2nd International Conference “From Scientific Computing to Computational Engineering”

2nd IC-SCCE
Athens, 5-8 July, 2006
© IC-SCCE

LAMINAR FREE CONVECTION IN VERTICAL AND INCLINED AIR CHANNELS

1 2 3 1
Evangelos C. Bacharoudis , Andronicos E. Filios , Michalis G. Vrachopoulos and Dionisios P. Margaris
1
Fluid Mechanics Laboratory, Mechanical Engineering and Aeronautics Department,
University of Patras, GR-26500, Patras, Greece
e-mail: margaris@mech.upatras.gr
2
Fluid Mechanics and Turbomachinery Laboratory
School of Pedagogical and Technological Education, 141 21 N. Heraclion, Greece
e-mail: aefilios@gmail.com
3
Department of Mechanical Engineering,
Technological Education Institute of Chalkis , 34400 Psachna Evias, Greece
e-mail: mvrachop@teihal.gr

Keywords: Inclined Channel, Asymmetric Heating, Natural Convection, CFD.

Abstract. The thermal control in many systems is widely accomplished applying natural convection process due
to its low cost, reliability and easy maintenance. Typical applications include the heat exchangers, cooling of
electronic equipment and nuclear reactors, solar chimneys and Trombe walls in building industry, etc.
Numerical and experimental studies of natural convective flows in two dimensional channels, opened to ambient
conditions at both end sections, are extensively reported in the literature but most of them are treated severely
idealised situations. Numerical studies of natural convective in vertical and inclined in a wide angle range air
channels and for different aspect ratio are presented in the present work. The 2D, laminar simulations are
obtained by solving the governing equations using a commercial CFD code. It is considered that the
temperature variations are not so high and the Boussinesq approximation is applied. The latter leads to the
simplification of the system of equations. The computed results for Nusselt number, velocity and temperature
profiles and mass flow rate are directly compared with those proposed in the bibliography letting therefore the
validation of the employed numerical procedure.

1 INTRODUCTION
Natural convection between parallel plates has been studied by many authors for different applications,
mostly for channels which are short and the driving temperature is not high. These applications are used in
electronic equipment for ventilation purposes [3, 4]. Channels with larger dimensions are frequently encountered
in other applications such as solar chimneys, heat exchangers etc. In such applications the channel may have
different forms. Sparrow and Azevedo [2] investigated the natural convection in open-ended vertical channels
for different aspect ratios both experimentally and computationally. They also took account the conduction in
one of the walls. Kaiser et al [5] studied natural convection in convergent channels and Desrayaud and Fichera
[1] worked with the case of the vertical air channel with two rectangular ribs attached to the walls of the
channel. Finally, Gadafalch et al. [6] investigated natural convection in large inclined channels for different
inclination angles. In all the previous cases the flow was laminar and the channel had symmetric or asymmetric
heating conditions.
In the present work an investigation in the transport phenomena in an inclined, open-ended, air channel,
which comprises the basic part of a solar chimney application attached to buildings, with asymmetric heating is
presented. The application of solar chimneys is an effective technique for reducing the temperature inside a
building as well as for the provision of natural ventilation, which helps in lowering the humidity and achieving
comfortable conditions inside the space. Its application in buildings may provide the required ventilation while
simultaneously covers part of the heating and cooling requirements. The buoyancy-induced air flow depends on
the difference in air density between the inside and outside of the solar chimney. The numerical study of the air
channel for different aspect ratios and different inclination angles is carried out. The governing elliptic equations
are solved in a two dimensional domain using a control volume method. Predicted velocity and temperature
profiles are presented for the different cases. Important parameters such as average Nusselt number are
calculated and compared with the proposed one from bibliography.
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.

