CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedía Engineering 105 (2015) 96 - 104

Procedía Engineering

www.elsevier.com/locate/procedia

6th BSME International Conference on Thermal Engineering (ICTE 2014)

Effect of magnetic field on natural convection in a C-shaped cavity

filled with ferrofluid

Satyajit Mojumder*, Sourav Saha, Sumon Saha, M.A.H. Mamun

Department of Mechanical Engineering, Bangladesh University of Engineering and Technology,Dhaka-1000.

Abstract

In this paper, natural convection heat transfer has been analyzed for a C-shaped cavity filled with ferrofluid. Cobalt-kerosene ferrofluid is selected here for different solid volume fractions = 0 ~ 0.15). Galerkin weighted residual method is applied to obtain the numerical solution. Simulations are carried out for a wide range of Rayleigh (Ra = 103 ~ 107) and Hartmann (Ha = 0 ~ 100) numbers for both the ferrofluid and the plain fluid. From this work, it is found that higher Rayleigh number enhances the heat transfer rate significantly, whereas presence of magnetic field tries to retard convection. Moreover, addition of 15% solid volume fraction of nanoparticle can augment the heat transfer rate up to 52.65% at moderate Rayleigh number compared to the base fluid. By an optimization of these parameters, effective heat transfer control equipment can be developed. Results are interpreted on the basis of streamline and isotherm patterns, and average Nusselt number of the heated wall and average fluid temperature of the cavity.

©2015TheAuthors.PublishedbyElsevierLtd.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-reviewunder responsibility of organizing committee of the 6 th BSME International Conference on Thermal Engineering (ICTE 2014) Keywords.■Ferrofluid; C-shaped cavity; Magnetic field; Natural convection.

1. Introduction

Nanofluid has been given enoromus importance for enhancement of heat transfer due to its wide application fields. It has application in solar thermal collector, desalination, cavity problem [1-4] and so on. It has been reported that nano-sized (0 ~ 100 nm) of metal, metal oxide and carbonaceous particle such as carbon nanotube (cnt) have

* Corresponding author. Tel.:+880-1737-434034 E-mail address: sjit018@gmail.com

1877-7058 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of organizing committee of the 6th BSME International Conference on Thermal Engineering (ICTE 2014) doi:10.1016/j.proeng.2015.05.012

higher thermal conductivity which enhance the effective thermal conductivity of the mixuture of the base fluid with it. Different types of nanofluid such as alumina-water, CM-water, TiO2-water are very common and widely used due to their low cost. On the other hand, cnt, diamond, silver, gold nanoparticles give better heat transfer when mixed with different base fluids such as water, glycol, mineral oil, etc. Among different nanofluids, ferrofluid is actually quite old. Though it has a great potential for better heat transfer, it has been overlooked in the recent years. Carbon nanotube based nanofluids are stirring the scientific world at the moment [5-7]. In most of the cases of ferrofluids, the nano particles are ferrous particles. Besides, some other materials like nickel and cobalt are also used. Moshen and Mofid [8] recently analyzed free convection of nickel ferrofluid in a square cavity which was heated from the below and subjected to a magnetic field. They concluded that larger solid volume fraction of nanoparticles in the base fluid provided better heat transfer. Basically, ferrofluid can be very much effecive when it is applied in a cavity under magnetic field because the nanoparticle of the ferrofluid has magnetic susceptibility. Recently, some studies show that ferrofluid is a very good alterantive to conventional nanofluids [9-10].

Shape of the cavity is an important factor which basically depends on the actual physical problem. Different shapes of cavity have been analyzed for nanofluid in presence of magnetic field extensively. Among them square, rectangular, circular and trapezoidal cavities are very common [11-12]. Bakier [13] analyzed the C-shaped cavity filled with nanofluid, whose sides are open and side walls are discretely heated and reported that heater length had significant effect on the heat transfer.

The objective of this paper is to investigate the effect of implementation of a ferrofluid in a C-shaped cavity subjected to magnetic field and has one of its side walls isothermally heated. Special attention is given to analyze the effect of Rayleigh and Hartmann numbers on the characteristics of convection heat transfer inside the cavity.

