Inf. J. Heat Mass Transfer. Vol. 37. No. 17, DD. 2601-2613.

1994
Copynght 0 1994~&emer Science Ltd
Pergamon Printed in Great Britain. All rights reserved
0017-9310194 $7.N+-0.00

0017-9310(94)EOO99-G

Developing laminar flow and heat transfer in a

rectakgilar duct with one-walled injection
and suction
Y. C. CHENG and G. J. HWANGt
Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan
100. Republic of China

and

M. L. NG
Energy and Resources Laboratories, Industrial Technology Research Institute, Chutung, Taiwan
310, Republic of China

(Received 9 September 1993 and in revisedform 30 March 1994)

Abstract-This paper presents a numerical investigation on the developing laminar flow and forced
convection in the entrance region of a rectangular duct with one wall subjected to a constant heat flux and
a uniform fluid injection or suction. Using the pressure deviation and vorticity-velocity method, the three-
dimensional Navier-Stokes equations and energy equation are solved simultaneously. Typical velocity and
temperature distributions along the flow direction are reported, and both the hydrodynamic and thermal
entrance lengths are examined. The phenomenon of flow reversal in suction flow is illustrated by variations
of the axial pressure gradient. Friction factors and Nusselt numbers are also described for various wall
Reynolds numbers and aspect ratios. Finally, a comparison of this numerical study with the existing data
is made.

INTRODUCTION developed flow of an incompressible fluid in a one-

porous-wall channel with uniform injection or
FLOW and heat transfer in a duct with wall fluid
suction, Donoughe [12] showed that equations of
injection or suction has been studied extensively in the
motion could be reduced to a fourth-order nonlinear
past. Most of investigations dealt mainly with uniform
ordinary differential equation by a similarity trans-
transpiration for porous tubes or symmetric tran-
formation, and obtained a third-order perturbation
spiration for porous parallel plates, and a few studies
solution for the cases of suction or injection,
reported the results of semiporous ducts. There are
- 4 < Re, < 4. Eckert et al. [ 131 performed a numeri-
numerous engineering applications for semiporous
cal analysis by employing the Runge-Kutta method
ducts, e.g. the vapor flow through a metal screen for
for a wider range of the wall Reynolds number,
food processing [l], and the solar air collectors [2, 31
- 13 < Re, < 20. The non-existence of a solution for
using transpired metal plates. Jorne [4] and Lessner
Re, < - 13 was observed due to flow reversal at the
and Newman [5] studied the mass transfer along a
one-porous-wall vertical channel in electrochemical impermeable wall. White [14] selected the power-series
systems. For transpiration cooling of turbine rotor method to obtain an exact solution for the whole
blades, Soong and Hwang [6] worked on the laminar range of Re, except for Re, < - Il. The flow reversal
mixed convection in a radially rotating semiporous was observed at Re, = - 11. Recently, Soong and
channel. In fuel cell stacks [7, 81, both the gaseous Hwang [6] applied ttie similarity transformation to
reactant flows of cathode and anode along an exo- the conjugated momentum and energy equations for
thermic porous electrode are subjected to wall heat laminar mixed convection in a radially rotating semi-
flux with fluid injection or suction. Among different porous channel with constant wall heat flux. The
types of cell stacks, the gas flow in one-porous-wall fully developed velocity and temperature profiles were
rectangular ducts for phosphoric acid [9], polymer obtained by using the fourth-order Runge-Kutta
electrolyte membrane [lo] and molten carbonate fuel scheme. Flow reversal was also observed and dis-
cells [ 1 l] can be found. cussed in their results.
For two-dimensional, steady, laminar fully In the entrance region, Massey [l] solved the energy
equation with weak injection, 0 < Re, < 4, by using
an integral method. Rhee and Edwards [2] introduced
tAuthor to whom correspondence should be addressed. the assumption of boundary-layer flow and a concept

2601
2602 Y. C. CHENG et al.

NOMENCLATURE

A cross-sectional area, ab u, u, w dimensionless velocity components in

a height of a rectangular duct the x, y and z directions,
b width of a rectangular duct respectively, U/U,, VD,/v, and WD,,/v
hydraulic diameter, 4A/S X, Y. Z rectangular coordinates
.r” friction factor, 2r,/(pU*) X, I’, z dimensionless coordinates, X/(Re,D,):
h heat transfer coefficient Y/D,,, and Z/D,,.
k thermal conductivity
Lhyr 4! hydrodynamic entrance length and
dimensionless hydrodynamic Greek symbols
entrance length, I,,,,= L,,/(ReD,) c? thermal diffusivity, k/(pc,)
Ltlu 4, thermal entrance length and Y aspect ratio, a/b
dimensionless thermal entrance length, t, dimensionless temperature,
&h= -&I(Pe&) (T- T,)l(@&
M, N number of grids in the JJ and z V kinematic viscosity
directions 5 vorticity defined by equation (9)
NM local Nusselt number, hDJk P density
P, P pressure and dimensionless pressure, r shear stress.
P = P/w%
P’, P' pressure deviation and dimensionless
pressure deviation, p’ = P’Dz/(@) Subscripts
P,P pressure and dimensionless pressure b bulk fluid condition
averaged over a cross section, point of complete mass extraction
p = P/(plJ$ ;d value in the fully developed region
Pe Peclet number, PrRe t value at the upper wall
Pr Prandtl number, V/E W wall condition
Re local Reynolds number, C?D,/v 0 inlet condition.
Re, wall Reynolds number, V,D,/v
ReO inlet Reynolds number, ~oDh/v
S perimeter, 2(a + b) Superscripts
_
T temperature average value
+ definition for x+ = x/Pr
U, V, W velocity components in the X, Y
* normalized value.
and Z directions, respectively

