateral boundaries, Comput. Methods Appl. Mech. Engrg. 123 (1995)
309–316.
surface pressure signifying that the flow is essentially under [9] S. Mittal, S.P. Singh, B. Kumar, R. Kumar, Flow past bluff bodies: Effect of
favourable pressure gradient. Similar to the steady unbounded blockage, Int. J. Comput. Fluid Dyn. 20 (3–4) (2006) 163–173.
flow, the base pressure for highly confined flow increases with Re. [10] S. Sen, S. Mittal, G. Biswas, Steady separated flow past a circular cylinder
at low Reynolds numbers, J. Fluid Mech. 620 (2009) 89–119.
The values of Cp0 and Cpb being unity and very small, respectively
[11] O. Posdziech, R. Grundmann, A systematic approach to the numerical
and with Cp0 ≫ Cpb , the drag is expectedly large. The decay of calculation of fundamental quantities of the two-dimensional flow over
drag with increasing Re is an outcome of pressure recovery at a circular cylinder, J. Fluids Struct. 23 (2007) 479–499.
the base, i.e. the base pressure continues to become less negative [12] F.H. Shair, A.S. Grove, E.E. Petersen, A. Acrivos, The effect of confining walls
on the stability of the steady wake behind a circular cylinder, J. Fluid Mech.
with increasing Re. The (Cp0 − Cpb ) difference yields quite a close
17 (4) (1963) 546–550.
estimate of the pressure drag. As expected, the variation of nor- [13] M. Coutanceau, R. Bouard, Experimental determination of the main fea-
malized vorticity and normalized shear stress is identical along tures of the viscous flow in the wake of a circular cylinder in uniform
the cylinder surface. Depending on the value of Re, the sidewall translation. Part 1. Steady flow, J. Fluid Mech. 79 (2) (1977) 231–256.
[14] P.K. Stansby, A. Slaouti, Simulation of vortex shedding including blockage
surface pressure aids or opposes the cylinder surface pressure. by random-vortex and other methods, Internat. J. Numer. Methods Fluids
At low Re, such as 4, the pressure gradient along the sidewalls 17 (1993) 1003–1013.
and cylinder surface is both favourable. The favourable pressure [15] J.H. Chen, W.G. Pritchard, S.J. Tavener, Bifurcation for flow past a cylinder
gradient on the channel walls bolsters the favourable pressure between parallel planes, J. Fluid Mech. 284 (1995) 23–41.
[16] P. Anagnostopoulos, G. Iliadis, S. Richardson, Numerical study of the
gradient of the cylinder and flow remains attached. Over 6 ≤ blockage effects on viscous flow past a circular cylinder, Internat. J. Numer.
Re < 15, a competition continues between the adverse pressure Methods Fluids 22 (1996) 1061–1074.
gradient along the cylinder opposing the flow and favourable [17] J.H. Chen, Laminar separation of flow past a circular cylinder between two
parallel plates, Proc. Natl. Sci. Counc. Rep. China A 24 (2000) 341–351.
pressure gradient along the channel walls aiding the flow. The
[18] J. Chakraborty, N. Verma, R.P. Chhabra, Wall effects in flow past a circular
pressure gradient along the sidewalls turn locally adverse at cylinder in a plane channel: A numerical study, Chem. Eng. Process. 43
Re = 20 and thus, aids the adverse pressure gradient along the (2004) 1529–1537.
cylinder. This enhancement of adverse pressure gradient leads to [19] B. Kumar, S. Mittal, Effect of blockage on critical parameters for flow past
a circular cylinder, Internat. J. Numer. Methods Fluids 50 (2006) 987–1001.
the eventual separation of the laminar boundary layer from the
[20] S. Singha, K.P. Sinhamahapatra, Flow past a circular cylinder between
cylinder surface at Re = 27.8. Two-term empirical relations for parallel walls at low Reynolds number, Ocean Eng. 37 (8–9) (2010)
predicting the variation of Cd and Cdp with Re are proposed. It is 757–769.
found that Cd and Cdp vary as Re−1 . The rate of decay of Cd for [21] M.K. Rao, A.K. Sahu, R.P. Chhabra, Effect of confinement on power-law fluid
flow past a circular cylinder, Polym. Eng. Sci. 51 (10) (2011) 2044–2065.
highly confined flow is therefore much higher than unbounded [22] T.E. Tezduyar, S. Mittal, S.E. Ray, R. Shih, Incompressible flow com-
