Академический Документы
Профессиональный Документы
Культура Документы
www.elsevier.com/locate/ijhmt
Abstract
In this work, the heat transfer in a channel composed of a smooth and a corrugated wall is studied under laminar
ow conditions. The velocity and temperature distributions are determined with the help of a nite element model.
The heat transfer performance of the corrugated wall channel is compared with that of a smooth wall duct. The
numerical model is utilized in a genetic algorithm to maximize the heat transfer by optimizing the corrugation
prole, for given volume of the corrugated wall and pressure drop in the channel. Some optimum corrugation
proles are presented at the end. 7 2000 Elsevier Science Ltd. All rights reserved.
0017-9310/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 1 7 - 9 3 1 0 ( 0 0 ) 0 0 0 5 4 - 5
4300 G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310
Nomenclature
c specic heat capacity of the uid (J/kg K) tb0 bulk temperature in the reference channel
d distance between the external surfaces of the (K)
channel (m) Tb0 normalized bulk temperature in the reference
f corrugated wall thickness (m) channel
F normalized corrugated wall thickness u, v velocity components in the x- and y-direc-
-
F average normalized corrugated wall thick- tions (m/s)
ness U, V normalized velocity components
hx local heat transfer coecient (W/m2 K) w average velocity in the corrugated channel
kf , ks thermal conductivity of the uid and solid (m/s)
(W/m K) w0 average velocity in the reference channel (m/
l period of the corrugation function (m) s)
L normalized period of the corrugation func- x, y longitudinal and transversal coordinates
tion X, Y normalized coordinates
Nue equivalent Nusselt number
Nux local equivalent Nusselt number Greek symbols
p sum of the pressure and the gravity potential a convergence parameter
contribution (Pa) g ratio between solid and uid thermal con-
P normalized pressure ductivity
Pr Prandtl number Dp0 pressure drop in the reference channel for a
q0 heat ux per unit of surface (W/m2) length equal to l (Pa)
Re Reynolds number DP normalized pressure drop for a length equal
tf , ts temperature of the uid and solid (K) to l (Pa)
Tf , Ts normalized temperature of the uid and m dynamic viscosity of the uid (kg/m s)
solid r density of the uid (kg/m3)
factors concerning the geometry of the walls, the determined with the help of a nite element model.
properties of the coolant uid, and the nature of The performance of the corrugated wall channel is
the ow. Moreover, it can only be correctly com- compared with that of a conduit composed of two
pared with the heat transfer eectiveness of at zero thickness at walls having the same external size.
wall channels by also considering the external size, Moreover, the nite element model is utilized in an
the wall volume or weight and the pressure drop. original genetic algorithm to determine the values of
Most of the studies performed on the uid dy- the polynomial prole parameters which optimize the
namical and thermal phenomena occurring in corru- heat transfer performance of the corrugated channel
gated wall channels consider corrugations having a for given volume of the wall or pressure drop in the
periodical pattern which is described by simple func- channel. Finally, some optimum corrugation proles
tions such as rectangular, triangular, or sinusoidal. are presented for dierent situation.
However, due to the variety of thermal and uid
dynamical characteristics described in the literature
under dierent conditions, the study of more com- 2. The mathematical model
plex corrugation proles can be useful to better
evaluate the convenience of assigning to the channel Let us consider a channel composed of two parallel
walls corrugated rather than at proles. at walls. One wall is smooth and thermally insulated,
In the present work, the problem of optimizing the and the other is corrugated and crossed by a heat ux
heat transfer is studied in a channel composed of a at q0 which is uniformly imposed on its external surface
insulated wall and a corrugated one crossed by a heat (Fig. 1). Moreover, the thickness of the insulated wall
ux uniformly imposed. The analysis is limited to the is ideally zero, while that of the corrugated wall is a
region where the dynamic and thermal prole is fully periodical function f of the longitudinal coordinate x,
developed, under laminar ow conditions. To the cor- whose period is l. Between the two walls, a coolant
rugated wall prole, a periodical pattern is assigned, uid passes through in laminar ow in the x direction.
which is described by a fth order polynomial func- Let us introduce the following hypotheses:
tion. The velocity and temperature distributions are . the system is in steady state;
G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310 4301
! !