of stream function. Quasi-linear equations of momen- The interesting phenomena of fully developed flow,
tum and an energy equation with isothermal boundary complete mass extraction, flow reversal as well as the
conditions were solved using a finite-difference results of friction factors and Nusselt numbers are
method. In addition, Sorour and Hassab [3] used the presented.
perturbation solution of axial velocities and the
method of superposition to solve the energy equation
THEORETICAL ANALYSIS
for a semiporous channel with a constant wall heat
flux. The eigenvalues were calculated using an implicit Consider a steady laminar flow of an incompressible
finite-difference scheme. Later, Sorour et al. [15] fluid with constant physical properties flowing
reworked the same problem using a quasi-linear tech- through a horizontal rectangular duct. The physical
nique for solving both the momentum and energy configuration and coordinate system are shown in Fig.
equations with non-uniform wall transpiratioh. 1. At the inlet of this duct, both the axial velocity and
Again, flow reversal appeared for strong suction rates. temperature distributions of the fluid are uniform.
For a three-dimensional flow, Hwang et al. [7, 81 The lower wall of the duct is porous with uniform
used the vorticity-velocity method to obtain the flow injection or suction and subjected to a constant heat
and temperature fields in a square duct with one wall flux, while the other three walls are impermeable and
subjected to uniform fluid injection or suction and adiabatic. This injection fluid is the same as that of
constant heat flux. However, aspect ratios other than the mainstream and has the same temperature of the
1 for small one-porous-wall rectangular ducts are fre- heated porous wall. For the case of suction flow, the
quently used in fuel cell stacks as shown in Fig. 1. above conditions are fulfilled automatically. In
The present paper investigates both the flow and heat addition, the external force, compression work and
transfer characteristics in a developing region of rec- viscous dissipation are all neglected.
tangular ducts with aspect ratios of 0.2, 0.5, 2 and 5. In the fuel cell system in Fig. 1, a uniform chemical
 Laminar flow and heat transfer in a rectangular duct

lel
Flo
Air Flow
Fuel Cell Stack

.’
/ “W ’ Heated Porous Wall , Q

FIG. 1. Schematic diagram for fuel cell stack and coordinate system

reaction rate is assumed in the plate between the anode in the energy equation can be neglected. Hence the
and cathode ; thus constant wall heat flux is used by governing equations can be obtained in dimensionless
many researchers [7-l 11. In our laboratory, an exper- forms as :
iment has been carried out for the thermal boundary
condition by using an electric screen heater on the d”+!?+;““=o
bottom duct wall. The injected air is heated separately ax ay aZ
to the wall temperature by a group of heaters before
a~ au au dp a% a%
the air flows into the duct. U~+u-+w~=--+,+-$ (4)
With the above assumptions, conservation equa- ala dx a,
tions for mass, momentum and energy can be listed ac au apc a2u a%
u~++t~-+wz=--+I+t,
a~
as follows : (5)
a,v ay ay cz
v-v=0
aw aMl ih apj azw al,*,
u-++-++---_--+++_
(V*V)V = - ;vP- ++vv2v ax ay aZ a,- ay* a22 (6)

(V*V)T = aV2T (1) .;+,.;+,;=;($+$). (7)

where P(X) is the averaged - .pressure over a cross
section at axial location X, and P’(X, Y, Z) is the pres- The term, ap’jax, in equation (4) is already eliminated
sure deviation induced by the entrance effect and fluid under the assumption of high entrance Reynolds num-
injection or suction. Referring to the coordinate sys- ber. The initial and boundary conditions are
tern in Fig. 1, the following dimensionless variables
and parameters can be introduced : u-l =tl=b~=H=O atentrance,x=O

x = X/(D,Re,) y = Y/Dh z = Z/D,, LI= c- Re, =w=c?~/ay+i=o

u= u/u, v = VD,/v w = WD,/v at porous wall, y = 0

j? = p/(pU;) p’ = p’D;/(pv’) u = l3 = M’= aOpy = 0

I,
6 = (T- T,)/(qD,/k) Reo = UODh/v at impermeable wall, y = l$
Re, = V,D,lv Pr = v/u. (2)
u = c = w = L%/az = 0
At high inlet Reynolds number, Re,, >> 1, and high
Peclet number, Pea >> 1, all axial diffusion terms in the 1,
at impermeable walls, z = 0 and li-i (8)
momentum equations and the axial conduction term 8
2604 Y. C. CHENG et al