2. Problem formulation

2.1. Physical modelling

The details of the C-shaped cavity is shown in Fig. 1. The top and the bottom walls along with some portions of the outer-right vertical walls are kept adiabatic. Left side wall is maintained at a high temperature (Th) and the notched portion of the right side is maintained at a low temperature Tc (< Th). The entire cavity is filled with cobalt-kerosene ferrofluid. The Cartesian co-ordinate system of the problem is well defined and the gravity is acting in the negative direction of the y-axis. The cavity has an equal height and width of L. Aspect ratio of the notch (a/b) is kept unity and a/2L = c/L = 0.25. A magnetic field with strength Bo is applied on the right side of the cavity along the negative horizontal direction. The radiation effect and viscous dissipation are neglected for the simplification of the problem.

Cobalt- kerosene ferrofluid

m B„ -

u=v=0.adiabatic

Magnetic field b

u = v = 0,T = Tc

Fig. 1. Schematic diagram of a C-shaped cavity filled with cobalt-kerosene ferrofluid.

2.2. Mathematical formulation

The two-dimensional steady state continuity, momentum and energy equations are applied to model this problem for flow and thermal fields. The working fluid is assumed to be incompressible Newtonian with constant properties and thermal equlibrium between the ferromagnetic particles and the base fluid are assumed. Boussinesq approximation is applied for the density variation of the fluid. From the above stated assumptions, the non-dimensional governing equations take the form as

5X dY dX \ dX) dY { dY, Here non-dimensional dependent variables are designated by S and corresponding diffusion and source term are defined by rS and SS, respectively and those are summarized in Table 1.

Table 1. A summary of the terms of the non-dimensional governing equations (1).

Equations S rs Ss

Continuity 1 0 0

U-momentum U f Pfflf -dP/dX

V-momentum V f Pfflf -dP/dY + (pß)ffRaPr& / (pt ßf) - Ha2PrV

Thermal energy & af/ af 0

Scales which are adopted to obtain the above non-dimensional governing equation are presented below

x= y,=uL,y=v,P>f )g, e=(Lg>. (2)

L L af af pffaf (— Tc)

In the aforementioned equations, U, V, P and & are non-dimensional velocities, pressure and temperature respectively. Here subscripts ' ff' and f stand for the properties of the ferrofluid and the base fluid respectively.

The non-dimensional governing parameters Rayleigh number (Ra), Prandtl number (Pr) and Hartmann number (Ha) can be defined as

eBfL3 (— T) vf I crf

Ra = ^-^-c,Pr = =BL\—^, (3)

ocfvf af ° \pffvf

where Of indicates the electrical conductivity of the ferrofluid.

The density of ferrofluid which is assumed to be constant can be expressed as,

Pff =(1 ~V)Pj . (4)

Here property of solid ferromagnetic particles (cobalt) is represented by subscript ' s'. In the above equation, solid volume fraction (<) has significant effect on thermal diffusivity of ferrofluid which is quite different from the base fluid and can be modeled as,

where heat capacitance of ferrofluid (pcp)f can be found by,

[PCP)ff = i1 " V){PcP)/ + v{PcP), • (6)

In addition, the thermal expansion coefficient (fif) of the ferrofluid can be obtained as

(pP)ff =(l-q>)(pP)f + q>(pP)s. (7) Moreover, dynamic viscosity of the ferrofluid (uf) and can be expressed as,

Effective thermal conductivity of the ferrofluid can be described as,

f = k, + 2kf -2p(kf -k ) (9)

kf ks + 2kf +p(kf -ks) '

The thermophysical properties of the base fluid (kerosene) and the ferromagnetic particle (cobalt) are given in Table 2.

Table 2. Thermophysical properties of kerosene and cobalt [8].

cP P k ß M

(Jkg-'K"1) (kgm-3) (Wm-'K"1) (K"') (PaS)

Fluid ; : Kerosene 2090 780 0.149 9.9 x 10"4 0.00164

Solid ; : Cobalt 420 8900 100 1.3 x 10"5 -

Boundary conditions in the dimensionless form for the present problems are given in Table 3.

