International Journal of Thermal Sciences 50 (2011) 1887e1899

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Buoyancy driven flow within an inclined elliptic enclosure

F.M. Mahfouz 1
Mechanical Engineering Department, UET, Taxila, Pakistan

a r t i c l e i n f o a b s t r a c t

Article history: Buoyancy driven flow and associated heat convection in an elliptical enclosure has been investigated. The
Received 10 October 2010 enclosure which is the space between two horizontal concentric confocal elliptic tubes is heated through
Received in revised form its inner tube surface which is maintained at either uniform temperature or uniform heat flux. The
29 April 2011
induced buoyancy driven flow and the associated heat convection are predicted at different enclosure
Accepted 2 May 2011
orientations. The full governing equations in terms of vorticity, stream function and temperature are
Available online 15 June 2011
solved numerically using Fourier Spectral Method. Beside Rayleigh and Prandtl numbers the heat
convection process in the enclosure depends on the geometry of the enclosure and the angle of incli-
Keywords:
Natural convection
nation with respect to gravity vector. The geometry of the enclosure is represented in terms of major axes
Uniform wall temperature ratio and axis ratio of inner tube. The study considered a moderate range of Rayleigh numbers between
Uniform heat flux 5
103 and 1
105 while Prandtl number is fixed at 0.7. The inner tube axis ratio is considered between
Fourier spectral method 0 and 1 while the ratio between the two major axes is considered up to 3. The angle of inclination of the
Wall temperature minor axes with respect to gravity vector is varied from 0 to 90 . The results for local and average Nusselt
Nusselt number numbers as well as temperature distribution are obtained and discussed together with the details of both
flow and thermal fields. For isothermal heating conditions, the study has shown an optimum value for
major axes ratio that minimizes the rate of heat transfer in the enclosure. While in case of heating at
uniform heat flux the study revealed existence of major axes ratio at which the mean temperature of the
inner wall is maximum. Another aspect of this paper is the prediction of global flow circulation around
the inner tube in case of asymmetrical orientation of the enclosure with respect to the gravity vector.
Ó 2011 Elsevier Masson SAS. All rights reserved.

1. Introduction
Buoyancy driven flow and associated heat convection in an
annular enclosure has long been investigated because of its perti-
nence to many practical engineering applications. These applica-
tions include solar collector receivers, aircraft cabin insulation,
cooling of electronic equipment, nuclear reactor systems, thermal
storage systems and many others. The annular enclosure can be
geometrically formed by the region between either concentric or
eccentric two elliptical tubes. The region between two concentric
elliptical tubes can represent different annuli configurations
ranging from the annulus formed between two concentric circular
tubes to annulus formed by a flat plate surrounded by an elliptical
tube. In such annular spaces the induced flow is initiated due to
temperature difference between the two walls of the enclosure. The
temperature difference, in this study, is developed as a result of
heating the inner wall and cooling the outer wall. The inner wall is
heated and kept either at uniform temperature, UWT, or at uniform
E-mail address: fmahfouz64@hotmail.com.
1
On leave from Department of Mechanical Power Engineering, Menoufia
University, Egypt.
heat flux, UHF while the outer wall is kept at uniform temperature
higher than that of the ambient cooling medium.
In case of UWT natural convection in a horizontal enclosure
between two concentric/eccentric circular tubes, the volume of
previous research is considerable. Kuehn and Goldstein [1] studied
the natural convection between two concentric cylinders both
experimentally and theoretically. Beside providing good, thorough
literature review, the study covered the most details of steady heat
transfer characteristics and flow patterns developed in the enclo-
sure. The authors presented heat transfer results in terms of Nusselt
number and equivalent conductivity. The equivalent conductivity is
defined as the actual heat flux divided by the heat flux that would
occur by pure conduction. The authors have further extended their
experimental study [2] to include the natural convection between
two eccentric isothermal circular tubes; the case which has been
considered later by a number of researchers [3e12]. In case of UHF
natural convection in a horizontal enclosure between two
concentric/eccentric circular tubes only few studies are reported in
the literature [13e18]. In case of eccentric circular enclosures
(including both UWT and UHF), some authors considered the case
of symmetrical enclosure with respect to gravity vector, while some
others considered the case of asymmetric enclosure. In case of

1290-0729/$ e see front matter Ó 2011 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ijthermalsci.2011.05.001
1888 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899

symmetry no global flow circulation is reported, while in case of
asymmetry the global circulation has been reported in the works of
Wang and Bau [9], Guj and Stella [10], Shu et al. [11] and Shu and
Wu [12]. In case of asymmetrical elliptic enclosure such
a phenomenon is expected to take place which, however, has not
been previously reported.
In comparison with the circular concentric and eccentric cases,
a little work has been done on non-circular enclosures. In case of
elliptic enclosure few studies considered the case of UWT [19e22].
Elshamy et al. [19] used the finite-volume approach to investigate
numerically the steady natural convection in horizontal symmetric
annulus formed between two confocal elliptic tubes. They pre-
sented the results in terms of the local and mean Nusselt numbers
together with correlations for the mean Nusselt number. Lee and
Lee [20] have investigated experimentally and theoretically,
through a few test cases, steady natural convection of air in
a symmetric annulus formed between two concentric, confocal
elliptic tubes. They used the conventional finite-difference method
to solve the governing equations that were written in stream
function-vorticity formulation. Cheng and Chao [21] investigated
numerically the steady natural convection in some horizontal
concentric and eccentric elliptical configurations using the finite-
volume method. They presented their heat transfer results for
a number of different symmetrical elliptic configurations,
expressing their results in terms of equivalent conductivity. Zhu
et al. [22] used the Differential Quadrature (DQ) method to study
numerically the steady natural convection in a number of
symmetrical elliptic annuli configurations.
Though its practical importance, heat convection from elliptical
surfaces that heated and kept at uniform heat flux has not been
given the enough attention. The existing studies mostly considered
natural convection from elliptical surfaces that are being placed in
infinite medium [23e25]; the case which is analogous to the elliptic
enclosure with very high major axes ratio. Raithby and Hollands [23]
studied the natural convection from elliptic tube with its major axis
vertical and with different axis ratio and placed in infinite medium.
Merkin [24] solved the boundary layer equations for the case of
natural convection from elliptic tube with major axis either hori-
zontal or vertical. He presented results for local and averaged heat
transfer rates for both constant surface temperature and constant
surface heat flux. Mahfouz and Kocabiyik [25] considered unsteady
free convection from elliptic tube with major axis horizontal.
The previous studies on free convection within elliptic enclo-
sures, as shown above, have focused mainly on symmetric config-
urations and UWT heating. On the other hand, the available studies
on natural convection in case of UHF heating are those related to
elliptic cylinders placed in infinite medium. The only attempts to
investigate the case of asymmetric elliptic enclosure are those made
by Schreiber and Singh [26], Mahfouz and Badr [27] and Sakr et al.
[28]. Schreiber and Singh [26] investigated natural convection in
a horizontal elliptic enclosure placed at different orientations with
respect to gravity vector using stream function vorticity formula-
tion. The study considered only the case of UWT heating with the
numerical results using some commercial CFD software. In their
results, the flow field is presented in terms of velocity vectors from
which the global circulation cannot be ascertained.
The purpose of this piece of work is to predict the steady flow
and thermal fields within elliptic enclosure placed at different
orientations and heated through its inner wall which is maintained
either at uniform wall temperature or at uniform heat flux. More-
over, the study pays attention to the possibility of having global
flow circulation in the enclosure; the phenomenon which, to the
best of author knowledge, has not been reported in case of elliptical
enclosures. The enclosure under consideration is the space that
formed between two concentric, confocal elliptic tubes and its
geometry is decided in terms of the major axes ratio (Mr) and the
axis ratio of inner tube (Ari).
2. Problem statement
The elliptic enclosure geometry, shown in Fig. 1, is the horizontal
two-dimensional annular space between two concentric, long
elliptical tubes of the same focal distance. The minor axes of the
two tubes are aligned and inclined at an arbitrary angle, l with
respect to the gravity vector. The inner tube surface is heated and
maintained at either uniform temperature, Ti or uniform heat flux,
qi while the outer tube surface is cooled and kept at uniform
temperature To (To < Ti). The resulting buoyancy driven flow in the
enclosure is assumed to be laminar and two dimensional. The
density of the fluid is considered constant except in the buoyancy
term where it varies according to Buossensq approximation.
2.1. The governing equations
The induced flow is initially time dependent until it reaches the
steady state after a period of time. The steady state solution can be
directly obtained by solving the steady state governing equations.
However, this study considers the solution of the time dependent
equations as easier, more convenient and less expensive route to
the steady state solution. The time dependent governing equations
of mass, momentum and energy can be written in Cartesian coor-
dinates in terms of stream function, vorticity and temperature as:
!
0 0 0 0 0  
vz v2 z v2 z vj0 vz vj0 vz 1 vFy0 vFx0
¼ n þ  þ þ  (1)
vs vx02 vy02 vy0 vx0 vx0 vy0 r vx0 vy0
ξ
Fb
η
'
x
y'

