Vous êtes sur la page 1sur 20

J. Non-Newtonian Fluid Mech.

165 (2010) 14421461

Contents lists available at ScienceDirect

Journal of Non-Newtonian Fluid Mechanics


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

Laminar ow of power-law uids past a rotating cylinder


Saroj K. Panda, 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

a b s t r a c t
In this work, the continuity and momentum equations have been solved numerically to investigate the ow of power-law uids over a rotating cylinder. In particular, consideration has been given to the prediction of drag and lift coefcients as functions of the pertinent governing dimensionless parameters, namely, power-law index (1 n 0.2), dimensionless rotational velocity (0 6) and the Reynolds number (0.1 Re 40). Over the range of Reynolds number, the ow is known to be steady. Detailed streamline and vorticity contours adjacent to the rotating cylinder and surface pressure proles provide further insights into the nature of ow. Finally, the paper is concluded by comparing the present numerical results with the scant experimental data on velocity proles in the vicinity of a rotating cylinder available in the literature. The correspondence is seen to be excellent for Newtonian and inelastic uids. 2010 Elsevier B.V. All rights reserved.

Article history: Received 9 May 2010 Received in revised form 17 July 2010 Accepted 20 July 2010

Keywords: Rotating cylinder Power-law uid Drag Lift Reynolds number

1. Introduction Flow past a cylinder constitutes a classical problem in the domain of uid mechanics. The interest in this ow conguration stems from both fundamental and pragmatic considerations. For instance, a circular cylinder denotes the simplest bluff body shape which is free from geometrical singularities and it has indeed provided valuable insights into the underlying physics of the ow including the wake phenomena, vortex shedding, drag and lift characteristics. On the other hand, the ow past a cylinder also represents idealization of several industrially important applications such as the ow in tubular and pin-type heat exchangers, ltration screens, membrane based separation modules, etc. Additional examples are found in the use of obstacles of various shapes to form weld-lines in polymer processing applications, in the resin transfer process of producing ber-reinforced composites, and as a model for lungs. The imposition of rotation is used to delay and/or suppress the propensity for vortex shedding thereby extending the steady ow regime. Indeed, the ow over a cylinder is inuenced by a large number of parameters such as the nature of the far ow eld (uniform, or shear, or extensional), conned or unconned cylinder, type of uid (compressible, or incompressible, or nonNewtonian), stationary or rotating cylinder, etc. For the simplest case of the incompressible uniform ow of Newtonian uids over an unconned stationary cylinder, the ow undergoes several transitions with a gradual increase in the value of the Reynolds number. Thus, for instance, the ow remains attached to the surface of the

Corresponding author. Tel.: +91 512 259 7393; fax: +91 512 259 0104. E-mail address: chhabra@iitk.ac.in (R.P. Chhabra). 0377-0257/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.jnnfm.2010.07.006

cylinder up to about the Reynolds number value of 45. As the value of the Reynolds number is gradually incremented, the occurrence of an adverse pressure gradient at some point on the surface of the cylinder results in the separation of ow from the surface thereby leading to the formation of the so-called wake region. This region is characterized by the loss of fore and aft symmetry of the ow eld, albeit the ow is still steady and two-dimensional. The attached twin-vortices grow in size with further increase in the value of the Reynolds number and the ow continues to be symmetric about mid-plane. At about Re = 4647, the wake becomes asymmetric and vortices begin to break off alternately from the upper and lower halves of the cylinder respectively thereby resulting in the so-called laminar vortex shedding regime. Under these conditions, the ow is still two-dimensional, but no longer steady and one must seek solutions to the time-dependent equations. This ow regime occurs up to about the Reynolds number of about 165, at which the wake itself becomes turbulent and early signatures of 3D ow also begin to manifest. Evidently, the underlying changes in the kinematics of the ow associated with each ow regime also impact the macroscopic ow characteristics like the force coefcients (drag and lift), Strouhal number and Nusselt number, etc., especially in the way these scale with the Reynolds number. Due to concentrated research efforts expended in elucidating the aforementioned as well as many other associated aspects, a wealth of information has accrued in the literature as far as the ow of Newtonian uids past a stationary cylinder is concerned and this body of information has been reviewed in several excellent sources [15]. Similarly the ow imposed over a rotating cylinder is not only another fundamental ow of the bluff body family, but the rotation of cylinder is also used to increase lift, to stabilize ow and/or to control boundary layers on aerofoils. For Newtonian uids, this ow con-

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1443

Nomenclature CD CL CDN CLN CDP CDF CLP CLF D De FD FL H I2 L m n Ni p R Re Ux Uy U0 x* y* Drag coefcient Lift coefcient Total drag coefcient in a Newtonian uid Total lift coefcient in a Newtonian uid Pressure component of drag coefcient Frictional component of drag coefcient Pressure component of lift coefcient Frictional component of lift coefcient Diameter of the cylinder (m) Deborah number Drag force per unit length of the cylinder (N/m) Lift force per unit length of the cylinder (N/m) Height (and width) of the square domain (m) Second invariant of the rate of deformation tensor (s2 ) Length of the cylinder (m) Power-law consistency index (Pa sn ) Power-law index Number of points on the surface of the cylinder Pressure (Pa) Radius of the cylinder (m) Reynolds number x-Component of velocity (m/s) y-Component of velocity (m/s) Uniform velocity of the uid at the inlet (m/s) Stream wise co-ordinate, x* = x/R Transverse co-ordinate, y* = y/R

Greek letters Rotational velocity Angular velocity of the cylinder (rad/s) Density of uid (kg/m3 ) Extra stress tensor (Pa) ij Viscosity (Pa s) ij Component of the rate of strain tensor (s1 ) Angular displacement from the front stagnation point Relaxation time (s)

guration has also been explored, though the resulting literature is neither as extensive nor as coherent as that for a stationary cylinder. Much of the pertinent literature has been reviewed in [16]. It is thus fair to say that an adequate body of information is available on the ow of Newtonian uids past a stationary and a rotating cylinder. On the other hand, it is readily acknowledged that many uids of industrial signicance exhibit non-Newtonian ow characteristics including shear-dependent viscosity (shear-thinning), yield stress, visco-elasticity, etc. Typical examples include polymer melts and solutions, foams, emulsions and suspensions, etc. [7]. The ow of non-Newtonian uids over a rotating cylinder denotes an idealization of some engineering applications encountered in coating operations, rotary drum lters used for non-Newtonian slurries, roller bearing applications, in oil drilling operations, in mixing vessels with novel impeller designs, etc. In spite of their fundamental and pragmatic signicance, very little is known about the uid mechanical aspects of the non-Newtonian uid ow over a rotating cylinder and this work endeavours to ll this gap in the current literature. However, in order to facilitate the presentation of the new results obtained in this study, it is instructive to briey review the body of knowledge available on the ow of non-Newtonian (especially power-law uids) past a sta-