2 PROBLEM FORMULATION

2.1 The physical problem

The geometry of the air channel is shown in Figure1. It consists of a large channel where L, is the height of
the channel and b, is the spacing of the plates. It can be vertical or inclined and with different aspect ratios b/L.
The inclination of the channel referenced by the angle θ and varies from θ=0 deg (vertical channel) to θ=75 deg.
The aspect ratio, b/L, is also changed for three representative cases as concerning the thermofluid phenomena
which are formed in the channel. The three cases of b/L=0,01, 0,1 and 0,05 represent the fully developed
regime, the boundary-layer regime and an intermediate case respectively. The channel is asymmetric heated with
the left wall adiabatic. In all the cases considered heat transfer through the walls causes buoyancy-driven flow. It
is also assumed that outside the air channel the fluid is at rest.
outlet

adiabatic Tw

y
inlet 1 0
To,p To,po
x

Figure 1. Geometry of the problem

The flow inside the channel based on RαL is laminar :

gβ ( Tw1 − To )L3
RaL = Pr (1)
ν2
where β, is the volumetric thermal expansion coefficient, g is the gravitational acceleration, ν is the kinematical
viscosity, Tw1 is the right wall temperature and T0 is the inlet air temperature. Another important dimensionless
parameter which is traditionally used in problems of open-ended channels is the Rayleigh number based on the
plate spacing and the modified Rayleigh number. Particularly, Rαb

gβ ( Tw1 − T0 )b 3
Ra b = Pr (2)
ν2
and the modified Rayleigh number, Rab* :

⎛b⎞
Ra*b = Ra b ⎜ ⎟ (3)
⎝L⎠
An important characteristic of the flow is the rate of heat transfer through the air channel walls. Using
Newton’s law of cooling for the local convection coefficient h the Nusselt number for the right wall based on the
plate spacing of Figure 1 may be expressed as:

hb q ∂T
Nu br =
k
, h=
Tw1 − T0
, q=
∂x ∫ x =0
dy (4)
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.

2.2 Mathematical model

2.2.1 The governing equations

A numerical investigation of the natural buoyancy-driven fluid flow and heat transfer in the open-ended
channel has been attempted. The simulations were conducted using the commercial CFD code, Fluent® [7]. The
steady, laminar, incompressible and two-dimensional form of the conservation equations was solved for the fluid
flow in the channel using the Boussinesq approximation [10]. The corresponding set of differential equation
may be written:
∂u ∂v
+ =0 (5)
∂x ∂y

∂u ∂u ∂p ′ ⎛ ∂ 2u ∂ 2u ⎞
ρu + ρv =− + ρν ⎜⎜ 2 + 2 ⎟⎟ − ρβ ( T − T0 )g x (6)
∂x ∂y ∂x ⎝ ∂x ∂y ⎠

