Sie sind auf Seite 1von 15

International Journal of Thermal Sciences 50 (2011) 2027e2041

Contents lists available at ScienceDirect

International Journal of Thermal Sciences


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

Flow of power-law fluids past an equilateral triangular cylinder: Momentum


and heat transfer characteristics
A. Prhashanna, Akhilesh K. Sahu, 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

Article history: Extensive numerical simulations for the 2-D laminar flow of power-law fluids over an equilateral
Received 2 July 2010 triangular cylinder are performed to elucidate the role of power-law index (0.2  n  1.8) on the critical
Received in revised form Reynolds number denoting the onset of flow separation and of vortex shedding. Results are presented for
14 February 2011
both orientations of the triangular cylinder, namely, its apex facing in the upstream and the downstream
Accepted 22 April 2011
directions. For shear-thinning fluids, the onset of wake formation and vortex shedding both are seen to
Available online 28 May 2011
be delayed than that in Newtonian and shear-thickening fluids. After delineating the limits of the steady
flow regime, the effects of the Reynolds and Prandtl numbers and the power-law index on drag
Keywords:
Triangular cylinder
phenomena and heat transfer characteristics of triangular cylinder for shear-thinning fluids have been
Vortex shedding studied in the steady flow regime. The results reported herein embrace the following ranges of condi-
Critical Reynolds number tions: 0.2  n  1; 1  Re  30 and 1  Pr  100. Detailed results on isotherm contours, local Nusselt
Power-law number and its surface-averaged values are presented. Finally, the present numerical values of the
Wake formation critical Reynolds numbers and Nusselt number have been correlated using simple forms which are
Strouhal number convenient for interpolating the present results for the intermediate values of the governing parameters
Nusselt number in a new application.
Ó 2011 Elsevier Masson SAS. All rights reserved.

1. Introduction information clearly shows that the bulk of the literature pertains to
the case of a circular cylinder, followed by that relating to cylinders
The flow of fluids over cylinders of different cross-sections of elliptic and square cross-sections [1,2]. However, only very
constitutes an important class of problems within the field of limited work has been reported on the flow over semi-circular
fluid mechanics from both theoretical as well as pragmatic cylinders [3e6], triangular [7,8], polygonal cylinders [9] and
considerations. Such model flows are often used to gain useful square cylinders (e.g., see [10,11]). In addition to the viscosity and
insights into the nature of the underlying physical processes such density of the fluid and the diameter of the cylinder, the flow past
as the detailed structure of the flow field, wake phenomenon, a circular cylinder is influenced by a large number of parameters
vortex shedding, etc. as well as the prediction of hydrodynamic including its orientation with respect to the oncoming fluid, nature
forces acting on submerged objects (drag and lift), the rates of heat of the faraway flow field (oscillating, shear, uniform, angle of inci-
and mass transfer. In addition to such an overwhelming intrinsic dence, for instance), type of fluid (compressible or incompressible,
importance, the flow past cylinders of various cross-sections is also Newtonian or non- Newtonian), confined or unconfined flow,
encountered in numerous industrial settings. Typical examples aspect ratio of the cylinder, etc. Even for the simplest case of the
include tubular and pin type heat exchangers, use of thin wires as parallel uniform cross-flow of an incompressible fluid past an
measuring probes, support structures, cooling of electronic unconfined circular cylinder, the flow exhibits a range of flow
components, use of submerged obstacles to form weld lines in regimes. For instance, at very low Reynolds numbers, the flow
polymer processing applications, etc. Consequently, over the years, remains attached to the surface of the cylinder thereby showing the
a vast body of knowledge has accrued on the flow of Newtonian complete fore and aft symmetry. Under these conditions, the flow is
fluids past cylinders of different cross-sectional shapes over wide two-dimensional and steady provided the length to diameter ratio
ranges of conditions. A cursory inspection of the available of the cylinder is large to avoid the end effects. This flow regime
occurs up to about Re w 4e5. As the Reynolds number is
progressively increased, the flow begins to separate in the rear part
* Corresponding author. Tel.: þ91 512 2597393; fax: þ91 512 2590104. of the cylinder leading to the formation of two symmetric vortices.
E-mail address: chhabra@iitk.ac.in (R.P. Chhabra). Both the length and width of the separated flow region increase

1290-0729/$ e see front matter Ó 2011 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ijthermalsci.2011.04.018
2028 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

Nomenclature n Power-law flow behaviour index


ns Unit vector normal to the surface of the cylinder
b Side of the triangular cylinder, m Np Number of grid points on the surface of the cylinder
cp Specific heat of the fluid, J/kg K Nu Nusselt number, Nu ¼ hb=k ¼ ðvT=vns Þ
CD Drag coefficient, CD ¼ 2FD =rUN 2 b p Pressure, Pa
CD Time averaged drag coefficient Pr Prandtl number, Pr ¼ cp mðUN =bÞn1 =k
CDp Pressure component of drag coefficient, Re Reynolds number, Re ¼ rUN 2n bn =m

CDp ¼ 2FDp =rUN 2 b St Strouhal number, St ¼ fb=UN


CL Lift coefficient, CL ¼ 2FL =rUN2 b T Temperature of fluid, K
Clrms Root mean square value of lift coefficient TN Temperature of fluid at inlet, K
dt Time step, s Tw Temperature of surface of body, K
f Frequency of vortex shedding, s1 UN Free stream velocity, m s1
FD Drag force per unit length, N m1 Ux ; Uy x- and y- components of the velocity, m s1
FDp Pressure force per unit length, N m1 x, y Stream-wise and transverse coordinates, m
FL Lift force per unit length, N m1
g Acceleration due to gravity, m/s2 Greek symbols
h Local convective heat transfer coefficient, W/m2 K h Viscosity, Pa s
H Lateral height of the computational domain, m r Density of the fluid, kg m3
I2 Second invariant of the rate of the strain tensor, s2 3 Component of the rate of the strain tensor, s1
k Thermal conductivity of fluid, W/m K s Extra stress tensor, Pa
Lu Upstream length of computational domain, m
Ld Downstream length of computational domain, m Superscript/Subscript
Lr Recirculation length, m c Critical
m Power-law consistency index, Pa sn s surface of the cylinder

