Heat TransferAsian Research, 41 (5), 2012

Lattice Boltzmann Simulation of Mixed Convection in an Inclined Cavity with a Wavy Wall
Mohammad Jafari, Mousa Farhadi, Kurosh Sedighi, and Ehsan Fattahi Faculty of Mechanical Engineering, Babol University of Technology, Babol, Islamic Republic of Iran

In the present study, the effect of inclination on mixed convection heat transfer and fluid flow in a lid-driven cavity with a wavy wall is investigated using the lattice Boltzmann method. The double-population approach with second-order accuracy at velocity and temperature fields is used to simulate the curved boundary in the lattice Boltzmann method. The problem is investigated for different Richardson numbers (0.1 Ri 10), curve amplitudes (0.05 A 0.25), and inclination angles (0 180) when the Reynolds number is equal to 100. Results show that the inclination phenomenon has important effects on both flow and temperature fields at high Richardson numbers. It is also found that the inclination loses its role on mixed convection heat transfer from the wavy wall by the increase of the curve amplitude of the wavy wall for all Richardson numbers. 2012 Wiley Periodicals, Inc. Heat Trans Asian Res, 41(5), 371387, 2012; Published online 4 May 2012 in Wiley Online Library (wileyonlinelibrary.com/journal/htj). DOI 10.1002/htj.21005 Key words: lattice Boltzmann method (LBM), inclination, wavy surface, Richardson number, Nusselt number, lid-driven cavity

1. Introduction The lattice Boltzmann method (LBM) is a numerical technique based on kinetic theory for simulating fluid flows and modeling the physics in fluids. The progress of using the lattice Boltzmann method as a powerful numerical technique to simulate heat transfer and fluid flow problems is obvious in the last decade [110]. The lattice Boltzmann method has well-known advantages such as easy implementation, possibility of parallel coding, and simulating of complex fluid dynamic problems such as multiphase flow [9, 10], porous media [11, 12], nanofluids [13, 14], etc. Convection heat transfer from wavy surfaces is observed in many engineering and scientific applications such as electronic devices, solar collectors, and wavy-plate condensers in refrigerators. In recent years many studies about convection heat transfer and fluid flow in wavy enclosures have been done using traditional CFD methods [1521]. In these studies both natural and mixed convection were analyzed and important parameters such as curve amplitude and number of undulations were investigated in detail. 2012 Wiley Periodicals, Inc. 371

Al-Amiri et al. [15] have conducted a study on mixed convection heat transfer in a lid-driven cavity with a sinusoidal bottom wall using a finite element approach based on the Galerkin method. They studied the effect of the Richardson number, amplitude of surface, and number of undulations on the fluid flow and heat transfer. Their results show an increase in the amplitude of a wavy surface and Reynolds number causes the average Nusselt number to increase. They also reported that the trend of the local Nusselt number has a wavy pattern. Adjlout et al. [16] presented a numerical simulation of natural convection in an inclined cavity with a wavy wall. They used partial differential equations in their work and illustrated the Nusselt number decreases in comparison with the square cavity. They also investigated the effect of inclination on natural convection for different rotation angles. Al-Zoubi and Brenner [22] investigated fluid flow over a sinusoidal surface using LBM. The effect of Reynolds number and geometrical dimensions on the velocity distributions and flow factors were studied in their article. The present study focuses on using LBM to investigate the effect of inclination on mixed convection in a lid-driven cavity with a wavy wall. The effect of the inclination angle on the local Nusselt number distribution and average Nusselt number for various Richardson numbers and amplitudes is investigated. Furthermore, streamline and temperature contours are presented in the results for different conditions. Nomenclature cs: cp: e: F: feq : g: Gr: k: L: Nu: Nuave: Pr: Re: Ri: T: Tref: u0 : w: speed of sound in lattice scale specific heat at constant pressure, kjkg1k1 discrete lattice velocity in direction external force in direction of lattice velocity equilibrium distribution acceleration due to gravity, ms2 Grashof number, g(Th Tc)L3/n2 thermal conductivity, Wm1k1 characteristic length, m local Nusselt number average Nusselt number Prandtl number, n/ Reynolds number, u0L/n Richardson number, Gr/Re2 temperature, k bulk temperature, k, Tref = (Th + Tc) / 2 velocity of lid wall, ms1 weighting factor Greek Symbols : t: : : thermal expansion coefficient, 1k1 lattice time step inclination angle density, kgm3 372