∂v ∂v ∂p′ ⎛ ∂2v ∂2v ⎞

ρu + ρv =− + ρν ⎜⎜ 2 + 2 ⎟⎟ − ρβ (T − T0 )g y (7)
∂x ∂y ∂y ⎝ ∂x ∂y ⎠

∂T ∂T λ ⎛ ∂ 2 T ∂ 2 T ⎞
ρu + ρv = ⎜ + ⎟ (8)
∂x ∂y c p ⎜⎝ ∂x 2 ∂y 2 ⎟
⎠
where x and y are the coordinates in the Cartesian-coordinate system indicated in Fig. 1, p΄ is the ‘reduced’ static
pressure, T is the temperature, T0 is the reference pressure, u and v are the velocity vectors expressed in the
reference system (x,y), gx and gy the gravitational acceleration vectors and the physical properties of the air ρ, ν,
β, λ and cp are respectively the density, the kinematical viscosity, the thermal expansion coefficient, the thermal
conductivity and the specific heat at constant pressure, which are assumed constant (Pr=0,713).
The non dimensional form of the governing equations and boundary conditions shows that Nub depends on
the Rayleigh number, the aspect ratio and the inclination of the two plates.
2.2.2 Boundary conditions
Boundary conditions have been specified at the inlet, outlet and walls of the air channel. At the inlet section,
it is obvious that there is a specific velocity profile because the fluid is moving with a specific mass flow rate. It
is well-known that this velocity profile is the result of the pressure difference between two points inside and
outside of the channel at the same height. Let us consider the points 0, 1 in Figure 1. The fluid is motionless at
the point 0 in ambient temperature T0=20oC and static pressure po. In point 1 the fluid obtains an unknown
velocity profile which produces mass flow rate m  at inlet temperature T0 and static pressure p. According to the
literature [9,11-14] it is assumed that the air moves from point 0 to point 1 with an adiabatic and reversible way.
Specifically the Bernoulli equation holds at the entrance region outside the channel and the pressure difference
between the two points is converted to kinetic energy, i.e.
1 2 1
p+ ρu = p o ⇒ p ′ = − ρu 2 (9)
2 2
where p΄, is the ‘reduced’ static pressure and p0=-ρ0gy, is the ambient pressure. Moreover, it is assumed that the
streamwise variations of temperature are neglected. Finally, under certain conditions, conduction effects at inlet
could be important, but for Rab above 1, it can be assumed that these are negligible [5].
At the outlet section the streamwise variations of velocity components and temperature are neglected. In
addition, it is assumed that the fluid’s pressure becomes equal to the ambient pressure [9,11-14].The static
pressure in an arbitrary point can be written as:
p = p ′ + p o ⇒ p = p ′ − ρ o gy (10)

In order to be satisfied the preceding condition p=p0 at the outlet region, we have to impose p′=0. In this way
according to Gadafalch et al [6] all the kinetic energy of the air is assumed to be converted to heat. Finally, it
was defined a backflow total temperature 20οC in case the fluid entered to the chimney from the outlet. In this
case we consider the incoming air to be fresh air in a temperature Τ0=20οC.
The right wall of the channel is isothermal and has a temperature, i.e. Tw1=40oC. The left wall is considered
adiabatic,
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.

∂T
=0 (11)
∂x x =0

Additional non-slip conditions for velocities are imposed at the channel walls.
2.2.3 Grid Independence
A structure, mapped mesh with quadrilateral 2D elements has been built in the code. In order to ensure the
accuracy of the numerical results, a detailed grid independence study was performed by changing the number of
the nodes in the horizontal (H) and in the vertical (V) direction (Table 1). The purpose was to define the best
mesh resolution to capture the boundary layer regime. In the simulations a grid with 40×240 quadrilateral cells
has been used giving grid independent solutions for all the cases considered. In Table 1, the control of the size of
the grid for the case of a vertical channel is presented.
V/H=4 Nub V/H=6 Nub
20×80 12,44 20×120 12,40
40×160 11,59 40×240 11,56
50×200 11,49 50×300 11,47
60×240 11,45 60×360 11,43
Table 1. Average Nusselt number Nub for different grids in the vertical channel with b/L=0,1

2.2.4 Numerical solution details

The solution of the set of the non-linear equations together with the boundary and internal conditions has
been made with the segregated steady-state solver embodied in Fluent® commercial software. Several iterations
of the solution loop must be performed before a converged solution is obtained. For pressure-velocity coupling
Fluent provides three methods in the segregated solver: SIMPLE, SIMPLEC, and PISO. SIMPLE has been used
in all cases studied. Because of the non-linearity of the problem the solution process is controlled via relaxation
factors that control the change of the variables as calculated at each iteration. The convergence is checked by
several criteria (e.g. the conservation equations should be balanced; the residuals of the discretised conservation
equations must steadily decrease).

