Professional Documents
Culture Documents
www.elsevier.com/locate/ijhmt
Abstract
An analytical method has been developed for two-dimensional inverse heat conduction problems by using the
Laplace transform technique. The inverse solutions are obtained under two simple boundary conditions in a nite
rectangular body, with one and two unknowns, respectively. The method rst approximates the temperature changes
measured in the body with a half polynomial power series of time and Fourier series of eigenfunction. The expressions
for the surface temperature and heat ux are explicitly obtained in a form of power series of time and Fourier series.
The verications for two representative testing cases have shown that the predicted surface temperature distribution is
in good agreement with the prescribed surface condition, as well as the surface heat ux.
2003 Elsevier Science Ltd. All rights reserved.
Keywords: Inverse solution; Two-dimensional heat conduction; Laplace transform; Transient
1. Introduction
A procedure to solve Inverse Heat Conduction
Problems (IHCP) is to derive surface heat ux and
temperature from temperature changes inside a solid.
The method proves to be very powerful and useful when
a direct measurement of surface heat ux and temperature is dicult, due to severe working conditions, such
as those in space vehicle atmosphere re-entry, in accidents involving coolant breaks in the plasma-facing
components, and in the quenching of a high temperature
surface. Recently, the IHCP has been numerically treated and extended to multiple dimensions with the help of
the development of computer technologies related to
software and hardware and the improvement of computer capability. For example, Hsieh and Su [1] and Bell
[2] employed a dierential method; Lithouhi and Beck
[3] employed a nite element method and Shoji and Ono
0017-9310/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0017-9310(02)00510-0
2136
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Nomenclature
coecients in Eq. (7)
Aj
a
thermal diusivity
Bj
coecients in Eq. (7)
bk
coecients in Eq. (8)
Cj
coecients in Eq. (6)
cj
coecients in Eq. (8)
C10 , C20 , C30 , C40 coecients in Eq. (3)
n
coecients for approximating measured
Dj;k
temperature variation in Eq. (9)
1
Gj;k
coecients in Eq. (18) for set 1
l;m
Hj;k
l;m
Hj;k
Gj;k
Ly
mj
nj
N
Nj
Nsf
s
T
t
Tc
x; y
U
h
s
si
n
g
4a
oh
0 at g 1
og
4b
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Adiabatic
1
Adiabatic
= 1
1()
Solid
1
Adiabatic
1
X
j0
Bj cosnj g g2 =Lg
Adiabatic
Nj
X
cj cosmj n
j0
N
X
k0
bk
s sn k=2
Ck=2 1
8
=1
n
=2
2()
1()
Solid
2137
Adiabatic
Fig. 2. Illustration of the B.C in the boundary set 2.
2138
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Nj X
N
X
j0
k1
Dj;k
s sn k=2
Ck=2 1
9
f n; gn ; s essn
cosmj n
j0
N
X
13
12
cosmj n
hn; g; s ess1
N
X
Dj;k =sk=21
14
k0
Nj
X
PN
Dj;k
k0 sk=21
sinnj =Lg2 g
cosmj n
sinfnj =Lg2 g1 g
j0
ss2
e
Nj
X
PN
Dj;k
k0 sk=21
sinfnj =Lg1 gg
cosmj n
sinfnj =Lg2 g1 g
j0
15
q
Substitution of nj i s m2j into cosnj =L and
hq i
sinfnj =Lg2 gg turns them into cosh s m2j =L
hq i
and i sinh s m2j =L , respectively. Taking into account these relations and setting g 0 in Eqs. (14) and
(15), we can obtain the surface temperature (g 0) as
For the boundary set 1,
hw n; s ess1
Dj;k =sk=21 ;
Nj
X
j0
p2
smj
cosmj n cosh
L
p2
smj
cosh
1
g
l
L
k0
n 1; 2
10
N
X
Dj;k =sk=21
16
k0
Cj
ess1
PN
1 k=21
k0 Dj;k =s
n 1g