1
flow where Cd varies as Re− 2 . Though the values of Cd for B = 0.9 putations with stabilized bilinear and linear equal-order-interpolation
at low Re are of comparable magnitude with those of very low Re velocity-pressure elements, Comput. Methods Appl. Mech. Engrg. 95 (1992)
221–242.
flow for unconfined flow, it is emphasized that the low Re flow
[23] C.H.K. Williamson, Oblique and parallel modes of vortex-shedding in the
under large blockage is strictly distinct from the unbounded flow wake of a circular cylinder at low Reynolds numbers, J. Fluid Mech. 206
at very low Re. (1989) 579–627.
149
[24] G.K. Batchelor, An Introduction to Fluid Dynamics, Cambridge University [32] B. Fornberg, Steady viscous flow past a circular cylinder up to Reynolds
Press, Cambridge, United Kingdom, 1967. number 600, J. Comput. Phys. 61 (2) (1985) 279–320.
[25] M. Brons, B. Jakobsen, K. Niss, A.V. Bisgaard, Streamline topology in the [33] S.C.R. Dennis, G.Z. Chang, Numerical solutions for steady flow past a
near wake of a circular cylinder at moderate Reynolds numbers, J. Fluid circular cylinder at Reynolds numbers up to 100, J. Fluid Mech. 42 (1970)
Mech. 584 (2007) 23–43. 471–489.
[26] M.J. Lighthill, in: L. Rosenhead (Ed.), Laminar Boundary Layers, Dover [34] B. Fornberg, A numerical study of steady viscous flow past a circular
Publications, Inc., 1963, pp. 46–113. cylinder, J. Fluid Mech. 98 (4) (1980) 819–855.
[27] A.K. Dhiman, S. Kumar, Non-Newtonian power-law flow across a confined [35] F.T. Smith, Comparisons and comments concerning recent calculations for
triangular bluff body in a channel, Korean J. Chem. Eng. 30 (1) (2013) flow past a circular cylinder, J. Fluid Mech. 113 (1981) 407–410.
33–44. [36] S.K. Jordan, J.E. Fromm, Laminar flow past a circle in a shear flow, Phys.
[28] R.D. Henderson, Details of the drag curve near the onset of vortex Fluids 15 (6) (1972) 972–976.
shedding, Phys. Fluids 7 (1995) 2102–2104. [37] A. Pantokratoras, A note on the drag coefficient of steady flow of non-
[29] M. Braza, P. Chassaing, Ha Minh, Numerical study and physical analysis of Newtonian, power-law fluids across unbounded two-dimensional bodies
the pressure and velocity fields in the near wake of a circular cylinder, J. at low Reynolds numbers, Fluids 2 (1) (2017) 5.
Fluid Mech. 165 (1986) 79–130. [38] P. Sivakumar, R.P. Bharti, R.P. Chhabra, Steady flow of power-law flu-
[30] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere ids across an unconfined elliptical cylinder, Chem. Eng. Sci. 62 (2007)
Publication, 1980. 1682–1702.
[31] S.C.R. Dennis, M. Shimshoni, The Steady Flow of a Viscous Fluid Past a [39] R.L. Underwood, Calculation of incompressible flow past a circular cylinder
Circular Cylinder, Current Papers. Paper No. 797, Aeronautical Research at moderate Reynolds numbers, J. Fluid Mech. 37 (1) (1969) 95–114.
Council, 1965.
150

1 s2.0 S0997754621000212 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0997754621000212 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0997754621000212 Main

Uploaded by

Copyright:

Available Formats

European Journal of Mechanics / B Fluids 87 (2021) 135–150

Contents lists available at ScienceDirect

European Journal of Mechanics / B Fluids

Steady flow past a circular cylinder under large blockage

1. Introduction numerical treatment also differentiates the unbounded flow from

domain of size 95D × 10 9

is prescribed as the essential boundary condition at the inlet.

Table 3 For Re = 100, additional computations of the unsteady Navier–

For flow past a circular cylinder at low blockage, the onset of

transport is important even at high Re. The transport of vorticity

5.2. The flow

To illustrate the development of flow with Re, streamline

increases non-linearly with Re for B = 0.125, 0.25 and 0.5. The

dominant as the Re is increased (see Fig. 10d for Re = 60, 10e

5.6. The large drag in confined flow

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.