Fig. 1. (a) Schematic representation of the physical model, (b) computational domain, and (c) close up view of the grid in the vicinity of cylinder.

tionary cylinder and that of Newtonian uids past a rotating cylinder. 2. Previous work As noted earlier, since a succinct summary of the pertinent literature on the ow of Newtonian uids over a rotating cylinder is available in numerous references, e.g., see [6,8], only the salient features are recounted here. For the simplest case of an unconned cylinder, the ow is now governed by two dynamic parameters, namely, Reynolds number and dimensionless rotational velocity, , of the cylinder. While many studies are available at moderate to high Reynolds numbers at which the ow eld is three-dimensional (e.g., see [912]), since the present work is concerned with the steady ow regime, the ensuing discussion is limited to the steady ow regime. At Re = 4647, the ow becomes unsteady even for a

1444 Table 1 Selection of optimum domain. H/D Re = 0.1, n = 1, = 6 120 220 240 Re = 0.1, n = 0.2, = 6 120 220 240 Re = 5, n = 1, = 6 100 140 160 Re = 5, n = 0.2, = 6 100 140 160 CDP 31.404 29.290 29.078 132.154 132.050 132.049 3.390 3.389 3.389 5.089 5.088 5.089

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

CDF 32.805 30.600 30.371 16.382 16.370 16.371 3.150 3.149 3.149 0.185 0.185 0.185

CD 64.209 59.890 59.450 148.536 148.42 148.42 0.240 0.239 0.240 5.274 5.273 5.274

CLP 13.230 12.304 12.218 22.580 22.427 22.427 20.955 20.953 20.953 7.893 7.892 7.892

CLF 8.469 7.860 7.806 0.141 0.155 0.154 4.575 4.574 4.574 0.043 0.043 0.043

CL 21.698 20.164 20.024 22.438 22.271 22.273 25.530 25.527 25.527 7.936 7.935 7.935

stationary cylinder, i.e., = 0. Broadly speaking, small values of the rotational velocity do not seem to inuence this transition. However, once the value of exceeds a critical value (1 ) for Re > 4647, the tendency for vortex shedding is somewhat suppressed [13]. However, subsequent more detailed studies [6,8] also point to the occurrence of two more transitions. At high Reynolds numbers and for > 2 , the ow again becomes unsteady and nally, again the ow reverts to a steady ow regime at > 3 at Re = 200, albeit the second transition 2 3 occurs over a rather narrow range of conditions. As expected, for a xed value of the Reynolds number, drag and lift coefcients decrease with the increasing value of [14] and so does the rate of heat transfer [6,15] in the range 5 Re 166 and 4. In contrast to the aforementioned studies based on the solution of the complete governing equations, Kendoush [16] has employed the boundary layer approximation to predict the value of Nusselt number for a rotating cylinder. All in all, the ow is known to be steady up to Re 40 and 6 over a rotating cylinder [6]. In contrast, the literature on the ow of power-law uids over a stationary cylinder is not only of recent vintage, but is also much less extensive than that for Newtonian uids. Most of these studies have been reviewed recently [17,18]. Broadly, reliable results are now available on the momentum transfer characteristics for a nonrotating cylinder in unconned power-law uids in the steady ow regime [1923] for both shear-thinning and shear-thickening uids. At low Reynolds numbers, shear-thinning behaviour enhances drag above its Newtonian value and as expected, shear-thickening

Table 2 Details of grids used for grid independence study. Grid H/D = 220 G1 G2 G3 H/D = 100 G1 G2 G3 Ni 100 200 400 100 200 400 /D 0.005 0.005 0.005 0.005 0.005 0.005 Ncells 117,600 151,100 218,100 86,496 151,100 178,896

has the opposite effect. However, it is appropriate to add here that whether the drag is reduced or augmented with respect to the Newtonian value is also linked to the choice of the characteristic viscosity used to dene the Reynolds number [24]. Furthermore, the role of power-law index gradually diminishes with the increasing value of the Reynolds number. Similarly, heat transfer is facilitated in shear-thinning uids both in forced and mixed convection regimes [2527]. Qualitatively, similar trends are observed for a cylinder conned symmetrically in between two parallel walls [28,29]. Finally, there has been only one set of study on the ow and heat transfer phenomena from a cylinder submerged in powerlaw uids in the laminar vortex shedding regime [17,18]. For the sake of completeness, it is worthwhile to add here that scant results are also available for square [30,31] and elliptic cylinders [32,33]

Table 3 Effect of grid details on the results at Re = 0.1 (H/D = 220) and Re = 5 (H/D = 100). Grid Re = 0.1, n = 1, = 6 G1 G2 G3 Re = 0.1, n = 0.2, = 6 G1 G2 G3 Re = 5, n = 1, = 6 G1 G2 G3 Re = 5, n = 0.2, = 6 G1 G2 G3 CDP 29.318 29.290 29.289 132.130 132.050 132.048 3.380 3.391 3.392 5.095 5.090 5.082 CDF 30.618 30.600 30.586 16.380 16.370 16.366 3.141 3.150 3.151 0.1850 0.1855 0.1858 CD 59.936 59.890 59.875 148.510 148.420 148.414 0.232 0.240 0.241 5.280 5.275 5.266 CLP 12.37 12.304 12.285 22.491 22.427 22.416 20.920 20.959 20.960 7.892 7.893 7.893 CLF 7.907 7.86 7.845 0.153 0.155 0.156 4.560 4.575 4.577 0.046 0.044 0.044 CL 20.276 20.164 20.13 22.338 22.271 22.261 25.480 25.531 25.537 7.930 7.936 7.937

