## Abstract

The impact of placing curved obstacles on natural convection in enclosures with differentially heated side walls is analyzed in the current study using the lattice Boltzmann method (LBM). A method to choose characteristic velocity based on Knudsen number is implemented which eradicates the need of arbitrarily guessing characteristic velocities to proceed with simulations. In addition, a less computationally intensive probability distribution function for equilibrium temperature is used. For validation, a standard natural convection problem with left wall at high temperature, right wall at low temperature, and top and bottom adiabatic walls is considered. A grid independence test is conducted and the code is validated with existing results for various Rayleigh numbers, which shows a good agreement. The problem is then modified by including circular and elliptical obstacles of adiabatic, hot, and cold nature. A boundary interpolation technique is used to implement the velocity and temperature boundary conditions at the inner boundaries. The streamline patterns and temperature contours show interesting observations such as dependence of location of vortices on the type of obstacle boundary used, and formation of low or high temperature zones around obstacle at high Rayleigh numbers. Results show that the change in the shape of the obstacle contributes to the Nusselt number variations at the high temperature boundary and low Rayleigh numbers.

## Introduction

Natural convection is known to have a lot of industrial and engineering applications as well in solar dryers, cooling of small electronic systems, and battery cooling systems [1,2]. Natural convection in a cavity has been studied in detail using computational fluid dynamics (CFD) techniques by multiple authors. De Vahl Davis [3] provided a benchmark solution for the case with natural convection in a closed cavity with top and bottom walls with adiabatic boundary condition, the left wall at high temperature and the right wall at low temperature. This study is widely used for validating newly developed natural convection codes. Aydin et al. [4] studied steady natural convection in a two-dimensional (2D) enclosure with a cooled ceiling and isothermal heating from one side and found that in shallow enclosures, the Rayleigh number had a greater bearing on heat transfer. Dalal and Das [5] conducted a study of natural convection inside a rectangular cavity with bottom wall having spatially varying temperature and other walls having constant low temperature using the SIMPLE algorithm with a higher order upwind scheme.

Study on natural convection using lattice Boltzmann method (LBM) have been conducted in the past decades due to the various advantages such as suitability of parallel computing and lesser complexity of equations being solved. Dixit and Babu [6] simulated natural convection in a square cavity using LBM with a very fine grid of wall y + < 0.3 for higher Rayleigh numbers. Arumuga Perumal and Dass [7] reviewed the development of lattice Boltzmann method for macro fluid flows and heat transfer problems. Mejri et al. [8] studied the LBM simulation of natural convection in an inclined triangular cavity filled with water, with a horizontal hot wall, vertical cold wall, and inclined insulated wall. Mansouri et al. [9] used MRT-LBM to study natural convection characteristic in a cavity micropolar fluid in it. Abouricha et al. [10] modeled turbulent natural convection in a heated house where the source of heat was placed at the ground. The lattice Boltzmann simulation of incompressible thermal flows in two and three dimensions has been done by Perumal and Dass [11]. Karki et al. [12] studied laminar natural convection and entropy generation in Rayleigh–Benard (R–B) convection with various nanofluids. Bhopalam et al. [13] have used lattice Boltzmann method to compute incompressible flows in a two-sided oscillating lid-driven cavity. Bhopalam et al. [14] have used lattice Boltzmann method to compute flows in double-sided cross-shaped lid-driven cavities.

Natural convection with obstacles has attracted researchers due to its various applications. Kim et al. [15] studied the natural convection for the difference in temperature between a cold outer square enclosure and a hot inner circular cylinder over a range of Rayleigh numbers 103 to 106. Kim et al. [16] conducted numerical simulations of natural convection in a square enclosure with a cylinder placed inside at a Prandtl number of 0.7 with different bottom wall temperatures using immersed boundary method. They found that a large Rayleigh number caused a significant variation in streamline patterns and isotherms in the enclosure with increase in the bottom temperature. Liao and Lin [17] studied the influence of Prandtl number on instability of natural convection flows within an enclosure with a constant temperature circular cylinder placed at the center for moderate Rayleigh numbers. Karki et al. [18] studied the effects of placing one, two, and four adiabatic square obstacles inside a bounded enclosure at various Rayleigh numbers and studied the influence of the obstacles on Nusselt number at cold wall. Sen et al. [19] have studied mixed convection in a single as well as double-sided lid-driven cavity using lattice Boltzmann method. Mei et al. [20] have used lattice Boltzmann method for three-dimensional flows with curved boundaries. Bararnia et al. [21] have used lattice Boltzmann method to simulate natural convection in a square enclosure having a horizontal elliptical cylinder. Yan and Zu [22] have simulated heat transfer and fluid flow past a rotating isothermal cylinder using lattice Boltzmann method.