assumption of zero stream function at both inner and outer walls of

the enclosure. Such an assumption prevents the numerical predic-
tion of the global flow circulation in the enclosure. The study, λ g
however, focused on the special case of steady natural convection
from a flat plate surrounded by an elliptical tube. Mahfouz and Badr ξο ξi
[27] numerically investigated unsteady natural convection in
a horizontal isothermal elliptic enclosure placed at different orien-
tations using Fourier spectral method without paying attention to
the global flow circulation. While Sakr et al. [28] experimentally
investigated the natural convection from an elliptic tube placed at
different orientations and subject to UHF heating within somehow
large circular enclosure. The authors have also provided some Fig. 1. Physical domain and coordinate system.
 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899 1889

!
v2 j0 v2 j0 v2 j v2 j
z0 ¼  þ 02 (2) Jz ¼ þ (6)
vx02 vy vx
2 vh2

! !
vT v2 T v2 T vj0 vT vj0 vT vf v2 f v2 f vj vf vj vf
¼ a þ  0 0þ 0 (3) J ¼ þ þ  (7)
vs vx02 vy02 vy vx vx vy0 vt vx
2 2
v h vx vh vh vx
0
where s is the time, j0 is the stream function, z is the vorticity, where
0
T is the temperature, r is the density, n and a are the momentum x ¼ x0 =ai , y ¼ y0 =ai , t ¼ sa=a2i , j ¼ j0 =a, z ¼ z a2i =a and
and thermal diffusivities respectively. Fx0 ¼ rg bðT  To Þsin ðlÞ, f ¼ ðT  To Þ=DTref .
Fy0 ¼ rg bðT  To Þcos ðlÞ are the components of the net upward J ¼ xx yh  xh yx is the determinant of the Jacobian of coordinate
driving force, Fb. transformation matrix, 3i is the eccentricity of inner tube and is
defined as the ratio of the distance between the ellipse foci to the
2.2. Initial and boundary conditions length of the major axis, Pr ¼ n/a is the Prandtl number,
Ra ¼ g bð2ai Þ3 DTref =na is the Rayleigh number (called modified
To start the time dependent solution an initial condition is Rayleigh number in case of UHF), DTref is the reference temperature
needed. At time t ¼ 0, the inner surface is assumed to be suddenly difference and is equal to Ti  To in case of UWT while it is equal to
heated to either uniform temperature Ti or uniform heat flux, qi while ai qi =k in case of UHF.
the outer tube surface temperature is kept unchanged at To. The fluid In the new coordinates, the steady boundary conditions (4) can
in the enclosure is initially at rest and at the same uniform temper- now be expressed as:
ature, To, of outer wall. The hydrodynamics boundary conditions at
the inner and outer walls are the no-slip condition and imperme- - On the inner tube surface, x ¼ xi
ability. The thermal boundary condition at the inner wall is the
8 9
uniform wall temperature, UWT (T ¼ Ti) or uniform heat flux, UHF <f ¼ 1 in case of UWT =
(q ¼ qi) while the temperature of outer wall is kept uniform at To. vj vj
¼ 0; ¼ 0 and vf (8a)
The steady state boundary conditions can be expressed math- vx vh : ¼ J 1=2 in case of UHF ;
vx
ematically as follows:

- On the inner surface - On the outer tube surface, x ¼ xo

8 vj vj
vj0 v j0 < T ¼ Ti in case of UWT j¼ ¼ 0; ¼ 0 and f ¼ 0 at (8b)
¼ 0; ¼ 0; and vT (4a) vx vh
vx 0 vy 0 : q ¼ k ¼ qi in case of UHF
vSn where xi defines the inner tube surface ð ¼ tanh1 Ari Þ while
Sn is the direction of heat flux normal to the surface. xo defines the outer tube surface ð ¼ tanh1 Aro Þ, Aro ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1  ð1  Ar2i Þ=Mr2 , Mr is the outer to inner tube major axes ratio
- On the outer surface ð ¼ ao =ai Þ.

vj0 vj0
j0 ¼ 0
¼ 0; ¼ 0 and T ¼ To (4b) 3. The method of solution
vx vy0
The stream function, as indicated above, is set to zero on the The method of solution is based on Fourier spectral method in
outer wall while its value on the inner wall will be calculated. The which the dependent variables stream function j, vorticity z and
calculated value of the stream function at the inner wall will temperature f are approximated using Fourier series expansion.
determine both intensity and direction of the global flow circula- The approach is similar to that used by Collins and Dennis [29] and
tion around the inner tube, if exists. Badr and Dennis [30] and Mahfouz and Badr [31]. The dimension-
less stream function j, vorticity z and temperature f are approxi-
2.3. Coordinate transformation mated as:

In order to set the boundary conditions properly without inter- 1 X

N
jðx; h; tÞ ¼ Fo ðx; tÞ þ ½fn ðx; tÞsinðnhÞ þ Fn ðx; tÞcosðnhÞ (9a)
polation, a coordinate system that makes boundary lines coincide 2 n¼1
with coordinates lines should be used. In this problem the elliptic
coordinates x; h defined as x ¼ 3i cosh x cos h, y ¼ 3i sinh x sin h is
adopted. In elliptic coordinates the governing equations read: 1 X
N
zðx; h; tÞ ¼ Go ðx; tÞ þ ½gn ðx; tÞsinðnhÞ þ Gn ðx; tÞcosðnhÞ (9b)
2 n¼1
!  
J vz v2 z v2 z 1 vj vz vj vz X
N
¼ þ 2 þ  1
Pr vt vx
2 vh Pr vx vh vh vx fðx; h;tÞ ¼ Ho ðx;tÞþ ½hn ðx;tÞsinðnhÞþHn ðx;tÞcosðnhÞ (9c)
2 n¼1
3i Ra vf
þ ½coshx sinh sinl  sinhx cosh cosl where Fo, fn, Fn, Go, gn, Gn, Ho, hn and Hn are the Fourier coefficients
8 vx
3i Ra vf and N represents the number of terms considered in the Fourier
þ ½sinhx cosh sinl þ coshx sinh cosl (5) series. Substitution of j, z and f defined in (9a)e(9c) in equations
8 vh
(5)e(7) and integrating these equations (after multiplying each by
1890 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899

1, sinnh and cosnh respectively) from 0 and 2p results in the - At x ¼ xo

following set of differential equations:

2 v2 Fo Fo ¼ Fn ¼ fn ¼ vFn =vx ¼ vfn =vx ¼ vFo =vx ¼ 0 and Ho ¼ Hn ¼ hn ¼ 0

¼ Go cosh2x  G2 (10a)
32i vx2 (13b)
where Do and Dn are the coefficients that result from the expansion
P
! of J 1=2 as a Binomial Series in the form Do þ N 1 Dn cosðnhÞ. From
2 v2 fn 1h i
the condition at inner wall (13a) it can be concluded that the value
n2 fn ¼ gn cosh2x  g þsgnðn2Þgjn2j (10b)
32i vx
2 2 ðnþ2Þ of stream function at inner wall will be equal to jwall ¼ FoðwallÞ =2.
The value of Fo at the wall is determined from the solution of
equation (10a).
! Integrating both sides of equation (10a) with respect to x
2 v2 Fn 1h i
n2 Fn ¼ Gn cosh2x  dn2 Go þGðnþ2Þ þGjn2j (10c) between x ¼ xi and x ¼ xo and using the boundary conditions in
32i vx
2 2 equation (13) gives the following integral condition:

