You are on page 1of 9

Conjugated heat transfer in unsteady channel ows

Amilcare Pozzi, Renato Tognaccini

Dipartimento Ingegneria Aerospaziale, Universit di Napoli Federico II, Piazzale V. Tecchio 80, 80125 Napoli, Italy
a r t i c l e i n f o
Article history:
Received 16 October 2009
Accepted 8 April 2011
Available online 18 May 2011
Keywords:
Conjugated heat transfer
Thermo-Fluid Dynamics
Analytical methods
a b s t r a c t
The exact analytical solution of the unsteady impulsive Thermo-Fluid Dynamic eld arising in a two-
dimensional channel with thick solid walls is presented when the thermal eld in the uid is coupled
with the thermal eld in the solid (conjugated heat transfer). The cases studied in this paper depend
on the boundary conditions imposed on the unwetted sides of the channel walls: assigned temperature
and adiabatic condition. Moreover the case of a given heat loss at the unwetted wall is also considered in
an appendix. The temperature and heat ux at the soliduid interface are analyzed as function of time
and of the nondimensional parameters governing the problem.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
Analytical solutions of exact problems of physical relevance are
very useful, even if they, obviously, are limited to simple geome-
tries. Indeed, they clearly identify the role of the physical parame-
ters governing the phenomenon. In addition, they are fundamental
test cases for the verication and validation of numerical methods
required for the analysis of geometrically complex problems.
In Fluid Dynamics, the renewed interest in analytical solutions
is documented by the recent book of Drazin and Riley [1] collecting
the exact analytical solutions of the NavierStokes equations.
Concerning Heat Transfer in Fluid Dynamics, there is a lack of
exact solutions of the thermal eld when the Eckert number is
taken into account due to the mathematical difculties introduced
by the non homogeneous term in the energy equation, see Pop and
Ingham [2] for a recent book with focus on analytical models.
Nonetheless, recent advances can be found, among others, in [3]
(FalknerSkan ows), [4], [5] (unsteady pipe ows), [6] (impulsive
Rayleigh ow).
Another topic with limited analytical solutions is the analysis of
conjugated heat transfer effects. This term was introduced by
Perelman [7] to identify problems in which both temperature
and heat ux at a soliduid interface are unknown and are deter-
mined by simultaneous and coupled solutions of the Thermo-Fluid
Dynamic equations in the uid and the energy equation in the
solid.
Conjugated phenomena are usually neglected in Fluid
Dynamics: either the temperature or the heat ux are assigned as
boundary conditions. These effects are, however, relevant in many
applications such as aerospace and cooling technologies.
The lack of exact analytical solutions of the Thermo-Fluid
Dynamic eld and the difculty in the solution of the coupled
problem itself implies that conjugated problems are usually stud-
ied by numerical or approximate methods [8,9]. Just in the recent
years exact analytical solutions describing conjugated effects have
been published. Pozzi and Tognaccini [10] found the solution for
the conjugated heat transfer in the case of an impulsively acceler-
ated ow from rest to a constant speed over an innite plate of
nite thickness in the case of imposed temperature and of adia-
batic condition on the unwetted side of the plate. Pozzi et al.
[11] showed that, in the case of plate of innite thickness the solu-
tions in both the uid and in the solid are self-similar with very
simple analytical expressions.
Weigand and Gassner [12] studied the conjugate extended
Graetz problem which takes into account the effects of the Peclet
number. The same problem, but including a model of a simulated
ame has been very recently analysed, again by analytical meth-
ods, by Veeraragavan and Cadou [13]. A key role in conjugated ther-
mal effects is played by the thermal activity ratio dened as the ratio
between the thermal effusivities in the uid and in the solid.
Pozzi and Tognaccini [14] have recently added another family to
the short list of exact analytical solutions of the NavierStokes
equations of practical relevance. Indeed they presented the
solution of the completely developed unsteady ow in a two-
dimensional channel when an arbitrary time varying pressure gra-
dient is imposed (unsteady Poiseuille ow). The solution has been
straightforward extended to the case of a moving wall with arbi-
trary time dependent velocity (unsteady Couette ow). In addition,
for vanishing thickness of the wall (i.e. without conjugated effects)
the temperature eld has also been analytically computed in the
case of assigned time varying temperature at the walls for an arbi-
trary Prandtl number in the case of Eckert number equal to zero and
for the relevant case of Prandtl number equal to one in the case of
0017-9310/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2011.04.019