where y is the aspect ratio of the rectangular duct. 1

NM= - (15)
Referring to the coordinate system shown in Fig. 1, &J-Q,
the value of Re, is positive for injection flow, negative
for suction flow and zero for impermeable flow. The where t& is the bulk mean temperature difference
solution of these equations depends on three inde- defined as
pendent parameters : the aspect ratio y, the wall Reyn- 4Y 1
&-- u0 dy dz.
olds number Re, and the Prandtl number Pr. (16)
(l+y)* ii
In common practice, the magnitude of Re, is
smaller than 1 in fuel cells [9] and electrochemical cells
[4, 51. In addition, Re, is between 1 and 10 for drying
METHOD OF SOLUTION
processes [l] and solar air collectors [2, 31. For the
purpose of engineering applications, - 20 < Re, < 20 To find a solution analytically for the unknowns, u,
is selected in the present investigation. Therefore, v, w, [ and 0 in this problem is beyond contemplation.
Rr, cc Rq, is assumed in this study for fast duct flow Hence, a numerical finite-difference scheme based on
(Re, >> 1). the vorticity-velocity method [16] is employed to
To simplify further the governing equations, a vor- obtain the simultaneous solutions at each desired axial
ticity function in the axial direction is defined as location with different y and Re,. The three parabolic-
elliptic partial differential equations (4, 12, 7) for II, [
and 0 are solved by the Du Fort-Frankel matching
scheme, and the other two elliptic equations (IO, 11)
By differentiating equation (9) and combining with are solved by the alternating direction implicit scheme.
continuity equation (3), the following equations for u The average axial pressure term, -dp/dx, is adjusted
and UJcan be derived : iteratively to meet the constraint of overall mass bal-
ance at each axial location. It usually takes only three
_+e=
d2V -3
aZ
a2u
ayax (‘0)
steps to get an accurate value for this axial pressure
ay* az* gradient. The detailed numerical procedures can be
2 found in Hwang et al. [7].
C!+zW=~_- a% For a better numerical accuracy, different cross-
ay azax’ (11)
ay* az* sectional meshes were used for different aspect ratios.
By using a cross differentiation of pressure terms in Numerical experiments were carried out to ensure the
equations (5) and (6) and combining with equation accuracy of the results on the step and mesh sizes,
(3), a single equation for the axial vorticity can now even at strong injection and suction conditions which
be obtained : may cause stiff velocity and temperature gradients.
Table 1 presents the results of numerical experiments.
ai ai ai ,av gaw For the cases listed in this table, most of the variations
U~+v-+w~+i-+~
aY aY in the friction factors, fRe, and Nusselt numbers, Nu,
are less than 1% at all axial positions. Although the
auaw auao ay ay
+ _____ =-+g (12) differences of Nu for the cases y = 0.2, 0.5 and 1.0
(Tyax azax ay2 are slightly higher at x = 0.001 for injection flow
Re, = 20, the values are still less than 2%.
Then equations (4), (lo), (11), (12) and (7) for u, v,
As a partial verification of the computation, the
w, <, and 0 can be solved. The pressure gradient,
hydrodynamically developing flow was calculated
- dp/dx, in equation (4) is determined by the overall
without mass transpiration. These results are com-
mass balance,
pared with Curr et al. [17], and the differences in
friction factors are within 1% at all axial locations for
(13) y = 1.O. Due to the lack of simultaneously developing
flow data for rectangular ducts with one heated wall
where tl is the averaged axial velocity obtained from
in the available literature, only the fully developed
the solution of equation (4).
friction factors and Nusselt numbers for various
The local friction factor can be computed by using
aspect ratios are compared as shown in Table 2. The
the wall velocity gradients,
peripherally averaged friction factors deviate within
1.O% from those of Shah and London [18], and the
averaged Nusselt numbers deviate within 1.7% from
those of Schmidt and Newell [19] except for the case
of y = 0.2 which is 4.3% because of the coarse 10 x 10
finite-difference grids used. All the present cal-
culations were performed on an IBM 560 workstation.
and the local Nusselt number can also be calculated Typical computation time is approximately 6690
from an average temperature difference between the CPU minutes for computation down to the axial dis-
heated wall and the bulk mean fluid, tance x = 0.1. It is noted that the computation time
 Table 1. Numerical experiments for grids and step sizes (Pr = 0.72)

Rl 0.2 0.5 1.0 2.0 5.0

Aspect ratio ...~
X
Grid. No., M x N 31 x81t 41 x91 41 x6lt 51 x71 51 x51t 61 x 61 61 x41? 71 x51 81x31? 91 x41
Step size x I O5 5.0 2.5 5.0 2.5 5.0 2.5 5.0 2.5 5.0 2.5
X