with Reynolds number. The flow is still 2-D and steady, but the fore incompressible fluids over an equilateral triangular object with its
and aft symmetry is lost. This flow regime persists up to about apex facing the flow. While the major thrust of their work was on
Re w 45e46. Further increase in the value of the Reynolds number the prediction of fluid mechanical [7] and heat transfer [12] aspects,
leads to the instability of wake and the vortices are shed alternately based on the global mode analysis, they reported the critical value
from the upper and lower-half of the cylinder thereby transiting of Re ¼ 39.9 to indicate the cessation of the steady flow regime for
to the so-called laminar vortex shedding regime. Under these the unconfined flow past a triangular cylinder which compares
conditions, the flow is periodic in time, but it continues to be two- favorably with the value of 36.37 reported by Jackson [13]. Simi-
dimensional. This flow regime persists up to about Re w 190e200 larly, Zielinska and Wesfreid [14] investigated numerically the
and beyond this value, the flow transits to three- dimensional and transition from the steady to time-periodic flow from a triangular
turbulence in wake, etc. [1,2]. Naturally, the underlying differences cylinder (with apex aligned with the flow) confined in a planar
in the flow field adjacent to the cylinder in different flow regimes channel (blockage ratio of 1/15). They reported a value of Re ¼ 38.3
also manifest at the macroscopic level in the way the drag coeffi- which was corroborated subsequently by experimental observa-
cient and Nusselt number scale with the Reynolds number. It is tions [15]. This value is also in broad agreement with that reported
perhaps fair to say that the second transition, i.e., from steady to by De and Dalal [7] and Jackson [13]. In another study, Abbassi et al.
periodic in time, has received much more attention than the other [16] studied flow and heat transfer from a triangular prism (base to
transitions. Similar transitions are also observed with cylinders of height ratio 0.5) confined in a planar channel (blockage ratio 1/4).
other shapes, albeit the values of the Reynolds number denoting They reported the wake to become asymmetrical at Re ¼ 45 which
the transition from one flow regime to another are strongly influ- is also consistent with the general idea that the confinement has
enced by the shape of the cylinder. Thus, for instance, for circular a tendency to stabilize the flow, albeit more recent work suggests
and elliptic cylinders, the point of flow separation is not fixed and this value to be in the vicinity of Re ¼ 58e59 [17]. Faruquee and
generally keeps moving forward with the increasing Reynolds Olatunji [8] have also studied the unconfined flow past a two-
number. Intuitively, it is reasonable to expect that the flow sepa- dimensional triangular cylinder with its apex facing/opposing the
ration is delayed for streamlined bluff bodies and/or for confined flow in the Reynolds number range 30  Re  150. They reported
flows, and this is indeed borne out by the few studies available on a value of the critical Reynolds number in the range 40  Re  42
this aspect [1,2]. Unlike in the case of circular and elliptic cylinders, for both orientations which is somewhat larger than the previous
the flow separation occurs at fixed points for cylinders with corners values. It is, however, appropriate to add here that it is not at all
such as square and triangular cylinders. Thus, for instance, the flow uncommon to encounter deviations of this order due to the
separation for the flow of a Newtonian fluid past a square bar differences in domain, grid, convergence criterion, solution meth-
occurs at about Re w 2 whereas the second transition delineating odology, etc. [18]. Furthermore, not all investigators have reported
the steady/unsteady flow regimes occurs at about Re w 53. Simi- the criterion they used to delineate the value of the critical Rey-
larly, for an equilateral triangular cylinder oriented normal to the nolds number including the loss of mid-plane symmetry, value of
direction of flow, the transition further depends upon its orienta- lift coefficient, distinct onset of the vortex shedding, etc. While all
tion, e.g., whether the vertex is facing upstream or downstream these phenomena are inter-related, but the values of the critical
direction. There has been very little activity on the prediction of the Reynolds number are likely to somewhat differ from one criterion
critical Reynolds number denoting the onset of flow separation and to another. Thus, for instance, intuitively it appears that the wake
on the commencement of vortex shedding regime from an equi- will become asymmetric first and the vortex shedding will occur at
lateral triangle, even in Newtonian fluids [7,8]. For instance, De and a slightly larger value of the Reynolds number. Srikanth et al. [17]
Dalal [7] considered the steady and time-dependent flow of have studied the effect of symmetric planar confinement on drag
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2029

and heat transfer parameters of a triangular cylinder (apex oriented computational requirements. The influence of the boundary walls
upstream) in the Reynolds number range 1  Re  80, but for was further reduced by treating them as slip boundaries so that
a single value of the Prandtl number value of 0.71 and for a fixed these are free from dissipative effects.
blockage of (1/4). They reported the effect of confinement to be The continuity and momentum equations for this flow are
more dramatic in the time-dependent flow regime than that in the written as:
steady flow regime. Similarly, the effect of asymmetric confinement Continuity Equation:
has been studied recently over the range of Reynolds number
V$U ¼ 0 (1)
(100  Re  450) which clearly pertains to the time-dependent
flow conditions [19]. It is thus fair to say that only preliminary Momentum Equation:
results are available on the momentum and heat transfer charac-  
teristics (only for air) for a triangular cylinder even in Newtonian DU
r g ¼ Vp þ V$s (2)
fluids. Dt
On the other hand, many industrial substances of macromo-
Energy Equation:
lecular and multi-phase nature (like polymer melts and their
solutions, foams, emulsions and suspensions) display a range of DT
non-Newtonian flow characteristics including shear-dependent rcp  kV2 T ¼ 0 (3)
Dt
viscosity, visco-elasticity, yield stress, etc. [20]. Such fluids are
encountered routinely in polymer, food, pharmaceutical and In the energy equation, the viscous dissipation effects have been
mineral process engineering applications [20]. Amongst different neglected. Furthermore the thermo-physical properties (thermal
types of non-Newtonian characteristics, the simplest is the shear- conductivity, heat capacity, density and viscosity) of the fluid are
dependent viscosity, i.e., shear-thinning and shear-thickening, assumed to be temperature-independent. While these two
which is often approximated by the simple power-law model. In simplifications lead to the decoupling of the momentum and
spite of their wide occurrence in a variety of settings, very limited energy equations, the latter assumption also limits the applicability
results are available on the flow of power-law fluids past a circular of the present results to the situations wherein the maximum
cylinder [21e25], or an elliptic cylinder [26,27] or a square cylinder temperature difference, DT ¼ TW  T0 is small.
[28e30]. Furthermore, most of these studies are limited to the so- The extra stress tensor s, is related to the rate of deformation
called steady flow regime, except for the recent studies which have tensor 3ðuÞ via the scalar viscosity of the fluid as
dealt with the laminar vortex shedding regimes for a circular
[31,32] and a square cylinder [33,34]. While the limits of the steady
flow regime for the cross-flow of power-law fluids past a circular
A
cylinder [35], an elliptic cylinder [36] and a square cylinder [37]
have been delineated only recently, no such results are available
for triangular cylinders. Similarly, as far as known to us, there has
been no prior study on the flow of power-law fluids past a trian-
gular cylinder. This work aims to fill this gap in the literature. In
particular, reported herein are the critical values of the Reynolds
number denoting the onset of flow separation and of laminar
vortex shedding from a two-dimensional equilateral triangular
cylinder, oriented normal to the direction of flow, with its apex
facing in the upstream or downstream direction. Furthermore after
delineating the limits of the steady flow regime, momentum and
heat transfer characteristics are investigated over the following
ranges of conditions: power-law index: 0.2 to 1; Reynolds number:
1 to 30 and Prandtl number: 1e100.

2. Problem statement and governing equations

Consider the two-dimensional transverse flow of an incom- B


pressible power-law fluid with a uniform velocity UN and at
a temperature TN over a long (in the z-direction) equilateral
triangular cylinder (maintained at a constant temperature Tw and
of side b) with its apex facing upstream (Fig. 1(A)) or downstream
(Fig. 1(B)). Naturally, it is not possible to simulate truly unconfined
flow condition in a numerical study, and it is mimicked here by
enclosing the triangular cylinder in a rectangular domain, as shown
schematically in Fig. 1(A) and (B). The inlet is located at a distance
Lu from the tip of the cylinder and the outlet at a downstream
distance of Ld from the rear vertical side of the object as shown in
Fig. 1. The lateral walls were separated by a distance H. While one
would like to choose the values of Lu , Ld and H as large as possible to
study the truly unconfined flow, but this would require exorbitant
computational resources. On the other hand, their small values Lu ,
Ld and H will unduly influence the flow phenomena in the vicinity
of the submerged cylinder. Hence, their optimal values denote
a tradeoff between the accuracy of the results vis-à-vis the Fig. 1. Schematics of the flow around an equilateral triangular cylinder.
2030 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