Corresponding author.
E-mail address: renato.tognaccini@unina.it (R. Tognaccini).
International Journal of Heat and Mass Transfer 54 (2011) 40194027
Contents lists available at ScienceDirect
International Journal of Heat and Mass Transfer
j our nal homepage: www. el sevi er . com/ l ocat e/ i j hmt
arbitrary Eckert number (whose effects are usually neglected in
literature).
With the help of these results, we now present a further step in
the analysis of conjugated heat transfer. Indeed these effects are
analytically investigated here when the unsteady channel ow
with assigned impulsive pressure gradient is developing between
walls of nite thickness and the thermal boundary conditions are
assigned on the unwetted side of the channel. The physical prob-
lems in the uid and in the solid are coupled enabling the continu-
ity of the temperature and of the heat ux at the soliduid
interface.
It will be given the analytical solution of the exact problem with
the Fluid Dynamic eld governed by the NavierStokes equations
and non homogeneous heat equation taking into account for the
effects of the Eckert number and the temperature eld in the solid
governed by the homogeneous heat equation.
The main emphasis will be addressed to the analysis of the time
varying temperature and heat ux at the soliduid interface,
which are the result of the introduction of the conjugated effects
in the study. The solution is given in a series representation, which
is very useful to obtain results with any prescribed accuracy.
Finally in Appendix D is considered the case in which the unw-
etted solid wall is not adiabatic, but a known heat loss is imposed
as proposed in [13].
2. The physical problem
We consider an innite two-dimensional channel (on both
sides) with half-section length d. Both solid walls of the channel
have thickness b. We shall consider problems symmetrical with
respect to the channel centerline, see Fig. 1. At time t = 0 the uid
is impulsively accelerated by imposing a constant pressure gradi-
ent dp/dx. The boundary conditions for the thermal eld are im-
posed on the unwetted side of the thick walls and are constant
along the channel walls. Both the velocity and the temperature
eld are completely developed, therefore the solution only de-
pends on time and on the spatial coordinate orthogonal to the wall.
We assume that an incompressible, laminar ow with constant
properties (kinematic viscosity m and thermal conductivity k) arises
in the channel. In this case the dynamic eld is not coupled with
the temperature one and the momentum equation is linear since
the ow is parallel.
The energy equations in the uid and in the solid assume the
following forms:
@h
@s

1
Pr
@
2
h
@g
2
E
@u
@g

2
; 1a
@h

@s
t
fs
@
2
h

@Y
2
; 1b
where g is the nondimensional spatial coordinate in the uid (refer-
enced to d) and origin placed on the channel centerline, Y is the spa-
tial coordinate in the solid lower wall (referenced to b) with origin
placed on the unwetted side of the plate (Y = 1 corresponds to
g = 1), s is the non dimensional time referenced to t
f
= d
2
/m,
h = (T
f
T
0
)/T
0
is the temperature in the uid (T
0
is the initial tem-
perature, h

= (T
s
T
0
)/T
0
is the temperature in the solid and u is the
axial uid velocity referenced to V
ref
= (dp/dx) d
2
/(2l), the usual
reference velocity in a Poiseuille ow (l is the dynamic viscosity).
Pr = lc
p
/k is the Prandtl number and E V
2
ref
=c
p
T
0
is the Eckert
number (c
p
is the specic heat at constant pressure in the uid).
t
fs
= t
f
/t
s
is the ratio between the reference time in the uid and in
the solid t
s
= b
2
/a
s
(a
s
is the thermal diffusivity in the solid).
The analytical solution of the velocity eld has been given in
[14]:
uq; s 2s u
h
q; s; 2
where
u
h
q; s hq; s

1
k1
1
k
hk q; s hk q; s; 3a
ha; s 2s 2a
2
erfc
a

s
p

p
p a

s
p
e
a
2
=s
; 3b
q = (g + 1)/2 and erfc(z) species the complementary error function.
The coupling between the thermal eld in the uid and in the
solid is obtained by imposing the continuity of the temperature
and of the heat ux at the soliduid interface in this way:
hs; 1 h

s; 1 h
w
s; 4a
K

Pr
p
t
fs
p
@h
@g
s; 1
@h

@Y
s; 1; 4b
where K b=dk=k
s

t
fs
p
Pr
p
is the thermal activity ratio we shall
discuss later (k
s
is the thermal conductivity in the solid).
The initial conditions for the temperature elds are
h0; g 0; h

0; Y 0; 5
at the initial time the temperature in the uid and in the solid is
constant. Three cases are studied here, depending on the condition
imposed on the unwetted sides of the channel walls:
(a) isothermal case, the temperature is kept at a constant value
h

s; 0 h

e
;
(b) adiabatic case, the heat ux @h

/@Y(s, 0) = 0;
(c) assigned heat loss, see Appendix D.
It is worth to note that in this case of completely developed
temperature eld, the solution we are looking for solves the com-
plete energy equation including axial conduction and convection.
3. The transformed solution
We propose here the solution of the coupled Eq. (1), including
the effects of the dissipation of kinetic energy in internal energy,
in the relevant case of Pr = 1. The Laplace transform with respect
to time of these equations provides simple ordinary linear differen-
tial equations. However both solutions will depend on the un-
known interface temperature that should be obtained by
imposing the coupling condition Eq. (4). Fig. 1. Sketch of the channel with thick walls.
4020 A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027
3.1. The solution in the uid
Denoting with Hs; g L
s
hs; g the Laplace transform of the
uid temperature with respect to time, Eq. (1a) with initial condi-
tion (5) reduces to
@
2
H
@g
2
sH EL
s
@u
@g