cos j L 1
11
hw n; s
nq o
1
N
s m2j =L g2 cosmj n X
Dj;k
nq
o
e
k=21
s
j0 sinh
k0
s m2j =L g2 g1
nq o
2
Nj sinh
N
s m2j =L g1 cosmj n X
X
Dj;k
nq
o
ess2
sk=21
j0 sinh
k0
s m2 =L g g
ss1
Nj sinh
X
17
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
1
Nj X
N
X
Gj;k cosmj n
s s1 k=2
Ck=2
1
j0 k1
18
hw n; s
1;2
Nj X
N
X
Gj;k cosmj n
s s1 k=2
Ck=2
1
j0 k1
2;1
Nj X
N
X
Gj;k cosmj n
s s2 k=2
Ck=2 1
j0 k1
19
j0
smj
L
N
X
smj
cosmj n sinh
L
p2
smj
cosh
1 gl
L
1
Dj;k =sk=21
20
Uw n; s
1;2
Nj X
N
X
Hj;k cosmj n
s s1 k=2
Ck=2
1
j0 k1
2;1
Nj X
N
X
Hj;k cosmj n
s s2 k=2
Ck=2
1
j0 k1
23
j0
1
N
X
Dj;k
cosmj n ess2
sk=21
k0
q
nq o
Nj
s m2j =L cosh
s m2j =L g1
X
nq
o
j0
sinh
s m2j =L g2 g1
2
N
X
Dj;k
cosmj n
k=21
s
k0
Nj
X
q
nq o
s m2j =L cosh
s m2j =L g2
nq
o
sinh
s m2j =L g2 g1
22
p2
k0
Uw n; s ess1
1
Nj X
N
X
Hj;k cosmj n
s s1 k=2
Ck=2
1
j0 k1
Nj
X
p2
Uw n; s ess1
In the same way that we derived the surface temperature, in achieving the surface heat ux, we also need
to expand the hyperbolic functions in Eqs. (20) and (21)
in series around s 0, and then perform the inverse
Laplace transform. The unknown surface heat uxes are
nally expressed as
Uw n; s
2139
21
24
2140
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Testing case 2:
U 1 for 0 < n < 0:5 and U 0 for 0:5 < n < 1:0
at g 0
26
Fig. 4. Temperature reproduced from Eq. (9) for the temperature distribution in the Fig. 3.
Fig. 5. Surface heat ux estimated with Eq. (23) for the testing
case 2 at g1 0:02 and g2 0:05.
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Fig. 6. Surface temperature estimated by Eq. (18) for the testing case 1 at g1 0:02 and g2 0:05.
2141
To both the testing cases 1 and 2, the inverse solutions of Eqs. (19) and (23) are capable of predicting the
surface temperature and heat ux within an error bound
from 3% to 5%. In general, the prediction of the surface
temperature is better than that of the surface heat ux.
The main reason is related to the temperature disturbance imposed by Eq. (24): For the heat ux, the temperature dierence between the measuring points
becomes important and it makes level of signicant
gure degrade, that is the imposed disturbance be amplied.
It is necessary to mention that for the testing case 1,
the temperature on the surface possesses a discontinuity
point at n 0:5 so that the surface heat ux obtained by
dierentiating the temperature with respect to g does not
converge uniformly at n 0:5. Consequently, the predicted surface heat ux at n 0:5 is, of course, divergent. Under such circumstances, only the temperature
estimation is possible from the inverse solution, which is
shown in Fig. 6. Although the surface temperature for
the testing case 2 predicted from Eq. (18) is slightly inferior to that predicted from Eq. (19), it is much better
than the predicted surface heat ux shown in Fig. 7.
It can be concluded from a direct comparison of Figs.
57 that the inverse solutions obtained from two measuring lines can estimate the surface temperature, as well
as the surface heat ux, with much higher accuracy than
that obtained from only one measuring line. Therefore,
we recommend Eq. (23), instead of Eq. (22), to be used
in the unknown surface heat ux predictions even under
the boundary set 1 condition.
4. Discussions
4.1. Inuence of discontinuous point
As shown in Figs. 4 and 5, a degradation of prediction accuracy is observed near the discontinuity point.