s ¼ 2h3ðuÞ (4) respectively. In this work, the first transition (i.e., Rec ) was located
by examining the values of the dimensionless stream function,
where 3ðuÞ, the rate of deformation tensor is related to the velocity vorticity and velocity vector near the surface of the bluff body to
field as check for the onset of separation. The typical values of the
1h i dimensionless stream function and vorticity near the cylinder are
3ðuÞ ¼ VU þ ðVUÞT (5) of the order of 105e108. The other critical point corresponding to
2
Rec denoting the cessation of the steady flow regime was located by
and the power-law viscosity is given by examining the lift coefficient-time history. For steady flow, the lift
n1 coefficient will gradually approach zero or very small values (the
h ¼ mð2I2 Þ 2 (6) cut off value of the order of 104 was used here). This inference was
further substantiated by examining the asymmetry of the wake
where I2 is the second invariant of the rate of deformation tensor, about the center line, y ¼ 0. Thus, in the range Re < Rec , the flow is
and n is the flow behavior index. Evidently, n < 1 denotes the shear- steady with two standing symmetric vortices. Finally, the local
thinning behavior; n > 1 corresponds to the shear-thickening value of the Nusselt number on the surface of the isothermal
behavior and n ¼ 1 represents the standard Newtonian fluid cylinder is simply given by the temperature gradient on the surface
behavior. of the cylinder and this, in turn, can be integrated over the surface
The physically realistic boundary conditions for this flow may be of the cylinder to obtain the mean Nusselt number.
written as follows:
3. Numerical solution methodology
 At the inlet boundary, the uniform flow in the x- direction is
prescribed, i.e., The present study has been carried out using FLUENT (version
6.3.26). The unstructured ‘quadrilateral’ and ‘triangular’ cells of
Ux ¼ UN ; Uy ¼ 0; T ¼ TN (7) non-uniform grid spacing were generated using the commercial
grid tool GAMBIT (version 2.3.16). Triangular cells are used only for
 On the surface of the triangular cylinder: The usual no-slip the case of delineating the onset of wake formation near the surface
condition for velocity and that of constant temperature are of the cylinder. For Re < Rec, the flow is expected to be symmetric
implemented, i.e., about the mid-plane and the solution was sought only in half
domain. On the other hand, full domain computations were carried
Ux ¼ 0; Uy ¼ 0; T ¼ Tw (8) out to delineate the value of Rec . The two-dimensional, laminar,
segregated solver was used to solve the incompressible flow on the
 At the exit-boundary: The default outflow boundary condition collocated grid arrangement. The unsteady solver was used for the
in FLUENT is used which assumes a zero diffusion flux for all vortex shedding case ðRe > Rec Þ while the steady solver was used
flow variables, and this is similar to the fully developed flow for the wake formation ðRezRec Þ case. The second order upwind
condition. It needs to be emphasized here that it ensures the scheme was used to discretize the convective terms in the
gradients in the axial direction to be zero, but the gradients momentum and energy equations. The semi-implicit method for
may still exist in the lateral direction. the pressure linked equations (SIMPLE) scheme was used for
 At the center line for steady flows, the symmetric flow condi- solving the pressure-velocity coupling. The second order time-
tion is used, i.e., discretization together with an implicit scheme has been used for
time integration in this work. The ‘constant density’ and ‘non-
vUx vT Newtonian power-law’ viscosity models were used to input the
Uy ¼ 0; ¼ 0; ¼ 0 (9) physical properties of the fluid. This is, however, of no particular
vy vy
significance as the final results are presented in a dimensionless
 The lateral walls are assumed to be slip boundaries and adia- form. FLUENT solves the system of algebraic equations using the
batic, i.e., Gauss-Seidel (G-S) point-by-point iterative method in conjunction
vUx vT with the algebraic multi-grid (AMG) method. This approach greatly
Uy ¼ 0; ¼ 0; ¼ 0 (10) reduces the number of iterations (thereby accelerating conver-
vy vy
gence) and thus leading to significant savings in CPU time, partic-
The governing equations (1)e(6) together with the boundary ularly when the model equations contain a large number of control
conditions outlined in equations (7)e(10), have been solved volumes. The relative convergence criteria of 108 for the conti-
numerically to map the flow domain in terms of the velocity, nuity and x- and y-components of the momentum and energy
pressure and temperature fields over wide ranges of physical and equations were prescribed in this work. Also, the solution was
kinematic parameters. This information, in turn, is used to deduce deemed to have converged when there was no change (at least up
the global characteristics like force coefficients (drag and lift), to the fourth decimal place) in the value of the total drag coefficient
Strouhal number and Nusselt number. At low Reynolds numbers, and the corresponding changes in the value of the lift coefficient
the flow is expected to be steady and symmetric (about the mid- (for vortex shedding case) are of the order of 105e106 for more
plane), and the scaling considerations suggest the flow to be gov- than 1000 time steps or when it shows more than 10 constant
erned by three dimensionless parameters, namely, power-law periodic cycles in time history of the lift and drag coefficients.
index, Reynolds number and drag coefficient, with both lift coeffi-
cient and Strouhal number being zero. In the time-dependent 3.1. Choice of numerical parameters
periodic flow regime, Strouhal number and lift coefficient achieve
non-zero values which are expected to be functions of the power- The accuracy and reliability of the numerical results is naturally
law index and Reynolds number. The Nusselt number shows contingent upon a prudent choice of optimal domain, grid and time
additional dependence on the Prandtl number. step. In this work, the domain is characterized by the values of Lu , Ld
The two critical Reynolds numbers denoting the onset of flow and H. These parameters were systematically varied in order to
separation and onset of vortex shedding are denoted as Rec and Rec arrive at their optimum values. Similarly, a satisfactory grid should
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2031

Table 1 Table 3
Domain independence study for vertex facing upstream at Re ¼ 0.01. Domain independence study for vertex facing upstream at Re ¼ 10.

n¼1 n ¼ 0.2 n ¼ 1.4 n¼1 n ¼ 0.2 n ¼ 1.8


Lu Lu
120 CD 503.68 2540.5 140.34 10 CD 2.8854 3.0653 2.8110
CDp 318.37 1991.9 82.984 CDp 1.4910 2.3829 1.1756
150 CD 499.95 2538.8 138.54 12 CD 2.8706 3.0683 2.7886
CDp 316.11 1990.6 81.957 CDp 1.4835 2.3855 1.1665
180 CD 498.80 2541.6 137.89 15 CD 2.8608 3.0701 2.7735
CDp 315.41 1992.7 81.590 CDp 1.4784 2.3871 1.1603

H Ld
500 CD 464.70 2546.1 128.13 20 CD 2.8714 3.0554 2.7889
CDp 293.71 1996.2 75.636 CDp 1.4840 2.3739 1.1666
600 CD 460.35 2544.3 127.28 25 CD 2.8714 3.0877 2.7888
CDp 290.89 1994.8 75.119 CDp 1.4840 2.4026 1.1666
700 CD 458.30 2543.9 126.82 30 CD 2.8714 3.0886 2.7888
CDp 289.56 1994.5 74.838 CDp 1.4840 2.4033 1.1666

Ld H
120 CD 509.31 2537.9 139.50 30 CD 2.8130 3.0643 2.7003
CDp 322.33 1989.9 82.569 CDp 1.4478 2.3835 1.1222
150 CD 510.40 2543.2 139.54 40 CD 2.7968 3.0370 2.6756
CDp 323.06 1994.0 82.597 CDp 1.4384 2.3595 1.1106
180 CD 510.72 2544.7 139.55 50 CD 2.7912 3.0347 2.6670
CDp 323.27 1995.1 82.601 CDp 1.4351 2.3577 1.1066

be sufficiently fine in the vicinity of the submerged obstacle (where from domain effects.. The corresponding results for the opposite
the gradients are expected to be steep) without being prohibitively orientation are shown in Table 2 for Re ¼ 1. Evidently, the values of
expensive in terms of computational resources. Both these aspects Lu ¼ 50; Ld ¼ 40 and H ¼ 140 are seen to be adequate in this case.
have been investigated here. Since the flow field decays very slowly at low Reynolds numbers,
shorter domain should suffice at high Reynolds numbers. The
results shown in Tables 3 and 4 lend support to this assertion.
3.2. Effect of domain size
Therefore, for the case of vertex oriented in upstream direction,
Lu ¼ 12, Ld ¼ 25 and H ¼ 40 were used whereas for the vertex
In this work, three values of Lu and Ld (non-dimensionalized
facing downstream, Lu ¼ 20, Ld ¼ 30 and H ¼ 70 have been found
using b) as 120,150 and 180 were used at Re ¼ 0.01 for three
to be satisfactory. Suffice it to say here that broadly, the domain size
representative values of the power-law index, namely, n ¼ 0.2, 1
used here is much larger than that used in previous studies [7,8,13].
and 1.4. It is clearly seen that the results change very little when the
value of Lu is increased from 150 to180, albeit the CPU time
increased significantly (Table 1, vertex oriented upstream). Simi- 3.3. Grid independence study
larly, the results change by even smaller amounts when Ld is
increased from 150 to 180. It therefore appears reasonable to fix Several grids were generated to select an optimal grid which
both Lu and Ld as 150. The results included in Table 1 also suggest would be sufficiently accurate without necessitating excessive
that H ¼ 600 denotes a satisfactory choice for the values to be free computational resources in terms of CPU time and memory. While

