International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

3D conjugate heat transfer with thermal radiation in a hollow

cube exposed to external flow
C. Albanakis, D. Bouris *
Department of Engineering and Management of Energy Resources, University of Western Macedonia, Bakola and Sialvera, 50100 Kozani, Greece

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

Article history: The interaction of an asymmetrically heated cube envelope exposed to turbulent external flow is numer-
Received 23 July 2007 ically studied using a three dimensional computational fluid dynamics control volume approach with
Available online 24 May 2008 conjugate heat transfer, including thermal radiation effects. An analytical approach is used for thermal
radiation modelling with an implicit boundary condition for convective and radiative heat transfer at
Keywords: solid–fluid interfaces. For external flow Reynolds numbers of 2  105–106 and heat flux values of
3D RANS q = 20, 40 W m2, a weak but turbulent (Ra = 1.5–5  109) buoyant flow is induced inside the cube. The
Conjugate heat transfer
temperature distribution of the inner surfaces is significantly affected by heat flux orientation with
Thermal radiation
Cube envelope
regard to external flow as well as the relative influence of convective and thermal radiation heat transfer.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction dent of the relative placement of the two cubes. Turbulent flow
past a surface mounted cube has been a challenging problem for
Natural convection is found in a number of applications such as numerical studies due to the complex flow pattern that includes
electronics, aeronautics, building engineering, solar energy collec- recirculation regions, stagnation points and unsteady vortex shed-
tors, aerospace systems, etc, where forced convection and thermal ding. Cheng et al. [5] compared large eddy simulations to standard
radiation might also be present. Buoyant flows in indoor spaces Reynolds averaged modelling using the k–e model and Tsuchiya
have been receiving much attention recently and understanding et al. [6] investigated the applicability of modifications to the
the mechanisms that link them with external wind flow is valuable standard k–e model for bluff body flows where excessive turbu-
information for natural or hybrid ventilation applications [1]. An lence energy production is usually predicted in the stagnation re-
interesting situation arises when a building is subject to external gions. Iaccarino et al. [7] and Rodi [8] attribute the superior
airflow while simultaneously exposed to solar radiation since it predictions of LES for this type of flow to correct modelling of
combines a number of heat transfer modes: natural and forced both turbulence energy production and vortex shedding. Unfortu-
convection, thermal radiation and conduction through the enve- nately, since the computational cost of this type of simulation is
lope. The purpose of the present study is to present a numerical high and even unsteady RANS are overwhelming for parametric
methodology that is being developed to study the fundamental studies [7], some of the inaccuracies of RANS models must still
3D configuration, for all involved heat transfer modes. be tolerated [8] although knowledge of the shortcomings is
Turbulent flow and heat transfer past a surface mounted cube important for extracting dependable conclusions.
has been the subject of both experimental and numerical studies. In a differentially heated cubic enclosure, natural convection
Meinders et al. [2] and Nakamura et al. [3] performed experimen- dominates and this has also been a subject of interest for many
tal measurements of the local and mean Nusselt number distribu- years. Markatos et al. [9] numerically studied the buoyant flow of
tion on the surfaces of a heated cube exposed to turbulent smoke in a square enclosure and the numerical modelling of the
external flow and the latter developed correlations for the exter- buoyancy forces in gases is still being refined [10]. A common con-
nal surface heat transfer as a function of Reynolds number. The in- figuration for natural convection in enclosures is when one side of
side of the cube was solid and at steady heating conditions. the enclosure (usually a cube) is heated and the opposite side is
Meinders and Hanjalic [4] extended these studies to heat transfer cooled. The laminar flow experiment by Bajorek and Lloyd [11],
from two surface mounted cubes, finding that the distribution of which involved a 2D enclosure with obstacles (baffles) extending
the heat transfer coefficient depends on the vortex shedding of from the top and bottom walls, has been used as a validation test
the upstream cube but that the average value remains indepen- case for both enclosure natural convection and conjugate heat
transfer [12]. Tian and Karayiannis [13] experimentally studied a
* Corresponding author. Tel.: +30 24610 56675. 2D enclosure with differentially heated side walls, but for a moder-
E-mail address: [email protected] (D. Bouris). ately turbulent flow (Ra  109), without obstacles. Dol and Hanjalic

0017-9310/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2008.01.038
6158 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

Nomenclature

A surface area, m2 To reference temperature for buoyancy force calculation, K

Cp specific heat, J kg1 K1 u velocity, m s1
Cl, C1, C2, C3 k–e model constants x, y, z Cartesian directions
d, L, H geometric dimensions of configurations, m
dn normal distance from solid surface, m Greek symbols
e emissivity a thermal diffusivity, m2 s1
Ft interpolation coefficient b thermal expansion coefficient, K1
Fij view factor between surface i and j e turbulence energy dissipation rate, m2 s3
g acceleration of gravity, m s2 DT temperature difference, K
Gr Grasshoff number, bgH3(Th  Tc)/m2 g, n secondary Cartesian directions in Fig. 1
h convective heat transfer coefficient, W m2 K1 l, lt, leff dynamic, turbulent and effective viscosity, kg m1 s1
hri,j linearised radiative heat transfer coefficient for surfaces m kinematic viscosity, m2 s1
i and j, W m2 K1 q density, kg m3
kt thermal conductivity, W m1 K1 r Stefan–Boltzmann constant (5.67  108 W m2 K4)
k turbulence kinetic energy, m2 s2 rT turbulent Prandtl number
Nu convective Nusselt number, hL/kt
P pressure, Pa Subscripts
Pk, Gk turbulence kinetic energy production terms, kg m1 s3 i = 1, 2, 3 Cartesian notation corresponding to x, y, z directions,
q heat flux, W m2 respectively
qj,i radiative heat flux between surfaces i and j, W m2 sf solid–fluid interface
qr radiative heat flux term, W m2 s solid
Ra Rayleigh number, gb(Th  Tc)H3/ma f fluid
Re Reynolds number, U0H/m c cold wall
ST source term in energy equation, W m3 h hot wall
T temperature, K o upstream flow

