Documente Academic
Documente Profesional
Documente Cultură
References
Cao, Y Beam, J., and Donovan, B., 1996, "Air-Cooling System for Metal Oxide Nomenclature
Semi-conductor Controlled Thyristors Employing Miniature Heat Pipes," J. Thermo- a/s = specific area of fluid-solid interface per unit volume
physics and Heat Transfer, Vol. 10, No. 3. pp. 484-489.
Faghri, A., 1995, Heat Pipe Science and Technology, Taylor & Francis, London. d = diameter
Fisher, L., and Martin, H., 1997, "Friction Factors for Fully Developed Laminar
Flow in Ducts Confined by Corrugated Parallel Walls," Int. J. Heat and Mass
Transfer, Vol. 40, pp. 635-639. Department of Mechanical Engineering, Hong Kong University of Science and
Hopkins, R., 1996, "Flat Miniature Heat Sinks and Heat Pipes with Micro Capillary Technology, Clear Water Bay, Kowloon, Hong Kong, e-mail: mechthsu@usthk.ust.hk.
Grooves: Manufacturing, Modeling and Experimental Study," Masters thesis, The Mem. ASME.
University of Connecticut, Storrs, CT. Contributed by the Heat Transfer Division for publication in the JOURNAL OF HEAT
Hopkins, R., Faghri, A., and Khrustalev, D., 1997, "Flat Miniature Heat Pipes with TRANSFER. Manuscript received by the Heat Transfer Division, Mar. 31, 1997; revision
Micro Capillary Grooves," Proc. of the 1997 National Heat Transfer Conference, received, Mar. 25, 1999. Keywords: Conduction, Porous Media, Transient and Un-
HTD-Vol. 349, Vol. 11, pp. 71-80. steady Heat Transfer. Associate Technical Editor: M. Kaviany.
Journal of Heat Transfer Copyright 1999 by ASME AUGUST 1999, Vol. 121 / 733
represents the interfacial heat transfer. The last equalities in Eqs. Q = M0(t, - tf) + M, (V7> - aVt,) (19)
(7) and (8) are evident from the interfacial boundary conditions (3)
and (4). The negative sign in the last two terms of Eq. (6) reflects where
the fact that the source terms in Eq. (5) have to become the sink
terms in Eq. (6), or vise versa. These last two integral terms
M kfVf'o- ds and M, = kfVt\ ds.
represent the additional unknowns to the equation system (5) and =V
(6). Closure modeling for these integral terms then becomes inev-
itable. (20a, b)
The above derivations are basically in-line with those of Quintard
and Whitaker (1993); however, here we provide a more concise
3 Closure Modeling
treatment in obtaining the relations (17) and (19) which in form are
To close the equation system (5) and (6), we need to construct slightly different from those of Quintard and Whitaker (1993). It is
the constitutive equations which relate the integral terms to the important to note that the closure coefficients, G0, G,, M0, and
macroscopically phase-averaged temperatures, Tf and Ts. To this M,, depend only on the detailed local geometry of particles, but
end, we first decompose Tf and Ts into not on the macroscopic quantities. For randomly distributed spher-
ical particles, the volumetric averaged tensor G, has to be axially
Tf=Tf+T} (9) symmetric to exhibit an isotropic property, i.e., G, = Gl where /
is the unit matrix and G is a scalar quantity. According to New-
and ton's law, the locally interfacial heat transfer is proportional to the
total fluid-solid interfacial area Afs in REV. Therefore, we have
Ts=ts+ T, (10)
M0 = hfsafs where a{s (= AfJV) is the specific interfacial area and
where T'f and T's represent the microscopic temperature variations hfs is the interfacial heat transfer coefficient. From the physical
from the phase-averaged values. With the REV length scale being point of view, a nonzero value in G 0 or M, will lead to a
I, the time scale of the macroscopic conduction in fluid phase is convective behavior associated with the macroscopic temperature
l2/af which is much larger than the time scale of the microscopic gradient, where the vectors G 0 and M, resemble the convection
conduction given by d2pla{ since I > dp; similar argument applies velocities. Quintard and Whitaker (1993) had argued that this
to solid phase. With respect to the macroscopic process of long convective behavior does not have a physical ground since all
time scale, it is plausible to assume that the local microscopic heat these transfer processes are essentially from conduction. They also
conduction process is quasi-steady. Under this quasi-steady as- demonstrated mathematically that G0 = M, = 0. Equations (17)
sumption, the substitution of (9) and (10) into (1) and (2) leads to and (19) then reduce to
The predictions of kjk{ based on Eq. (29) for <J> = 0.36 are shown
Qfs = hfsafs(Ts - Tf). (22) in Fig. 3. For comparison, the predictions based on Hsu et al.'s
Equations (5) and (6) with the tortuosity and interfacial transfer (1995) model for in-line two-dimensional array of cylinders are
terms given by Eqs. (21) and (22), respectively, are the governing also included in Fig. 3.
equations for the transient heat conduction in porous media. The
evaluation of the closure coefficients G and hfs then becomes a 5 Evaluation of Closure Coefficients
critical task. We want to reemphasize that the values of G and hfs We now return to the transient heat conduction problem of local
depend only on the microscopic interfacial geometry and the thermal nonequilibrium. To render Eqs. (21) and (22) applicable,
thermal properties of solid and fluid. analytical expressions for evaluating G and hfs are required. They
are described below.