2

: 6
The solution is
Hs; g H
w
s
4E
s
3

cosh

s
p
g
cosh

s
p H
p
s; g; 7
where H
w
s L
s
h
w
s and H
p
s; g L
s
h
p
s; g, with h
p
a partic-
ular integral of Eq. (1a). As shown in [14] a particular integral is
h
p
s; g
E
2
u
h
s; g
2
; 8
where u
h
(s, g) is given by Eq. (3).
The calculation of the interface temperature will only require
the knowledge of H
p
at the soliduid interface where u = 0 and
the Laplace transform of h
p
is simple.
3.2. The solution in the solid
The Laplace transform of Eq. (1b) with initial condition (5) is
@
2
H

@Y
2

s
t
fs
H

0; 9
where H

is the Laplace transform of the temperature in the solid.


The solution depends on the assigned boundary condition imposed
on the unwetted side of the channel wall.
Isothermal case
H

s; Y H
w
s
sinhrY
sinhr

h

e
s
sinhr1 Y
sinhr
; 10
where r

s=t
fs

.
Adiabatic case
H

s; Y H
w
s
coshrY
coshr
: 11
3.3. Coupling condition
Both transformed solutions in the uid and in the solid are un-
known since depend on the unknown transformed interface tem-
perature, which can be obtained by coupling the solutions in the
uid and in the solid. In the previous equations the continuity of
the temperature across the soliduid interface has been implicitly
taken into account by setting H

w
H
w
. In transformed variables
the continuity of the heat ux is:
K

t
fs
p
@H
@g

w
s
@H

@Y

w
s: 12
The temperature at the soliduid interface can be obtained substi-
tuting the spatial derivatives of equations (7), (10) or (11) in Eq.
(12). We have:
Isothermal case
H
w
2EK
H
E1
s
3

H
E2
s
5=2

h

e
H
ee
s
; 13
where
H
E1

tanh

s
p
D
a
; H
E2

1
cosh
2

s
p
1
D
a
; H
ee

1
cosh
2
r
1
D
a
; 14
with D
a
cothr Ktanh

s
p
.
Adiabatic case
H
w
2EK
H
Eb1
s
3

H
Eb2
s
5=2

; 15
where
H
Eb1

tanh

s
p
D
b
; H
Eb2

1
cosh
2

s
p
1
D
b
; 16
with D
b
tanhr Ktanh

s
p
.
The transformed heat ux at the soliduid interface are
obtained differentiating Eqs. (10) and (11).
Isothermal case
@H

@Y

w
s r H
w
s cothr
1
sinhr
h

e
s

: 17
Adiabatic case
@H

@Y

w
s rH
w
s tanhr: 18
4. Initial and asymptotic behavior
The analysis of the initial (s ?0) and asymptotic (s ?1) tem-
peratures and heat ux at the soliduid interface can be per-
formed without computing the inverse Laplace transforms by
means of the Abelian and Tauberian theorems, see Appendix A. In-
deed, the Abelian theorem allows for the analysis of the solution in
the physical space for s ?0 by studying the transformed solution
for s ?1. On the contrary, the Tauberian theorem provides the
physical solution for s ?1 using the transformed solution for
s ?0.
4.1. The isothermal case
The initial behavior of the interface temperature is obtained by
evaluating the leading term of Eq. (13) for s ?1 and then per-
forming the inverse transform according to the Abelian theorem
s !1: H
w
s % 2E
K
1 K
1
s
3
; 19
s !0 : h
w
s % E
K
1 K
s
2
: 20
In this case, ow driven by an imposed impulsive pressure gradient,
the temperature at the soliduid interface is continuous and grows
as s
2
.
Similarly, analyzing Eq. (17), the initial behavior of the interface
heat ux is
s !1:
@H

@Y

w
%
2EK

t
fs
p
1 K
1
s
5=2
; 21
s !0 :
@h

@Y

w
%
8EK
3

p
p
t
fs
p
1 K
s
3=2
: 22
A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027 4021
At the initial time the heat ux is also continuous. Noteworthy, for
small time values, due to the dissipation of kinetic energy, the uid
is heating the walls of the channel, even if h

e
> 0.
The asymptotic behavior of the interface temperature is ob-
tained determining the leading term of Eq. (13) for s ?0:
s !0 : H
w
s % h
wa1
1
s

k
1
k
1
s 1

; 23
s !1: h
w
s % h
wa1
1 e
s=k
1

; 24
where h
wa1

4EK
3

t
fs
p h

e
and k
1
K=

t
fs
p
1=3t
fs
. Therefore, for
large time values, the interface temperature exponentially tends to-
wards a constant value h
wa1
, higher than the value imposed on the
external wall.
The asymptotic behavior of the interface heat ux is
s !0 :
@H