A Knudsen number based approach as proposed by Kao et al. [23] is used to select characteristic velocity, thereby eliminating the need to randomly select relaxation times. A modified curved boundary interpolation technique inspired by Yu et al.’s [24] approach is implemented to model the inner boundaries. In the present work, different types of circular and elliptical obstacles are placed inside a cavity. These obstacles are maintained adiabatically, at constant high temperature and at constant low temperature. The adiabatic obstacle helps us understand how the presence of an obstacle influences the flow characteristics within the enclosure without transferring heat to the flow around it. This would provide a case to study purely the effect of geometrical blockage in the domain. The obstacle is then given a constant high and constant low-temperature condition to study the effect of obstacle temperature on the flow and observe how it affects the movement of the fluid on its either side. Physically, contact with a hot wall makes fluid relatively lighter and move upward whereas contact with a cold wall makes the liquid relatively heavier and move downward. Our aim would thus be to search for these flow traits in our results, and check if contact with a hot or cold obstacle would create diagonal asymmetries in flow patterns. The effect of Rayleigh number, obstacle size, and shape on the extent of shift in flow features such as primary vortex is also studied. Laminar flow in steady-state condition is considered for the present study. Results are obtained for various Rayleigh numbers at a fixed Prandtl number of 0.71.

## Problem Specification

The schematics that have been modeled are shown in Fig. 1. The square enclosure walls are stationary. The top and bottom walls are considered adiabatic. The left wall is considered to be at high temperature Th, whereas the right wall is maintained at low temperature Tc. In the first case, a square cavity with no obstacle is considered for validation. The results produced by the developed code are compared with results of De Vahl Davis [3], and a grid independence test is also conducted. Second, circles with diameter d = 0.125L and d = 0.25L, where L represents the length which is taken equal to the number of elements to make sure each element is unit size, are placed at the center of the cavity. After that, an ellipse with horizontal major axis a = 0.25L and vertical minor axis b = 0.125L is placed at the center of the domain. Lastly, an ellipse with horizontal minor axis a = 0.125L and vertical major axis b = 0.25L is placed at the center of the domain. All these obstacles are assigned three different types of boundary conditions: adiabatic, hot, and cold boundary conditions.

Fig. 1
Fig. 1
Close modal

### Methodology.

The thermal LBM model for natural convection uses two probability distribution functions f and g for the flow and temperature field, respectively. The two distributions they obey with respect to the BGK approximation are as follows:
$fi(x+ciΔt,t+Δt)−fi(x,t)=1τ(fieq(x,t)−fi(x,t))+Fi$
(1)
$gi(x+ciΔt,t+Δt)−gi(x,t)=1τs(gieq(x,t)−gi(x,t))$
(2)
Here, Fi is the momentum input due to buoyant flow as expressed later. τ is the relaxation time of the flow field and τs is the relaxation time of the temperature field given by
$τ=3ν+0.5$
(3)
$τs=3νPr+0.5$
(4)
τ and τs should be greater than 0.5 for the algorithm to be stable. The equilibrium distribution functions $fieq$ and $gieq$ are chosen as
$fieq=ρwi[1+3(ci.u)c2+9(ci.u)22c4−3(u.u)2c2]$
(5)
$gieq=Twi[1+3(ci.u)c2]$
(6)
Here, the weights for each direction in the D2Q9 model are computed as
$wi={49ifi=019ifi=1,2,3,4136ifi=5,6,7,8$
(7)
The flow and thermal properties are computed as
$ρ=∑i=08fi$
(8)
$u=1ρ(∑i=08fici)$
(9)
$T=∑i=08gi$
(10)

The aim of the simulation is to reduce the difference between the equilibrium distribution function and probability distribution function for both velocity and temperature. To start the simulation, three important parameters need to be determined:

1. The characteristic velocity for the simulation (unat): It is fixed as a function of Knudsen number Kn, Rayleigh number Ra, and Prandtl number Pr and specific heat γ
$unat=2RaKn2c2πγPr$
(11)

Here, Kn = 10−4, Pr = 0.71, and γ = 1.4 yielding a characteristic velocity such that Mach number is in the incompressible region.