Zxo
 
32i
vGo vG 2
v Go ðGo cosh2x  G2 Þdx ¼ 0 (14a)
32i cosh2x  2 ¼ 2 Pr þ So ðx; tÞ (11a) 2
vt vt vx
2 xi

Similarly, multiplying both sides of equations (10b) and (10c) by

   enx and integrating from x ¼ xi to x ¼ xo and using boundary
vgn 1 vgðnþ2Þ vgjn2j
32i cosh2x  þ sgnðn  2Þ conditions (13), one can obtain the following integral conditions
vt 2 vt vt
!
2
v gn vGo vFo Zxo  i
¼ 2Pr  n2 gn þ nFn  nGn þ Sn1 ðx; tÞ (11b) 32i 1h
vx
2 v x vx cosh2xgn  gðnþ2Þ þsgnðn2Þgjn2j enx dx ¼ 0; (14b)
2 2
xi
  
vGn 1 vG vg vg
32i cosh2x  dn2 o þ ðnþ2Þ þ jn2j 
vt 2 vt vt vt Zxo
32i 1h i
! cosh2xGn  dn2 Go þGðnþ2Þ þGjn2j enx dx ¼ 0; (14c)
v2 G n vG o vF o 2 2
¼ 2Pr  n2 Gn  nfn þ ngn þ Sn2 ðx; tÞ (11c) xi
vx
2 vx vx
where dn2 ¼ 1 if n ¼ 2 and dn2 ¼ 0 if n s 2.
  2 The above integrals are carried out numerically using Simpson’s
vHo vH v Ho
32i cosh2x  2 ¼ 2 2 þ Zo ðx; tÞ (12a) rule to calculate the values of the functions Go, gn and Gn on the
vt vt vx surface of the inner wall ðx ¼ xi Þ. These functions are then used to
predict the vorticity distribution along the inner wall. The vorticity
  
vhn 1 vhðnþ2Þ vhjn2j distribution along the outer wall is calculated from the same three
32i cosh2x  þ sgnðn  2Þ
vt 2 vt vt functions but calculated at the outer wall (x ¼ xo) from the equa-
! tions (10a)e(10c).
2
v hn vHo vFo
¼ 2  n2 hn þ nFn  nHn þ Zn1 ðx; tÞ (12b) The numerical analysis proceeds further by discretizing
vx
2 v x vx equations (10a)e(12c) using finite difference method. The Crank-
Nicolson scheme is used to discretize the time dependent equa-
   tions, (11a)e(12c). The discretization is carried out over uniform
vHn 1 vH vH vH
32i cosh2x  dn2 o þ ðnþ2Þ þ jn2j girds in x direction and with equal time steps till reaching the
vt 2 vt vt vt steady state. The time step is selected very small, especially for
!
2
v Hn vHo vFo higher Ra cases to ensure the stability of the numerical scheme. In
¼ 2  n2 Hn  nfn þ nhn þ Zn2 ðx; tÞ (12c) discretization process all spatial derivatives are approximated using
vx
2 vx vx
second order accurate central finite difference scheme. The first
derivatives at the inner and outer walls are approximated using one
where So, Sn1, Sn2 , Zo, Zn1 and Zn2 are all easily identifiable functions. side second order accurate forward (at inner wall) or backward (at
Equations (10a)e(12c) define a system of differential equations that outer wall) three point finite difference scheme. At every time step
should be solved simultaneously at every time step to get the the coupled non-linear system of equations has been linearized
details of the flow and thermal fields. through an iterative procedure, resulting in tri-diagonal system of
The steady boundary conditions for all Fourier functions pre- equations that has been solved by Tri-Diagonal Matrix Algorithm,
sented in equations (10)e(12) are deduced from (8) and can be TDMA.
expressed as:
3.1. Heat transfer parameters
- At x ¼ xi
After obtaining the stream function and temperature distribu-
Fn ¼ fn ¼ vFn =vx ¼ vfn =vx ¼ vFo =vx ¼ 0 and tions, the flow and heat transfer characteristics are easily deter-
mined. The heat transfer results are presented in terms of local and
½Hn ¼ hn ¼ 0; Ho ¼ 2 in case of UWT while
  mean Nusselt numbers in case of UWT and in terms of local Nussult
vHo vHn vhn number distribution at outer wall and local and mean temperatures
¼ 2Do ; ¼ Dn ; ¼ 0 in case of UHF ð13aÞ
vx vx vx of inner wall in case of UHF.
 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899 1891

In case of UWT local and mean Nussult numbers are the direct Table 1
indicator of the local and mean rates of heat transfer at the inner Grid and number of terms independency test in case of UHF for Ra ¼ 5
104
Ari ¼ 0.5, Mr ¼ 2.25 and l ¼ 45.
and outer walls. The local Nusselt numbers at the inner and outer
walls are defined as Item Np Independence test N Independence test

Np
N fmi Np
N fmi
Nui ¼ hi Li =k; Nuo ¼ ho Lo =k (15)
1 25
15 0.4146 51
05 0.3975
where k is the fluid thermal conductivity and Li and Lo are char- 2 41
15 0.3966 51
10 0.3964
3 51
15 0.3963 51
15 0.3963
acteristic lengths related to the inner and outer walls, hi and ho 4 61
15 0.3963 51
20 0.3963
are the local heat transfer coefficients at inner and outer walls
and are defined as:

hi ¼ qi =DTref ; qi ¼ Q_ i =Pi ¼ kðvT=vSn Þi Z

qo dpo  
a p vHo
THR ¼ ¼  i (20)
ho ¼ qo =DTref ; qo ¼ Q_ o =P o ¼ kðvT=vSn Þo qi pi pi vx o

where, Qi and Qo are the total rate of heat transfer per unit length Numerically, When THR ¼ 0.999 the steady state solution is
of inner and outer walls, W is the width of enclosure and Sn is the reached and the calculation is terminated.
local direction of heat flux normal to the wall. From the above
definitions one can deduce 4. Validation of the numerical results
   
