You are on page 1of 13

Ind. Eng. Chem. Res.

2005, 44, 5815-5827 5815

Flow and Forced Convection Heat Transfer in Crossflow of


Non-Newtonian Fluids over a Circular Cylinder
A. A. Soares and J. M. Ferreira
Departamento de Fı́sica, Universidade de Trás-os-Montes e Alto Douro, Apartado 1013,
5000-911 Vila Real, Portugal

R. P. Chhabra*
Department of Chemical Engineering, Indian Institute of Technology, Kanpur 208016, India

The steady and incompressible flow of power-law type non-Newtonian fluids across an unconfined,
heated circular cylinder is investigated numerically to determine the dependence of the individual
drag components and of the heat transfer characteristics on power-law index (0.5 e n e 1.4),
Prandtl number (1 e Pr e 100), and Reynolds number (5 e Re e 40). The momentum and
energy equations are expressed in the stream function/vorticity formulation and are solved using
a second-order accurate finite difference method to determine the pressure drag and frictional
drag as well as the local and surface-averaged Nusselt numbers and to map the temperature
field near the cylinder. The accuracy of the numerical procedure is established using previously
available numerical and analytical results for momentum and heat transfer in Newtonian and
power-law fluids. The results reported herein provide fundamental knowledge of the flow and
heat transfer behavior for the flow of non-Newtonian fluids over a circular cylinder; these results
further show that the effect of the power-law index on such behavior is strongly conditioned by
the kinematic conditions and less so by the type of thermal boundary condition prescribed at
the cylinder surface.

1. Introduction pressure and frictional drag, as well as Nusselt and/or


Sherwood numbers, for the flow of Newtonian fluids
The flow of fluids and forced convection across a across a single cylinder over most conditions of interest.
heated cylinder has been the subject of considerable On the other hand, it is readily acknowledged that many
research interest because of its relevance in many materials (e.g., polymer solutions, melts, muds, emul-
engineering applications. For instance, a knowledge of sions, and suspensions) encountered in chemical and
the hydrodynamic forces experienced by submerged processing applications often exhibit complex non-
cylindrical objects such as off-shore pipelines is essential Newtonian behavior including shear thinning and shear
for the design of such structures. On the other hand, thickening, viscoelasticity, and yield stress.13,14 More-
because of changing process and climatic conditions, one
over, reliable knowledge of the hydrodynamic forces and
also needs to determine the rate of heat transfer from
heat transfer characteristics of submerged cylindrical
such structures. Furthermore, present economic and
objects is needed in connection with engineering pro-
environmental concerns have raised an interest in
cesses which include the use of wires and thin cylinders
methods to determine and control forced convection heat
transfer from horizontal cylindrical structures. Ad- as measurement probes and sensors in non-Newtonian
ditional industrial processes where heat/mass transfer flow15-18 and in the design of slurry pipelines where
from an isolated cylinder plays an important role large particles are conveyed in a non-Newtonian vehicle.
include anemometry and chemical or radioactive con- Additional examples are found in polymer processing
tamination/purification, glass cooling, plastics and in- operations such as the use of submerged surfaces to
dustrial devices, and other processes from turbine form weld lines. Despite such overwhelming importance
blades to electronic circuits. Because of the importance and frequent occurrence of non-Newtonian behavior,
of these applications, there has been a great deal of very little work has been reported in the literature on
interest in the flow of Newtonian fluids and the heat the crossflow and forced convection heat transfer to non-
transfer across a single cylinder from the experimental, Newtonian fluids from a long, heated cylinder, and the
analytical, and numerical standpoints (e.g., see refs present work aims to fill this gap in the existing
1-8), and excellent reviews are indeed available (e.g., literature. It is, however, instructive and useful to
refs 9-12). On the basis of a combination of such briefly summarize the previous scant results available
analytical/numerical simulations and the available ex- in the literature prior to undertaking the formulation
perimental results, it is perhaps fair to say that satis- of the present problem.
factory methods are now available which enable the From a theoretical standpoint, because of the rheo-
prediction of gross engineering parameters such as logical complexities of non-Newtonian flow across a long,
heated cylinder, most of the work in this field relies on
* Author to whom correspondence should be addressed. the use of a simple, two-parameter power-law model,
Tel.: 00 91 512 259 7393. Fax: 00 91 512 259 0007/0104. which must be combined with governing equations of
E-mail: chhabra@iitk.ac.in. momentum, continuity, and energy to obtain the pres-
10.1021/ie0500669 CCC: $30.25 © 2005 American Chemical Society
Published on Web 06/11/2005
5816 Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 1. Real (x, y) plane and the computational (, θ) plane. Variables include the distance (H) from the external boundary to the
cylinder surface, the free stream fluid temperature (T∞), and the uniform approach velocity (U∞).