1. The momentum due to buoyant flow [25] gives three ways to approximate the gravitational force term of which the following method is chosen:
$Fi=3wigβT(x,t)ρ(x,t)ciy$
(12)
where
$gβ=unat2H$
(13)
2. The kinematic viscosity ν: The parameter used to evaluate the relaxation time is calculated using the expression:
$ν=unat2H2PrRa$
(14)

### Outer Wall Boundary Conditions.

The boundary conditions for flow are considered to be bounce-back boundary conditions. For instance at the left wall, directions pointing inward to the domain that have a nonzero velocity component perpendicular to the wall are evaluated in terms of known quantities as follows:
$f1(0,y)=f3(0,y)$
(15)
$f5(0,y)=f7(0,y)$
(16)
$f8(0,y)=f6(0,y)$
(17)
The temperature boundary conditions at the left wall (Th = 1) at constant high temperature are evaluated as follows:
$g1(0,y)=w1+w3−g3(0,y)$
(18)
$g5(0,y)=w5+w7−g7(0,y)$
(19)
$g8(0,y)=w6+w8−g6(0,y)$
(20)
The temperature boundary conditions at the right wall (Tc = 0) at constant low temperature are evaluated as follows:
$g1(L,y)=−g3(L,y)$
(21)
$g5(L,y)=−g7(L,y)$
(22)
$g8(L,y)=−g6(L,y)$
(23)
The adiabatic boundary condition at top and bottom walls is evaluated with the bounce-back boundary conditions. For instance, at the top wall
$g4(x,H)=g2(x,H)$
(24)
$g7(x,H)=g5(x,H)$
(25)
$g8(x,H)=g6(x,H)$
(26)

### Inner Wall Boundary Conditions.

The boundary for curved obstacles will not coincide with the nodal points in the domain calling for the need of an interpolation scheme to satisfy appropriate conditions at the boundary. The bounce-back condition for velocity needs to be implemented at the boundary node xb as shown in Fig. 2 as implemented by Yu et al. [24]. During the collision step, value at xff and xf are computed and during the streaming step, these values are streamed along α direction to xf and xb, respectively. Using the values of $fα(xf)$ and $fα(xb)$, the value of $fα(xw)$ is computed as follows:

$fα(xw)=fα(xf)+Δ(fα(xb)−fα(xf))$
(27)
where Δ is computed as
$Δ=|xb−xf||xw−xf|$
(28)
Fig. 2
Fig. 2
Close modal
The bounce-back condition at the boundary is implemented by the following relation:
$fβ(xw)=fα(xw)$
(29)
The values of $fβ(xf)$ are computed by linearly interpolating between $fβ(xb)$ and $fβ(xff)$ as follows:
$fβ(xf)=fβ(xb)+ΔΔ+1(fβ(xff)−fβ(xw))$
(30)
For temperature, a similar treatment is used. The first step remains preserved
$gα(xw)=gα(xf)+Δ(gα(xb)−gα(xf))$
(31)
For adiabatic boundary condition, the bounce-back condition is used
$gβ(xw)=gα(xw)$
(32)
For the cold and warm wall condition, respectively, modified versions of the equation are used
$gβ(xw)=−gα(xw)$
(33)
$gβ(xw)=w(β)+w(α)−gα(xw)$
(34)
The third interpolation step remains similar to the flow field interpolation
$gβ(xf)=gβ(xb)+ΔΔ+1(gβ(xff)−gβ(xw))$
(35)

## Results and Discussion

### Validation.

For validation, the case to be studied is shown in Fig. 1(a). The length of domain L and height of domain H is the number of lattice elements in each direction. This is done to ensure that each discretised cell has unit length and height. Under the section “Results and Discussion”, all figures are plotted in a nondimensional sense, i.e., the x dimension is nondimensionalized by L, whereas the y dimension is nondimensionalized by H. A lattice size independence test is performed with 80 × 80, 100 × 100, and 120 × 120 configurations. The average Nusselt number for simulation is computed as
$Nuavg=1+L⟨uxT⟩αΔT$
(36)
where ΔT = 1 represents the temperature difference between domain walls. Average Nusselt number obtained from the current simulations are compared with those obtained by De Vahl Davis [3] and the results are presented in Table 1 showing a good agreement with published data. The error showed an increase with the increase in Rayleigh number. However as expected the error decreased with the increase of number of nodes. The maximum error for simulation using a 100 × 100 grid was found to be 2.477% at Rayleigh number of 106. Since the simulations are fast and accurate using this lattice size, the lattice is used to proceed with further endeavors. Figure 3 shows the velocity streamlines and temperature contours obtained.
Fig. 3
Fig. 3
Close modal
Table 1

Average Nusselt number for validation case

Ra103104105106
Nu by Vahl Davis1.1182.2434.5198.825
81 × 811.10612.19314.40788.5271
% Error1.06442.22472.46073.3756
101 × 1011.10912.20724.43868.6064
% Error0.79611.59611.77922.477
121 × 1211.11122.2164.45658.6549
% Error0.60821.20371.3831.9274
Ra103104105106
Nu by Vahl Davis1.1182.2434.5198.825
81 × 811.10612.19314.40788.5271
% Error1.06442.22472.46073.3756
101 × 1011.10912.20724.43868.6064
% Error0.79611.59611.77922.477
121 × 1211.11122.2164.45658.6549
% Error0.60821.20371.3831.9274