: :

lattice relaxation time dimensionless stream-function Subscripts

ave c: f: h: max:

average cold fluid hot maximum 2. Lattice Boltzmann Method

2.1 Basic theory of LBM By increasing the use of LBM as an alternative numerical method for traditional CFD methods, there is no need to introduce LBM as a new numerical method for simulating fluid flow and heat transfer problems, thus a brief review of LBM technique is presented here. The main form of the lattice Boltzmann equation after LBGK assumptions for both flow and temperature fields can be written as follows [1, 2]: For the flow field: (1)

For the temperature field: (2) where f(x, t) is a distribution function on the mesoscopic level. Here c = x / t is a streaming speed where x and t are the lattice cell size and the lattice time step size, respectively; while e and F are, respectively, the discrete lattice velocity in direction and the external force term in the direction of discrete velocity. This work used the D2Q9 model which has the following values of e for various directions:

(3) Then m and t reveal the lattice relaxation time for the flow and temperature fields, respectively. The kinetic viscosity (n) and the thermal diffusivity () are defined in terms of their respective relaxation 2 times, i.e., n = c2 s (m 1 / 2) and = cs (t 1 / 2), respectively. The relaxation times must be greater eq than 0.5 to satisfy that viscosity and thermal diffusivity are positive, while feq and g are equilibrium distribution functions that are calculated for flow and temperature fields, respectively, as follows: (4)

373

(5) where w is a weighting factor, and is the lattice fluid density. In the D2Q9 model the values of w in different directions are (6)

The Boussinesq approximation is applied for natural convection and radiation heat transfer and viscous dissipation are neglected. Therefore, the force term in Eq. (1) needs to be assumed as below in the needed direction: (7) In LBM Eqs. (1) and (2) are solved in two important steps that are called collision and streaming steps. The collision step is as follows for flow field and temperature field, respectively: (8) (9) The streaming step can be written as follows: (10) (11) ~ ~ where f and g denote the post-collision distribution function. Macroscopic variables can be calculated in terms of these variables, with the following formula for flow density, momentum, and temperature, respectively.

(12)

2.2 Curved boundary modeling For simulating a curved boundary condition, the method used is the same as those presented by Mei et al. [24] for the velocity field and those reported by Guo et al. [25] for the temperature field. These treatments supply second-order accuracy for both the velocity and temperature fields. Figure 1 reveals an arbitrary part of a curved wall boundary. The black points indicate points on the boundary xw, the white points represent the boundary nodes in the fluid region xf, and the grey 374

Fig. 1. Schematic of curved boundary. points show those in the solid region xb. The fraction of an intersected link in the fluid region is defined as follows: (13) ~_ To calculate the distribution function in the solid region f_ (xb, t) based upon the boundary nodes in the fluid region, the procedure used is the same as reported by Mei et al. [23]. The ChapmanEnskog expansion for the post-collision distribution function is

(14) where (15) where ubf and defined the aspect of the value of as follows:

(16)

where uw is the velocity of the solid wall, ubf is the assumed velocity for interpolations, and _ e_ e. For the temperature field in the curved boundary in the present study the applied method is based on a combination of the methods presented by Guo et al. [25]. The distribution function for temperature is divided into two parts, equilibrium and nonequilibrium: (17) So the collision step is obtained as follows:

375

(18)
neq Obviously to calculate ~ g(xb, t + t), both geq (xb, t) and g (xb, t) are required. Equilibrium and non-equilibrium parts of Eq. (11) are defined as

(19) Then T b is determined by linear extrapolation using either:

