SOLUTION OF AN INVERSE PROBLEM OF HEAT

CONDUCTION BY ITERATION METHODS

0. M. Alifanov UDC536.24.02

An i t e r a t i o n s c h e m e f o r d e t e r m i n i n g n o n s t a t i o n a r y heat flux is c o n s t r u c t e d f o r the l i n e a r c a s e .

The s c h e m e is b a s e d on the two g r a d i e n t slope m e t h o d s .

We c o n s i d e r the following i n v e r s e p r o b l e m f o r the h o m o g e n e o u s l i n e a r equation of heat conduction in

the r e g i o n {0 < x < b, $ > 0}. The heat flux function q(r) at the left b o u n d a r y of the r e g i o n is to be d e t e r -
m i n e d f r o m the known t e m p e r a t u r e f(T) and the h e a t flux Q(T) at the r i g h t b o u n d a r y . Any p r o b l e m of d e t e r -
m i n a t i o n of a b o u n d a r y function can be r e d u c e d to this f o r m when f(r) is known at an inner point 0 < x 1 < b
and the b o u n d a r y condition is s p e c i f i e d at the point x = b. F o r this p u r p o s e it is n e c e s s a r y to solve the
c o r r e s p o n d i n g b o u n d a r y value p r o b l e m in the r e g i o n x 1 < x < b.
Thus we have (it is a s s u m e d that the r e d u c t i o n to the initial condition has been made)
OT(L Fo) _ 0 2 T ( L F o ) , 0<~<~, Fo>O, (I)
0 Fo 0~2
T(~, 0) = 0, (2)
OT(1, Fo) = Q(Fo), (3)
b 0~
T(l, Vo) = f(Fo), (4)
)~ OT (O, Fo)
= q (Fo) - ?, (5)
b 0~
w h e r e Fo = a T / b 2, ~ = x / b .
We i n t e r p r e t the i n c o r r e c t l y f o r m u l a t e d p r o b l e m (1)-(5) as a p r o b l e m of o p t i m u m c o n t r o l , i . e . , to
c h o o s e the c o n t r o l q(Fo) f r o m the condition of m i n i m u m deviation of T(q(Fo), 1, Fo) = T*(Fo) f r o m the
given function f(Fo) in the m e t r i c of s p a c e L 2 (square integrable functions):
For~
J (q) = .([T*(Vo) - - [ (Fo)] 2 dFo - - rain. (6)
0 q

We shaI1 solve the f o r m u I a t e d e x t r e m u m p r o b l e m by the g r a d i e n t m e t h o d s . F o r this p u r p o s e we

w r i t e out the f o r m u l a f o r the g r a d i e n t of functional (6).
Let us a s s u m e that q(Fo) u n d e r g o e s an i n c r e m e n t Aq(Fo). Then the t e m p e r a t u r e T(~, Fo) c h a n g e s
by an a m o u n t AT(~, Fo) which s a t i s f i e s the c o n d i t i o n s
OAT (~, Fo) 02AT ~, Fo) (7)
-- , 0 < ~ < 1 , Fo~0,
0 Fo 0~z
AT (~, 0) = 0, (8)
0AT(l, Fo) -- 0, (9)
o~
OAT (0, Fo) _ Aq (Fo),
O0)
b O~
S. O r d z h o n i k i d z e M o s c o w Aviation Institute. T r a n s l a t e d f r o m I n z h e n e r n o - F i z i c h e s k i i Zhurnal, VoI.
26, No. 4, pp. 682-689, April, 1974. O r i g i n a l a r t i c l e s u b m i t t e d June 17, 1973.

9 19 75 Plenum PublishhTg Corporation, 227 West 17th Street, New York, N. Y. 1001 l. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in arty titan or by any means, electronic, mechanical, photocoto,ing, microfilming,
recording or otherwise, without written permission of the publisher. A copy of this article is m,ailable from the publisher for $15.00.