[14] performed numerical studies of turbulent natural convection attention. Bouris [21] and Petridou and Bouris [22] have previously
in a side-heated enclosure with a number of turbulence models performed studies of the flow and heat transfer past a hollow sur-
and found that the low-Reynolds number k–e model variants fail face mounted cube, examining both the effects of turbulence mod-
to represent details of the flow but are capable of providing quali- elling assumptions as well as the variation of the thermal fields
tative information. Radiation effects in these configurations can be- through the cube envelope. However, radiation effects were not in-
come important for large temperature differences, or when cluded in the study and thermal sources were uniformly distributed
convection is not prominent and Kasemi and Naraghi [15] per- inside the cube. In the present study, a numerical methodology is
formed 2D numerical studies of the laminar flow in a differentially presented that allows for the simultaneous solution of forced and
side heated square enclosure at low values of gravitational acceler- buoyant turbulent flow with conjugate convective, conductive
ation. The effects of radiation were modelled using the discrete ex- and radiative heat transfer through solid–fluid interfaces. Special
change factor method and became more important as convection attention is placed on the thermal boundary condition at the so-
was reduced at low-g. Mezrhab et al. [16] used the radiosity ap- lid–fluid interface so that it is implicitly included in the solution
proach to include radiation effects in their study of 2D laminar nat- rather than through a separate iterative procedure.
ural convection in a cavity with a square body at its center and
more recently Dharma Rao et al. [17] used an ADI solver to perform 2. Numerical methodology
a detailed conjugate heat transfer analysis of 2D conduction, natu-
ral convection and radiative heat transfer in a horizontal fin array 2.1. Flow field
considered as a two-fin enclosure. Sharma et al. [18] studied turbu-
lent flow in a differentially heated square enclosure with thin The numerical method is based on the Reynolds averaged Na-
walls, for which conjugate heat transfer of convection and conduc- vier–Stokes equations (RANS) with the mass, momentum (veloci-
tion was considered at the interface but the walls themselves were ties: u1, u2, u3, pressure: P) and enthalpy (Cp: specific heat, T:
modelled as one dimensional fins. The presence of three dimen- temperature) equations written in Cartesian notation (i = 1, 2, 3
sional solid objects was considered by Consalvi et al. [19] who pre- correspond to x, y, z directions, respectively):
sented a methodology for solving turbulent flow and heat transfer
in enclosures under conditions of fire. Sharma et al. [18] and Cons- o
ðqui Þ ¼ 0 ð1Þ
alvi et al. [19], used the k–e model for turbulence effects, the radi- oxi
  
osity method for radiation and the harmonic mean of the solid and o oP o oui ouj
ðquj ui Þ ¼  þ leff þ þ qbg i ðT  T o Þ ð2Þ
fluid thermal conductivities was used at the interface. However, oxj oxi oxj oxj oxi
  
Chen and Han [20], who focused on modelling conjugate heat o o l C P oT
transfer with SIMPLE based algorithms, found that the application ðquj C P TÞ ¼ kt þ t þ ST ð3Þ
oxj oxj rT oxj
of the harmonic mean across the interface might lead to erroneous
solutions. where kt is the thermal conductivity, q the density, l the dynamic
There are a number of complex physical phenomena present in viscosity, b the thermal expansion coefficient and gi is the accelera-
both the flow past a surface mounted cube and in the buoyant flow tion of gravity (g1 = g3 = 0, g2 = 9.81 m s2) with To the reference
inside a differentially heated cubic enclosure that have been ad- temperature for buoyancy force calculation. This system of equa-
dressed separately in the past, as shown in the preceding discus- tions applies to the whole domain whether fluid or solid (velocities
sion, but the combined configuration has not received much become zero), dominated by natural or forced convection (the
 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168 6159

buoyancy term becomes negligible in areas of no temperature dif- through openings in the enclosure. The reflected radiation depends
ference) and it is solved on a Cartesian grid with a collocated vari- on the emissivity of the surfaces but for the most common sur-
able arrangement using the SIMPLE pressure correlation algorithm faces found in building enclosures the emissivity is high (>0.8),
[23] with corrections for the cell face velocities [24]. For calculation leading to a limited number of reflections. Based on this observa-
of the turbulent viscosity (lt): tion, an analytic solution can be derived, that considers reflected
2 radiation between surface pairs, including reflections from a third
k
leff ¼ l þ lt ¼ l þ C l ð4Þ participating surface i.e. direct and indirect radiation resulting
e from one intermediate reflection. The heat flux being absorbed
the transport equations for turbulent kinetic energy (k) and turbu- at surface 2 that originates at surface 1 or results from infinite
lent energy dissipation (e) are: reflections between the two surfaces or any other single interme-
  diate surface (n) is given as a geometrical progression:
o o ok
ðquj kÞ ¼ leff þ P k þ Gk  qe ð5Þ
oxj oxj oxj qðnÞ1!2 ¼ e2 ð1  en Þe1 rA1 T 41 F 1!n F n!2 þ e2 ð1  e1 Þð1  e2 Þ
 
o o oe e  ð1  en Þ2 e1 rA1 T 41 F 21!n F 2n!2 F 2!1 þ    1
ðquj eÞ ¼ l þ ½C 1 ðPk þ C 3 Gk Þ  C 2 e ð6Þ
oxj oxj eff oxj k
  e2 ð1  en Þe1 rA1 T 41 F 1!n F n!2
oui ouj oui l oT ¼ ð9Þ
P k ¼ lt þ ; Gk ¼  t g i b ð7Þ 1  ð1  e1 Þð1  e2 Þð1  en ÞF 1!n F n!2 F 2!1
oxj oxi oxj rT oxi
so the flux arriving at 2 from 1, through any other surface, would be
The constants in (3)–(7) are Cl = 0.09, C1 = 1.44, C2 = 1.92, rT = 1.0, calculated as the sum of (9) for all surfaces n = 1, N (except 1 and 2).
rk = 1.0, re = 1.3. Test calculations using C 3 ¼ tanhðu2 =ðu21 þ u23 Þ0:5 Þ, The total radiation exchange between surface (i = 1) and surface
as used by Sharma et al. [18], did not show any significant differ- (j = 2) can be linearised in the form
ences compared to the suggestion of Markatos et al. [9] that it is qj;i ¼ q2;1 ¼ hr2;1  A1  ðT 2  T 1 Þ ð10Þ
acceptable to take the buoyancy constant C3 to be zero for enclo-
sures – possibly because of the low (only moderately turbulent) where hr2,1 is the linearised radiative heat transfer coefficient that
Rayleigh number of the flows considered. The k–e turbulence model can be calculated from the infinite geometrical progression of
was used [25–27] but it is important to mention that although the three-surface interactions (9) and is [30]:
best results for turbulent heat transfer with the k–e model are ob- hr2;1 ¼ ðT 22 þ T 21 Þ  ðT 2 þ T 1 Þ
tained with its low-Reynolds number variants [28], these are very 8
< e1  e2  F 1!2
demanding in terms of grid density near the walls and convergence  h i þ e1  e2  r  A2
stability so the standard wall function approach has been used here. : 1  ð1  e Þ  ð1  e Þ  F 2 A =A
1 2 1!2 1 2
Buoyancy forces were calculated based on density differences aris- )
X
N
ð1  e1 Þ  F 1!i F 2!i
ing from temperature variations with respect to a reference temper- 
ature-density value. This is based on the commonly used i¼3
Ai  ½1  ð1  e1 Þ  ð1  e2 Þ  ð1  ei Þ  F 1!i F i!2 F 2!1 
Boussinesq approximation that has recently been re-examined by ð11Þ
Sparrow and Abraham [10], who concluded that, as long as the fluid
In (11), the view factor reciprocity relation F2?1 = F1?2A1/A2 has
properties and reference density are considered at the mean sur-
been taken into account and summation refers to all of the remain-
rounding temperature, it is acceptably accurate compared to the
ing surfaces (other than the surface pair being considered) involved
computationally expensive alternative of directly inserting the
in radiation exchange.
P Netradiation for any surface (i) is calculated
buoyancy term. All terms are discretised using second order differ-
from the sum j¼1;N1 qj;i where q1,2 = q2,1 and hr1,2 = hr2,1A1/A2,
ences while second order upwind differencing of the convective
so that for all surface pairs being considered, only half of the heat
terms was implemented using the fully bounded second order
transfer rates (q) need to be calculated. The term in brackets in
BSOU scheme [29].
(11) involves computationally expensive summation for all surfaces
but, since it is only geometric and hence temperature independent,
2.2. Radiative heat transfer
it can be calculated once at the beginning of the calculation proce-
dure and then the expression for hrj,i is easily updated depending on
The last term in (3) is a heat source term that can be added to
the local surface temperatures before being inserted in the source
the temperature equation to represent the effects of heat sources,
terms in (3). Calculation of the view factors (Fij) in (11) is facilitated
such as thermal radiation. First, a finite number of surfaces (N)
by the use of the Cartesian grid and can be based on the analytical
are chosen as a radiative heat transfer system. In the present study,
expressions given by Howell [31] (see Fig. 1 for notation):
these are the solid surface control volume cell faces that surround
the enclosure and are determined by the grid resolution. Calcula- 1 X2 X 2 X 2 X 2