### Effects of Placing Obstacles in Enclosure

#### Circular Obstacle With d = 0.125H.

Figure 4 shows the impact of placing an adiabatic circular obstacle with diameter d = 0.125H. From the figure it can be seen that at low Rayleigh number (103 − 105), adiabatic obstacle marginally hinder the fluid flow characteristics displayed by a differentially heated block without obstacles (which is shown in Fig. 3). We observe that similar to Fig. 3, the fluid moves up the hot wall and down the cold wall and the obstacle. For all values of Ra, a change in temperature contours is observed surrounding the obstacle and centro-symmetry of velocity streamline patterns can also be seen from the figures. At a higher Rayleigh number of 106, the obstacle causes the development of more than two primary vortices, and a secondary vortex being clearly visible on the right-hand side of the obstacle which can be observed in Fig. 4(g). This is a clear deviation in flow behavior inside the cavity with and without obstacles which strictly forms additional vortices.

Fig. 4
Fig. 4
Close modal

Figure 5 shows the impact of placing a cold circular obstacle with diameter d = 0.125H in the enclosure. In Fig. 5(b), when the cold obstacle is placed at low Rayleigh number, Ra = 103, an extended region of low temperature is formed to the right side of the domain, engulfing the obstacle itself. Unlike the adiabatic case, centro-symmetry of velocity streamlines is not observed for the cold obstacle case. At Ra = 104, the flow field develops a vortex to the left of the obstacle which can be seen in Fig. 5(c). The cold obstacle causes the fluid around it to move downward by increasing its density. This will lead to the formation of a primary vortex on the left side of the circular obstacle as the fluid is moving up along the hot wall. A bounded region of low temperature (T < 0.1) is formed around the circular portion indicating the penetration of slightly higher temperature fields between the obstacle and the cold wall. At Ra = 105, two vortices are noted on either side of the obstacle in Fig. 5(e) but due to the cold nature of the obstacle, the left vortex is larger in size. This occurs as the fluid on the left wall tends to move up the hot wall, whereas the flow close to the obstacle tends to move downward, and thus establishes a more powerful vortex on the left side. This dissimilarity in vortex sizes is not expected in the case of adiabatic obstacles as the obstacle itself just acts as blockage and uniformly changes flow features in both the right half and left half of the domain. The bounded region of low temperature (T < 0.1) as described above are observed for this case in the form of smaller circle. In Fig. 5(g), at Ra = 106, the flow turns chaotic with the development of multiple vortices on both sides of the obstacle. The bounded region of low temperature (T < 0.1) contracts for the obstacle indicating a tendency of higher heat transfer similar to the tendency of compression of isotherms near external walls at constant temperature.

Fig. 5
Fig. 5
Close modal

Figure 6 shows the effect of placing a hot circular obstacle with diameter d = 0.125H which is similar to the cold obstacle with certain exceptions. Since the obstacle now has a high temperature boundary condition, it tends to push the fluid upward by reducing its density. The dominant primary vortex will now move to the right side with the fluid being pushed down by the cold obstacle interacting with the fluid being pushed up by the hot obstacle, thus establishing a dominant loop. On the left side, the vortex becomes weaker as the flow direction of the fluid near the cold left wall and the central obstacle are similar in direction, nullifying the strength of the left vortex, and making it a secondary vortex. The bounded regions of low temperature are replaced with bounded regions of high temperature (T > 0.9). At low Rayleigh number the primary vortex is formed toward the right instead of the left as seen in the cold wall case. Also, the streamline patterns and isotherms look like a mirror image along 45 deg line drawn from the lower left end to the top right end of the enclosure, especially at low Rayleigh numbers of 103 and 104. For isotherms, the value assigned after taking the mirror image should be subtracted from 1 due to the change in type of obstacle.

Fig. 6
Fig. 6
Close modal

#### Circular Obstacle With d = 0.25H.

Figure 7 shows the streamline patterns and isotherms when an adiabatic circular obstacle with diameter d = 0.25H, which is two times the size of the previous obstacle, is placed in the enclosure. The effect is identical to the adiabatic smaller obstacle case where at low Rayleigh numbers (103 − 105), the obstacle has marginal effect on the fluid flow characteristics shown by a differentially heated block in the absence of obstacles. From the figures, we can also observe centro-symmetry of velocity streamline patterns which was also seen in the smaller circular obstacle case. In Fig. 7(g), for a higher Rayleigh number like 106, the development of more than two primary vortices and a secondary vortex are clearly visible on the right-hand side of the circular obstacle.