471
 The functional J gets the corresponding increment
AJ ~ J (q -~ Aq) d (q) = J (T -k AT) ~ J (T).
Hence neglecting quantities of the second order of smallness we obtain
Forn
hJ = .I 2iT* (Fo) -- f (Fo)] AT (1, Fo) dFo. (115
0

Following [1] we write the boundary value problem conjugate to (7)-(10):

0,r (~, Fo5 02~ (g, Fo~)
O<g<l, 0 < Fo <Fo,,, (12)
0Fo Og.2
(~, Fore) = 0, (13)
0~2 (1, Fo}
= 9. IT* (Fo) - - f (Fo)l, (14)
0~ (0, Fo) (15)
-~O.
og
Then for the increment of functional (11) we get

AJ = ~ AT(~, Fo) O,(g, Fo) dFo --- dg4 AT(~, zo) O@(g' Vo) dFo
o~ 9 g o~
0 ~= 0 0 "

from which we obtain

1 1

(16)
8 0
where
Fcrn

I~ (~5 = S AT (~, Fo) 02~ (~'(9~

=Fo')dFo,
0

Fo m
fOAT (~, Fo) a~ (~, Fo') dFo
L (%) = o~ at

Next making use of the conjugate equation (12) and integrating by parts with (8) and (135 taken into considera-
tion, we have
Fo m Fo m

l'(g) = -- 3" hr(~, Fo) O~)(~,0

FoV~ dFo = ,f ~b(~,. Fo)ohr(g,OFoF~ gV~
0 0

Changing the order of integration and passing on to the spatial derivative AT(~, Fo) from (7) we ob-
tain
1 Forn 1

W-
0 0 0

Integrating by parts with the use of (9) and (10) we get

1 Fo m 1
b
~ 11(~)d~-~-;*(O, Vo)Aq(Fo)dFo--s (z7)
0 0 0

F r o m (16) and (17) we get

Fort~

AJ = .bbL.f ~ (0, Fo) Aq (Fo) d Fo.

0

In Hilbert space L 2 [1]

472
 Uom
AJ = .[ S' (q)Aq(FoidFo.
0

Thus the e x p r e s s i o n f o r the g r a d i e n t of functional (6) is

j , (q) = b ~ (0, Fo). (18)

F o r constructing the i t e r a t i o n a p p r o x i m a t i o n s to the d e s i r e d heat flux function q(r) we shall u s e the

m e t h o d s of s t e e p e s t d e s c e n t and conjugate g r a d i e n t s .
In the f i r s t c a s e s t a r t i n g f r o m a c e r t a i n " t e s t " solution q~ the subsequent a p p r o x i m a t i o n s a r e
found f r o m the f o r m u l a

qk+l = q~ __ [3hj, (q~), k = 0, 1. . . . (19)

The coefficient/3 k, which d e t e r m i n e s the step in going f r o m qk to qk+l, is found f r o m the condition of
m i n i m u m of j(qk+l).
Fo m Forn
rain f IT (qk _ ~S,k) _ fled Fo = rain I" t-- ~AT (S'~) -I- T (qk) __ []z d Fo.
6j f~ 8
Hence we obtain

S AT (J'#) iT (r -- [I d Fo
0
Fo m
.I AT2 (j,k) d Fo
0

In u s i n g the m e t h o d of c o n j u g a t e g r a d i e n t s [2] the s e q u e n c e {qk} is sought in the following f o r m :

r = r __ [~hpr~, k = 0, 1. . . . . (20)
where
p1~= j,t~ _, ?lj~-,,
F~

[ [J"~'(Fo)12d Fo
]'h = : Fom (~0 = O),

.I [j,i~,-1(Fo)]~ d Fo
0

The c o e f f i c i e n t flk is found f r o m the condition

rain J (qk _ 13p~).