L vf MrLo 1=2 vf The validity of the obtained numerical results depends on the
Nui ¼  i J 1=2 ; Nuo ¼  J ; (16)
ai vx i ao vx o correctness of the mathematical model as well as the accuracy of
Z
1 P the numerical solution. In this study the accuracy of the numerical
The mean Nusselt number is defined as Nu ¼ NudP where
P 0 technique depends on the number of grid points in x direction, Np
and the number of terms N considered in the Fourier series. Both Np
P is the perimeter of the elliptic section. Taking the characteristic
and N in every case are decided after conducting grid and number
lengths Li and Lo as Pi/p and Po/p respectively, the mean Nusselt
of terms independency tests. Table 1 shows one of these tests for
number at inner and outer walls can be expressed as:
both Np and N in case of UHF for Ra ¼ 5
104, Mr ¼ 2.25, Ari ¼ 0.5
    and l ¼ 45 . It can be seen from the table that when Np and N are
vHo vHo equal to or more than 51 and 15 respectively, the computed values
Nui ¼  ; Nuo ¼  (17)
vx i vx o of fmi are almost the same. The number of grid points in x direction
Np is taken as 51 in cases of small Ra and/or Mr while for cases with
Based on the above selection of characteristic lengths Li and Lo,
high Ra and/or Mr this number is increased till it reaches 121. The
the ratio of mean Nusselt number at inner and outer walls is related
number of terms N in Fourier series is taken 10 for small Ra and/or
to the total heat transfer rate from the inner and outer walls as:
Mr and this number increases as Ra and/or Mr increases. The
Nui =Nuo ¼ Q_ i =Q_ o (18) maximum number of terms considered in most cases is 30. It
should be noted that during time dependent computation the
Now, if the whole enclosure is considered as a closed sys- number of terms is taken as 5 at the start of computations
tem bounded by inner and outer walls, the application of the
P _ (time ¼ 0) and then more terms are added as the time increases till
first law of thermodynamics entails that Q W _ ¼ dEsys=dt. reaching N terms at the steady state.
It can be shown that there is no work interaction between In order to ascertain the validity of the mathematical model and
the system and its surroundings and it can also be shown that in accuracy of the numerical solution, the present results are first
the steady state condition there is no change in the system compared with the most relevant results in the literature. The most
energy, Esys with respect to time. Therefore, it can be easily relevant study in case of UWT is that conducted by Elshamy et al.
concluded that in the steady state condition the heat transfer [19]. In that study the authors numerically investigated natural
rate added to the enclosure through the inner wall must be equal convection in air bounded by two confocal horizontal elliptic
to the heat transfer rate rejected from the enclosure through the cylinders. The local and mean Nusselt numbers are determined at
outer wall. Thus when the mean Nussult number at the outer some eccentricities of the inner and outer elliptic cylinders and the
wall approaches that at the inner wall (within tolerance of 0.001) results were validated by comparison with some experimental
the steady state solution is reached and the calculation is data. Thus, the results of Elshamy et al. [19] are used to validate the
terminated. present numerical results. The present result is compared with one
In case of UHF the total heat transfer rate from the inner wall is of the cases considered in the that study in which the geometry of
known and the most important parameter from practical point of the elliptical enclosure is given in terms of the eccentricity of inner
view is the surface temperature of the inner wall, fi which should and outer walls as 3i ¼ 0:9; 3o ¼ 0:4, respectively. In terms of the
be inspected to avoid surface over heating. The value of fi is present geometry parameters it gives Mr ¼ 3i =3o ¼ 2:25 and
qffiffiffiffiffiffiffiffiffiffiffiffiffi
calculated from (9c) and the mean dimensionless temperature of
Ari ¼ 1  32i ¼ 0:436. The comparison is carried out at
the inner wall is then calculated as:
Ra ¼ 1.86
104 and Ra ¼ 3.72
105 (which correspond to Ra ¼ 104
ZP Z2 p and 2
105 in [19] respectively). Due to difference in scaling the
1 1
fmi ¼ fi dp ¼ fi Ji1=2 dh (19) parameters in [19] are modified to match the present parameters.
Pi Pi The comparison of both flow and thermal fields in terms of
0 0
streamlines and isotherms at Ra ¼ 1.86
104 is shown in Fig. 2.
The steady state solution in case of UHF is reached when the Since the case is symmetrical, only one half is considered for
total heat ratio, THR ¼ Q_ o =Q_ i approaches unity. This ratio can be comparison. The figure shows very good agreement of both flow
expressed as and thermal fields at that Ra.
1892 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899

35
Ra=3.72 x 105, Pr=0.7
30 Mr=2.25, Ari=0.436
Present
25 Num. [19]

20
outer wall
Nu
15

10 inner wall
Fig. 2. The streamlines (left) and isotherms (right) patterns for the case of Ra ¼ 18600,
Mr ¼ 2.25 and Ari ¼ 0.436 and comparison with Ref.[19] a) Present b) Ref [19]. 5

0
The comparison of local Nussult number distribution along the 0 30 60 90 120 150 180
inner and outer walls is presented in Fig. 3 in case of Ra ¼ 1.86
104 θ (deg.)
and in Fig. 4 in case of Ra ¼ 3.72
105. The angle q is measured from
the vertical axis counter clock-wise and is related to the coordinate Fig. 4. Local Nu distribution along inner and outer walls for the case of Ra ¼ 3.72
105,
angle h as q ¼ tan1 ðAri tanhÞ  90 where h starts from major axis Mr ¼ 2.25 and Ari ¼ 0.436 and comparison with numerical results of Ref.[17].

counter clock-wise. Fig. 3 shows very good agreement in case of

Ra ¼ 1.86
104 for both inner and outer walls Nu distribution.
In case of UHF the present work is compared with that of yoo
While in case of Ra ¼ 3.72
105, Fig. 4 shows good agreement for
[17] who considered natural convection in concentric circular
Nu along the inner wall but the agreement at outer wall is good
enclosure of radius ratio 2 (Mr ¼ 2) and filled with a fluid of Pr ¼ 0.7.
along most of the surface except20 < q < 20 where noticeable
In that work, Yoo [17] investigated two steady solutions that
difference can be seen. This difference reaches its maximum,
obtained with and without initial flow perturbation. The latter
almost 30%, at q ¼ 0. It seems that at high Ra the work of Elshamy
solution (without perturbation) is relevant to the present solution
et al. [19] underestimates the value of local Nu at outer wall around
and thus is considered for comparison. The comparison is carried
q ¼ 0 which definitely leads to underestimation of the Nu. Such
out at two modified Ra of 45 600 and 8
105 which correspond to
a conclusion is verified by inspecting the Nu at the two walls in [19]
5700 and 105 respectively in Yoo [17]. The comparison is carried out
after modifying them to match the present ones. These modified
for distribution of both the dimensionless temperature along the
values along with present corresponding ones are shown in Table 2.
inner wall, fi and heat flux ratio along the outer wall, qr ¼ qo =qi .
The table shows little overestimation at inner wall but some how
Fig. 6 shows very good agreement for both fi and qr in case of
significant underestimation at outer wall especially at higher Ra. As
Ra ¼ 45 600 while at high Ra ¼ 8
105, Fig. 7 shows good
a further test of the accuracy, the present results for Nu distribution
agreement.
in case of concentric circular enclosure are compared with the
experimental results of Kuehn and Goldstein [1] and numerical
results of Elshamy et al. [19]. The comparison is presented in Fig. 5 5. Results and discussion
which shows good agreement with the experimental results of [1]
while the comparison with the numerical results of [19] is Having the present numerical technique validated, a set of runs
reasonable except around q ¼ 0. is carried out to investigate the effect of Rayleigh number, enclosure
geometry (in terms of Ari and Mr), orientations as well as different
inner surface heating on the heat convection in the enclosure. The
12
study considered a moderate range of Ra (modified Ra in case of
Ra=1.86 x 104, Pr=0.7 UHF) between 5
103 and 1
105 while Prandtl number is fixed at
Mr=2.25, Ari = 0.436
10 0.7. The inner tube axis ratio is considered between 0 and 1 while
Present the ratio between the two major axes is considered up to 3. The
Num. [19] angle of inclination of the minor axes with respect to gravity vector
8 is varied from 0 to 90 . As previously indicated the steady state
solution is reached when the total heat rejected from the outer wall
Nu 6 of the enclosure approaches that pumped into the enclosure
through the inner wall. In case of UWT this solution is obtained
when Nui is almost equal to Nuo while in case of UHF the steady
inner wall
4 solution is reached when the total heat ratio THR approaches one.

2 Table 2
outer wall Comparison of Nu at inner and outer walls with that of Elshamy et al. [19].

Ra Nu Present Nu [19]
0
0 30 60 90 120 150 180 Inner wall Outer wall Inner wall [19] Outer wall [19]
θ (deg.) Observed Observed Observed Modified Observed Modified
1.86
104 3.179 3.178 3.53 3.233 1.19 3.159
Fig. 3. Local Nu distribution along inner and outer walls for the case of Ra ¼ 18600,
3.72
105 6.326 6.325 7.25 6.641 2.22 5.889
Mr ¼ 2.25 and Ari ¼ 0.436 and comparison with numerical results of Ref.[19].
 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899 1893

14 2.5 2.5
Ra=8x105 Mr=2
Ra=9.75*104, Pr=0.7
12 Mr=2.6 2.0 Present 2.0
Num.[17]
Present (Ari=0.998)
10
Exp. [1] 1.5 1.5
qr qr
8 Num.[19] φ
Keq 1.0 1.0
6
φ
0.5 0.5
4

Inner wall 0.0 0.0

2
0 30 60 90 120 150 180
Outer wall θ
0
Fig. 7. Dimensionless temperature distribution along inner and heat flux distribution
0 30 60 90 120 150 180 along outer wall for the case of UHF at Ra ¼ 8
105, Mr ¼ 2.0 and Ari ¼ 0.998 and
θ comparison with numerical results of Ref. [17].