@Y

w
% h
wa1
1
s

k
1
k
1
s 1

e
s
; 25
s !1:
@h

@Y

w
% h
wa1
1 e
s=k
1

h

e
: 26
For large time values the heat ux becomes constant. Due to the
conjugated effects, the uid will heat the channel walls whatever
the value of the imposed temperature h

e
.
4.2. The adiabatic case
The analysis by the Abelian theorem shows that in the adiabatic
case for s ?0 we have the same behavior obtained in the isother-
mal case, Eqs. (19) and (21); i.e. for small time values the interface
temperature does not depend on the condition imposed on the
unwetted side of the plate. This result does not surprise, because
in the solid, near the interface, as s ?0 (small time scale), the
unwetted side of the plate looks innitely far.
The asymptotic behavior of the interface temperature is
s !0 : H
w
s %
4
3
E
K
K1=

t
fs
p
1
s
2
; 27
s !1: h
w
s %
4
3
E
K
K1=

t
fs
p s: 28
In the case of adiabatic external wall of the channel, due to the dis-
sipation of kinetic energy, for large time values the temperature at
the soliduid interface (and therefore in the uid and in the solid
too) indenitely and linearly grows.
The asymptotic behavior of the interface heat ux is
s !0 :
@H

@Y

w
%
4
3
E
K
t
fs
K

t
fs
p
1
s
; 29
s !1:
@h

@Y

w
%
4
3
E
K
t
fs
K

t
fs
p : 30
For large time values the heat ux asymptotically tends to a con-
stant value.
5. The temperature at the soliduid interface
The main mathematical problem is the knowledge of the in-
verse Laplace transform of the temperature at the soliduid inter-
face. Indeed the analytical expressions in Eqs. (13) and (15) do not
have a direct inverse Laplace transform. However, the use of the
binomial series
1 x
1

1
h0
1
h
x
h
; 1 < x < 1 31
allows for a series representation of the interface temperature in
which each term has a simple analytical inverse transform.
5.1. The isothermal case
Specifying with
x e
2

s
p
2r
be
2

s
p
e
2r
; 32
where b = (1 K)/(1 + K), D
a
(see Eq. (14)) can be written as
D
a

1 K
4cosh

s
p
sinhr
e

s
p
r
1 x 33
and
H
E1

1
1 K
1 1 be
2

s
p
r
1 x
1

: 34
In order to substitute the binomial series, Eq. (31), in Eq. (34) it is
necessary that jxj < 1. This relation is satised for s )0. For small
values of s, jxj < 1 for s < s
0
(K, t
fs
); small s corresponds to large s,
therefore in this case the solution can be determined by the asymp-
totic formulae. When jxj < 1, it is
H
E1

1
1 K
1 1 be
2

s
p
r

1
h0
1
h
x
h

: 35
The nal series representation of H
E1
is obtained by applying an
extension of the Newton formula, Eq. (B.2):
H
E1

1
1 K
1 1 b

1
h0
1
h

h
r0
b
r
h
r

r
j0
r
j

s
p
b
h
2v
e

s
p
b
h
2

;
where b
h
= 2[v(1 j) + h + j r] and v 1=

t
fs
p
.
With a similar procedure, assuming
x 1 be
2

s
p
1 be
2

s
p
2r
be
2r
be
4

s
p
e
4

s
p
2r
37
and taking into account for Eq. (B.4) it is possible to obtain:
H
E2

4
1 K

1
h0
1
h

h
i0
h
i

i
j0
b
ij
1 b
j
i
j

ij
k0
i j
k

j
r0
j
r

e

s
p
e
h
e

s
p
e
h
2v

;
38
where e
h
= 2[v(h i + k + j r) + 2h j 2k + 1]. Finally, by Eqs.
(32) and (B.2), it is:
H
e

2
1 K

1
h0
1
h

h
r0
b
r
h
r

r
j0
e

s
p
b
h
v
e

s
p
b
h
v2

;
39
The inverse Laplace transform of each term in these series can be
now analytically performed. We have
h
w
2EKh
E1
h
E2
h

e
h
ee
; 40
where
h
E1
L
1
s
H
E1
s
3

; h
E2
L
1
s
H
E2
s
5=2

; h
ee
L
1
s
H
ee
s

: 41
With the help of the inverse transforms reported in Appendix C, we
obtain
4022 A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027
L
1
s
H
E1
s
3