32.64 32.60 31.52 31.68 31.62 31.83 32.15 32.93 35.02 35.01
8.358 8.508 7.957 8.094 7.531 7.645 7.169 7.224 6.860 6.853
24.17 24.23 22.27 22.33 21.38 21.44 21.44 21.48 22.48 22.50
2.040 2.039 1.695 1.690 1.384 1.378 1.108 1.102 0.857 0.850
24.12 24.17 21.94 21.98 20.72 20.83 20.24 20.28 21.06 21.11
0.920 0.921 0.632 0.630 0.407 0.405 0.241 0.240 0.127 0.126

42.57 42.19 39.20 39.53 37.10 37.48 36.11 36.43 36.48 36.56
23.53 23.66 23.17 23.34 22.86 23.04 22.57 22.75 22.61 22.73
= 0.01 30.66 31.09 25.38 25.61 22.36 22.51 21.28 21.39 21.96 22.04
16.62 16.50 16.12 16.08 15.81 15.78 15.71 15.66 16.11 15.97
= 0.1 20.77 20.85
15.20 15.07
____ ~~_____
tThe number of grids and step sizes used in the present study.
$Flow reversal at x = 0.011 for ‘J = 0.2, x = 0.0156 for y = 0.5, and x = 0.0294 for y = 1.0, and complete mass extraction at x = 0.075 for 7 = 2.0 and x = 0.225 for Y = 5.0.
2606 Y. C. et al.
CHENG

Table 2. Comparison of fully developed friction factors and Nusselt numbers of impermeable
duct flow

Y fRe .fRet Nu NUS

0 23.955 24.00 5.374 5.385
0.2 19.01 19.07 4.223(31 x 8 1) 4.411
0.5 15.50 15.55 3.489(41x 61) 3.539
1.0 14.19 14.23 2.681(51x 51) 2.712
2.0 15.50 15.55 1.830(61x41) 1.854
5.0 19.01 19.07 0.948(81 x 31) 0.964

t Shah and London [18].

$ Schmidt and Newell [19],grid size IOx 10.
5 The values were obtained for y = 0.001.

depends mainly on the length of axial distance, the and -5, respectively. It is known that the maximum
aspect ratio and the magnitude of injection or suction axial velocity for the case of y = 1 is the largest one
velocity through the porous wall. among those for y $ 1 with a fixed mass flow, but it
is not true in a duct with either injection or suction.
For injection flow, the mass additions for y = 0.2 and
RESULTS AND DISCUSSION
y = 0.5 are larger than for y = 1 ; thus their axial vel-
For exploring the flow and heat transfer charac- ocities are larger than for y = 1 along the axial
teristics, the effects of fluid injection or suction and distance. However, due to smaller mass additions, the
the aspect ratio on the development of axial velocity velocity profiles for y = 2 and 5 are flatter than that
and temperature profiles in the rectangular duct are for y = 1 as depicted in Fig. 2(e). In addition, the
discussed. Both the hydrodynamic and thermal upward shifts of axial velocity peaks for y = 0.2 and
entrance lengths are examined for a further under- 0.5 are slightly lower than for y = 1. As noted in Fig.
standing of the problem. The interesting phenomena 2(f) for suction flow, the effect of mass subtraction on
of fully developed flow, complete mass extraction and u is more significant when y is smaller, and the peaks
flow reversal are illustrated. The friction factors and of velocity profiles along the duct are higher when y
Nusselt numbers in the entrance as well as the fully is larger. It is also shown by the dotted lines in Figs.
developed regions are described for various wall 2(a)-2(f) that the normalized velocities, U* = U/o,
Reynolds numbers and aspect ratios. Finally, a com- are very similar to each other for different wall Reyn-
parison of this numerical study with existing data is olds numbers and also similar to those of constant
included. mass flow for different aspect ratios.
In Fig. 3, the constant-velocity maps of u, v and w
Velocity development in the transverse cross sections for y = 0.5 are plotted.
The axial velocity profiles along the rectangular All values of u, u and w on four walls are zero except
duct with injection for different wall Reynolds num- v = Re, on the lower porous wall. The location
bers are presented in Figs. 2(a) and 2(c) for y = 0.5 marked with a positive or negative sign represents the
and 2.0, respectively. The dimensionless velocities point with the highest absolute value of velocity. The
shown by solid lines are on the center plane per- flow is in the hydrodynamically developed region for
pendicular to the porous wall. The fluid injection case (a), near the section of complete mass extraction
induces both mass addition and blowing force from for case (b), and in the region of flow reversal for case
the porous wall ; thus it increases the axial velocity and (c). Between these three cases, the flow patterns of u
shifts its peak away from the porous wall. Due to and w are similar ; however, a significant difference is
more fluid injection, for larger Re,, larger x or smaller observed for u. The shifts of axial velocity peak are in
;‘, the increment of axial velocity is bigger, and the opposite directions for injection and suction flow as
velocity shift is more pronounced. Figures 2(b) and shown in cases (a) and (b), and a pair of reverse flows
2(d) show the cases of suction flow. In contrast, fluid occur on the upper corners of the cross section as
suction induces both mass subtraction and withdrawal shown in case (c).
force from the porous wall ; thus it decreases the axial
velocity and shifts its peak towards the porous wall. Hydrodynamic entrance length, complete mass extrac-
The velocity components shrink along the axial dis- tion andjow reversal
tance until the flow reaches the point of complete mass When the fluid with uniform inlet velocity dis-
extraction or flow reversal. This shrinking of axial tribution flows into the entrance of a one-porous-
velocities and the shift of velocity peak are more pro- wall rectangular duct, the inertia force, friction force,
nounced with stronger suction, larger x or smaller y. injection or suction force and pressure force all inter-
The effect of aspect ratio on the axial velocity devel- act with one another. The velocity, U,,,,,/ir, will be
opment is illustrated in Figs. 2(e) and 2(f) for Re, = 5 constant in the fully developed region if a normalized
 Laminar Row and heat transfer in a rectangular duct 2607