(20) where Tb1 and Tb2 are defined as follows with respect to : (21a) (21b) where Tf and Tff denote the fluid temperature in node xf and xff, respectively. The next task is to neq determine the gneq (xb, t). Second-order approximation is also used and g (xb, t) is evaluated as (22) 3. Problem Description The vertical walls of the cavity are assumed to be insulated while the wavy bottom surface is maintained at a uniform temperature (Th) higher than the top lid temperature (Tc). The lid wall has a constant velocity motion equal to 0.1 of sound speed in a left to right direction (Fig. 2). The working fluid is assumed to be a Newtonian fluid with constant fluid properties; the flow is considered to be laminar, incompressible, steady, and two-dimensional and viscous dissipation was neglected. The effect of different inclination angles (0 180) on both fluid flow and heat transfer is investigated for different Richardson numbers (0.1 Ri 10), amplitudes of wavy wall (0.05 A 0.25) while the Reynolds number is equal to 100 and the Prandtl number is fixed to 0.71. The wavy wall is defined as follows: (23) where A is the curve amplitude. The local Nusselt number is defined along the wavy wall: (24) where n is the coordinate direction normal to the wavy wall and L is the characteristic length that is calculated as follows: (25) By integrating the local Nusselt number along the wavy surface, the average Nusselt number is calculated as follows: (26)

376

Fig. 2. Geometry of the problem. where s shows the integral is calculated along the wavy line. The numerical code is validated with the results presented by Al-Amiri et al. [15] for mixed convection in a lid-driven cavity with a wavy sinusoidal bottom surface at Gr = 104, Ri = 10, Pr = 1, and A = 0.05 (Fig. 3). In addition, Table 1 shows the comparison of the values of average Nusselt number of this study and those presented by Al-amiri et al. [15] which show good agreement.

Fig. 3. Streamline and temperature contour for Gr = 104, Ri = 10, Pr = 1, A = 0.05, = 0. (Top: Al-amiri et al. [15]; bottom: present study). 377

Table 1. Average Nusselt Number Along the Wavy Wall

A uniform regular Cartesian mesh with 160 160 grid points is used. This number of nodes is selected with a mesh independency treatment that is indicated in Table 2 for two special issues. The value of Nuave for 130 130 grid points has a 2% difference from a 100 100 grid while the difference between 160 160 and 190 190 grid points is less than 0.1; therefore the 160 160 grid satisfies mesh independency in the numerical solution. To get a steady state result in the numerical solution, the results are compared at different time steps. Figure 4 shows the local Nusselt number along the wavy wall for various numbers of time steps. As indicated, when the number of time steps is more than 104 the results converge to a steady state and increasing of the time step after this condition has no clear effect on the values of needed parameters. 4. Result and Discussion In convection heat transfer, mixed convection is defined as a combination of forced convection and natural convection heat transfer. In thermal convection problems, the Richardson number (Ri = Gr/Re2) is a dimensionless parameter that represents the importance of natural convection relative to forced convection. If the Richardson number is much less than unity, the buoyancy force term is unimportant and natural convection is negligible (Ri < 0.1), for high Richardson numbers (Ri > 10), natural convection is dominant and forced convection is negligible, if the Richardson number is of order unity (0.1 < Ri < 10), both forced and natural convections are important. Forced convection is independent from the gravity force thus the inclination has no effect on forced convection. On other hand, natural convection depends on the buoyancy force that is related with the gravity force direction; therefore the inclination plays an important role on both fluid flow and temperature fields when natural convection is dominant. Figures 5 and 6 show the streamlines and temperature contours for different inclination angles at Ri = 0.1 and Ri = 10 when A = 0.05. It is observed when inclination has no visible effect on flow Table 2. Average Nusselt Number Along the Wavy Surface for Different Mesh Grids