1
1 K
s
2
2
1 b


1
h0
1
h

h
r0
b
r
h
r

r
j0
r
j

/s; b
h
2v /s; b
h
2

;
42
L
1
s
H
E2
s
5=2

4
1 K

1
h0
1
h

h
i0
h
i

i
j0
b
ij
1 b
j
i
j

ij
s0
i j
s

j
r0
j
r

ws; e
h
ws; e
h
2v;
43
L
1
s
H
e
s

2
1 K

1
h0
1
h

h
r0
b
r
h
r

r
j0
erfc
b
h
v
2

s
p

erfc
b
h
v 2
2

s
p

;
44
where /(s, a) and w(s, a) are reported in Appendix C.
5.2. The adiabatic case
The inverse Laplace transform of Eq. (15) is obtained by means
of a similar procedure. In particular choosing
x e
2

s
p
2r
be
2

s
p
e
2r
45
and
x be
4

s
p
e
2r

2
1 K
e
2

s
p
1 e
2r
e
4

s
p
2r
; 46
with the help of Eqs. (32) and (45) it is:
H
Eb1

1
1 K
1 1 b

1
h0

h
r0
b
r
h
r


r
j0
1
2hj
r
j

e

s
p
b
h
2v
e

s
p
b
h
2

;
47
H
Eb2

4
1 K

1
h0

h
i0
h
i

i
j0
2
1 K

ij
b
j
i
j

ij
k0
i j
k

j
r0
1
2hikr
j
r

e

s
p
c
h
2u
e

s
p
c
h

;
48
where c
h
= 2[u(r + k + h i) + j i + 2h 2r + 1].
In this case the inverse transform of Eq. (15) is:
h
w
2EKh
Eb1
h
Eb2
; 49
where
h
Eb1
L
1
s
H
E1
s
3

; h
Eb2
L
1
s
H
E2
s
5=2

: 50
With the help of Appendix C these inverse transforms are:
L
1
s
H
Eb1
s
3

1
1 K
s
2
2
1 b


1
h0

h
r0
b
r
h
r

r
j0
1
2hj
r
j

/s; b
h
2v /s; b
h
2

;
51
L
1
s
H
Eb2
s
5=2

4
1 K

1
h0

h
i0
h
i

i
j0
2
1 K

ij
b
j
i
j

ij
k0
i j
k

j
r0
1
2hikr
j
r

ws; c
h
2v ws; c
h
:
52
6. The heat ux at the soliduid interface
6.1. The isothermal case
From Eq. (17), the heat ux is
@H

@Y

w

2EK

t
fs
p cothr
H
E1
s
5=2
cothr
H
E2
s
2

t
fs
p
cothrH
e

s
p

t
fs
p
1

s
p
sinhr
: 53
The inverse transform of the last term in this equation can be ob-
tained due to the property of the Jacobi h
3
function [15] reported
in Appendix C, Eq. (C.8) applied for q = 1/2.
In addition, taking again into account for the properties of the
binomial series we have:
cothr 1 2

1
p0
e
2v

s
p
p1
; 54
therefore:
@H

@Y

w

2EK

t
fs
p
H
E1
s
5=2

H
E11
s
5=2

H
E2
s
2

H
E21
s
2

t
fs
p
H
e

s
p
H
e1

s
p

t
fs
p
1

s
p
sinhr
; 55
where
H
E11
2

1
p0
e
2v

s
p
p1
H
E1
; 56a
H
E21
2

1
p0
e
2v

s
p
p1
H
E2
; 56b
H
e1
2

1
p0
e
2v

s
p
p1
H
e
: 56c
Using the series expressions of H
E1
, H
E2
and H
e
, the physical heat
ux (@h/@Y)
w
can be now straightforward obtained with the help of
the inverse transforms given in Appendix C.
6.2. The adiabatic case
Eq. (15) gives
@H

@Y

w
2
EK

t
fs
p tanhr
H
Eb1
s
5=2
tanhr
H
Eb2
s
2

: 57
Again, since
tanhr 1 2

1
p0
1
p
e
2v

s
p
p1
; 58
we obtain
@H

@Y

w
2
EK

t
fs
p
H
Eb1
s
5=2

H
Eb11
s
5=2

H
Eb2
s
2

H
Eb21
s
2

; 59
A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027 4023
where
H
Eb11
2

1
p0
1
p
e
2v

s
p
p1
H
Eb1
; 60a
H
Eb21
2

1
p0
1
p
e
2v

s
p
p1
H
Eb2
60b
and the physical heat ux at the soliduid interface can be ob-
tained as in the isothermal case with the help of Appendix C.
7. The solution in the uid and in the solid
Once the temperature at the soliduid interface is known, the
solutions in the uid and in the solid are uncoupled. The energy
equations can be independently solved with the interface temper-
ature given as boundary condition for both elds.
Taking into account for the convolution theorem and for a prop-
erty of the Jacobi h
2
function, see Eq. (C.6), the solution in the uid
is
hs; g
1
2

