You are on page 1of 23

See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/259128217

Steady flow of Bingham plastic fluids past an


elliptical cylinder
ARTICLE in JOURNAL OF NON-NEWTONIAN FLUID MECHANICS DECEMBER 2013
Impact Factor: 1.94 DOI: 10.1016/j.jnnfm.2013.09.006

CITATIONS

DOWNLOADS

VIEWS

11

28

81

2 AUTHORS:
Swati Patel

Raj Chhabra

Indian Institute of Technology Kanpur

Indian Institute of Technology Kanpur

6 PUBLICATIONS 19 CITATIONS

209 PUBLICATIONS 3,152 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Swati Patel


Retrieved on: 15 August 2015

Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Contents lists available at ScienceDirect

Journal of Non-Newtonian Fluid Mechanics


journal homepage: http://www.elsevier.com/locate/jnnfm

Steady ow of Bingham plastic uids past an elliptical cylinder


S.A. Patel, R.P. Chhabra
Department of Chemical Engineering, Indian Institute of Technology, Kanpur 208016, India

a r t i c l e

i n f o

Article history:
Received 24 July 2013
Received in revised form 16 September 2013
Accepted 19 September 2013
Available online 26 September 2013
Keywords:
Elliptical cylinder
Bingham plastic uid
Reynolds number
Bingham number
Yielded/unyielded zones
Drag coefcient

a b s t r a c t
In the present work, the ow of Bingham plastic uids past an elliptical cylinder has been investigated
numerically elucidating the effect of yield stress and uid inertia on the momentum transfer characteristics at nite Reynolds numbers for a 100-fold variation in the aspect ratio. The governing differential
equations have been solved over wide ranges of Reynolds number (0.01 6 Re 6 40) and Bingham number
(0.01 6 Bn 6 100) in the laminar ow regime employing the nite element method. Furthermore, the
effect of the aspect ratio (E) of the elliptical cylinder on the detailed ow characteristics has been studied
by varying it from E = 0.1 to E = 10 thereby spanning varying levels of streamlining of the submerged
object. In particular, new extensive results on streamline contours, shape and size of yielded/unyielded
regions, shear rate proles, surface pressure distribution and drag coefcient as functions of the Reynolds
number, Bingham number and aspect ratio are presented and discussed. The functional dependence of
the individual and total drag coefcients on the governing dimensionless parameters, aspect ratio, Reynolds number and Bingham number, is explored. The present results reveal a signicant inuence of the
shape of the cylinder, i.e., aspect ratio on the detailed ow patterns and the overall hydrodynamic ow
behavior of elliptical cylinders.
2013 Elsevier B.V. All rights reserved.

1. Introduction
Owing to the wide occurrence of viscoplastic uid behavior in
suspensions, foams and multiphase systems encountered in scores
of industrial settings including food, pharmaceutical, personal-care
product sectors, polymer composites, geological applications, etc.,
there has been a renewed interest in studying their uid mechanical behavior in various congurations [13]. One of the main distinguishing aspects of viscoplastic uids is the fact that the ow
domain is spanned by the so-called yielded (uid-like) and unyielded (solid-like) regions depending upon the prevailing stress levels
vis-a-vis the value of the uid yield stress. From a theoretical/
numerical standpoint, not only this aspect itself poses enormous
challenges in resolving such regions but such dual nature of the
ow eld also has a deleterious effect on the degree of mixing
and convective transport of heat and mass, for diffusion is the chief
mode of heat and mass transfer operating in the unyielded regions.
Thus, the current interest in studying the behavior of such media in
complex geometries stems from both pragmatic and fundamental
considerations. Consequently, over the past fty years or so, significant advances have been made in the behavior of viscoplastic uids in internal ows [1,3], porous media ows [4], mixing vessels
[5], etc., though the uid mechanical aspects have been studied
much more thoroughly than the corresponding heat and mass
Corresponding author. Tel.: +91 512 2597393; fax: +91 512 2590104.
E-mail address: chhabra@iitk.ac.in (R.P. Chhabra).
0377-0257/$ - see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jnnfm.2013.09.006

transfer phenomena. In contrast, the currently available body of


knowledge on the so-called external or boundary layer type ows
is very limited indeed [6]. The bulk of the available literature relates to the prediction of drag and stability of spherical particles
settling in such uids in the creeping ow regime, e.g., see
[712] or on interactions between them [13]. Detailed discussion
and cross-comparisons between various numerical and/or experimental studies have been presented elsewhere [14,15]. Sufce it
to add here that based on a combination of the experimental and
numerical studies, reliable results are now available on the wall effects, drag coefcient and the size/shape of the yielded regions for
spherical particles undergoing steady translation in viscoplastic
uids in the creeping ow regime. These comparisons clearly reveal that the predictions and experiments for drag on a single
sphere are in reasonable agreement in the creeping ow regime.
Indeed, the effect of nite Reynolds numbers (up to 100) on drag
and heat transfer characteristics of a heated sphere in Bingham
plastic and Herschel Bulkley uids has been reported only very recently [14,15]. Broadly, while the uid yield stress acts to stabilize
the ow eld by postponing the ow detachment to higher values
of the Reynolds number than that in Newtonian uids, it obviously
increasingly restricts the size of the yielded uid-like regions close
to the surface of the sphere where the stress level exceeds the uid
yield stress. On the other hand, with the increasing Reynolds
number, the uid-like domains tend to expand spatially thereby
facilitating convective transport [14,15]. In contrast, much less
attention has been accorded to the other two-dimensional shapes

33

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Nomenclature
a
b
Bn
Bnc
CD
C D;1
CDF
CDP
Cp
C p
Cpo
D
D1
E
FD
FDF
FDP
lR
lw

semi-axis of the elliptical cylinder along the direction of


ow, m
semi-axis of the elliptical cylinder normal to the direction of ow, m 

Bingham number  slo 2b
V 1 , dimensionless
B

critical Bingham number denoting the disappearance of


ow separation, dimensionless
drag coefcient, dimensionless
limiting plastic drag coefcient, dimensionless
frictional drag coefcient, dimensionless
pressure drag coefcient, dimensionless
pressure coefcient, dimensionless
modied pressure coefcient, Eq. (16), dimensionless
pressure coefcient at the front stagnation point,
dimensionless
diameter of circular cylinder, m
diameter of the computational domain, m
aspect ratio of the elliptical cylinder, (=a/b), dimensionless
drag force per unit length of the cylinder, N m1
frictional component of drag force per unit length of the
cylinder, N m1
pressure component of drag force per unit length of the
cylinder, N m1
length of the static rigid zone (Zr2) from the center of
the cylinder, m
distance from the center of the cylinder to the point of
reattachment of the near closed streamline along the
x-axis, m
length of the cylinder in the z-direction, m

such as circular cylinders [1622] and square bars [23,24]. While


the currently available results for a circular cylinder are restricted
to the creeping ow only (zero Reynolds number), limited results
for a square cylinder are available at nite Reynolds numbers up
to Re = 40 [24]. Indeed, not only these studies reveal the existence
of different types of yielded/unyielded domains, but their shapes
and sizes are also modulated by the shape of the object as well
as by the values of the governing parameters. The simplest deviation from a circular cylinder is an elliptical shape which not only
allows the varying levels of streamlining simply by varying its aspect ratio but it is also free from geometric singularities such as a
square cylinder. Therefore, this work is concerned with the twodimensional ow of Bingham plastic uids past an elliptical cylinder oriented with its long axis transverse to the ow. At the outset,
it is instructive to briey recount the available results on the ow
of Newtonian uids past elliptical cylinders and the analogous results for viscoplastic uids which, in turn, facilitate the presentation and discussion of the new results obtained in this work.
1.1. Previous work
The ow past elliptical cylinders denotes a classical problem
in the realm of uid mechanics and transport phenomena and
has been studied widely over the past 100 years or so for Newtonian uids. Early attempts at studying the ow of Newtonian
uids past elliptical cylinders are invariably based on the use
of the Oseens linearized form of the NavierStokes equations
to obviate the so-called Stokes paradox. This approach is exemplied by the works of Tomotika and Aoi [25], Imai [26] and
Hasimoto [27]. Subsequent results [28,29] based on the numeri-

LR
Lw
m
n
ns
nx, ny
p
ps
p1
Re

Re
S
V
V1



a
,
length of the unyielded rigid static zone (Zr2)  lR2a
dimensionless


a
, dimensionless
recirculation length  lw2a
regularization parameter, dimensionless
power-law ow behavior index, dimensionless
unit vector normal to the surface of cylinder, dimensionless
x- and y-components of the unit vector normal to the
surface of cylinder, dimensionless
pressure, dimensionless
local pressure on the surface of cylinder, Pa
reference pressurefar awayfrom the cylinder, Pa
Reynolds number  qV 1l 2b , dimensionless
B

modied Reynolds number, Eq. (17), dimensionless


surface area of the cylinder, m2
velocity vector, dimensionless
free stream velocity, m s1

