Chemical Engineering Science 62 (2007) 1682 – 1702 www.elsevier.com/locate/ces Steady flow of power-law fluids across an unconfined elliptical cylinder P. Sivakumar, Ram Prakash Bharti, R.P. Chhabra ∗ Department of Chemical Engineering, Indian Institute of Technology, Kanpur 208016, India Received 3 July 2006; received in revised form 25 November 2006; accepted 27 November 2006 Available online 14 December 2006 Abstract The momentum transfer characteristics of the power-law fluid flow past an unconfined elliptic cylinder is investigated numerically by solving continuity and momentum equations using FLUENT (version 6.2) in the two-dimensional steady cross-flow regime. The influence of the power-law index (0.2 n 1.8), Reynolds number (0.01 Re 40) and the aspect ratio of the elliptic cylinder (0.2 E 5) on the local and global flow characteristics has been studied. In addition, flow patterns showing streamline and vorticity profiles, and the pressure distribution on the surface of the cylinder have also been presented to provide further physical insights into the detailed flow kinematics. For shear-thinning (n < 1) behaviour and the aspect ratio E > 1, flow separation is somewhat delayed and the resulting wake is also shorter; on the other hand, for shear-thickening (n > 1) fluid behaviour and for E < 1, the opposite behaviour is obtained. The pressure coefficient and drag coefficient show a complex dependence on the Reynolds number and power-law index. The decrease in the degree of shear-thinning behaviour increases the drag coefficient, especially at low Reynolds numbers. While the aspect ratio of the cylinder exerts significant influence on the detailed flow characteristics, the total drag coefficient is only weakly dependent on the aspect ratio in shear-thickening fluids. The effect of the flow behaviour index, however, diminishes gradually with the increasing Reynolds number. The numerical results have also been presented in terms of closure relations for easy use in a new application. 䉷 2007 Elsevier Ltd. All rights reserved. Keywords: Power-law fluids; Elliptical cylinders; Drag coefficients; Pressure coefficients; Reynolds number; Shear-thinning; Shear-thickening 1. Introduction Owing to its fundamental and pragmatic significance, considerable research efforts have been devoted to the study of cross-flow of fluids past cylinders of circular and non-circular cross-sections. Typical examples include the flow in tubular and pin heat exchangers, in the RTM process of manufacturing fibre reinforced composites, in filtration screens and aerosol filters, etc. Consequently, a voluminous body of knowledge is now available on the transverse flow of Newtonian fluids over a circular cylinder, e.g., see, Zdravkovich (1997, 2003). In contrast, much less is known about the role of the cross-sectional shape of the cylinder on the flow phenomenon. For instance, only limited results available for the flow of Newtonian fluids past a cylinder of square and rectangular cross-sections have been investigated numerically in the low-to-moderate Reynolds ∗ Corresponding author. Tel.: +91 512 2597393; fax: +91 512 2590104. E-mail address: [email protected] (R.P. Chhabra). 0009-2509/$ - see front matter 䉷 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2006.11.055 number range during the last 10–20 years (Dhiman et al., 2005, 2006a; Sharma and Eswaran, 2004, 2005; Chhabra, 2006, and references therein). Likewise, the tubes of elliptic crosssection are used commonly in tubular heat exchangers as these offer good space economy without sacrificing heat transfer-flow resistance benefits. It is readily acknowledged that many substances encountered in industrial practice (pulp and paper, food, polymer and process engineering applications) display shear-thinning and/or shear-thickening behaviour (Chhabra and Richardson, 1999). Owing to their high viscosity levels, non-Newtonian substances are generally processed in laminar flow conditions. Though many non-Newtonian fluids display viscoelastic behaviours, the available scant literature both for the flow past a single cylinder and a periodic array of cylinders seems to suggest the viscoelastic effects to be minor in this flow configuration (Chhabra, 2006; Liu et al., 1998; Talwar and Khomami, 1995). Furthermore, the fluid relaxation time often shows a dependence on the shear rate, which is similar to the sheardependence of viscosity. Thus, the relaxation time will also P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 decrease with the increasing value of the Reynolds number and hence, a suitably defined Deborah number would also be small. On this count, the viscoelastic effects are not expected to be significant in this case. Unfortunately, most numerical simulations examine the role of viscoelasticity in the absence of sheardependent viscosity and predict very little change in the value of drag coefficient. Thus, it appears that the steady flow resistance is determined primarily by the flow geometry (shape of the cylinder) and the viscous properties of the fluid. Therefore, it seems reasonable to begin with the flow of purely viscous power-law type fluids as long as the power-law constants are evaluated in the shear rate range appropriate for the flow over a elliptical cylinder, and the level of complexity can be built up gradually to accommodate other non-Newtonian characteristics. Admittedly, the power-law model does not predict the socalled zero shear viscosity in the limit of zero shear rate, but the available numerical and experimental studies for the flow over a sphere and a circular cylinder clearly show that despite this deficiency, the use of this model yields satisfactory predictions of the flow parameters (Chhabra, 2006; Schowalter, 1978). As far as known to us, there has been no prior study of the cross-flow of power-law fluids over an elliptic cylinder. This work aims to fill this gap in the literature. At the outset, it is instructive, however, to briefly recount the available limited work on the flow of power-law fluids past circular cylinders and on Newtonian fluid flow past elliptic cylinders to facilitate the subsequent presentation of the new results for the flow of power-law fluids past an elliptic cylinder. 2. Previous work The flow of Newtonian fluids past an unconfined circular cylinder has been studied extensively, e.g., see Zdravkovich (1997, 2003). All in all, a reasonable body of experimental and/or numerical results is now available on almost all aspects of the cross-flow of Newtonian fluids past a circular cylinder over most conditions of interest. In contrast, only limited information is available for the Newtonian flow over an elliptical cylinder, even in the steady cross-flow regime. Imai (1954) and Hasimoto (1958) studied the steady flow past an elliptic cylinder at very low Reynolds numbers by using the Oseen’s linearized equations and presented an expression for drag. Lugt and Haussling (1974) numerically solved the flow equations for the laminar flow past an abruptly started elliptic cylinder at 45◦ incidence. In particular, time required to reach the steady or quasi-steady state flow was investigated by following the evolution of streamlines and lines of constant vorticity and of drag, lift and moment coefficients with time. Meller (1978) numerically solved the vorticity-stream function formulation of the Navier–Stokes equations using the finite difference method and the matrix pivotal condensation method for only two values of the Reynolds number of 20 and 40 and for an aspect ratio of 10 at an angle of attack of 30◦ . Subsequently, Patel (1981) presented a semianalytical solution of the Navier–Stokes equations to investigate the viscous, incompressible flow around an impulsively started elliptic cylinder at various angles of incidence (0, 30, 45 1683 and 90◦ ) at Re = 200. D’Alessio and Dennis (1994) employed the vorticity-stream function form of the Navier–Stokes equations in terms of transformed coordinates to enforce the correct decay of vorticity at large distances from the elliptic cylinder. They presented the values of drag and lift coefficients for Re=5 and 20 for different inclinations of the elliptic cylinder ranging from 0◦ to 90◦ . Subsequently, this flow was revisited by Dennis and Young (2003) who studied the steady two-dimensional flow past an elliptic cylinder at various inclinations to the oncoming stream by solving the Navier–Stokes equations in the range of the Reynolds number (1 Re 40) and for a range of inclinations (0–90◦ ) for one value of the aspect ratio E = 5. Using the spectral-element method, Johnson et al. (2001) reported the effects of the Reynolds number (30 Re 200) and of aspect ratio (0.01 E 1) on the Strouhal number, drag coefficients and on the onset of vortex shedding. They reported the conditions for the steady flow regime to be relatively insensitive to the value of aspect ratio E, whereas the span of the other regimes narrows down with a decrease in the value of E. For E < 1, the steady regime was reported to end at Re ≈ 35–40 which is comparable to the commonly used value of ∼ 40–45 for circular and square cylinders. Apart from the above-referenced numerical studies, Khan et al. (2005) have recently employed the boundary layer approximation of the momentum and energy equations using the Von Karman–Pohlhausen