INJECTION SUCTION

FIG. 2. Dimensionless axial and normalized velocity distributions at different axial positions.

Y =os u V W 8
(a) Re,=5
x=0.07

(b) Re,,,=-5
x=0.14

(c) Renp-10
x=0.07

A=0.03 A=l.O A=O.6 A=0.05

FIG. 3. Constant velocities and isotherm maps for y = 0.5 : (a) injection, Re, = 5, x = 0.07, (b) suction,
Re, = -5, ,x = 0.14, (c) suction, Re, = - 10, x = 0.07.
2608 Y.C. CHENG~~~.

@ Entronce Length
51 Complete Moss Critical Point

A Flow Reversal

---ldp
iidx

FIG. 4. Wall Reynolds number effects on normalized axial

pressure gradients along the axial direction.

axial pressure gradient is constant. As shown in Fig. FIG. 5. Entrance lengths, locations of complete mass extrac-
4, the pressure gradient cannot drop to a constant tion and flow reversal under different aspect ratios and suc-
tion conditions.
value for y = 0.5 and Re, = -5, - 10, -20, as well
as y = 2 and Re, = - 10, - 20 ; hence the flow cannot
reach its fully developed region. The hydrodynamic
entrance length, I,,,,, is usually defined as the duct occurs earlier with larger suction rate and lower aspect
length required to achieve a maximum axial velocity ratio. However, flow reversal has never been observed
of 99% of the corresponding fully developed value. numerically for y = 2 and 5 for Re, = 0 to -40. When
Using this definition, the variations of entrance the fluid is extracted slowly from the porous wall, its
lengths for different Re, and y are shown in Fig. 5. mass in a local cross section diminishes gradually
In general, the hydrodynamic entrance length is along the axial direction and finally vanishes at the
decreased by fluid injection (Re, > 0) for y < 2, point of complete mass extraction. But, with a strong
except for y = 5. A duct with lower aspect ratio and suction, the fluid in the cross section cannot fill the
stronger injection means a higher mass input to the vacancies caused by the extraction and thus flow rever-
channel, hence a shorter & can be expected. In most sal is created locally at that specific axial position.
of the cases for suction flow, the hydrodynamically Obviously, flow reversal occurs because no more axial
fully developed region cannot be reached due to the momentum can be withdrawn (u, = 0). Thus the
complete mass extraction or flow reversal. decreasing axial momentum must be balanced by a
It is understandable that the point of complete mass pressure rise. This adverse pressure gradient reverses
extraction, x,, in a suction flow along the rectangular the flow where its velocity is small.
duct can always be achievable. Using the mass The normalized axial pressure gradients,
balance, one has the value x, : - (l/n)(dp/dx), along the axial direction can be illus-
trated by using Fig. 4 with various wall Reynolds
l+Y
(17) numbers for y = 0.5 and 2. For the cases of Re, 2 0,
Xc=2IRe,J. this normalized quantity gradually decreases in the
The values of x, are plotted with solid lines in Fig. 5 developing region, and an asymptotic value is attained
for various y and Re,, and the axial distances of flow in the fully developed region. The fully developed
reversal detected numerically are plotted by dashed value is larger for higher Re, to overcome the flow
lines. As pointed out by Quaile and Levy [20], the resistance induced by injection. In most cases of suc-
laminar flow in a porous tube may be changed to tion flows, the pressure gradient reduces continuously
turbulence with a sufficiently large wall suction. without an asymptotic value along the duct due to
Therefore, their results for complete mass extraction mass subtraction. However, the pressure gradient for
and flow reversal are also included in this figure. Due y = 2 and Re, = - 5 drops gradually to a fully
to the stronger outward suction force, flow reversal developed constant value before the complete mass
 Laminar flow and heat transfer in a rectangular duct 2609

- Injection

fRe

A Flow Reversal

\ - Injection
\
2. \ r Rew=-20 --- c. .r+:h”

A Flow Reversal Al Complete Mass

Extraction

FIG. 6. Variations of friction factors along the axial direction for y = 0.2, 0.5, 2.0 and 5.0 with various
values of Re,.