Table 2 Table 4
Domain independence study for vertex facing downstream at Re ¼ 1. Domain independence study for vertex facing downstream at Re ¼ 10.

n¼1 n ¼ 0.2 n ¼ 1.8 n¼1 n ¼ 0.2 n ¼ 1.8

Ld Lu
30 CD 10.922 25.536 6.2340 15 CD 2.9537 3.3381 2.8898
CDp 7.6312 19.999 4.6387 CDp 2.5904 3.1139 2.4757
40 CD 10.928 25.718 6.2974 20 CD 2.9475 3.3388 2.8803
CDp 7.6355 20.156 4.6410 CDp 2.5844 3.1141 2.4767
50 CD 10.930 25.724 6.2978 25 CD 2.9463 3.3145 2.8783
CDp 7.6364 20.161 4.6413 CDp 2.5832 3.0843 2.4657

Lu H
40 CD 10.563 25.719 6.0310 60 CD 2.7924 3.3056 2.6414
CDp 7.3666 20.158 4.4435 CDp 2.4437 3.0701 2.2678
50 CD 10.521 25.715 5.9997 70 CD 2.7866 3.3078 2.6318
CDp 7.3352 20.154 4.4202 CDp 2.4385 3.0730 2.2598
60 CD 10.496 25.701 5.9827 80 CD 2.7833 3.3113 2.6262
CDp 7.3001 20.134 4.3921 CDp 2.4355 3.0775 2.2552

H Ld
120 CD 10.383 25.557 5.9188 20 CD 2.8079 3.3525 2.6721
CDp 7.2214 20.009 4.3467 CDp 1.4573 3.1313 2.2930
140 CD 10.355 25.485 5.9026 30 CD 2.8104 3.3582 2.6764
CDp 7.2019 19.947 4.3351 CDp 2.4596 3.1390 2.2965
160 CD 10.338 25.469 5.8924 40 CD 2.8108 3.3624 2.6769
CDp 7.1896 19.933 4.3278 CDp 2.4599 3.1446 2.2969
2032 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

Table 5
Effect of grid size on flow parameters with vertex facing upstream.

Grid No. of cells Np CDp CD

n ¼ 0.2 n ¼ 1.0 n ¼ 1.4 n ¼ 0.2 n ¼ 1.0 n ¼ 1.4


Re ¼ 0.01
G1 17053 150 1944.3 287.70 74.418 2484.6 459.59 127.99
G2 67909 300 2014.6 295.28 76.512 2543.0 459.77 127.37
G3 209758 450 2025.1 297.70 77.010 2547.9 460.33 127.19

Re ¼ 10
G1 31170 114 2.4112 1.3686 1.0318 3.1076 2.7763 2.6568
G2 51000 150 2.3999 1.4186 1.2210 3.0827 2.7934 2.6722
G3 122798 225 2.3371 1.4735 1.1522 3.0815 2.8089 2.6872

a range of composite grids consisting of triangular and/or quadri- The adequacy of the aforementioned parameters employed here
lateral cells were used in this work. Suffice it to say here that the is further demonstrated by presenting some benchmark compari-
finest mesh ðd ¼ 0:003bÞ was used over a distance of 3b in both x- sons with the prior literature results in the next section.
and y-directions enclosing the triangular cylinder. The grid was
made progressively coarse by using a successive ratio of 1.04. For 4. Results and discussion
instance, for Re ¼ 1 (Table 6), we have used an unstructured
triangular grid near the surface of the cylinder which decreases In this work, the time marching numerical scheme was used
rapidly away from the cylinder. On the other hand, at Re ¼ 10, we with an initial velocity profile corresponding to a higher Reynolds
have used structured quadrilateral grid. Therefore, the total number and/or using a steady flow field to initiate the calculations.
number of cells is only one parameter. The number of cells on the While in experimental work, the flow can become unstable by
surface of the cylinder ranged from 50 to 450 while the total several factors including non-uniform flow conditions, vibrations
number of cells varied from w14,000 to w2, 25,000. Tables 5 and 6 due to mechanical devices (pump or compressor), irregularities of
summarize the results elucidating the effect of grid for both the physical boundaries, surface roughness, etc., these factors are
orientations of the triangular object. An examination of these two irrelevant in numerical simulations. On the other hand, even when
tables suggests that G2 is adequate in each case. It needs to be the initial and boundary conditions are symmetric, the truncation,
emphasized here that the label G2 is only symbolic as it varied in rounding off and discretization errors act as destabilizing factors
detail from one case to another. While the pressure drag coefficient which ultimately destabilize the flow thereby leading to the vortex
changes by up to w4e5%, the total drag coefficient rarely changes shedding regime under appropriate conditions of the Reynolds
by more than 0.5% in going from Grid G2 to G3. Therefore, the grid number. Naturally this approach requires significantly large time
G2 (different in each case) was finally used in this work. Same grids for these errors to grow to the extent required for the flow to
were tested to study the heat transfer from the cylinder to shear- become unstable ultimately leading to the onset of time-dependent
thinning fluids in the steady flow regime and it was found that periodic flow regime. This process can be accelerated by perturbing
the grid G2 is sufficiently fine for heat transfer study also (Table 7). the flow artificially, as that used by Braza et al. [38] and others [39]
relating to the case of a circular cylinder. However, in the present
study, no external stimulus was used for this purpose and the time-
3.4. Effect of time step dependent calculations were continued for a long time until either
the flow became steady or fully periodic in time. Prior to presenting
Since the delineation of Rec necessitates the solution of the the new results, it is worthwhile to validate the numerical meth-
time-dependent equations, the choice of an optimal time step is odology employed here.
also germane to obtain reliable results. Essentially, two different
values of time step, dt ¼ 0.01 and dt ¼ 0.001 (non-dimensionalised 4.1. Validation of results
using b/UN) were used. Table 8 presents a summary of the results
obtained with these two values and the choice of the value of dt The present results are compared with the scant literature
seems to have virtually no influence on the resulting values of the values for the flow of Newtonian fluids past an equilateral triangle
lift and drag coefficients and of the Strouhal number. Consequently, with vertex facing upstream (Table 9) in terms of drag coefficient
the values of dt in this range only have been used in this study and and recirculation length Lr for a range of values of the Reynolds
the exact value of dt is of no particular significance. number. Except at Re ¼ 10, the two values of drag are within w2.5%

Table 6 Table 7
Effect of grid size on flow parameters with vertex facing downstream. Grid independence study for heat transfer.