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461 Table 4 Comparison between the present and literature results for a rotating cylinder in Newtonian uids. Source Present Stojkovic et al. [8] Badr et al. [15] Present Stojkovic et al. [8] Present Paramane and Sharma [6] Stojkovic et al. [8] Badr et al. [15] Present Paramane and Sharma [6] Stojkovic et al. [8] Present Paramane and Sharma [6] Present Paramane and Sharma [6] Re CDP =1 1.9531 =6 =1 1.01 1 =6 2.734 2.791 =1 0.7774 0.775 =6 2.0841 1.9 CDF 1.9043 0.8312 0.8368 2.4227 2.4467 0.5458 0.54 2.0661 2.117 CD 3.8574 3.8013 3.81 1.8412 1.8368 1.91 0.3112 0.3443 1.3233 1.315 0.018 0.217 CLP 2.315 2.29 20.9557 2.387 2.3634 2.424 2.44 28.86 28.8764 28.86 2.3345 2.3405 32.894 32.9103 CLF 0.6076 0.55 4.575 0.3614 0.357 0.3501 0.35 2.9412 2.9314 2.9461 0.2645 0.2608 2.2621 2.248 CL

1445

2.922 2.9022 2.84 25.5308 25.4764 2.7483 2.7204 2.7741 2.79 31.8003 31.8078 31.806 2.599 2.6013 35.1562 35.1583

20

40

submerged in conned and unconned power-law uids, most of which are based on the assumption of the steady ow regime except for the recent results reported in [34] on the ow over a square cylinder in the laminar vortex shedding regime. On the other hand, a vast literature exists on the ow of visco-elastic uids past a non-rotating circular cylinder [24]; however, most of it relates to the creeping (zero Reynolds number) ow regime and endeavours to elucidate the role of uid visco-elasticity on the uid mechanical aspects in the absence of shear-thinning behaviour. While most experimental studies report signicant changes in the detailed ow eld which are attributed to visco-elasticity, but the numerical simulations predict only minor changes. At low Reynolds numbers, visco-elasticity reduces the drag on a cylinder below its Newtonian value and it increases the drag at high Reynolds numbers. Therefore, the literature is inundated with conicting reports regarding the role of elasticity in this case. As far as known to us, there have been only two studies on the ow of non-Newtonian uids past a rotating cylinder. Townsend [35] numerically studied the uniform ow of an Oldroyd model uid past a rotating cylinder over the range of conditions as 5 and 10 Re 60. His results suggest that the visco-elasticity tends to increase both the drag and lift experienced by a rotating cylinder whereas the shear-thinning behaviour tends to reduce their values. On the other hand, Christiansen [36,37] has reported detailed experimental data on the local velocity eld for

a cylinder rotating in Newtonian and polymer solutions of varying levels of shear-thinning and visco-elasticity. It is thus safe to conclude that very little is known about the ow of non-Newtonian uids past a rotating cylinder. Admittedly, complex uids exhibit a range of rheological features such as shearthinning, visco-elasticity and yield stress, it seems reasonable to begin with the simplest type of non-Newtonian behaviour, namely, shear-thinning viscosity and the level of complexity can gradually be built up to incorporate other aspects of real uid behaviour in a systematic manner. This work is thus concerned with the ow of power-law uids past a rotating cylinder over the range of conditions Re 40, 6 and 0.2 n 1 over which the ow is expected to be steady and two-dimensional. In particular, extensive numerical results are presented elucidating the effect of rheological and dynamic parameters on the detailed kinematics of ow and on drag and lift coefcients. Finally, the present predictions are compared with the experimental results available in the literature [36,37]. 3. Problem statement and governing equations Let us consider the incompressible steady ow of a powerlaw uid (with uniform velocity U0 ) past a cylinder of radius R or diameter D (innitely long in z-direction) rotating with an angular velocity of in the anti-clockwise direction, as shown schematically in Fig. 1(a). Since it is not possible to simulate numerically truly an unconned ow, it is customary to introduce an articial domain in the form of a box as shown schematically. Following the previous studies [6], the cylinder is placed at the center of a square of size H, as shown in Fig. 1(b). The value of H is chosen in such a manner that it does not unduly inuence the ow eld and at the same time it necessitates only modest computational resources. The ow phenomenon is described by the continuity and momentum equations written in their compact forms as follows: Continuity:

Table 5 Comparison between the present and literature values for a stationary cylinder in shear-thinning uids (n = 0.2). Source Re =0 CDP Present Sivakumar et al. [41] Present Sivakumar et al. [41] Present Sivakumar et al. [41] Present Sivakumar et al. [41] 0.1 1 10 40 201.0391 195.1465 20.0084 19.6615 2.4251 2.4388 0.9915 0.9954 CDF 69.5039 72.3863 7.0551 7.2447 0.7318 0.7178 0.1450 0.1443 CD 270.5431 267.5328 27.0636 26.9062 3.1570 3.1566 1.1365 1.1397

.V = 0
Momentum: DV = p + Dt + g

(1)

(2)

1446

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 2. Streamline contours near the cylinder for n = 0.2 for = 1 and 6.

where , V, p, and g, respectively, are the uid density, velocity vector, isotropic pressure, extra stress tensor and gravity. For power-law uids, the components of the extra stress tensor ij are related to the rate of deformation tensor as

where the rate of deformation tensor is given by ij = 1 2 Uj Ui + xj xi (4) , for a power-law uid is given (5)

And the scalar viscosity function, by


ij

= 2 ij

(3)

= m(2I2 )(n1)/2

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1447

Fig. 3. Streamline contours near the cylinder for n = 0.6 for = 1 and 6.

where n is the so-called power-law index. Evidently, n = 1 corresponds to the Newtonian uid behaviour and n < 1 denotes shear-thinning behaviour. It is customary to introduce dimensionless variables. In this work, the free stream velocity, U0 , and diameter of cylinder, D, have been used as scaling variables. Thus, the pressure has been scaled 2 using U0 , stress components using m(U0 /D)n , time with D/U0 , etc.

The physically realistic boundary conditions for this ow are written as follows: (1) At the inlet plane: It is the uniform ow in x-direction, i.e., Ux = U0 , Uy = 0

1448

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 4. Streamline contours near the cylinder for n = 1 for = 1 and 6.

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1449

Fig. 5. Vorticity contours near the cylinder for n = 0.2 for = 1 and 6.