Fig. 5. Local equivalent thermal conductivity distribution along inner and outer walls
for the case of Ra ¼ 9.75
104, Mr ¼ 2.6 and Ari ¼ 0.998 and comparison with
experimental results of Ref.[1] and numerical results of Ref.[19].
Fig. 8 shows the time development of both THR and mean
temperature of inner wall in case of UHF for modified Ra ¼ 5
104,
Ari ¼ 0.5, Mr ¼ 2.25 and at l ¼ 45 . The figure shows that the steady
state mean surface temperature fmi , is reached when the THR is
almost equal to one.
The general characteristics of steady flow and thermal fields for
both UWT and UHF at the same above controlling parameters
(Ra ¼ 5
104, Ari ¼ 0.5, Mr ¼ 2.25 and at angle of l ¼ 45 ) are
studied and the results are presented in Figs. 9 and 10 for heating at
UWT and Figs 11 and 12 for heating at UHF. The streamline patterns
in case of UWT, plotted in Fig. 9a, show that the flow field is formed
of two circulating eddies, one on the right side and another on the
left side. Each eddy is comprised of flow circulating upward along
2.0
Ra=45600, Mr=2
1.6 Present
Num.[17]
the inner wall and downward along the outer wall. The right side
eddy circulates clockwise while the left side eddy circulates counter
clock wise. The two eddies are separated by the zero stream
function line (dotted line in the figure). This line intersects with
outer wall at two points; the upper point which is the flow stag-
nation point on the wall and the lower point which is the flow
detachment point from the wall. In case of symmetry, as presented
in Fig. 2, this line coincides with inner wall and vertically divides
the enclosure into two similar eddies with no global flow circula-
tion. While in case of asymmetry, as in this particular case, the zero
stream function line is no longer vertical nor does it coincide with
inner wall but rather shifted away from the wall, indicating the
existence of global flow circulation. The intensity of global circu-
lation depends on the value of wall stream function (i.e., depend on
the value of function Fo at the wall) which in turn depends on the
controlling parameters. The direction of the global flow circulation
depends on controlling parameters and is decided by the eddy to
which the global flow will be entrained. In this particular case the
global flow has joined the right side eddy and thus circulates clock
2.0
1.6

1.2 qr 1.2
qr
φ
0.8 0.8
φ

0.4 0.4

0.0 0.0
0 30 60 90 120 150 180
θ

Fig. 6. Dimensionless temperature distribution along inner wall and heat flux distri-
bution along outer wall for the case of UHF at Ra ¼ 45,600, Mr ¼ 2.0 and Ari ¼ 0.998 Fig. 8. Time development of THR and fmi in case of UHF at Ra ¼ 5
104, Mr ¼ 2.25,
and comparison with numerical results of Ref. [17]. Ari ¼ 0.5 and l ¼ 45.
1894 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899

a b a b

Fig. 9. Streamlines and isotherms patterns in case of UWT at Ra ¼ 5
104, Mr ¼ 2.25, Fig. 11. Streamlines and isotherms patterns in case of UHF at Ra ¼ 5
104, Mr ¼ 2.25,
Ari ¼ 0.5 and l ¼ 45, a) Streamlines b) isotherms. Ari ¼ 0.5 and l ¼ 45, a) Streamlines b) isotherms.

Fig. 11 shows the streamlines and isotherms in case of UHF for

wise. The isotherms plot for the same case (Fig. 9b) shows the
formation of thermal plume at the top of the inner wall as a result of
the heated upward flow along the two sides of inner wall. The
thermal plume in this case is not vertically symmetric but rather
inclined in the direction of the convected flow towards the upper
part of the outer wall.
In the above case of UWT the characteristics of steady thermal
field near the walls is determined by inspecting local Nusselt
number distribution at the walls. The heat flux distribution along
the inner and outer walls in terms of local Nu is shown in Fig. 10. It
can be observed that the local Nu assumes higher values on the
outer wall in the region of flow impingement on the wall. Such flow
is already heated during its upward motion along the two sides of
hot inner wall and as it hits the cold outer wall high negative
temperature gradient is created and so high heat flux. The highest
heat flux occurs at the point of stagnation, around q ¼ 30. On the
other side, the minimum heat flux (i.e., Nu) at the outer wall occurs
at the point of flow detachment from the lower part of the wall,
around q ¼ 222. In case of the inner wall the figure shows that the
lowest local Nu takes place at the point of flow detachment from
the top part of the wall around q ¼ 12 while the maximum Nu,
unlike the outer wall, does not take place at the point of stagnation
(around q ¼ 198) but rather it occurs at q ¼ 180 where the enclosure
gap is minimum. At that point the heat flux is high due to minimum
resistance to conduction and due to higher velocity gradient (and
so heat convection coefficient) near the walls as well.
the same above values of controlling parameters considered in
UWT. It can be seen that the general streamlines pattern (Fig. 11a) is
almost similar to that in case of UWT. The isotherms plot in this case
(Fig. 11b) is similar only in the formation of thermal plume at the top
of the inner wall. In case of UWT all isotherms (from zero to one)
encircle the inner wall while in case of UHF heating not all isotherms
encircle the inner wall. The isotherms that encircle the inner wall
are only those with values between zero and minimum inner wall
temperature while the isotherms with values higher than this
minimum emanate from the inner wall and do not encircle it. The
inner wall temperature fi , for the same case is shown in Fig. 12.
Shown also in the same figure the distribution of Nu along the outer
wall. It can be seen from the figure that maximum f is reached at the
point of upward flow detachment from the top part of inner wall
around q ¼ 6 while it can be observed that the minimum temper-
ature is reached at almost q ¼ 180. The Nu distribution is qualita-
tively similar to that in case of UWT, that is the highest values are in
the region of impingement around q ¼ 25 while the minimum is
obtained within the lower stagnation region around q ¼ 214. Quick
comparison of Figs. 10 and 12 for the Nu distribution along the outer
wall leads to a conclusion that the Nuo in case of UHF is smaller than
that in case of UWT. This means that for the same Ra the value of
total heat transfer through the same enclosure in case UHF is smaller
than that in case UWT which in turn causes less intensive convec-
tion currents in case of UHF.

1.2 1.2
Mr=2.25 Ra=5x104
16 Ari = 0.5 , λ=45
Mr=2.25 Ra=5x104 1.0 UHF 1.0
Ari = 0.5 , λ=45
UWT Inner wall

12 Inner wall
0.8 Outer wall 0.8
Outer wall
Nuo
0.6 φi 0.6 φi
Nu 8
0.4 0.4
Nuo
4 0.2 0.2

0.0 0.0
0 0 60 120 180 240 300 360
0 60 120 180 240 300 360 θ
θ
Fig. 12. Dimensionless temperature distribution along inner wall and Local Nu
Fig. 10. Local Nu distribution along inner and outer walls in case of UWT at distribution along outer wall in case of UHF at Ra ¼ 5
104, Mr ¼ 2.25, Ari ¼ 0.5 and
Ra ¼ 5
104, Mr ¼ 2.25, Ari ¼ 0.5 and at l ¼ 45. l ¼ 45.
 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899 1895

Table 3
Steady mean Nusselt number Nu at different values of Ra, Ari , l and Mr in case of UWT.