Table 3. Nondimensional boundary conditions of the present problem.

All walls X = 0 0 < Y < 1 0 < X < 1 Y = 0, 1 X=1 0 < Y < 0.25, 0.75 < Y < 1 0.5 < X < 1 Y = 0.25, 0.75 X = 0.5 0.25 < Y < 0.75

U = V = 0 & = 1 cS/êY = 0 dS/dX = 0 & = 0 & = 0

Average Nusselt number is evaluated for the vertical heated wall and calculated from the following expression

kf 1 9©

Nu = --f- J — dY. (10)

kf 0 dX

Average fluid temperature inside the cavity can also be found by the following expression

= A\®dA, (11)

Where A is the non-dimensional area which can be evaluated by A = 1 - ab / I2.

Flow field of the present problem is visualized through streamline which is obtained from stream function. Stream function is defined from velocity components U and V. Relation between the stream function and velocity components for a two-dimensional flow is given by,

U = —, V = -—. (12)

3. Numerical procedure and grid independency test

The entire domain of C-shaped cavity is discritized into triangular mesh elements of different size. Galerkin weighted residual method of finite element analysis is applied to obtain the numerical solution. Non-dimensional governing equations are transformed into a set of algebric equations and iteration technique is employed to find the converged solution. The convergence criterion is set to 10-5, so that \ym+1-ym\ < 10-5 where y is the general dependent variable. A grid sensitivtiy test is performed to ensure the numerical accuracy of the computation. The details of the grid test is shown in Table 4. It can be found that with the increment of number of the elements, the average Nusselt number is decreased and average fluid temperature is increased simultanously. After reaching the grid size having element number 4526, the average Nusselt number and average fluid temperature becomes less sensitive to the number of elements. So for the present problem, grid having element number 4526 is taken as the optimum grid and all the simulations are carried out at this specified grid.

Table 4. Grid sensitivity test for Ra = 105, Ha = 25 with base fluid (y = 0).

Number of elements 136 544 2176 4526 8704

Average Nusselt number (Nu) 3.9209 3.6475 3.5818 3.5689 3.5699

Average fluid temperature (&av) 0.4025 0.4027 0.4027 0.4027 0.4027

4. Code validation

A code validation is performed for checking the reliability of the present code. The present code is compared with the results of Ghasemi et al. [14] on the basis of average Nusselt number for two different Hartmann numbers at Ra = 105. The details of the code validation is presented in Fig. 2. From the figure, it is found that the present code is in very close agreement with the published result. Thus the numerical accuracy of the present work is ensured.

5.0 4.5 r4.0 3.5 3.0

Present (Ra =105& Ha=0) Ghasemi et al. [14] ( Ra =105& Ha=0) Present (Ra =105& Ha=30) Ghasemietal. [14](Ra =105& Ha=30)

0.02 0.04

Fig. 2. Comparison of average Nusselt number as a function of solid volume fraction for Ha = 0 and 30 at Ra = 105 with Ghasemi et al. [14].

5. Results and Discussions

Goal of this investigation is to find out the effect of Rayleigh and Hartmann numbers on natural convection inside a C-shaped cavity containing cobalt-kerosene ferrofluid. Consequently, Rayleigh and Hartmann numbers are varied to compare heat transfer performance between plain fluid and cobalt-kerosene ferrofluid in terms of streamline and isotherm contours, average Nusselt number and average temperature of the fluid. Influence of Rayleigh number is observed by varying its value from 103 to 107 and Hartmann number is varied from 0 to 100 for the present analysis. To intensify the comparison, 15% solid volume fraction of ferromagnetic particles (y = 0.15) is selected as representative of ferrofluid.