integral method for flow and heat transfer from an elliptical cylinder (E 1). Scant experimental results are also available for the flow and heat transfer from elliptic cylinder, but most of these relate to high Reynolds number, e.g., see Ilgarbuis and Butkus (1988), Nishiyama et al. (1988a,b), Ota et al. (1987), Seban and Drake (1953), etc. and these have focused on time-dependent flow characteristics. Apart from these single cylinder studies, some results are also available for Newtonian fluid flow and heat transfer in bundles of elliptic cylinders (Epstein and Masliyah, 1972; Masliyah, 1973). Therefore, only limited information is available even for the flow of Newtonian fluids over an elliptic cylinder. Even less is known about the power-law fluid flow over a circular cylinder and as far as known to us, no prior results are available for elliptic cylinders. Since the detailed discussions of these for a circular cylinder are available elsewhere (Bharti et al., 2006, 2007; Sivakumar et al., 2006), only the salient features are recapitulated here. The three creeping flow studies (Ferreira and Chhabra, 2004; Tanner, 1993; Whitney and Rodin, 2001) are in excellent agreement with each other for n < 1, as far as the values of the total drag coefficients are concerned; the corresponding values of the limiting Reynolds number denoting the cessation of the creeping flow regime have been delineated recently (Sivakumar et al., 2006). Likewise, the two independent numerical studies in the two-dimensional steady flow regime for finite values of the Reynolds number (Re 40) (Bharti et al., 2005, 2006, 2007; Chhabra et al., 2004; Soares et al., 2005) are also in good agreement with each other. For the sake of completeness, it is also appropriate to mention here that limited results for the steady flow of powerlaw fluids past confined and unconfined square (Dhiman et al., 2006b; Gupta et al., 2003; Paliwal et al., 2003) and rectangular 1684 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 (Nitin and Chhabra, 2005) cylinders are also available. In the latter cases, the drag coefficient normalized using the corresponding Newtonian values shows only weak additional dependence on the power-law index and over the Reynolds number range of 5–40. Before leaving this section, it is appropriate to add here that there has been one study for the creeping flow of power-law fluids through periodic arrays of elliptic cylinders (Woods et al., 2003). In principle, these results should reduce to the limit of a single elliptic cylinder as the solid volume fraction approaches zero. However, since the drag coefficient shows a non-linear and rather strong dependence on volume fraction, this approach yields reliable values only when the results are available for extremely small values of the solid volume fraction. Thus, it is safe to conclude that there has been no prior study on the steady flow of power-law fluids past an unconfined elliptical cylinder. The present work is concerned with the determination of the detailed kinematics and macroscopic hydrodynamic characteristics in the two-dimensional steady cross-flow regime for the range of Reynolds number as 0.01Re 40, aspect ratio as 0.2 E 5, and power-law index as 0.2 n 1.8, thereby covering both shear-thinning (n < 1) and shear-thickening (n > 1) fluid behaviours. Consider the two-dimensional, steady flow of an incompressible power-law liquid with an uniform velocity U∞ across a long elliptical cylinder of aspect ratio E. The unconfined flow condition is simulated here by enclosing the elliptic cylinder in a circular outer boundary of diameter D∞ , as shown in Fig. 1. The diameter of the outer circular boundary D∞ is taken to be sufficiently large to minimize the boundary effects (Fig. 1). The continuity and momentum equations for this flow configuration are written as continuity equation : ∇.u = 0, (1a) momentum equation : (u.∇u − f) − ∇. = 0, (1b) where, , u, f and are the density, velocity (Ux and Uy ), body force and the stress tensor, respectively. The stress tensor, sum of isotropic (pressure, p) and the deviatoric stress tensor (), is given by (2) The rheological equation of state for power-law fluids is given by = 2(u), (3) where (u), the components of the rate of strain tensor, is given by (u) = 21 {(∇u) + (∇u)T }. For a power-law fluid, the viscosity, , is given by (n−1)/2 I2 =m , 2 I2 = (2xx + 2yy ) + (2xy + 2yx ) (4) (5) (6) and the components of the rate of strain tensor are related to the velocity components as follows: jUy jUy jUx 1 jUx , yy = , xy = yx = + . xx = jx jy 2 jy jx (7) The physically realistic boundary conditions for this flow may be written as follows: • At the inlet boundary: The condition of the uniform flow is imposed, i.e., Ux = U∞ and Uy = 0. (8a) • On the surface of the elliptical cylinder: The standard no-slip condition is used, i.e., Ux = 0 3. Problem statement = −pI + . where m is the power-law consistency index and n is the powerlaw index of the fluid (n < 1: shear-thinning; n = 1: Newtonian; and n > 1: shear-thickening fluids) and I2 is the second invariant of the rate of strain tensor () given by and Uy = 0. (8b) • At the exit boundary: The default outflow boundary condition in FLUENT, which assumes a zero diffusion flux for all flow variables, was used. Physically this condition implies that the conditions of the outflow plane are extrapolated from within the domain and have negligible impact on the upstream flow conditions. The extrapolation procedure used by FLUENT updates the outflow velocity and pressure in a manner that is consistent with a fully developed flow assumption, when there is no area change at the outflow boundary. It is important to note here that gradients in the cross-stream direction may exist at the outflow boundary. Only the diffusion fluxes in the direction of flow at the exit plane are assumed to be zero. Also, the use of this condition obviates the need to prescribe any boundary condition for pressure. This is similar to the homogeneous Neumann condition, given by jUx =0 jx and jUy = 0. jx (8c) • At the plane of symmetry, i.e., centre line (y = 0): The symmetric flow condition has been used. It can be written as follows: jUx =0 jy and Uy = 0. (8d) Owing to the symmetry of the flow over the range of conditions studied herein, the solution is obtained only in the upper half of the domain y 0 (Fig. 1). The numerical solution of Eqs. (1a) and (1b) along with the above-noted boundary conditions yields the velocity (Ux and Uy ) and pressure (p) fields and these, in turn, are used to deduce the local and global characteristics like P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 a b c D∞ D∞ D∞ a a Uniform Velocity U∞ 1685 b Outflow a b Uniform Velocity U∞ Uniform Velocity U∞ Outflow b Outflow Cylinder Cylinder y Cylinder y y x Outer Boundary Outer Boundary x x Outer Boundary Fig. 1. Schematic representation of the unconfined flow around a elliptical cylinder: (a) E < 1(a > b), (b) E = 1 (a = b, circular cylinder), (c) E > 1(a < b). stream-function, vorticity, wake length, pressure coefficients, drag coefficients, as outlined below. At this point, it is appropriate to introduce some definitions. • The Reynolds number (Re) for power-law fluids is defined as follows: Re = 2−n (2a)n U∞ . m (9) • Surface vorticity, , is computed by the following expression: jUr = , js (10) where Ur = (Ux sin + Uy cos ), and is the angular displacement from the front stagnation point and s is the unit vector normal to the surface of the cylinder. • Surface pressure coefficient, Cp , the ratio of the static to dynamic pressure on the surface of the cylinder, is defined as follows: p() − p∞ Cp = 2 , (11) 2 U∞ where p() is the surface pressure at an angle and p∞ is the free stream pressure. • Recirculation (or wake) length, Lw , is measured from the rear stagnation point ( = 180◦ ) to the point of reattachment for the near closed streamline (Ux = Uy = 0) on the line of symmetry (y = 0) in the downstream section. It is expressed by the following relation: Lw = l−b , 2b (12) where l is the distance from the centre of the cylinder to the point of reattachment for the near closed streamline. • Total drag coefficient, CD , is defined as CD = 2FD = CDP + CDF , 2 a U∞ (13) where FD is the drag force per unit length of the cylinder. The pressure and friction drag coefficients, CDP and CDF , are calculated using the following definitions: 2FDP CDP = (14) = Cp nx dS ∗ , 2 a U∞ S where S ∗ is the surface area. 2FDF 2n+1 = (∗ .ns ) dS ∗ CDF = 2 a U∞ Re S = 2n+1 Re S (∗xx nx + ∗xy ny ) dS ∗ , (15) where ns is the direction vector normal to the surface of the cylinder, given as (x/a 2 )ex + (y/b2 )ey ns = = n x ex + n y e y , (x/a 2 )2 + (y/b2 )2 (16) where ex and ey are the x- and y-components of the unit vector, respectively, and ∗ , the dimensionless shear stress, is expressed as ∗ij = ∗ jUj jUi + jj ji = I2∗ 2 (n−1)/2 jUj jUi + , jj ji (17) where ∗ is the dimensionless viscosity and I2∗ is the dimensionless second invariant of the rate of strain tensor. The shear stress, viscosity and second invariant of the rate of strain tensor are scaled with m(U∞ /a)n , m(U∞ /a)(n−1)/2 and (U∞ /a)2 , respectively. Since the governing equations in the Cartesian coordinates have been solved using FLUENT, hence all the above-noted expressions are written in Cartesian coordinates only. Corresponding expressions, especially for the individual drag coefficients, in other coordinate systems are available in the literature, e.g., see Dennis and Young (2003), D’Alessio and Dennis (1994), Khan et al. (2005), Kiya and Arie (1975), Sugihara-Seki (1993), etc. 1686 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 4. Numerical methodology The present numerical study has been carried out using FLUENT (version 6.2). The unstructured ‘quadrilateral’ cells of non-uniform spacing were generated using GAMBIT. The grid near the surface of the cylinder was sufficiently fine to resolve the boundary layer over the cylinder. Furthermore, the two-dimensional, steady, laminar segregated solver was used with second order upwinding scheme for the convective terms in the momentum equation. The semi-implicit method for the pressure linked equations (SIMPLE) scheme was used for pressure–velocity coupling and non-Newtonian power-law model was used for viscosity. A convergence criterion of 10−10 was used for x- and y-components of the velocity and the residuals of continuity. 