Ra l Nu

Ari ¼ 0.25

Mr ¼ 1.2 (4.724)a Mr ¼ 1.5 (2.671)a Mr ¼ 2.0 (1.823)a Mr ¼ 2.5 (1.491)a Mr ¼ 3.0 (1.292)a
104 0 4.724 2.688 2.361 2.573 2.685
25 4.724 2.750 2.431 2.723 2.892
75 4.724 2.771 2.710 2.975 3.053
90 4.724 2.777 2.721 2.998 3.259

105 0 4.724 3.912 4.175 4.251 4.481

25 4.724 3.939 4.461 4.656 4.716
75 4.724 4.345 5.065 5.321 5.409
90 4.724 4.381 5.127 5.352 5.452

Ari ¼ 0.75
Mr ¼ 1.2 (8.717)a Mr ¼ 1.5 (4.114)a Mr ¼ 2.0 (2.513)a Mr ¼ 2.5 (1.943)a Mr ¼ 3.0 (1.641)a
4
10 0 8.717 4.115 2.798 3.125 3.415
25 8.717 4.116 2.818 3.153 3.452
75 8.717 4.119 2.898 3.245 3.553
90 8.717 4.120 2.904 3.259 3.562

105 0 8.718 4.472 5.247 5.612 5.732

25 8.719 4.512 5.261 5.672 5.806
75 8.721 4.667 5.461 5.893 5.968
90 8.725 4.680 5.478 5.915 6.042
a
Nucond .

The effect of controlling parameters on steady flow and heat inner tube as Ari increases. The increase of the surface area for the
transfer characteristics in the enclosure under different heating same Mr decreases the space between the two tubes which
conditions is discussed in the following sections. Section 5.1 decreases the resistance to heat conduction and so increases the
discusses the case of heating the inner wall at UWT, while section heat transfer rate and so Nu. The effect of Mr (for given Ra, l and Ari)
5.2 discusses the case of heating at UHF. on Nu can be detected by close inspection of Table 3. For narrow
enclosure, Mr < 1.5 and wide enclosure, Mr > 2.5 the Nu assumes
5.1. Heating at uniform wall temperature, UWT higher values while, otherwise it assumes smaller values. Such an
observation gives possibility of having value for Mr at which Nu is
The effect of Rayleigh number, Ra enclosure orientation, l and
enclosure geometry (in terms of Ari and Mr) on the steady mean
Nusselt number at the inner wall, Nu are presented in Table 3. It can
be seen that the effect of Ra on Nu is clear and as expected, that is as
Ra increases the Nu increases. The increase of Nu with Ra becomes
more significant in case of wider enclosure (or as Mr increases)
while in case of narrow enclosure (small Mr) the effect of Ra on Nu
is insignificant. This is because the convection currents in case of
small Mr are very weak and the conduction mode is dominant
regardless the value of Ra. This can be easily observed in case of
Mr ¼ 1.2 where Ra shows negligible effect on Nu which is almost
equal to Nucond . The value of Nucond is the value of Nu obtained due
to pure conduction (when Ra is set to zero) and is only dependent
on the enclosure geometry (Ari and Mr).
At certain value of Ra and enclosure geometry, Table 3 also
aλ =0 b λ = 25
shows that as the angle of inclination l increases the Nu increases.
This effect is similar to that concluded for the case of natural
convection from elliptic tube placed in infinite medium (see Badr
[32] and D’Alessio et al. [33]). As the projected area of inner tube
normal to the buoyancy driven vertical flow decreases the resis-
tance to the flow decreases, resulting in an increase in the velocity
of flow adjacent to the inner wall. This leads to a decrease in the
thickness of both velocity and thermal boundary layers and so
causes high rate of heat transfer (and so high Nu). It can be also
observed that the effect of l on Nu is more effective for wide
enclosure while its effect becomes insignificant in case of narrow
enclosure.
Table 3 further shows the effect of enclosure geometry in terms
of inner tube axis ratio Ari and major axes ratio of the two tubes Mr
c λ = 75 d λ = 90
on the Nu. It can be inferred that as the Ari increases the Nu Fig. 13. Streamlines (left) and isotherms (right) patterns in case of UWT at
increases, this is simply due to the increase of surface area of the Ra ¼ 5
104, Mr ¼ 2.5, Ari ¼ 0.75 and at different orientations.
1896 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899
minimum. The detailed calculation with small increment of Mr
assured the existence of Mr value at which Nu is minimum. Such
a conclusion has been reported and explained in the work of
Mahfouz and Badr [27].
The steady flow field at different orientations for the case of
Ra ¼ 5
104, Mr ¼ 2.5, Ari ¼ 0.75 is shown in Fig. 13. The figure
shows that in case of symmetrical enclosure l ¼ 0 and 90 the zero
stream function line coincides with the inner and outer walls and
passes through the vertical line of symmetry, resulting no global
flow circulation around the inner wall. While in cases of asymmetry
l ¼ 25 and 75, the global flow circulation around the inner wall can
be observed. In this particular case the figure shows that the global
flow circulates with the right side eddy in clock wise direction for
asymmetric cases as can be observed from the dashed line in the
figure.
The effect of the major axis ratio, Mr on the steady streamlines
and isotherms is shown in Fig. 14 for the case of Ra ¼ 5
104,
Ari ¼ 0.75 and l ¼ 30. The figure shows that the streamlines for
these asymmetrical cases show the global flow circulation around
Mr=1.5
Mr=2.0
Mr=2.5
Mr=3.0
Fig. 14. Streamlines (left) and isotherms (right) patterns in case of UWT at
Ra ¼ 5
104, Ari ¼ 0.75 and at different Mr.
the inner wall. For Mr ¼ 1.5 and Mr ¼ 2 the global flow is following
the left side eddy and rotating around the inner wall in counter
clock wise direction and the correspondence thermal plume is no
longer vertical but rather inclined to the left side. As Mr increases to
Mr ¼ 2.5 and further to Mr ¼ 3 the global flow changes its direction
and follows the right side eddy that rotates around the inner wall in
clock wise direction together with the thermal plume that also
inclined to the right side. The effect of Mr on intensity of convection
currents and isotherms can be also observed from the figure. At
Mr ¼ 1.5 (narrow gap enclosure) the flow currents are weak and the
conduction mode of heat transfer dominates as can be inferred
from the isotherms which appear as almost concentric ellipses. For
Mr ¼ 2 the enclosure gap becomes a bit wider and the fluid currents
become more noticeable with clear signs of developing of thermal
plume at the top part of the enclosure. As Mr increases to Mr ¼ 2.5
and further to Mr ¼ 3, the convection currents get much faster,
causing vigorous flow motion and more developed thermal plume.
The intensive fluid motion within the enclosure especially near
the walls causes high temperature gradient at the walls and leads to
high local heat flux and so high local Nu. Fig. 15 shows the effect of
Mr on local Nu at outer wall. It can be observed that as Mr increases
the value of maximum Nu at the two walls increases. Also it can be
seen as Mr increases, the angle qat which maximum Nu occurs
along the outer wall decreases. The value of q at Mr ¼ 1.5 is 62 and
decreases to 61.2, 50.8 and 50.2 as Mr increases to 2.0, 2.5 and 3,
respectively. The decrease of q means the movement of the point of
maximum Nuo in the clockwise direction, indicating more incli-
nation of thermal plume in the clockwise direction as Mr increases.
5.2. Heating at uniform heat flux, UHF
The results for the case of UHF are presented in Table 4 and
Figs. 16e19. Table 4 shows the effect of the controlling parameters
on the steady mean dimensionless inner wall temperature fmi . It
can be seen from the table that at a fixed value of Ari, Mr and l as
the modified Ra increases the fmi decreases. It should be kept in
mind that decreasing of fmi with the increase of modified Ra does
not necessarily mean that the mean wall temperature itself
decreases, it only means that the dimensionless mean temperature
fmi ¼ ðTmi  To Þ=DTref decreases. From the definition of modified
Ra, for a given fluid and geometry, higher value of modified Ra
20
Ra=5x104
Ari = 0.75 , λ=30
16
Mr=3.0
Mr=2.5
12 Mr=2.0
Nuo Mr=1.5
4
0
0 60 120 180 240 300 360
θ
Fig. 15. Local Nu distribution along outer wall in case of UWT at Ra ¼ 5
104,
Ari ¼ 0.75 and l ¼ 30 and at different Mr.
 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899 1897

Table 4
Steady mean temperature, fmi at inner wall in case of UHF.