tion of the net radiative heat fluxes between these surfaces could F 12 ¼ ð1ÞðiþjþkþlÞ Gðxi ; yj ; gk ; nl Þ
ðx2  x1 Þ  ðy2  y1 Þ l¼1 k¼1 j¼1 i¼1
be calculated using the commonly applied radiosity approach,
which however requires a matrix inversion that may become ð12Þ
excessively demanding for a large number of surfaces. Here, ther- where G is given for parallel surfaces
mal radiation between solid surfaces separated by a non-partici- 8 2 3
pating medium (air) is modelled according to an analytical >
< qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 6 yg 7
approach proposed by Clarke [30] that involves simpler calcula- G¼  ðy  gÞ  ðx  nÞ2 þ z2  tan1 4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5
2p > :
tions. The emissive power (qi) from any surface of area (Ai), emis- ðy  gÞ2 þ z2
sivity (ei) and temperature (Ti) is 2 3
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
6 xn 7
qi ¼ ei  r  Ai  T 4i ð8Þ þ ðx  nÞ  ðy  gÞ2 þ z2  tan1 4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5
ðy  gÞ2 þ z2
with (r) the Stefan–Boltzman constant. This will be distributed 9
2
>
=
among all other surfaces in an enclosure, depending on the view z
  ln½ðx  nÞ2 þ ðy  gÞ2 þ z2  ð13Þ
factor between surface pairs (Fij), and will be absorbed or reflected 2 >
;
by them, possibly returning to the original surface or escaping
6160 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

a b

Fig. 1. Notation for calculation of view factors in (12)–(14): (a) parallel surfaces; (b) perpendicular surfaces.

and perpendicular surfaces: condition, an implicit procedure is introduced. At any solid–fluid

( " # interface, the heat balance requires that:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi    
1 ðy  gÞ oT oT
G¼  ðy  gÞ  x2 þ n2  tan1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A  kt ¼ A  kt þ qr ð15Þ
2p x2 þ n2 on solid on fluid
)
1 where qr is the radiation term calculated from the contribution of
  ½x2 þ n2  ðy  gÞ2   ln½x2 þ n2 þ ðy  gÞ2  ð14Þ all surfaces, according to (11). An assumption is made that the
4
interface temperature (Tsf) will be a linear function of the solid
and fluid temperatures neighbouring the interface:
As already mentioned, (11) does not take into account radiation that
might have reached the surface after more than one reflection (i.e. T sf ¼ F t  T s þ ð1  F t Þ  T f ð16Þ
four or more participating surfaces). For situations with low emis- It will subsequently be shown that it is in fact not linear but the ini-
sivity surfaces, this could lead to errors in the calculated radiative tial hypothesis facilitates analysis, without loss of generality. Sub-
heat transfer so a numerical check is performed at the end of the stitution of (16) into (15) leads to
procedure in order to verify that the total radiative energy leaving  
any surface is equal to that given by (8). Deviations were insignifi- 1 kts qr
Ft ¼   ð17Þ
cant in the present study (<105%) but, when present, they can be kts =dnsf—s þ ktf =dnf—sf dnsf—s AðT s  T f Þ
overcome by distributing the remaining energy among all the sur- and (16) is rewritten as:
faces, on a view factor weighted basis. A similar procedure is per-  
formed to verify that the summation of view factors for any kts
T þ dnktf T f A  qr
dnsf—s s f—sf
surface equals unity, deviations (here, <107%) to which must be T sf ¼ ð18Þ
ðkts =dnsf—s þ ktf =dnf—sf ÞA
attributed only to numerical errors since there are no assumptions
in the methodology. where (kts, ktf) are the solid and fluid thermal conductivities respec-
The view factors can also be exploited to implement shading be- tively, (A) is the surface area of the control volume cell faces that
tween any two surfaces (A) and (B) using the following procedure: neighbour the interface and (dnf–sf, dnsf–s) are the normal distances
(a) the view factor is calculated according to (12)–(14) but only if between the control volume cell center and the interface on the
the angles between the surface direction vectors ð~ nA ; ~
nB Þ and the fluid and solid side respectively. Notice that the interpolation
line segment joining the surface centers ðABÞ are both <90°. If parameter (Ft) is a function of the temperatures neighbouring the
not, then the surfaces are not facing each other and the view factor interface and the radiation heat flux. (18) is used for calculation
is set to zero; (b) the surface pair (A, B) is examined with relation to of the surface temperatures appearing in (10) and (11) while for
the remaining surfaces in the system (i = 3, N) and an obstructing the discretisation of the enthalpy equation radiation is included as
(shading) surface is determined when only one of FAi, FBi is non- a source term for the solid side control volume and the following,
zero and the intercept point between the surface (i) and the line equivalent but numerically more stable, condition is applied:
segment ðABÞ is between the two surfaces and within the limits    
of surface (i). Finding these co-ordinates is based on basic 3D vec- T0  Ts T f  T 0sf kts =dnsf—s  T s þ ktf =dnf—sf  T f
kt sf ¼ kt ; T 0sf ¼
tor analysis. If an obstructing surface is found then FAB = 0 but (11) dn s dn f kts =dnsf—s þ ktf =dnf—sf
must still be used since there will be indirect radiation exchange ð19Þ
between the surfaces.
For inclusion of radiation effects in the calculation and consid- where T 0sf is the surface temperature that would arise if (15) were
eration of heat flow through solid–fluid interfaces, a conjugate heat applied without radiation effects i.e. the purely conductive boundary
transfer boundary condition must be applied at all solid–fluid condition appearing in conjugate heat transfer without radiation
interfaces and the enthalpy equation must be solved in both fluid [32]. In fact, this approach is very similar to that of Li and Durbetaki
and solid domains, the only difference being in the velocities (zero [33] and Consalvi et al. [19], except they both used a harmonic mean
for solid domain) and the diffusion coefficients appearing in (3). value for the thermal conductivity instead of calculating the surface
Here, for implementation of the conjugate heat transfer boundary temperature as is done here. Using the harmonic mean of (kt/Cp) on
 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168 6161