5. Choice of numerical parameters Naturally, the accuracy and reliability of the numerical results is contingent upon a prudent choice of optimal domain and grid. In this work, the domain is characterized by the diameter (D∞ ) of the faraway cylindrical envelope of fluid. Evidently, it is not possible to simulate truly unconfined flow in any numerical study. A very large value of D∞ will warrant enormous computational resources and a small value will unduly influence the results. The effect of this parameter is examined in the next section, followed by that of the grid size to arrive at their optimal values. 5.1. Domain independence study The available scant literature shows that the smaller the value of the Reynolds number, the larger the value of D∞ required for the results to be free from domain effects. This is so presumably due to the thickness of the boundary layer which decreases with the increasing Reynolds number. Bearing in mind this feature, several values of D∞ /(2a) ranging from 200 to 1200 have been used in this study to minimize the so-called domain effects. For three values of n = 0.2, 1 and 1.8, three values of Re = 0.01, 5 and 40 and for the extreme values of E = 0.2 and 5, the results are shown in Table 1. It can be seen that an increase in the domain size from 1000 to 1200 at Re = 0.01 alters the drag values by 1.14% and 1.35% at n = 1 for E = 0.2 and 5, respectively. The increases in the domain size from 700 to 1200 and from 900 to 1200 at n = 0.2 and 1.8 show the maximum change in the drag values of 0.004% and 0.34% for E = 0.2, the corresponding values are 0% and 0.47% for E = 5 at Re = 0.01. Furthermore, it was found that for Re = 5 and 40, the change in the computational domain size from 200 to 300 and 250 to 300 yields very small change in the drag values (maximum value being 0.47% and 0.60% for E = 0.2 and 5, respectively) for n = 0.2, 1 and 1.8, respectively. These results clearly suggest a faster decay in velocity with distance in both shear-thinning and shear-thickening fluids than that in a Newtonian fluid. Therefore, the domain sizes of 1200 and 300 are believed to be adequate in the Reynolds number range of 0.01Re5 and 5Re 40, respectively, over the power- law index range (0.2 n 1.8) considered here, to obtain the results which are essentially free from domain effects. 5.2. Grid independence study Having fixed the domain size, the grid independence study was carried out using three non-uniform grids for the three values of the power-law index of 0.2, 1 and 1.8, at Re = 0.01 and 40 for E = 0.2 and 5, as shown in Table 2. Obviously, in moving from the grid G2 to G3, the results remain virtually unchanged (except for the friction drag for E = 0.2, n = 0.2 at Re=40, which changes by 3.8%) accompanied by an enormous increase (up to 2–3 fold in extreme cases) in the computational time. In spite of this seemingly large change, the corresponding change in the value of the total drag coefficient is less than 1%. Therefore, the grid G2 is believed to be sufficiently refined to resolve the flow phenomena with acceptable levels of accuracy within the range of conditions of interest here. 6. Results and discussion In this work, steady flow computations have been carried out by systematically varying the value of the power-law index in the range as 0.2 n 1.8 in the steps of 0.2, for the Reynolds number of 0.01, 1, 5, 10, 20, 30, 40 and for five values of E = 0.2, 0.5, 1, 2 and 5. The results presented herein are based on the domain size, D∞ /(2a), of 1200 and 300, respectively, for 0.01 Re 5 and 5 Re 40 and with the grid size of G2. The flow of Newtonian fluids is known to be two-dimensional and steady with or without two symmetric vortices over this range of Reynolds number (Dennis and Young, 2003; Lugt and Haussling, 1974; Williamson, 1996; Zdravkovich, 1997) and this is assumed to be so for power-law fluids also. However, prior to presenting the new results, it is appropriate to validate the solution procedure to ascertain the accuracy and reliability of the results presented herein. 6.1. Validation of results Since extensive validation for the flow of Newtonian and power-law fluids over a circular cylinder (E =1) has been dealt with elsewhere (Bharti et al., 2006; Sivakumar et al., 2006), only the additional comparison for the flow of Newtonian fluids past an elliptic cylinder is reported here (Table 3). The agreement between the present and prior results is seen to be excellent, except that for E = 0.2 and Re = 40. While the exact reasons for this discrepancy are not known, it needs to be emphasized here that the present results are based on the use of a much larger domain, i.e., D∞ /(2a) of 300 (5 Re40) and 1200 (0.01 Re 5), whereas the corresponding values are 25 (Dennis and Young, 2003) and 7.5 (values due to Dennis and Young (2003) in Dennis and Chang, 1969) and the grid employed in the present work (G2 grid) is also finer than that used by Dennis and Young (2003) and D’Alessio and Dennis (1994). This is particularly puzzling since the agreement for Re = 5 and P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 1687 Table 1 Domain independence study for the elliptic cylinders (E = 0.2 and 5) Domain size CDP CDF D∞ /(2a) n = 0.2 500 700 900 1000 1200 E = 0.2, Re = 0.01 2367.0033 – 2366.9577 360.7814 – 348.4848 – 344.0414 2366.8682 340.1520 100 200 300 E = 0.2, Re = 5 5.3891 5.3891 5.3891 n=1 n = 1.8 n = 0.2 CD n=1 n = 1.8 n = 0.2 n=1 n = 1.8 – 37.9013 37.5075 37.3819 – 145.4178 143.8668 – – 143.8606 – 75.6175 73.0404 72.1091 71.2940 – 12.9681 12.8333 12.7903 – 2512.3756 2510.8274 – – 2510.7288 – 436.399 421.5253 416.1505 411.446 – 50.8695 50.3408 50.1723 – 3.2602 3.2268 3.2161 2.5220 2.4837 2.4720 0.2757 0.2757 0.2757 0.6196 0.6138 0.6119 0.7699 0.7583 0.7547 5.6648 5.6649 5.6649 3.8799 3.8407 3.8280 3.2919 3.2421 3.2268 250 300 E = 0.2, Re = 40 1.5311 1.4747 1.5308 1.4735 1.4367 1.4351 0.01916 0.01916 0.1594 0.1593 0.3086 0.3082 1.5503 1.5499 1.6341 1.6328 1.7453 1.7433 500 700 900 1000 1200 E = 5, Re = 0.01 1286.0257 – 1286.0439 84.2902 – 80.8679 – 79.5933 1286.0439 78.5353 – 9.7702 9.6437 9.5987 – 2911.2636 2911.2591 – – 2911.2591 – 427.8410 410.4807 404.0138 398.6456 – 12.9681 12.8333 12.7903 – 4197.2893 4197.303 – – 4197.303 – 512.1312 491.3487 483.6072 477.1810 – 55.9885 55.2629 55.0051 – 100 200 300 E = 5, Re = 5 2.8317 2.8318 2.8318 – 1.0137 1.0085 0.7418 0.7272 0.7228 5.9005 5.9006 5.9005 – 4.0817 4.063 3.0461 2.9877 2.9700 8.7323 8.7324 8.7324 – 5.0954 5.0715 3.7880 3.7149 3.6929 200 300 E = 5, Re = 40 0.5395 0.5395 0.4412 0.4403 0.4050 0.4038 0.8068 0.8068 1.2025 1.2003 1.3025 1.2987 1.3463 1.3463 1.6437 1.6406 1.7076 1.7026 n=1 n = 1.8 Table 2 Grid independence study for Re = 0.01 and 40 CDP Grid details Grid Ncell Nc /(2a) n = 0.2 CDF n=1 n = 1.8 n = 0.2 CD n=1 n = 1.8 n = 0.2 G1 G2 G3 10 000 24 000 30 000 100 200 200 0.0100 0.0018 0.0013 Re = 40, E = 0.2 1.5094 1.4621 1.5327 1.5044 1.5333 1.5071 1.4166 1.4810 1.4849 0.0369 0.0218 0.0210 0.2042 0.1671 0.1650 0.3888 0.3269 0.3238 1.5463 1.5545 1.5543 1.6663 1.6716 1.6721 1.8055 1.8079 1.8087 G1 G2 G3 16 000 22 000 33 600 100 200 240 0.0070 0.0015 0.0010 Re = 40, E = 5 0.5472 0.4411 0.5393 0.4444 0.5390 0.4442 0.4031 0.4094 0.4088 0.8017 0.8067 0.8076 1.2131 1.2097 1.2103 1.3235 1.3157 1.3168 1.3489 1.3460 1.3467 1.6543 1.6541 1.6545 1.7266 1.7251 1.7256 G2 G3 107 800 128 000 200 200 0.0018 0.0013 Re = 0.01, E = 0.2 2374.6020 333.5823 2377.2439 334.5221 36.4118 36.3234 134.6189 132.3215 70.9433 70.0080 12.6726 12.6461 2509.221 2509.5654 404.5256 404.5302 49.0845 48.9696 G2 G3 61 200 65 200 200 240 0.0015 0.0010 Re = 0.01, E = 5 1295.9740 77.6424 1302.2717 78.4666 9.5332 9.5578 2901.1760 2894.9470 394.1143 393.2763 45.0962 45.0709 4197.1510 4197.2187 471.7567 471.7429 54.6294 54.6288 Ncells is the number of cells in the computational domain; Nc is the number of points on the surface of the cylinder and is the grid spacing in the vicinity of the cylinder. 