. the velocity and temperature proles are fully devel- @ vu @ v 2 @p @ 2v @ 2v
r m 2
oped; @x @y @y @x 2 @y2
. the uid properties are uniform;
. the viscous dissipation within the uid is negligible.
@u @v
Let y be the transversal coordinate. Due to the sym- 0 3
metry in the normal to plane (x, y ) direction, the heat @x @y
transfer performances of the channel can be studied by
where r and m are the density and the dynamic vis-
only determining the velocity and temperature distri-
cosity of the uid, respectively, u and v are the velocity
butions on the plane (x, y ). Moreover, due to the
components in the x- and y- direction, respectively,
periodicity of the channel shape, the study can be lim-
and p is the sum of the pressure and the gravity poten-
ited to a portion whose length is equal to l.
tial contribution.
The velocity distribution must satisfy the force and
The temperature distributions in both the uid and
mass balance equations, which for a steady-state
the solid must satisfy the following energy balance
incompressible ow can be reduced to the following
equation:
form:
! ! !
@ u 2 @ uv @p @ 2u @ 2u @ tf u @ tf v @ 2 tf @ 2 tf
r m 1 rc k 4
@x @y @x @x 2 @y2 @x @y @x 2 @y2
@ 2 ts @ 2 ts
0 5
@x 2 @y2
x y l fXd u
X Y L FX U
d d d d w0
6
v p tf ts
V P Tf Ts
w0 Dp0 Dt Dt
Fig. 1. (a) Corrugated channel. (b) Grids of the temperature @ VU @ V 2 24L @ P 2 @ 2V @ 2V
8
(), pressure (), and velocity (). @X @Y Re @ Y Re @ X 2 @ Y 2
4302 G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310
@U @V @ Ts 1
0 9 RePr 22
@X @Y @ Y X, 0 2g
0 12 X X X jk Y aik bik X
E Ek 23
k
Xk X j k ak aik bk bik X
U0, Y UL, Y 13 where Xk and Xjk are node coordinates in the longi-
tudinal direction, ak bk X is the value which Y
V0, Y VL, Y 14 assumes on the line passing through the oblique side
where node k is, and aik bik X is the value which Y
assumes on the opposite side. In this way, E linearly
P0, Y PL, Y DP 15 changes on the element sides and on the segments join-
ing the middle points of each couple of opposite sides.
Eqs. (7)(11) can now be integrated after having
@ V @ V
16 divided each trapezoidal element in four subelements
@ X 0, Y @ X L, Y by joining the middle points of the opposite sides. For
each node where U, V, or P is unknown, Eqs. (7)(9),
DP being the normalized pressure drop between the
respectively, must be integrated on the four subele-
inlet and the outlet section of the corrugated channel
ments surrounding the node. By taking boundary con-
portion. Moreover, the value of the pressure in one
ditions (12)(16) into account, the following system of
point of the channel is required.
equations are obtained:
Eqs. (10) and (11) must be integrated by imposing
h i
that temperatures and heat uxes in the normal to the
A Bx diagU By diagV
surface direction are identical in the solid and in the
uid on the contact surface, and: U Sx P Mx DP Nx Pn 24
Tf L, Y Tf 0, Y 1 17 h i
A Bx diagU By diagV
Ts L, Y Ts 0, Y 1 18 V Sy P My DP Ny Pn 25
@ Tf @ Tf
19
@ X 0, Y @ X L, Y
@ Ts @ Ts
20
@ X 0, Y @ X L, Y
@ Tf Fig. 2. Optimum channel geometry for Re = 500 and Pr =
0 21
@ Y X, 1 5.
G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310 4303
where a is a parameter ranging from 0 to 1. The pro- An equivalent Nusselt number for the corrugated
cedure can be iterated until relative changes in vectors channel can then be dened as the average of Nux:
U a and V a become smaller than an established value.