Grid No. of cells Np CDp CD Grid No. No. of cells Np Nu

n ¼ 0.2 n ¼ 1.0 n ¼ 1.8 n ¼ 0.2 n ¼ 1.0 n ¼ 1.8 n ¼ 0.3 n¼1


Re ¼ 1 Vertex upstream Re ¼ 30, Pr ¼ 100
G1 16695 75 20.0198 7.1254 4.3196 25.4999 10.2161 5.8608 G1 20540 50 25.832 17.895
G2 46047 150 20.4666 7.3957 4.5721 25.6185 10.3091 5.8965 G2 26785 75 25.492 17.006
G3 80038 300 20.6392 7.5087 4.4884 25.6166 10.3087 5.9006 G3 38250 150 25.383 16.936

Re ¼ 10 Vertex downstream Re ¼ 30, Pr ¼ 100


G1 14032 51 3.0681 2.2036 2.0380 3.4099 2.7672 2.6292 G1 97150 50 17.392 12.805
G2 56448 105 3.1264 2.4021 2.2374 3.3699 2.8045 2.6708 G2 117600 80 16.989 12.499
G3 225792 210 3.0600 2.5579 2.4054 3.3609 2.8232 2.6887 G3 124200 100 16.964 12.498
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2033

Table 8 4.2. Onset of flow separation


Effect of time step.

n Re dt ¼ 0.01 dt ¼ 0.001 When flow approaches the triangle, the flow comes to rest at the
CL CD;avg St CL CD;avg St
front stagnation point and then goes around the obstacle. At low
Reynolds numbers, the fluid inertia is negligible and therefore
Vertex facing upstream
0.3 37 0.0348 1.383 0.1729 0.0348 1.383 0.1729 a fluid element is able to closely follow the surface of the
1 41 0.022 1.516 0.1284 0.022 1.516 0.1284 submerged body. However, as the Reynolds number is progres-
1.8 33 0.00775 1.814 0.1053 0.00775 1.814 0.1053 sively increased, fluid inertia increases and the stabilizing viscous
Vertex facing downstream forces diminish. This, in turn, leads to the establishment of adverse
0.3 33 0.2232 1.797 0.1277 0.2232 1.797 0.1277 pressure gradient on the surface of the bluff body at a critical value
1 37 0.0507 1.73 0.1067 0.0507 1.73 0.1067 of Reynolds number. This leads to the separation of flow from the
1.8 30 0.02 1.978 0.093 0.02 1.978 0.093
surface of the bluff body. For a power-law fluid, one can postulate
that the viscous force scales as wUN n whereas the inertial force
2
scales as wUN . Thus for a shear-thinning fluid, (n < 1), the viscous
of each other whereas that of Lr , St, CD and Clrms are even closer than
this. While the exact reasons for relatively large differences at force shows weaker dependence on the velocity than in Newtonian
Re ¼ 10 are not immediately obvious, this is likely due to the fluids. Also, conversely, the fluid viscosity is minimum near the
relatively short domain used by De and Dalal [7] than that used surface of the cylinder (as the rate of shearing is maximum here)
here. Similarly, the present value of (for vertex facing upstream) which progressively increases away from the cylinder. In other
Rec ¼ 40  0:5 is remarkably close to the values of 39.9 reported words, there is a region of low viscosity fluid close to the
by De and Dalal [7] and of 42 reported by Faruquee and Olatunji [8], submerged cylinder which is encapsulated by a highly viscous mass
albeit the precision of their values is not known. This value also of the fluid. This is likely to postpone the onset of flow separation to
compares favorably with the value of 38.3 for a triangular cylinder higher Reynolds number than that in Newtonian fluids. By a similar
as reported by Zielinska and Wesfreid [14,15]. At Re ¼ 50, the reasoning, one would expect to see the reverse trend in shear-
present values of CD ¼ 1:53 and Strouhal number, St ¼ 0:147 are thickening fluids. Indeed, the results shown here lend support to
also very close to the corresponding values of CD ¼ 1:53 and St ¼ this line of argument. In this work, the onset of wake formation is
0:1498 due to De and Dalal [7]. While the differences of this order judged by examining the local values of vorticity, direction of
may seem large, but it needs to be emphasized here that such velocity vectors and streamline patterns near the triangular
differences are not at all uncommon in such numerical studies [18]. cylinder. Fig. 2(A) and (B) show the streamline profiles near the
Such deviations stem from the inherent differences in numerics submerged cylinder for extreme values of power-law index (n ¼ 0.5
(domain, grid size, time step and convergence criterion, for and 1.4 for the case of apex facing the upstream direction and for
instance) and the solution methodology. For instance, De and Dalal n ¼ 0.2 and 1.8 for the case of apex facing downstream) including
[7] used a domain with Lu ¼ 9, Ld ¼ 20 and H ¼ 20 which is much n ¼ 1, for at least two close by values of the Reynolds number which
shorter than the one used here. Similarly, their convergence crite- bracket the transition point. In this figure, for all values of power-
rion of 105 is much less stringent than the value of 108 used in law index, the flow around the triangular cylinder is not sepa-
this study. On the other hand, Faruquee and Olatunji [8] used rated at the lower Reynolds number (left column) while at the
a cylindrical domain. Finally, the present solution methodology was higher Reynolds number (right column), separation is seen to have
thoroughly calibrated against the benchmark problem of the occurred either on the face or on the edge of the cylinder. For
laminar flow of Newtonian and power-law fluids for the lid-driven instance, in Fig. 2(A) for n ¼ 0.5, there is no flow separation around
cavity flow configuration [40,41]. The present values of the center the cylinder at Re ¼ 0.55, while at Re ¼ 0.6 separation has clearly
line velocities were within 2% of that reported by Ghia et al. [40] occurred near the edge of the cylinder. So for power-law index,
for Newtonian and by Neofytou [41] for power-law fluids. Such n ¼ 0.5, the critical Reynolds number ðRec Þ lies in the range
a close correspondence inspires further confidence and lends 0.55  Re  0.6. Similarly, the critical values of the Reynolds
credibility to the reliability and accuracy of the present results. number lie between Re ¼ 0.3 and 0.35 for n ¼ 1.0 and between
Notwithstanding all these factors, the present results are believed Re ¼ 0.02 and 0.025 for n ¼ 1.4. This approach was used to ascertain
to be reliable to within 2e3%. the value of Rec for the other values of power-law index and/or for
the second orientation of the apex facing downstream direction
(Fig. 2(B)). For the case of apex facing the downstream direction, as
Table 9
the value of power-law index increases from 0.2 to 1.8, the sepa-
Comparison of the present and literature results.
ration point moves from near to the edge towards the apex
Re CDp CD Lr (Fig. 2(B)). On the other hand, for the apex facing the upstream
Present De and Present De and Present De and direction (Fig. 2(A)), separation occurs somewhere between the
Work Dalal [7] Work Dalal [7] Work Dalal [7] edge and the middle of the vertical face; the separation point
10 2.793 2.68 1.419 1.35 0.834 0.75 moves towards the edge as the power-law index is gradually
20 2.018 1.97 1.075 1.05 1.514 1.5 increased. It is also observed that the size of the wake increases
30 1.699 1.68 0.949 0.94 2.234 2.25
with the power-law index for both orientations of the cylinder at
35 1.597 1.58 0.911 0.91 2.594 2.625
40 1.536 1.52 0.880 0.88 e e a fixed value of the Reynolds number. Fig. 3 shows the influence of
the power-law index on the critical Reynolds number ðRec Þ
Onset of Vortex Shedding For Vertex Placed upstream
Rec
denoting the point of wake formation for both orientations. The
Present Work 41 critical Reynolds number values reported herein are believed to be
De and Dalal [7] 39.9 accurate to within a band of 0.5 and 0.05 for the apex facing the
Faruquee and Olatunji [8] 42 upstream and downstream directions respectively. For shear-
For Re ¼ 50 thinning fluids, the critical Reynolds number curve goes through
St CD Clrms a maximum at n ¼ 0.4 and 0.6 for these two configurations
Present Work 0.1473 1.527 0.0704 respectively, whereas for shear-thickening fluids, it continuously
De and Dalal [7] 0.1498 1.53 0.074
decreases for both orientations of the triangular cylinder. This is
2034 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

qualitatively similar to that seen for a circular [35], elliptic [36] and upstream configuration is less streamlined than the other way
square cylinder [37] and is ascribed to the inherently different around. Conversely, due to the sudden termination of the bluffbody
scaling of the viscous and inertial forces with velocity and power- surface in the case of vertex facing upstream, the adverse pressure
law index, as alluded to earlier. It is also seen that the value of gradient is set up at lower value of the Reynolds number than that
Rec for the apex facing the downstream direction is higher than in the opposite case. The relationship between the critical Reynolds
that for the apex facing the upstream direction by an order of number, Rec , and the power-law index, n, can be best represented
magnitude. This is probably due to the fact that the apex facing by the following expressions:

Fig. 2. Streamline patterns near the cylinder at low values of Reynolds number for n ¼ 0.2, 0.5, 1, 1.4 and 1.8, (A) vertex facing the upstream direction (B) vertex facing the
downstream direction.
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2035

Fig. 2. (continued).

 apex facing the upstream direction (0.5  n  1.4) Rec ¼ 0:214 þ 89:33n  177:77n2 þ 153:08n3  63:29n4
þ 10:13n5 (12)
Rec ¼ 8:22þ39:45n61:03n2 þ39:25n3 9:1n4 (11)
The range of power-law index for the apex facing the upstream
 apex facing the downstream direction (0.2  n  1.8) direction is limited to 0.5 and 1.4 due to very low expected values of
2036 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

detected easily. In the present study, the onset of vortex shedding is


judged by examining the vorticity contours and streamline patterns
near the cylinder. Fig. 4(A) and (B) show the vorticity contours near
the cylinder for three values of the power-law index of 0.3, 1 and 1.8,
and for two values of Reynolds number for both orientations of the
triangular cylinder. In both figures, for all values of power-law index,
at the first Reynolds number vortices are either standing (steady
flow) or flapping (seen in highly shear-thinning fluids only) while at
the second value of the Reynolds number shedding is seen to have
been initiated. For instance, in Fig. 4(A) for n ¼ 0.3, the flow is seen to
be asymmetric because of flapping of vortices at Re ¼ 34, while at
Re ¼ 35 vortex shedding is clearly seen to have started. So for the
power-law index of 0.3, the critical Reynolds number ðRec Þ for the
apex of the cylinder facing the upstream direction lies in the range
34  Re  35. This methodology was used to obtain the critical
values of the Reynolds number for the other values of the power-law
index and also for the cylinder with its apex facing in the down-
stream direction. The critical Reynolds numbers reported here are
thus accurate to within the band of 0.5. The dependence of the
critical Reynolds number on the power-law index is shown in Fig. 5.
The critical Reynolds number Rec goes through a maximum value
Fig. 3. Variation of critical Reynolds number ðRec Þ with power-law index. for shear-thinning fluids while it continuously decreases with
power-law index for shear-thickening fluids. Results for triangular
critical Reynolds number ðOð103 ÞÞ. For this range of Reynolds cylinder presented here are qualitatively similar to that for a circular
number, much larger computational domain is required than that cylinder [35]. The dependence of the critical Reynolds number on
used in this study. Lastly, Eq. (11) predicts the present results with power-law index, n, can be described by the following fits:
an average error of 1% which rises to a maximum of 2.5%. The cor-
responding values for Eq. (12) are 1.1% and 4% respectively.  apex facing the upstream direction

4.3. Onset of vortex shedding Rec ¼ 22:47þ53:46n52:02n2 þ19:48n3 2:97n4 (13)

At Re ¼ Rec , the pair of steady state eddies formed as seen above  apex facing the downstream direction
grow in size with the increasing Reynolds number up to Re  Rec .
Thus, the flow is steady up to Re  Rec . For Re  Rec, the eddies Rec ¼ 24:85 þ 28:41n  19:67n2 þ 2:96n3 (14)
become unstable and vortices start to shed from the body and The average and maximum deviations for the apex of the cylinder
thereby making the flow to be unsteady which is periodic in nature facing upstream direction are 0.74% and 1.9% and for the apex facing
Also, the alternate vortex shedding leads to unbalanced pressure downstream direction are 0.41% and 0.83% respectively. Further-
distribution over the surface of the cylinder. Thus, for instance, more, for both orientations, the critical Strouhal number (at
surface pressure drops significantly at the corner of the cylinder as Re ¼ Rec ) decreases as the power-law index is gradually increased
vortex shedding occurs, but this is accompanied by a concomitant from 0.2 to 1.8. Similarly, the critical drag coefficient (time aver-
increase in skin friction. Consequently, alternating viscous and aged) is seen to attain minimum values in shear-thinning fluids at
pressure forces of varying magnitudes are exerted on the cylinder at about nx0:6 whereas, on the other hand, it increases mono-
different instants of time. Naturally owing to the shear-dependent tonically for shear-thickening fluids.
viscosity of power-law fluids, additional complications arise and The non-monotonic behaviour of both critical Reynolds number
the role of skin friction gets accentuated in this case. In the present Rec and Rec as seen in, Figs. 3 and 5, is believed to arise from the
work, it was found that on increase in the Reynolds number, the interaction between the two non-linear terms (inertial and viscous)
standing eddies behind the cylinder in the steady flow primarily flap present in the momentum equations which scale differently with
only up to some Reynolds number and then start to shed at a slightly velocity. For power-law fluids, as argued above, the viscous term
higher values of the Reynolds number, Re. In the flapping zone, scales as UNn whereas the inertial term will scale as U 2 . So at low
N
simulations were run for sufficiently long time to ensure that the Reynolds numbers, the viscous term diminishes with the decreasing
vortices flap only without shedding. However, this phenomenon value of power-law index for shear-thinning fluids but it increases
was observed only in highly shear-thinning fluids, i.e., small values with velocity in shear-thickening fluids. Also, the viscosity of
of n. It is readily conceded that the spatial decay of velocity field is a shear-thinning fluid becomes very large as the shear rate
very rapid in shear-thinning fluids and therefore, as remarked decreases and it thus tends to an infinite value far away from the
earlier, the effective viscosity of fluid increases rapidly away from cylinder where the shear rate tends to zero. Therefore, the viscous
the cylinder. Thus, in the initial stages, the instabilities induced due effects always dominate the flow characteristics even far away from
to the low viscosity in the vicinity of the cylinder are suppressed by the cylinder under these conditions. Conversely, the viscosity of
the surrounding highly viscous mass of fluid. Naturally, the rate of shear-thickening fluids is a maximum close to the submerged
change of viscosity with shear rate is strongly influenced by the obstacle and it gradually decreases away from the object.
value of n, such stabilization occurs only when the rate of decrease
of the effective viscosity is steep, i.e., small values of the power-law
index. The change from flapping to shedding of vortices occurs over 5. Drag phenomena and heat transfer in steady flow regime
a very small range of Reynolds number and this range depends
primarily on the value of power-law index. Furthermore, this range It is clear now that the flow of shear-thinning fluids (n < 1)
is too narrow for shear-thickening and Newtonian fluids to be across a triangular cylinder is steady and symmetric for the
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2037

Fig. 4. Vorticity contours near the cylinder for n ¼ 0.3, 1 and 1.8, (A) vertex facing the upstream direction (B) vertex facing the downstream direction.

following range of conditions 0.3  n  1 and Re  30. The effect of covered here is believed to be sufficient to delineate the scaling of
power-law index, Reynolds number and Prandtl number on pres- the Nusselt number with the Prandtl number. The variation of CDp
sure and total drag coefficient and heat transfer from triangular and CD with Reynolds number is shown in Figs. 6 and 7 respectively.
cylinder for shear-thinning fluid in the steady flow regime is dis- Similar to the trends seen for a circular [23] and a square cylinder
cussed in this section. The computations have been carried out for [29], the pressure and total drag coefficients show an inverse
1  Re  30, 0.3  n  1 and Pr ¼ 1, 5, 10, 50 and 100. These results dependence on the Reynolds number in the range of conditions
are restricted to shear-thinning fluids primarily because shear- studied herein. Furthermore, a cylinder with its apex facing
thinning behaviour is encountered much more frequently than downstream has a little higher values of drag coefficient CD and of
the shear-thickening behaviour in industrial practice [20]. the pressure component CDp than that for a cylinder with its apex
Furthermore, owing to their high viscosity levels, the Prandtl facing the upstream direction, at the same value of the Reynolds
number can reach as high a value as 1000, but the 100-fold range number. The difference between the two values is very small
2038 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