A c c o r d i n g to the s u p e r p o s i t i o n p r i n c i p l e the solution of the s e c o n d b o u n d a r y value p r o b l e m f o r the

l i n e a r h o m o g e n e o u s heat conduction equation with b o u n d a r y functions ql(Fo) and %(Fo) can be w r i t t e n in
the f o r m
Eo Fo
O(g, Fo)= q2(q) O0(L Fo--q) dn-!- q~(n)OO(1--L Fo--n) dn,
i
0
OFo ,
0
i O Fo

w h e r e ~(Fo) is the solution of p r o b l e m (1), (2), (3), (5) with unit heat flux at one b o u n d a r y of the plate and
z e r o at the o t h e r .

Using a suitable a p p r o x i m a t i o n m e t h o d [3] we obtain the following f o r m u l a f o r c o m p u t i n g | at the

n - t h instant of t i m e :

= q..iO,(..... +0 (~) -i- "~ 7- x},~ - (~), n=l, 2. . . . m,

i~l i=1 (21)
where
- b - b "

473
 J Below we shall need On(0) and On(1 ) . In this c a s e
N~
1 i
40
\ 1
J

!
t
e7~(o) = - 2 V a Fo (m P) Z
]-o
i~*
]/- A Fo p)

v A Fo (m - - p) p=i-1

@:,,.(6) = - - 4 V" a Fo (m - - p)
>: i0" [ 2 I f A2,+,
N2

Fo (m P)
]t-p=i-I '
1=0

.0 ~ 8 12 f8 k
~7~ (1) = ~ 2i (0),
Fig. 1. E x a m p l e of c o n v e r g e n c e
of the i t e r a t i o n s e q u e n c e s (J, deg 2, a~m2~(1) = ~ (0).
AFo = 0.04, q ~ 0, fn, u n p e r - where
t u r b e d values): 1) c o m p u t a t i o n of AFo = aTm
a p p r o x i m a t i o n s by f o r m u l a (19); 2) b~m
1
by f o r m u l a (20). i(l)* [u] = .-7=- exp [ u'z] - - u (1 - - (D [u]) (iq3* [oo] = 0),
VZ
u

and r [u] = 2A&r.fexp [--r}2]d~/is the e r r o r i n t e g r a l .

0
The n u m b e r s N1 and N2 a r e c h o s e n f r o m the conditions

~ e, i(I:)* [ 2N~+t -]~e,

ir V A Fo-(m - - p) LV-A F-~(m - - V) 2 V A Fo (m p) "

w h e r e e > 0 is a p r e s p e c i f i e d s m a l l quantity.
In c o n s t r u c t i n g the s e q u e n c e of e l e m e n t s qk a c c o r d i n g to (19) o r (20) in e a c h i t e r a t i o n it is n e c e s s a r y
to solve t h r e e p r o b l e m s of d e t e r m i n a t i o n of the t e m p e r a t u r e T*(Fo), the function ~b(0, Fo), and the i n c r e -
m e n t AT(u k) (u = j , k o r pk). Atl t h e s e q u a n t i t i e s can be found f r o m (21) with c o r r e s p o n d i n g v a l u e s of ~,
ql, and q2. F o r Tn*k we have

~ = 1, ql~:=q~( -~ b ) - b Q , ,.

f o r ~bkn(0)

~=0, ~,.=0, ~,.=z~.~

~n= T n-1 + • n~ = T m - n -

f o r A T k (u)
-- k
= 1, ql, = un, ~,, = 0 (AT~ =0).
F r o m t h e s e quantities we c o m p u t e
= ~m--. (0),

~ k ~-- n~l
ttZ
w E(ATL1)2 + (aTb ~]

Z.a t(AL,)~ + (J;~)'l

~'h ~ .,,=1 (~'o = 0),
m
[(j;k_-]l)~ -t- (j;~-l)=]

w h e r e F k = ATk~r
The g r a d i e n t m e t h o d s of solving the i n v e r s e p r o b l e m of h e a t conduction (1)-(5) w e r e v e r i f i e d taking a
s e r i e s of m e t h o d o l o g i c a l e x a m p l e s and showed good r e s u l t s . As e x p e c t e d [2] the s e q u e n c e (20) c o n v e r g e s

474
 '(~mGX

f
~a

"', !