s
0
h
w
s s
@h
2
@q
q; sd s E

s
0
s s
2
@h
2
@q
q; sd s h
p
s; g; 61
where q = (g + 1)/2 and h
w
is given by Eq. (40) for the isothermal
case and by Eq. (49) in the adiabatic case. Pozzi and Tognaccini
[14] discussed in detail the solution of the dynamic and thermal
elds arising in an unsteady channel ow with prescribed temper-
ature at the wall. They also give the explicit analytical expression of
the integrals in Eq. (61) in the case of power law for the wall
temperature.
The solution in the solid depends on the assigned boundary
condition on the unwetted side of the channel. This is a standard
problem of thermal conduction in a solid strip widely discussed
in literature, see [16] for instance.
Isothermal case
Taking into account for a property of the Jacobi h
3
function, see
Appendix C, the inverse transform of Eq. (10) is
h

s; Y
t
fs
2

s
0
h
w
s s
@h
3
@q
1
q
1
; t
fs
sd s
t
fs
h

e
2

s
0
@h
3
@q
2
q
2
; t
fs
sd s; 62
where q
1
= (Y + 1)/2, q
2
= (2 Y)/2 and h
w
is given by Eq. (40).
Adiabatic case
In the same way it is
h

s; Y
t
fs
2

s
0
h
w
s s
@h
2
@q
1
q; t
fs
sd s; 63
where q = (Y + 1)/2 and h
w
is given by Eq. (49).
8. Analysis of the results
The temperature and the heat ux at the soliduid interface
are respectively plotted in Figs. 2(a) and 2(b) for the isothermal
case. In the gures there is a zoom of the pictures for small s val-
ues. The exact solutions (20 terms of the series have been com-
puted) are compared with the initial and asymptotic behaviors.
In the plotted case h

e
> 0 the evolution of the temperatures and
heat ux is characterized by four main stages.
(1) For s ?0 there is a local behavior independent of the
boundary condition imposed on the external wall; in this
stage the uid is heating the solid wall due to the dissipation
of kinetic energy, and the interface temperature grows as s
2
,
such as predicted by the local analysis.
(2) As time grows, the effects of the imposed h

e
appear, the
sign of the heat ux can change and the solid wall is heating
the uid, but the temperature at the interface is still
growing.
(3) When h
w
becomes greater than h

e
, the main role is again
taken by the dissipation of kinetic energy, the uid is again
heating the solid.
(4) For s ?1 there is the asymptotic stage, the interface tem-
perature reaches the asymptotic value h
w
> h

e
and the uid
is heating the solid wall at a constant rate given by the
asymptotic solution, Eq. (25).
This asymptotic solution also gives the possibility to quantify
the time required, in practice, to reach the asymptotic state. For in-
stance the time required to obtain h
w
/h
wa1
= 0.99 is
Fig. 2. Solution at the soliduid interface versus time, isothermal case, (a):
temperature, (b): heat ux, h

e
1; E 1; K 1; t
fs
0:5, : exact solution; - --:
initial solution; - -: asymptotic solution.
4024 A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027
s
99%
K=

t
fs

1=3t
fs
log
1
0:01

: 64
The nondimensional heat ux plotted in Fig. 2(b) is essentially
the Nusselt number, given in this case by
Nu
2d

2

K
T
0
T
e
T
0

@h

@Y

w
; 65
where the channel height 2d has been chosen as reference length. In
addition, the nondimensional temperature does not explicitly de-
pends on T
e
if it is referred to (T
e
T
0
) instead of T
0
.
In the Figs. 3(a) and 3(b) the interface temperature and heat ux
are respectively plotted for the adiabatic case (20 terms of the ser-
ies have been used for the representation of the exact solution).
For small time values the local stage can be recognized, the
temperature follows the same behavior (s
2
) of the isothermal case,
such as the heat ux (s
3/2
). Following a properly transient stage, in
the asymptotic evolution, the temperature linearly grows, while
the heat ux reaches a constant value given by Eq. (29).
As in the case of Rayleigh ow over a thick plate [10], K is the
main additional parameter ruling the conjugated effects. K is the
thermal activity ratio:
K

qc
p
k

q
s
c
p
s
k
s
; 66
given by the ratio between the thermal effusivities in the uid and
in the solid. It is the same parameter ruling the interface tempera-
ture of two semi-innite solids at different initial temperatures. It is
interesting to note that it can vary of many orders of magnitudes
changing the solid or the uid, see [11] for instance.
In this case of adiabatic wall, the adoption of V
2
ref
=c
p
as reference
temperature leads to an expression of the solution independent of
the Eckert number.
Figs. 4 and 5 show the interface temperature versus time
respectively for different values of E and K. In each gure both
the isothermal and adiabatic cases are plotted. It is interesting to
note that a similar behavior of the temperature can be obtained
by varying E or K.
Fig. 3. Solution at the soliduid interface versus time, adiabatic case, (a):
temperature, (b): heat ux, E = 1, K= 1, t
fs
= 0.5, : exact solution; ---: initial
solution; - -: asymptotic solution.
Fig. 4. Temperature at the solid uid interface versus time for different values of
the Eckert number, (a): isothermal case h

