Professional Documents
Culture Documents
Grupo 9
Grupo 9
a r t i c l e i n f o a b s t r a c t
Article history: Two approaches to predicting the sphere cooling process by laminar natural convection were compared
Received 9 November 2010 in terms of the accuracy of the volume averaged sphere temperature and the heat transfer rate between
Accepted 4 February 2011 the sphere and the surrounding fluid. The first approach is based on the formulation of conjugate heat
Available online 4 March 2011
transfer (heat conduction in the sphere and laminar natural convection in the fluid). The second approach
includes the lumped capacity method based on the assumption that the temperature in the sphere is spa-
Keywords: tially uniform and on the Churchill correlation function. The solution to the problem depends on the Ray-
Sphere cooling
leigh number (Ra), the Biot number (Bi), the Prandtl number (Pr), and the sphere-to-fluid thermal
Natural convection
Lumped capacitance method
diffusivity ratio (A). The lumped capacitance method gives fairly accurate results with respect to the con-
jugate heat transfer method (discrepancy in the volume averaged sphere temperature less than 5%) when
A Bi/Ra0.452 < 0.05, for Bi < 0.15 and Pr > 1.
Ó 2011 Published by Elsevier Ltd.
Nomenclature
Pr ¼ lkc Prandtl number
q0 c ks
A¼ k qs cs sphere-to-fluid thermal diffusivity ratio
0 2
D2
~ ¼ p q0 c2
p dimensionless pressure
c specific heat capacity k
q2 cbðT s;0T 0 ÞgD3
D sphere diameter Ra ¼ 0 Rayleigh number
kl
g gravitational acceleration ~t ¼ tk 2 dimensionless time
k thermal conductivity of the fluid q cD0
~
n outward normal to the sphere surface V ¼ v qk0 cD dimensionless fluid velocity
~ ~
p pressure
p0 pressure without hydrostatic pressure, see Eq. (3)
q~ ¼ qq0 dimensionless density
Q_ heat transfer rate H ¼ TTT 0
s;0 T 0
dimensionless temperature
t time
T temperature ~ ¼ Dr non-dimensional del operator
r
v
~ fluid velocity
a heat transfer coefficient Subscripts
b coefficient of thermal expansion s sphere
l viscosity 0 initial condition
q density sa surface averaged
r del operator va volume averaged
eNu ¼ 100 NuChNuNu FL
relative difference of Nusselt numbers Ch from lumped model and Churchill’s correlation function
FL
T va;FL T sa;Ch FL from FLUENT solution
eT ¼ 100 T s;0 T 0 relative temperature difference
_ _
eQ ¼ 100 Q ChQ_Q FL relative heat transfer rate difference Superscript
FL
T transpose
Bi ¼ aD=ks Biot number
Nu = aD/k Nusselt number
The integral on the right hand side of Eq. (9) denotes the heat trans- in which the fluid temperature changes from the sphere surface
fer rate (Q_ Þ from the surrounding fluid to the sphere defined by temperature Ts to the far field fluid temperature T0. The initial heat
transfer rate (Q_ ) between the sphere and the fluid is defined by the
Q_ ¼ aSs ðT 0 T sa Þ ð10Þ steady state natural convection. Physical properties of the sphere
where a is the heat transfer coefficient, Ss is the area of the sphere and the fluid are assumed to be constant.
surface, and Tsa is the averaged temperature along the sphere sur- The following two approaches have been used to solve the
face. In the sphere, this heat transfer rate can be defined as problem of sphere cooling:
DT
Q_ ¼ ks Ss ð11Þ (1) Conjugate heat transfer formulation (conduction in the
D sphere and laminar natural convection in the fluid) solved
where DT denotes the temperature difference within the sphere by the CFD code FLUENT. The FLUENT solution is obtained
needed to drive Q_ . The definition of the Biot number follows from in a spherical calculation domain with the size of 45D. The
Eqs. (10) and (11): problem is considered as an axially symmetric with respect
to the vertical midline. At the outer boundary, the slip iso-
aD DT
Bi ¼ ¼ ð12Þ thermal wall was prescribed. The computational grid is
ks T 0 T sa
refined near the sphere according to the expected solution
representing the ratio of the temperature difference within the for the greatest value of the Grashof number as described
sphere to the temperature difference between the sphere and the in [5]. Such a grid ensures a grid-independent solution. The
surrounding fluid. If the Biot number is small (i.e. the heat transfer total number of grid cells was 67,273. The segregated SIM-
resistance within the sphere is small compared to the resistance to PLE algorithm and the second order implicit time integration
the heat transfer between the sphere and its surroundings), the were used, together with a variable integration time step.