378

Fig. 4. Local Nusselt distribution for different time steps at A = 0.05, Ri = 1, and = 0. and temperature fields in Ri = 0.1, different inclination angles make various flow shapes and temperature contours for Ri = 10. At Ri = 0.1, the flow field has a mono-cellular area which is shown in Fig. 5. The lid-driven motion begets a big vortex in the cavity area. The dominant mechanism is forced convection; therefore, the gravitational force does not have any sensible effect on flow field and subsequently on temperature field. Thus the inclination phenomenon is unimportant for this low Richardson number. It can be observed from the minimum and maximum values of streamlines in Fig. 5 which is slightly different for various inclination angles (Fig. 5). On the other hand, when natural convection is dominant (Ri = 10), the flow and temperature fields are completely depended on inclination angles (Fig. 6). Figure 6(a) shows that when the inclination angle is equal to zero, the natural convection flow and the flow created by lid-driven motion are in the same directions which makes a big vortex in the enclosure. Also there is a weak vortex with the opposite direction at the furrow of the wavy wall near the right vertical wall [Fig. 6(a)]. The other parts of Fig. 6 present that for inclination angles bigger than zero [Figs. 6(b)6(f)], the direction of the buoyancy force changes and leads the flow field to have two different patterns in the cavity domain. The lid motion tries to create a clockwise circulation (with negative stream-function value) at the flow domain when the location of hot and cold walls causes a counter clockwise circulation (with positive stream-function value) to be created. Thus under this condition, natural convection creates a big vortex (counter clockwise) and the lid-driven motion makes a smaller vortex (clockwise) under the lid-wall. The values of stream-function of the counter-clockwise circulation decrease significantly with an increase of the inclination angle which is due to the loss of natural convections importance in the flow pattern by increasing the inclination angle from the 0 to 180 [Figs. 6(b)6(g)]. The flow pattern for = 180 in the enclosure domain is presented in Fig. 6(g), because the heated surface is located in the upper position and cold wall location is at the bottom of the domain, there is almost no natural convection. At this inclination angle heat transfer is dominated by 379

Fig. 5. Streamline and isotherm contours for A = 0.05 and Ri = 0.1 at different inclination angles. (a) = 0; (b) = 30; (c) = 60; (d) = 90; (e) =120; (f) = 150; (g) =180. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com/journal/htj.] 380

Fig. 6. Streamline and isotherm contours for A = 0.05 and Ri = 10 at different inclination angles. (a) = 0; (b) = 30; (c) = 60; (d) = 90; (e) =120; (f) = 150; (g) =180. Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com/journal/htj.] 381

conduction heat transfer and the temperature field is completely conductive at this condition which is due to the opposite direction of gravity force created by hot-cold wall direction at = 180 [Fig. 6(g)]. At this inclination angle, there are three circulation vortexes in the enclosure; the main flow vortex is created by lid motion in a clockwise direction, as can be found by the values of the stream-function. The other two vortexes are weak at the bottom of the cavity and their forms are in relation to the wavy wall shape. By comparison between these figures it can be illustrated that the inclination phenomenon has a significant effect on mixed convection for high Richardson numbers where natural convection is dominant. For more consistency, the variations of Nusselt number distribution along the wavy wall for different Richardson numbers are presented in Fig. 7. The results emphasize that the inclination has a greater effect on the rate of heat transfer at a high Richardson number (Ri = 0.1) [Fig. 7(b)] in comparison with a low Richardson number (Ri = 0.1) [Fig. 7(a)]. As presented in this figure, on the wavy-wall surface, the harmonic curve for the local Nusselt number distribution on the surface of wavy wall has the same frequency as that of the wavy surface, and the maximum value occurs

Fig. 7. Local Nusselt number along the wavy wall for different inclination angles at A = 0.05. (a) Ri = 0.1; (b) Ri = 10.