Ram l fmi
Ari ¼ 0.25 Ari ¼ 0.75

Mr ¼ 1.5 Mr ¼ 2.0 Mr ¼ 2.5 Mr ¼ 3.0 Mr ¼ 1.5 Mr ¼ 2.0 Mr ¼ 2.5 Mr ¼ 3.0

(0.555)a (0.796)a (0.965)a (1.098)a (0.431)a (0.706)a (0.911)a (1.076)a
5
103 0 0.5531 0.7113 0.6890 0.6549 0.4310 0.6834 0.6930 0.6435
30 0.5514 0.6821 0.6351 0.5959 0.4309 0.6794 0.6838 0.6315
60 0.5478 0.6423 0.5879 0.5510 0.4309 0.6721 0.6681 0.6161
90 0.5463 0.6279 0.5636 0.5371 0.4309 0.6687 0.6616 0.6095

5
104 0 0.4703 0.4555 0.4396 0.4360 0.4260 0.4554 0.4207 0.4095

30 0.4549 0.4137 0.3961 0.3936 0.4245 0.4482 0.4137 0.4023
60 0.4547 0.3799 0.3610 0.3541 0.4216 0.4361 0.4022 0.3899
90 0.4187 0.3709 0.3493 0.3414 0.4203 0.4311 0.3973 0.3849
a
fcond
mi .

means higher value of DTref and so higher value of qi and accord-
ingly and most probably higher value of Tmi . The table also shows
that at fixed modified Ra, Mr and l as Ari increases fmi decreases. Of
course as Ari increases the enclosure space decreases and accord-
ingly the resistance for conduction decreases. The decrease in
conduction resistance increases the heat transfer coefficient and
hence decreases the mean surface temperature.
Table 4 also shows the effect of angle of inclination l, on fmi . It
can be seen that as the angle l increases the fmi decreases. The
increase of l (from zero to 90) decreases the resistance to the
upward flow and so causes high velocities near the inner wall
which enhances the heat transfer coefficient and decreases the
local and mean surface temperature. It can be also observed in case
of small values of Mr the effect of angle of inclination is almost
negligible. The reason is that for such a narrow enclosure the
conduction mode of heat is dominant and the role of convection
mode is minimal for all orientations. In case of relatively ample
enclosure Mr  2 increasing of inclination allows the vertical
convection currents to move easily and intensively, enhancing the
mean heat transfer coefficients and so decreasing mean surface
temperature.
Further inspection of Table 4 shows the effect of Mr on the fmi . It
can be observed that (at fixed Ra, Ari and l) fmi takes higher values
at Mr ¼ 2 and smaller values otherwise. Such an observation gives
rise to the possibility of having a maximum fmi close to Mr ¼ 2
which is similar to the observation of having minimum Nu at
certain Mr in case of UWT. To attest such an observation a detailed
calculation is carried out with smaller increment of Mr for the case
of Ra ¼ 104, Ari ¼ 0.5 and l ¼ 0 and the results are plotted in Fig. 16.
The figure shows the existence of maximum value for fmi at
Mr ¼ 1.95. Such maximum temperature must be avoided in the
enclosure design to preclude the possibility of surface overheating.
The existence of such maximum mean temperature can be attrib-
uted to that the effect of Mr on convection heat transfer is different
from its effect on conduction heat transfer. As Mr increases the
resistance to conduction increases, leading to an increase of fmi as
can be observed from the values of fcond mi (in parenthesis) in the
table. On the other side as Mr increases the convection currents
become more active and so enhance the heat transfer coefficient on
the inner wall and decreases the wall mean temperature. This
opposite effect of Mr on fmi due to conduction and convection
results in maximum value for fmi at certain Mr.
The effect of Ari on the streamlines and isotherms is shown in
Fig. 17 for the case of Ra ¼ 5
104, Mr ¼ 2.5 and l ¼ 0. The
streamlines and isotherms are symmetrical with respect to the
vertical axis and therefore only half of every field is considered. For
this symmetrical case the streamlines plot reveals no global flow
circulation. Also it can be observed that the fluid close to the outer
a b

0.7

0.6

0.5

φmi Ra=104 Ari = 0.5

λ=0
c d
0.4

0.3

0.2
1.0 1.4 1.8 2.2 2.6 3.0
Mr
Fig. 17. Streamlines (left) and Isotherms (right) in case of UHF at Ra ¼ 5
104,
Fig. 16. Effect of Mr on the mean inner wall temperature. Mr ¼ 2.5, Ari ¼ 0.5 and at different orientations.
1898 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899
Fig. 18. Local Nu distribution along outer wall in case of UHF at Ra ¼ 5
104, Mr ¼ 2.5,
l ¼ 30 and at different Ari.
wall in the lower part of the enclosure is almost stagnant and with
slight temperature variation as can be inferred from the isotherms
plot. Also the figure shows that as Ari increases the enclosure space
gets narrower and the isotherms get closer, resulting in higher
temperature gradients within the enclosure specially in the upper
region close to the outer wall and the lower region close to the
inner wall. Such higher temperature gradients near walls cause
higher heat flux at the outer wall and lower temperature at the
inner wall. The effect of Ari on heat flux represented by local Nu at
outer wall is given in Fig. 18. The figure shows that at any point
(same q) on the outer wall as Ari increases the Nu increases.
The increase in Nu is maximum at the point of maximum local Nu
at q ¼ 0.
The steady flow and thermal fields at l ¼ 0 and l ¼ 60 for the
case of Ra ¼ 5
104, Mr ¼ 2.5, Ari ¼ 0.5 are shown in Fig. 19. The
figure shows that the effect of inclination on the streamlines and
isotherms are generally similar to those in case of UWT. That is no
λ =0
λ = 60
Fig. 19. Streamlines (right) and isotherms (left) patterns for case of UHF at
Ra ¼ 5
104, Mr ¼ 2.5, l ¼ 0 and at different Ari, a) Ari ¼ 0.1, b) Ari ¼ 0.4, c) Ari ¼ 0.6
and d) Ari ¼ 0.9.
global flow circulation is found in symmetrical case l ¼ 0 while in
cases of asymmetry l ¼ 60 the global flow circulation around the
inner wall is easily observed. In this particular case the global flow
circulates with the right side eddy in clock wise direction as can be
observed from the dashed lines in the figure. The isotherms plot
(right) confirm that the thermal plume is vertically symmetric in
case of l ¼ 0 while for asymmetrical case l ¼ 60 it inclines in the
direction of flow coming upward from the top of inner wall and
heading the outer wall. The figure shows that for this particular
case the thermal plume inclines to the right. The inclination of
thermal plume has been observed in the work of [28].
6. Conclusions
The problem of steady natural convection in an enclosure that
formed between two confocal elliptical tubes has been numerically
investigated using the Fourier Spectral Method. The enclosure is
heated through its inner wall which is maintained either at uniform
temperature, UWT or uniform heat flux, UHF. The heat convection
within the enclosure is dependent on Rayleigh number (modified
Ra in case UHF), Prandtl number, angle of inclination of minor axes
to gravity vector, ratio of outer tube major axis to inner tube major
axis and the axes ratio of the inner tube. The results are obtained for
Rayleigh numbers from 5
103 to 105. The major axes ratio is varied
up to 3 while the angle of orientation of the minor axes is varied up
to 90 . The Prandtl number is fixed at 0.7 and the inner tube axis
ratio varies from 0 to 1. The results for the special case of natural
convection in concentric circular and elliptic enclosures are first
obtained and compared with the earlier reported data and very
good agreement has been found. The results for local and mean
Nusselt numbers are presented and discussed together with the
details of both flow and thermal fields at various enclosure orien-
tations. The study has shown that as the angle of enclosure incli-
nation increases the total rate of heat transfer increases in case of
UWT while in case of UHF the increase of inclination decreases the
mean temperature of the inner wall. In case of UWT the study has
shown that for a certain value of Rayleigh number there is a value of
the major axes ratio at which the heat transfer through the
enclosure reaches its minimum. This value decreases as Rayleigh
number increases. While in case of UHF the study has shown that at
a certain modified Rayleigh number there is a value of the major
axes ratio at which the inner wall mean temperature is maximum.
Moreover, the study has confirmed the existence of global flow
circulation around the inner wall in case of asymmetrical
enclosures.
Acknowledgements
The support received from HEC and UET Taxila, Pakistan is very
highly appreciated.
References
[1] T.H. Kuehn, R.J. Goldstein, An experimental and theoretical study of natural
convection in the annulus between horizontal concentric cylinders, J. Fluid
Mech. 74 (1976) 695e719.
[2] N. H Kuehn, R.J. Goldstein, An experimental study of natural convection heat
transfer in concentric and eccentric horizontal cylindrical annuli, ASME J. Heat
Transfer 100 (1978) 635e640.
[3] U. Projahn, H. Rieger, H. Beer, Numerical analysis of laminar natural convec-
tion between concentric and eccentric cylinders, Numer. Heat Transfer 4
(1981) 131e146.
[4] C.H. Cho, K.S. Chang, K.H. Park, Numerical simulation of natural convection in
concentric and eccentric horizontal annuli, ASME, J. Heat Transfer 104 (1982)
624e630.
[5] H.M. Badr, Study of laminar free convection between two eccentric horizontal
tubes, Trans. Can. Soc. Mech. Eng. 7 (4) (1983) 191e198.
 F.M. Mahfouz / International Journal of Thermal Sciences 50 (2011) 1887e1899 1899