Greek symbols
c_
rate of strain tensor, dimensionless
lB
plastic viscosity, Pa s
ly
yielding viscosity, Pa s
q
density of the uid, kg m3
h
angular position on the surface of the cylinder measured
from the front stagnation point,
s
extra stress tensor, dimensionless
s0
yield stress, Pa
Subscripts
i, j, x, y Cartesian coordinates

cal solutions of the complete NavierStokes equations revealed


the results obtained in [2527] to be grossly inadequate for
Re > 2 for unconned ow conditions. Since the rst numerical
study of Epstein and Masliyah [28], numerous numerical studies
pertaining to the steady ow regime [29], elucidating the inuence of incidence [30], etc. have been reported in the literature
which are mutually consistent as far as the values of the drag,
recirculation length, etc. are concerned. Depending upon the values of the Reynolds number and aspect ratio, the ow past a cylinder exhibits a variety of ow regimes, akin to that seen for a
circular cylinder. Thus, for instance, Faruquee et al. [31] have
extensively studied the inuence of aspect ratio on the wake
characteristics at a xed Reynolds number of 40. At Re = 40,
the critical aspect ratio was reported to be 0.34 for the onset
of ow separation. Subsequently, Stack and Bravo [32] presented
the critical Reynolds number denoting the onset of ow separation for aspect ratios ranging from 0 (plate normal to ow) to 1
(circular cylinder) by solving the complete NavierStokes equations. As the value of E becomes increasingly larger than unity,
the degree of streamlining increases and the ow remains attached to the surface of the cylinder up to much larger values
of the Reynolds number than the oft reported value of Re = 56
for a circular cylinder. The effect of connement on the vortex
shedding characteristics of an elliptical cylinder has been investigated using the lattice Boltzmann method recently [33]. At the
other extreme, the high Reynolds number limit has also been approached by employing the standard integral boundary layer
analysis for the prediction of skin friction and Nusselt number
for an elliptical cylinder [34]. More detailed reviews of the pertinent studies are available elsewhere [3537].

34

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

In contrast, as far as known to us, within the framework of the


generalized Newtonian uids, there have been only three studies
dealing with the ow of power- law uids past elliptical cylinders
[3537]. Sivakumar and co-workers [35,36] reported extensive results on the momentum and forced convection heat transfer characteristics in the steady ow regime (Re 6 40) for shear-thinning
and shear- thickening uids. Both the drag and Nusselt number
values were found to be enhanced in shear-thinning uids and
these were suppressed in shear-thickening media with reference
to that in Newtonian uids otherwise under identical conditions.
However, these results are based on a priori assumption of the
steady ow regime to prevail over the range of conditions spanned
in their study. Indeed, the limits of the steady ow regime for elliptical cylinders of various aspect ratios have been delineated only
recently [37]. Based on these ndings, some of the results reported
by Sivakumar and co-workers [35,36] might be less reliable than
initially thought. Also, as expected, for blunt shapes (E < 1) ow
separation occurs at lower values of the Reynolds number than
that for a circular (E = 1) cylinder and the critical Reynolds number
increases with the increasing value of E. This nding is consistent
with that of Faruquee et al. [31].
Even less is known about the ow of viscoplastic uids past
elliptical cylinders. Putz and Frigaard [38] presented very limited
results for a two-dimensional planar ow over an elliptical cylinder using the standard TaylorHood nite element method in the
creeping ow regime. Similarly, in an attempt to mimic the behavior of an articial lung, Zierenberg et al. [39] have considered the
pulsatile ow of Casson model uid (blood) over a circular cylinder. While they have considered three values of the Reynolds number (5, 10 and 40), the yield stress values are extremely small
(corresponding to blood) and therefore very small deviations from
the corresponding Newtonian kinematics are predicted in their
study. From the aforementioned discussion, it is thus fair to conclude that there is only scant information available on the ow
of viscoplastic uids over an elliptical cylinder. For a given value
of the aspect ratio (E), it is expected that with the increasing Reynolds number, the uid-like yielded domains must grow in size,
but this tendency is countered to some extent by the uid yield
stress. Intuitively therefore, it appears that for a given Reynolds
number and aspect ratio, there must be a critical value of the Bingham number above which the ow remains attached due to the
equilibrium between the yield stress and viscous forces on one
hand and the inertial forces on the other. Conversely, for a given aspect ratio of the cylinder and Bingham number, it is expected that
the ow would remain attached to the surface of the cylinder up to
higher Reynolds numbers than that in Newtonian uids. This work
endeavors to ll this gap in the literature.
In particular, the main objective of the present work is to solve
the eld equations (continuity and momentum) numerically for
the ow of Bingham plastic uids past an elliptical cylinder elucidating the effect of uid yield stress and inertia on the uid
mechanical aspects in the range of conditions as: Reynolds number
0.01 6 Re 6 40, Bingham number 0.01 6 Bn 6 100 and aspect ratio
0.1 6 E 6 10. This work also reports the limiting values of the Bingham number above which the ow does not detach itself from the
surface of the elliptical cylinder. The present results are compared
with the previous studies wherever possible.

i.e., Vz = 0 and @
0. The unconned ow condition is reached
@z
here by enclosing the elliptical cylinder in a hypothetical concentric cylindrical envelope of uid of diameter D1 as shown schematically in Fig. 1b. The diameter of the outer circular boundary D1 is
taken to be sufciently large to minimize the boundary effects.
While no information exists about the ow regimes in Bingham
plastic uids for elliptical cylinders, by analogy with the transitions
observed in Newtonian uids [37,40], the ow is expected to be
steady and symmetric about the mid plane (y = 0) over the range
of conditions spanned here and therefore the computations have
been carried out only in half-domain (y P 0) to economize on
the computational effort.
For 2-D, incompressible and steady ow, the continuity and
momentum equations in their dimensionless forms are given by:
Continuity:

rV 0

Momentum:

V  rV rp

For a Bingham plastic uid, the deviatoric part of the stress tensor s is given by

c_ 0 if jsj 6 Bn



Bn

s 1 _ c_ if jsj > Bn
jcj

q
2
where jc_ j 12 trc_ is the magnitude of rate of deformation tensor
q
and jsj 12 trs2 is the magnitude of deviatoric stress tensor. In

these equations, the two dimensionless parameters are the familiar


Reynolds number (Re) and Bingham number (Bn) which are dened
a little later in Eqs. (9) and (10).
The rate-of-strain tensor c_ is given by

c_ rV rV T

There have been several approaches developed to obviate the


discontinuity inherent in the Bingham constitutive equation [41].
However, the two such approaches have gained wide acceptance,
namely that of Papanastasiou [42] and bi-viscosity [43] in the literature. While primarily the former is used in this work, limited results were also obtained with the latter to corroborate the
reliability of our results. Papanastasiou [42] modied the classical
Bingham model by introducing an exponential term for the stress
growth. The proposed BinghamPapanastasiou model which transforms the solid regions to a viscous one of high viscosity is given
by:

s 1


Bn1  expmjc_ j
c_
jc_ j

where m, the regularization parameter, controls the exponential


growth of the stress. Evidently, in the limit of m ? 1, this model
coincides with the Bingham model. Similarly, the bi-viscosity model
approach [43] postulates:

ly 
c_ for jsj 6 Bn
lB

2. Problem statement and formulation


The ow of an incompressible Bingham uid with uniform
velocity V1 over a long elliptical cylinder of aspect ratio E = a/b oriented transverse to the direction of ow is considered here, as
shown schematically in Fig. 1a. Since the cylinder is innitely long
in the z-direction, the ow is considered to be two-dimensional,

1
r:s
Re

s Bn c_ 

!
Bn
for jsj > Bn
ly =lB

The relative merits and demerits of different regularization


methods and cross-comparisons between their predictions based
on different regularization techniques for specic geometries like

35

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

E = 0.1
E=1
E = 10

x
a

Cylinder

(a)
y

Uniform velocity

Out flow

E = 0.1

No slip- wall

E=1
E = 10

symmetry

(b)
Fig. 1. Schematics of the ow past an elliptical cylinder: (a) physical model (b) computational domain.

the creeping ow over a sphere are available elsewhere [41,44].


Potential difculties in locating the yield surfaces through such
regularization methods have also been discussed in Ref. [44].
In order to complete the problem statement, the following
boundary conditions have been used in this work.
The front-half of the uid envelope (of diameter D1) is designated as the inlet and at this surface, a uniform ow in the x-direction is prescribed, i.e., Vx = 1 and Vy = 0.
The rear-half of the surrounding uid envelope is designated as
the outlet and here the disturbance to the ow eld caused by the
elliptical cylinder is assumed to have subsided and thus, zero-diffusion ux condition for the both velocity components, i.e.,
@V
@V x
0 and @xy 0 is used here on this plane.
@x
On the surface of the cylinder: The standard no-slip boundary
condition, i.e., Vx = Vy = 0 is used.
Over the range of conditions spanned here, the ow is expected
to be symmetric about y = 0 plane and therefore, the symmetry
x
conditions are implemented here, i.e., @V
0 and Vy = 0.
@y
The preceding governing equations and the boundary conditions have been rendered dimensionless by using V1 and 2b as
the characteristic velocity and length scales respectively. These,
 
in turn, can be used to obtain the corresponding scales as lB V2b1 ,

qV 21 and

2b
V1

for the stress components, pressure and regularization parameter respectively. Naturally, one could have chosen 2a
instead of 2b as the characteristic linear scale, but since the aspect
ratio E is dimensionless on its own, one can convert these results
from one format (based on the choice of 2b) to another (based on
the choice of 2a). Evidently, in this case, the momentum characteristics are governed by the following three dimensionless
parameters:
Bingham number: This represents the ratio of the yield stress
to viscous forces, i.e.,