(2) The top and bottom walls are assumed to be slip boundaries so that there is no dissipation at these walls. The corresponding mathematical description is given by: Ux = 0; y

(3) On the surface of the solid cylinder: The standard no-slip boundary condition is used, i.e., Ux = sin and Uy = cos

Uy = 0

(4) At the exit plane: The default outow boundary condition option in FLUENT (a zero diffusion ux for all ow variables) was used

1450

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 6. Vorticity contours near the cylinder for n = 0.6 for = 1 and 6.

in this work. This choice implies that the conditions of the outow plane are extrapolated from within the domain and as such have negligible inuence on the upstream ow conditions. The extrapolation procedure used by FLUENT updates the outow velocity and the pressure in a manner that is consistent with the fully developed ow assumption, when there is no area change at the outow boundary. However, the gradients in the cross-

stream direction may still exist at the outow boundary. Also, the use of this condition obviates the need to prescribe a boundary condition for pressure. This is similar to the homogeneous Neumann condition, that is, Ux =0 x and Uy =0 x

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1451

Fig. 7. Vorticity contours near the cylinder for n = 1 for = 1 and 6.

The numerical solution of the governing Eqs. (1) and (2) together with the aforementioned boundary conditions maps the ow domain in terms of Ux , Uy and p which, in turn, can be post processed to evaluate the derived quantities such as stream function, vorticity, drag and lift coefcients as functions of the pertinent governing parameters, namely, the Reynolds number (Re), power-law index (n) and the non-dimensional rotational velocity of the cylinder (). At this juncture, it is appropriate to introduce the pertinent denitions of some of these parameters as follows:

Reynolds number (Re)


2n U0 Dn

Re =

Rotational velocity () R U0

1452

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 8. Representative pressure proles on the surface of the rotating cylinder.

Drag coefcient (CD ) 2FD


2 U0 D

CD =

where FD is the drag force in the direction of ow exerted on the cylinder per unit length. It is also customary to split the total drag force into two components arising from the shear and pressure forces thereby giving rise to the corresponding drag coefcient components as CDF and CDP respectively.

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1453

Fig. 9. Dependence of CD , CDP , CL on Reynolds number for = 1 and 6.

1454

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 10. Variation of CD , CDP , CL with for Re = 0.1 and 40.

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461 Table 6 Numerical values of drag and lift coefcients as functions of , Re and n. CD Re =0 n=1 0.1 1 10 40 Re 61.1275 10.5663 2.7983 1.5136 =4 n=1 0.1 1 10 40 CL Re =0 n=1 0.1 1 10 40 Re 0 0 0 0 =4 n=1 0.1 1 10 40 13.4117 12.6919 14.5583 16.1329 n = 0.8 6.3120 8.3219 12.7017 15.2928 n = 0.6 5.6369 6.2170 10.7257 14.1065 n = 0.4 151.086 7.6299 8.6285 12.4331 n = 0.2 14.1945 11.9027 7.5009 10.7610 n = 0.8 0 0 0 0 n = 0.6 0 0 0 0 n = 0.4 0 0 0 0 n = 0.2 0 0 0 0 =2 n=1 6.6968 6.1190 5.9608 5.7205 =6 n=1 20.1638 20.2742 28.6372 35.1562 n = 0.8 9.6371 12.855 22.633 25.557 n = 0.6 8.9071 9.371 15.137 19.400 n = 0.4 12.8874 10.935 10.080 15.069 n = 0.2 n = 0.8 3.0726 4.0993 5.3654 5.3832 n = 0.6 2.6343 3.0404 4.8587 5.0545 n = 0.4 3.9018 3.8614 4.5096 4.7375 60.5715 8.9340 0.5891 0.1440 n = 0.8 84.4540 10.8012 1.3992 0.3397 n = 0.6 119.1578 12.9498 2.2384 0.7686 n = 0.4 151.0860 15.7119 2.9974 1.0976 n = 0.2 169.8143 17.3883 3.3288 1.3608 n = 0.8 100.8522 13.2468 2.8609 1.4378 n = 0.6 164.2520 17.4697 2.9348 1.3455 n = 0.4 231.3182 23.3118 3.0268 1.2393 n = 0.2 270.5431 27.0635 3.1570 1.1365 =2 n=1 60.9866 10.1578 2.1340 0.8453 =6 n=1 59.8808 6.9582 0.7747 0.0180 n = 0.8 80.6602 10.0769 1.5118 0.6089 n = 0.6 108.9539 12.5536 2.8136 0.9679 n = 0.4 133.0632 15.0304 3.3974 1.2528 n = 0.8 89.9372 11.7208 2.2173 0.8517 n = 0.6 133.9509 14.2309 2.3359 0.8592 n = 0.4 176.8087 17.8904 2.5245 0.8798

1455

n = 0.2 199.819 20.2951 2.6436 0.9115

n = 0.2 148.426 16.6408 3.5269 1.3297

n = 0.2 7.2012 6.6750 4.5950 4.4573

22.269 15.432 7.9919 12.260

Lift coefcient (CL ) CL = 2FL


2 U0 D

where FL is the lift force acting in the y-direction on the cylinder per unit length. Here also, it is not uncommon to split the lift force into two components due to pressure and shear as CLP and CLF respectively. The scaling of the governing equations and the boundary conditions suggests the integral ow parameters, CD and CL to be functions of the Reynolds number, power-law index and . This functional relationship is explored and developed in this work. 4. Numerical solution method Since detailed descriptions of the numerical solution procedure are available elsewhere [17,18,28], only the salient features are recapitulated here. In this study, the eld equations have been solved using FLUENT (version 6.2.26). The structured quadrilateral cells of uniform and non-uniform grid spacing were generated using the commercial grid tool GAMBIT (version 2.3.16). The twodimensional, laminar, segregated solver was used to solve the incompressible ow on the collocated grid arrangement. Both steady and unsteady solvers have been used in this study. While the major thrust of this work is on the steady ow regime, limited time-dependent simulations for extreme values of the governing parameters such as Re = 40, n = 0.2 and = 6, etc. were also conducted to ascertain that the ow regime indeed was steady over the range of conditions covered in this study. The second order upwind scheme has been used to discretize the convective terms in the momentum equations. The semi-implicit method