Fig. 7
Fig. 7
Close modal

Figure 8 shows the impact of placing a cold circular obstacle with diameter d = 0.25H in the enclosure. Similar to the previous cold obstacle case, a vortex on the left side of the larger circular obstacle is formed due to its cold nature which pushes the fluid downward and this combined with the upward motion of the fluid along the hot wall leads to the formation of a primary vortex on the right side. In Fig. 8(b), similar to the smaller circular obstacle at Ra = 103, a region of low temperature is formed on the right side of the domain. At Ra = 104, the streamline patterns in Fig. 8(c) shows a vortex to the left of the obstacle. Also from the isotherms, it is seen that the obstacle is encapsulated by a constant temperature contour region which becomes smaller in area, indicating higher levels of heat transfer near the top portion of the cold outer wall. At Ra = 105, one vortex is formed to the left of the obstacle. In this case, physically a similar trend is observed with respect to placing the smaller cooled circular obstacle. However, it is noticed that the secondary vortex almost disappears. This can be attributed to the size of the cooled obstacle, which cools more fluid down, and thus on the left side, matches the flowrate of fluid flowing down near the right cold wall. Similar amount of cooling results in nullification of the right vortex. This tendency is in direct contrast to the case with adiabatic obstacles, where two clearly distinguishable vortices of equal size are noticed on the right and left side, which implies the effect is not influenced by geometrical blockage, but by the virtue of a larger sink at the center of the domain. At Ra = 106, the flow becomes chaotic. A notable shift of cold region from being concentrated along the low temperature to being dominant near the horizontal adiabatic wall is noticed for the larger obstacle at Ra = 106 (Fig 8(h)).

Fig. 8
Fig. 8
Close modal

Figure 9 shows the effect of placing a hot circular obstacle with diameter d = 0.25H in the enclosure. From the streamline patterns and isotherms, it can be seen that the results obtained are similar to the cold circular obstacle but with some differences. In Fig. 9(b), when Ra = 103, a region of high temperature is formed on the left side of the domain. The hot nature of the obstacle increases the temperature of the fluid around it making it lighter. This combined with the effect of the cold wall leads to a vortex being formed on the right side of the domain, and similar to the colder case, the secondary vortex is suppressed. At Ra = 104, the streamline patterns show a vortex to the right of the obstacle (Fig. 9(c)). At Ra = 105, one vortex is formed to the right of the obstacle (Fig. 9(e)) and at Ra = 105, the flow becomes chaotic (Fig. 9(g)).

Fig. 9
Fig. 9
Close modal

#### Horizontal Elliptical Obstacle.

The streamline patterns and isotherms developed, when an adiabatic horizontal elliptical obstacle is placed in the enclosure, is shown in Fig. 10. For low Rayleigh numbers (103−105), the adiabatic horizontal elliptical obstacle does little to hinder the flow characteristics obtained in differentially heated block without obstacles. For Ra = 103, the circular vortices which were observed in the circular obstacle case have now become elliptical in shape and like the other adiabatic cases, centro-symmetry of velocity streamline patterns can also be seen. In Fig. 10(g), when Ra = 106, the development of more than two primary vortices is noticed with the secondary vortex on the left-hand side of the obstacle unlike the circular obstacles where the secondary vortex was on the right-hand side.

Fig. 10
Fig. 10
Close modal

Figure 11 shows the streamline patterns and isotherms obtained when a cold horizontal elliptical obstacle is placed in the domain. At Ra = 103, an extended region of low temperature is formed on right side of the obstacle (Fig. 11(b)), engulfing it. Just like the circular obstacle cases, the cold wall of the obstacle makes the surrounding fluid colder and heavier which causes the formation of the vortex on its left side as the fluid is moving up the hot wall. For Ra = 104, development of a vortex to the left of the obstacle is observed (Fig. 11(c)). Similar to the larger circular obstacle, a constant temperature contour region is observed which encloses the obstacle and becomes smaller in area. In Fig. 11(e), at Ra = 105, the horizontal ellipse forms one vortex to the left of the obstacle. The flow turns chaotic at Ra = 106 which is indicated by the development of multiple vortices on either side of the obstacle (Fig. 11(g)).

Fig. 11
Fig. 11
Close modal

Figure 12 shows the effect of placing a hot horizontal elliptical obstacle in the enclosure. The hot wall of the obstacle combined with the cold right wall of the enclosure allow the formation of a vortex on the right side of the elliptical obstacle which is a feature also observed in the hot circular obstacles. The bounded regions of low temperature observed in the cold elliptical case are replaced with bounded regions of high temperature (T > 0.9). At low Rayleigh number, the primary vortex is formed toward the right instead of the left as seen in the cold wall case.

Fig. 12
Fig. 12
Close modal

#### Vertical Elliptical Obstacle.

Figure 13 shows the streamline patterns and isotherms developed when an adiabatic vertical elliptical obstacle is kept in the enclosure. For low Rayleigh numbers (103 − 105), just like the other adiabatic obstacles, the adiabatic vertical elliptical obstacle does little to hinder the flow characteristics obtained in differentially heated block in the absence of obstacles and centro-symmetry of velocity streamline patterns is observed for all values of Ra. However in Fig. 13(g), at Ra = 106, unlike the other adiabatic obstacles, the vertical elliptical obstacle does not have a secondary vortex, indicating the decreased tendency with an increased width to height ratio of the object placed.