Bn

so
 
lB V2b1

Reynolds number: This denotes the ratio of the inertial to viscous forces, i.e.,

Re

qV 21
 
lB V2b1

10

Of course, the aspect ratio, E = a/b, which describes the shape of


the cylinder cross-section, is the third dimensionless parameter.
The preceding denitions of the Reynolds and Bingham numbers

36

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

are based on the assumption that the characteristic shear rate is of


the order of (V1/2b) and the effective viscous stress is given simply
as lB(V1/2b) thereby disregarding the inuence of the uid yield
stress. However, the inclusion of the yield stress in estimating
the representative viscosity will only rescale the Reynolds number
by incorporating the effect of the Bingham number, as seen in Eq.
(17) here and elsewhere [14,24].
It is customary to present the detailed kinematics of the ow in
terms of the streamlines in the vicinity of the cylinder and the distribution of pressure coefcient along the surface of the cylinder.
The overall gross behavior is denoted in terms of the recirculation
length, individual and total drag coefcients. In the case of viscoplastic uids, the size and shape of the yielded zones also depend
on the values of three parameters, namely, Re, Bn and E. Some of
these characteristics are dened here.
Drag coefcient (CD): This is a measure of the net hydrodynamic force exerted by the uid on the immersed cylinder along
the direction of ow. The drag coefcient is made up of two components, namely, friction drag (CDF) due to the shearing forces and
form drag (CDP) due to the normal forces acting on the cylinder.
These are dened as follows and are essentially evaluated by the
surface integrals as shown below:

C D C DF C DP
C DF 1
2

F DF

V 21 2b

11

2
Re

sxx nx sxy ny dS

12

where nx and ny are the components of the unit normal vector, ns,
normal to the surface of the cylinder given as
2

ns

x=a2 ex y=b ey
r
nx ex ny ey
 x 2  y 2

2
2
a
b

13

The form drag is dened and evaluated as follows

C DP 1
2

F DP

V 21 2b

C p nx dS

14

In Eq. (14), Cp is the dimensionless pressure coefcient dened


as the ratio of the static to dynamic pressure on the surface of the
cylinder, i.e.,

Cp

ps  p1
1
2

qV 21

15

In Eq. (15), ps is the local pressure at a point which varies along


the surface of the cylinder and p1 is the reference pressure far
away from the cylinder.
Further insights into the nature of this ow can be gained by
rescaling the pressure coefcient (ratio of the static pressure to
yield stress) on the surface of the cylinder. The modied pressure
coefcient is dened as:

C p C p  Re

16

where

Re

Re
1 Bn

17

Recirculation (or wake) length (Lw): It is the dimensionless


distance measured from the rear of the cylinder to the point of
reattachment for the near closed streamline Vx = Vy = 0 on the line
of symmetry (y = 0).

Lw

lw  a
2a

18

where lw is the distance from the center of the cylinder to the point
of reattachment for the near closed streamline as shown schematically in Fig. 2a. In the context of Newtonian uids, this is a direct
measure of the wake length. However, in the present situation, as
will be seen in Sections 5.2 and 5.3, there is an unyielded zone attached in the rear of the cylinder which is engulfed in the recirculating region. Therefore, it is not uncommon to introduce another
characteristic parameter to describe the length of this static zone.
Length of the unyielded rigid zone (LR): It is the dimensionless
length of the static rigid zone Zr2 measured from the rear of the
cylinder.

LR

lR  a
2a

19

where lR is the length of static zone downstream of the cylinder


measured from the center of the cylinder (Fig 2b).
Finally, the scaling considerations suggest that the detailed
kinematics and macroscopic momentum characteristics in the
present case are inuenced to varying extents by three dimensionless groups or combinations thereof, namely, Reynolds number
(Re), Bingham number (Bn) and the aspect ratio of the cylinder
(E). This work endeavors to understand and develop this
relationship.
3. Numerical methodology
The preceding partial differential equations subject to the aforementioned boundary conditions have been solved here numerically using the nite element method based solver, COMSOL
Multiphysics (Version 4.2a). The computational domain was
meshed using a non-uniform grid structure created by the builtin meshing function employing quadrilateral and triangular elements. Owing to the expected steep gradients close to the surface
of the cylinder and near the interface demarcating the yielded and
unyielded regions, a ne mesh was used in these regions which
was progressively made coarse to economize on the required computational effort. The resulting system of equations is solved using
the steady, 2-D, laminar ow module with parallel direct linear solver (PARDISO). The deviatoric part of the stress tensor in the

Fig. 2. Schematic representation of (a) recirculating wake and (b) static zone characteristics.

37

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

momentum equation is approximated using the Papanastasiou


modied Bingham model. Based on our previous studies
[14,15,23,24], the relative convergence criterion of 105 for the
equations of continuity and momentum is used in this work. Within the framework of this criterion, the drag values had also stabilized at least up to four signicant digits. Besides, the results
obtained using more stringent convergence criterion were virtually
indistinguishable from the present results. Appropriate values of
the uid characteristics like density (q), yield-stress (so), plastic
viscosity (lB) and the geometric parameters were specied to
achieve the desired values of the three dimensionless parameters,
namely, Re, Bn and E. However, these specic values of the physical
properties of the uid are of no particular signicance because the
nal results are presented here in their nondimensional form.
Finally, the yield-surfaces denoting the boundaries between the
yielded and unyielded regions were located by manually rening
the computational mesh in this region and by comparing the magnitude of the dimensionless extra stress tensor with the uid yield
stress (Bingham number) within a tolerance of 106107 at each
point. Further reduction in the tolerance criterion did not produce
any noticeable changes in the shape or size of the unyielded
regions.
4. Choice of numerical parameters
Undoubtedly, the accuracy and reliability of the numerical results is strongly inuenced by the choice of numerical parameters,
namely, domain size, quality of grid, convergence criterion, the value of the regularization parameter, m and of the yielding viscosity,
ly. Much has been written about these aspects elsewhere
[14,23,24], and therefore only the salient points are recapitulated
here. Bearing in mind the fact that the velocity eld decays slowly,
at low Reynolds numbers and/or Bingham numbers, several values
of (D1/2b) were used in this study to choose its optimum value
without a signicant loss of accuracy and keeping the required

computational effort at a reasonable level (Table 1a). An inspection


of Table 1a suggests that the domain sizes of D1/(2b) = 300, 520,
900 and 800 are believed to be adequate for E = 0.1, 0.2, 0.5 and
1 respectively while for E = 2, 5 and 10, the domain size of D1/
(2b) = 1000, 2500 and 5000 respectively are seen to be adequate
for the ranges of conditions of interest here: 0.01 6 Re 6 40 and
0.01 6 Bn 6 100.
Having xed the domain size, a grid independence study has
been carried out using three non- uniform grids (G1,G2 and G3)
with the increasing level of renement for a range of values of
the aspect ratio, E at Re = 40 and Bn = 100. The grid used in the
present work is divided into two subregions. The rst zone in the
vicinity of cylinder where the mesh is highly concentrated consists
of triangular elements, as shown in Fig. 3 for extreme values of the
aspect ratio E = 0.1 and E = 10. Otherwise non-uniform quadrilateral elements were employed for the remaining intermediate values of the aspect ratio. The second sub-region away from the
cylinder employed non-uniform quadrilateral elements with manual rening near the yield surfaces. Fig. 3 shows the schematics of
the grid structures used for aspect ratio E = 0.1 and 10 with their
close-up view near the cylinder expanded for the three grids G1,
G2 and G3 tested in this work. These results for grid tests are summarized in Table 1b which clearly show that the values of the drag
coefcient change negligibly as one goes from grid G2 to G3. Furthermore, Fig. 4 shows the inuence of grids on the velocity prole
in the x- and y-directions for the extreme values of the aspect ratio
E = 0.1 and E = 10 at Re = 40 and Bn = 100. Detailed examination of
the results shown in Table 1b and Fig. 4 reveals that grid G2 is adequate to resolve the thin boundary layers under the extreme conditions considered herein. Indeed, the comparison shown in Fig. 4
constitutes a more stringent test of the grid effects than the results
shown in Table 1b. Thus, on both counts, G2 seems to be adequate
in the present study.
A reliable prediction of the unyielded zones also depends
strongly on an appropriate choice of the regularization parameter,

Table 1
Choice of computational parameters: (a) Domain effects. (b) Grid effects.
(a) Domain independence test
Re = 0.01
E

a
b

(b) Grid independence test


Re = 40

Domain size

Bn = 0.01

D1/(2b)

CD

Grid

Elementsb

CDP

Bn = 0.01

Bn = 100

CD

CDP

CD

CDP

0.1

260
300
350

833.54
834.01
834.07

737.43
737.47
737.40

0.1a

G1
G2
G3

26,878
35,722
40,322

1.6600
1.6596
1.6596

1.5735
1.5771
1.5773

61.333
61.278
61.267

58.879
59.508
59.567

0.2

360
520
700

836.30
836.61
836.70

686.88
688.49
690.14

0.2

G1
G2
G3

12,000
23,640
26,000

1.6355
1.6313
1.6303

1.4994
1.4768
1.4720

62.231
61.749
61.672

58.591
58.342
58.088

0.5

400
900
1300

848.31
848.16
848.34

622.01
573.45
578.92

0.5

G1
G2
G3

12,000
23,640
26,000

1.5631
1.5626
1.5624

1.2335
1.2367
1.2376

63.031
62.835
62.815

52.856
53.537
53.708

500
800
1800

874.29
875.50
874.64

441.23
447.36
446.33

G1
G2
G3

12,000
23,640
28,000

1.5078
1.5076
1.5076

0.98117
0.98593
0.98663

66.247
65.989
65.970

49.281
49.704
49.819

600
1000
2000

929.57
929.46
929.39

317.62
313.20
313.90

G1
G2
G3

12,000
23,640
28,000

1.4940
1.4945
1.4943

0.7237
0.7185
0.7186

74.920
74.286
74.260

47.333
46.000
46.006

1500
2500
5000

1076.7
1077.5
1077.8

183.72
186.00
187.68

G1
G2
G3

13,200
18,200
21,000

1.6515
1.6513
1.6513

0.4515
0.4489
0.4485

108.09
107.36
107.20

46.269
45.545
45.387

10

3000
5000
8000

1270.2
1270.4
1270.3

122.24
127.91
127.50

10a

G1
G2
G3

7527
16,332
20,609

1.95041
1.95506
1.95338

0.2964
0.2802
0.2808

164.40
161.53
161.51

43.862
43.700
43.332

Free triangular grid in the vicinity of cylinder.