382

approximately at the locations of maximum cross-section of the wavy wall. This is due to the high velocity gradient in this section and subsequently increases the temperature gradient. Conversely, the flow has a low velocity gradient near the wall surface in each corner, which decreases the heat transfer ratio. Therefore, the minimum local Nusselt number is located in a corner of the wavy wall for all inclination angles for each Richardson number (Fig. 7). The compactness of isotherms (presented in Fig. 6) in each section can explain the variation of Nusselt number along the wavy wall for all inclination angles (Fig. 7). Sections with high compact isotherms have a high Nusselt number. Conversely when the isotherms lose their compactness the value of Nusselt number decreases. The overall shape of Nusselt number distribution at = 180 is completely similar to the wavy wall form [Fig. 7(b)] which is due to dominant conduction heat transfer and the same compactness of isotherms near the wavy wall at this inclination angle [Fig. 6(g)]. Decreasing the natural convection term by increasing the inclination angles leads to a decrease of heat transfer rate at a high Richardson number. The values of average Nusselt number with respect to different inclination angles are plotted in Fig. 8 for different Richardson numbers and amplitudes. At a high Richardson number (Ri = 10), the decreasing of Nuave due to the increase of inclination angle is obvious. Making a comparison between different parts of Fig. 8 shows that an increase of the curve amplitude leads to a decrease in the role of inclination on the rate of heat transfer. To make the relation between curve wall amplitude and heat transfer rate at different inclination angles, the values of the average Nusselt number with respect to different curve amplitudes are presented in Fig. 9 for different inclination angles and Richardson numbers. Different parts of Fig. 9 confirm that with an increase of curve amplitude the difference between maximum and minimum values of Nuave at various inclination angles becomes lower for all tested Richardson numbers. Also, it is observed that variation of the amplitude does not have a sensible effect on the heat transfer rate for all inclination angles at low Richardson number (Ri = 0.1) [Fig. 9(a)]. On other hand, by increasing the Richardson number, the amplitude role on the heat transfer rate increases for various inclination angles [Fig. 9(b), 9(c)]. It is observed that an increase of curve amplitude leads to the decrease of heat transfer rate for 90 and an increase for > 90 at Ri = 10. To explain the reason for this obtained result, the values of stream-function at the core of the main vortex (|max|) in the cavity domain are tabulated in Table 3. The presented values show that flow strength is reduced by increasing the amplitude especially when natural convection is dominant. It is logical that the reduction of flow strength causes a decrease of convection heat transfer from the hot wall. On the other hand, an increase of the amplitude of the wavy wall causes the distance between hot and cold walls to decrease. Reduction of this distance increases the rate of conduction heat transfer which eliminates the effects of flow strength reduction on the heat transfer rate. Therefore the values of Nuave have no sensible variation by increasing the amplitude for all investigated inclination angles at Ri = 0.1 [Fig. 9(a)]. This indicates that under this condition enhancement of heat conduction counterbalances the effect of flow strength reduction. At Ri = 10, for 90 increasing the amplitude leads to an extreme flow strength decrease (Table 3). The flow strength reduction decreases the rate of heat transfer. Because of dominant natural convection at high Richardson number, enhancement of the conduction heat transfer cannot recompense the effects of flow strength reduction on the heat transfer rate. Thus, it is seen that at Ri = 10 the increase of the curve amplitude leads to a decrease of the heat transfer rate at 90 [Fig. 9(c)]. 383

Fig. 8. Average Nusselt number for different Ri and . (a) A = 0.05; (b) A = 0.15; (c) A = 0.25. On other hand, by increasing the inclination angle up to 90, the heat transfer pattern is likely to be purely conduction. Thus the role of conduction heat transfer becomes more important in the heat transfer rate. Under this condition, it is observed that an increase of the amplitude causes the heat transfer rate to increase sensibly [Fig. 9(c)] although an increase of the amplitude has a negative effect on flow strength.

384

Fig. 9. Average Nusselt number for different A and . (a) Ri = 0.1; (b) Ri = 1; (c) Ri = 10.

Table 3. The Value of Stream-Function at the Core of Main Vortex (|max|) Aspect to A, , and Ri