4 Lumped Mixture Model Under Local Thermal Equi-
Tortuosity Parameter G. Since the value of G depends only
librium Condition on the local interfacial geometry and on the solid and fluid thermal
Before evaluating the coefficients G and hf we now examine properties, it is plausible to assume that the expression for G
the heat conduction in porous media under the local thermal obtained under the local thermal equilibrium condition can be
equilibrium condition. This condition implies that extended to the thermal nonequilibrium regime. This bears the
resemblance to the closure modeling of turbulent flows where a
T=T,= Tf . (23) simple flow is usually used to evaluate the closure coefficient in a
closure relation. Therefore, G is now obtained by solving Eq. (26):
The constraints for invoking the local thermal equilibrium assump-
tion were discussed in detail by Quintard and Whitkaer (1995).
Equations (5) and (6), after invoking Eqs. (21), (22), and (23), are G = - 4 > - c r ( l -</>) /(I - vV (30)
added together to lump into the following transient heat conduc-
tion equation for the solid-fluid mixture: where kjkf is then determined by Eqs. (27) and (28).
To demonstrate the characteristics of G, the values of G were
(24) calculated from Eq. (30) over a wide range of (j>, yc, and a. Figure 4
(pCp)md^=V-[keVT] shows the results of G for different yc with <j>fixedat 0.36. The
distinct feature is that G is always negative. This has to be the
where situation since 4> + o(\ - (f>), which represents the result of a lumped
parallel two-layer model, is the upper limit of kjkf. This implies that
(pcp)m = <j>pfCpf+ (i - 4>)PsCps (25)
is the heat capacitance and
10"
ke = 4>k{+ (1 - <j>)ks + kf{\ - a)2G (26)
experiments 0.02
the effective stagnant thermal conductivity of the mixture. Again, 103 3D model c/a = 0.00
the value of ke depends only on the microscopic interfacial geom- 3D model c/a = 0.13
2D model, c/a= 0.00, 0.01 & 0,02
'/,
etry and the solid-fluid properties. Equation (24) resembles to the 10 2
K/VO.OI
transient heat conduction of a pure substance whose solutions
ke/kf /, 4&7/
subjected to different initial and boundary conditions have been
obtained extensively in open literature. Therefore, the main task
becomes the determination of the effective stagnant thermal con- c/a = 0.00
ductivity.
There exist several models for the effective stagnant thermal ,J^ $=0.36
conductivity as mentioned in Section 1. For details of them, the ^^
readers are referred to the review article of Kaviany (1991) and 10-2 10-1 10 10 2 10 3 10" 10 5
Cheng and Hsu (1998). Here, we adopt the lumped composite ks/kf
layer model of Hsu et al. (1995) for in-line periodic array of
three-dimensional cubes with the unit cell shown in Fig. 2. Using Fig. 3 The predictions of the effective stagnant thermal conductivity for
media of in-line periodic array of three-dimensional cubes and two-
a composite layer scheme, they obtained the following expression dimensional cylinders, and their comparison to experimental data of
for kjkf. Nozad etal. (1985)
I
-t 7
j
tjII
II
-0.05
M'-/
/
it' i
\ L*
"i
T
Q -0.10 rJ* = J II
"T if -*''*'
1
f
-"
1it1
tt TKS
jT tt
' If tt %'
T
I . -' fl l
tt itj*
T tt ^
1 1
tt1
_.I ~ jj'07
0.001 0.01 0.1 10 100 1000
0.1 1000
a = ks/kf
a = ks/kf
Fig. 4 The effect of particle-touching parameter on the tortuosity coef-
ficient when the porosity is 0.36 Fig. 5 The effect of porosity on the tortuosity coefficient when the
particle-touching parameter is 0.1
Nu = (37)
A(l + Jfita)
The other limit case occurs when j3 is very small, i.e., when the
solid is thermally much less diffusive than fluid. Then B ap-
proaches zero and Nu 2(1 + 1/A). This case behaves quite
similarly to the case of a o; it also resembles to the mass
Fig. 7 The dependence of the Nusselt number of interfacial heat trans- diffusion process in fluid phase, which depends solely on the mass
fer on the local response time scale A and on the thermal properties of
two phases, a and ji. The results of Quintard and Whitaker (1995) for </> =
diffusivity of fluid.
0.36 are shown to be a special case of A = 0.2 and /3 = 5.4.