20 is excellent, but the two values show a significant difference only for Re = 40. No more explanation can be given for this discrepancy at this stage. Bearing in mind the aforementioned factors coupled with the fact that the numerical predictions for power-law fluids tend to be intrinsically less accurate (owing to the interaction between the two non-linear terms), it is perhaps 1688 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 Table 3 Comparison of Newtonian (n = 1) drag results for elliptical cylinders (E = 0.2 and 5) E = 0.2 Re Present Dennis and Young (2003) D’Alessio and Dennis (1994) 5 20 40 3.8289 2.0802 1.6319 3.854 2.119 1.876 3.862 2.140 – Dennis and Young (2003) 8.096 2.712 1.7650 1.169 0.7890 Dennis and Chang (1969)a ) 8.222 – 1.848 1.228 0.794 E=5 Re Present 1 8.1105 5 2.7361 10 1.7681 20 1.1680 40 0.7860 a Values as cited by Dennis and Young (2003). reasonable to expect the present results for power-law fluids to be probably reliable to within ±2–3%. 6.2. Drag phenomena For unconfined flow, the scaling of the field equations and the boundary conditions suggests the individual and total drag coefficients to be functions of the Reynolds number (Re), flow behaviour index (n) and the aspect ratio (E) of the elliptical cylinder. These dependences have been explored in the ensuing sections. 6.2.1. Pressure drag coefficient (CDP ) Table 4 shows the dependence of the pressure and friction drag coefficients on the Reynolds number, power-law index and the aspect ratio of the elliptical cylinder. For a fixed value of the Reynolds number, the value of the pressure drag coefficient, CDP decreases as the fluid behaviour changes from shear-thinning (n < 1) to Newtonian (n = 1), and finally to shear-thickening (n > 1), irrespective of the shape of the cylinder. Furthermore, the influence of the power-law index gradually diminishes as the value of the Reynolds number is progressively increased. Conversely, the effect of the flow behaviour index is more prominent at low Reynolds numbers than that at high Reynolds numbers, a trend which is also consistent with that seen for spheroidal and spherical particles (Chhabra, 2006). This is simply due to the fact that with the increasing value of the Reynolds number, inertial forces far outweigh the viscous forces whence it hardly matters whether the fluid is shear-thinning or shear-thickening or Newtonian. For a fixed value of the power-law index, the value of CDP is seen to increase with a decrease in the Reynolds number for all values of aspect ratio E. For instance, the value of CDP at n = 0.2 increases by a factor of 1537, 1960 and 2392 for E = 0.2, 1 and 5 as the Reynolds number is decreased from 40 to 0.01. The corresponding changes in the value of CDP at n = 1 are almost an order of magnitude smaller than these values and even smaller changes are seen for an extremely shear-thickening fluid with n = 1.8. It can clearly be seen that the dependence of the pres- sure drag coefficient on the Reynolds number, power-law index and the aspect ratio is highly non-linear. The present numerical data for the pressure drag coefficient (CDP ) over the range of conditions (0.01 Re 40, 0.2 n 1.8 and 0.2 E 5) can be best represented by the following relationship: log(10CDP Rem ) = f1 (n)f2 (Rem , n)f3 (Rem , Em ). (18) The functions f1 (n), f2 (Rem , n) and f3 (Rem , Em ) are given as f1 (n) = 7.55 − 1.35n2 + 0.40n3 , f2 (Rem , n) = 0.998 1.145 − 0.32n + 0.26n2 0.118 0.0655 , −0.055n + − X X2 3 f3 (Rem , Em ) = − 0.366 + 0.103X + 0.032X 2 0.758 0.032 − 0.002X 3 + − , 2 Em Em (19) (20) (21) where X = log(10Rem ) and the modified Reynolds number (Rem ) and modified aspect ratio (Em ) are represented as follows: √ 100 Rem = (22) and Em = 1 + E. Re Eq. (18) shows maximum and average deviations (max and avg ) of 14.72% and 4.37%, respectively, for the 360 individual data points. There are, however, only 39 data points exhibiting deviations greater than 10%, generally relating to combination of high Re, low n and small value of E. 6.2.2. Friction drag coefficient (CDF ) The variation of the friction drag coefficient with the Reynolds number, power-law index and aspect ratio is also included in Table 4. For a fixed value of the Reynolds number, the dependence of the friction drag coefficient on the power-law index shows a non-monotonous behaviour in both shear-thinning and shear-thickening fluids. The key trends can be summarized as follows: for a fixed value of the aspect ratio and for weakly non-Newtonian fluids, the friction drag coefficient is nearly independent of the power-law index at low Reynolds numbers. As the Reynolds number is gradually increased, the friction drag coefficient transits via an almost power-law index independent behaviour to an upward trend wherein the value of CDF increases slowly with the increasing value of the power-law index over the entire range. Broadly, the larger the Reynolds number, weaker is the dependence on the power-law index. These trends are clearly due to the nonlinear nature of the viscosity over the surface of the cylinder (Bharti et al., 2006). The viscosity for a shear-thinning fluids (n < 1) becomes very large as the shear rate decreases and hence it tends to infinity faraway from the cylinder. Therefore, the viscous effects dominate even faraway from the cylinder. It is, however, appropriate to add here that this stems from the unrealistic fluid behaviours predicted by the simple powerlaw fluid in the limit of zero shear rate. However, on the P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 1689 Table 4 Dependence of the pressure and friction drag coefficients on the Reynolds number (Re), power-law index (n) and aspect ratio (E) n Re = 0.01 20 30 40 Re = 0.01 E = 0.2, pressure drag coefficient (CDP ) 0.2 2375.9478 237.5592 23.9572 5.3884 3.1428 0.6 1346.3304 134.8532 14.3732 3.9490 2.5917 1 333.5823 44.7364 8.1027 3.1779 2.3274 1.4 86.8510 19.9672 5.5220 2.7318 2.1454 1.8 36.4118 11.5126 4.2218 2.4355 1.9996 2.0690 1.8738 1.8079 1.7450 1.6763 1.7258 1.6119 1.5917 1.5628 1.5197 1.5453 1.4645 1.4615 1.4475 1.4169 E = 0.2, friction drag coefficient (CDF ) 134.7223 13.4686 1.3479 0.2593 0.1067 0.0401 0.0257 0.0191 170.2879 17.0571 1.8124 0.4610 0.2619 0.1436 0.1008 0.0788 70.9433 9.5105 1.7035 0.6121 0.4004 0.2570 0.1952 0.1596 24.6542 5.6621 1.5386 0.6971 0.4984 0.3503 0.2805 0.2373 12.6726 4.0046 1.4298 0.7533 0.5709 0.4257 0.3532 0.3064 E = 0.5, pressure drag coefficient (CDP ) 0.2 2186.3272 218.6630 22.0423 4.8996 2.8187 0.6 1145.1463 114.6440 12.2512 3.3947 2.2306 1 266.1707 36.0942 6.5975 2.6169 1.9253 1.4 68.3911 15.8394 4.4109 2.1982 1.7364 1.8 28.5659 9.0728 3.3411 1.9397 1.6009 1.7678 1.5974 1.5037 1.4255 1.3544 1.4390 1.3677 1.3311 1.2872 1.2383 1.2866 1.2411 1.2289 1.2015 1.1639 E = 0.5, friction drag coefficient (CDF ) 367.2389 36.7219 3.6769 0.7288 0.3424 0.1375 0.0805 0.0568 398.4208 39.8847 4.2423 1.0725 0.6143 0.3448 0.2430 0.1895 138.0381 18.7110 3.3707 1.2111 0.7945 0.5166 0.3973 0.3278 43.2766 10.0226 2.7454 1.2482 0.8960 0.6372 0.5165 0.4419 20.6677 6.5797 2.3684 1.2551 0.9564 0.7214 0.6054 0.5309 E = 1, pressure drag coefficient 0.2 1951.2650 195.1465 19.6615 0.6 924.8634 92.6154 9.9524 1 206.7774 27.9680 5.1847 1.4 52.8495 12.1910 3.4327 1.8 21.9883 7.0059 2.5970 (CDP ) 4.3221 2.8040 2.0897 1.7245 1.5125 2.4388 1.8440 1.5411 1.3657 1.2508 1.4842 1.3040 1.2007 1.1240 1.0616 1.1516 1.1003 1.0607 1.0174 0.9742 0.9954 0.9899 0.9790 0.9524 0.9195 E = 1, friction drag coefficient 723.8972 72.3863 7.2447 681.9876 68.2901 7.2864 210.5813 28.4644 5.1708 61.8726 14.2860 3.9494 28.0521 8.9808 3.2567 (CDF ) 1.4534 1.8465 1.8613 1.7967 1.7264 0.7178 1.0650 1.2241 1.2911 1.3171 0.3394 0.6139 0.8047 0.9249 0.9997 0.2069 0.4406 0.6262 0.7567 0.8457 0.1443 0.3469 0.5223 0.6537 0.7479 E = 2, pressure drag coefficient 0.2 1645.7317 164.5780 16.5779 0.6 674.0869 67.5313 7.3550 1 141.3820 19.5365 3.7597 1.4 36.4728 8.6388 2.4962 1.8 15.6546 5.0639 1.9058 (CDP ) 3.6262 2.1456 1.5597 1.2730 1.1143 2.0132 1.4153 1.1597 1.0068 0.9195 1.1840 0.9900 0.8893 0.8235 0.7767 0.8965 0.8220 0.7781 0.7414 0.7102 E = 2, friction drag coefficient 0.7461 1369.7311 136.9786 13.7023 0.7271 1075.3721 107.7219 11.5905 0.7126 287.8216 39.7249 7.4070 0.6912 80.0214 18.9939 5.3515 0.6689 35.5244 11.5752 4.2452 (CDF ) 2.7638 2.9717 2.6877 2.4364 2.2443 1.3956 1.7262 1.7714 1.7496 1.7087 0.7020 1.0134 1.1734 1.2581 1.2999 0.4657 0.7424 0.9219 1.0359 1.1053 0.3455 0.5943 0.7761 0.9012 0.9834 E = 5, pressure drag coefficient 0.2 1295.9744 129.6308 13.0755 0.6 402.9162 40.4019 4.5800 1 77.6424 11.0763 2.3102 1.4 20.9627 5.1494 1.5731 1.8 9.5332 3.1774 1.2359 (CDP ) 2.8535 1.4367 1.0085 0.8210 0.7228 1.5725 0.9571 0.7456 0.6430 0.5871 0.9017 0.6645 0.5668 0.5142 0.4839 0.6653 0.5436 0.4879 0.4549 0.4347 E = 5, friction drag coefficient (CDF ) 0.5417 2901.1761 290.1007 29.0330 5.8819 2.9996 0.4737 1736.2927 174.0588 19.1987 5.0654 2.9700 0.4403 394.1143 56.0330 10.9978 4.0630 2.6872 0.4183 103.8442 25.5858 7.4535 3.4049 2.4412 0.4038 45.0962 15.1635 5.6513 2.9700 2.2509 1.5469 1.7731 1.7913 1.7569 1.7091 1.0549 1.3172 1.4166 1.4510 1.4555 0.8053 1.0687 1.2003 1.2673 1.2987 0.1 1 5 10 other hand, even if one uses the other fluid models (like Ellis or Carreau), the value of the viscosity will approach the so-called zero shear viscosity faraway from the cylinder. Of course, the shear-thickening fluid will exhibit maximum viscosity near the cylinder. Due to this large viscosity (faraway from the cylinder) in shear-thinning fluids, the value of the friction drag coefficient was seen to be larger than that in shear-thickening fluids