to flow separation. From another vantage point, one can argue as


the drag here is dominated by form drag (e.g., see Tables 1e6)
therefore if one were to push a triangle at a constant velocity into
a fluid which is at rest, the required force will be lower when the
apex coincides with the forward stagnation point than that for the
opposite orientation. This is simply due to the fact that the fluid will
be gradually pushed away in the first case as opposed to that a large
amount of fluid will get suddenly displaced right at the start in the
latter case. Furthermore, the effect of power-law index is seen to be
much more pronounced at low Reynolds numbers than that at high
Reynolds numbers. This is simply due to the fact that at high Rey-
nolds numbers, the inertial forces far outweigh the viscous forces
under these conditions. Consequently, the viscous characteristics
are of little relevance here.
Fig. 8 shows representative isotherms close to the surface of the
triangular cylinder for n ¼ 0.3 at Pr ¼ 1 and 100. For the sake of
comparison, the number of contours is maintained same in all
plots. As expected, it can clearly be seen that the thermal boundary
layer is much thicker at Pr ¼ 1 than that at Pr ¼ 100. For a cylinder
with its apex oriented in the upstream direction, Fig. 8(a) and (b),
isotherms are crowded near the apex while they are widely spread
Fig. 5. Variation of critical Reynolds number ðRec Þ with power-law index. out near the apex (Fig. 8(c) and (d)) when the apex faces the
downstream direction. The variation of the local Nusselt number
Oð102 Þ at low Reynolds numbers and it increases with Reynolds over the faces of the triangular cylinder, at extreme values of
number. This is simply due to the fact that at low Reynolds numbers power-law index and Reynolds numbers are shown in Fig. 9. Since
the fluid elements are able to follow closely the contour of the the flow is symmetric about the mid-plane, the variation at the top
cylinder whereas at high Reynolds numbers higher drag results due half of the cylinder only is shown here. For the cylinder with apex

a b

Fig. 6. Variation of pressure drag coefficient with Reynolds number for shear-thinning fluids.

a b

Fig. 7. Variation of total drag coefficient with Reynolds number for shear-thinning fluids.
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2039

is obviously so due to the relatively thick boundary layer at low


a values of the Prandtl number, such as Pr ¼ 1.
As noted earlier, while detailed variation of the Nusselt number
along the surface of the cylinder affords useful physical insights
into the relative contribution of each surface, it is the overall
surface-averaged value of the Nusselt number which is frequently
b needed in process design calculations. The dependence of the
surface averaged Nusselt number on the Reynolds number, Re for
a range of values of the power-law index, n ¼ 0.3, 0.6, 0.8 and 1 and
for Pr ¼ 1, 5, 10, 50 and 100 is shown in Fig. 10. It can be seen in all
cases that the average Nusselt number for the case of cylinder with
c apex facing the upstream is higher than that for the cylinder with
apex facing the downstream direction, otherwise under identical
conditions. This is presumably so due to the significant differences
in the extent of streamlining in two orientations, and the resulting
changes in the drag behaviour (hence flow field) described in the
d preceding section. In particular, for the case of vertex facing
upstream, just after the onset of flow separation, wake forms over
the entire face of the cylinder and its size increases with Reynolds
number. On the other hand, for the opposite orientation, the first
signature of the flow separation is seen near the corner and the
center of the recirculation region gradually moves towards the face
of the cylinder with a gradual increase in the Reynolds number.
Fig. 8. Isotherm patterns around the triangular cylinder for n ¼ 0.3 at Re ¼ 30. Thus, for a fixed value of the Reynolds number, the overall wake
region is larger for the apex oriented upstream than the down
stream case. This leads to the reduction in drag and enhancement in
facing the upstream direction, the Nusselt number goes through heat transfer in the case of the vertex oriented upstream. This
a maximum value at the apex (point A) while it is so at the front conjecture is further supported by the fact that, the difference
corner point B for the cylinder with apex facing the downstream between the two values is strongly dependent on the values of the
direction. Also, the variation in Nusselt number over the individual Reynolds and Prandtl numbers. This can safely be ascribed to the
faces is more significant at high Prandtl numbers in comparison to differences in the growth of boundary layers in these two cases.
nearly the same value of the Nusselt number over each face at low Irrespective of the orientation of the triangular cylinder, the Nusselt
Prandtl numbers including the sharp changes at corner points. This number increases with increasing Reynolds and/or Prandtl number

a b

c d

Fig. 9. Local Nusselt number variation over the top half of the triangular cylinder for n ¼ 0.3 and 1 at Re ¼ 30.
2040 A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041

a b

c d

Fig. 10. Variation of average Nusselt number with Reynolds and Prandtl number for n ¼ 0.3, 0.6, 0.8 and 1.

or both. It is also seen that the Nusselt number increases as the 6. Conclusions
degree of shear-thinning increases at fixed values of the Reynolds
and Prandtl numbers. This is clearly due to the lowering of the The influence of power-law index on the appearance of wake
effective viscosity of the fluid elements near the cylinder. and on the beginning of vortex shedding for the flow across an
Conversely, it is thus possible to realize enhancements of up to equilateral triangular cylinder has been investigated numerically. In
about 50e60% in the rate of heat transfer under appropriate addition, the effects of power-law index, Reynolds number and
conditions. Broadly, the degree of enhancement increases with the Prandtl number on the drag coefficients and heat transfer from
increasing Reynolds and Prandtl numbers or both and with the a triangular cylinder to shear-thinning fluids in the steady flow
decreasing value of the power-law index. All these factors lead to regime have also been investigated. The critical Reynolds number
the thinning of the thermal boundary layers. This finding is for the onset of both wake formation and vortex shedding
consistent with that reported for circular, square and elliptic decreases as the value of power-law index increases whereas in
cylinders in the steady flow regime [24,27,29]. shear-thinning fluids both critical Reynolds numbers (Rec and Rec
From an engineering application viewpoint, it is desirable to vs n) go through a maximum as the power-law index increases. For
develop a simple correlating equation which can also be useful to the case of apex facing the upstream direction, the critical Reynolds
interpolate the results for the intermediate values of the Reynolds number for the onset of wake formation is much smaller ðOð102 
and Prandtl numbers and power-law index. The following corre- 101 ÞÞ than that for the case of apex facing the downstream
lation fits the present numerical data for the average Nusselt direction ðOð100  101 ÞÞ. On the other hand, the critical Reynolds
number (apex facing downstream): number ðRec Þ for the onset of vortex shedding for the apex facing
  the flow is a little higher than that for the apex facing the down-
3n þ 1 0:51
Nu ¼ 0:66  Pr0:33  Re0:41  (15) stream direction. Furthermore, for both orientations, the wake
4n
separation and vortex shedding phenomena are seen to be delayed
Eq. (15) has an average deviation of 7.3% which rises to as the fluid behaviour changes from shear-thickening to New-
a maximum of 24.9%. Eq. (15) also fits the corresponding data for tonian, and finally to shear-thinning. In the steady flow regime, the
apex facing upstream provided the numerical value of 0.66 is pressure and total drag coefficients decrease with Reynolds number
replaced by 0.894. In this case, the average error is 12% which reaches and the Nusselt number increases with Prandtl and Reynolds
maximum value of 28%. The index of 1/3 of the Prandtl number in Eq. numbers irrespective of the value of power-law index. The effect of
(15) is consistent with the generally accepted value in the literature power-law index gradually diminishes with the increasing Rey-
and the value of 0.41 for Reynolds number is also in line with that nolds number. Furthermore, shear-thinning is seen to promote heat
cited for a circular cylinder in this range of Reynolds number. transfer under otherwise identical conditions. Thus, it is possible to
A. Prhashanna et al. / International Journal of Thermal Sciences 50 (2011) 2027e2041 2041