5.1. Effect of Rayleigh number

Fig. 3 depicts influence of Rayleigh number on streamline and isotherm contours and compares between flow scenarios for ferrofluid (y = 0.15) and plain fluid (y =0). The analysis shown here is presented for Ha = 25, while varying Rayleigh numbers (Ra = 105, 106 and 107). Figs. 3(a), (b) and (c) show streamline patterns for Ra = 105, 106 and 107 respectively. A large vortex is observed inside the cavity with clockwise rotation for both ferrofluid and plain fluid. The vortex has its center near the middle of the enclosure and adjacent to the left wall. The vortex seems stretched in the narrower regions of the cavity. This particular shape of the vortex is due to the result of thermal buoyant force generated by the heated vertical left wall. Fluid inside the cavity gets heat from the left wall and the density of the fluid becomes lower adjacent to the heated wall. As a result, fluid close to the heated left wall becomes lighter at first and then moves in the upward direction along the wall. Eventually, the heat is transferred to the longest corner of the cavity. On the other hand, fluid near the cold wall has higher density comparatively. Due to its higher density, fluid near the cold wall moves downward creating a clockwise current inside the cavity. This tendency of moving downward is higher near the vertical portion of the cold wall than the horizontal portions of it. As the Rayleigh number is increased, strength of the vortex becomes higher and at Ha = 25, Ra = 107 and y = 0, it

has a value of ¥min = -28.9862. Introducing ferrofluid (< = 0.15) the strength of the central vortex reduce to Wmin = -27.4823. Moreover, the vortex tends to elongate in the horizontal corner direction. This is due to higher temperature difference between the side walls as indicated by high Rayleigh number. By the introduction of ferrofluid, a slight shift in the position of streamlines can be attained without any significant improvement in the strength of the flow. However, more streamlines with higher value of stream function are observed near the middle of the cavity. This occurs because ferrofluid has improved thermal properties, especially higher thermal conductivity compared to plain fluid. From Figs. 3(d), (e) and (f), a comparison between isotherm contours for ferrofluid and plain fluid can be made at Ra = 105, 106 and 107 respectively for Ha = 25. Near the hot and the cold walls, the isotherms are nearly parallel to each other. This pattern indicates the development of thermal boundary layers near the hot and the cold walls. Isotherms near the middle of the enclosure are distorted showing evidence of increased convection. The pattern of isotherm contours is the same for both ferrofluid and plain fluid. However, at Ra = 105 and 106, there is a clear difference in the formation of isotherm contours between ferrofluid and plain fluid.

Fig. 3. Comparison of streamline and isotherm contours for (a), (d) Ra = 105, (b), (e) Ra = 106; (c), (f) Ra = 107 at Ha = 25. Figures in the top row are for streamline, whereas those in the bottom row represent isotherm contours. The solid line in each figure represents the data for plain fluid (p = 0), whereas the dashed line represents ferrofluid with (p = 0.15.

Fig. 4. Variation of (a) average Nusselt number of the heated wall and (b) average temperature of fluid inside the cavity as a function of Ra for Ha = 25. The solid line indicates results for the plain fluid and the dashed line for the ferrofluid.

For ferrofluid, the isotherm contours are less distorted in general and more isotherms with higher values are seen at the middle of the enclosure. This occurs because of the improved thermal conductivity of ferrofluid which ensures increased heat transfer. With higher value of Rayleigh number, the isotherm patterns are greatly distorted at the middle causing better convection. There is no remarkable difference in the isotherms between ferrofluid and plain fluid at Ra = 107 because convection becomes strong for both ferrofluid and plain fluid. At Ra = 107, isotherms become almost horizontal at the middle of the cavity because of very steep temperature difference between the walls.

Figs. 4(a) and (b) show the influence of Rayleigh number on average Nusselt number at the heated wall and average temperature of the fluid for Ha = 25. From Fig. 4(a), it can be observed that with increasing Rayleigh number, average Nusselt number increases for both ferrofluid and plain fluid. Rate of this increment is low in the range of 103 < Ra < 105. When Ra > 10 , average Nusselt number increases at an elevated rate with Ra. Ferrofluid always gives better heat transfer compared to plain fluid except in the region of Ra = 105. In this region, both ferrofluid and plain fluid show equal performance. Fig. 4(b) shows that, average temperature of the ferrofluid inside the cavity increases with increasing Rayleigh number. Rate of increment of average temperature is low up to Ra = 104 and after that average temperature increases very rapidly. Throughout the entire range of Rayleigh number, average temperature of plain fluid is higher than that of the ferrofluid. Both of these outcomes suggest that use of ferrofluid in place of plain fluid might result in better heat transfer.