6 Concluding Remarks
In this study, a two-equation model is proposed for the transient
hA 2a heat conduction in porous media under the nonthermal equilibrium
Nu = f (35) condition. Macroscopic equations for fluid and solid phases were
B aAa + aB
obtained from the microscopic transient heat conduction equations
1 + A CT(1 - B)
through a phase averaging procedure. The averaging procedure leads
where Nu is the Nusselt number for the interfacial heat transfer to the closure problem where the number of unknowns become more
coefficient, aA = A/(l + A) and aB = B/(l - B). From the than the number of equations. Closure relations for the two additional
conservation of thermal energy, A and B are related by unknowns, namely, the tortuosity and interfacial heat transfer terms,
are derived from the microscopic equations. The tortuosity constant G
3 + A B which appears in the closure relation is evaluated based on the lumped
j8A2 B2 (36) parameter model for the effective thermal conductivity proposed early
1 +A 1 -B
by Hsu et al. (1995), while the interfacial heat transfer coefficient hfi
where j3 = ajaf is the thermal diffusivity ratio of solid to fluid. is based on a microscopic quasi-steady assumption.
The values of Nu are evaluated from Eqs. (35) and (36) for The present result of G shows mat G is always negative, i.e., the
different values of A, cr, and 0. The results are plotted in Fig. 7 as tortuosity effect is to reduce the effective thermal conductivities of
a function of cr for A = 0.1, 0.2, and 0.4, with |3 varying. Figure fluid and solid. There is a minimum value of G at certain porosity
7 indicates that Nu approaches zero when cr approaches zero. We because G approaches zero when the porosity approaches zero or one,
also find that Nu approaches a constant when cr approaches infi- i.e., when the media become pure substances. If the porosity is fixed,
nite. When cr approaches zero, the solid particles behave as being the absolute values of G appear to be largest as cr approaches zero,
nonconductive and little heat can transfer across the interface. On reduce monotonically as cr increases, and become zero as cr ap-
the other hand, when <x approaches infinite the particles become proaches infinite. This characteristics suggests that the results of G
highly transmittable to thermal conduction and the heat flux across shown in Figs. 4 and 5 should apply only to media with packed solid
the interface depends solely on the thermal conductivity of fluids. particles as originally implied from the model of Hsu et al. (1995). For
In fact, the interfacial transfer coefficient hfs becomes linearly materials of different microscopic geometry such as sponges with
proportional to the thermal conductivity of fluids and the propor- phase symmetric property, other appropriate model for kjks such as
tional constant depends solely on A. The dependence of the inter- the one proposed by Hsu et al. (1994) should be used. It is expected
facial transfer coefficient on the thermal diffusivity ratio j8 in the that the present two-equation model for transient heat conduction is
middle range of cr is interesting. From the quasi-steady model as applicable to a variety of media once the appropriate lumped param-
depicted in Fig. 6, the effective thermal conductivity of the con- eter model for kjkf is chosen.
duction layers near the solid-fluid interface resembles that of two The model for the interfacial heat transfer as proposed in this
layers in-series. This is implied from Eq. (35) where ajkf and study indicates that Nu depends on the microscopic geometric
aB/ks are the effective thermal resistances for fluid and solid parameter (or local response time scale parameter) A and the
conduction layers, respectively. With the normalization of hfs with material thermal parameters cr and j3. While the values of cr and j3
respect to kf, /3 is a measure of the relative thickness of the thermal are determinative once the materials are chosen, the value of A
layer inside the sphere. Higher )3 implies thicker thermal layer depends on how the particles are packed, i.e., on the porosity. For
inside the sphere, and therefore lower value of Nu as implied by dispersed particles immersed in fluid of large extent, the assess-
the layer-in-series model. ment of A is more subtle. However, for densely packed media it is
Quintard and Whitaker (1995) obtained an analytical expression expected that the thermal conduction in the fluid phase will be
for hfi based on the unit cell of Chang (1983). Their expression for largely confined by the neighboring solid particles. The inverse
Nu can be rewritten into the same form as Eq. (35), but the dependence of Nu on A as shown in Eqs. (35) and (37) suggests
coefficients aA and aB depend solely on the porosity. For com- that the interfacial transfer coefficient hfs is sensitive to the time
parison, the values of Nu are computed from their expression with scale chosen for quasi-steady analysis. The equivalent radius r1e3 for
4> = 0.36 and are also plotted in Fig. 7. We found that the results the unit cell of Chang (1983) is given by rjr = (1 - cb)" ' . If
predicted from our present model become identical to those from the microscopic interfacial heat transfer is assumed to reach a
Quintard and Whitaker's (1995) model if A = 0.2 and j3 = 5.4. quasi-steady state when the diffusion length of thermal boundary 1/3
However, it should be noted that the present model given by (35) layer in fluid becomes re, we have 1 + A = (1 - cb)~ . For a
and (36) is more general than Quintard and Whitaker's since their medium with a typical value of cb = 0.36, we find that A = 0.16.