Fig. 13
Fig. 13
Close modal

Figure 14 shows the effect of placing a cold vertical elliptical obstacle in the domain. At Ra = 103, the obstacle is engulfed by a region of low temperature which is developed on the right side (Fig. 14(b)). Just like the circular and horizontal elliptical obstacle cases, for Ra = 104, a vortex is developed to the left side of the obstacle (Fig. 14(c)) due to the combined effect of the cold obstacle wall which pushes the fluid downward and left hot wall of the enclosure which pushes the fluid upward. Similar to the larger circular obstacle and the horizontal ellipse, a constant temperature contour region is formed which encloses the obstacle and becomes smaller in area.

Fig. 14
Fig. 14
Close modal

At Ra = 105, the vertical ellipse forms two vortices on either side of the obstacle (Fig. 14(e)). For Ra = 106, the flow turns chaotic just like the horizontal ellipse case.

Figure 15 shows the effect of placing a hot vertical elliptical obstacle in the enclosure which is similar to the other hot obstacle cases as the vortex is formed on the right side of the elliptical obstacle as the fluid moves downward along the cold wall and upward close to the hot obstacle due to effect of temperature on density of the fluid. From the streamline patterns and isotherms, it can be seen that the results obtained are similar to the cold wall elliptical obstacle with small differences. At low Rayleigh number the primary vortex is formed toward the right instead of the left as observed in the cold wall case.

Fig. 15
Fig. 15
Close modal

Comparing the results of horizontal and vertical obstacles at Ra = 105, we see that the horizontal extent of the source or sink plays a more dominant effect in suppression of the secondary vortex in comparison with the vertical extent of the source or sink. This could be attributed to the fact that the distance between a point horizontally furthest from the center of the obstacle, and a vertical wall of the domain are lower in the case of the horizontal ellipse in comparison with the vertical obstacle. This implies that a lowering of this particular distance affects flow closer to the wall, thus suppressing the space available to form a secondary vortex. Another trend that is repeatedly observed is the flow features observed for domains with cold obstacles are a mirror image about a diagonal joining the lower right corner to the upper left corner, of flow features observed for domains with warm obstacles. This is justified as the obstacle produces opposing effects as far as movement of the fluid near the obstacle is concerned.

#### Variation in Nusselt Number.

As for Nusselt number variations as shown in Figs. 16, 17, and 18, there is an increase in the hot wall Nusselt number with increase in Rayleigh number and a decrease in the same with the increase of height as can be implied from the crowding of contour lines near the bottom of the hot wall. The variations between Nusselt number of each obstacle decrease with the increase of Rayleigh number clearly indicating the diminishing importance of type of obstacle on the heat transfer near the hot wall due to the crowding of contour lines irrespective of obstacle. It is noted from the figures that the highest hot wall Nusselt numbers at Ra = 103, for cold wall and hot/adiabatic walls, are noted for the small circular and large circular obstacles, respectively. The difference between hot wall Nusselt number is least for the adiabatic case followed by the cold and warm obstacles.

Fig. 16
Fig. 16
Close modal
Fig. 17
Fig. 17
Close modal
Fig. 18
Fig. 18
Close modal

## Conclusion

The lattice Boltzmann method is used to compute the fluid flow and heat transfer characteristics in a bounded enclosure with curved obstacles of different shapes and sizes. An interpolation method is presented to satisfy the adiabatic and constant temperature boundary conditions. Streamline patterns, isotherms, and Nusselt number variations along hot wall are presented and discussed in detail.

When an adiabatic obstacle is placed within the enclosure, very little difference is observed when compared with the results obtained in an enclosure with no obstacles for low Rayleigh numbers. However at higher Rayleigh numbers, the obstacle causes the formation of more than two primary vortices. A secondary vortex is visible on the right-hand side for circular obstacles of both diameters. In the horizontal ellipse configuration, the secondary vortex is noticed instead on the left-hand side. The vertical ellipse does not have a visible secondary vortex, indicating the decreased tendency with an increased width to height ratio of object placed.

With the cold obstacle placed at low Rayleigh number, Ra = 103, an extended region of low temperature is formed to the right side of the domain, engulfing the obstacle itself for all cases. The presence of a cold obstacle in the middle of the domain causes the fluid to move downward due to the increase in its density. This combined with the movement of the fluid up along the hot wall leads to the formation of a dominant vortex on the left side of the obstacle. At Ra = 104 the flow field develops a vortex to the left of the obstacle for all cases. At Ra = 105 two vortices are noted on either side of the obstacle for the small circle and vertical ellipse, while the horizontal ellipse and larger circle forms just one vortex to the left of the obstacle. At Ra = 106 the flow turns chaotic for almost all cases with the development of multiple vortices on both sides of the obstacle.