the fluid and solid side is a potential source of error in many steady (d/L = 0.1, H/L = 0.25) protruding from the middle of the top and
state applications according to Chen and Han [20] and the present bottom walls. The side walls were maintained at a constant tem-
methodology avoids this problem. perature difference of 15.5 K, leading to a Rayleigh number of
The procedure of including conjugate heat transfer with radia- Ra = 3.5  105. Zimmerman and Acharya [12] studied this configu-
tion effects can be summarised as: ration numerically and found an important influence of the per-
fectly conducting wall boundary condition, as opposed to the
(a) calculation of the radiation term (qr) depending on the wall more commonly used adiabatic one. In the present study, the
surface (not cell center) temperatures that appear in (10). configuration was extended by 3L in the third direction and discre-
These are calculated through (18) based on values from tised with a 50  58  21 Cartesian grid. As expected, no 3D effects
the previous iteration. were observed. There were three grid cells defining the 0.1L thick
(b) (qr) is inserted as an energy equation source term (ST) in (3) top and bottom walls, which were assumed to have the thermal
for the control volume on the solid side of the interface. conductivity of the experimentally used plexiglass (kt = 0.665
(c) implementation of (19) is achieved by using (T 0sf ) for calculat- W m1 K1). The same material was considered for the two baffles
ing the normal diffusion terms of both solid and fluid control of the enclosure. The calculated flow and temperature field is pre-
volumes neighbouring the interface. Wall functions for tur- sented in Fig. 2 with conjugate heat transfer considered for the baf-
bulent boundary layers can be included in the usual manner, fles and the top and bottom walls. The favourable comparison with
as presented for non-radiative conjugate heat transfer in [32]. the experimental measurements of Bajorek and Lloyd [11] and the
numerical predictions of Zimmerman and Acharya [12] is evident
Updating of the radiation heat fluxes is performed at every iter- in Fig. 3. Inclusion of radiation effects (not shown) did not lead
ation of the SIMPLE solution procedure and interface temperatures to any discernible differences, also in agreement with the findings
are updated accordingly. The implementation has proven to be ro- of Zimmerman and Acharya [12], most probably because of the
bust although some underrelaxation for the enthalpy equation was relatively small temperature differences and the constant tem-
found beneficial. In the present approach, a single grid spans fluid perature walls.
and solid domains and the conjugate heat transfer boundary condi- Radiation effects were examined by comparison with the
tion, including radiative effects, is implicitly applied in the solution numerical calculations of Kasemi and Naraghi [15] who used the
of the system of equations. This is facilitated by the assumption of discrete exchange factor method to calculate radiative heat transfer
(16) and only the temperatures and heat flux required for the (Ft) in a square (H  H) enclosure for normal and low levels of gravita-
coefficients are lagged in the iterative process. tional acceleration. The advantage of this type of validation is that
radiation effects and natural convection effects can be considered
3. Results separately. A 22  28  15 cartesian grid was used, with the depth
direction exhibiting no 3D effects. As before, the top and bottom
The numerical procedure was first validated through the simu- walls were discretised using three grid cells but assuming negligi-
lation of relevant configurations in the published literature and ble thermal conductivity in order to simulate adiabatic boundary
then applied to parametrically study the situation of a hollow cube conditions. Results are presented in Fig. 4 for H = 0.021 m, a Grass-
exposed to external fluid flow and a constant heat flux on one of its hoff number of Gr = 700, and cold–hot wall temperature ratio of
external surfaces. Tc/Th = 0.5. Radiation effects are initially omitted, leading to an
For the flow inside the enclosure, validation was first per- almost conductive, linear, behavior, where the weak natural con-
formed for low-Rayleigh number flows in side heated enclosures, vection plays a negligible role. With radiation present, natural con-
with and without radiation and internal obstacles, and then for a vection remains at low levels but the temperature distribution
moderately high Rayleigh number flow. For the effect of the exter- along the hot and cold walls is significantly altered. Agreement
nal flow on the cube envelope, the most common approach is to
set a constant heat transfer coefficient on the external walls (e.g.
[18]). Here, the turbulent flow past a surface mounted cube is cal-
culated leading to local values of convective heat transfer on the
external walls. For calculation of turbulent flow past a surface
mounted cube, eddy viscosity models, such as the k–e model ap-
plied here, admittedly lead to inaccuracies that are difficult to
overcome but although variants to the model have been suggested
their effectiveness is still questionable [8,5,22]. Large eddy simula-
tions and even unsteady RANS have proven to yield the best re-
sults when compared to experiments but the computational cost
is much higher. To the authors’ knowledge, similar analysis of
the heat transfer from a surface mounted cube has not been pub-
lished but a previous attempt to examine it [22] indicated that
similar conclusions should be drawn. Since the purpose here is
to provide a more accurate estimation of the external wall heat
transfer rather than applying a uniform constant value, and due
to the excessive computational requirements for a better ap-
proach, the standard k–e model was used but the accuracy of the
calculation will be kept in mind and its effect on the results and
conclusions will be discussed.

3.1. Isolated cubical enclosure

Bajorek and Lloyd [11] measured the flow and temperature field Fig. 2. Flow field vectors (m/s) and temperature field contours (K) for buoyant flow
in a two dimensional square enclosure (L  L) with two baffles in an enclosure with two baffles (Ra = 3.5  105).
6162 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

a 0.5 250 b 1.0

Present Numerical: T*
0.4 Bajorek&Lloyd (1982): T* 200 0.9
Zimmerman & Acharya (1991): V/(v/L)
0.3 Present Numerical: V/(v/L) 150 0.8
0.2 100 0.7
0.1 50 0.6

V/ (v/L)
T*

Y/L
0.0 0 0.5
-0.1 -50 0.4 Zimmerman & Acharya (1991)
Present Numerical
-0.2 -100 0.3
-0.3 -150 0.2
-0.4 -200 0.1
-0.5 -250 0.0
0.0 0.2 0.4 0.6 0.8 1.0 -150 -100 -50 0 50 100 150
X/L U/ (v/L)

Fig. 3. Comparison with experimental and numerical results: (a) for the non-dimensional form T* = [T  (Th + Tc)/2]/(Th  Tc) and V/(m/L) along the centreline between the hot
and cold wall; (b) U/(m/L) vertically, midway between the hot and cold wall.