temperature difference within the sphere will be small. Thus, for The minimum non-dimensional time step was 106 (at the
small Bi numbers it is appropriate to neglect the temperature vari- beginning of the cooling process) and the maximum was
ation within the sphere (on the assumption that the uniform tem- 103 (at the last stage of the process).
perature field is Ts = Tsa), and, consequently, Eq. (9) takes the form of (2) Lumped capacitance method. It was assumed that the tem-
perature inside the sphere was uniform and the heat transfer
dT sa
qs V s cs ¼ aAs ðT 0 T sa Þ ð13Þ between the sphere and the fluid was defined by the Chur-
dt
chill correlation function with a time-varying Rayleigh num-
or, in a non-dimensional form, of ber defined according to Eq. (16).
1 dHsa
¼ Hsa ð14Þ
6A Bi d~t
Generally speaking, a is a time-varying quantity that depends on Table 1
the strength of natural convection in the fluid; therefore, the Biot Relative difference of the Nusselt numbers obtained by the FLUENT and from the
Churchill correlation function in steady state conditions for an isothermal sphere.
number changes during the cooling process. In the case of heat
transfer between an isothermal sphere and a fluid in the steady Ra Pr NuFL NuCh eNu =%
state natural convection, the Churchill correlation function [2] is 1 104 5 7.03 7.31 3.98
recommended for Pr P 0:7 and Ra 6 1011 : 1 105 5 11.08 11.44 3.25
1 106 1 16.50 16.90 2.42
0:589Ra1=4 1 106 5 18.44 18.78 1.84
Nu ¼ 2 þ ð15Þ 1 106 25 19.44 19.80 1.85
½1 þ ð0:469=PrÞ9=16 4=9 1 107 5 31.42 31.85 1.37
where the Nusselt number is defined as Nu = aD/k. The use of this
steady state correlation function for predicting the unsteady cooling
process, assuming that the time varying Rayleigh number is defined
as Table 2
2 3
Maximal relative temperature difference eT , %.
qðtÞ cb½T sa ðtÞ T 0 gD
RaðtÞ ¼ ð16Þ Bi A Ra = 10 4
Ra = 105 Ra = 106 Ra = 107
kl
0.05 5 0.92 0.51 0.53 0.39
implies that the temperature field changes so slowly that the natu- 50 3.87 1.87 1.03 0.41
500 25.8 11.7 4.99 1.88
ral convection is fully developed all the time (quasi-steady
conditions). 0.15 5 1.48 1.25 0.83 0.47
50 11.7 4.15 2.26 1.02
Eqs. (13) and (15) were solved with an initial condition of
500 43.2 26.3 12.4 4.84
Tsa(0) = Ts,0 employing the fourth order Runge–Kutta method [7].
It is clear that both the non-dimensional sphere temperature (here,
the surface averaged temperature corresponds to the volume
averaged one), and the heat transfer rate are functions of non-
dimensional time and four non-dimensional parameters Pr, Ra, A, Table 3
and Bi. Maximal heat transfer rate difference eQ =%.
4
Bi A Ra = 10 Ra = 105 Ra = 106 Ra = 107
3. Results and discussion 0.05 5 5.5 2.8 2.3 1.6
50 12.5 6.8 3.4 1.8
500 42.5 25.8 12.2 5.8
The cooling process of a resting sphere immersed in an un-
bounded fluid was considered. Initially, the sphere is at a uniform 0.15 5 9.3 4.9 3.3 2.6
50 25 14.5 7.1 3.8
temperature Ts,0 and the fluid is in steady state motion due to nat-
500 76 49 27 14.6
ural convection. Close to the sphere there is a thin boundary layer
2306 Z. Virag et al. / International Journal of Heat and Mass Transfer 54 (2011) 2303–2309
The results from the two approaches will be compared in terms parameters: Pr, Ra, A, and ks/k, while in the LCM, the Biot number
of the volume averaged sphere temperature and the heat transfer appears instead of the parameter ks/k. Since the Nusselt number
rate between the sphere and the fluid. In the first approach these can be calculated from the heat transfer rate and the parameter