sure and friction drag coefficients and/or heat transfer reviewed elsewhere.26 Similarly, there have been a few
rate. The use of this simple model enables the degree experimental studies dealing with the forced convection
of complexity of the non-Newtonian effects to be kept heat transfer from a circular cylinder to streaming non-
at a tractable level, well-suited for the study of the Newtonian fluids,27-32 but most of these relate either
simplest and also the most common type of non- to very high values of Reynolds number and/or to
Newtonian behavior, i.e., shear-thinning behavior (n < the flow of viscoelastic or drag-reducing polymer solu-
1) and shear-thickening behavior (n > 1). Thus, the tions.33 Admittedly, these studies are not of direct
creeping flow of power-law fluids past an unconfined interest here, but these are mentioned here for the sake
circular cylinder has been investigated by Tanner,19 who of completeness. Finally, it is appropriate to mention
demonstrated that, in the limit of zero Reynolds num- here that limited results on drag and heat transfer from
ber, the Stokes paradox is removed for shear-thinning square cylinders to power-law fluids were reported
fluids but not for shear-thickening fluids. Subsequently, recently.34
a similar inference was also reported by Marǔsić- In summary, while scant numerical results are avail-
Paloka.20 Tanner19 also supplemented his analytical able on the total drag of circular cylinders in power-
results with numerical predictions for values of the law fluids at finite Reynolds numbers, no information
power-law index from 0.4 to 0.9; the correspondence is available on the individual drag components and on
between the two results was found to be reasonable. the forced convection heat transfer from circular cylin-
Subsequently, these results have been extended to ders to power-law fluids. The present work thus aims
values of the power-law index as low as n ) 0.2,21 and to fill these gaps in the existing literature and to
these new results are consistent with the previous numerically solve the momentum and heat transfer
results for spheres and cylinders. Apart from these equations for steady power-law flow across a long,
creeping flow analyses, as far as is known to us, there heated circular cylinder over a range of power-law
have been only two studies of power-law fluid flow over indices (0.5 e n e 1.4), Reynolds numbers 5 e Re e 40,
a circular cylinder at finite Reynolds numbers.22,23 and Prandtl numbers Pr ) 1, 5, 25, 50, and 100, in order
D’Alessio and Pascal22 numerically investigated the to determine the Nusselt number and related param-
steady power-law flow around a cylinder at Reynolds eters, map the isothermal lines, and also extend the
numbers Re ) 5, 20, and 40 and examined the depen- work of Chhabra et al.23 and investigate the effect of
dence of critical Reynolds number, wake length, separa- the power-law index on both the pressure and frictional
tion angle, and drag coefficient on the value of the drag coefficients. A blockage value of β ) 0.037 was
power-law index for a fixed blockage ratio β ) 0.037, selected for this work because it corresponds to asymp-
where β is defined as the ratio between the cylinder totic boundary conditions at 54.6 radii away from the
diameter and the distance H from the external boundary cylinder surface,23 a distance which is regarded to be
to the cylinder surface (Figure 1). They reported that a sufficient to ensure that the flow and heat transfer at
progressive increase in Reynolds number (Re) from 5 this surface are insensitive to boundary effect and,
to 40 resulted in a decrease in the degree of convergence, therefore, models the flow of an unbounded fluid across
which restricted the range of power-law indices for the heated cylinder surface.
which a fully converged numerical solution was possible
from 0.65 to 1.2 at Re ) 5 to 0.95-1.1 at Re ) 40. 2. Basic Equations
Subsequently, Chhabra et al.23 extended the work of
D’Alessio and Pascal22 to include a wide range of Consider the steady flow of an incompressible power-
conditions (β ) 0.037, 0.082, and 0.164; Re ) 1, 20, and law fluid with a constant far away streaming velocity
40; and 0.2 e n e 1.4). However, in both these studies, (U∞) and temperature (T∞), along the x direction normal
only the values of the total drag coefficient were re- to the axis of a circular cylinder (Figure 1a). Since the
ported, and hence, further work is still needed to deter- present study is restricted to long cylinders and flow
mine the effect of the power-law index on the individual conditions of Re e 40, the flow across the cylinder is
pressure and frictional drag coefficients. Coelho and steady and two-dimensional; i.e., all flow variables are
Pinho24,25 undertook a detailed experimental study of independent of the z- coordinate and are, therefore,
the flow characteristics of various vortex shedding functions of cylindrical coordinates r and θ alone.
regimes for the flow of non-Newtonian fluids around a Furthermore, the thermophysical properties (density F,
cylinder at Re ) 50-9000 and reported that shear heat capacity Cp, thermal conductivity k, and power-
thinning resulted in decreased cylinder boundary-layer law parameters K and n) are assumed to be independent
thickness and diffusion length. Aside from these studies, of temperature. Under these conditions, the momentum
some effort has also been directed at studying the mo- and energy equations are not coupled. The equation of
mentum and thermal boundary-layer flows of power- continuity, the r and θ components of the equations of
law fluids over a circular cylinder, and these have been motion, and the thermal energy (in the absence of
Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5817

viscous dissipation) in cylindrical coordinates35 can be (U∞/a)2I2, respectively. In eq 3, the temperature T is


expressed in their dimensionless stream function/vor- scaled in two different ways depending on the thermal
ticity formulation in terms of the polar coordinates (, boundary condition imposed at the cylinder surface, i.e.,
θ) with  ) ln(r/a), where a is the radius of the cylinder, at  ) 0 (see Figure 1). The two commonly used thermal
as follows, boundary conditions at  ) 0 are that of either a
constant temperature (Ts) or a constant heat flux (qs).
continuity equation Thus, for the constant temperature boundary condition,
1 ∂ ∂ψ ∂ ∂ψ the dimensionless temperature T is related to its
e ∂ ( )
e
∂θ
-
∂θ
ψ+ (
∂
)0 ) (1) dimensional counterpart as e-T(Ts - T∞), whereas, for
the constant heat flux boundary condition, the dimen-
sionless temperature T is related to its dimensional
-momentum component counterpart as e-Tqsa/k. The Reynolds number is

( )( ) defined as
2
∂ψ ∂ ψ ∂ψ ∂2ψ ∂ψ
- +ψ + +ψ )
∂θ ∂∂θ ∂ ∂θ2 ∂ F(2a)nU∞2-n
-
1 ∂p 2n - ∂ 
-
2 ∂ Re
e
∂ [
(e τrr) +
∂τrθ
∂
- τθθ (2a) ] Re )
K
where K denotes the power-law consistency index and
(8)

θ - momentum component n denotes the power-law index.

( ) ( )
The Prandtl number is defined as
∂ψ ∂2ψ ∂ψ ∂ψ ∂2ψ

( )
- + + + ψ )
∂θ ∂2 ∂ ∂ ∂∂θ CpK U∞ n-1

[ ]
Pr ) (9)
1 ∂p 2n -2 ∂ 2 ∂τθθ k 2a
- - e (e τrθ) + (2b)
2 ∂θ Re ∂ ∂θ Eliminating the pressure in eq 2 by the usual method
energy equation of crossdifferentiation with the introduction of the

( )
vorticity ω, followed by some rearrangement, leads to
2 2
eRePr ∂ψ

( )
∂ T ∂ T ∂T
+ - 2 + + ∂2ω ∂2ω ∂ω ∂ω
∂2 ∂θ2 ∂ 2 ∂θ η + 2 + 2λ + 2µ + γω ) F (10a)
( )) ( ) ∂2 ∂θ ∂ ∂θ
∂T eRePr ∂ψ eRePr ∂ψ
∂θ 2
ψ+ (
∂
+T 1+
2 ∂θ
) 0 (3)
where
where p, Re, and Pr are the dimensionless pressure, the
∂η Re‚e ∂ψ
Reynolds number, and the Prandlt number, respec- λ) - η - n+1 (10b)
tively. ∂ 2 ∂θ
The dimensionless components of the extra stress
∂η Re‚e ∂ψ
tensor for a power-law fluid are written as
µ)
∂θ
+ n+1
2 ∂(+ψ ) (10c)
τij ) -ηij (4)
∂η Re‚e ∂ψ
where η is the dimensionless viscosity and ij are the γ)-2 +η+ n (10d)
∂ 2 ∂θ
dimensionless components of the rate-of-deformation

( )( )
tensor (e.g., ref 35).
The equation for the dimensionless power-law viscos- ∂2ψ ∂2ψ ∂2η ∂2η ∂η
F) ψ- 2
+ 2 2
- 2+2 -
ity is ∂ ∂θ ∂θ ∂ ∂