for the pressure linked equations (SIMPLE) scheme was used for solving the pressurevelocity coupling. The constant density and non-Newtonian power-law viscosity modules were used to input the physical properties of the ow. However, the input values of the physical properties , m, n and kinematic parameters like D, U0 , , etc. are of no consequence as the nal results are reported in a dimensionless form. FLUENT solves the system of algebraic equations using the GaussSiedel (GS) point-by-point iterative method in conjunction with the algebraic multi-grid (AMG) method solver. The use of the AMG scheme greatly reduces the number of iterations (thereby accelerating convergence) and thus economizing the CPU time required to obtain a converged solution, particularly when the model contains a large number of control volumes. Relative convergence criteria of 108 for the residuals of the continuity and x- and y-component of the momentum equations were prescribed in this work. Furthermore, a simulation was deemed to have converged when the values of the global parameters had stabilized to at least four signicant digits. 5. Choice of numerical parameters Undoubtedly, the reliability and accuracy of the numerical results is strongly inuenced by the choice of domain size (H), grid characteristics (number of cells on the surface of the cylinder, grid spacing, stretching, etc.) and to some extent by the convergence criterion, etc. In this work, the values of these parameters have been selected after extensive exploration. The values of (H/D) ranging from 80 to 240 have been employed here to arrive at an optimal value of this parameter. Intuitively, it appears that a larger domain is required at low Reynolds numbers than that at high Reynolds numbers. This is so simply due to the slow spatial decay of the velocity eld at low Reynolds numbers. Therefore, the domain independence study has been carried out at the lowest value of Re = 0.1

1456

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 11. Effect of power-law index on the normalized values of CD and CL with for Re = 0.1 and 40.

used in this study. After a detailed examination of the results (drag and lift coefcients), sufce it to say here that, the value of (H/D) used in this work is 220 at low Reynolds numbers (Re < 5), whereas (H/D) = 100 was used for Re 5 for the domain effects to be negligible as can be seen in Table 1. These ndings are consistent with that reported by others, e.g., see [6] over this range of Reynolds numbers. Having xed the value of (H/D), an optimal grid should meet two conicting requirements: namely, it should be sufciently ne to resolve the thin boundary layers and the steep gradients near the rotating cylinder, and of course without being prohibitively computation intensive. For this purpose, the relative performance of three grids has been studied in detail to arrive at the choice of an optimal grid. Each grid was characterized in terms of the number of points (Ni ) on the surface of the cylinder and the value of (/D) near the cylinder, as summarized in Table 2. A typical grid is shown in

Fig. 1(c). For the same combinations of the values of Re, n and as that used in the domain independence study, Table 3 summarizes the relative performance of the three grids in terms of the resulting values of the individual components of drag and lift coefcients. A detailed examination of these results reveals that very little is gained in terms of accuracy of the present results as one moves from G2 to G3 whereas the CPU time required for G3 is many folds larger than that needed for G2 to satisfy the same criterion of convergence. Thus, grid G2 denotes a good compromise between the accuracy and computational efforts and hence all results reported herein are based on the use of grid G2. Finally the adequacy of the values of (H/D) and grid G2 is further demonstrated in the next section by benchmarking the present results against the literature values for a few limiting cases like for a rotating cylinder in Newtonian media, and for a non-rotating cylinder in power-law uids.

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1457

Fig. 12. Comparison between the predicted and experimental velocity proles for 86% aqueous glycerin solution. Re = 0.411, n = 1: (a) = 0, (b) = 1.18, and (c) = 1.18.

6. Results and discussion As noted earlier, the present ow is governed by ve dimensionless parameters, namely, CD , CL , Re, n and and the principal aim of the present study is to develop this functional relationship. However, prior to embarking upon the presentation of new results, it is useful to demonstrate the adequacy of the numerical computations which will also help ascertain the accuracy of the new results reported herein. 6.1. Validation of results Since detailed comparisons are available elsewhere for the ow of Newtonian uids past a stationary cylinder [22,23], sufce it to say here that the present results are within 12% of the literature values. Next, Table 4 presents a comparison for a rotating cylinder in Newtonian uids; both the values of drag and lift coefcients are included here for = 1 and 6. Broadly speaking, the present results are within 12% of that reported by Stojkovic et al. [8] and Paramane and Sharma [6] for Re 1,

though the present values are seen to deviate by up to 56% at Re = 0.1 from that reported in [8]. Furthermore, at = 6 and Re = 40, while the present values of CDP and CDF are in excellent agreement with that reported in Ref. [6], but the total drag values differ by an order of magnitude. This is so due to two reasons: both CDP and CDF are of similar magnitudes but of opposite sign. Secondly, the values of CDP and CDF corresponding to Ref. [6] have been read off their gures. Thus, even a small error incurred in extracting these values gets accentuated in this case. While this may seem like a large difference, it needs to be emphasized here that the differences of this magnitude are not at all uncommon in such numerical studies due to underlying variations in different studies arising from the choices of solution methodologies, size and shape of domain, convergence criterion, etc. [38]. Similarly, Table 5 reports a representative comparison between the present and literature values for a stationary cylinder in shear-thinning uids. While more extensive comparisons are reported elsewhere [23], it is fair to say that the present results are seen to be in excellent agreement with the literature values in Table 5.

1458

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

Fig. 13. Comparison between the predicted and experimental velocity prole for 1% cellosize hydroxyethyl cellulose (HEC) solution. Re = 0.079, n = 0.43: (a) = 0, (b) = 1.706, and (c) = 1.706.

Aside from the aforementioned benchmark comparisons, the ow of Newtonian and power-law uids was also studied in the standard lid driven square cavity. The present results for the centerline velocities were found to be within 2% of the corresponding Newtonian results [39] and within 2.5% for power-law uids as reported by Neofytou [40]. These comparisons lend further support to the credibility of the numerical solution methodology employed herein. Based on the above-noted comparisons, the present results for a cylinder rotating in power-law uids are believed to be reliable to within 2%. 6.2. Detailed kinematics Figs. 24 show representative results elucidating the inuence of Re, and n on the streamlines in the vicinity of the cylinder. For the range of Reynolds number encompassed here, the ow is known to be steady for all values of the power-law index at = 0, i.e., for a

non-rotating cylinder [41]. As expected, the ow eld is symmetric about the mid-plane, and at low Reynolds numbers (Re 1), it also exhibits fore and aft symmetry. This is so due to the fact that the viscous forces far outweigh the inertial forces under these conditions. However, as the Reynolds number is gradually increased, the ow detaches itself from the surface of the cylinder which leads to the formation of the wake region in the rear of the cylinder. This marks the formation of a pair of standing vortices which grow in size with the increasing Reynolds number up to a critical value of the Reynolds number. For Newtonian uids, the rst signature of ow separation is seen at about Re = 45 whereas this transition is delayed in shear-thinning uids [41]. For instance, it occurs at about Re 1112 in a highly shear-thinning uid (n = 0.3) for a non-rotating cylinder [41]. A detailed examination of Figs. 24 reveals the following overall trends: As soon as the rotation of the cylinder is superimposed on the uniform ow, the ow becomes asymmetric as can clearly be seen even at Re = 0.1, albeit there is no