two parameters are functions of non-dimensional time and four ks/k can be defined as ks/k = Nu/Bi, it is possible to choose the Biot
4 4
Ra=10 , Bi=0.05 Ra=10 , Bi=0.15
10 10
A=5 A=5
5 A=50 5 A=50
A=500 A=500
100(Θsa,Ch - Θva,FL), %
100(Θsa,Ch - Θva,FL), %
0 0
-5 -5
-10 -10
-15 -15
-20 -20
-251 0.8 0.6 0.4 0.2 0 -25 1 0.8 0.6 0.4 0.2 0
Θva,FL Θva,FL
100(Θsa,Ch - Θva,FL), %
0 0
-5 -5
-10 -10
-15 -15
-20 -20
-251 0.8 0.6 0.4 0.2 0 -25 1 0.8 0.6 0.4 0.2 0
Θ va,FL Θ va,FL
6 6
Ra=10 , Bi=0.05 Ra=10 , Bi=0.15
10 10
A=5 A=5
5 A=50 5 A=50
A=500 A=500
100(Θsa,Ch - Θva,FL), %
100(Θsa,Ch - Θva,FL), %
0 0
-5 -5
-10 -10
-15 -15
-20 -20
-251 0.8 0.6 0.4 0.2 0 -25 1 0.8 0.6 0.4 0.2 0
Θva,FL Θva,FL
7 7
Ra=10 , Bi=0.05 Ra=10 , Bi=0.15
10 10
A=5 A=5
5 A=50 5 A=50
A=500 A=500
100(Θsa,Ch - Θva,FL), %
100(Θsa,Ch - Θva,FL), %
0 0
-5 -5
-10 -10
-15 -15
-20 -20
-251 0.8 0.6 0.4 0.2 0 -25 1 0.8 0.6 0.4 0.2 0
Θ va,FL Θ va,FL
Fig. 1. Relative differences of the sphere temperature obtained rom the lumped capacitance method (Hsa;Ch ) with respect to the volume averaged temperature from the
FLUENT solution (Hva;FL ), for different values of the Rayleigh number.
Z. Virag et al. / International Journal of Heat and Mass Transfer 54 (2011) 2303–2309 2307
Ra=104 Ra=104
Ra=105 Ra=105
Ra=106 Ra=106
100(QCh - QFL)/QFL, %
100(QCh - QFL)/QFL, %
5 Ra=107 5 Ra=107
0 0
-5 -5
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
Θva,FL Θva,FL
Fig. 2. Relative differences of the heat transfer rate obtained from the lumped model with respect to the FLUENT solution for selected values of the Rayleigh number.
6 6
2
Ra=10 , Bi=0.05 2
Ra=10 , Bi=0.15
A=5 A=5
A=50 A=50
A=500 A=500
1.5 1.5
100(Θva,FL - Θsa,FL), %
100(Θva.FL - Θsa,FL), %
1 1
0.5 0.5
0 0
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
Θva,FL Θva,FL
Fig. 3. Difference of the surface and the volume averaged sphere temperature for Ra = 106 and for two values of the initial Biot number.
number instead of ks/k for the independent variable. The difference (b) Variation of Pr = 1, 5, and 15 and A = 5, 50, and 500 at initial
between the two solutions will be analyzed for different values of Bi = 0.05 and Ra = 106. These calculations should prove the
four parameters: Pr, Ra, A, and the initial value of Bi. In natural assumption that the Prandtl number is not relevant for the
convection with Pr > 1, the buoyancy and viscous forces are in difference between the two solutions.
balance, while the inertial forces in Eq. (5) can be neglected. If so,
the Prandtl number can be eliminated from the equation, and it The results from the two approaches will differ at least for the
can be expected that the Prandtl number has not a great influence following two reasons:
on the difference between the two solutions. Because of that, the
analysis was performed in two steps: (I) In the first approach the temperature field in the sphere is not
uniform, and the sphere surface temperature (which is
(a) Variation of three parameters: Ra = 104, 105, 106, and 107, responsible for the heat transfer rate from the sphere to the
A = 5, 50, and 500 and initial Bi = 0.05 and 0.15, while Prandtl fluid) is different from the volume averaged temperature.
number is kept constant Pr = 5.
6
Ra=10 , Bi=0.05, A=5
6 10
1
Ra=10
8
A*Bi=0.25 Pr=1
A*Bi=2.5 6 Pr=5
100(Θsa,Ch - Θva,FL), %
A*Bi=25 Pr=25
0.8 4
A*Bi=0.75
A*Bi=7.5 2
A*Bi=75
0.6 0
Θva,FL
-2
0.4 -4
-6
0.2 -8
-10
1 0.8 0.6 0.4 0.2 0
0
0 0.2 0.4 0.6 Θva,FL
∼
t
Fig. 5. Influence of the Prandtl number on the relative temperature difference
Fig. 4. The time-varying volume averaged sphere temperature for Ra = 106. obtained from the two solutions at Ra = 106, Bi = 0.05, and A = 5.