Refer to half-domain (y P 0).

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

G1

G2

G1

G2

G3

E = 0.1

38

E = 10

G3

Fig. 3. Schematic representation of non-uniform computational grid structure with their expanded view near the cylinder.

Fig. 4. Inuence of grid size on the variation of velocity proles in x- and y-directions at Re = 40 and Bn = 100.

m. Naturally, its unduly small values do not correctly capture the


behavior of the uid whereas its excessively high values lead to a
very stiff coefcient matrix thereby leading to oscillations in the
predicted contours of the unyielded zones and convergence problems [44,45]. In this study, the values of m ranging from 104 to
107 have been examined for 0.1 6 E 6 10 at Bn = 10 and Bn = 100.
Fig. 5 shows the inuence of this parameter on the predictions of
the yielded/unyielded zones for extreme values of the aspect ratio,
i.e., E = 0.1 and E = 10. Evidently, the results change very little for
m P 107. Based on these observations, the results reported herein

are based on the value of m = 107. Similarly, in the case of the biviscosity model, one needs to examine the effect of the value of
the yielding viscosity (ly) on the accuracy of the solution. Table 2
summarizes the results showing the inuence of this parameter on
the values of the pressure and drag coefcients at the minimum
and maximum values of the Bingham numbers used in this work.
Evidently, the value of (ly/lB) = 105 is seen to be satisfactory over
the range of conditions spanned in the present work. Finally, the
adequacy of these choices is demonstrated in the next section by
presenting a few benchmark comparisons.

39

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Fig. 5. Inuence of the regularization parameter, m, on the location of unyielded zones at Re = 40 (a) E = 0.1 (b) E = 10.

Table 2
Inuence of the yielding viscosity ly, on the total drag (CD) and pressure drag (CDP) coefcients.
ly
lB

CD

CDP

Re = 0.01

Re = 40

Re = 0.01

Re = 40

Bn = 0.01

Bn = 100

Bn = 0.01

Bn = 100

Bn = 0.01

Bn = 100

Bn = 0.01

Bn = 100

E = 0.1
104
105
106

835.47
835.49
835.50

242,882
242,903
242,906

1.6633
1.6644
1.6652

61.525
61.531
61.532

732.17
732.16
732.16

234,144
234,164
234,167

1.5559
1.5570
1.5578

59.149
59.156
59.156

E = 10
104
105
106

1269.4
1269.4
1269.4

646,055
646,056
646,056

1.9555
1.9555
1.9555

161.53
161.53
161.53

98.131
98.131
98.131

174,693
174,693
174,693

0.2802
0.2802
0.2802

43.700
43.700
43.700

Table 3
Comparison of drag coefcients (CD) for elliptical cylinders (E = 0.2 and E = 5) in Newtonian uids.
Re

Dennis and Young [30]

DAlessio and Dennis [29]

Sivakumar et al. [35]

Present

E = 0.2
0.01
0.1
1
5
20
40

3.854
2.119
1.876

3.862
2.140

404.53
54.247
9.806
3.790
2.065
1.621

409.97
54.748
9.839
3.782
2.062
1.618

E=5
0.1
1
5
10
20
40

8.096
2.712
1.765
1.169
0.789

8.222

1.848
1.228
0.794

67.109
8.110
2.7361
1.768
1.168
0.786

66.586
8.014
2.665
1.730
1.147
0.774

5. Results and discussion


In this study, the governing differential equations of ow have
been solved numerically to examine the effects of aspect ratio

(0.1 6 E 6 10) on the uid ow around an elliptical cylinder over


wide ranges of dimensionless parameters as: 0.01 6 Re 6 40 and
0.01 6 Bn 6 100. In particular, the present results endeavor to elucidate the role of these parameters on streamline patterns, yielded/

40

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Table 4
Comparison of front stagnation point pressure coefcient Cpo(h = 0) and drag coefcient of elliptical cylinders in Newtonian uids.
E

Re

Cpo

CD

Masliyah and Epstein [48]

Present

Masliyah and Epstein [48]

Present

0.2

1
5
15
40

1.634
1.212
1.049

1.619
1.226
1.088

10.810
3.942
2.586
1.814

9.839
3.782
2.309
1.618

0.5
5
15
40

2.047
1.468
1.200

2.004
1.417
1.176

18.820
4.298
2.379

17.835
4.225
2.370

5
10
20
40

2.649
2.037
1.656
1.436

2.408
1.839
1.481
1.262

5.019
3.490
2.424
1.771

5.037
3.417
2.350
1.637

Table 5
Comparison of drag coefcient between the present results for E = 0.1 and that of a
vertical at plate in Newtonian uids.

Table 7
Comparison of the present and literature values of drag coefcient at nite Reynolds
numbers for a circular cylinder (E = 1).

Re

Dennis et al. [51]

Present

CD  Re

0.5
1
5
10
20
30
40

15.08
9.66
3.75
2.75
2.09
1.82
1.68

15.99
9.95
3.81
2.76
2.10
1.82
1.66

Re

0.0083
0.0833
0.8333
4.1667
8.3333
16.6667
33.3333

Re

Bn = 0.2
Mossaz et al. [20]

Present

25.628
25.723
26.678
30.921
36.225
46.834
68.050

24.720
24.717
25.291
30.879
37.429
48.732
67.826

0.005
0.05
0.5
2.5
5
10
20

Bn = 1
Mossaz et al. [20]

Present

59.279
59.281
59.478
62.920
68.522
78.912
97.052

59.239
59.317
60.103
63.593
67.955
76.680
94.130

ature values for a few limiting cases, as this will help ascertain the
level of the accuracy of the new results for Bingham plastic uids
for elliptical cylinders presented herein.

5.1. Validation of results

Fig. 6. Comparison of drag coefcient values for a vertical at plate and an elliptical
cylinder (E = 0.1).

Table 6
Validation of the present results (Bn = 105) for elliptical cylinders in the fully plastic
limit.
Ref.

C D;1
E = 0.5

E=1

E=2

E = 10

Randolph and Houlsby [53]


Mitsoulis [16]
Tokpavi et al. [19]
Putz and Frigaard [38]
Present

13.1
13.205

11.94
11.7
11.94
11.94
11.939

11.56
11.581

11.35
11.331

unyielded zones, ow kinematics and drag coefcient. At the outset, it is, however, important to validate the solution methodology
used in this study by comparing the present results with the liter-

Initially, a few results were obtained for the ow of Bingham


plastic uids in a lid-driven square cavity and the resulting values
of the centerline velocities at the horizontal and vertical positions
of the vortex center were found to be within 2% of the literature
values for Bingham plastic uids [9,46] and the results of centerline velocities for Newtonian uids show deviations around 3%
with those reported by Neofytou [47]. Next, as reliable results
are now available for the ow of Newtonian uids (Bn = 0) past
elliptical cylinders over the range of conditions spanned here, Table 3 shows a comparison between the present and literature values culled from a few sources employing different numerical
solution schemes, domains, etc. With the exception of one data
point of Dennis and Young [30], the present values are within 3
4% of the previous results [29,30,35]. Furthermore, Table 4 compares the present values of the pressure coefcient at the front
stagnation point and drag coefcient for the Newtonian uids
[48]. Barring the results for E = 5, the other values are seen to be
within 4% of each other. While no prior results are available for
an elliptical cylinder with E = 0.1, these are expected to be very
close to that for a plane surface oriented normal to the oncoming
uid stream. Table 5 and Fig. 6 show comparisons between the
present results for E = 0.1 and that for a plate with the literature results culled from several sources [4952]. The close correspondence seen in Table 5 and in Fig. 6 is particularly instructive and
lends credibility to the present solution methodology. Finally,
Table 6 shows a comparison between the present and literature
values for elliptical and circular cylinders in terms of the limiting

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

41

Fig. 7. Representative streamline proles for an elliptical cylinder (a) E = 0.1 (b) E = 0.2 (c) E = 1 (d) E = 5 (e) E = 10.

drag values (Bn ? 1) while Table 7 compares the values of the


drag coefcient at nite Reynolds numbers for a circular cylinder

in a Bingham plastic uid. Once again, an excellent match is seen


to exist in these tables. Similar extensive comparisons for the drag

42

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Table 8
Effect of Reynolds number and Bingham number on the recirculation length.
Bn