Fig. 5. Comparison of streamline and isotherm contours for (a), (d) Ha = 10; (b), (e) Ha = 25; (c), (f) Ha = 100 at Ra = 105. Figures in the top row are for streamline, whereas those in the bottom row represent isotherm contours. The solid line in each figure represents the data for plain fluid whereas the dashed line represents ferrofluid.

5.2. Effect of Hartmann number

Effect of Hartmann number on the streamline and the isotherm contours is presented in Fig. 5. Here, Hartmann number is varied from 0 to 100 at Ra = 105. As representative, comparison between ferrofluid and plain fluid in terms of streamlines and isotherms is shown for Ha = 10, 25 and 100. For Ha = 10 and 25, a clockwise rotating vortex is observed inside the cavity. At Ha = 100, this vortex splits into two weak vortices at the top and the bottom of the cavity. The most noticeable change in the streamline pattern is the reduction of stream function value with higher Hartmann numbers. Increasing Hartmann number induces stronger magnetic field, which produces Lorentz force acting on the flow domain. This force reduces the strength of the flow current inside the cavity. Since ferrofluid contains magnetic particle in it, it has higher magnetic susceptibility than plain fluid. Consequently, high Hartmann number influences ferrofluid more intensely than plain fluid. At Ha = 100, weakening of convective current introduces the two weaker vortices at the top and the bottom of the cavity as mentioned earlier. Figs. 5(d),

(e) and (f) show isotherm contours for Ha = 10, 25 and 100 respectively at Ra = 105. For both ferrofluid and plain fluid at Ha = 10 and 25, isotherms are parallel to each other near the heated and the cold walls, while those are distorted in the center region of the cavity. This pattern evolves due to the development of boundary layers near the heated and the cold walls as well as strong convection in the middle of the cavity. Analysis of isotherm contours reveals one interesting point at higher Hartmann number (Ha = 100). When Ha = 100, isotherms for both plain fluid and ferrofluid have similar distribution and are almost vertical. This particular pattern suggests very weak convection where conduction effect is more dominant.

Figs. 6(a) and (b) depict variation in average Nusselt number at the heated wall and average temperature of the fluid with Hartmann numbers at Ra = 105. Fig. 6(a) shows that value of average Nusselt number at the heated wall decreases with increasing Hartmann numbers. It agrees with the previous observation where it was found that Hartmann number has an adverse effect on the flow strength inside the cavity. Average Nusselt number decreases very rapidly up to Ha = 20 for both plain fluid and ferrofluid. At higher value of Hartmann number (Ha > 40), value of Nu becomes more or less constant for ferrofluid, whereas it is decreasing for the plain fluid. Average Nusselt number is always greater for ferrofluid than plain fluid except within the region of 20 < Ha < 30, where the values of Nu for both ferrofluid and plain fluid are almost same. Fig. 6(b) exhibits the change in average temperature of the fluid with Hartmann number at Ra = 105. One can notice from the figure that average temperature is always greater for plain fluid. Increasing Hartmann number causes a reduction in the value of average temperature of the fluid inside the cavity. This reduction occurs at an elevated rate up to low Hartmann number (Ha < 22) for ferrofluid and comparatively moderate Hartmann number (Ha < 40) for plain fluid. At higher Hartmann number (Ha > 40), average temperature of the fluid is almost constant for ferrofluid.

Fig. 6. Variation of (a) average Nusselt number of the heated wall and (b) average temperature of fluid inside the cavity as a function of Ha for Ra = 105. The solid line indicates results for the plain fluid and the dashed line for the ferrofluid.