η ) I2(n-1)/2 (5) 4
∂2ψ ∂η
-
∂2 η
∂θ∂ ∂θ ∂θ∂ (
(10e) )
where I2 is the dimensionless second invariant of the The exponential scaling for the stream function, the
rate-of-deformation tensor given as vorticity, and the temperature is appropriate since the
stream function is exponentially large far from the

[( ∂2ψ ∂2ψ
) ( )] cylinder, the vorticity is exponentially small everywhere
2
∂2 ψ 2
I2 ) e-2 ψ - + 2 +4 (6) except in the region of the wake, and the temperature
∂2 ∂θ ∂θ∂
decreases rapidly away from the cylinder. This scaling
procedure, used by D’Alessio and Pascal22 and Chhabra
The vorticity in its scaled form is given as et al.,23 is also employed in the present study because
it suppressed the numerical instabilities which typically
∂2ψ ∂2ψ ∂ψ occur for lower values of n.
2
+ 2+2 +ψ+ω)0 (7)
∂ ∂θ ∂ Due to the two-dimensional nature of the flow prob-
lem (x-y plane) and because the oncoming flow is in
In eqs 1-7, the dimensionless stream function ψ, the x direction, it is sufficient to consider the region y
vorticity ω, and pressure p are related to their dimen- g 0 and x2 + y2 g 1 only. Thus, the corresponding region
sional counterparts as eU∞aψ, e-(U∞/a)ω, and (U∞2F/ in the (, θ) plane is defined by  g 0 and 0 e θ e π
2)p, respectively. The dimensionless components of the (Figure 1).
extra stress tensor τij and the dimensionless second The physically realistic boundary conditions for this
invariant of the rate of deformation tensor I2 are related flow are expressed as follows. On the cylinder surface,
to their dimensional counterparts as K(U∞/a)nτij and i.e., at  ) 0, the usual no-slip condition is applied
5818 Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

∂ψ ∂ψ j ) is
where the dimensionless viscosity in real space (η
) )0 (11a) given as
∂ ∂θ
which together with eq 7 gives j ) hI2(n-1)/2
η (13)

∂2ψ and the dimensionless second invariant of the rate of


ψ ) 0 and ω ) - (11b)
∂2 deformation tensor in real space (Ih2) is obtained through
the insertion into eq 6 of ψ ) e-ψ h , where ψ
h is the
The two commonly used thermal boundary conditions dimensionless stream function in real space. Inserting
at the surface of the solid cylinder are scaled as eq 13 into eq 12a gives

T)1 (11c) 2n+1


CDF ) -
Re
∫0π (Ih2(n-1)/2 ω))0 sin(θ) dθ (14)
for the constant temperature condition and
From ref 38,
∂T
)T-1
[ ]
(11d)
∂ 2n+1 jωj)
∫0θ
∂(η
p|)0 ) p0 + dθ (15)
for the constant heat flux condition, where the dimen- Re ∂ )0

sionless temperature (T) used in the present study is where p0 is the dimensionless stagnation pressure and
related to the previously used34,36,37 dimensionless tem- j is the dimensionless vorticity in real space, i.e.,
ω
perature in real space (Th ) as T ) eTh.
On the plane of symmetry at θ ) 0, π, we use the
j ) e-ω
ω (16)
following:
Equations 13 and 16 are inserted into eq 15 which is
∂T
ψ)ω) )0 (11e) then used to integrate eq 12b by parts, giving
∂θ
2n+1
Far away from the cylinder surface, for ∞ ) 4, we use CDP )
Re
∫0π [∂∂ (e-hI2(n-1)/2 ω)])0 sin(θ) dθ (17)
the asymptotic approximation for stream function and
vorticity given by Chhabra et al.,23
which is rearranged as

[
CD - θ
ψ ≈ sin(θ) +
2
e
π (
- erf(Q) ) (11f)
CDP )
2n+1
Re
∫0π (Ih2(n-1)/2))0 (∂ω
∂
- ω) +

ω≈-
CDReI2
2n+1xπ
(1-n)/2
Qe-Q
2
(11g)
n - 1 1 ∂Ih2
2 hI2 ∂
ω ])0
sin(θ) dθ (18)

where CD is the drag coefficient, The total drag coefficient,

CD ) CDF + CDP (19)

x
Re (1-n)/4 θ
Q)e /2

2
I
n 2
sin
2 () (11h)
is obtained from eqs 14 and 18.
The rate of heat transfer is usually expressed in terms
and erf(Q) is the standard error function. of the Nusselt number. The local Nusselt number on
The far away stream temperature boundary condi- the cylinder surface is defined by36,37
tions are

T ) 0 and
∂T
∂
)0 (11i)
Nu(θ) ) -2 (∂T∂rjh ) rj)1
) -2 (∂T∂ - T) )0
(20)

for the constant temperature boundary condition (iso-


The system of elliptic partial differential eqs 3, 7, and thermal cylinder surface) and by
10, together with the corresponding boundary conditions
(eq 11), has been discretized using the finite difference
method. The resulting system of algebraic equations has
been solved using an iterative Gauss-Seidel relaxation
Nu(θ) )
h(2a) 2
k
)
T | )0
(21)

method. Once the values of ω, ψ, and T are computed for the constant heat flux boundary condition, where rj
in the flow domain, the coefficients of frictional drag ) e and T ) eTh.
CDF, pressure drag CDP, total drag CD, local Nusselt The surface averaged value of the Nusselt number is
number Nu(θ), and average Nusselt number Nu can be given by
determined from the corresponding equations. The
equations for the pressure and frictional drag coef-
ficients are obtained as follows Nu )
1
π
∫0π Nu(θ) dθ (22)

2n+1
CDF ) -
Re
∫0π (ηj ω))0 sin(θ) dθ (12a)
3. Numerical Solution Method
The numerical solution procedure used here is an
CDP ) - ∫0π
(p))0 cos(θ) dθ (12b)
iterative Gauss-Seidel relaxation method already used
in an earlier study,23 and to avoid redundancy only the
Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5819