1.0

0.9

0.8
T/Th

0.7

Present, Rad, Top

Kasemi&Naraghi (1993), Rad,Top
Present, Rad, Bottom
0.6 Kasemi&Naraghi (1993), Rad,Bottom
Present, No Rad,Top
Kasemi&Naraghi (1993), No Rad, Top
Present, No Rad, Bottom
Kasemi&Naraghi (1993), No Rad, Bottom
Fig. 5. Flow field vectors (m s1) and temperature field contours (K) for buoyant
0.5
flow in an enclosure (Ra = 1.58  109).
0.0 0.2 0.4 0.6 0.8 1.0
X/H
623 grids so the representation of the general features of the flow
Fig. 4. Temperature distribution between hot and cold walls along top and bottom
walls, with and without radiation effects. is considered acceptable for the purpose of the study. Sharma
et al. [18] reached similar conclusions for grids ranging from
42  42 to 82  82 for higher Rayleigh numbers (Ra  1010) even
with the results of Kasemi and Naraghi [15] is considered accept- though they did not apply any type of wall boundary approxima-
able since differences are below 5% of (Th  Tc). tion such as wall functions or a low-Reynolds number variant of
Turbulent flow in a square (L  L) enclosure with hot (Th) and the k–e model.
cold (Tc) opposite vertical walls has been examined by comparison
with the results of Tian and Karayiannis [13]. The dimensions of 3.2. Hollow cube subject to external heating and flow
their experimental configuration have been retained in all direc-
tions (2L in the third direction) and the calculated flow field is pre- The chosen configuration is that of a hollow cube exposed to an
sented in Fig. 5. This is a moderately turbulent Rayleigh number external flow of cool air and simultaneously subjected to a heat
flow (Ra = 1.58  109), turbulence onset being at roughly 108, but flux on one of its external sides. The enclosure dimensions are
still poses a challenging numerical exercise. Results with a 323 grid, 3  3  3 m with 0.2 m thick walls and kt = 0.14 W m1 K1,
clustered near the walls, are presented in Figs. 6 and 7 for temper- q = 2300 kg m3, Cp = 650 J kg1 K1, e = 0.9. This gives emissivity
ature and velocity fields, respectively. Agreement is generally and a thermal permeability factor (0.7 W m2 K1) corresponding
acceptable and calculations with a 623 grid did not show signifi- to characteristic building materials. The computational domain ex-
cant improvements. According to Tian and Karayiannis [13], the tends 4H above the cube, 6H in the spanwise direction and 12H in
mean Nusselt number along the wall should be Nu = 64 and the the flow direction with the cube 3.5H from the inlet boundary. The
present calculations gave a 15% difference for both the 323 and upstream flow velocity was uniform and the temperature constant
 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168 6163

a 50
b
45

40
1.0
35

30 0.8
T (°C)

(T-Tc)/(Th-Tc)
25 0.6
Present, y/L=0.2
20
Tian&Karayiannis (2000): y/L=0.2
Present, y/L=0.4
0.4
15 Tian&Karayiannis (2000): y/L=0.4 Present, hotwall
Present, y/L=0.6 Present, coldwall
Tian&Karayiannis (2000): y/L=0.6 0.2
10 Tian&Karayiannis (2000), hotwall
Present, y/L=0.8
Tian&Karayiannis (2000), coldwall
Tian&Karayiannis (2000): y/L=0.8
5 0.0
0 0.2 0.4 0.6 0.8 1 0.00 0.05 0.10 0.15
X*=1-x/L X*=dwall/L

Fig. 6. (a) Temperature profiles between hot and cold walls at various heights of the enclosure. (b) Temperature boundary layer near the hot and cold wall at y/L = 0.5.

2250
Present, y/L=0.9, v*=(v+2)*1000

1750 Tian&Karayiannis (2000): y/L=0.9

Present, y/L=0.7, v*=(v+1.5)*1000

Tian&Karayiannis (2000): y/L=0.7

1250
v* (m/s)

Present, y/L=0.5, v*=(v+1)*1000

Tian&Karayiannis (2000): y/L=0.5

750
Present, y/L=0.3, v*=(v+0.5)*1000

Tian&Karayiannis (2000): y/L=0.3

250
Present, y/L=0.1, v*=(v+0)*1000

Tian&Karayiannis (2000): y/L=0.1

-2
50
0 0.2 0.4 0.6 0.8 1
X*=1-x/L

Fig. 7. Vertical velocity profiles between the hot and cold walls at various heights of the enclosure.

at To = 293 K, with the ground at the same constant temperature. was the same for all calculations: 62  50  76, with 30 points on
Symmetry boundary conditions were applied in the spanwise the cube edges but calculations were also performed with an
and vertical direction and fully developed flow at the outlet. For 82  60  91 grid and 39 points on the cube edges for the down-
the cube’s solid walls, conjugate heat transfer boundary conditions stream and top heat flux directions. Variations in the internal and
are applied and the standard wall function approach is used for the external face mean temperatures were between 3% and 12% of
momentum equations next to all solid boundaries. the maximum temperature difference.
Calculation of the flow past the hollow cube has been performed The Nusselt number is defined:
for three different external flow Reynolds numbers (Re = U0H/m =
qH ðdT=dnÞ  H
2  105, 5  105 and 1  106), two different heat flux values Nu ¼ ¼ ð20Þ
kt  ðT sf  T o Þ ðT sf  T o Þ
(q = 20 and 40 W m2) and four different outer face heat flux orien-
tations (top, upstream, downstream and one of the side faces). The where (dT/dn) is the temperature gradient normal to the surface, Tsf
external heat flux values are rather low to correspond to solar heat is the surface temperature and To is the upstream temperature. For
gains, but the present calculation is performed at steady state so the cube mean Nusselt number, the mean value of the five exposed
actual values would lead to unrealistic inner temperatures. The faces is calculated. Surface cooling of a heated cube exposed to
Reynolds numbers correspond to common wind velocity values of external flow has been studied in the past and external surface
U0 = 1, 2.5 and 5 m s1, respectively. The effect of the external and cube mean Nusselt number correlations all agree to the form
Reynolds number and heat flux value is examined for the upstream Nu = c  Ren with n < 1 (see e.g. [2,3]).
and downstream heat flux orientations and the full effect of heat Calculations were performed for a q = 20 W m2 heat flux, ini-
flux orientation as well as the relative influence of convective and tially on the upstream cube face and then on the downstream
radiative heat transfer is examined at Re = 2  105 and q = one, for Reynolds numbers in the range of Re = 2  105–1  106
20 W m2. During calculations, it was found beneficial for conver- and the cube mean Nusselt number and air temperature are pre-
gence to calculate external flow while maintaining internal cube sented in Fig. 8. With increasing Reynolds number, the Nusselt
temperature constant and to subsequently introduce internal number increases according to an n = 0.60 exponent (close to the
natural convection flow and then radiation effects. The grid size n = 0.68 value found by Nakamura et al. [3]) and thus leads to an
6164 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

Temperature (upstream heat source) heat flux value (q = 20 W m2) at the same Reynolds number as
Temperature (downstream heat source) 1200
Nusselt (upstream heat source)
well as a higher Reynolds number (Re = 1  106) are also shown.