achieve 50e60% increase in heat transfer over and above that in [20] R.P. Chhabra, J.F. Richardson, Non-Newtonian Flow and Applied Rheology,
second ed. Butterworth Heinemann, Oxford, 2008.
Newtonian fluids at the same values of the Reynolds and Prandtl
[21] S.J.D. D’Alessio, J.P. Pascal, Steady flow of a power-law fluid past a cylinder,
numbers. All these trends are qualitatively similar to that seen for Acta Mechanica 117 (1996) 87e100.
circular, square and elliptic cylinders. [22] R.P. Chhabra, A.A. Soares, J.M. Ferreira, Steady non-Newtonian flow past
a circular cylinder: A numerical study, Acta Mechanica 172 (2004) 1e16.
[23] R.P. Bharti, R.P. Chhabra, V. Eswaran, Steady flow of power-law fluids across
References a circular cylinder, Can. J. Chem. Eng. 84 (2006) 406e421.
[24] R.P. Bharti, R.P. Chhabra, V. Eswaran, Steady forced convection heat transfer
[1] M.M. Zdravkovich, Flow Around Circular Cylinders, Fundamentals, vol. 1. from a heated circular cylinder to power-law fluids, Int. J. Heat Mass Transfer
Oxford University Press, New York, USA, 1997. 50 (2007) 977e990.
[2] M.M. Zdravkovich, Flow Around Circular Cylinders, Applications, vol. 2. Oxford [25] A.A. Soares, J. Anacleto, L. Caramelo, J.M. Ferreira, R.P. Chhabra, Mixed
University Press, New York, USA, 2003. convection from a circular cylinder to power law fluids, Ind. Eng. Chem. Res.
[3] N. Boisaubert, M. Coutanceau, P. Ehrmann, Comparative early development of 48 (2009) 8219e8231.
wake vortices behind a short semicircular-section cylinder in two opposite [26] P. Sivakumar, R.P. Bharti, R.P. Chhabra, Steady flow of power law fluids across
arrangements, J. Fluid Mech. 327 (1996) 73e99. an unconfined elliptical cylinder, Chem. Eng. Sci. 62 (2007) 1682e1702.
[4] S.A. Nada, H. El-Batsh, M. Moawed, Heat transfer and fluid flow around semi- [27] R.P. Bharti, P. Sivakumar, R.P. Chhabra, Forced convection heat transfer from
circular tube in cross flow at different orientations, Heat Mass Transfer 43 an elliptical cylinder to power-law fluids, Int. J. Heat Mass Transfer 51 (2008)
(2007) 1157e1169. 1838e1853.
[5] T. Sophy, H. Sadat, R. Bouard, Calcul de I’ecoulement autour d’un cylindre semi- [28] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Steady flow of power-law fluids across
circulaire par une method de collocation, C R Mecanique 330 (2002) 193e198. a square cylinder, Chem. Eng. Res. Des. 84 (2006) 300e310.
[6] A. Chandra, R.P. Chhabra, Flow over and forced convection heat transfer in [29] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Heat transfer to power-law fluids from
Newtonian fluids from a semi-circular cylinder, Int. J. Heat Mass Transfer 54 a heated square cylinder, Num. Heat Transfer A 52 (2007) 185e201.
(2011) 225e241. [30] M. Bouaziz, S. Kessentini, S. Turki, Numerical prediction of flow and heat
[7] A.K. De, A. Dalal, Numerical simulation of unconfined flow past a triangular transfer of power-law fluids in a plane channel with a built-in heated square
cylinder, Int. J. Num. Meth. Fluids 52 (2006) 801e821. cylinder, Int. J. Heat Mass Transfer 53 (2010) 5420e5429.
[8] Z. Faruquee, T.V. Olatunji, Steady and unsteady laminar flow past an equi- [31] V.K. Patnana, R.P. Bharti, R.P. Chhabra, Two-dimensional unsteady flow of
lateral triangular cylinder for two different orientations, Proceedings of 5th power-law fluid over a cylinder, Chem. Eng. Sci. 64 (2009) 2978e2999.
joint ASME/JSME fluids engineering conference, San Diego, USA, 2007, pp. [32] V.K. Patnana, R.P. Bharti, R.P. Chhabra, Two-dimensional unsteady forced
1445e1450. convection heat transfer in power-law fluids from a heated cylinder, Int. J.
[9] Z.W. Tian, Z.N. Wu, A study of two-dimensional flow past regular polygons via Heat Mass Transfer 53 (2010) 4152e4167.
conformal mapping, J. Fluid Mech. 628 (2009) 121e154. [33] A.K. Sahu, R.P. Chhabra, V. Eswaran, Two dimensional unsteady laminar flow
[10] A. Sharma, V. Eswaran, Heat and fluid flow across a square cylinder in the two- of a power-law fluid across a square cylinder, J. Non-Newt. Fluid Mech. 160
dimensional laminar flow regime, Num. Heat Transfer Part A 45 (2004) 247e269. (2009) 157e167.
[11] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Flow and heat transfer across [34] A.K. Sahu, R.P. Chhabra, V. Eswaran, Forced convection heat transfer from
a confined square cylinder in the steady flow regime: effect of Peclet number, a heated square cylinder to power-law fluids in the unsteady flow regime,
Int. J. Heat Mass Transfer 48 (2005) 4598e4614. Num. Heat Transfer Part A 56 (2009) 109e131.
[12] A.K. De, A. Dalal, Numerical study of laminar forced convection fluid flow and [35] P. Sivakumar, R.P. Bharti, R.P. Chhabra, Effect of power-law index on critical
heat transfer from a triangular cylinder placed in a channel, J. Heat Transfer parameters for power-law flow across an unconfined circular cylinder, Chem.
129 (2007) 646e656. Eng. Sci. 61 (2006) 6035e6046.
[13] C.P. Jackson, A finite-element study of the onset of vortex shedding in flow [36] P.K. Rao, A.K. Sahu, R.P. Chhabra, Flow of Newtonian and power-law fluids
past variously shaped bodies, J. Fluid Mech. 182 (1987) 23e45. past an elliptical cylinder: a numerical study, Ind. Eng. Chem. Res. 49 (2010)
[14] B.J. Zielinska, J.E. Wesfreid, On the spatial structure of global modes in wake 6649e6661.
flow, Phys. Fluids 7 (1995) 1418e1424. [37] P.K. Rao, A.K. Sahu, R.P. Chhabra, Momentum and heat transfer from a square
[15] J.E. Wesfreid, S. Goujon- Durand, B.J. Zielinska, Global mode behaviour of the cylinder in power-law fluids, Int. J. Heat Mass Transfer 54 (2011) 390e403.
streamwise velocity in wakes, J. Phys. II 6 (1996) 1343e1357. [38] M. Braza, P. Chassaing, H. Ha Minh, Numerical study and physical properties of
[16] H. Abbassi, S. Turki, S.N. Nasrallah, Numerical investigation of forced the pressure and velocity fields in the near wake of a circular cylinder, J. Fluid
convection in a plane channel with a built-in triangular prism, Int. J. Therm. Mech. 165 (1986) 79e130.
Sci. 40 (2001) 649e658. [39] S. Mettu, N. Verma, R.P. Chhabra, Momentum and heat transfer from an
[17] S. Srikanth, A.K. Dhiman, S. Bijjam, Confined flow and heat transfer across asymmetrically confined circular cylinder in a plane channel, Heat Mass
a triangular cylinder in a channel, Int. J. Therm. Sci. 49 (2010) 2191e2200. Transfer 42 (2006) 1037e1048.
[18] P.J. Roache, Perspective: a method for uniform reporting of grid refinement [40] U. Ghia, K.N. Ghia, C.T. Shin, High-Re solutions for incompressible flow using
studies, J. Fluids Eng. 116 (1994) 405e413. the Navier-Stokes equations and a multigrid method, J. Comp. Phys. 48 (1982)
[19] M. Farhadi, K. Sedighi, A.M. Korayem, Effect of wall proximity on forced 387e411.
convection in a plane channel with a built-in triangular cylinder, Int. J Therm. [41] P. Neofytou, A 3rd order upwind finite volume method for generalized
Sci. 49 (2010) 1e9. Newtonian fluid flows, Adv. Eng. Softw. 36 (2005) 664e680.

Das könnte Ihnen auch gefallen