The reason is attributed to the employment of the
Fourier series in the temperature approximation given
by Eq. (9). In other words, the Gibbs phenomenon,
which is a characteristic phenomenon in the Fourier
series [13], becomes apparent near the discontinuity
point of n 0:5. But, for the location far from the discontinuity, the present inverse solution is still available.
4.2. Minimum predictable time
Eqs. (16) and (22) and Eqs. (19) and (23) give out the
surface temperature and heat ux explicitly. However,
based on study of one-dimensional IHCP [10], we expect
that for the Laplace operator, s, Eq. (11) could become
divergent. This characteristic takes place in any inverse
solution including numerical and analytical ones, that is
mathematically veried [6,7], since it takes a time for a
2142
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
4.4. Inuence of the relative positions of the two temperature measuring lines
To check the inuence of the relative positions of the
two temperature measuring lines, we further set the
positions of the two measuring lines to
Combination 1: g1 0:05 and g2 0:10;
Combination 2: g1 0:02 and g2 0:10.
With keeping Nj , N and Nsf the same as that used in
the estimation of Fig. 5, we can obtain the estimations
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
2143
Fig. 10. Surface heat ux estimated by Eq. (23) for testing case
2 with Nsf 2 (g1 0:02 and g2 0:05).
2144
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
27
w [-]
1 .2
1
0 .8
6. Conclusion
With using the Laplace transform technique, we
achieved inverse solutions for two-dimensional IHCP
for two simple boundary sets, the results are summarized
as the following:
1. Except in the vicinity of discontinuity point, surface
temperature and heat ux can be predicted well over
the whole surface with an error less than a few percent.
2. The minimum predictive time for the proposed IHCP
solution is about smin 0:02.
3. The position of the rst measuring line g1 in the solid
is recommended to be as near as possible to the interested unknown boundary surface and not to be farther than 0.02.
4. Eq. (23), instead of Eq. (22), is recommended for estimating surface heat ux even for the boundary set 1
condition.
5. In order to predict the surface temperature and heat
ux from the proposed IHCP solution successfully,
a high precision instrument that can ensure measured
temperature at uncertainty level less than 0.1%,
namely Nsf 3 at least, is recommended.
0 .6
0 .4
0 .2
[-]
-0 .2
1
[-]
0 .8
0 .6
0 .4
0 .2
1 .1 26
0 .8 41
0 .5 56
0 .2 71
0 .0 20
0
Fig. 12. Surface heat ux calculated from Eq. (23) for moving
heat source using time partition method.
1
j0
k0
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
P
n
n
If we dene Pj;k 1
j0 cosmj ni Dj;k , the above equation is then rewritten to
f ni ; gn ; s
Nk
X
k0
Pj;k
s sn k=2
Ck=2 1
Nj
N X
X
k0
2145
K1;j s Co1;j
1
X
cn;j sn
n0
and
l
K2;j s Co2;j
1
X
dn;j sn ;
l 1; 2
n0
l
cn;j
and
n
1 X
l
y hni;j ;
g0 i0 i;j
dn;j
l 1; 2
where hi;j is calculated from the following listed equations. The calculation to the employed item gi;j and col
l
ecients xi;j , yi;j are listed in Table 2.
Dj;k Qj;k n; s
j0
n
h0;j 1
h1;j g1;j
2
h2;j g2;j g1;j
3
h3;j g3;j 2g1;j g2;j g1;j
2
2
4
h4;j g4;j 2g1;j g3;j g2;j
3g1;j
g2;j g1;j
Appendix B
2
2
h5;j g5;j 2g1;j g4;j g2;j g3;j 3g1;j
g3;j g1;j g2;j
3
5
g2;j g1;j
4g1;j
..
.