extraction. The curves drop to negative values for effect of mass reduction. For the case of y = 2, the
y = 0.5 and Re, = - 10 and - 20. As can be seen also values of fRe decrease and then increase after x = 0.04
in Fig. 5, flow reversal occurs at specific axial positions and 0.025 for Re, = - 10 and -20, respectively. The
with these negative pressure gradients. existence of the inversion points is due to the effect of
strong mass extraction. For the cases of y = 0.2 and
Developing and developedfriction factors 0.5, fluid suction induces flow reversal at a specific
The variations of skin friction, f Re, along the axial point before the flow reaches the point of complete
direction in a rectangular duct for both injection and mass extraction for Re, = - 10 and -20, and all the
suction flows with y = 0.2, 0.5, 2.0 and 5.0 are shown fluid can be completely extracted without flow reversal
in Fig. 6. For injection flow, the friction factor for Re, = - 5. The friction factor decreases con-
decreases with the increase in Re, in the inlet region. tinuously for Re, = - 5 and - 10 along the axial
In contrast, it increases with the increase in Re, at direction, but increases slightly after x = 0.01 for
x > 0.003. As indicated by equation (14) the local Re, = -20.
friction factor is affected by both the average axial The fully developed friction factors for injection
velocity and the wall velocity gradient for a specific flow with various aspect ratios are presented in Fig.
aspect ratio. In the inlet region, axial velocity gradi- 7. This figure does not include the cases of suction
ents are almost the same for different Re,, but the flow in which the hydrodynamically developed regions
average axial velocity is larger for larger Re,. As a exist only for weak suction. As the aspect ratio is
result, friction factors are dominated by the average specified, the friction factor is always larger for a
axial velocity. However, when the flow is further stronger injection flow due to a larger axial velocity
developed along the axial direction, the wall velocity gradient. In this figure, Shah and London’s data [18]
gradient dominates the variation of friction factor. for impermeable flow in a rectangular duct are also
Consequently, increasing wall velocity gradients due included for comparison ; the differences between these
to larger wall Reynolds numbers results in higher data and calculated results of this study are small.
values of friction factors.
For suction flow, the friction factor increases with Development of temperature profile
increasing rate of suction in the whole region as shown The effect of Re, on dimensionless temperature pro-
in Fig. 6. However, at large x, the variations in the files at axial distances x+ = 0.001, 0.01, 0.1 and 1 for
friction factor become more complicated due to the both injection and suction flows are shown in Fig. 8.
2610 Y. C. CHENG et al.

__ in the upper part, and the temperature of injection

0 Shah and London [18] flow is always higher than that at the same position
- Present Study of the suction flow.
24 For injection flow, the temperature of the fluid in a
rectangular duct arises by both wall heat flux and fluid
(fRe)fd
injection. Since the upper wall is adiabatic and the
center portion of the duct is cooled by upstream fluid,
the increment of temperature is more pronounced for
the fluid at the lower part of a cross section. For a
12 I I specified aspect ratio, this increment is more sig-
0.0 0.5 1.0 0.5 0.0 nificant due to a larger amount of heat input by the
*r-l/Y+ injected fluid for larger Re, and X. As less fluid is
FIG. 7. Fully developed friction factors of various injection injected for larger y, the thermal resistance along the
flows in the rectangular ducts. J direction is increased. This explains that the tem-
perature difference along the y direction for larger y
is greater than that for small y. It is also noted that
These data are all on the center plane perpendicular the temperature is out of scale in Fig. 8(c) for Re, = 20
to the porous wall. As a common practice, the coor- and is not plotted here. For suction flow, the fluid
dinate, X’ = x/Pr, is used in these figures. Because temperature can arise by heat conduction from the
heat is transferred from the porous wall, the tem- porous wall. As part of the thermal energy is carried
perature in the lower part is always higher than that away by the extracted fluid, the fluid temperature

INJECTION SUCTION
(a>Rew=

(c) ReW=

- 8=(q~k)*~o ---- e*=(~)z=o

FIG. 8. Axial dimensionless and normalized temperature distributions at different axial distances
 Laminar flow and heat transfer in a rectangular duct 2611

1.9
tion cases, but cannot approach an asymptotic value
for the cases of strong suction.
The dimensionless thermal entrance length can be
1.6 expressed as l,, = L,,/(PeD,) and is plotted in Fig. 9.
Its value is determined by observing 8:reaching an
asymptotic value. The entrance length is longer for
1.3 higher aspect ratio in impermeable flow, and the same
result is also obtained in injection flow. Usually the
thermally fully developed region cannot be achieved
1.0 for suction flow, except for the weak suction case.