Nusselt Number (Nu)

Air Temperature (K)

295
Nusselt (downstream heat source)
1000 The higher heat flux value leads to higher inner temperatures
and the higher Reynolds number induces a stronger cooling effect

Mean Cube
Mean Cube

800 but the relative spatial temperature distribution remains the same:
when the upstream cube face is heated, the inner surface temper-
294 600 atures and the cube mean temperature are higher. The tempera-
ture difference between the hottest and coldest face is between
400 0.6 and 2 K for the cases presented in Fig. 9 but all Rayleigh number
values are in the range Ra = 1.5–5  109 i.e. above the turbulent
293 200
threshold (Ra > 108).
0 2 50000 500000 750000 1000000
The relative ‘‘asymmetry” between upstream and downstream
ExternalFlow Reynolds Number(Re)
heating is further demonstrated through Figs. 10 and 11. Fig. 10a
Fig. 8. Effect of flow Reynolds number on mean cube air temperature and mean shows the temperature distribution on the external surfaces of
cube Nusselt number. the cube when the upstream face is heated with a heat flux of
q = 20 W m2 and the external flow Reynolds number is Re =
2  105. Streaklines indicate the presence of the upstream stagna-
asymptotic decrease in cube mean temperature. The opposite tion zone and horseshoe vortex, as well as the recirculation zones
effect can be observed if, for comparison purposes, the heat flux on the top, side and downstream faces. The ‘‘hot spot” at the front
is doubled to q = 40 W m2 at an external Reynolds number of stagnation region is prominent, as are the cooler edges, but the hot
Re = 2  105. The resulting distribution of internal cube face tem- fluid captured within the top and side recirculation zones, com-
peratures is shown in Fig. 9 for the two cases of upstream and bined with conduction through the envelope, leads to a tempera-
downstream orientation of external heat flux. Results for the lower ture increase on the rest of the surfaces as well. Fig. 10b shows
the corresponding buoyancy induced inner flow: the warmer up-
stream face causes a recirculation zone of air that rises, follows
Re=2E5, Q=20 W/m2, Upstream Side the ceiling and declines along the back but turns upward again
2 298
Re=2E5, Q=20 W/m2, Downstream midway along the bottom face. This is explained by the radiatively
2 Upstream
Re=1E6, Q=20 W/m2,
297
heated area near the front of the bottom face which lifts the flow
2
Re=1E6, Q=20 W/m2, Downstream
2 Upstream
before it reaches the upstream face. A radiatively heated region
Re=2E5, Q=40 W/m2,
2 Downstream 296 is also evident on the outer surface of the downstream face, which
Re=2E5, Q=40 W/m2,
does not appear when radiation effects are not included. At the
295 same Reynolds number (Re = 2  105) but at double the heat flux
Downstream Top
value (q = 40 W m2) the surface temperatures and the tempera-
294
ture differences become larger leading to a doubling of Rayleigh
293 number (Ra = 4.5  109) and the recirculation zone is now pushed
Side
towards the top and downstream face (Fig. 11a). The asymmetry of
Downstream
the inner and outer flow and heat transfer for upstream and down-
Top
stream heating is clear in Fig. 11 for the higher heat flux value. The
Upstream
Bottom cube’s inner surface temperature range is roughly the same when
upstream (Fig. 11a) and downstream faces (Fig. 11b) are heated
but the variation in temperature difference induces a different in-
Bottom Upstream
ner flow, possibly due to the slightly higher inner surface temper-
ature differences for downstream heating (see also Fig. 9):
Fig. 9. Inner cube face temperatures for heat flux on upstream and downstream Ra = 5  109 as opposed to Ra = 4.5  109 for upstream heating.
orientated external faces. On the outer surfaces of the cube envelope, an iso-temperature

Fig. 10. Temperature and flow field (a) around and (b) inside cube envelope when heat flux is on upstream side (Re = 2  105, q = 20 W m2).
 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168 6165

Fig. 11. Temperature and flow field in and around the cube envelope when heat flux is on (a) upstream and (b) downstream side. Streaklines have been projected onto the
center-plane and the outer temperature iso-surface value is T = 293.1 K (Re = 2  105, q = 40 W m2).

surface has been plotted at the value of T = 293.1 K. These zones are cussed. On the other hand, this introduces a uniformity in the heat
indicative of the heat dissipated by the flow and the extensive transfer among the cube surfaces that actually makes the present
warm air zones in Fig. 11b show the heat leaving the cube. In discussion conservative in its points and conclusions.
Fig. 11a, these regions are thinner, implying that more heat is being Calculation of the inner flow, induced by the non-uniform cool-
transferred to the other cube faces, leading to higher surface ing of the envelope, has already been shown for upstream and
temperatures. downstream cooling. The situation with the heat flux on the top
The relative variation of mean temperature values and Nusselt face is presented in Fig. 12 where the effects of convective and
numbers on the cube’s external faces are presented in Table 1 radiative heat transfer on the inside of the cube are evident.
and Table 2 for all possible heat flux orientations for Re = Fig. 12a shows the inner temperature distribution when the air is
2  105 and q = 20 W m2. The temperatures of the cube faces considered stagnant (i.e. only conductive heat transfer) while in
show a minimal variation of DT = (Tface  To) while the heated face Fig. 12b air motion induces convective effects and leads to a cooler
is 3–4 K hotter than the unheated faces for all cases. One would ceiling but hotter side walls. The full effect of conduction, convec-
have expected a stronger dependence on orientation but this is tion and radiation is presented in Fig. 12c where the induced flow
actually more evident in the convective Nusselt numbers (Table (Vmax = 0.03 m s1) and the higher temperatures on the floor and
2). Comparison with the experimental correlation of Nakamura sidewalls are evident. The differences in the sidewall temperatures
et al. [3] is also shown and should be compared with the heated that are induced by the effect of the outer flow lead to a slight
face for each situation. Upstream and downstream values are in asymmetry with the center of the recirculation zone being pushed
relatively good agreement with the correlations and, although towards the downstream face.
the top and side values are appropriately predicted higher than The relative influence of convection and radiation on the inner
the upstream and downstream ones, they are underestimated by temperature field is further demonstrated in Fig. 13 where the ra-
about 30%, due to the turbulence modelling difficulties already dis- dial graphs have been plotted for the four different heating orien-
tations and results are shown for stagnant air, pure convection and
convection–radiation. Temperature differences of over 4 K arise
when convection and radiation are taken into account instead of
Table 1
Mean temperatures on the outer faces of the cube, depending on the direction of the stagnant air, especially on the heated face. Even more interesting
heat flux (Re = 2  105, q = 20 W m2) is the fact that thermal radiation leads to a more uniform internal
temperature distribution among the surfaces whereas pure con-
Temperatures (K) External heat source direction
vection is not as effective in this regard. The heated surface re-
Upstream Downstream Top Side
mains at a much higher temperature without radiation effects,
Side 293.48 293.46 293.39 296.85 which tend to cool it and simultaneously heat the others, leading
Top 293.44 293.43 296.73 293.40
to the uniformity in temperature. This effect is a direct conse-
Upstream 296.64 293.33 293.41 293.42
Downstream 293.35 297.03 293.43 293.41 quence of keeping the heat flux constant and letting conduction,
convection and radiation determine the wall’s temperature. In