[6] J. Prusa, L.S. Yao, Natural convection heat transfer between eccentric hori- [28] R.Y. Sakr, N.S. Berbish, A.A. Abd-Alziz, A.S. Hanafi, Experimental and numerical
zontal cylinders, ASME J. Heat Transfer 105 (1983) 108e116. investigation of natural convection heat transfer in horizontal elliptic annuli,
[7] D. Naylor, H.M. Badr, J.D. Tarasuk, Experimental and numerical study of J. Appl. Sci. Res. 4 (2) (2008) 138e155.
natural convection between two eccentric tubes, Int. J. Heat Mass Transfer 32 [29] W.M. Collins, S.C.R. Dennis, Flow past an impulsively started circular cylinder,
(1) (1989) 171e181. J. Fluid Mech. 60 (1973) 105e127.
[8] D.W. Pepper, R.E. Cooper, Numerical solution of natural convection in [30] H. M Badr, S.C.R. Dennis, Time-dependent viscous flow past an impulsively started
eccentric annului, AIAA J. 21 (1983) 1331e1337. rotating and translating circular cylinder, J. Fluid Mech. 158 (1985) 447e488.
[9] Y.Z. Wang, H.H. Bau, Low Rayleigh number convection in horizontal, eccentric [31] F.M. Mahfouz, H.M. Badr, Flow structure in the wake of a rotationally oscil-
annuli, Phys. Fluids 31 (1988) 2467e2473. lating cylinder, ASME J. Fluids Eng. 122 (2000) 290e301.
[10] G. Guj, F. Stella, Natural convection in horizontal eccentric annuli: numerical [32] H.M. Badr, Laminar natural convection from elliptic tube with different
study, Numer. Heat Transfer 27 (1995) 89e105. orientations, ASME J. Heat Transfer 119 (1997) 709e718.
[11] C. Shu, Q. Yao, K.S. Yeo, An efficient approach to simulate natural convection [33] S.J.D. D’Alessio, L.A. Finlay, J.P. Pascal, Free convection from elliptic cylinders at
in arbitrarily eccentric annuli by vorticity-stream function formulation, small Grashof numbers, Int. J. Heat Mass Transfer 51 (2008) 1379e1392.
Numer. Heat Transfer A 38 (7) (2007) 739e756.
[12] C. Shu, Y.L. Wu, Domain-free discretization method for doubly connected
domain and its application to simulate natural convection in eccentric annuli, Nomenclature
Comput. Meth. Appl. Mech. Eng. 191 (2002) 1827e1841.
[13] E. Van de Sande, B.J.G. Hamer, Steady and transient natural convection in ai : length of semi-major axis of inner tube
enclosures between horizontal circular cylinders (constant heat flux), Int. J. ao : length of semi-major axis of outer tube
Heat Mass Transfer 22 (3) (1979) 361e370. Ari: axis ratio of inner tube
[14] E.K. Glakpe, C.B. Watkins, J.N. Cannon Jr., Constant heat flux solutions for Aro: axis ratio of outer tube
natural convection between concentric and eccentric horizontal cylinders, fn, Fn: Fourier coefficients
Numer. Heat Transfer 10 (3) (1986) 279e295. g: gravitational acceleration
[15] A. Castrejon, D.B. Spalding, An experimental and theoretical study of transient gn, Gn: Fourier coefficients
free-convection flow between horizontal concentric cylinders, Int. J. Heat h; h: local and average heat transfer coefficients
Mass Transfer 31 (2) (1988) 273e284. hn, Hn: Fourier coefficients
[16] R. Kumar, Study of natural convection in horizontal annuli, Int. J. Heat Mass J: Jacobian of coordinate transformation
Transfer 31 (6) (1988) 1137e1148. k: thermal conductivity
[17] J.-S. Yoo, Dual free-convective flows in a horizontal annulus with a constant Mr: major axes ratio ¼ ð ao =ai Þ
heat flux wall, Int J. Heat and Mass Transfer 46 (2003) 2499e2503. Nu; Nu: local and mean Nusselt numbers
[18] J.-S. Yoo, Flow pattern transition of natural convection in a horizontal annulus Pr: Prandtl number ð ¼ n=aÞ
with constant heat flux on the inner wall, Int. J. Numer. Meth. Heat Fluid Flow Ra: Rayleigh number ð ¼ gbð2ai Þ3 ðDTref =anÞÞ
15 (7) (2005) 698e709. t: dimensionless time
[19] M.M. Elshamy, M.N. Ozisik, J.P. Coulter, Correlation for laminar natural T: temperature
convection between confocal horizontal elliptical cylinders, Numer. Heat x0 ; y0 : Cartesian coordinates
Transfer A 18 (1990) 95e112. x, y: dimensionless Cartesian coordinates
[20] J.H. Lee, T.S Lee, Natural convection in the annuli between horizontal confocal
elliptic cylinders, Int. J. Heat Transfer 24 (1995) 1739e1742. Greek symbols
[21] C.H. Cheng, C.C. Chao, Numerical prediction of the buoyancy driven flow in the a: thermal diffusivity
annulus between horizontal eccentric elliptical cylinders, Numer. Heat b: coefficient of thermal expansion
Transfer A 30 (1996) 283e303. DTref : reference temperature difference ð ¼ Ti  To Þ in case of UWT and ð ¼ ai qi =kÞ in
[22] Y.D. Zhu, C. Shu, J. Qiu, J. Tani, Numerical simulation of natural convection case of UHF qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

between two elliptical cylinders using DQ method, Int. J. Heat Transfer 47 3i : inner ellipse dimensionless eccentricity 1  Ar2i
¼
(2004) 797e808.  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
[23] G.D. Raithby, K.G.T. Hollands, Laminar and turbulent free convection from 3o : outer ellipse dimensionless eccentricity ¼ 1  Ar2o
elliptic cylinders with a vertical plate and horizontal circular cylinder as f: dimensionless temperature ¼ T  To =DTref
special cases, ASME J. Heat Transfer 98 (1976) 72e80. h ; x: elliptical coordinates
[24] J.H. Merkin, Free convection boundary layers on cylinders of elliptic cross l: annulus major axes angle of inclination to the horizontal
section, ASME J. Heat Transfer 99 (1977) 453e457. n: kinematics viscosity
[25] F.M. Mahfouz, S. Kocabiyik, Transient numerical simulation of buoyancy r: density
driven flow adjacent to an elliptic tube, Int. J. Heat Fluid Flow 24 (2003) s: time
864e873. j0 ; j: dimensional and dimensionless stream functions
[26] W.C. Schreiber, S.N. Singh, Natural convection between confocal horizontal z0 ; z: dimensional and dimensionless vorticity
elliptical cylinders, Int. J. Heat Mass Transfer 28 (1985) 807e822.
[27] F.M. Mahfouz, H.M. Badr, Heat convection between two confocal elliptic tubes, Subscripts
placed at different orientations, AAMM, Adv. Appl. Math. Mech. 1 (5) (2009) i: at the inner tube surface
639e663. o: at the outer tube surface