main features are included here. For the discretization Table 1. Comparison of Present Results with Literature
of eqs 3, 5, 7, and 10, a finite difference method was Values of Drag Coefficient (CD), Separation Angle (θs),
employed. A second-order upwind differencing technique Dimensionless Wake Length (L/a), and Surface-Averaged
Nusselt Number (Nu) for Newtonian Fluids (Reynolds
is used to solve eqs 3 and 10 with one-sided difference Number Re and Prandtl Number Pr ) 0.715)
approximations to the convective terms of T and ω,
whereas for the diffusion terms, the central difference Re CD θs L/a Nu authors
approximation was used. For all other terms in these 5 3.95 1.46 present results
equations, central difference approximations have been 4.116 Dennis and Chang2
employed, except for the first derivative of ψ in the 4.0a 1.39b Lange et al.6
10 2.76 29.2° 0.50 1.86 present results
direction of , which was determined using the second- 2.846 29.6° 0.53 Dennis and Chang2
order forward difference approximation. This forward 1.86 Mandhani et al.36
difference approximation is necessary to obtain conver- 2.8c 1.81b Lange et al.6
gence for n e 0.6. Equations 5 and 7 are rewritten as 20 1.99 43.5° 1.85 2.43 present results
finite difference equations using the central difference 2.045 43.7° 1.88 Dennis and Chang2
of second-order accuracy. The resulting system of equa- 2.0001 1.82 Fornberg3
2.0c 2.41b Lange et al.6
tions is solved using a Gauss-Seidel iterative method 1.99 43.24° 1.79 2.4121 Juncu7
with an under-relaxation factor of 0.8 to the tempera- 30 1.67 49.5° 3.20 2.85 present results
ture and vorticity variables. To obtain consistent ap- 1.7167 49.6° 3.223 Takami and Keller8
proximations for ψ, ω, and T, for each iteration a sweep 2.88 b Lange et al.6
is made through all mesh points and updated values of 40 1.49 53.4° 4.55 3.20 present results
1.522 53.8° 4.69 Dennis and Chang2
the drag coefficient CD and the Nusselt number Nu,
1.4980 4.48 Fornberg3
which correspond to each of the two thermal surface 1.5c 3.28b Lange et al.6
boundary conditions, are determined by numerical a Evaluated from Figure 6 of ref 6. b Evaluated from eq 18 of
integration of eqs 19-21 on the cylinder surface using
ref 6. c Evaluated from Figure 8 of ref 6.
Simpson’s rule. Note that the solutions of the temper-
ature fields defined by the two thermal surface bound-
ary conditions are obtained simultaneously. Conver- shown herein, it is necessary to validate the numerical
gence was achieved when, for the same iteration, the solution procedure used in this study. The case of
variation in CD and Nu values in two successive itera- Newtonian flow, n ) 1, past a cylinder was used as the
tions was less than a preset value of 10-8. To accelerate initial test of the numerical solution procedure. For the
the convergence of the numerical solution, the Newto- range of Re numbers used in this study and a Prandtl
nian steady solution values of ψ, ω, and T at every point number Pr ) 0.715 (corresponding to the flow of air),
of the grid were used as the initial guesses for non- the numerical values of CD, separation angle θs, dimen-
Newtonian flow. sionless wake length L/a, and average Nusselt number
The outer boundary was positioned at ∞ ) 4, corre- Nu for the constant temperature boundary condition
sponding to asymptotic boundary conditions at a dis- were all found to be in excellent agreement with
tance of ∼54.6 radii away from the cylinder. For a (N + previous results available in the literature for Newto-
1) × (M + 1) computational grid, the spacings in the nian fluids (Table 1). Because for a blockage ratio β )
radial  and angular θ directions are ∞/N and π/M, 0.037 and Re ) 20 and 40 the dependences on the
respectively. For the choice of appropriate grid, ad- power-law index of the total drag and the streamline
ditional tests were carried out for the largest Reynolds and isovorticity patterns are identical to those already
and Prandtl numbers (Re ) 40 and Pr ) 100) for 0.5 e reported in a previous study,23 their detailed comparison
n e 1.4. Initial tests carried out using three grids (mesh with the original results of D’Alessio and Pascal22 is not
sizes 101 × 51, 201 × 101, and 401 × 101) showed that repeated here. As far as is known to us for power-law
the gain in accuracy of the drag coefficient and average fluids, no numerical heat transfer results for flow across
Nusselt number which resulted from the use of the a circular cylinder using a fixed blockage ratio are
finest grid (401 × 101) was overall <2%. Weighing the available for validation. However, to strengthen the
marginal differences in the values of the drag coef- reliability of our results, additional tests for a collection
ficients and average Nusselt numbers obtained with the of cylinders within the framework of the free surface
401 × 101 grid against the disproportionately large cell model were carried out to validate the numerical
increase in CPU time, the 201 × 101 grid was consid- method used in the present study to determine the
ered adequate for the present study since it provided solution of the energy equation. For the first test, the
sufficient numerical resolution. Newtonian flow over a collection of cylinders at constant
temperature was simulated using the conditions in
4. Results and Discussion Table 1 of Mangadoddy et al.,37 i.e., n ) 1; Re ) 1; 10 e
In the present study, the momentum and thermal Pe e 500; voidages e ) 0.4, 0.5, and 0.6; and a 101 × 51
energy equations have been solved numerically in the grid. A discrepancy of <2.5% was observed relative to
following range of conditions: 5 e Re e 40; Pr ) 1, 5, the average Nusselt number values reported by Man-
25, 50, and 100; and 0.5 e n e 1.4 for the two commonly gadoddy et al.37 For the second test, the power-law flow
used thermal boundary conditions. Extensive results on over a collection of cylinders was simulated to determine
the detailed temperature fields, the local (Nu(θ)) and the average Nusselt number values for conditions of Re
surface-averaged (Nu) values of the Nusselt number, ) 1, Pe ) 500, and e ) 0.9. The discrepancy between
and the drag coefficients (CD, CDF, and CDP) have been the values obtained from the present numerical method
obtained to help elucidate the non-Newtonian effects of and those obtained from the analytical expressions of
the power-law fluids in the aforementioned range of Ferreira and Chhabra39 for creeping flow increased from
conditions. 0.4% at n ) 1 to 3.8% at n ) 0.6. This is perhaps not
4.1. Validation of Results. Prior to undertaking the surprising, owing to the assumptions inherent in their
detailed presentation and discussion of the new results approximation analysis.
5820 Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 2. Variation of (a) CDF, (b) CDP, and (c) CD with Re and n.