at low Reynolds numbers. On the other hand, for shear-thickening (n > 1) fluids, the inertial effects dominate faraway from the cylinder, which therefore shows an opposite dependence of the drag coefficient on the power-law index in the low Reynolds number region. The present numerical data for the friction drag coefficient (CDF ) over the range of conditions (0.01 Re 40, 0.2 n 1.8 and 0.2 E 5) can be best represented by the following closure relationship: log(50CDF Rem ) = f1 (n)f2 (Rem , n)f3 (Rem , Em ). (23) The functions f2 (Rem , n) and f3 (Rem , Em ) are given as f2 (Rem , n) = 0.586[−0.329 + (0.42 − 0.015n)X + 0.007X 2 + 0.19n + 0.01n2 ], (24) 0.1 1 5 10 20 30 2.495 0.0868 − f3 (Rem , Em ) = − 0.386 + X X2 3.44 2.295 1 + 3.59 − − , 2 Em X Em 40 (25) where X = log(10Rem ) and the function f1 (n), the modified Reynolds number (Rem ) and the modified aspect ratio (Em ) are given by Eqs. (19) and (22), respectively. Eq. (23) shows maximum and average deviations (max and avg ) of 60.26% and 3.68%, respectively, for the 360 individual data points. There are, however, 15 data points exhibiting deviations greater than 15%, only one data point with > 50%, relating to Re = 40, n = 0.2 and E = 0.2, respectively. Further examination of the residuals revealed these points showing large deviations relate to certain combinations involving high values of Re (∼ 30–40) and/or small values of n (∼ 0.2) and/or small values of E (∼ 0.2). If these 15 data points are excluded, the maximum and average deviations reduce drastically to 14.92% and 3.05%, respectively, for the remaining 345 individual data points. 6.2.3. Total drag coefficient (CD ) The total drag coefficient (sum of the pressure and friction drag coefficients, Eq. (13)) is plotted in Fig. 2 for a range of 1690 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 Fig. 2. Effect of Reynolds number (Re), power-law index (n) and aspect ratio (E) on the total drag coefficient (CD ). Fig. 3. Effect of Reynolds number (Re), power-law index (n) and aspect ratio (E) on the drag ratio (CDP /CDF ). P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 1691 values of the Reynolds number, power-law index and aspect ratio. Once again, the dependence is seen to be similar to that of CDP and CDF shown in Table 4. The dependence of the total drag coefficient, CD on the power-law index is seen to be stronger in shear-thinning (n < 1) fluids than that in shearthickening (n > 1) fluids. Consequently, at low Reynolds numbers (0.01 Re 1), significant variation in the drag values can be seen in shear-thinning fluids than that in Newtonian and in shear-thickening fluids. At high values of the Reynolds number (20 Re 40), the value of CD increases with an increase in the power-law index for 0.2 E 5; however, this effect gets accentuated at E = 2 and 5. These trends are qualitatively consistent with the literature values for a circular cylinder (Bharti et al., 2006; Chhabra et al., 2004; Soares et al., 2005). The total drag coefficient (CD ) over the range of the Reynolds number (0.01 Re 40), power-law index (0.2 n1.8) and aspect ratio (0.2 E 5) can be best represented by the following relation: log(CD Rem ) = f1 (n)f2 (Rem , n)f3 (Rem , Em ), (26) where the functions f2 (Rem , n) and f3 (Rem , Em ) are given as f2 (Rem , n) = 0.26X + 0.03n − 0.06, 2 f3 (Rem , Em ) = 1.74 − 1.17Em + 0.41Em + Fig. 4. Effect of Reynolds number (Re), power-law index (n) and aspect ratio N ). (E) on the normalized total drag coefficient (CD (27) 0.106 , X (28) where X = log(Rem ). The function f1 (n), the modified Reynolds number (Rem ) and the modified aspect ratio (Em ) are given by Eqs. (19) and (22), respectively. Eq. (26) reproduces 360 numerical data points with the maximum and average deviations (max and avg ) of 33.93% and 4.77%, respectively. There are, however, only 11 data points exhibiting deviations greater that 10%. Further examination of the residuals revealed these points to relate to the same combinations of high values of Re (∼ 30–40) and/or small values of n (∼ 0.2) and/or small values of E (∼ 0.2), as noted earlier. If these 11 data points are excluded, the maximum and average deviations reduce to 8.83% and 3.48%, respectively, for the remaining 349 individual data points. It needs to be stressed here that no prior predictive equation exist in the literature even for the flow of Newtonian fluids over an elliptical cylinder over these ranges of Re and E. Admittedly, some attempts have been made to consolidate the results for power-law fluid flow in tube bundles by Bruschke and Advani (1993) and Woods et al. (2003) by using different length scales with limited success. Not only the length scale depends upon the solid volume fraction of arrays and/or on n, but these approaches have been tested for n < 1 and/or E < 1 and in the creeping flow limit (Re = 0). In the present case, the results relate to n < 1 and n > 1, Reynolds number up to 40 and 0.2 E 5, it is far from obvious whether a single choice of length scale will work under all conditions. On the other hand, the correlations developed here based on the length and velocity scales which emerge naturally by the non-dimensionalization of the governing equations and boundary conditions. Also, these expression embrace much 1692 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 a b Y 0.6 n=0.2 c 0.6 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 1 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=0.2 -0.5 0 0.5 n=0.2 -0.2 -0.1 1 0 0.1 0.2 0.3 X 0.6 0.6 n=1.0 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 0.6 -0.5 1 n=1.8 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 1 0 0.5 0.6 0.4 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=1.0 1 n=1.8 -0.5 0 0.5 1 n=1.0 -0.2 0 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 0.2 n=1.8 -0.2 0 0.2 Fig. 5. Streamline (upper half) and vorticity (lower half) profiles for Re = 0.01. (a) E = 0.2, (b) E = 1, (c) E = 5. wider range of Reynolds number (Re) than that in the study of Bruschke and Advani (1993), Re = 0, E = 1 and of Woods et al. (2003), Re = 0, E < 1. Further examination of the present results in terms of the variation of the drag ratio (CDP /CDF ) on the Reynolds number (Re), power-law index (n) and aspect ratio (E) revealed this ratio (CDP /CDF ) to decrease with an increase in the power-law index (Fig. 3) for the ranges of conditions studied in this work. Qualitatively, this effect can be explained as follows: for a fixed n . For a shearvalue of E and Re, the viscous forces scales as U∞ thinning fluid (n < 1), as the value of n is gradually increased, the viscous forces will rise and hence the friction will increase. Similarly, for a shear-thickening fluid (n > 1), viscous forces will rise even more steeply. In either event, one would expect the ratio (CDP /CDF ) to decrease as seen in Fig. 3. A rapid decrease in the ratio with an increase in power-law index was seen at low Reynolds numbers and/or aspect ratio in shear-thinning fluids thereby suggesting rather strong dependence on the aspect ratio, E. Broadly, the value of the ratio (CDP /CDF ) was > 1 for E 1, which changes to (CDP /CDF ) 1 as the value of the aspect ratio E exceeds 1. This shows the role of shape. Thus, the pressure drag dominates for E 1 which acts as a more of a bluff body than a circular cylinder thereby resulting in the sudden and significant bending of streamlines and showing poor pressure recovery. On the other hand, E > 1 behaves more like a streamlined surface wherein a fluid element can easily follow the cylinder surface whence reducing the role of pressure forces. Finally, in order to elucidate the role of power-law rheology on the cylinder drag in an unambiguous manner, drag coefficient for an elliptic cylinder in power-law fluids is normalized N =C using the corresponding Newtonian value (CD D,n /CD,n=1 ) at the same values of the Reynolds number and aspect ratio. Fig. 4 shows the dependence of the normalized drag coefficient on the Reynolds number and power-law index. It is seen that the shear-thinning behaviour causes an increase in the value of the drag coefficient which diminishes with the increasing Reynolds number. As expected, not only this trend flips over in shearthickening fluids, but the effect is also seen to be somewhat suppressed. This figure clearly shows the effect of power-law index to be much more prominent at low Reynolds numbers than that at high Reynolds numbers. Also, the value of the drag coefficient is seen to be enhanced much more than that for a sphere where this ratio reaches a maximum value of ∼ 1.5 at n ≈ 0.4 as opposed to the value of 9 seen here for the extreme case of n = 0.2. P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 a b Y 0.6 n=0.2 c 0.6 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 1693 1 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=0.2 -0.5 0 0.5 n=0.2 -0.2 -0.1 1 0 0.1 0.2 0.3 X 0.6 0.6 n=1.0 0.4 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 0.6 -0.5 1 n=1.8 0.4 0.2 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 1 0 0.5 0.6 0.4 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=1.0 1 n=1.8 -0.5 0 0.5 1 n=1.0 -0.2 0 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 0.2 n=1.8 -0.2 0 0.2 Fig. 6. Streamline (upper half) and vorticity (lower half) profiles for Re = 1. (a) E = 0.2, (b) E = 1, (c) E = 5. 