Re

Lw
E = 0.1

E = 0.2

E = 0.5

E=1

E=2

E=5

E = 10

0.01

1
5
10
20
40

0.8915
5.2663
10.127
21.151
47.325

2.0811
4.3815
9.4296
21.201

0.3125
1.1269
2.7842
6.3974

0.2106
0.8683
2.1867

0.1271
0.5706

0.1

1
5
10
20
40

0.2940
4.2747
8.477
17.786
39.876

1.5548
3.5784
7.8883
17.877

0.1121
0.8010
2.2418
5.3753

0.0619
0.6057
1.7784

0.3886

5
10
20
40

1.4640
3.6493
8.1874
18.254

1.1840
3.1989
7.8966

0.0552
0.5397
1.9409

0.2825

10
20
40

0.5789
2.0011
5.1902

0.4853
1.7259

0.1212

10

20
40

0.5175
2.2307

Table 9
Values of critical Bingham number for elliptical cylinders.
Re

1
5
10
20
30
40

E = 0.1

E = 0.2

E = 0.5

E=2

Bnc (no wake)

Bn (wake)

Bnc (no wake)

Bn (wake)

Bnc (no wake)

Bn (wake)

Bnc (no wake)

Bn (wake)

Bnc (no wake)

0.5
4.5
7.75
13.5
18.75
24

0.75
4.75
8
13.75
19
24.24

1.75
4.25
8.25
11.75
15.5

2
4.5
8.5
12
15.75

0.2
1.25
3
4.75
6.50

0.3
1.5
3.25
5
6.75

0.2
0.8
1.4
2

0.25
0.85
1.45
2.25

0.075
0.30
0.45

0.08
0.35
0.50

Table 10
Comparison of recirculation length Lw for Bingham plastic uid ow past circular
cylinder.
Lw
Mossaz et al. [20]
Bn

Re = 20

0.08
0.19

0.6310
0.3793

Present

%error

0.6567
0.4016

4.08
5.88

1.7954
1.4857
1.1931
0.9809
0.6195

3.34
1.47
3.95
2.54
3.60

Re = 40
0.08
0.18
0.28
0.39
0.59

E=1

Bn (wake)

1.8574
1.5079
1.2422
1.0064
0.5980

of a sphere in Bingham plastic and HerschelBulkley uids have


been recently presented elsewhere [14,15] and therefore, these
are not repeated here. Based on the foregoing extensive comparisons, the new results for elliptical cylinders reported herein are believed to be reliable to within 23%.

5.2. Streamlines contours and recirculation length


Representative streamline proles close to the surface of an
elliptical cylinder (E = 0.1, 1 and 10) are shown in Fig. 7 for a range
of values of the Reynolds number and Bingham number. At low
Reynolds numbers, the uid inertia is small and therefore a uid
element is able to negotiate the body contour without incurring

any loss of kinetic energy and thus the ow remains attached to


the surface of the cylinder. Similarly, the yield stress of the uid
also tends to delay the onset of ow detachment from the surface
of the cylinder. This is ascribed to the fact that away from the submerged cylinder, the material is by and large unyielded which acts
as a virtual wall and it is thus tantamount to that the ow occurs in
a conned geometry. This, in turn, tends to suppress the propensity
for ow separation, in line with the available results in Newtonian
uids. Thus, while the tendency for ow separation increases with
the increasing Reynolds number, it is suppressed with the increasing Bingham number for a given shape, i.e., value of E. Naturally
both these mechanisms are modulated by the shape of the object.
Thus, for instance, at E = 0.1 which behaves like a plane surface oriented normal to ow, due to sudden changes in the ow direction,
ow separation is likely to occur at low Reynolds numbers; the
critical value being Re = 0.08 for Newtonian uids [32]. Thus, for instance, at Re = 1, there is a visible separated region in the form of
twin counter rotating vortices at Bn = 0.01 which seems to disappear completely at Bn P 0.1. Intuitively, it appears that higher
the Reynolds number, larger would be the value of the Bingham
number needed to prevent the ow separation. This observation
is clearly borne out by the results shown in Fig. 7 irrespective of
the value of E 6 1. However, for E > 1, ow separation occurs at
much larger Reynolds numbers even in Newtonian and powerlaw uids [37] and with the introduction of yield stress, this trend
is likely to continue even up to higher Reynolds numbers, as can be
seen in Fig. 7 and in Table 8. These trends are qualitatively consistent with that reported for a circular cylinder [20] and a sphere
[14]. Table 8 summarizes the functional dependence of the recirculation length Lw on the Reynolds number and Bingham number for

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Bn = 100 Re = 40

Zr3

Zr2

Zr1

Zr3

43

the present values of the recirculation length, Lw, with that of


Mossaz et al. [20] in the limit of E = 1 and n = 1, and the two values
are seen to be in good agreement.
Finally, attention is drawn to the fact that in one case corresponding to Re = 10, Bn = 0.1 and E = 1, there is a second recirculation region, smaller than the primary wake present, while no wake
was observed at Bn = 0.2 and only one recirculating region was
seen at Bn = 0.080.09. Therefore, it is likely that this point is just
too close to the critical point corresponding to the suppression of
the wake formation at Bn = 0.2. It is likely that the primary recirculating region splits into smaller regions before disappearing altogether. On the other hand, the presence of the second
recirculation region is not a numerical artifact because this case
was repeated at least with two different meshes and for a few values of the Bingham number in the vicinity of Bn = 0.1. No more
explanation can be given at this stage for this effect.

Zr4
5.3. Delineation of yielded/unyielded zones
Fig. 8. Schematic representation of the rigid zones around a circular cylinder (E = 1)
(ow is from left to right).

a range of values of E. For a xed value of the Bingham number and


aspect ratio, the recirculation length shows a positive dependence
on the Reynolds number which is in line with the behavior seen in
Newtonian and power-law uids. On the other hand, for a xed
Reynolds number and aspect ratio, the recirculation length decreases with the increasing Bingham number. The decreasing wake
size and disappearance of the standing vortices is also expected
with the increasing value of the aspect ratio (E) due to the increasing degree of streamlining of the bluff body. Thus for instance, no
ow separation is observed for the range of Bingham and Reynolds
numbers considered in this study for E > 2 which is also in line with
the previous results [37]. Similarly, no ow separation was observed at Re 6 0.1 for the ranges of Bingham number, Bn and aspect
ratio, E embraced in this study. Table 9 summarizes the critical values of the Bingham number (within 0.13) as a function of the aspect ratio and Reynolds number above which the ow remains
attached to the surface of the submerged body. It is worthwhile
to add here the values of the critical Bingham number (for a xed
Reynolds number) reported here (denoting the cessation of the
ow separation) are complimentary to the critical values of the
Reynolds number, for a xed Bingham number, reported by Mossaz et al. [20] which denote the onset of the formation of the recirculating regions in the rear of the cylinder. Therefore, while it is not
possible to contrast these two results, however, Table 10 contrasts

One of the distinct features of the ow of viscoplastic media is


the simultaneous existence of the uid-like (yielded) and solid-like
(unyielded) regions, both in the vicinity of the submerged object
and far away from it, as have been reported for a sphere, circular
cylinder and a square bar. Similarly, in the present case, three distinct unyielded zones are observed, shown schematically in Fig. 8,
where the unyielded zones are shaded while the unshaded regions
represent the deforming uid zones; however, these differ in shape
and size depending upon the aspect ratio (especially Zr1 and Zr2)
of the cylinder from that seen for a circular cylinder (E = 1). These
are briey described below:
Two triangular shaped unyielded zones (Zr1 and Zr2) attached
to the front and rear of the cylinder at the stagnation points
which are static in nature. The triangular shape of these zones
observed in this study was also reported by Mossaz et al. [21].
Two symmetric rigid cores (Zr3), equidistant from the cylinder
on the either side about the horizontal axis of symmetry. These
are dynamic in nature, i.e., these are undergoing a rigid bodylike rotation without deformation of the uid.
A rigid envelope enclosing the uid zone, far away from the cylinder referred to here as Zr4. This is also dynamic in nature in so
far that it is moving as a solid plug with a uniform velocity V1,
without deforming.
The existence of the above-mentioned rigid zones has been also
conrmed by comparing the location of the yielded/unyielded

Fig. 9. Comparison of unyielded zones of (a) Tokpavi et al. [19] (creeping ow) with that of (b) present work (Re = 0.01) for Bingham plastic uid.