4.2. Dependence of the Drag Coefficient on Re Re ) 40, a result which is consistent with their numer-
and n. The hydrodynamic drag force exerted by the ical findings of a much less significant frictional con-
fluid on the solid cylinder is determined by the indi- tribution to the overall drag for a square cylinder
vidual contributions due to the pressure and frictional (30-45% at Re ) 5 and 0.5-0.21% at Re ) 40). Such a
forces acting on the cylinder and can be expressed in significant difference in the frictional contribution can-
terms of three dimensionless groups, namely, the drag not be attributed to the higher blockage ratio of β )
coefficients (frictional drag CDF, pressure drag CDP, and 0.143 used by Gupta et al.,34 because (although the
total drag CD), the Reynolds number Re, and the power- present results for a circular cylinder were obtained
law index n. The dependence of the drag coefficients on using β ) 0.037) additional checks using a higher
the two other dimensionless groups is shown in Figure blockage ratio of β ) 0.164 showed that the relative
2. The results showed that an increase in the power- weight of the frictional contribution remained almost
law index caused an overall increase in the frictional unchanged (36.3-48.7% at Re ) 5 and 21.7-39.7% at
drag as the fluid behavior transitioned from shear- Re ) 40). It can, thus, be concluded that the more
thinning to Newtonian and finally to shear-thickening significant frictional contribution to the overall drag
behavior. Also, the dependence of the frictional drag on which was observed in the present study of a circular
the power-law index was more pronounced at higher cylinder was due to its different geometrical configura-
Reynolds numbers (Figure 2a). Such an increase in the tion relative to a square cylinder as used by Gupta et
dependence has also been reported for a square cylin- al.,34 where pressure forces play a more dominant role
der34 and is consistent with an increase in the general than that in the case of a circular cylinder.
level of shearing for higher Reynolds number values, 4.3. Heat Transfer Characteristics. Besides the
thereby resulting in higher levels of effective viscosity dependence on the power-law index n and Reynolds
for larger values of n, which in turn influences the number Re, the heat transfer characteristics are also
frictional component of the drag. Conversely, an in- dependent on the Prandtl number Pr and on the type
crease in the power-law index caused a decrease in of thermal boundary condition applied to the cylinder’s
pressure drag, and the dependence of the pressure drag surface, i.e., constant temperature or constant heat flux.
on the power-law index became less pronounced at For the constant heat flux boundary condition, the
higher Reynolds numbers (Figure 2b). Similar effects effects of the power-law index on the local Nusselt
have previously been reported for square cylinders34 and number Nu(θ) were more pronounced but otherwise
for spherical and spheroidal particles translating in similar to those for the constant temperature boundary
power-law fluids.40,41 The present results further showed condition; therefore, only the latter are shown herein.
that the increased dependence of CDF on the power-law The study of these effects showed that the local Nusselt
index at higher Re, combined with the decreased number always increased with Pr and/or Re (Figures 3
dependence of CDP under the same conditions, resulted and 4). A maximum value in Nu(θ) was located close to
in a Re (Re ) 15 ( 5) at which the dependence of the the front stagnation point, and its peak value was found
total drag on the power-law index was minimal (Figure to increase with shear thinning (Figures 3 and 4). For
2c). Increased shear thinning resulted in an increase highly shear-thinning fluids (e.g., n ) 0.5), an increase
in total drag for Re < 15 ( 5, whereas for Re > 15 ( 5, in Prandtl number from Pr ) 1 to Pr ) 100 always
the opposite effect was observed (Figure 2c). Such effects resulted in a downstream shift in the position of this
stem from the significant frictional contribution to the maximum (Figures 3 and 4). Moreover, under the same
overall drag, which at Re ) 5 ranged from 36.1% at n shear-thinning condition (i.e., n ) 0.5), an increase in
) 0.5 to 48.8% at n ) 1.4 and remained significant even Reynolds number from Re ) 5 to Re ) 40 also resulted
at Re ) 40, where it ranged from 21.8% at n ) 0.5 to in a downstream shift in the position of this maximum
40.1% at n ) 1.4. However, a similar effect was not at lower Prandtl numbers (e.g., Pr ) 1), whereas at
observed by Gupta et al.,34 who reported that increased higher Prandtl numbers (e.g., Pr ) 100), the position of
shear thinning resulted in increased total drag up to the maximum shifted upstream (Figures 3 and 4). It was
Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5821

Figure 3. Effect of the power-law index (n) on the local Nusselt


number Nu(θ) at Re ) 5, using the constant temperature boundary
condition on the cylinder surface for (a) Pr ) 1 and (b) Pr ) 100 Figure 4. Effect of the power-law index (n) on the local Nusselt
(downstream θ ) 0° and upstream θ ) 180°). number Nu(θ) at Re ) 40, using the constant temperature
boundary condition on the cylinder surface for (a) Pr ) 1 and (b)
also found that, at Re ) 5, the minimum value of the Pr ) 100 (downstream θ ) 0° and upstream θ ) 180°).
local Nusselt number was located close to the rear of
the cylinder for both Pr ) 1 and Pr ) 100 (Figure 3). cylinder length with the presence of wall effects, a
However, an increase in both the Reynolds and Prandtl voidage ratio of 0.99, etc. Probably because of these
numbers up to Re ) 40 and Pr ) 100 resulted in an factors, the local Nusselt number curves obtained in the
upstream shift in the position of this minimum, which present study for Pr ) 1 differed in shape from those of
also became more pronounced (Figures 3 and 4). At the the aforementioned studies, although some general
point at which Nu(θ) was at its maximum for highly features present therein4,36 and shown in Figure 8 of
shear-thinning fluids, the dependence of Nu(θ) on the ref 36, e.g., maximum and minimum Nusselt number
power-law index was also strongest. Conversely, at the values close to the front and rear ends of the cylinder,
point where Nu(θ) was at its minimum, this dependence respectively, upstream shift in position of the minimum
was quite weak (Figures 3 and 4). Unfortunately, there with increasing Reynolds number, and increase in local
are no results from the literature with which to compare Nusselt number with Reynolds number, were also
the dependence of the local Nusselt number on the observed in the present investigation of the variation
power-law index, and even for the Newtonian case, it of local Nusselt number across the cylinder surface.
is difficult to establish a meaningful comparison be- However, the extent of this variation is determined by
tween the present theoretical results obtained at Pr ) a complex interplay between the kinematic (Re and Pr)
1 using a blockage ratio of β ) 0.037 and the previous and physical characteristics (β and n) of the system, and
experimental (e.g., ref 4) or theoretical (e.g., ref 36) to better understand the effect of these characteristics
results obtained using a different Prandtl number (Pr on heat transfer, it is both useful and convenient to use
) 0.74) and other different conditions such as finite the mean value of the Nusselt number (averaged over
5822 Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 5. Variation of surface-averaged Nusselt number (Nu)