With rearrangement, hw s and /w s are derived as:
(1) For the boundary set 1,
hw s ess1
ess1
1
Nj X
N
1
X
Dj;k cosmj n X
Co1;j cn;j sn
k=21
s
j0 k0
n0
Nj X
N
X
Gj;k sn cosmj n
j0 k1
Table 1
Boundary set 1
Boundary set 2
Surface temperature
PNj
f1 sK1;j s
h j0
o
PNj n
2
1
h j0 f1 sK1;j s f2 sK1;j s
Surface heat ux
PNj
U j0
f1;j sK2;j s
o
PNj n
2
1
U j0 f1;j sK2;j s f2;j sK2;j s
2146
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Uw s ess1
ss1
Nk
X
1
Nj X
N
1
X
Dj;k cosmj n X
Co2;j dn;j sn
k=21
s
j0 k0
n0
Hj;l
Nj X
N
X
and Gj;l and Hj;l are coecients of Eqs. (18) and (22),
respectively, and Ej;k Co1;j ck;j and Fj;k Co2;j dk;j
Dj;2k1 Fj;k ;
1
Hj;k sn
0 6 l 6 N ; Nk IntfN l=2g
k0
cosmj n
j0 k1
Gj;1
Nk
X
hw s ess1
Dj;2k1 Ej;k1 ;
k0
Gj;l
Nk
X
ess2
1
Dj;2kl Ej;k ;
0 6 l 6 N; Nk IntfN l=2g
ess1
k0
Hj;1
Nk
X
1
Nj X
N
1
X
Dj;k cosmj n X
2 2
Co1;j cn;j sn
k=21
s
j0 k0
n0
1;2
Nj X
N
X
Gj;k
cosmj n
sk=21
j0 k1
ess2
Dj;2k1 Fj;k1 ;
2
Nj X
N
1
X
Dj;k cosmj n X
1 1
Co1;j cn;j sn
k=21
s
n0
j0 k0
k0
2;1
Nj X
N
X
Gj;k
cosmj n
sk=21
j0 k1
Table 2
Kernel Ks and its coecients
(1) For the boundary set 1
Kernel Ks
mj 0 (j 0)
Co1;0 1
Co2;0 L1
gn;0
Common item
mj > 0 (j > 0)
xn;j
Coecients Coj
Co1;j
1g1 2n
1
2n!
L
k Cn 1 2k m2kn
j
kn 2k! L
cosh mj L1
cosh mj L1
cosh
1g
mj
gn;j
Common item
mj 0 (j 0)
1
xn;0 2n1!
mj > 0 (j > 0)
1
yn;0 2n!
Co1;j
gl 2n
L
gn;0
l
xn;j
mj
gl 2n1
Co1;0 1
Common item
cosh
2k
2kn
k Cn 1g1
mj
L
kn 2k!
1g1
cosh mj L
y0;j 1,
P
k1 Cn 1 2k1 2k1n
L mj
n P 1
yn;j kn1 2k1!
mj sinh mj L1
1
m
sinh
m
j
j
1gL
Co2;j L1
1
gl 2k1 2kn1
k Cn
mj
kn 2k1! L
g
sinh mj Ll
g
sinh mj Ll
sinh m g2 g1
f j L g
gn;j
Co2;0 L1
yn;j
l
k Cn gl 2k m2kn
j
kn 2k! L
g
coshmj Ll
g
f
mj cosh mj Ll
Co2;j L1 sinh
g2 g1
L
fmj
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
Uw s ess1
1
Nj X
N
1
X
Dj;k cosmj n X
2 2
Co2;j dn;j sn
k=21
s
j0 k0
n0
ess2
ess1
2
Nj X
N
1
X
Dj;k cosmj n X
1 1
Co2;j dn;j sn
k=21
s
j0 k0
n0
1;2
Nj X
N
X
Hj;k
cosmj n
sk=21
j0 k1
ess2
2;1
Nj X
N
X
Hj;k
cosmj n
sk=21
j0 k1
where
Nk
X
l;m
Gj;1
2
1
1 X
L2 coskpg eLkp t
2
2
p k1
k
1
2L X
cosf2n 1png
1n
2
p n0
2n 1
coshf2n 1p 1 g=Lg
sinhf2n 1p=Lg
1
2n12 p2 t
2L2 X
n cosf2n 1png e
1
p2 n0
2n 12
1 X
1
4L2 X
1n
3
p n0 k1
0 6 l 6 N ; Nk IntfN l=2g
Dj;2k1 Ej;k1 ;
k0
l;m
Gj;l
Nk
X
Dj;2kl Ej;k ;
l;m
Nk
X
Dj;2k1 Fj;k1 ;
k0
l;m
Hj;l
Nk
X
Lk2 p2 t
Dj;2kl Fj;k ;
h4 n; g; s
0 6 l 6 N ; Nk IntfN l=2g
k0
l;m
f2n 12 Lk2 g 2n 1
k0
Hj;1
2147
where
l;m
h1 n; g; s C1 q0 b
h2 n; g; s
1
X
s2
2
2C1 q0 L2
cosjpn
4
jp jpb
2
cosjpbs bejp s
jpsinjpbs b
j1
h3 n; g; s
1
X
h4 n; g; s
1 X
1
2 X
sinjp=2 cosjpn
2
3
p j1 k0
j
L2 p 2k 1 sinf2k 1gp=2g e
1
Lkp2
1 X
1
X
j1
1
2X
sinjp=2 cosjpn coshf1 g jp=Lg
p j1
j
coshjp=L
L
4
kp
k1
s
eLkp
4C1 q0 L2
!