385

5. Conclusion The double population treatment in the lattice Boltzmann method (LBM) has been used to simulate fluid flow and heat transfer in a lid-driven cavity with a wavy wall. The effect of inclination phenomenon on mixed convection was investigated for different Richardson numbers (0.1 Ri 10), curve amplitudes of the wavy wall (0.05 A 0.25) and rotation angles (0 180). Some of the most important results that have been achieved in this study are as follows: (a) Richardson number plays an important role in mixed convection of fluids from wavy surfaces. (b) When forced convection is dominant, the inclination has no clear effect on both flow and temperature fields. (c) The effect of inclination on mixed convection rate from the wavy wall decreases by increasing the curve amplitude of the wavy wall at all Richardson numbers. (d) At high Richardson number (Ri = 10), the increase of the curve amplitude leads to a decrease of the heat transfer rate for 90 and an increase for > 90. Literature Cited 1. Succi S. The lattice Boltzmann equation for fluid dynamics and beyond. Oxford; 2001. 2. Mohamad A. Applied lattice Boltzmann method for transport phenomena. Momentum, Heat and Mass Transfer. Sure: 2007. 3. Tsushima S, Moriyama K, Hirai S. Lattice Boltzmann simulation on flow fields connected with multiple side-channels. Heat TransAsian Res 2007;36:96104. 4. Verhaeghe F, Luo LS, Blanpain B. Lattice Boltzmann modeling of microchannel flow in slip flow regime. J Comput Phys 2009;228:147157. 5. Fattahi E, Farhadi M, Sedighi K. Lattice Boltzmann simulation of natural convection heat transfer in eccentric annulus. Int J Therm Sci 2010;49:23532362. 6. van der Sman RGM. MRT lattice Boltzmann schemes for confined suspension flows. Comput Phys Commun 2010;181:15621569. 7. Gao D, Chen Z. Lattice Boltzmann simulation of natural convection dominated melting in a rectangular cavity filled with porous media. Int J Therm Sci 2011;50:493501. 8. Aghajani Delavar M, Farhadi M, Sedighi K. Numerical simulation of direct methanol fuel cells using lattice Boltzmann method. Int J Hydrogen Energy 2010;35:93069317. 9. Park J, Li X. Multi-phase micro-scale flow simulation in the electrodes of a PEM fuel cell by lattice Boltzmann method. J Power Sources 2008;178:248257. 10. Huang H, Li Z, Liu S, Lu XY. Shan-and-Chen-type multiphase lattice Boltzmann study of viscous coupling effects for two-phase flow in porous media. Int J Numer Meth Fluids 2009;61:341354. 11. Jeong N. Advanced study about the permeability for micro-porous structures using the lattice Boltzmann method. Trans Porous Media 2010;83:271288. 12. Xuan YM, Zhao K, Li Q. Investigation on mass diffusion process in porous media based on Lattice Boltzmann method. Heat Mass Transf 2010;46:10391051. 13. Nemati H, Farhadi M, Sedighi K, Fattahi E, Darzi AAR. Lattice Boltzmann simulation of nanofluid in lid-driven cavity. Int Commun Heat Mass Transf 2010;37:15281534. 386

14. Zhou L, Xuan Y, Li Q. Multi-scale simulation of flow and heat transfer of nanofluid with Lattice Boltzmann Method. Int J Multiphase Flow 2010;36:364374. 15. Al-Amiri A, Khanafer K, Bull J, Pop I. Effect of sinusoidal wavy bottom surface on mixed convection heat transfer in a lid-driven cavity. Int J Heat Mass Transf 2007;50:17711780. 16. Adjlout L, Imine O, Azzi A, Belkadi M. Laminar natural convection in an inclined cavity with a wavy-wall. Int J Heat Mass Transf 2002;45:21412152. 17. Nishimura T, Bian Y, Morega AM. Fluid flow and mass transfer in a sinusoidal wavy-walled tube at moderate Reynolds numbers. Heat TransAsian Res 2003;32:650661. 18. Das PK, Mahmud S. Numerical investigation of natural convection inside a wavy enclosure. Int J Therm Sci 2003;42:397406. 19. Kumar BVR. A study of free convection induced by a vertical wavy surface with heat flux in a porous enclosure. Numer Heat Transf Part A 2000;37:493510. 20. Saha S, Sultana T, Saha G, Rahman MM. Effects of discrete isoflux heat source size and angle of inclination on natural convection heat transfer flow inside a sinusoidal corrugated enclosure. Int Commun Heat Mass Transf 2008;35:12881296. 21. Wang CC, Chen CK. Forced convection in a wavy-wall channel. Int J Heat Mass Transf 2002;45:25872595. 22. Al-Zoubi A, Brenner G. Simulating fluid flow over sinusoidal surfaces using the lattice Boltzmann method. Comput Math Appl 2008;55:13651376. 23. Mei R, Yu D, Shyy W, Luo LSh. Force evaluation in the lattice Boltzmann method involving curved geometry. Phys Rev 2002;65:114. 24. Mei R, Luo LS, Shyy W. An accurate curved boundary treatment in the Lattice Boltzmann method. J Comput Phys 1999;155:307330. 25. Guo ZL, Zheng C, Shi BC. An extrapolation method for boundary conditions in lattice Boltzmann method. Phys Fluids 2002;14:20072010.

"F F F"

387