L
After having determined the discrete velocity distri- 1
bution, Eqs. (10) and (11) can be integrated on the sur- Nue Nux dX 34
L 0
face of the four subelements surrounding each node of
the temperature grid where values are unknown. By Such an equivalent Nusselt number corresponds to the
taking boundary conditions (17)(22) into account, the average Nusselt number which would be calculated for
following system of equations is obtained: the reference channel if on its heat ux crossed surface
h i occurred the same temperature distribution as on the
H Gx diag U t Gy diag V t T corrugated channel, for given heat ux and inlet bulk
temperature. It depends on the channel geometry, Re,
Mt Nt Tn 30 Pr, and g:
4304 G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310
Table 1
Characteristic parameters of the wall corrugations which maximize Nue under dierent conditions
Constraint Re Pr F0 F1 F2 F3 F Nue DP
DP 100 1 7.75 3
DP 100 5 0.307 0.307 0.307 0.307 0.307 7.75 3
DP 500 1 7.75 3
amplitude. Such a prole coincides with that obtained wall to the main ow due to conduction, so that the
with the genetic algorithm by maximizing the contact isothermal curves are denser. In the cavities, the heat
surface between uid and solid. When Re is equal to removed from the wall is transferred to the main ow
500 and Pr to 5, the increment of the heat transfer sur- through the eddies. This convective heat transfer mech-
face between uid and solid overcome the reductions anism is less ecient when the eddy velocity is lower
of the local convective heat transfer coecient due to (Re = 100) or the conduction is prevalent (Pr = 1),
the channel enlargements caused by the wavy prole so that, in these cases, the local convection coecient
with respect to the case of a maximum thickness wall. is noticeably reduced in the cavities.
On the contrary, when Re or Pr is lower, such re- It is interesting to observe that, for Re equal to 500
ductions are stronger and cannot be overcome by any and Pr to 5, the equivalent Nusselt numbers of chan-
extension of the contact surface. nels with maximum amplitude sinusolidal corrugation
In Figs. 3 and 4 the velocity and temperature distri- prole having period equal to l and l/2 are equal to
butions in the channel of Fig. 2 are reported for Re 8.07 and 8.28, respectively. Therefore, under these con-
equal to 500 and Pr to 5. One larger eddy occurs in ditions, sinusoidal corrugation proles perform worse
the rst cavity and smaller eddies in the second one. than fth order polynomial and at proles.
Near the peaks the heat is directly transferred from the It must be noticed that the volume of the corrugated
Fig. 7. Isothermal curves in the channel of Fig. 5 for Re equal to 100 and 500 and Pr equal to 1 and 5. Curves are drawn every
5% of the dierence between the maximum and the minimum normalized temperature.
G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310 4307
wall of Fig. 2 is much smaller than that of the maxi- respect to the case of a maximum thickness wall.
mum thickness wall obtained as the optimum one for Therefore, even if the convection in the cavities is
low Reynolds or Prandtl numbers. If needed, the thick- reduced for low values of Reynolds or Prandtl num-
ness of the at wall can be reduced, by decreasing the bers, the wavy prole results in being more ecient.
external size of the channel. In this way, for a given In Figs. 6 and 7 the velocity and temperature distri-
drop between the wall and bulk temperatures, the dis- butions in the channel of Fig. 5 are reported. For Re
sipated heat slightly increases and the system becomes equal to 500, only one larger eddy occurs in the two
more compact. However, if the external size of the cavities, while for Re equal to 100 two smaller eddies
channel is constrained, it can be interesting to optimize occur. When Re is lower, for given q0, the temperature
the corrugation prole for a given value of the average increment Dt is higher, since the mass ow rate is
wall thickness F, which is representative of the wall lower. Therefore, the temperature drop between the
volume and weight. wall and the uid is relatively smaller, mainly when Pr
By constraining the volume of the corrugated wall, is low and the conduction prevails.
it has been found that a wavy prole maximizes Nue By looking at Table 1, it is evident that the corru-
even if Re and/or Pr are low. In particular, by con- gated wall of Fig. 5 provides a much larger pressure
straining F to 0.2, the optimum wall prole reported in drop than the at one having the same volume, while
Fig. 5 has been obtained in every case. In particular, the corrugated wall of Fig. 2 provide a smaller press-
with respect to the case of a maximum thickness wall ure drop than the at one having the same maximum
(F(X ) = 0.2; Nue 6:72; DP 1:95), the wavy wall of height. Therefore, it is interesting to nd the wall ge-
Fig. 5 provides a 30.5% increment in the heat transfer ometry which maximizes Nue for a given pressure drop
when Re is equal to 500 and Pr to 5, a 8.9% increment between the inlet and the outlet sections.