2308 Z. Virag et al. / International Journal of Heat and Mass Transfer 54 (2011) 2303–2309
6 6
Ra=10 , Bi=0.05, A=50 Ra=10 , Bi=0.05, A=500
10 10
8 8
Pr=1 Pr=1
6 Pr=5 6 Pr=5
100(Θsa,Ch - Θva,FL), %
100(Θsa,Ch - Θva,FL), %
Pr=25 Pr=25
4 4
2 2
0 0
-2 -2
-4 -4
-6 -6
-8 -8
-10 -10
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
Θva,FL Θva,FL
Fig. 6. Influence of the Prandtl number on the relative temperature difference obtained from the two solutions at Ra = 106, Bi = 0.05 and A = 50, and 500.
(II) In the lumped capacitance method the quasi-steadiness is Fig. 2 shows the relative difference of heat transfer rates as a
assumed (i.e. the correlation function for a steady state iso- function of non-dimensional volume averaged sphere temperature.
thermal sphere is used in the calculation of unsteady cool- The maximal value of this difference occurs at the beginning of the
ing). This introduces some discrepancies into the solution, cooling process when the sphere surface temperature Hs decreases
especially if the cooling process is fast. rapidly, while the volume averaged temperature Hva at the very
beginning of the cooling process changes similarly to the tempera-
ture from the lumped capacitance method. In all the cases at the
3.1. Comparison of Nusselt numbers from the steady FLUENT solution beginning of the cooling process, the heat transfer rate from the
and the correlation function FLUENT solution is slower than the heat transfer rate from the
lumped capacitance method, while, after a short time, the opposite
Before a comparison of results for the unsteady sphere cooling is true – the heat transfer rate obtained from FLUENT is slightly fas-
was made, the Nusselt numbers in steady state conditions (ob- ter than the one obtained from the lumped capacitance method.
tained by the FLUENT and those obtained by the lumped capaci- The influence of Bi is shown in Fig. 3. As the values of Bi
tance method and the Churchill correlation function) for a increase, the difference between the volume and the surface
selected Rayleigh number range had been compared. The FLUENT averaged sphere temperature becomes greater. For three times
steady state solutions were used as initial conditions for unsteady higher values of Bi, the maximal difference between the tempera-
calculations. Table 1 shows the relative difference eNu = 100 tures is increased about three times. In the lumped capacitance
(NuCh NuFL)/NuFL between the Nusselt numbers obtained numer- method it is assumed that these two temperatures are the same,
ically by the FLUENT and the values defined by the Churchill corre- so it is clear that the accuracy of the LCM will decrease with an in-
lation function. The maximal relative difference is 3.98% at the crease in the value of the Biot number.
lowest value of the Rayleigh number. This is also the difference be- It is clear from Eq. (14) that the time constant for the sphere
tween the initial heat transfer rates in the two approaches. Thus, cooling is 1/(6A Bi), so the process is faster for a greater product
the acceptable difference between the LCM solution and the FLU- A Bi. This can be noted in Fig. 4 which shows the change in the
ENT solution is chosen to be 5%. volume averaged sphere temperature from the FLUENT solution
versus non-dimensional time. In the fast cooling process, the
assumption introduced into the LCM that the process is quasi-
3.2. Comparison of unsteady solutions steady is no more valid, and the difference between the FLUENT
and the LCM solution becomes greater.
The unsteady calculations were performed for different values Figs. 5 and 6 illustrate the influence of the Prandtl number on
of Bi and A using the steady state solutions obtained for the the relative temperature difference obtained from the two
constant temperature of the sphere (for the chosen Ra) as initial
conditions. Results of each FLUENT calculation are compared with
the results of the lumped capacitance method in terms of maximal
relative temperature difference of the sphere and the relative
difference of heat fluxes from the two models eQ ¼ 100 ðQ_ Ch
Q_ FL Þ=Q_ FL . The relative temperature difference is defined as
eT = 100 (Tva,FL Tsa,Ch)/(Ts,0 T0), where Tva,FL is the volume aver-
aged sphere temperature obtained from the FLUENT solution and
Tsa,Ch is the sphere temperature obtained from the LCM (using
Churchill’s correlation function). Table 2 summarizes the maximal
relative temperature difference and Table 3 the maximal relative
difference of heat transfer rate. In all calculations the Prandtl num-
ber was constant (Pr = 5).
Fig. 1 depicts the relative temperature differences versus the
non-dimensional volume averaged sphere temperature for four se-
lected values of the Rayleigh number, with the initial Bi and A as
parameters. The maximal difference occurs almost in the middle
of the temperature interval and increases with increasing initial
Bi and A, and decreases with an increasing Ra. Fig. 7. Relative temperature difference between the two solutions, %.
Z. Virag et al. / International Journal of Heat and Mass Transfer 54 (2011) 2303–2309 2309
4. Conclusion