44

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

regions for a circular cylinder (E = 1) at low values of the Reynolds


number (Re = 0.01) in the present study with that of Tokpavi et al.
[19] for the creeping ow regime at Bn = 10 and Bn = 100 (Fig. 9).
Notwithstanding the inherently different values of the Reynolds
number in the two cases and the numerical solution methodologies, the two predictions are seen to be qualitatively similar.
Naturally, the size of each of these unyielded segments will
vary not only with the kinematic parameters (Re and Bn), but
also with the aspect ratio of the cylinder. For the extreme values
of E = 0.1 and E = 10 considered here, the shape of the cylinder
corresponds to a vertical (E ? 0) or to a horizontal (E ? 1) at
plate. The static zones Zr1 and Zr2 are observed to be the largest
corresponding to E = 0.1 (Fig. 10a). These regions shrink gradually as the aspect ratio increases and there is no evidence of
the formation of these static zones for E > 1 (Fig. 11b). On the
other hand, the size of zone Zr3 is observed to be the largest
for the extreme geometry given by E = 10 (Fig. 11b) due to the
increased extent of streamlining of the cylinder. Their size decreases progressively, as the body shape becomes increasingly
blunt, due to the enhanced levels of deformation and it vanishes
altogether for aspect ratio, E < 0.5, as shown in Fig. 10. Similarly,

the kinematic parameters, Reynolds number and Bingham number, also exert signicant inuence on the size of these zones.
With the increasing Reynolds number, the size of zone Zr3 decreases for a given value of the aspect ratio at low Bingham
numbers, while at high values of Bn, this effect is not so significant, as can be seen clearly in Fig. 11. The size of the static zone
Zr1 (in the front side of cylinder) decreases as the Reynolds
number increases at low Bingham numbers while Zr2 (formed
in the rear of the cylinder) increases and this is discussed more
later. However, the role of Reynolds number is somewhat countered by the increasing Bingham number in suppressing these
regions. Finally, irrespective of the value of the aspect ratio,
the far away rigid uid envelope Zr4, surrounding the uid zone
increases in size as the value of Bn increases, attaining a limiting
behavior corresponding to the fully plastic limit reaching at a
limiting value of Bingham number. Included in these gures
are also the predictions of the bi-viscosity model (with
ly/lB = 105) where the two results are seen to be in very good
agreement thereby suggesting that it is possible to use either
of these approaches with suitably chosen values of m or ly. This
nding is also in line with our previous studies [14,23,24].

Fig. 10. Unyielded uid zones (shaded): (a) E = 0.1 (b) E = 0.5 (dashed lines represent bi-viscosity model predictions) (ow is from left to right).

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

45

Fig. 11. Unyielded uid zones (shaded): (a) E = 1 (b) E = 10 (dashed lines represent bi-viscosity model predictions) (ow is from left to right).

Before leaving this section, it is worthwhile to analyze the functional relationship between the size of static zone Zr2 on one hand
and the Reynolds number and Bingham number on the other.
Fig. 12 shows the representative results for 0.1 6 E 6 1. These
trends are seen to be qualitatively similar to that for a circular cylinder [20]. However, for E P 2, this zone was not observed due to
the streamlining of the cylinder shape.
5.4. Flow kinematics
Figs. 13 and 14 show the variation of the x-component of velocity, Vx, along the positive x-axis and y-axis at the extreme values of
the Reynolds number, Re = 0.01 and Re = 40 for a range of values of
the Bingham number and for representative values of the aspect
ratio. An inspection of the velocity proles along the y-axis for
E = 10 (Fig. 13) shows that there are four different segments of
curve in the case of high Bingham numbers. These segments are
characterized as:
Fig. 12. Dependence of the size of the static rigid zone Zr2 on the Reynolds number
and Bingham number.

III: Rapid change in velocity Vx where uid experiences relatively a high rate of deformation.

46

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

IIIII: Solid body rotation, representing unyielded zone Zr3.


IIIIV: Corresponds to a ow region with very high strain rate.
IVV: Corresponds to a dynamic zone Zr4 moving with a constant velocity without shearing.
As the aspect ratio of the elliptic cylinder decreases, the
size of zone Zr3 shrinks and ultimately it vanishes. So only
the segments III, IIIIV and IVV are observed for aspect ratio
E 6 0.5 which suggest altogether the disappearance of the zone

Zr3 for this conguration of elliptical cylinders as shown in


Fig. 14. On the other hand, for E = 0.1, an examination of the
velocity prole along the x-direction shows three different
regions irrespective of the value of the Bingham number, Bn,
spanned here (Fig. 14). These segments are characterized as
follows:
III: Static (Vx = 0), corresponds to the rigid zone (static zone
Zr2) adhering to the surface of the cylinder.

Fig. 13. Velocity prole along (i) y = 0, x > 0 (ii) x = 0, y > 0 for E = 10 and E = 1.

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

IIIII: Velocity changes from 0 to Vx corresponding to the uidlike zone between the rigid envelope Zr4 and static zone Zr2.
IIIIV: Constant velocity Vx = 1, corresponding to the translation
of the rigid envelope Zr4.
As the aspect ratio of the cylinder increases, the static zone Zr2
decreases in size (Fig. 14) and disappears above aspect ratio E = 1

47

as shown in Fig. 13, hence one only observes the segments IIIII
and IIIIV in this case.
Fig. 15 shows the proles of the second invariant of the strain
rate tensor at the equator and on the vertical axis of the symmetry
at Re = 5 for a range of Bingham numbers and for the extreme values of aspect ratio (E = 0.1 and E = 10). For an elliptical cylinder
with E = 0.1 shown in Fig. 15a, for very small values of Bingham

Fig. 14. Velocity prole along (i) y = 0, x > 0 (ii) x = 0, y > 0 for E = 0.5 and E = 0.1.

48

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Fig. 15. Shear rate magnitude proles at the equator (y = 0) and on the vertical axis (x = 0) at Re = 5: (a) E = 0.1 and (b) E = 10.

Fig. 16. Inuence of the regularization parameter, m on the velocity proles in x- and y-directions at Re = 5 and Bn = 100.

number (Bn = 0.01 and Bn = 0.1), i.e., small deviations from the
Newtonian uid behavior, two peaks (at x = 2.5 and x = 5) are present at Re = 5. Under these conditions, the yield stress effects are
rather weak and the uid behaves nearly like a Newtonian uid
and there is a well formed wake in the rear of the plate which
probably does not extend up to the top edge of the cylinder. Hence,
the two peaks probably correspond to the sharp turning of the
streamlines at the two points along the wake contour. With the

increasing Bingham number (at a xed Reynolds number), as the


unyielded zone Zr2 appears and grows which behaves like a solid-region thereby extending the body contour in the downstream
direction. This, in turn, leads to a gradual turning of the streamlines
and hence, the rst minor peak disappears altogether. Thus, there
is only one maximum in the shear rate plot in x-direction located in
the uid zone between Zr2 and Zr4 for Bn P 1. While for aspect
ratio E = 10, Fig. 15b, the presence of one peak in the x-direction

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

49

Fig. 17. Variation of the modied pressure coefcient along the surface of cylinder for (i) E = 0.1 (ii) E = 0.5 (iii) E = 1.

conrms the uid zone between the cylinder and Zr4. On the other
hand, in the case of an elliptical cylinder with E = 10 (shown in
Fig. 15b) there are two zones of high shear rate in the positive ydirection which manifest in the form of two peaks of the velocity
prole in the y-direction. As aspect ratio approaches E = 0.1, only
one peak located in the uid zone between cylinder and Zr4 is observed (Fig. 15a). It is, however, appropriate to mention that the
shear rate is scaled here using (V1/2b) as the characteristic shear
rate. The only other possibility is to employ (V1/2a) as the characteristic shear rate. These two values are, however, inter-related via
the value of the aspect ratio, E. Both these choices approximate the
shear rate in an average sense, as actual shear rate could be significantly higher than this value in some parts of the ow domain.
However, since the values of the regularization parameter (m) have

been varied here by 23 orders of magnitude accompanied by a


very little change in the detailed velocity prole (shown in
Fig. 16) and/or in the value of drag coefcients clearly demonstrates the robustness of the values of m used here. This, as such,
lends further credibility to the reliability of the present results.
Figs. 17 and 18 show the pressure variation along the surface of
an elliptical cylinder for a range of values of the aspect ratios spanning the range 0.1 6 E 6 10 and Bingham number 0.01 6 Bn 6 100
at Re = 10 and Re = 40 in terms of the modied pressure coefcient,
C p . Evidently, the aspect ratio is seen to have a strong inuence on
the pressure coefcient distribution along the surface of the cylinder, similar to the case of Newtonian uids. These results conrm
that as the aspect ratio increases, the pressure decrease becomes
sharper in the front part of the cylinder. For each conguration

50

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Fig. 18. Variation of the modied pressure coefcient along the surface of cylinder for (i) E = 2 (ii) E = 5 (iii) E = 10.

of the elliptical cylinder, it is clear from these gures that the magnitude of the pressure on the surface of the cylinder shows a positive dependence on the both Reynolds number and Bingham
number.
5.5. Drag coefcients
The drag coefcient is a gross parameter which describes the
macroscopic uid mechanical behavior and it consists of two components, i.e., viscous drag due to shear stress and form drag (CDP)
due to the pressure eld, as dened in Eqs. (12) and (14). Fig. 19
shows the dependence of the total (CD) and pressure (CDP) drag
coefcients on the Reynolds number and Bingham number for a
range of values of the aspect ratio considered in this study. Both

drag coefcients exhibit the classical inverse dependence on the


Reynolds number while positive dependence on the Bingham
number irrespective of the shape of the cylinder. The relative contributions of the friction and form drag depend upon the shape of
the cylinder, as can be clearly seen in Fig. 20. For E 6 1, the elliptical cylinder acts more like a bluff body and thus the total drag is
dominated by the form drag drawing little contribution from the
viscous drag. As the aspect ratio E increases above unity, the object
becomes more streamlined where the total drag is dominated by
the viscous drag. Fig. 20 also reveals that the ratio CDP/CDF becomes
independent of the Reynolds number above the value of Bingham
number 50 for a given value of the aspect ratio while the total
drag coefcient increases with the increasing Bingham number
(Fig. 19). It is desirable to correlate the present numerical results

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

51

Fig. 19. Dependence of drag coefcient (CD) and pressure drag coefcient (CDP) on Reynolds number and Bingham number.

using regression which will facilitate the interpolation of the


present results for the intermediate values of the parameters.
The present numerical values of the total (CD) and pressure (CDP)
drag coefcients for elliptical cylinders have been correlated over
the range of conditions (0.01 6 Re 6 40, 0.01 6 Bn 6 100 and
0.1 6 E 6 10) as follows:

C D I1

m1 1  m2 Ref
1 m3 Bna m4 Bnb
Re

C DP I2

k1
Rek

1 k3 Bna k4 Bnb

20

21

Table 11 summarizes the values of the empirically tted


constants in Eqs. (20) and (21) for the total and pressure drag
coefcients respectively which exhibit additional dependence on
the aspect ratio E. Eqs. (20) and (21) reproduce the present numerical data (343 data points) for Reynolds number (1 6 Re 6 40) and
aspect ratio 0.1 6 E 6 10 with an average deviation (davg) < 6% except at the lowest Bingham number values of 0.01 and 0.1 where the

maximum deviations are of the order of 33% as shown in Table 11.