The behavior of enclosure with hot wall obstacle is similar to that of the cold obstacle with certain exceptions. Since the cold temperature of the obstacle is replaced with a constant high temperature, the dominant vortex is formed on the right side of the obstacle instead of the left. The bounded regions of low temperature are replaced with bounded regions of high temperature (T > 0.9). At low Rayleigh number, the primary vortex is formed toward the right instead of the left as seen in the cold wall case.

The study reveals many interesting observations which include the dependence of location of vortices on the type of obstacle boundary used and formation of low or high temperature zones around the obstacle at high Rayleigh numbers.

## Conflict of Interest

There are no conflicts of interest. This article does not include research in which human participants were involved. Informed consent not applicable. This article does not include any research in which animal participants were involved.

## Data Availability Statement

The authors attest that all data for this study are included in the paper.

## References

1.
Malekshah
,
M. H.
,
Malekshah
,
E. H.
,
Salari
,
M.
,
Rahimi
,
A.
,
Rahjoo
,
M.
, and
Kasaeipoor
,
A.
,
2018
, “
Thermal Analysis of a Cell of Lead-Acid Battery Subjected by Non-Uniform Heat Flux During Natural Convection
,”
Therm. Sci. Eng. Prog.
,
5
, pp.
317
326
.
2.
Piratheepan
,
M.
, and
Anderson
,
T. N.
,
2015
, “
Natural Convection Heat Transfer in Facade Integrated Solar Concentrators
,”
Sol. Energy
,
122
, pp.
271
276
.
3.
De Vahl Davis
,
G.
,
1983
, “
Natural Convection of Air in a Square Cavity: A Bench Mark Numerical Solution
,”
Int. J. Numer. Methods Fluids
,
3
(
3
), pp.
249
264
.
4.
Aydin
,
O.
,
Ünal
,
A.
, and
Ayhan
,
T.
,
1999
, “
Natural Convection in Rectangular Enclosures Heated From One Side and Cooled From the Ceiling
,”
Int. J. Heat Mass Transfer
,
42
(
13
), pp.
2345
2355
.
5.
Dalal
,
A.
, and
Das
,
M. K.
,
2006
, “
Natural Convection in a Rectangular Cavity Heated From Below and Uniformly Cooled From the Top and Both Sides
,”
Numer. Heat Transfer Part A Appl.
,
49
(
3
), pp.
301
322
.
6.
Dixit
,
H. N.
, and
Babu
,
V.
,
2006
, “
Simulation of High Rayleigh Number Natural Convection in a Square Cavity Using the Lattice Boltzmann Method
,”
Int. J. Heat Mass Transfer
,
49
(
3
), pp.
727
739
.
7.
Arumuga Perumal
,
D.
, and
Dass
,
A. K.
,
2015
, “
A Review on the Development of Lattice Boltzmann Computation of Macro Fluid Flows and Heat Transfer
,”
Alexandria Eng. J.
,
54
(
4
), pp.
955
971
.
8.
Mejri
,
I.
,
Mahmoudi
,
A.
,
Abbassi
,
M. A.
, and
Omri
,
A.
,
2016
, “
LBM Simulation of Natural Convection in an Inclined Triangular Cavity Filled With Water
,”
Alexandria Eng. J.
,
55
(
2
), pp.
1385
1394
.
9.
Mansouri
,
A. E.
,
Hasnaoui
,
M.
,
Amahmid
,
A.
,
Dahani
,
Y.
,
Alouah
,
M.
,
Hasnaoui
,
S.
,
Khaoula
,
R.
,
Ouahas
,
M.
, and
Bennacer
,
R.
,
2017
, “
MRT-LBM Simulation of Natural Convection in a Rayleigh-Benard Cavity With Linearly Varying Temperatures on the Sides: Application to a Micropolar Fluid
,”
Front. Heat Mass Transfer
,
9
, p.
28
;
1
14
.
10.
Abouricha
,
N.
,
El Alami
,
M.
, and
Gounni
,
A.
,
2019
, “
Lattice Boltzmann Modeling of Natural Convection in a Large-Scale Cavity Heated From Below by a Centered Source
,”
ASME J. Heat Transfer-Trans. ASME
,
141
(
6
), p.
62501
;
1
9
.
11.
Perumal
,