3 RESULTS AND DISCUSSION

A parametric study was carried out varying the aspect ratio of the channel by changing the plate spacing b
and keeping the channel length constant and the inclination angle θ. Typical results from the performed
simulations for the various cases considered are presented and discussed in this section.
In Figure 2, the variation of the rate of the heat transfer Nub is presented as a function of the inclination angle
for three different aspect ratios. The Nusselt number Nub increases as the width b of the channel becomes larger
and diminishes as the channel deviates from the vertical position. Specifically when the inclination angle
increases the rate of heat transfer becomes smaller. The numerical results are directly compared with the results
of the equation of Gadafalch [6]. It is an empirical heat transfer relation for the Nub, based on the proposed
empirical equation of Bar-Cohen and Rohsenow. The maximum relative error is below 13% and the average
relative one below 4%. In small aspect ratios there is an excellent agreement with the proposed empirical results
but in bigger aspect ratio there is an obvious deviation.
Many researchers have proposed different empirical relations for Nusselt number Nub for vertical or inclined
or converging channels with asymmetric heating. For the general case of a vertical channel all the relations give
values for the Nub which varies slightly each other. Taking into account this fact and considering the maximum
and the average relative error of the empirical Gadafalch’s equation we can conclude that the numerical results
are in good agreement with the results of the bibliography.
In Figure 3, the computed values of Nusselt number are plotted vs aspect ratio with angle of inclination as a
parameter. It is clear that an increase in the aspect ratio of the channel causes an increase in the heat transfer rate
and an increase in the inclination angle results in a decline of the heat transfer rate respectively.
Similar to the variations of the heat transfer rate, Nub is the behaviour of the mass flow rate when the
inclination angle varies. The mass flow rate for the examined aspect ratios and inclination angles is shown in
Table 2. When the channel is rotated from the vertical position, less fluid comes out from the top of the channel
and the mass flow rate is diminished for all aspect ratios. This fact is also explained from the behaviour of the
dimensionless velocity for different inclination angles in Figure 6.
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.

14 14
Gadafalch, b/L=0,10 Νumerical, b/L=0,10
Gadafalch, b/L=0,05 Νumerical, b/L=0,05 θ= 0 deg
13 Gadafalch, b/L=0,01 Νumerical, b/L=0,01 13
θ=15 deg
12 12 θ=30 deg
θ=45 deg
11 11 θ=60 deg
θ=75 deg
10 10

9 9

Nusselt number, Nub

Nusselt number, Nub

8 8

7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0
0 10 20 30 40 50 60 70 80 0,00 0,02 0,04 0,06 0,08 0,10
Angle, θ (deg) Aspect ratio, b/L

Figure 2. Variation of Nuselt number with the Figure 3. Variation of Nusselt number with aspect ratio
inclination angle for different aspect ratios b/L and inclination angle
However the behaviour of mass flow rate is completely different from the one of Nusselt number Nub for the
case of different aspect ratios. While the aspect ratio b/L of the channel increases, the mass flow rate also
increases until b/L=0,05 for the specific modified Rayleigh numbers Rab* . For aspect ratio b/L=0,1 a decrease in
the rate takes place. In Figure 4, the dimensionless velocity profile for b/L=0,05 is more developed, has a greater
maximum value and extends farther into the channel from the hot wall in comparison with the one of b/L=0,1.
This fact results in the reduction of the velocity gradients near the hot wall and this may produce an appreciable
decrease of the viscous drag, which means that more fluid comes out from the channel for aspect ratio b/L=0,05.
So, the larger the plate spacing does not mean the more the mass flow rate.
On the other hand, in Figure 5, the slope of the function of the temperature is steeper for larger aspect ratios
which explain the fact that the heat transfer rate increases with the aspect ratio of the channel. Concluding, the
distance between the two plates is very important parameter for the heat and mass transfer of fluid through the
channel.
Finally, the inlet boundary condition comprises a significant factor for the calculation of the mass flow rate
[9]. With such inlet boundary condition as is described in 2.2.2 it is assured that the fluid is moving inside the
channel in the beginning, only by natural convection.
Angle
b/L
0o 15o 30o 45o 60o 75o
0,01 0,00266052 0,00259877 0,00241180 0,00209220 0,00162204 0,00095751
0,05 0,00632778 0,00618253 0,00619966 0,00594561 0,00575755 0,00506380
0,10 0,00452035 0,00448795 0,00436613 0,00418612 0,00392853 0,00342578
Table 2. The mass flow rate (kg/s) for different cases of aspect ratios and inclination angles.
A thorough examination of the thermofluid phenomena inside the duct is presented below. The flow in the
channel is driven by two different forces, the buoyancy forces and the pressure forces. On the other hand, the
viscous forces act against the flow direction,. All the following results are presented in dimensionless format.
Specifically the dimensionless velocity v* is given:
v* = v / v max (12)
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.
where vmax is the maximum velocity among the compared cases each time. And the dimensionless temperature
ΔT* is