6. Conclusion

Problem of steady, laminar natural convection inside a C-shaped cavity at varying Rayleigh and Hartmann numbers is investigated thoroughly in this paper for two types of fluid (plain fluid containing no ferromagnetic particle and ferrofluid containing up to 15% solid volume fraction of ferromagnetic particles). It is found that ferrofluid produces higher heat transfer rate at low and high value of Rayleigh numbers. At moderate Rayleigh number, ferrofluid has no significant improvement in heat transfer over plain fluid. Increment in Rayleigh number increases heat transfer rate for both ferrofluid and plain fluid. Rate of increase in average Nusselt number is higher at high value of Rayleigh numbers. On the other hand, increasing Hartmann numbers has the opposite effect than that of increasing Rayleigh numbers. Higher Hartmann number weakens convection for both ferrofluid and plain fluid. However, ferrofluid provides higher value of Nu than plain fluid even under strong magnetic field (very high Ha).

Acknowledgements

The authors would like to thank Department of Mechanical Engineering, Bangladesh University of Engineering and Technology (BUET) for the support provided during this research work.

References

[1] J. M. Wu, J. Zhao, A review of nanofluid heat transfer and critical heat flux enhancement-Research gap to engineering application, Prog. Nucl. Energ. 66 (2013) 13-24.

[2] A. M. Hussein, K.V. Sharma, R. A. Bakar, K. Kadirgama, A review of forced convection heat transfer enhancement and hydrodynamic characteristics of a nanofluid, Renew. Sust. Energy Rev. 29 (2014) 734-743.

[3] L. SyamSundar, M. K. Singh, I. Bidkin, A. C. M. Sousa, Experimental investigations in heat transfer and friction factor of magnetic Ni nanofluid flowing in a tube, Int. J. Heat Mass Transfer 70 (2014) 224-234.

[4] Z. H. Liu, Y. Y. Li, A new frontier of nanofluid research - Application of nanofluids in heat pipes, Int. J. Heat Mass Transfer 55 (2012) 67866797.

[5] M. Ashouri, B. Ebrahimi, M. B. Shafii, M. H. Saidi, M. S. Saidi, Correlation for Nusselt number in pure magnetic convection ferrofluid flow in a square cavity by a numerical investigation, J. Magn. Magn. Mater. 322 (2010) 3607-3613.

[6] T. C. Jue, Analysis of combined thermal and magnetic convection ferrofluid flow in a cavity, Int. Commun. Heat Mass Transfer 33 (2006) 846-852.

[7] H. Aminfar, M. Mohammadpourfard, S. A. Zonouzi, Numerical study of the ferrofluid flow and heat transfer through a rectangular duct in the presence of a non-uniform transverse magnetic field, J. Magn. Magn. Mater. 327 (2013) 31-42.

[8] S. Mohsen, G. B. Mofid, Free convection of ferrofluid in a cavity heated from below in the presence of an external magnetic field, Powder Technol. 256 (2014) 490-498.

[9] G. H. R. Kefayati, Natural convection of ferrofluid in a linearly heated cavity utilizing LBM, J. Mol. Liq. 191 (2014) 1-9.

[10] M. Goharkhah, M. Ashjaee, Effect of an alternating nonuniform magnetic field on ferrofluid flow and heat transfer in a channel, J. Magn. Magn. Mater. 362 (2014) 80-89.

[11] F. Selimefendigil, H. F. Öztop, Effect of a rotating cylinder in forced convection of ferrofluid over a backward facing step, Int. J. Heat Mass Transfer, 71 (2014) 142-148.

[12] S. M. Snyder, T. Cader, B. A. Finlayson, Finite element model of magnetoconvection of a ferrofluid, J. Magn. Magn. Mater. 262 (2003) 269-279.

[13] M. A. Y. Bakier, Flow in open C-shaped cavities: How far does the change in boundaries affect nanofluid?, Engineering Science and Technology, an International Journal, Available online 3 June 2014, ISSN 2215-0986, http://dx.doi.org/10.10167j.jestch.2014.04.007.

[14] B. Ghasemi, S. M. Aminossadati, A. Raisi, Magnetic field effect on natural convection in a nanofluid-filled square enclosure, Int. J. Therm. Sci. 50 (2011) 1748 -1756.