with Re and n, using the constant temperature boundary condition Figure 6. Variation of surface-averaged Nusselt number (Nu)
on the cylinder surface for (a) Pr ) 1 and (b) Pr ) 100. with Re and n, using the constant heat flux boundary condition
on the cylinder surface for (a) Pr ) 1 and (b) Pr ) 100.
the surface), Nu, and, thus, to eliminate one variable
from the process. The results for the surface-averaged 24.2% at Re ) 5 and 26% at Re ) 40 (Figure 5b). For
Nusselt number showed that, from the standpoint of the constant heat flux boundary condition, the corre-
heat transfer, the constant heat flux boundary condition sponding decreases in Nu for Pr ) 1 are 11.4% at Re )
was more efficient than the constant temperature 5 and 20.5% at Re ) 40 (Figure 6a), while for Pr ) 100,
boundary condition (Figures 5 and 6), consistent with they are 23.6% at Re ) 5 and 26.7% at Re ) 40. Since
results already obtained for the local Nusselt number. heat transfer by conduction is independent of fluid
For both the aforementioned boundary conditions, an viscosity, the aforementioned observations showing an
increase in Re and/or Pr, a decrease in the power-law increased dependence of the Nusselt number on the
index, or both caused an increase in the value of Nu power-law index for larger values of Pe are consistent
(Figures 5 and 6). Such an increase in Nu is also with the decrease in the contribution of conduction to
consistent with that already observed for Nu(θ) and can heat transfer, which results from an increase in the
be explained in terms of the increase in the contribution Peclet number. Notwithstanding the importance of the
of convection to the heat transfer, which results from surface-averaged Nusselt number (Nu) as a design
an increase in Peclet number Pe ) Re‚Pr. Likewise, a parameter, to further elucidate the effect of the power-
decrease in the value of n causes a decrease in the law index on the heat transfer process at different
effective viscosity and, therefore, increased convection, Prandtl numbers, the surface-averaged Nusselt number
again resulting in a larger Nu. The present results also for power-law fluids was normalized with respect to the
showed an increase in the dependence of Nu on the corresponding value for Newtonian fluids at the same
power-law index for larger values of Pe (Figures 5 and values of Pr and Re. The results for the normalized sur-
6). For instance, for the constant temperature boundary face-averaged Nusselt number Nu* using both the con-
condition and Pr ) 1, an increase in the level of shear stant temperature and constant heat flux boundary
thickening from n ) 0.5 to n ) 1.4 resulted in a decrease conditions (Figures 7 and 8) paralleled those already
in Nu of 9.3% at Re ) 5 and 18.1% at Re ) 40 (Figure obtained for Nu under the same conditions (Figures 5
5a); whereas for Pr ) 100, the value of Nu decreases by and 6), namely, the increase in both the value of Nu*
Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5823

Figure 8. Effect of the power-law index (n) on the normalized


surface-averaged Nusselt number (Nu*) corresponding to various
Figure 7. Effect of the power-law index (n) on the normalized values of Pr, using the constant heat flux boundary condition on
surface-averaged Nusselt number (Nu*) corresponding to various the cylinder surface for (a) Re ) 5 and (b) Re ) 40.
values of Pr, using the constant temperature boundary condition
on the cylinder surface for (a) Re ) 5 and (b) Re ) 40. of the power-law index on heat transfer. This difficulty,
however, can be obviated by using the Reynolds and
and its dependence on the power-law index with in- Peclet (Pe ) Re × Pr) numbers instead of Re and Pr.
creased Pe and the increase in Nu* which resulted from Figures 7 and 8 relate to the fixed values of Re and show
an increase in the level of shear thinning. The depen- the effect of increasing Prandtl number. It can readily
dence of Nu* on Pr was highly nonlinear. For instance, be seen from the definition of Pe that the increasing Pr
from Pr ) 1 to Pr ) 50, there was a strong variation of implies increasing Pe, and for fixed values of F, K, and
Nu* with Pr away from the Newtonian limit, whereas a, the only way to achieve this is by increasing the
from Pr ) 50 to Pr ) 100, Nu* became almost indepen- velocity. This in turn leads to enhanced levels of
dent of Pr at Re ) 5 (Figures 7a and 8a); it became even shearing (≈ v/2a). For n < 1, this lowers the effective
more independent at Re ) 40 where, from Pr ) 50 to viscosity, which facilitates heat transfer, and of course,
Pr ) 100, the slight variation in Nu* was only noticeable the reverse is true for n > 1. Finally, the use of Nu* as
under highly shear-thickening conditions (see Figures a benchmarking criteria to study the effect of non-
7b and 8b). Figures 7 and 8 further show that, for n < Newtonian behavior on heat transfer characteristics
1, Nu* was an increasing function of Pr, whereas for n showed that the choice of the type of thermal boundary
> 1, Nu* was a decreasing function of Pr, i.e., for n < 1, condition (constant temperature or constant heat flux)
the rate of increase of Nu with Pr was larger than that did not produce a significant impact on heat transfer
for Newtonian fluids, whereas for n > 1, this rate of in power-law fluids (Figures 7 and 8).
increase was smaller than that for Newtonian fluids. Representative plots of the isotherms for both the
The trends seen in Figures 7 and 8 can qualitatively be constant temperature (top half) and the constant heat
explained as follows: since both the Reynolds and flux (bottom half) boundary conditions (Figures 9 and
Prandtl numbers are functions of the kinematic and 10) showed that an increase in the Prandtl and/or
rheological properties due to the power-law viscosity, Reynolds numbers increased the asymmetry and com-
defined as K(v/2a)n-1, it is not possible to isolate the role plexity of the aft contours relative to the fore contours
5824 Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

Figure 9. Effects of n and Pr on the isothermal lines for Re ) 5, using the constant temperature (top) and the constant heat flux (bottom)
boundary conditions on the cylinder surface (flow direction ) left to right).

and also compacted the isothermal lines, which resulted viscosity and the inertial terms. Hence, the assumption
in an increased temperature gradient overall. Moreover, of temperature-independent physical properties affords
since the increase in the Prandtl and/or Reynolds a great simplification, as it does allow the momentum
numbers increased the compactness of the isothermals and energy equations to be decoupled. Clearly, in the
toward the upstream direction (Figures 9 and 10), it also absence of this assumption, this decoupling is not
resulted in a more pronounced upstream increase in the possible. Besides, it should be borne in mind that these
local temperature gradient along the cylinder surface, are the very first set of results on this topic, and
consistent with the corresponding variation in Nusselt naturally, these will get superseded by more realistic
number (Figures 3 and 4). The effects of the power-law fluid models and/or by considering temperature-depend-
index on the temperature profiles also became more ent properties and/or by including viscous dissipation
pronounced at higher Prandtl and/or Reynolds numbers effects in future studies. It is suggested that, in the first
(Figures 9 and 10). For both the constant temperature instance, the effect of the temperature-dependent vis-
and constant heat flux boundary conditions, increased
cosity could be incorporated by using the same correc-
shear thinning resulted in an overall increase in the
tion as that used for Newtonian fluids. Second, admit-
compactness of the isothermals (Figures 9 and 10),
tedly most non-Newtonian materials exhibit much more
which increased the temperature gradient overall. Both
the present results for a circular cylinder and those of complex rheological behavior (especially viscoelasticity)
Gupta et al.34 for the square cylinder showing an than that captured by the simple power-law fluid model
increase in the overall temperature gradient with shear used herein. However, viscoelastic effects are generally
thinning are in line with the previously predicted42,43 significant in transient flows and/or in severely confined
decrease in boundary layer thickness with increased flows where extensional deformation can occur. In most
levels of shear thinning, which has recently been other situations, viscous effects dominate the macro-
confirmed experimentally for a circular cylinder.25 Be- scopic flow phenomena such as drag and heat transfer,
fore closing the discussion, it is appropriate to mention e.g., see refs 17, 18, and 44. For instance, for flow around
the two main limitations underlying the results which a sphere, when both shear thinning and viscoelasticity
have been presented in this study. First, it needs to be are present, shear-thinning behavior dominates the
emphasized here that the governing equations are drag and convective heat- and mass-transfer behav-
highly nonlinear due to both the shear dependent ior.26,45 Therefore, it seems reasonable to start with the
Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5825