T − T0
ΔT* = (13)
Tw − T0

1,0 1,0

b/L=0,01
0,9 b/L=0,05 0,9
b/L=0,10

0,8 0,8

Dimensionles temperature, (T-To)/(Tw-To)

0,7
Dimensionless v-velocity, v*

0,7

0,6 0,6

0,5 0,5

0,4 0,4

0,3 0,3

0,2 0,2
b/L=0,01
b/L=0,05
b/L=0,10
0,1 0,1

0,0 0,0
0,0 0,2 0,4 0,6 0,8 1,0
x/b
-0,1
0,0 0,2 0,4 0,6 0,8 1,0
x/b

Figure 4. Velocity profiles within the channel for Figure 5. Temperature profiles within the channel for
three different aspect ratios three different aspect ratios
In Figures 4 and 5, the dimensionless velocity and the dimensionless temperature profile at the outlet section
are displayed for different aspect ratios. It is clear the appearance of two representative flow field regimes for
b/L=0,01 the fully developed regime and for b/L=0,1 the boundary layer regime. In the first, the distance
between the two plates is small enough and the formed boundary layer near the right wall approaches the left
wall. This fact gives a fully developed velocity profile. No reverse flow occurs in this flow regime.
For aspect ratios b/L=0,05 and 0,1 the boundary layer regime prevails. The velocity and the thermal
boundary layer are created near the right wall far away from the other wall. An important remark is the reverse
flow occurring from the middle of the air channel and upwards. The appearance of this recirculation region is
easily explained. Since buoyancy pushes the flow towards the hot wall new fluid needs to be admitted through
the exit, in order to satisfy mass conservation. In Figure 4 the reverse flow at the outlet section is presented.
Further details about the formation of the velocity and temperature boundary layer across the duct’s walls are
presented in previous work for vertical channels [11].
Considering the channel with aspect ratio b/L=0,1 a parameterised investigation of the velocity and the
thermal boundary layer are presented in Figures 6 and 7. With fixed dimensions of the channel, the inclination
angle changes from θ=0 deg to θ=75 deg for six different cases. The boundary layer regime is formed for all
inclination angles. The maximum in v-velocity profile is diminished as the angle of the rotation of the air
channel becomes larger. That coincides with the previous predicted results about the Nusselt number Nub and
the mass flow rate. Reverse flow occurs in all the cases. The velocity and temperature profile were captured at
the outlet section.
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.

1,0 1,0

θ= 0 deg θ= 0 deg
0,9 0,9
θ=15 deg θ=15 deg
θ=30 deg θ=30 deg
0,8 0,8
θ=45 deg θ=45 deg

Dimensionless temperature, (T-To)/(Tw-To)

θ=60 deg θ=60 deg
0,7 0,7
Dimensionless v-velocity, v*

θ=75 deg θ=75 deg

0,6 0,6

0,5 0,5

0,4 0,4

0,3 0,3

0,2 0,2

0,1 0,1

0,0 0,0
0,6 0,7 0,8 0,9 1,0
-0,1 x/b
0,0 0,2 0,4 0,6 0,8 1,0
x/b

Figure 6. Velocity profiles within the channel spacing for the Figure 7. Temperature profiles within the
examined inclination angles channel spacing for the examined inclination
angles