6.3. Detailed flow kinematics Evidently, the dependence of the individual and total drag coefficients seen in the previous section should also be reflected in similar differences in the detailed kinematics of the flow close to the cylinder. In particular, the streamline and vorticity profiles, recirculation length, variation of pressure coefficients over the surface of the cylinder, and their values at the stagnation points are analysed to gain further physical insights into the nature of the flow. 6.3.1. Streamline and vorticity profiles Representative streamline (upper half) and vorticity (lower half) profiles close to the elliptical cylinder (E = 0.2, 1 and 5) for four values of the Reynolds number of 0.01, 1, 10 and 40 and for three values of the power-law index of 0.2, 1 and 1.8 are shown in Figs. 5–8. For a fixed value of power-law index (n), flow is seen to be somewhat retarded as the value of the aspect ratio E increases. This is clearly due to fact that with increasing deviation in the value of E from 1, the shape of the cylinder corresponds to a vertical (E → 0) or to a horizontal (E → ∞) flat plate-like obstructions to the flow. For a fixed value of E, the flow is seen to accelerate with an increase in the power-law index and/or Reynolds number and/or both. The effect is more visible in shear-thinning (n < 1) fluids and/or for E < 1 than that in shear-thickening (n > 1), or in Newtonian (n=1) and/or for E 1. As expected. there is no evidence of flow separation at Re = 0.01 for the range of power-law index and aspect ratios considered here. These figures also show the decreasing values of the Reynolds number, at which the separation occurs, with a decrease in the value of the aspect ratio E for both shear-thinning and shear-thickening fluids. For instance, for a cylinder with E = 0.2, a significant wake can clearly be seen at Re = 1, while no flow separation is seen for E = 1 and 5 at this value of the Reynolds number, Fig. 6. Irrespective of the type of fluid behaviour, based on the loss of fore and aft symmetry and/or flow separation, the limiting value of the Reynolds number denoting the end of the creeping flow regime shows a drastic reduction as the value of the aspect ratio E is decreased. As noted earlier, this behaviour is clearly due to the fact that a cylinder with E < 1 behaves like a bluff body whereas E > 1 configuration shows some features of streamline body. Furthermore, shear-thinning behaviour delays the separation 1694 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 a b n=0.2 0.2 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=0.2 0.4 0.2 0 Y c 0.6 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 1 -0.5 0 0.5 n=0.2 -0.2 -0.1 1 0 0.1 0.2 0.3 X 0.6 n=1.0 0.2 0.4 0 0.2 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=1.0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 n=1.8 0.2 -0.5 1 0 0.5 0.6 1 n=1.8 0.4 0.2 0 0 -0.2 -0.2 -0.4 -0.4 -0.5 0 0.5 1 -0.5 0 0.5 1 n=1.0 -0.2 0 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 0.2 n=1.8 -0.2 0 0.2 Fig. 7. Streamline (upper half) and vorticity (lower half) profiles for Re = 10. (a) E = 0.2, (b) E = 1, (c) E = 5. while it occurs earlier in shear-thickening fluids than that in Newtonian fluids for a fixed value of the aspect ratio. This feature is qualitatively consistent with that seen for a circular cylinder. Qualitatively, it can be rationalized as follows: Since the effective shear rate is maximum close to the cylinder, the viscosity is minimum in this region for a shear-thinning fluids (and of course, maximum for a shear-thickening fluid). The velocity field is known to decay faster in a shear-thinning fluid which leads to a drastic reduction in shear rate, i.e., high viscosity. In other words, even slightly away from the cylinder, highly viscous slow moving liquid meets with the low viscosity fluid in a thin layer encapsulating the cylinder. Thus, the separation is somewhat delayed in this case. On the other hand, for n > 1, the viscosity is maximum near the obstacle and lower elsewhere so this less viscous fast moving fluid stream leads to an early flow separation. 6.3.2. Recirculation length (Lw ) The phenomena of flow recirculation and separation in the rear of the cylinder are often quantified in terms of the recirculation (or wake) length and the angle of separation, respectively. Fig. 9 shows the dependence of the recirculation length on the power-law index (0.2 n1.8), Reynolds num- ber (0.01 Re 40) and aspect ratio (0.2 E 5) of elliptical cylinders. For a fixed value of the Reynolds number, the length of the recirculation zone decreases with an increase in the value of the aspect ratio E, irrespective of the type of flow behaviour of the fluid. No flow separation was observed at Re = 0.01 for any value of power-law index (n) and aspect ratio (E) studied here (Figs. 5 and 9), at Re = 1 for all values of value of powerlaw index n and the aspect ratio E 1 (except at n = 1.8 and E = 1, see Figs. 6 and 9) and for E = 5 in shear-thinning fluids (E = 5, Fig. 9) for all values of the Reynolds number. For a fixed value of the aspect ratio E, the value of the recirculation length (Lw ) decreases with the decreasing power-law index for shear-thinning fluids, whereas it increases with the increasing degree of shear-thickening. The wake grows faster with an increase in the value of E for E > 1 than that for E < 1 as the fluid behaviour changes from Newtonian to shear-thickening; the change in behaviour from Newtonian to shear-thinning shows the opposite dependence on E. In summary, the wake tends to be smaller in shear-thinning fluids and/or for E < 1 than that in shear-thickening fluids and/or E > 1 under otherwise similar conditions. These trends are obviously consistent with the plausible considerations presented in the preceding section. P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 a b 1 n=0.2 0.4 0.5 Y c 0.6 n=0.2 0.2 0 0 -0.2 -0.5 1695 -0.4 n=0.2 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 -1 0 0.5 1 1.5 2 2.5 -0.5 0 0.5 -0.2 -0.1 1 0 0.1 0.2 0.3 X 1 0.6 n=1.0 n=1.0 0.4 0.5 0.2 0 0 -0.2 -0.5 -0.4 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 n=1.0 -1 0 0.5 1 1.5 1 2 2.5 -0.5 0 0.5 0.6 n=1.8 n=1.8 0.4 0.5 0.2 0 0 -0.2 -0.5 -0.2 -0.1 1 -0.4 0 0.1 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 0.2 0.3 n=1.8 -1 0 0.5 1 1.5 2 2.5 -0.5 0 0.5 1 -0.2 -0.1 0 0.1 0.2 0.3 Fig. 8. Streamline (upper half) and vorticity (lower half) profiles for Re = 40. (a) E = 0.2, (b) E = 1, (c) E = 5. 6.3.3. Surface pressure coefficients (Cp ) Representative variation of the pressure coefficient over the surface of an elliptic cylinder is shown in Figs. 10–12 for a range of values of the power-law index, for four values of the Reynolds number as Re = 0.01, 1, 10 and 40 and for three values of the aspect ratio as E = 0.2, 1 and 5, respectively. These figures suggest the stronger dependence of the pressure profiles on the shape (E) of the cylinder than that on Reynolds number (Re) and/or on the power-law index (n). This finding is also consistent with that for spheroidal particles (Chhabra, 2006). The effect of power-law index (n) on the value of Cp is seen to be stronger in shear-thinning (n < 1) fluids and/or in flow without separation than that in shear-thickening (n > 1) fluids and/or the high Reynolds number flow. The curves show qualitatively similar patterns irrespective of the aspect ratio of the elliptical cylinder and the behaviour of the fluid. At low Reynolds numbers and for E < 1, it can be seen from Fig. 10 that surface pressure is almost constant in the front and the rear half of cylinder with a step change at = /2. In contrast for E = 1 (circular cylinder), the pressure profile patterns are significantly different (Fig. 11) than that for E < 1 (Fig. 10) for the flow with separation in the rear of the cylinder. In this case, the pressure is seen to decrease from its maximum value at the front stagnation point ( = 0)) along the surface of the cylinder followed by an increase due to the flow recirculation in the rear side of the cylinder. In the absence of separation, pressure continues to decrease up to the rear stagnation point (=). For E > 1, the pressure decreases rapidly near the front stagnation point to a minimum (Fig. 12), which varies very little thereafter in the rear side thereby following the surface of the streamlined shape of the elliptical cylinder. In the front of the cylinder, the rate of decrease of Cp increases with an increase in the value of E which is clearly due to the varying extents of the bending of streamlines. The fore and aft symmetry in the variation of Cp in the rear and front of the cylinders can be seen under the conditions without flow separation, Figs. 10–12, respectively, i.e., the crossover (Cp =0) occurs at =/2 of the cylinder. This crossover point gradually shifts forward with the increasing Reynolds number thereby suggesting early separation. For a fixed value of the aspect ratio (E) and Reynolds number (Re), the shear-thinning (n < 1) fluids always show a higher value of the pressure in the front of the cylinder than that for Newtonian fluids with a flip over in the downstream side of the cylinder, e.g., see Figs. 10–12 (left side). The shear-thickening (n > 1) fluids show the opposite dependence of the Cp profiles on the power-law