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

1459

indication of ow separation at this stage. In fact, a distinct saddle point is seen to form in Newtonian and mildly shear-thinning uids (n = 0.6) which disappears in highly shear-thinning uids (such as at n = 0.2). This is simply due to the rapid decay of the ow eld in highly shear-thinning uids and it is almost tantamount to a cylinder rotating in a cavity formed by low-viscosity uid which itself is surrounded by a high viscosity uid, akin to the rotation of an impeller in a highly shear-thinning and/or viscoplastic uid [42]. Owing to the weak advection, this effect is particularly striking at Re = 0.1, where no saddle point is seen in the streamlines plot. The viscous effects diminish with the increasing Reynolds number, or rotational velocity (due to enhanced levels of deformation), or a combination of both. Thus, as the value of the Reynolds number is progressively increased, one can see the formation of a streamline enclosing the cylinder at lower and lower values of . For instance, while this phenomenon occurs at = 6 for a weak uniform ow such as that at Re = 1 for n = 0.2, it was found to occur at = 4 for Re = 10 and n = 0.2. Similarly, the saddle point is seen to shift a little in the anti-clockwise direction with the increasing Reynolds number, or decreasing value of power-law index. For Newtonian uids, at vanishingly small values of the Reynolds numbers (Re 0), while the effect of rotation and uniform ow is easily delineated by superimposing the two individual solutions due to the linearity of the NavierStokes equations. This is clearly not possible for powerlaw uids due to the non-linearity of the viscous term even when the non-linear inertial terms are altogether neglected. As Reynolds number is gradually increased, the inertial effects progressively become important. While for = 0, the ow eld is found to be symmetric about the mid-plane, but with the increasing value of , the symmetry is lost and the upper vortex detaches from the cylinder (e.g., see Fig. 3 for Re = 10). Similarly, the lower vortex also dislodges itself and eventually disappears with increasing . Furthermore, the streamlines near the cylinder are drawn closely around it until a ring of trapped uid is formed which rotates along with the cylinder. This behaviour is seen for all values of the power-law index considered herein (1 n 0.2), except for the fact that the size of the so-called separatrix [6] is seen to be smaller in shear-thinning uids than that in Newtonian uids otherwise under identical conditions. Thus, in summary broadly, qualitatively similar streamline patterns are seen in power-law shear-thinning and Newtonian uids in the steady ow regime. Typical value of the stream function ( * = /U0 D) range from 108 m2 /s at the bottom of the box (y = H/2) to 4000 m2 /s at top of the box (y = H/2) depending upon the values of n, Re, . Naturally, lower the value of Re, smaller is the value of * near the cylinder. Representative vorticity contours are shown in Figs. 57 for the same combinations of Re, n and which also reinforce the trends seen in streamline patterns. For instance, at low Reynolds numbers, the vorticity contours become increasingly asymmetric as the value of is increased. Indeed, larger is the value of Reynolds number, lower is the value of needed to break the symmetry. The crowding of the iso-vorticity contours underneath the cylinder show the intensity of the vorticity distribution in this region. With the increasing Reynolds number, the vortices are seen to be transported in the main direction of ow. The surface vorticity is naturally maximum on the surface of the cylinder (* = /U0 D) and it ranges from 1 to 220 at Re = 0.1 as the value of n is decreased from n = 1 to n = 0.2. Of course, it decays to very small values of the order of 1012 to 107 far away from the cylinder. Fig. 8 shows representative surface pressure proles for a range of combinations of Re, and n. Obviously, since the ow is not symmetric about the x- or y-axis, pressure proles are shown over the entire surface (0 360). At low Reynolds numbers, e.g., at Re = 0.1, the surface pressure in the front of the cylinder progressively increases with the increasing degree of shear-thinning (decreasing values of n), albeit the effect is most dramatic at the

front stagnation point. In the rear of the cylinder, the inuence of power-law index is not monotonic on the surface pressure. With the superimposition of rotation, the pressure is no longer maximum at the front stagnation point and it occurs at about = 2030 . At low Reynolds numbers, as the value of is gradually increased, the surface pressure prole exhibits increasing asymmetry. Similarly, for a xed value of , as the Reynolds number is gradually increased, the surface pressure drops progressively, and the effect of n also weakens up to about = 3. However, as expected, for = 6, the pressure is negative over most of the surface of the cylinder (e.g., see plot for Re = 40 and = 6), where as it undergoes a sign change at lower Reynolds numbers. The complex dependence seen in Fig. 8 obviously stems from different scaling of the viscous and inertial forces on velocity and power-law index. How these changes at the microscopic level inuence the values of macroscopic quantities like drag and lift coefcient will be seen in the next section, as one would expect the pressure drag to be negative under certain circumstances. 6.3. Drag and lift coefcients Figs. 9 and 10 show the inuence of , Re and n on pressure drag coefcient, total drag coefcient and lift coefcient. At low Reynolds numbers, both total drag coefcient and its pressure component are seen to decrease with the increasing value of rotational velocity which is consistent with the results of Stojkovic et al. [8] and Paramane and Sharma [6] for Newtonian uids (n = 1). While at low Reynolds numbers, the rotation of the cylinder augments both the total drag and its pressure component, but it is evident that the pressure drag contributes to the total drag signicantly under these conditions. On the other hand, at high Reynolds numbers, Re = 40, the dependence of CD and CDP on is more complex. For Newtonian uids, the total drag decreases with the increasing value of up to about 5 beyond which it shows an increase. It is interesting to see that under certain conditions both CD and CDP can be negative which is consistent with the pressure proles seen in the previous sections due to a very small contribution of friction component. Shear-thinning uids are also seen to exhibit qualitatively similar behaviour, except for the fact that while in some cases the value of CDP can become slightly negative but the total drag always remains positive thereby suggesting that friction probably contributes in greater proportion here than that in the case of Newtonian uids. Indeed, smaller is the value of power-law index (highly shear-thinning uid), higher is the total drag coefcient. Fig. 10 shows these results in a slightly different form where it is clearly seen that the effect of power-law index on both CDP and CD gradually weakens with the increasing Reynolds number. For instance, for = 1, the curves for different values of n collapse onto a single curve at about Re 12, whereas at = 6 this behaviour occurs at about Re 40. The lift coefcient is seen to be negative (i.e. acting in the downward direction) at all values of the Reynolds number considered in this work and it decreases almost linearly with the increasing rotational velocity. At low Reynolds numbers, it goes through a maximum value at about n 0.6, as can be seen clearly in Fig. 9. At low Reynolds numbers, the rate of change also increases with the increasing degree of shear-thinning behaviour, i.e., decreasing value of n. On the other hand, at Re = 40, the effect of n is rather insignicant up to about = 23 and beyond this value, the lift coefcient increases with the decreasing value of n. While these results are in line with that of Stojkovic et al. [8] for Newtonian uids, but these are at variance from that of Townsend [35] who predicted the lift to increase with to a maximum value, but then it falls steeply. This is an unusual result which has been attributed to the inadequacy of their computations [8]. A representative summary of the present numerical values of drag and lift coefcients as func-