e
1, (b): adiabatic case, K= 1, t
fs
= 0.5, -
-: E = 0.1; : E = 1; - - -: E = 10.
A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027 4025
In Appendix D it is proposed the solution of the present problem
when a heat loss is imposed on the unwetted wall. The initial and
asymptotic analysis of the interface temperature is discussed in de-
tail. It is noteworthy that, even with a small heat loss, the interface
temperature does not indenitely grows but reaches an asymptotic
value.
9. Conclusions
In this paper we have presented an exact solution of the unstea-
dy conjugated heat transfer problem of a ow arising in a 2D chan-
nel with thick walls. The ow is driven by a pressure gradient
impulsively set to a constant value at the initial time. The solution
of the thermal eld has been obtained in the relevant case of Pr = 1
taking into account for the effects of the dissipation of kinetic en-
ergy in the uid (Eckert number different than 0). Two problems
have been solved depending on the thermal condition imposed
on the unwetted side of the channel walls: isothermal and
adiabatic case.
The time evolution of the temperature and heat ux at the so-
liduid interface have been analyzed and discussed in terms of
the main parameters which are ruling the phenomena: the Eckert
number E and the thermal activity ratio K.
Very simple expressions of the initial and asymptotic behavior
of the interface temperature and heat ux have been derived. They
are in perfect agreement with the exact solution for s ?0 and
s ?1 and can be used to quickly quantify the conjugated effects
in a wider class of problems.
Finally, it has been analysed the case in which the external
unwetted solid wall is not perfectly adiabatic, showing that the
interface temperature reaches an asymptotic value and does not
indenitely grows.
Appendix A. Abelian and Tauberian theorems
The initial and asymptotic behavior of the temperature at the
soliduid interface can be better identied by analyzing the solu-
tion in the transformed space.
Specifying with F(s) the Laplace transform of f(t), the Abelian
and Tauberian theorems ensure that
1) if, for t ?0, f(t) ?A t
a
then, for s ?1, F(s) ?AC(a + 1)/s
a+1
;
2) if, for t ?1, f(t) ?A t
a
then, for s ?0, F(s) ?AC(a + 1)/s
a+1
;
where C(x) is the gamma function.
By theorem 1, the local behavior of h
w
(s) for s ?0 can be found
by looking for the behavior of H
w
(s) (Eqs. (13) and (15)) for s ?1
and then performing the inverse transform again as suggested by
theorem 1.
Similarly, by theorem 2, the asymptotic behavior of h
w
(s) for
s ?1 can be found by looking for the behavior of H
w
(s) for
s ?0 and then performing the inverse transform.
Appendix B. Some useful formulae
Newtons formula
a b
m

m
i0
m
i

a
i
b
mi
: B:1
Extension 1
A recursive application of Newtons formula, Eq. (B.1), gives
a b c
h

h
i0
h
i

c
hi

i
j0
i
j

a
j
b
ij
: B:2
Extension 2
A recursive application of Eq. (B.2), gives
a
1
a
2
b c
h

h
i0
h
i

c
hi

i
j0
i
j

b
ii

j
r0
j
r

a
r
1
a
jr
2
:
B:3
Extension 3
A recursive application of Eq. (B.3), gives
a
1
a
2
b
1
b
2
c
h

h
i0
h
i

c
hi

i
j0
i
j

ij
k0
i j
k

b
s
1
b
ijk
2

j
r0
j
r

a
r
1
a
jr
2
:
B:4
Fig. 5. Temperature at the solid uid interface versus time for different values of
the thermal activity ratio, (a): isothermal case h

e
1, (b): adiabatic case, E = 1,
t
fs
= 0.5, - -: K= 0.1; : K= 1; - --: K= 10.
4026 A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027
Appendix C. Some inverse Laplace transforms
L
1
s
e
a

s
p
s

erfc
a
2s

; C:1
L
1
s
e
a

s
p
s
2

/
0
s; a; C:2
where /
0
s; a s a
2
=2erfca=2

s
p
a

s=p

e
a
2
=4s
.
L
1
s
e
a

s
p
s
3

/s; a; C:3
where /s; a

s
0
/
0
z; adz.
L
1
s
e
a

s
p
s
3=2

w
0
s; a; C:4
where w
0
s; a 2

s
p
e
a
2
=4s
=

p
p
a erfca=2

s
p
.
L
1
s
e
a

s
p
s
5=2

ws; a; C:5
where ws; a

s
0
w
0
z; adz.
L
1
s
cosh2q 1

s
p

cosh

s
p



1
2
@h
2
@q
q; s; C:6
where h
2
(q, s) is the Jacobi h
2
function. One of its representations is
h
2
q; s
1

ps
p

1
k1
1
k
e
qk
2
=s
; C:7
with 0 6 q 6 1.
L
1
s
sinh2q 1

s
p

sinh

s
p

1
2
@h
3
@q
q; s; C:8
where h
3
(q, s) is the Jacobi h
3
function. One of its representations is
h
3
q; s