index. The smaller the value of the power-law index 1696 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 Fig. 9. Dependence of recirculation length (Lw ) on Reynolds number (Re), power-law index (n) and aspect ratio (E). (n), the higher is the value of Cp (see Figs. 10–12 (right side)). For a fixed value of the power-law index (n) and the aspect ratio (E), the value of the pressure coefficient is seen to be higher at low Reynolds numbers, which decreases with an increase in the Reynolds number (Re). Fig. 13(a) shows the dependence of the front stagnation pressure coefficient Cp (0) on the Reynolds number, power-law index and the aspect ratio. The maximum value of Cp (0) was seen at low Reynolds numbers and low power-law index. For a fixed value of the Reynolds number (Re) and the aspect ratio (E), Cp (0) is seen to decrease with an increase in the powerlaw index at low Reynolds numbers, irrespective of the type of the fluid behaviour. As the Reynolds number is progressively increased, the rate of increase in the value of Cp (0) with an increase in the power-law index was seen to slow down and only a slight change can be seen at higher Reynolds numbers. At higher values of Re, Cp (0) is seen to be weakly dependent on the power-law index. The rate of change in the front stagnation pressure coefficient with power-law index was seen to be almost constant at low value of the Reynolds number. P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 1697 Fig. 10. Effect of Reynolds number and power-law index on Cp over the surface of the elliptical cylinder (E = 0.2). For a fixed value of power-law index (n) and the Reynolds number (Re), Cp (0) shows an increase with an increase in the value of the aspect ratio (E), irrespective of the fluid behaviour. This increase was seen to be higher in shear-thickening fluids and/or E > 1 than that in shear-thinning and Newtonian fluids and/or E < 1. The functional dependence of the front stagnation pressure coefficient, Cp (0) on the Reynolds number (0.01 Re 40), power-law index (0.2 n 1.8) and aspect ratio (0.2 E 5) can be best represented by the following relation: log[Cp (0)Rem ] = f1 (n)f2 (Rem , n)f3 (Rem , Em ), (29) 1698 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 Fig. 11. Effect of Reynolds number and power-law index on the variation of Cp over the surface of the elliptical cylinder (E = 1). where the functions f2 (Rem , n) and f3 (Rem , Em ) are given as f2 (Rem , n) = 0.96 − 0.133n + 0.12n2 + 0.02 , X f3 (Rem , Em ) = 0.126 + 0.177X + 0.015X 2 0.31 0.174 − + , 2 Em Em (30) (31) where X=log(Rem ). The function f1 (n), the modified Reynolds number (Rem ) and the modified aspect ratio (Em ) are expressed by Eqs. (19) and (22), respectively. Eq. (29) shows maximum and average deviations (max and avg ) of 13.29% and 3.59%, respectively, for the 360 individual data points. Only 4 data points show > 10%. The dependence of the rear stagnation pressure coefficient, Cp () on the Reynolds number, power-law index and the aspect ratio, shown in Fig. 13(b) is qualitatively similar, except for the fact that it is negative. In addition, the crossover of the values can be seen with an increasing Reynolds number. The functional dependence of the rear stagnation pressure P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 1699 Fig. 12. Effect of Reynolds number and power-law index on the variation of Cp over the surface of the elliptical cylinder (E = 5). coefficient, Cp () on the Reynolds number (0.01 Re 40), power-law index (0.2 n1.8) and aspect ratio (0.2 E 5) can be best represented by the following relation: log[−10Cp ()Rem ] = f1 (n)f2 (Rem , n)f3 (Rem , Em ), (32) where the functions f2 (Rem , n) and f3 (Rem , Em ) are given as f2 (Rem , n) = 0.99 1.162 − 0.16n + 0.05n2 0.65 0.47 0.257n − + 2 + , (33) X X X 1700 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 is increased, the flow is governed by two non-linear terms, namely, inertial and viscous, which scale differently with velocity. For instance, one can argue that the viscous forces will n whereas the inertial forces as ∼ approximately scale as ∼ U∞ 2 U∞ . Thus, keeping everything else fixed, the decreasing value of power-law index (n) will suggest diminishing importance of the viscous effects for shear-thinning fluids, while the inertial 2 ). On the other hand, viscous efterm will still scale as (∝ U∞ fects are likely to grow with the increasing value of the powerlaw index (n) for a shear-thickening fluid. For the extreme case 1.8 , almost of n = 1.8, the viscous terms will also scale as ∼ U∞ identical to the inertial term. It is believed that these different kinds of dependencies on the velocity and power-law flow index are also responsible for the non-monotonous behaviour of drag coefficients. Finally, the highly non-linear form of the momentum equations also poses enormous challenges in choosing appropriate numerics and numerical parameters even in the steady flow regime. Hence, the drag and flow patterns show a much more intricate dependence on the Reynolds number (Re) and power-law index (n) than that seen for a Newtonian fluid. 7. Concluding remarks Fig. 13. Effect of Reynolds number, power-law index and aspect ratio of the elliptical cylinders on the (a) front and (b) rear stagnation pressure coefficient. f3 (Rem , Em ) = 0.06 + 0.153X + 0.0095X 2 2 + 0.03Em X − 0.143Em + 0.004Em , (34) where X = log(10Rem ). The function f1 (n), the modified Reynolds number (Rem ) and the modified aspect ratio (Em ) are expressed by Eqs. (19) and (22), respectively. Eq. (32) shows maximum and average deviations (max and avg ) of 7.52% and 1.78%, respectively, for the 360 individual data points. Only 7 data points show > 5%. In summary, the detailed kinematics of the flow and resulting values of the individual and total drag coefficients are seen to be influenced in an complex manner by the values of the Reynolds number, the power-law index and the aspect ratio of the elliptical cylinder. This interplay is further accentuated by the fact that even at low Reynolds numbers, the viscous terms are non-linear for power-law fluids. As the Reynolds number The flow characteristics of the two-dimensional steady flow of incompressible power-law fluids over an elliptical cylinder have been investigated numerically for the range of Reynolds number as 0.01 Re 40, power-law index as 0.2 n 1.8 and for the aspect ratio as 0.2 E 5 including the special case of a circular cylinder (E = 1). Extensive results on flow patterns in terms of the streamline and vorticity profiles adjacent to the cylinder are presented and discussed to gain physical insights into the nature of flow. The onset of wake formation (thereby marking the end of the creeping flow regime) was seen to be delayed with an increase in the aspect ratio (E), i.e., streamlining of the shape and/or with a decrease in the power-law index (n). The early separation was observed in shear-thickening fluids, as well as for the cylinders of E < 1. The shear-thinning behaviour (n < 1) of the fluid decreases the size of the wake; on the other hand, the shear-thickening fluids (n > 1) show the opposite behaviour. Distinct maximum values of the pressure coefficient were seen at the front stagnation point for the cylinders of E 1, while it was nearly constant in the front side of the cylinder for E < 1 for all values of the power-law index (n) and the Reynolds number (Re). At low Reynolds numbers, the front stagnation pressure coefficient was always higher in shear-thinning fluids than that in shear-thickening fluids. The pressure drag coefficient always decreases with an increase in the power-law index, irrespective of the value of Reynolds number, fluid behaviour, or shape of the cylinder. The friction drag coefficient shows a complex dependence on the Reynolds number (Re), power-law index (n) and the aspect ratio (E). Therefore, the total drag coefficient decreases with an increase in power-law index at low Reynolds numbers. In the creeping flow region, the shear-thinning behaviour results in higher value of the drag coefficient than that in shear-thickening fluids. The effect of aspect ratio, E on the drag coefficient in P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 shear-thickening fluids was also modest as compared to that in shear-thinning fluids. Overall, the aspect ratio of the cylinder plays a significant role in influencing the detailed flow patterns and the resulting gross behaviour. The functional dependence of the flow characteristics (drag and stagnation pressure coefficients) on the dimensionless parameters (Re, E and n) has also been presented. Notations a b CD CDF CDP N CD Cp Cp (0) Cp () D∞ e x , ey E FD FDF FDP I2 I2∗ Lw m n nx , ny ns Nc p p∞ p() Re U∞ Ux , Uy x, y semi-axis of the elliptical cylinder normal to the flow, m semi-axis of the elliptical cylinder along the flow, m drag coefficient, dimensionless frictional component of drag coefficient, dimensionless pressure component of drag coefficient, dimensionless normalized drag coefficient, CD,non-Newtonian / CD,Newtonian , dimensionless pressure coefficient, dimensionless value of Cp at front stagnation ( = 0), dimensionless value of Cp at rear stagnation ( = ), dimensionless diameter of the outer boundary, m x- and y-component of the unit vector, dimensionless aspect ratio of the elliptical