1460

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461

tions of Re, n and is given in Table 6. Intuitively, one would expect the product of drag coefcient and Reynolds number to approach a limiting value depending upon the values of n and as the value of the Reynolds number diminishes, but unfortunately the lowest value of Re = 0.1 used here is not sufciently small for this limiting behaviour to be observed in the present case. Finally, Fig. 11 shows the inuence of the power-law index, Reynolds number, and on the normalized values of CD , CDP and CL to delineate the role of power-law index in an unambiguous fashion. As noted earlier, at low Reynolds numbers, the total drag on a cylinder can be as much as 34 times of that in a Newtonian uid otherwise under identical conditions; however, this effect diminishes with the increasing value of . On the other hand, this effect is seen to be even more dramatic at Re = 40. In the extreme case, not only CD can undergo a change of sign depending upon the values of and n, but their ratio too can be as high as 70. In contrast, the lift coefcient is seen to display a more regular dependence on n and , both at high and low Reynolds numbers, though it peaks at about n 0.6 at low Reynolds numbers. 7. Comparison with experiments As noted earlier, while no experimental results are available on drag and lift coefcients for a cylinder rotating in Newtonian and in power-law uids in the steady ow regime, Christiansen [36,37] reported detailed velocity measurements for a rotating cylinder in a Newtonian glycerol solution and in two polymer solutions. Two test cylinders 102 mm long, but of two different diameters, namely, 1.59 mm and 4.76 mm were used. The (L/D) ratios are 22 and 65, which are modest for the end effects to be negligible. The two polymer solutions used in their study exhibited viscoelastic behaviour, with relaxation times of 0.041 s (HEC solution) and 8.6 s (Separan solution). The value of power-law index for both solutions is identical, as n = 0.43. The values of varied from 0 to 2.29 and of the Reynolds number are in the range 0.03 to 0.5. He reported the values of the x- and y-component of the velocity along (x = 0) and (y = 0) lines both upstream and downstream from the cylinder. Fig. 12 shows a comparison between the present predictions and the experimental data for their Newtonian uid at Re = 0.411 for a stationary cylinder ( = 0) and a rotating cylinder ( = 1.18). An excellent agreement is seen to exist between the present numerical results and the experimental results. Fig. 13 shows a similar comparison for the weakly elastic HEC solution wherein the correspondence is though less good, but may still be regarded as satisfactory thereby suggesting this solution to be virtually inelastic over this range of conditions. The corresponding value of Deborah number, dened as De = U0 /D, is 0.32 which is rather low and therefore, visco-elastic effects are expected to be negligible here. On the other hand, the value of Deborah number ranges from 48 to 150 for the Separan solution and therefore, it is not appropriate to compare these results with the present predictions. Overall, the experimental verication seen in Figs. 12 and 13 lends further credibility to the accuracy and reliability of the results presented herein. 8. Concluding remarks In this work, the ow of shear-thinning power-law uids past a rotating cylinder in the 2D laminar steady ow regime has been investigated numerically. The range of conditions encompassed include power-law index (1 n 0.2), Reynolds number (0.1 Re 40) and rotational velocity (0 6). The ow is visualized in terms of streamline contours, iso-vorticity proles and surface pressure proles for representative combinations of the values of Re, n and . The lift is found to be negative under most

conditions studied here, whereas the total drag force is usually positive but can be negative when the friction makes a little contribution and the pressure drag coefcient itself is negative such as that at high values of and Re. The power-law index exerts a much stronger inuence on drag and lift at low Reynolds numbers than that at high Reynolds numbers. Indeed, shear-thinning behaviour can augment drag values by up to a factor of 4 at low values of Re and , albeit the extent of increase decreases rapidly with the increasing values of . Finally, the present numerical predictions of velocity at x = 0 and y = 0 planes are shown to be in fair agreement with the scant experimental results available in the literature for Newtonian and relatively inelastic polymer solutions.