On the other hand, at low Reynolds numbers (0.01 6 Re < 1), Eqs.
(20) and (21) approximate the present numerical results with an
average error of less than 1% which rises to a maximum of 3.43%
over the range of values of E spanned here. Further statistical
examination of the results showed that the deviations between
the numerical and predicted, using Eqs. (20) and (21), values increased with the increasing values of the Reynolds number and aspect ratio and with the decreasing values of the Bingham number.
Therefore, there is a degree of self-cancelation of errors to some extent depending upon the combination of values of the parameters.
Unfortunately, Eqs. (20) and (21) do not seem to approach the expected Newtonian values as Bn ? 0. While the reasons for this are
not immediately obvious, similar difculties in reconciling the
numerical and experimental results in Bingham plastic media have
been observed for spheres [14,15] and for square cylinders [23,24].
Before leaving this section, it is worthwhile to compare the
present predictions with the numerical predictions of drag for a
plate oriented normal to the direction of ow available in the literature [54]. In terms of the present notations, their results

52

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

Fig. 20. Inuence of the Reynolds number, Bingham number and aspect ratio on the
drag ratio (CDP/CDF).

correspond to E = 0.15 but the two ends of the plate were chamfered at an angle of 30. Strictly speaking, therefore, it is neither
possible nor justied to make a comparison with their results.
However, limited results were obtained in the present study for
an elliptic cylinder with E = 0.15 for this purpose, and these are
plotted in Fig. 21 together with the results of Savreux et al. [54].
Notwithstanding the differences in the two geometries, the agreement is seen to be good in Fig. 21; the two values differ from each
other at most by 7.5%. All in all, the present results seem to be consistent with the previously available results for a circular cylinder
and a at plate oriented normal to the direction of ow.
6. Conclusions
In this work, extensive numerical results are reported for the
steady ow of Bingham plastic uids past an elliptical cylinder in
an innite medium over the range of conditions as: 0.01 6 Re 6 40,

Fig. 21. Comparison of the present drag coefcient results for E = 0.15 (hollow
symbols with solid lines) and for a normal at plate [54] (lled symbols with
dashed lines).

0.01 6 Bn 6 100, and 0.1 6 E 6 10. The extreme values of the aspect ratio, E, correspond to the limiting cases of a plane surface oriented normal and parallel to the direction of ow respectively.
Detailed results on the streamline contours, yielded/unyielded
zones, wake characteristics and drag coefcients are presented to
delineate the inuence of the inertial and yield stress forces on
the velocity and shear rate distribution in the close proximity of
the cylinder. Broadly speaking, while the increasing Reynolds number tends to eliminate the unyielded zones due to the increased
uid inertia, this tendency is strongly suppressed by the increasing
Bingham number due to the stronger yield-stress effects. In addition, the ow remains attached to the surface of the submerged
body up to higher Reynolds numbers in viscoplastic uids than
that in Newtonian uids. Indeed, for xed values of the Reynolds
number and aspect ratio, there exists a limiting Bingham number
beyond which the ow does not detach itself from the surface of

Table 11
Values of tted constants in Eqs. (20) and (21).
CD
E

CDP
0.1

0.2

0.5

10

0.2

0.5

10

0.01 6 Re < 1b
I2
0.434
k1
3.497
k3
5.841
k4
7.468
a
1.013
b
0.414
davg
0.07
dmax
0.42

0.439
3.049
6.616
7.809
1.014
0.401
0.08
0.59

0.318
2.434
7.735
7.786
1.012
0.401
0.07
0.60

0.247
1.924
9.284
7.739
1.008
0.409
0.11
0.77

0.198
1.457
6.979
11.932
0.414
1.003
0.16
0.87

0.169
0.617
10.087
28.561
0.222
1.005
0.72
3.29

0.148
1.800
0.056
9.540
0.50
1.004
0.94
3.43

1 6 Re 6 40
I2
1.234
k1
6.040
k3
3.141
k4
3.943
k
1.002
a
1.023
b
0.495
davg
2.70
dmax
18.57

1.172
5.591
3.362
3.842
1.002
1.023
0.489
2.51
17.56

0.973
4.536
3.767
3.888
1.002
0.494
1.021
2.87
17.12

0.774
3.601
4.684
3.736
1.001
1.016
0.503
2.97
18.16

0.560
2.785
3.286
5.947
1.001
0.520
1.009
3.64
20.34

0.350
1.023
17.181
4.318
1.001
1.006
0.294
4.54
24.58

0.240
1.800
9.524
0.093
1.001
1.004
0.271
3.91
21.47

0.01 6 Re < 1a
I1
0.370
m1
3.910
m3
5.348
m4
7.425
a
1.013
b
0.418
davg
0.08
dmax
0.48

0.411
3.973
5.286
7.359
1.013
0.421
0.09
0.43

0.382
4.051
7.388
5.363
0.424
1.010
0.10
0.45

0.407
4.351
7.347
5.289
0.440
1.006
0.13
0.54

0.501
4.412
6.155
8.127
0.997
0.444
0.12
0.43

0.534
4.698
8.352
10.123
0.998
0.460
0.13
1.00

0.863
4.834
12.427
12.950
0.997
0.468
0.15
1.69

1 6 Re 6 40
I1
1.315
m1
6.711
m2
0.003
m3
3.964
m4
2.888
f
0.412
a
0.497
b
1.023
davg
2.58
dmax
18.63

1.282
6.963
0.002
3.858
2.756
0.450
0.508
1.025
2.06
18.95

1.148
8.670
0.152
3.746
2.667
0.009
0.516
1.023
3.83
21.80

1.103
7.854
0.001
3.812
2.603
0.681
0.534
1.022
4.11
22.98

0.968
3.034
1.887
2.722
3.857
0.002
1.014
0.544
4.55
27.76

0.947
3.816
1.805
3.193
4.256
0.002
1.016
0.565
5.40
31.78

1.070
4.265
1.972
4.189
4.776
0.001
1.012
0.571
5.57
33.37

d: Percent relative r.m.s. deviation from the numerical data (Total data points = 49  7 = 343).
a
m2 = 0.
b
k = 1.

0.1

S.A. Patel, R.P. Chhabra / Journal of Non-Newtonian Fluid Mechanics 202 (2013) 3253