4 CONCLUSIONS
In this work research focuses on the study of the thermofluid phenomena occurring inside an air channel. A
numerical investigation of the buoyancy-driven flow field and heat transfer that take place inside open-ended
ducts is performed. The governing elliptic equations are solved in a two-dimensional domain using a control
volume method. Predicted dimensionless velocity and temperature profiles together with the average Nusselt
number are presented for different aspect ratios and for different inclination angles at the outlet of the channel.
First results show that the model predicts realistically the system behaviour.

REFERENCES
[1] Desrayaud G., Fichera A. (2002), “Laminar natural convection in a vertical isothermal channel with
symmetric surface-mounted rectangular ribs”, International Journal of Heat and Fluid Flow, Vol. 23, pp.
519-529.
[2] Sparrow E.M., Azevedo L.F.A. (1985), “Vertical-channel natural convection spanning between the fully-
developed limit and the single-plate boundary-layer limit”, International Journal of Heat and Fluid Flow,
Vol. 28, pp. 1847-1857.
[3] Zamora B., Hernandez J. (2001), “Influence of upstream conduction on the thermally optimum spacing of
isothermal, natural convection-cooled vertical plates arrays”, International Communications in Heat and
Mass Transfer, Vol. 28, pp. 201-210.
[4] Zamora B., Hernandez J. (1997), “Influence of variable property effects on natural convection flows in
asymmetrically-heated vertical channels”, International Communications in Heat and Mass Transfer, Vol.
24, pp. 1153-1162.
[5] Kaiser A. S., Zamora B., Viedma A. (2004), “Correlation for Nusselt number in natural convection in
vertical convergent channels at uniform wall temperature by a numerical investigation”, International
Journal of Heat and Fluid Flow, Vol. 25, pp 671-682.
[6] Gadafalch J., Oliva A., Van der Graaf G., Albets X. (2003), “Natural convection in a large, inclined
channel with asymmetric heating and surface radiation”, ASME Journal of Heat Transfer, 2003, Vol. 125,
pp 812-820.
 Evangelos C. Bacharoudis, Andronicos E. Filios, Michalis G. Vrachopoulos and Dionisios P. Margaris.
[7] Ozoe H., Yamamoto K., Sayama H., Churchill S.W. (1974), “Natural convection in an inclined rectangular
channel heated on one side and cooled on the opposing side”, International Journal of Heat and Mass
transfer,Vol. 17, pp. 1209-1217.
[8] Ozoe H., Sayama H., Churchill S.W. (1975), “Natural convection in an inclined rectangular channel at
various aspect ratios and angles-experimental measurments”, International Journal of Heat and Mass
transfer, Vol 18, pp. 1425-1431.
[9] Marcondes F., Maliska C.R. (1999), “Treatment of the inlet boundary conditions in natural-convection
flows in open-ended channels”, Numerical Heat Transfer, Part B, Vol. 35, pp 317-345.
[10] Gray D.D., Giorgini A. (1976), “The validity of the Boussinesq approximation for liquids and gases”,
International Journal of Heat and Mass transfer, Vol. 19, pp 545-551.
[11] Bacharoudis E.C., Vrachopoulos M.Gr., Koukou M.K., Filios A.E. (2006), “Numerical investigation of the
buoyancy-induced flow field and heat transfer inside solar chimneys”, International Conference on Energy
& Environmental Systems (EE 06), IASME/WSEAS, Evia Island, Greece, May 8-10, pp.1-6.
[12] Fluent 6.1 User's Guide 2003.
[13] Holman J. P. (1976), Heat transfer, Fourth Edition, International Student Edition.
[14] Bird B. R., Stewart W. E., Lightfoot E. N. (1960), Transport phenomena, John Wiley & Sons, Inc., New
York.