D. A.
, and
Dass
,
A. K.
,
2014
, “
Lattice Boltzmann Simulation of Two- and Three-Dimensional Incompressible Thermal Flows
,”
Heat Transfer Eng.
,
35
(
14–15
), pp.
1320
1333
.
12.
Karki
,
P.
,
Perumal
,
D. A.
, and
Yadav
,
A. K.
,
2022
, “
Comparative Studies on Air-, Water- and Nanofluids-Based Rayleigh–Benard Natural Convection Using Lattice Boltzmann Method: CFD and Exergy Analysis
,”
J. Therm. Anal. Calorim.
,
147
, pp.
1487
1503
.
13.
Bhopalam
,
S. R.
,
Perumal
,
D. A.
, and
Yadav
,
A.
,
2021
, “
Computational Appraisal of Fluid Flow Behavior in Two-Sided Oscillating Lid-Driven Cavities
,”
Int. J. Mech. Sci.
,
196
, p.
106303
.
14.
Bhopalam
,
R. S.
,
Perumal
,
D. A.
, and
Yadav
,
A.
,
2018
, “
Computation of Fluid Flow in Double Sided Crossshaped Lid-Driven Cavities Using Lattice Boltzmann Method
,”
European J. Mech. B/Fluids
,
70
, pp.
46
72
.
15.
Kim
,
B. S.
,
Lee
,
D. S.
,
Ha
,
M. Y.
, and
Yoon
,
H. S.
,
2008
, “
A Numerical Study of Natural Convection in a Square Enclosure With a Circular Cylinder at Different Vertical Locations
,”
Int. J. Heat Mass Transfer
,
51
(
7
), pp.
1888
1906
.
16.
Kim
,
M.
,
Doo
,
J.
,
Park
,
Y.
,
Yoon
,
H.
, and
Ha
,
M.
,
2014
, “
Natural Convection in a Square Enclosure With a Circular Cylinder According to the Bottom Wall Temperature Variation
,”
J. Mech. Sci. Technol.
,
28
(
12
), pp.
5013
5025
.
17.
Liao
,
C. C.
, and
Lin
,
C. A.
,
2015
, “
Influence of Prandtl Number on the Instability of Natural Convection Flows Within a Square Enclosure Containing an Embedded Heated Cylinder at Moderate Rayleigh Number
,”
Phys. Fluids
,
27
(
1
), p.
013603
.
18.
Karki
,
P.
,
Yadav
,
A. K.
, and
Arumuga Perumal
,
D.
,
2018
, “
Study of Adiabatic Obstacles on Natural Convection in a Square Cavity Using Lattice Boltzmann Method
,”
ASME J. Therm. Sci. Eng. Appl.
,
11
(
3
), p.
034502
.
19.
Sen
,
S.
,
Perumal
,
D. A.
, and
Yadav
,
A. K.
,
2020
, “
Numerical Study of Mixed Convection in Single
and
Double Sided Lid Driven Cavity Using Lattice Boltzmann Method
,”
International Conference on Computational Sciences—Modelling, Computing and Soft Computing (CSMCS-2020)
,
NIT Calicut
,
Sept. 10–12
, pp.
122
133
.
20.
Mei
,
R.
,
Shyy
,
W.
,
Yu
,
D.
, and
Luo
,
L.-S.
,
2000
, “
Lattice Boltzmann Method for 3-D Flows With Curved Boundary
, ”
J. Comput. Phys.
,
161
(
2
), pp.
680
699
.
21.
Bararnia
,
H.
,
Soleimani
,
S.
, and
Ganji
,
D.
,
2011
, “
Lattice Boltzmann Simulation of Natural Convection Around a Horizontal Elliptic Cylinder Inside a Square Enclosure
,”
Int. Commun. Heat Mass Transfer
,
38
(
10
), pp.
1436
1442
.
22.
Yan
,
Y.
, and
Zu
,
Y.
,
2008
, “
Numerical Simulation of Heat Transfer and Fluid Flow Past a Rotating Isothermal Cylinder—A LBM Approach
,”
Int. J. Heat Mass Transfer
,
51
(
9–10
), pp.
2519
2536
.
23.
Kao
,
P. H.
,
Chen
,
Y. H.
, and
Yang
,
R. J.
,
2008
, “
Simulations of the Macroscopic and Mesoscopic Natural Convection Flows Within Rectangular Cavities
,”
Int. J. Heat Mass Transfer
,
51
(
15–16
), pp.
3776
3793
.
24.
Yu
,
D.
,
Ren
,
W.
, and
Shyy
,
W.
,
2002
, “
A Unified Boundary Treatment in Lattice Boltzmann Method
,”
American Physical Society, Division of Fluid Dynamics 55th Annual Meeting
, AIAA 2003-953.
25.
Mohamad
,
A. A.
, and
Kuzmin
,
A.
,
2010
, “
A Critical Evaluation of Force Term in Lattice Boltzmann Method, Natural Convection Problem
,”
Int. J. Heat Mass Transfer
,
53
(
5
), pp.
990
996
.