Developing and developed Nusselt numbers

1.9
The effect of Re, on Nusselt numbers along the
axial direction for Pr = 0.72 withy = 0.2,0.5,2.0 and
1.6
5.0 is given in Fig. 10. As can be seen, the Nusselt
number is reduced by fluid injection and is increased
1.3 by fluid suction. This is similar to the solutions of
Pederson and Kinney [21] for porous tubes as well as
Doughty and Perkins 1221for porous parallel plates.
1 .o
0.001 0.01 0.1 1 10 According to equation (15), Nu is a reciprocal of
x &,,-Qb, and this value for injection flow gradually
POh
approaches an asymptotic value in the fully developed
FIG. 9. Wall Reynolds number effects on normalized tem- region. In the case of suction flow, the temperature
peratures on the upper wall along the axial direction.
difference of heated wall and bulk fluid is reduced by
fluid extraction, and hence the Nusselt number is
increases slowly along the axial distance, and tem- larger than that of injection flow at the same axial
perature is higher for lower suction rate. Due to larger position. The results show that a smaller aspect ratio
heat conduction area in a duct with smaller y, the fluid induces higher values of Nusselt number for both
temperature is higher. cases of injection and suction flow. The reason is that
The effect of aspect ratio on the temperature devel- the fluid mixes well in the duct with small y, as can be
opment is illustrated in Figs. 8(e) and 8(f) for Re, = 5 seen in Figs. 8(e) and (f).
and -5, respectively. Since the bottom porous wall Fully developed Nusselt numbers for injection flow
is longer, more heat is conducted and more fluid is with various aspect ratios are presented in Fig. 11.
extracted for small y ; then the fluid temperature This figure does not include the cases of suction flow,
increases quickly for y = 0.2, but slowly for y = 5. It because the developed regions only exist for weak
is also observed that, due to greater thermal resistance, suction. For a fixed aspect ratio, the Nusselt number
the temperature of the porous wall is higher for larger is always smaller with stronger injection. For large
aspect ratio in injection flow. injection such as Re, = 20, Nu is very small ; hence
the thermal energy input to the flow is mainly from
Thermal entrance length the injected fluid. In this figure, Schmidt and Newell’s
To determine the characteristics of the thermal data [19] for impermeable flow in a rectangular duct
entrance length more closely, a normalized tem- are also included for comparison ; the differences
perature Q* defined. 8* = (T- rW)/(Tb- i;J is used. between these data and calculated results of this study
The development of this normalized temperature pro- are small.
file is plotted by dashed lines in Fig. 8. For most cases,
the temperature profiles are not only similar to each Comparison with existing data for case y -+ 0
other but also invariant as the flow moves along the Among limited data available in the literature,
duct. Subsequently, the flow can be considered as Doughty and Perkins’ results [22, 231 can be selected
thermally fully developed. However, the temperature for comparison with lhe present data. Their results are
similarity and invariance cannot be found for the cases for hydrodynamically and simultaneously developing
of high aspect ratio and strong suction. In order to laminar flows between parallel porous plates subjected
identify the variation of temperature along the axial to constant heat flux. In the present numerical scheme,
distance, a specific local temperature has to be chosen. y = 0.001 and symmetric porous walls are taken to
Because the heat is input from the bottom porous obtain the comparable data. Figure 12(a) shows that
wall, the temperature of the upper wall is the slowest the results of the friction factor are in good agreement
to rise and the least sensitive. The variations of average with each other, even in the inlet region of the channel.
upper wall temperature, v, along the duct for y = 0.5 Moreover, near the location of flow reversal caused
and 2 is shown in Fig. 9. It is clearly noted that the by strong suction for Re, = - 20, the friction factors
normalized fluid temperature increases to an asymp- are also very close. In addition, the present Nusselt
totic constant at a specific axial location for all injec- numbers for y = 0.001 along the channel are plotted
2612 Y. C. CHENGet al

Nu

A Flow Reversal

Nu

X
p%iDh
FIG. 10. Variations of Nusselt numbers along the axial direction for Pr = 0.72, y = 0.2, 0.5, 2.0 and 5.0
with various values of Re,.

8 ,
(a) 56h I

0 Schmidt and Newell [lg] o Doughty and Perkins[22], Y=O

- Present Study, y=O.OOl
- Present Study

N”fd fRe

32

6.001 0.01 0.1

X
PeODh
FIG. I 1. Fully developed Nusselt numbers of various injec-
tion flows in the rectangular ducts. @I32

24
in Fig. 12(b) for comparison with the existing data
[23]. The present results fit Doughty and Perkins’ data
Nu 16
very well except those near the inlet region. However,
the present results for Re, = 0 also fit Hwang and
Fan’s data [24] well along the entire axial axis. 6

CONCLUSIONS
0:001 0.01 0.1
(1) The axial pressure drop in a rectangular duct is X
increased for the injection flow and decreased for the P%Dh
suction flow as compared with the impermeable flow. FIG. 12. Comparison of calculated results with the existing
However, the pressure recovery appears downstream data for (a) fRe, (b) Nu.
 Laminar flow and heat transfer in a rectangular duct 2613