1
k1
e
qk
2
=s
; C:9
with 0 6 q 6 1.
Appendix D. The case of heat loss at the unwetted wall
An interesting problem is the case in which the unwetted side
of the solid wall is not adiabatic, but a known heat loss is imposed.
Following [13], the new boundary condition is
@h

e
@Y

e
s Kh

e
s; D:1
where K is a given constant value.
The temperature at the solid uid interface can be obtained by
the same method proposed in Section 3:
H
w

H
w0
1 KD
c
; D:2
where H
w0
is the interface temperature in the adiabatic case (Eq.
(15)) and
D
c

1
cosh
2
rtanhr Ktanh

s
p
r K tanhr
: D:3
The analysis for small time values (s ?1) gives, as expected,
the same result, Eqs. (19) and (21) obtained in both the isothermal
and adiabatic cases.
The asymptotic behaviour of the interface temperature is
s !0 : H
w
s % h
wc1
1
s

1
s a
2


; D:4
s !1: h
w
s % h
wc1
1 e
a
2
s
; D:5
where h
wc1

4
3
E
K1K
K

t
fs
p and a
2
K

t
fs
p
=1 KK1=

t
fs
p
.
It is interesting to note that in the case of perfectly adiabatic
external wall, as s ?1 the interface temperature indenitely
grows with linear law. On the contrary, even with a small heat loss,
as s ?1the interface temperature tends to a constant asymptotic
value h
wc1
. The limits of Eqs. (D.4) and (D.5) for K ?0 can be com-
puted by the lHopitals rule and give the same asymptotic behav-
iors obtained in the perfectly adiabatic case, Eqs. (27) and (28).
The time necessary to reach the asymptotic state (99% of h
wc1
)
is
s
99%
K1=

t
fs


1 K

t
fs
p
K
log 0:01: D:6
The temperature on the unwetted side of the solid wall for large
time values also reaches an asymptotic value given by
h
e1

h
wc1
1 K
: D:7
References
[1] P. Drazin, N. Riley, The NavierStokes equations a classication of ows and
exact solutions, London Math. Soc. Lecture Note Series, vol. 334, Cambridge
University Press, 2006.
[2] I. Pop, D. Ingham, Convective Heat Transfer, Elsevier, 2001.
[3] S. Harris, D. Ingham, I. Pop, Unsteady heat transfer in impulsive FalknerSkan
ows: Constant wall temperature case, Euro. J. Mech./B Fluids 21 (2002) 447
468.
[4] D. Das, J. Arakeri, Unsteady unidirectional ow of second grade uid between
the parallel plates with different given volume ow rate conditions, Appl.
Math. Comput. 137 (2003) 437450.
[5] G. Brereton, Y. Jiang, Convective heat transfer in unsteady laminar parallel
ows, Phys. Fluids 18 (2006) 103602103614.
[6] A. Pozzi, R. Tognaccini, On the thermal eld in the impulsive Rayleigh ow,
Phys. Fluids 16 (12) (2004) 45324539.
[7] L. Perelman, On conjugated problems of heat transfer, Int. J. Heat Mass Transfer
3 (1961) 293303.
[8] K. Cole, Conjugate heat transfer from a small heated strip, Int. J. Heat Mass
Transfer 40 (1997) 27092719.
[9] A. Pozzi, R. Tognaccini, Symmetrical impulsive thermo-uid dynamic eld
along a thick plate, Int. J. Heat Mass Transfer 44 (2001) 32813293.
[10] A. Pozzi, R. Tognaccini, Time singularities in conjugated thermo-uid dynamic
phenomena, J. Fluid Mech. 538 (2005) 361376.
[11] A. Pozzi, G. Quaranta, R. Tognaccini, A self-similar unsteady ow with
conjugated heat transfer, Int. J. Heat Mass Transfer 51 (2008) 18041809.
[12] B. Weigand, G. Gassner, The effect of wall conduction for the extended Graetz
problem for laminar and turbulent channel ows, Int. J. Heat Mass Transfer 50
(2007) 10971105.
[13] A. Veeraragavan, C. Cadou, Theoretical study of conjugate heat transfer effects
on temperature proles in parallel ow with embedded heat sources, Int. J.
Heat Mass Transfer 53 (2010) 16991711.
[14] A. Pozzi, R. Tognaccini, Thermo-uid dynamics of the unsteady channel ow,
Euro. J. Mech./B Fluids 28 (2009) 299308.
[15] A. Ghizzetti, A. Ossicini, Trasformate di Laplace e calcolo simbolico, UTET,
1971.
[16] H. Carslaw, J. Jaeger, Conduction of Heat in Solids, second ed., Oxford
University Press, 1959.
A. Pozzi, R. Tognaccini / International Journal of Heat and Mass Transfer 54 (2011) 40194027 4027

You might also like