References
[1] M.M. Zdravkovich, Flow Around Circular Cylinders, vol. 1: Fundamentals, Oxford University Press, New York, 1997. [2] M.M. Zdravkovich, Flow Around Circular Cylinders, vol. 2: Applications, Oxford University Press, New York, 2003. [3] C.H.K. Williamson, Vortex dynamics in the cylinder wake, Annu. Rev. Fluid Mech. 28 (1996) 477539. [4] M. Coutanceau, J.R. Defaye, Circular cylinder wake conguration: a ow visualization survey, Appl. Mech. Rev. 44 (1991) 255305. [5] B.M. Sumer, J. Fredsoe, Hydrodynamics Around Cylindrical Structures, World Scientic, Singapore, 2006. [6] S.B. Paramane, A. Sharma, Numerical investigation of heat and uid ow across a rotating circular cylinder maintained at constant temperature in 2-D laminar ow regime, Int. J. Heat Mass Transfer 52 (2009) 32053216. [7] R.P. Chhabra, J.F. Richardson, Non-Newtonian Flow and Applied Rheology, 2nd ed., Butterworth-Heinemann, Oxford, 2008. [8] D. Stojkovic, M. Breuer, F. Durst, Effect of high rotation rates on the laminar ow around a circular cylinder, Phys. Fluids 14 (2002) 31603178. [9] Y.-M. Chen, Y.-R. Ou, A.J. Pearlstein, Development of the wake behind a circular cylinder impulsively started into rotary and rectilinear motion, J. Fluid Mech. 253 (1993) 449484. [10] M. Coutanceau, C. Menard, Inuence of rotation on the near-wake development behind an impulsively started circular cylinder, J. Fluid Mech. 158 (1985) 399446. [11] H.M. Badr, S.C.R. Dennis, Time-dependent viscous ow past an impulsively started rotating and translating circular cylinder, J. Fluid Mech. 158 (1985) 447488. [12] M.-H. Chou, Numerical study of vortex shedding from a rotating cylinder immersed in a uniform ow eld, Int. J. Num. Methods Fluids 32 (2000) 545567. [13] G. Hu, D. Sun, X. Yin, B. Tong, Hopf bifurcation in wakes behind a rotating and translating circular cylinder, Phys. Fluids 8 (1996) 19721974. [14] S. Kang, H. Choi, S. Lee, Laminar ow past a rotating circular cylinder, Phys. Fluids 11 (1999) 33123321. [15] H.M. Badr, S.C.R. Dennis, P.J.S. Young, Steady and unsteady ow past a rotating circular cylinder at low Reynolds numbers, Comp. Fluids 4 (1989) 579609. [16] A.A. Kendoush, An approximate solution of the convection heat transfer from an isothermal rotating cylinder, Int. J. Heat Fluid Flow 17 (1996) 439441. [17] V.K. Patnana, R.P. Bharti, R.P. Chhabra, Two dimensional unsteady ow of power-law uid over a cylinder, Chem. Eng. Sci. 64 (2009) 29782999. [18] V.K. Patnana, R.P. Bharti, R.P. Chhabra, Two dimensional unsteady forced convection heat transfer in power-law uids from a cylinder, Int. J. Heat Mass Transfer 53 (2010) 41524167. [19] R.I. Tanner, Stokes paradox for power-law ow around a cylinder, J. Non-Newt. Fluid Mech. 50 (1993) 217224. [20] M.J. Whitney, G.J. Rodin, Forcevelocity relationship for rigid bodies translating through unbounded shear-thinning power-law uids, Int. J. Non-Linear Mech. 36 (2001) 947953. [21] R.P. Chhabra, A.A. Soares, J.M. Ferreira, Steady non-Newtonian ow past a circular cylinder: a numerical study, Acta Mech. 172 (2004) 116. [22] A.A. Soares, J.M. Ferreira, R.P. Chhabra, Flow and forced convection heat transfer in cross ow of non-Newtonian uids over a circular cylinder, Ind. Eng. Chem. Res. 44 (2005) 58155827. [23] R.P. Bharti, R.P. Chhabra, V. Eswaran, Steady ow of power-law uids across a circular cylinder, Can. J. Chem. Eng. 84 (2006) 406421. [24] R.P. Chhabra, Bubbles, Drops and Particles in Non-Newtonian Fluids, 2nd ed., CRC Press, Boca Raton, FL, 2006. [25] R.P. Bharti, R.P. Chhabra, V. Eswaran, Steady forced convection heat transfer from a heated circular cylinder to power-law uids, Int. J. Heat Mass Transfer 50 (2007) 977990. [26] A.T. Srinivas, R.P. Bharti, R.P. Chhabra, Mixed convection heat transfer from a cylinder in power-law uids: effect of aiding buoyancy, Ind. Eng. Chem. Res. 48 (2009) 97359754. [27] A.A. Soares, J. Anacleto, L. Caramelo, J.M. Ferreira, R.P. Chhabra, Mixed convection from a circular cylinder to power-law uids, Ind. Eng. Chem. Res. 48 (2009) 82198231.

S.K. Panda, R.P. Chhabra / J. Non-Newtonian Fluid Mech. 165 (2010) 14421461 [28] R.P. Bharti, R.P. Chhabra, V. Eswaran, Two-dimensional steady Poiseuille ow of power-law uids across a circular cylinder in a plane conned channel: wall effect and drag coefcient, Ind. Eng. Chem. Res. 46 (2007) 38203840. [29] R.P. Bharti, R.P. Chhabra, V. Eswaran, Effect of blockage on heat transfer from a cylinder to power-law liquids, Chem. Eng. Sci. 62 (2007) 47294741. [30] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Flow and heat transfer across a conned square cylinder in the steady ow regime: effect of Peclet number, Int. J. Heat Mass Transfer 48 (2005) 45984614. [31] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Steady ow of power-law uids across a square cylinder, Chem. Eng. Res. Des. 84 (2006) 300310. [32] 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. [33] R.P. Bharti, P. Sivakumar, R.P. Chhabra, Forced convection heat transfer from an elliptic cylinder to power-law uids, Int. J. Heat Mass Transfer 51 (2008) 18381853. [34] A.K. Sahu, R.P. Chhabra, V. Eswaran, Two dimensional laminar ow of a powerlaw uid across a square cylinder, J. Non-Newtonian Fluid Mech. 165 (2010) 752763.

1461

[35] P. Townsend, A numerical simulation of Newtonian and visco-elastic ow past stationary and rotating cylinders, J. Non-Newtonian Fluid Mech. 6 (1980) 219243. [36] R.L. Christiansen, A laser-doppler anemometry study of visco-elastic ow past a rotating cylinder, Ind. Eng. Chem. Fundam. 24 (1985) 403408. [37] R.L. Christiansen, PhD Thesis, University of Wisconsin, Madison, WI, 1980. [38] P.J. Roache, Perspective:, A method for uniform reporting of grid renement studies, J. Fluids Eng. 116 (1994) 405413. [39] U. Ghia, K.N. Ghia, C.T. Shin, High-Re solutions for incompressible ow using the NavierStokes equations and a multigrid method, J. Comp. Phys. 48 (1982) 387411. [40] P. Neofytou, A 3rd order upwind nite volume method for generalized Newtonian uid ows, Adv. Eng. Softw. 36 (2005) 664680. [41] P. Sivakumar, R.P. Bharti, R.P. Chhabra, Effect of power law index on critical parameters for power law uid ow across an unconned circular cylinder, Chem. Eng. Sci. 61 (2006) 60356046. [42] R.P. Chhabra, Fluid mechanics and heat transfer with non-Newtonian liquids in mechanically agitated vessels, Adv. Heat Transfer 37 (2003) 77178.

Vous aimerez peut-être aussi