Figure 10. Effects of the power-law index (n) and Prandtl number (Pr) on the isothermal lines for Re ) 40, using the constant temperature
(top) and the constant heat flux (bottom) boundary conditions on the cylinder surface (flow direction ) left to right).

simplest type of non-Newtonian behavior and to build numbers such dependence was found to be less pro-
up the level of complexity gradually. nounced. The effects on isothermal patterns, which
result from an increase in Prandtl number and/or
5. Conclusions Reynolds number, are qualitatively similar to those
which result from a decrease in the power-law index.
The non-Newtonian flow across a heated circular The faster decay in the temperature field for higher
cylinder was investigated numerically to determine the Prandtl numbers and a lower power-law index suggests
hydrodynamic drag components and heat transfer char- a decrease of the thermal-boundary-layer thickness
acteristics for a wide range of power-law indices (0.5 e under the aforementioned conditions.
n e 1.4), Reynolds numbers (5 e Re e 40), and Prandtl
numbers (1 e Pr e 100). The study of flow parameters
Nomenclature
showed that the frictional drag component increased
with the flow-behavior index n and that the dependence a ) radius of the cylinder (m)
of this component on n increased with Re. For Re e 15, CD ) drag coefficient
increased shear thinning increased the total drag, CDF ) friction drag coefficient
whereas for Re g 15, it decreased the total drag. The CDP ) pressure drag coefficient
study of heat transfer showed that the constant heat Cp ) heat capacity (J/kg/K)
flux boundary condition was more efficient than the e ) voidage of cylinder assemblage
constant temperature condition. For the latter condition, F ) dimensionless function, eq 10e
the effect of the power-law index on the local Nusselt h ) heat transfer coefficient (W/m2 K)
number Nu(θ) was less pronounced than that for the H) distance from the external boundary to the cylinder
constant heat flux condition. A maximum value of Nu(θ) surface (m)
was located close to the front stagnation point, and its I2 ) dimensionless second invariant of the rate-of-deforma-
peak value was found to increase with shear thinning. tion tensor
At the location where the maximum peaked, the depen- hI2 ) dimensionless second invariant of the rate-of-deforma-
dence of Nu(θ) on n was found to be strongest. The tion tensor in real space
surface-averaged Nusselt number Nu was found to be K ) power-law consistency index (Pa sn)
a decreasing function of n, although for lower Peclet k ) thermal conductivity of the fluid (W/m K)
5826 Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005