Table 2
Mean Nusselt numbers on the outer faces of the cube, depending on the direction of the heat flux (Re = 2  105, q = 20 W m2)

Nu External heat source direction Correlations from [3]

Upstream Downstream Top Side
Side 368.41 331.35 358.24 390.44 Nus ¼ 0:12Re0:70 ¼ 598
Top 373.50 355.69 411.90 355.41 Nut ¼ 0:071Re0:74 ¼ 575
Upstream 380.61 377.40 380.02 379.74 Nuf ¼ 0:71Re0:52 ¼ 396
Downstream 309.58 332.46 296.13 309.19 Nur ¼ 0:11Re0:67 ¼ 380
6166 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

Fig. 12. Temperature distribution inside the cubicle considering (a) stagnant air without radiation effects, (b) natural convection without radiation effects, (c) natural
convection including radiation effects (Re = 2  105, q = 20 W m2).

Side
Downstream heat source 301 Upstream Heat Source Side
No internal flow 301
No internal flow
299 Convective internal flow
Convective internal flow 299
Convective_Radiative internal flow
Convective_Radiative internal flow 297
297
Downstream Top Downstream Top
295 295

293 293

Bottom Upstream Bottom Upstream

Side
Side Heat Source Side Top Heat Source 301
301 No internal flow
No internal flow
Convective internal flow Convective internal flow 299
299
Convective_Radiative internal flow Convective_Radiative internal flow
297
297
Downstream Top
Downstream Top 295
295

293 293

Bottom Upstream Bottom Upstream

Fig. 13. Variation of the mean temperatures of the inner surfaces of the cube, depending on the effects of convection and radiation for: (a) downstream heat source;
(b) upstream heat source; (c) side heat source; (d) top heat source (Re = 2  105, q = 20 W m2).

situations where studies focus on constant temperature walls, such depending on the heat source orientation with regard to the exter-
an effect cannot be observed. nal flow direction. The relative effects of upstream and down-
The interaction of the outer flow and the orientation of the stream heating have already been discussed with regard to
heated surface and their effect on the inner flow and temperature Fig. 11 and similar conclusions are drawn here, for the lower heat
distributions can be seen more clearly in Fig. 14. For each of the flux value. The highest temperatures, both the mean cube air and
four heat source orientations that were studied, the five surfaces’ the inner surfaces, arise when the heat source is on the upstream
mean temperatures are plotted on the corresponding axis. The side while heating on the top leads to the lowest inner tempera-
mean cube air temperature is also plotted on the axis that corre- tures. When the upstream face is heated, although the surface Nus-
sponds to the surface being heated. Here it can be seen that the in- selt number is low (see Table 2), the dissipated heat is transferred
ner temperature differences are not negligible (over 0.5 K) and that to the downstream surfaces, leading to their higher surface tem-
there is also a significant difference in the mean cube temperature, peratures when compared to the other heating orientations (Table
 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168 6167

Side cube and leads to higher inner temperatures than for other heat
Top heat source 295.5
source orientations.
Upstream heat source
Down stream heat source 295.0
Side heat source Acknowledgements
Cube mean air temperature 294.5