cylinder, =b/a, dimensionless drag force per unit length of the cylinder, N/m frictional component of the drag force per unit length of the cylinder, N/m pressure component of the drag force per unit length of the cylinder, N/m second invariant of the rate of the strain tensor, s−2 dimensionless second invariant of the rate of the strain tensor dimensionless, =I2 /(U∞ /a)2 recirculation length, dimensionless power-law consistency index, Pa sn power-law flow behaviour index, dimensionless x- andy-component of the direction vector normal to the surface of the cylinder, dimensionless direction vector normal to the surface of the cylinder, dimensionless number of grid points on the surface of the cylinder, dimensionless pressure, Pa free stream pressure, Pa surface pressure at an angle , Pa Reynolds number, dimensionless free stream velocity, m/s x- and y-components of the velocity, m/s streamwise and transverse coordinates, m 1701 Greek symbols ∗ ∗ ∗xx ∗xy component of the rate of the strain tensor, s−1 viscosity, Pa s dimensionless viscosity, =/[m(U∞ /a)(n−1)/2 ] angular displacement from the front stagnation, degree density of the fluid (kg/m3 ) shear stress, Pa dimensionless shear stress, =/[m(U∞ /a)n ] x- component of the dimensionless shear stress, =xx /[m(U∞ /a)n ] y-component of the dimensionless shear stress, =xy /[m(U∞ /a)n ] vorticity, s−1 References Bharti, R.P., Chhabra, R.P., Eswaran, V., 2005. Forced convection heat transfer to non-Newtonian fluids from a heated circular cylinder. In: Proceedings of 58th Annual Session of Indian Institute of Chemical Engineers (CHEMCON 2005), vol. 2, IIT Delhi, New Delhi, India, December 14–17, pp. 16–18. Bharti, R.P., Chhabra, R.P., Eswaran, V., 2006. Steady flow of power-law fluids across a circular cylinder. Canadian Journal of Chemical Engineering 84, 406–421. Bharti, R.P., Chhabra, R.P., Eswaran, V., 2007. Steady forced convection heat transfer from a heated circular cylinder to power-law fluids. International Journal of Heat and Mass Transfer 50(5–6), in press. doi: 10.1016/j.ijheatmasstransfer.2006.08.008. Bruschke, M.V., Advani, S.G., 1993. Flow of generalized Newtonian fluids across a periodic array of cylinder. Journal of Rheology 37, 479–498. Chhabra, R.P., 2006. Bubbles, Drops and Particles in Non-Newtonian Fluids. second ed.. CRC Press, Boca Raton, FL. Chhabra, R.P., Richardson, J.F., 1999. Non-Newtonian Flow in the Process Industries: Fundamentals and Engineering Applications. ButterworthHeinemann, Oxford. Chhabra, R.P., Soares, A.A., Ferreira, J.M., 2004. Steady non-Newtonian flow past a circular cylinder: a numerical study. Acta Mechanica 172, 1–16. D’Alessio, S.J.D., Dennis, S.C.R., 1994. A vorticity model for viscous flow past a cylinder. Computers and Fluids 23, 279–293. Dennis, S.C.R., Young, P.J.S., 2003. Steady flow past an elliptic cylinder inclined to the stream. Journal of Engineering Mathematics 47, 101–120. Dhiman, A.K., Chhabra, R.P., Eswaran, V., 2005. Flow and heat transfer across a confined square cylinder in the steady cross flow regime: effect of Peclet number. International Journal of Heat and Mass Transfer 48, 4598–4614. Dhiman, A.K., Chhabra, R.P., Sharma, A., Eswaran, V., 2006a. Effects of Reynolds and Prandtl numbers on heat transfer across a square cylinder in the steady flow regime. Numerical Heat Transfer, Part A 49, 717–731. Dhiman, A.K., Chhabra, R.P., Eswaran, V., 2006b. Steady flow of powerlaw fluids across a square cylinder. Chemical Engineering Research and Design 84, 300–310. Epstein, N., Masliyah, J.H., 1972. Creeping flow through clusters of spheroids and elliptical cylinders. Chemical Engineering Journal 3, 169–175. Ferreira, J.M., Chhabra, R.P., 2004. Analytical study of drag and mass transfer in creeping power-law flow across tube banks. Industrial and Engineering Chemistry Research 43, 3439–3450. Gupta, A.K., Sharma, A., Chhabra, R.P., Eswaran, V., 2003. Two-dimensional steady flow of a power-law fluid past a square cylinder in a plane channel: momentum and heat transfer characteristics. Industrial and Engineering Chemistry Research 42, 5674–5686. 1702 P. Sivakumar et al. / Chemical Engineering Science 62 (2007) 1682 – 1702 Hasimoto, H., 1958. On the flow of a viscous fluid past an inclined elliptic cylinder at small Reynolds numbers. Journal of Physical Society of Japan 8, 653–661. Ilgarbuis, U.R., Butkus, A., 1988. Hydraulic drag and average heat transfer coefficients of compact bundles of elliptical finned tubes. Heat Transfer Soviet Research 20, 12–21. Imai, I., 1954. A new method of solving Oseen’s equations and its applications to the flow past an inclined elliptic cylinder. Proceeding of Royal Society of London A 224, 141–160. Johnson, S.A., Thompson, M.C., Hourigan, K., 2001. Flow past elliptical cylinders at low Reynolds numbers. In: Proceedings of 14th Australasian Fluid Mechanics Conference, Adelaide University, December 9–14, pp. 343–346. Khan, W.A., Culham, J.R., Yovanovich, M.M., 2005. Fluid flow around and heat transfer from elliptical cylinders: analytical approach. Journal of Thermophysics and Heat Transfer 19, 178–185. Kiya, M., Arie, M., 1975. Viscous shear flow past small bluff bodies attached to a plane wall. Journal of Fluid Mechanics 69, 803–823. Liu, A.W., Bornside, D.E., Armstrong, R.C., Brown, R.A., 1998. Viscoelastic flow of polymer solutions around a periodic, linear array of cylinders: comparisons of predictions for microstructure and flow field. Journal of Non-Newtonian Fluid Mechanics 77, 153–190. Lugt, H.J., Haussling, H.J., 1974. Laminar flow past an abruptly accelerated elliptic cylinder at 45◦ incidence. Journal of Fluid Mechanics 65, 711–734. Masliyah, J.H., 1973. Viscous flow across tube banks of circular and elliptical cylinders: momentum and heat transfer. Canadian Journal of Chemical Engineering 51, 550–554. Meller, N.A., 1978. Viscous flow past an elliptic cylinder. Computational Methods and Mathematics in Physics 18, 138–149. Nishiyama, H., Ota, T., Matsuno, T., 1988a. Forced convection heat transfer from two elliptic cylinders in a tandem arrangement. Transactions of JSME B 52, 2677–2681. Nishiyama, H., Ota, T., Matsuno, T., 1988b. Heat transfer and flow around elliptical cylinders in a tandem arrangement. JSME International Journal, Series II 31, 410–419. Nitin, S., Chhabra, R.P., 2005. Non-isothermal flow of a power-law fluid past a rectangular obstacle (of aspect ratio 1 × 2) in a channel: drag and heat transfer. International Journal of Engineering Science 43, 707–720. Ota, T., NIshiyama, H., Taoka, Y., 1987. Flow around an elliptic cylinder in the critical Reynolds number regime. ASME Journal of Fluid Engineering 109, 149–155. Paliwal, B., Sharma, A., Chhabra, R.P., Eswaran, V., 2003. Power law fluid flow past a square cylinder: momentum and heat transfer characteristics. Chemical Engineering Science 58, 5315–5329. Patel, V.A., 1981. Flow around the impulsively started elliptic cylinder at various angles of attack. Computers and Fluids 9, 435–462. Schowalter, W.A., 1978. Mechanics of Non-Newtonian Fluids. Pergamon, Oxford, UK. Seban, R., Drake, R., 1953. Local heat transfer coefficients on the surface of an elliptic cylinder in a high speed air stream. Transaction of American Society of Mechanical Engineers 75, 235–240. Sharma, A., Eswaran, V., 2004. Heat and fluid flow across a square cylinder in the two-dimensional laminar flow regime. Numerical Heat Transfer, Part A: Applications 45, 247–269. Sharma, A., Eswaran, V., 2005. Effect of channel confinement on the twodimensional laminar flow and heat transfer across a square cylinder. Numerical Heat Transfer Part A Applications 47, 79–107. Sivakumar, P., Bharti, R.P., Chhabra, R.P., 2006. Effect of power-law index on critical parameters for power-law fluid flow across an unconfined circular cylinder. Chemical Engineering Science 61, 6035–6046. Soares, A.A., Ferreira, J.M., Chhabra, R.P., 2005. Flow and forced convection heat transfer in cross flow of non-Newtonian fluids over a circular cylinder. Industrial and Engineering Chemistry Research 44, 5815–5827. Sugihara-Seki, M., 1993. The motion of an elliptical cylinder in channel flow at low Reynolds numbers. Journal of Fluid Mechanics 257, 575–596. Talwar, K.K., Khomami, B., 1995. Flow of viscoelastic fluids past periodic square arrays of cylinders: inertial and shear-thinning viscosity and elasticity effects. Journal of Non-Newtonian Fluid Mechanics 57, 177– 202. Tanner, R.I., 1993. Stokes paradox for power-law fluids around a cylinder. Journal of Non-Newtonian Fluid Mechanics 50, 217–224. Whitney, M.J., Rodin, G.J., 2001. Force velocity relationships for rigid bodies translating through unbounded shear-thinning power-law fluids. International Journal of Non-Linear Mechanics 36, 947–953. Williamson, C.H.K., 1996. Vortex dynamics in the cylinder wake. Annual Reviews of Fluid Mechanics 28, 477–539. Woods, J.K., Spelt, P.D.M., Lee, P.D., Selerland, T., Lawrence, C.J., 2003. Creeping flows of power-law fluids through periodic arrays of elliptical cylinders. Journal of Non-Newtonian Fluid Mechanics 111, 211–228. Zdravkovich, M.M., 1997. Flow Around Circular Cylinders. Fundamentals, vol. 1. Oxford University Press, New York, USA. Zdravkovich, M.M., 2003. Flow Around Circular Cylinders. Applications, vol. 2. Oxford University Press, New York, USA.
© Copyright 2026 Paperzz