when Re is equal to 500 and Pr to 1, and a 7.7% By constraining the normalized pressure drop DP, it
increment when Re is equal to 100 and Pr to 5. has still been found that a wavy wall prole only maxi-
The optimum corrugation prole obtained in every mizes Nue when both Re and Pr are not too low. In
case by maximizing Nue for F constrained to 0.2 co- particular, by constraining DP to 3 and 2.5 the opti-
incides with that obtained with the genetic algorithm mum wall proles reported in Fig. 8 has been found
by maximizing either the contact surface between uid for the case where Re is equal to 500 and Pr to 5. The
and solid or the maximum thickness of the wall. Such optimum proles of Fig. 8 are very similar to that of
a prole provides larger values of the local convection Fig. 2, but the wave amplitude is reduced. In the other
coecient near the higher peak and a more extended three cases (Re = 100 and/or Pr = 1), by maximizing
heat transfer surface between uid and solid with Nue , a at wall has been obtained having normalized
Fig. 8. Optimum channel geometries for Re = 500, Pr = 5, and DP constrained to 2.5 and 3.
4308
G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310
Fig. 9. (a) Comparison between the equivalent Nusselt number analytically calculated for a at wall channel (continuous line) and the equivalent Nusselt number of the optimum
geometries found by the genetic algorithm for Re = 500, Pr = 5, and DP constrained to dierent values (stars). (b) Increments in Nue of the corrugated channel referred to the
value of the at wall channel.
G. Fabbri / Int. J. Heat Mass Transfer 43 (2000) 42994310 4309
thickness equal to 0.307 for DP constrained to 3 and In general, the relative improvements in the heat
to 0.263 for DP constrained to 2.5. However, referring transfer provided by the optimum fth order poly-
to the heat dissipated through a at wall causing the nomial corrugation prole increase with the Reynolds
same pressure drop, the optimum geometries obtained and Prandtl numbers and with the pressure drop in the
for the case where Re is equal to 500 and Pr to 5 pro- channel. Such a result is in accordance with the con-
vide a 4.4% increment when DP is constrained to 3 clusions of the previous analyses of the heat transfer in
and only a 0.7% increment when DP is constrained to corrugated channels considering simpler corrugation
2.5. Moreover, by constraining the normalized pressure proles such as rectangular [10] or sinusoidal [13].
drop to be less than 2.4, a at wall has been found to However, the adoption of a fth order polynomial
maximize Nue even in the case where Re is equal to prole in the present analysis allows more various and
500 and Pr to 5 (Fig. 9). Therefore, by assigning to the ecient corrugated wall geometries to be considered.
channel a wavy wall, improvements in the heat transfer
can only be obtained when the allowed pressure drop
is not too low.
References
rugata, in: Proc. 49th National Congress, Perugia, Italy, [16] G. Fabbri, A genetic algorithm for n prole optimiz-
1994, pp. 314. ation, Int. J. Heat and Mass Transfer 40 (9) (1997)
[14] E. Lorenzini, M. Spiga, G. Fabbri, A polynomial n 21652172.
prole optimization, Int. J. Heat and Technology 12 (1/ [17] G. Fabbri, Heat transfer optimization in internally
2) (1994) 137144. nned tubes under laminar ow conditions, Int. J. Heat
[15] N. Queipo, R. Devarakonda, J.A.C. Humphrey, Genetic and Mass Transfer 41 (10) (1998) 12431253.
algorithms for thermosciences research: application to [18] G. Fabbri, Optimum proles for asymmetrical longi-
the optimized cooling of electronic components, Int. J. tudinal ns in cylindrical ducts, Int. J. of Heat and
Heat and Mass Transfer 37 (6) (1994) 893908. Mass Transfer 42 (3) (1999) 511523.