L ) wake length (m) (10) Ahmad, R. A. Steady-state numerical solutions of the


n ) power-law index Navier-Stokes and energy equations around a horizontal cylinder
Nu ) average Nusselt number at moderate Reynolds numbers from 100 to 500. Heat Transfer
Eng. 1996, 17, 31-81.
Nu* ) average Nusselt number normalized with the
(11) Zdravkovich, M. M. Flow around circular cylinders: Fun-
corresponding value for the Newtonian fluid damentals; Oxford University Press: Oxford, U.K., 1997; Vol. 1.
Nu(θ) ) local Nusselt number (12) Zdravkovich, M. M. Flow around circular cylinders: Ap-
p ) dimensionless pressure plications; Oxford University Press: Oxford, U.K., 2003; Vol. 2.
p0 ) dimensionless stagnation pressure (13) Carreau, P. J.; De Kee, D. C. R.; Chhabra, R. P. Rheology
Pe ) Peclet number of polymeric systems; Hanser-Gardner: Munich, Germany, 1997.
Pr ) Prandtl number (14) Chhabra, R. P.; Richardson, J. F. Non-Newtonian flow in
r ) cylindrical coordinate (m) the process industries; Butterworth-Heinemann: Boston, MA,
rj ) dimensionless radial component in real space 1999.
qs ) heat flux at the surface of the cylinder (W/m2) (15) White, J. L.; Metzner, A. B. Constitutive equations for
viscoelastic fluids with applications to rapid external flows. AIChE
Re ) Reynolds number ()F(2a)nU∞2-n/K)
J. 1965, 11, 324-330.
U∞ ) uniform approach velocity (m/s) (16) Ruckenstein, E.; Ramagopal, A. Anomalous heat transfer
T ) dimensionless temperature and drag in laminar flow of viscoelastic fluids. J. Non-Newtonian
h ) dimensionless temperature in real space
T Fluid Mech. 1985, 17, 145-155.
Ts ) temperature on the surface of the cylinder (K) (17) Ng, M. L.; Hartnett, J. P. Natural convection in power law
T∞ ) free stream fluid temperature (K) fluids. Int. Commun. Heat Mass Transfer 1986, 13, 115-120.
vz ) z-component of velocity (m/s) (18) Ng, M. L.; Hartnett, J. P. Free convection heat transfer
from horizontal wires to pseudoplastic fluids. Int. J. Heat Mass
Greek Letters Transfer 1988, 31, 441-447.
γ ) dimensionless function, eq 10d (19) Tanner, R. I. Stokes paradox for power-law flow around a
cylinder. J. Non-Newtonian Fluid Mech. 1993, 50, 217-224.
θ ) angle (radians)
(20) Marǔsić-Paloka, E. On the Stokes paradox for power law
θs ) separation angle (degrees) fluids. Z. Angew. Math. Mech. 2001, 81, 31-36.
λ ) dimensionless function, eq 10b (21) Whitney, M. J.; Rodin, G. J. Force-velocity relationships
µ ) dimensionless function, eq 10c for rigid bodies translating through unbounded shear-thinning
ψ ) dimensionless stream function power-law fluids. Int. J. Nonlinear Mech. 2001, 36, 947-953.
h ) dimensionless stream function in real space
ψ (22) D’Alessio, S. J. D.; Pascal, J. P. Steady flow of a power-
η ) dimensionless viscosity law fluid past a cylinder. Acta Mech. 1996, 117, 87-100.
j ) dimensionless viscosity in real space
η (23) Chhabra, R. P.; Soares, A. A.; Ferreira, J. M. Steady non-
ω ) dimensionless vorticity Newtonian flow past a circular cylinder: a numerical study. Acta
Mech. 2004, 172, 1-16.
j ) dimensionless vorticity in real space
ω
(24) Coelho, P. M.; Pinho, F. T. Vortex shedding in cylinder flow
F ) fluid density (kg/m3) of shear-thinning fluids. I. Identification and demarcation of flow
 ) dimensionless polar coordinate ()ln(r/a)) regimes. J. Non-Newtonian Fluid Mech. 2003, 110, 143-176.
ij ) components of rate-of-deformation tensor (s-1) (25) Coelho, P. M.; Pinho, F. T. Vortex shedding in cylinder flow
τij ) dimensionless components of extra stress tensor of shear-thinning fluids. II. Flow characteristics. J. Non-Newtonian
Fluid Mech. 2003, 110, 177-193.
Subscripts (26) Chhabra, R. P. Heat and mass transfer in rheologically
θ ) angular component complex systems. In Advances in the rheology and flow of non-
r ) radial component Newtonian fluids; Siginer, D., DeKee, D., Chhabra, R. P., Eds.;
Elsevier: Amsterdam, The Netherlands, 1999; Chapter 39.
(27) Mizushina, T.; Usui, H.; Veno, K.; Kato, T. Experiments
Literature Cited of pseudoplastic fluid crossflow around a circular cylinder. Heat
TransfersJpn. Res. 1978, 7, 92-101.
(1) Tritton, D. J. Experiments on the flow past a circular
(28) Ghosh, U.K.; Gupta, S. N.; Kumar, S.; Upadhyay, S. N.
cylinder at low Reynolds numbers. J. Fluid Mech. 1959, 6, 547-
Mass transfer in crossflow of non-Newtonian fluid around a
567.
circular cylinder. Int. J. Heat Mass Transfer 1986, 29, 955-960.
(2) Dennis, S. C. R.; Chang, G. Z. Numerical solutions for steady
flow past a circular cylinder at Reynolds numbers up to 100. J. (29) Hoyt, J. W.; Sellin, R. H. J. Cylinder cross-flow heat
Fluid Mech. 1970, 42, 471-489. transfer in drag reducing fluid. Exp. Heat Transfer 1989, 2, 113-
127.
(3) Fornberg, B. A numerical study of steady viscous flow past
a circular cylinder. J. Fluid Mech. 1980, 98, 819-855. (30) Rao, B. K.; Phillips, B. J.; Andrews, J. Heat transfer to
(4) Eckert, E. R. G.; Soehngen, E. Distribution of heat transfer viscoelastic polymer solutions flowing over a smooth cylinder. Appl.
coefficients around circular cylinders in crossflow at Reynolds Mech. Eng. 1996, 1, 355-365.
numbers from 20 to 500. Trans. ASME 1952, 74, 343-347. (31) Rao, B. K. Heat transfer to non-Newtonian flows over a
(5) Bassam, A.; Abu-Hijleh, K. Numerical simulation of forced cylinder in crossflow. Int. J. Heat Fluid Flow 2000, 21, 693-700.
convection heat transfer from a cylinder with high conductivity (32) Rao, B. K. Heat transfer to two-phase air-viscoelastic fluid
radial fins in crossflow. Int. J. Therm. Sci. 2003, 42, 741-748. flows over a hot cylinder. Exp. Heat Transfer 2003, 16, 227-238.
(6) Lange, C. F.; Durst, F.; Breuer, M. Momentum and heat (33) Ghosh, U. K.; Upadhyay, S. N.; Chhabra, R. P. Heat and
transfer from cylinders in laminar crossflow at 10-4 e Re e 200. mass transfer from immersed bodies to non-Newtonian fluids. Adv.
Int. J. Heat Mass Transfer 1998, 41, 3409-3430. Heat Transfer 1994, 25, 251-319.
(7) Juncu, G. Unsteady conjugate heat/mass transfer from a (34) Gupta, A. K.; Sharma, A.; Chhabra, R. P.; Eswaran, V.
circular cylinder in laminar crossflow at low Reynolds numbers. Two-dimensional steady flow of a power-law fluid past a square
Int. J. Heat Mass Transfer 2004, 41, 2469-2480. cylinder in a plane channel: momentum and heat-transfer char-
(8) Takami, H.; Keller, H. B. Steady two-dimensional viscous acteristics. Ind. Eng. Chem. Res. 2003, 42, 5674-5686.
flow of an incompressible fluid past a circular cylinder. In High- (35) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport
Speed Computing in Fluid DynamicssThe Physics of Fluids phenomena, 2nd ed.; Wiley: New York, 2001.
Supplement II, Proceedings of International Symposium of the (36) Mandhani, V. K.; Chhabra, R. P.; Eswaran, V. Forced
International Union of Theoretical and Applied Mechanics, 1968; convection heat transfer in tube banks in crossflow. Chem. Eng.
Frenkiel, F. N., Stewartson, K., Eds.; American Institute of Sci. 2002, 57, 379-391.
Physics: New York, 1969; pp 51-56. (37) Mangadoddy, N.; Prakash, R.; Chhabra, R. P.; Eswaran,
(9) Morgan, V. T. The overall convective heat transfer from V. Forced convection in crossflow of power law fluids over a tube
smooth circular cylinders. Adv. Heat Transfer 1975, 11, 199-264. bank. Chem. Eng. Sci. 2004, 59, 2213-2222.
Ind. Eng. Chem. Res., Vol. 44, No. 15, 2005 5827

(38) Vijaysri, M.; Chhabra, R. P.; Eswaran, V. Power-law fluid (43) Skelland, A. H. P. Non-Newtonian flow and heat transfer.
flow across an array of infinite circular cylinders: a numerical Wiley: New York, 1967.
study. J. Non-Newtonian Fluid Mech. 1999, 87, 263-282. (44) Chhabra, R. P. Bubbles, drops and particles in non-
(39) Ferreira, J. M.; Chhabra, R. P. Analytical study of drag Newtonian fluids. CRC Press: Boca Raton, Fl, 1993.
and mass transfer in creeping power law flow across tube banks. (45) Westerberg, K. W.; Finlayson, B. A. Heat transfer to
Ind. Eng. Chem. Res. 2004, 43, 3439-3450. spheres from a polymer melt. Numer. Heat Transfer 1990, 17A,
(40) Tripathi, A.; Chhabra, R. P.; Sundararajan, T. Power law 329-345.
fluid flow over spheroidal particles. Ind. Eng. Chem. Res. 1994,
33, 403-410. Received for review January 18, 2005
(41) Tripathi, A.; Chhabra, R. P. Drag on spheroidal particles Revised manuscript received April 29, 2005
in dilatant fluids. AIChE J. 1995, 41, 728-731. Accepted May 10, 2005
(42) Serth, R. W.; Kiser, K. M. A solution of the 2-dimensional
boundary-layer equations for an Ostwald-de Waele fluid. Chem.
Eng. Sci. 1967, 22, 945-956. IE0500669

You might also like