the cylinder. Similarly, for highly streamlined shapes (E > 1), the
ow does not detach even at Re = 40, the maximum value of the
Reynolds number used in this study. The numerical drag values obtained in this work have been correlated using empirical expressions thereby enabling their interpolation for the intermediate
values of the governing parameters. The present results are consistent with the previous studies in the limits of E = 1 (circular cylinder) and corresponding to the ow transverse to a plane surface
(E = 0.15). Finally, this work also demonstrates that it is possible
to use either the bi-viscous or the exponential regularization method to predict the location of the yield surfaces with comparable
levels of precision.
Acknowledgement
We gratefully acknowledge the detailed and constructive comments made by the two anonymous reviewers.
References
[1] R.B. Bird, G.C. Dai, B.J. Yarusso, The rheology and ow of viscoplastic materials,
Rev. Chem. Eng. 1 (1983) 170.
[2] H.A. Barnes, Yield stress a review or palrahel everything ows?, J NonNewt. Fluid Mech. 81 (1999) 133178.
[3] R.P. Chhabra, J.F. Richardson, Non-Newtonian Flow and Applied Rheology,
second ed., Butterworth-Heinemann, Oxford, 2008.
[4] R.P. Chhabra, J. Comiti, I. Machac, Flow of non-Newtonian uids in xed and
uidised beds: a review, Chem. Eng. Sci. 56 (2001) 127.
[5] R.P. Chhabra, Fluid mechanics and heat transfer with non-Newtonian liquids in
mechanically agitated vessels, Adv. Heat Transfer 37 (2003) 77178.
[6] R.P. Chhabra, Bubbles, Drops, and Particles in Non-Newtonian Fluids, second
ed., CRC Press, Boca Raton, FL, 2006.
[7] A.N. Beris, J.A. Tsamopoulos, R.C. Armstrong, R.A. Brown, Creeping motion of a
sphere through a Bingham plastic, J. Fluid Mech. 158 (1985) 219244.
[8] M. Beaulne, E. Mitsoulis, Creeping motion of a sphere in tubes lled with
HerschelBulkley uids, J. Non-Newt. Fluid Mech. 72 (1997) 5571.
[9] Prashant, J.J. Derksen, Direct simulations of spherical particle motion in
Bingham liquids, Comput. Chem. Eng. 35 (2011) 12001214.
[10] D.D. Atapattu, R.P. Chhabra, P.H.T. Uhlherr, Wall effect for spheres falling at
small Reynolds number in a viscoplastic medium, J. Non-Newt. Fluid Mech. 38
(1990) 3142.
[11] D.D. Atapattu, R.P. Chhabra, P.H.T. Uhlherr, Creeping sphere motion in
HerschelBulkley uids: ow eld and drag, J. Non-Newt. Fluid Mech. 59
(1995) 245265.
[12] M. Hariharaputhiran, R.S. Subramanian, G.A. Campbell, R.P. Chhabra, The
settling of spheres in a viscoplastic uid, J. Non-Newt. Fluid Mech. 79 (1998)
8797.
[13] B.T. Liu, S.J. Muller, M.M. Denn, Interactions of two rigid spheres translating
collinearly in creeping ow in a Bingham material, J. Non-Newt. Fluid Mech.
(2003) 4967.
[14] N. Nirmalkar, R.P. Chhabra, R.J. Poole, Numerical predictions of momentum
and heat transfer characteristics from a heated sphere in yield-stress uids,
Ind. Eng. Chem. Res. 52 (2013) 68486861.
[15] N. Nirmalkar, R.P. Chhabra, R.J. Poole, Effect of shear-thinning behavior on heat
transfer from a heated sphere in yield-stress uids, Ind. Eng. Chem. Res. 52
(2013) 1349013504.
[16] E. Mitsoulis, On creeping drag ow of a viscoplastic uid past a circular
cylinder: wall effects, Chem. Eng. Sci. 59 (2004) 789800.
[17] T.H. Zisis, E. Mitsoulis, Viscoplastic ow around a cylinder kept between
parallel plates, J. Non-Newt. Fluid Mech. 105 (2002) 120.
[18] N. Roquet, P. Saramito, An adaptive nite element method for Bingham uid
ows around a cylinder, Comput. Methods Appl. Mech. Eng. 192 (2003) 3317
3341.
[19] D.L. Tokpavi, A. Magnin, P. Jay, Very slow ow of Bingham viscoplastic uid
around a circular cylinder, J. Non-Newt. Fluid Mech. 154 (2008) 6576.
[20] S. Mossaz, P. Jay, A. Magnin, Non-recirculating and recirculating inertial ows
of a viscoplastic uid around a cylinder, J. Non-Newt. Fluid Mech. 177 (2012)
6475.
[21] S. Mossaz, P. Jay, A. Magnin, Criteria for the appearance of recirculating and
non-stationary regimes behind a cylinder in a viscoplastic uid, J. Non- Newt.
Fluid Mech. 165 (2010) 15251535.

53

[22] B. Deglo de Besses, A. Magnin, P. Jay, Viscoplastic ow around a cylinder in an


innite medium, J. Non-Newt. Fluid Mech. 115 (2003) 2749.
[23] N. Nirmalkar, R.P. Chhabra, R.J. Poole, On creeping ow of a Bingham plastic
uid past a square cylinder, J. Non-Newt. Fluid Mech. 171172 (2012) 1730.
[24] N. Nirmalkar, R.P. Chhabra, R.J. Poole, Laminar forced convection heat transfer
from a heated square cylinder in a Bingham plastic uid, Int. J. Heat Mass
Trans. 56 (2013) 625639.
[25] S. Tomotika, T. Aoi, The steady ow of a viscous uid past an elliptic cylinder
and a at plate at small Reynolds number, Quart. J. Mech. Appl. Math. 6 (1953)
290312.
[26] I. Imai, A new method of solving Oseens equations and its applications to the
ow past an inclined elliptic cylinder, Proc. R. Soc. London A 224 (1954) 141
160.
[27] H. Hasimoto, On the ow of a viscous uid past an inclined elliptic cylinder at
small Reynolds numbers, J. Phys. Soc. Jpn. 8 (1958) 653661.
[28] N. Epstein, J.H. Masliyah, Creeping ow through clusters of spheroids and
elliptical cylinders, Chem. Eng. J. 3 (1971) 169175.
[29] S.J.D. DAlessio, S.C.R. Dennis, A vorticity model for viscous ow past a cylinder,
Comput. Fluids 23 (1994) 279293.
[30] S.C.R. Dennis, P.J.S. Young, Steady ow past an elliptic cylinder inclined to the
stream, J. Eng. Math. 47 (2003) 101120.
[31] Z. Faruquee, D.S.K. Ting, A. Fartaj, R.M. Barron, R. Carriveau, The effects of axis
ratio on laminar uid ow around an elliptical cylinder, Int. J. Heat Fluid Flow
28 (2007) 11781189.
[32] D. Stack, H.R. Bravo, Flow separation behind ellipses at Reynolds numbers less
than 10, App. Math. Model. 33 (2009) 16331643.
[33] D.A. Perumal, G.V.S. Kumar, A.K. Dass, Lattice Boltzmann simulation of viscous
ow past elliptical cylinder, CFD Lett. 4 (2012) 127139.
[34] W.A. Khan, J.R. Culham, M.M. Yovanovich, Fluid ow around and heat transfer
from elliptical cylinders: Analytical approach, J. Thermophys. Heat Transfer 19
(2005) 178185.
[35] P. Sivakumar, R.P. Bharti, R.P. Chhabra, Steady ow of power-law uids across
an unconned elliptical cylinder, Chem. Eng. Sci. 62 (2007) 16821702.
[36] R.P. Bharti, P. Sivakumar, R.P. Chhabra, Forced convection heat transfer from an
elliptical cylinder to power-law uids, Int. J. Heat Mass Trans. 51 (2008) 1838
1853.
[37] P.K. Rao, A.K. Sahu, R.P. Chhabra, Flow of Newtonian and power-law uids past
an elliptical cylinder: a numerical study, Ind. Eng. Chem. Res. 49 (2010) 6649
6661.
[38] A. Putz, I.A. Frigaard, Creeping ow around particles in a Bingham uid, J. NonNewt. Fluid Mech. 165 (2010) 263280.
[39] J.R. Zierenberg, H. Fujioka, R.B. Hirschl, R.H. Bartlett, J.B. Grotberg, Pulsatile
blood ow and oxygen transport past a circular cylinder, J. Biomech. Eng. 129
(2007) 202215.
[40] F.M. Najjar, S.P. Vanka, Simulations of the unsteady separated ow past a
normal at plate, Int. J. Numer. Methods Fluids 21 (1995) 525547.
[41] R. Glowinski, A. Wachs, On the numerical simulation of viscoplastic uid ow,
in: P.G. Ciarlet (Ed.), Handbook of Numerical Analysis, Elsevier, North-Holland,
2011, pp. 483717.
[42] T.C. Papanastasiou, Flow of materials with yield, J. Rheol. 31 (1987) 385404.
[43] E.J. ODonovan, R.I. Tanner, Numerical study of the Bingham squeeze lm
problem, J. Non-Newt. Fluid Mech. 15 (1984) 7583.
[44] B.T. Liu, S.J. Muller, M.M. Denn, Convergence of a regularisation method for
creeping ow of a Bingham material about a rigid sphere, J. Non-Newt. Fluid
Mech. 102 (2002) 179191.
[45] G.R. Burgos, A.N. Alexandrou, V. Entov, On the determination of yield surfaces
in HerschelBulkley uids, J. Rheol. 43 (1999) 463483.
[46] E. Mitsoulis, T. Zisis, Flow of Bingham plastics in a lid-driven square cavity, J.
Non-Newt. Fluid Mech. 101 (2001) 173180.
[47] P. Neofytou, A 3rd order upwind nite volume method for generalised
Newtonian uid ows, Adv. Eng. Softw. 36 (2005) 664680.
[48] J.H. Masliyah, N. Epstein, Steady symmetric ow past elliptical cylinders, Ind.
Eng. Chem. Fundam. 10 (1971) 293299.
[49] H. Coudeville, P. Trepaud, E.A. Braun, Drag measurements in slip and transition
ow, Rareed Gas Dyn. 1 (1965) 444466.
[50] K.O. Tamada, H. Miura, T. Miyagi, Low-Reynolds number ow past a cylindrical
body, J. Fluid Mech. 132 (1983) 445455.
[51] S.C.R. Dennis, W. Qiang, M. Coutanceau, J.L. Launay, Viscous ow normal to a
at plate at moderate Reynolds numbers, J. Fluid Mech. 248 (1993) 605635.
[52] K.M. In, D.H. Choi, M.U. Kim, Two-dimensional viscous ow past a at plate,
Fluid Dyn. Res. 15 (1995) 1324.
[53] M.F. Randolph, G.T. Houlsby, The limiting pressure on a circular pile loaded
laterally in cohesive soil, Geotechnique 34 (1984) 613623.
[54] F. Savreux, P. Jay, A. Magnin, Flow normal to a at plate of a viscoplastic uid
with inertia effects, AICHE J. 51 (2005) 750758.

You might also like