2
1
Lkp2
cosjpn coskpg
jp2 Lkp2 2 jpb2
jp2 Lkp2
sinjpbs b cosjpbs
jp
!
bejp
2
coskpg
k1
2C1 q0 b
Lkp2 s
2k12 j2 p2 t
fL2 2k 12 =4 j2 g
2
2
2
1
1 2X
sinf2k 1gp=2g eL 2k1 =4 p t
2k 1
2 p k0
2148
M. Monde et al. / International Journal of Heat and Mass Transfer 46 (2003) 21352148
References
[1] C.K. Hsieh, K.C. Su, A methodology of predicting cavity
geometry based on scanned surface temperature data
prescribed surface temperature at the cavity side, J. Heat
Transfer 102 (2) (1980) 324329.
[2] G.E. Bell, An inverse solution for the steady temperature
eld within a solidied layer, Int. J. Heat Mass Transfer 27
(12) (1984) 23312337.
[3] B. Lithouhi, J.V. Beck, Multinode unsteady surface
element method with application to contact conductance
problem, J. Heat Transfer 108 (2) (1986) 257263.
[4] M. Shoji, N. Ono, Application of the boundary element to
the inverse problem of heat conduction, Trans. JSME Ser.
B 54 (506) (1988) 28932900 (in Japanese).
[5] C.H. Huang, C.C. Tsai, An inverse heat conduction
problem of estimating boundary uxes in an irregular
domain with conjugate gradient method, Heat and Mass
Transfer 34 (1998) 4754.
[6] O.M. Alifanov, in: Inverse Heat Transfer Problems,
Springer-Verlag, 1994, pp. 58.
[7] J.V. Beck, B. Blackwell, C.R. Clair, Inverse Heat Conduction, A WileyInterscience Publication, 1985.
[8] M. Imber, J. Khan, Prediction of transient temperature
distributions with embedded thermo-couples, AIAA J. 10
(6) (1972) 784789.
[9] M. Imber, Temperature extrapolation mechanism for twodimensional heat ow, AIAA J. 12 (8) (1974) 10891093.
[10] M. Monde, Y. Mitsutake, Analytical method in inverse
heat transfer problem using laplace transform technique,
Int. J. Heat Mass Transfer 43 (21) (2000) 39653975.
[11] Monde et al., A new estimation method of thermal
diusivity using analytical inverse solution for one dimensional heat conduction, Int. J. Heat Mass Transfer 44 (16)
(2001) 31693177.
[12] H.S. Carslaw, Conduction of heat in solids, second ed.,
Oxford University Press, 1959.
[13] R.V. Churchill, Fourier Series and Boundary Value Problems, McGraw-Hill, 1963.
[14] J.V. Beck, K.D. Cole, A. Haji-Sheikh, B. Litkouhi, Heat
Conduction Using Green Functions, Hemisphere Publishing Corporation, 1992.