The research is co-funded by the European Union – European

294.0
Social Fund & Natural Resources – EPEAEK II and the Greek Minis-
Downstream Top
try of Education and Religious Affairs under the Pithagoras II
293.5
program.
Side 293.0
Downstream
Top References
Upstream
Bottom [1] P. Linden, The fluid mechanics of natural ventilation, Annu. Rev. Fluid Mech. 31
(1999) 201–238.
[2] E. Meinders, K. Hanjalic, R. Martinuzzi, Experimental study of the local
convective heat transfer from a wall mounted cube in turbulent channel flow,
J. Heat Transfer 121 (1999) 564–573.
[3] H. Nakamura, T. Igarashi, T. Tsutsui, Local heat transfer around a wall-mounted
Bottom Upstream cube in the turbulent boundary layer, Int. J. Heat Mass Transfer 44 (2001)
3385–3395.
Fig. 14. Mean temperatures of the inner surfaces and the cube for the different heat [4] E. Meinders, K. Hanjalic, Experimental study of the convective heat transfer
source directions that were examined. The mean cube temperature for each heating from in-line and staggered configurations of two wall-mounted cubes, Int. J.
orientation is placed on the radial axis that corresponds to the heated surface Heat Mass Transfer 45 (2002) 465–482.
(Re = 2  105, q = 20 W m2). [5] Y. Cheng, F.S. Lien, E. Yee, R. Sinclair, A comparison of large Eddy simulations
with a standard k–e Reynolds-averaged Navier–Stokes model for the
prediction of a fully developed turbulent flow over a matrix of cubes, J.
Wind Eng. Ind. Aerodynam. 91 (2003) 1301–1328.
1). On the other hand, the higher Nusselt numbers on the heated [6] M. Tsuchiya, S. Murakami, A. Mochida, K. Kondo, Y. Ishida, Development of a
top and side surfaces, which are already underpredicted (see Table new k–e model for flow and pressure fields around bluff body, J. Wind Eng. Ind.
Aerodynam. 67 and 68 (1997) 169–182.
2), explain the lower temperatures inside the cube, especially since [7] G. Iaccarino, A. Ooi, P.A. Durbin, M. Behnia, Reynolds averaged simulation of
the heat is carried away by the flow. The large recirculation zone unsteady separated flow, Int. J. Heat Fluid Flow 24 (2003) 147–156.
behind the cube leads to a low Nusselt number on the downstream [8] W. Rodi, Comparison of LES and RANS calculations of the flow around bluff
bodies, J. Wind Eng. Ind. Aerodynam. 69–71 (1997) 55–75.
heated face and the higher temperatures on the outer surface (Ta- [9] N.C. Markatos, M.R. Malin, G. Cox, Mathematical modelling of buoyancy-
ble 1) and the cube inner air temperature. induced smoke flow in enclosures, Int. J. Heat Mass Transfer 25 (1) (1982) 63–
75.
[10] E.M. Sparrow, J.P. Abraham, A new buoyancy model replacing the standard
4. Conclusions pseudo-density difference for internal natural convection in gases, Int. J. Heat
Mass Transfer 46 (2003) 3583–3591.
A study of combined fluid flow and heat transfer including [11] S. Bajorek, J. Lloyd, Experimental investigation of natural convection in
partitioned enclosures, J. Heat Transfer 104 (1982) 527–532.
forced and natural convection, thermal radiation and conduction [12] E. Zimmerman, S. Acharya, Free convection heat transfer in a partially divided
through solid–fluid interfaces has been performed for the heat vertical enclosure with conducting end walls, Int. J. Heat Mass Transfer 30 (2)
transfer through a cubic envelope subjected to external flow. The (1987) 319–331.
[13] Y. Tian, T. Karayiannis, Low turbulence natural convection in an air filled
numerical methodology that has been implemented includes all square cavity. Part I: the thermal and fluid flow fields, Int. J. Heat Mass Transfer
heat transfer modes in an implicit solution procedure through a 43 (2000) 849–866.
novel boundary condition at the solid–fluid interface for convec- [14] H.S. Dol, K. Hanjalic, Computational study of turbulent natural convection in a
side heated near cubic enclosure at a high Rayleigh number, Int. J. Heat Mass
tion and thermal radiation effects. Transfer 44 (2001) 2323–2344.
The methodology was first validated by comparison to experi- [15] M. Kasemi, M. Naraghi, Analysis of radiation–natural convection interactions
mental and numerical results from the literature for laminar and in 1-g and low-g environments using the discrete exchange factor method, Int
J. Heat Mass Transfer 36 (17) (1993) 4141–4149.
turbulent natural convection in enclosures, with and without radi-
[16] A. Mezrhab, H. Bouali, H. Amaoui, M. Bouzidi, Computation of combined
ation and conjugate heat transfer effects. There are some simplifi- natural-convection and radiation heat-transfer in a cavity having a square
cations applied, especially with regard to the turbulence modelling body at its center, Appl. Energ. 83 (2006) 1004–1023.
[17] V. Dharma Rao, S.V. Naidu, B. Govinda Rao, K.V. Sharma, Heat transfer from a
of the moderately turbulent (Ra  109) buoyant flow in the cavity
horizontal fin array by natural convection and radiation – a conjugate analysis,
and the heat transfer on the outer cube surfaces. However, valida- Int. J. Heat Mass Transfer 49 (2006) 3379–3391.
tion calculations show that the methodology can be used to derive [18] A.K. Sharma, K. Velusamy, C. Balaji, S.P. Venkateshan, Conjugate turbulent
qualitative conclusions with regard to a configuration that, to the natural convection with surface radiation in air filled rectangular enclosures,
Int J. Heat Mass Transfer 50 (2007) 625–639.
authors’ knowledge, has not been previously studied. [19] J.L. Consalvi, B. Porterie, J.C. Loraud, Method for computing the interaction of
It has been shown that radiation effects between the inner sur- fire environment and internal solid regions, Numer. Heat Transfer, Part A 43
faces of a cubic enclosure can play a significant role in determining (2003) 777–805.
[20] X. Chen, P. Han, A note on the solution of conjugate heat transfer problems
the temperature distribution when a constant heat flux is the heat using SIMPLE-like algorithms, Int. J. Heat Fluid Flow 21 (2000) 463–467.
source. An increase in the Reynolds number of the external flow af- [21] D. Bouris, A 3D calculation of the flow and heat transfer in and around a hollow
fects the cube Nusselt number and leads to an asymptotically cubicle exposed to external flow, in: 1st International Conference ‘‘From
Scientific Computing to Computational Engineering”, 1st IC-SCCE, Athens,
decreasing mean cube temperature. Different heat flux values sim- 8–10 September, 2004.
ilarly increase the mean temperatures although both cooling due [22] M. Petridou, D. Bouris, Experimental and numerical study of the effect of
to higher Nusselt numbers and heating due to higher heat flux val- openings on the surface pressure distribution of a hollow cube, WSEAS Trans.
Fluid Mech., 1790-5087 1 (6) (2006) 655.
ues seem to retain the inner surfaces’ relative temperature distri-
[23] S. Patankar, D. Spalding, A calculation procedure for heat mass and momentum
bution. The orientation of the heat source with regard to the transfer in three dimensional parabolic flows, Int. J. Heat Mass Transfer 15
external flow is also important. Due to the outer flow and buoy- (1972) 1787.
[24] C. Rhie, W. Chow, Numerical study of the turbulent flow past an airfoil with
ancy forces, a heat flux applied from different orientations will lead
trailing edge separation, AIAA J. 21 (1983) 1525–1532.
to different inner flow and temperature patterns. Heating on the [25] B.E. Launder, D.B. Spalding, The numerical computation of turbulent flows,
upstream surface causes much of the heat to remain within the Comput. Methods Appl. Mech. Eng. 3 (1974) 537.
6168 C. Albanakis, D. Bouris / International Journal of Heat and Mass Transfer 51 (2008) 6157–6168

[26] D.B. Spalding, Contribution to the theory of heat transfer across a [30] J.A. Clarke, Energy Simulation in Building Design, second ed., Butterworth,
turbulent boundary layer, Int. J. Heat Mass Transfer 7 (3) (1964) 743– Heinemann, 2001.
761. [31] J.R. Howell, A Catalog of Radiation Heat Transfer Configuration Factors,
[27] C.L.V. Jayatilaka, The influence of Prandtl number and surface roughness on the University of Texas, Austin second ed., 2001, Available online at: <http://
resistance of laminar sublayers to momentum and heat transfer, Prog. Heat www.me.utexas.edu/>.
Mass Transfer 1 (1969) 193–329. [32] D. Bouris, G. Bergeles, Numerical calculation of the effect of deposit formation
[28] H.B. Awbi, Calculation of convective heat transfer coefficients of room surfaces on heat exchanger efficiency, Int. J. Heat Mass Transfer 40 (17) (1997) 4073–
for natural convection, Energ. Build. 28 (1998) 219–227. 4084.
[29] G. Papadakis, G. Bergeles, A locally modified second order upwind scheme for [33] X. Li, P. Durbetaki, The conjugate formulation of a radiation induced transient
convection terms discretisation, Int. J. Numer. Methods Heat Fluid Flow 5 natural convection boundary layer, Int. J. Numer. Methods Eng. 35 (1992) 853–
(1995) 49–62. 870.