for suction flow with small aspect ratio due to flow national Symposium on Transport Phenomena in Thermal
Engineering, Vol. 1, pp. 685-690 (1993).
reversal, and a change to turbulent flow may be 9,
K. A. Alkasab and C. Y. Lu, Transient effects of chan-
expected. ging the electric load on the performance of phosphoric
(2) It is shown that fluid injection increases the acid fuel cell power plant, ht. J. Energy Systems 3,9%17
friction factor and reduces the hydrodynamic entrance (1985).
10. N. E. Vanderbough, J. R. Huff and J. Hedstrom, Heat
length. However, a fully developed region cannot
and mass transfer in PEM fuel cells, Proceedings of24th
always be achieved in suction flow due to mass extrac- IECEC, Vol. 3, pp. 1637-1640 (1989).
tion and flow reversal. 11. T. Okada, H. Ide, M. Tanaka, S. Narita and J. Ohtsuki,
(3) It has been verified that the wall heat transfer Study of temperature control in indirect internal reform-
inc. MCFC stack. Proceedings of 25th IECEC, Vol. 3.
decreases in injection flow and increases in suction
p< 207-212 (1990). ”
flow. Also, the thermal entrance length is increased 12. P. L. Donoughe, Analysis of laminar incompressible
with the fluid injection. flow in semiporous channels, TN-3759, NACA (1956).
(4) For the effect of aspect ratio, Nusselt number 13. E. R. G. Eckert, P. L. Donoughe and B. J. Moore,
Velocity and friction characteristics of laminar viscous
is decreased with the increase in the aspect ratio due
boundary-layer and channel flow over surfaces with ejec-
to larger thermal resistance through the cross section, tion and suction, TN-4102, NACA (1957).
and the flow finds it difficult to reach its thermally 14. F. M. White, Laminar flow in porous ducts, PhD Thesis,
fully developed region in the same situation. Georgia Institute of Technology, Atlanta, Georgia
(1959).
15. M. M. Sorour, M. A. Hassab and S. Estafanous,
Acknowledgements-This study is suDported by the Energy Developing laminar flow in a semiporous two-dimen-
Commissi& of the Ministry ofEcon&c Affaiis of Taiwan, sional channel with nonuniform transpiration, Int. J.
Republic of China through Fuel Cell Project No. 38K2100. Heat Fluid Flow lo,4454 (1987).
The authors also acknowledge the assistance of the Energy 16. F. C. Chou and G. J. Hwang, Vorticity-velocity method
and Resources Laboratories, ITRI for the use of a computer. for the Graetz problem and the effect of natural con-
vection in a horiiontal rectangular channel with uniform
wall heat flux. J. Heat Transfer 109.704-710 (1987).
REFERENCES 17. R. M. Curr, 6. Sharma and-D. G. Tanchell, I?um&ical
prediction of some three-dimensional boundary layers
1. W. M. Massey, Jr., Heat and mass transfer in semi- in ducts, Comput. Meth. Appl. Mech. Engng 1, 143-158
porous channels, PhD Thesis, Georgia Institute of Tech- (1972).
nology, Atlanta, Georgia (1969). - 18. R. A. Shah and A. L. London, Laminar Flow Forced
2. S. J. Rhee and D. K. Edwards. Laminar entrance flow Convection in Ducts, Suppl. 1 to Advanced Heat Transfer,
in a flat plate duct with asymmetric suction and heating, pp. 196222, Academicl%ess, New York (1978). _
Numer. Heat Transfer 4,85-100 (1981). 19. F. M. Schmidt and M. E. Newell. Heat transfer in fullv
3. M. M. Sorour and M. A. Hassab, Effect of sucking the developed laminar flow through’rectangular and iso;-
hot fluid film on the performance of flat plate solar celes triangular ducts, Int. J. Heat Mass Transfer 10,
energy collectors, Appl. Energy 14, 161-173 (1983). 1121-l 128 (1967).
4. J. Jorne, Mass transfer in laminar flow channel with 20. J. P. Quaile and E. K. Levy, Laminar flow in a porous
porous wall, J. Electrochem. Sot. 129,172771733 (1982). tube with suction, J. Heat Transfer 97, 66-71 (1975).
5. P. Lessner and J. Newman, Hydrodynamics and mass 21. R. J. Pederson and R. B. Kinney, Entrance-region heat
transfer in a porous-wall channel, J. Electrochem. Sot. transfer for laminar flow in porous tubes, Int. J. Hear
131,1828-1831 (1984). Mass Transfer 14, 159-161 (1971).
6. C. Y. Soong and G. J. Hwang, Laminar mixed con- 22. J. R. Doughty and H. C. Perkins, Jr., Hydrodynamic
vection in a radially rotating semiporous channel, Int. J. entry problems for laminar flow between parallel porous
Hear Mass Transfer 33, 1805- 18 16 ( 1990) plates, J. Appl. Mech. 92, 548-550 (1970).
I. G. J. Hwang, Y. C. Cheng and M. L. Ng, Friction 23. J. R. Doughty and H. C. Perkins, Jr., Thermal and
factors and heat transfer in a square duct with one- combined entry problems for laminar flow between par-
walled injection and suction, ht. J. Heat Mass Transfer allel porous plates, J. Heat Transfer 94,233-234 (1972).
36,2429-2440 ( 1993). 24. C. L. Hwang and L. T. Fan, Finite difference analysis of
8. G. J. Hwang, Y. C. Cheng and M. L. Ng, Friction factor forced convection heat transfer in entrance region of
and heat transfer correlations for gaseous reactant flow a flat rectangular duct, Appl. Sci. Res. 13A, 401422
in fuel cell power modules, Proceedings of 6th Znrer- (1964).