Y o/, o,a 0
i

L~ FC/FOrn 0 0,4 q8 FO/Fom

Fig. 2. R e c o n s t r u c t i o n of heat fluxes by the method of conjugate g r a d i e n t s (qmax is the l a r g e s t value
of the d e s i r e d function q(Fo)): a) AF = 0.01; fn, u n p e r t u r b e d values; continuous curve) exact solution;
dashes) initial approximation; points) 25th iteration; b) AFo = 0.005, q~ = 0, fn, u n p e r t u r b e d v a l -
ues; continuous curve) exact solution; dashes) l l t h iteration; e) AFo = 0.03, q0(Fo) = 0, fn, u n p e r -
turbed values; continuous curve) exact solution; dashes) 1st iteration; 1) second iteration; 2) 3rd
iteration; 3) a p p r o x i m a t i o n chosen a c c o r d i n g to d i s c r e p a n c y principle (4th iteration); 4) 50th iteration.

somewhat f a s t e r and p e r m i t s a c l o s e r a p p r o a c h to the m i n i m u m of functional (6) c o m p a r e d to (19). This is

i l l u s t r a t e d in Fig. 1 f o r one of the e x a m p l e s .
We note that a disadvantage of the above a l g o r i t h m of d e t e r m i n i n g the heat flux is the absence of a
u n i f o r m c o n v e r g e n c e of the a p p r o x i m a t i o n s (19) and (20) (the c o n v e r g e n c e o c c u r s only in L2). Because of
condition (13) the g r a d i e n t of the functional J' (q) deviates f r o m the exact value in s o m e neighborhood of the
end point Fo = Fom. This c a u s e s a distortion of the solution q(Fo) also in this neighborhood (q(Fo m)
= q0(FOm), q, (Fore) = q0, (FOm)) (see Fig. 2a).
However, if the b e h a v i o r of the function q(Fo) n e a r the right boundary of the t i m e interval is p r e d i c t e d
beforehand, then the p r o p o s e d a l g o r i t h m is v e r y efficient. F o r example, if it is known that the d e s i r e d
heat flux v a n i s h e s at Fo = FOm, then specifying the initial a p p r o x i m a t i o n q~ we obtain q(Fo) v e r y close
to the exact function (Fig. 2b).
The solution of the i n v e r s e p r o b l e m is quite stable to p e r t u r b a t i o n s of the input data both in the c a s e
of the method of s t e e p e s t descent andthe method of c o n j u g a t e g r a d i e n t s . This is mainly accounted f o r by a
significant d e c r e a s e of the r a t e of c o n v e r g e n c e of these methods of gradient slope with the i n c r e a s e in the
n u m b e r of i t e r a t i o n s which in a way d a m p s the "oscillations" of the a p p r o x i m a t i o n s qk(Fo). The r e s u l t s
f o r the f i r s t t h r e e i t e r a t i o n s obtained f r o m f o r m u l a (20) a r e shown in Fig. 2c f o r sufficiently l a r g e f l u c t u a -
tion p e r t u r b a t i o n s of the input data (fSn = fn + 6~n, where 6 = 107c fmax; ~n is a r a n d o m quantity u n i f o r m l y
d i s t r i b u t e d in the s e g m e n t [--1, 1]. The l a s t a p p r o x i m a t i o n is v e r y close to the sought c u r v e of q(Fo).
If the p r o c e s s of a p p r o x i m a t i o n s is continued f u r t h e r , the r e s u l t s change v e r y insignificantly in the
next t h r e e Lterations. A gradual oscillation of the solution is o b s e r v e d s t a r t i n g f r o m 7 - 8 - t h iteration (50-th
a p p r o x i m a t i o n f o r this example is shown in Fig. 2c).
Computations show that it is advisable to use the d i s c r e p a n c y principle [4, 5] for t e r m i n a t i n g the
i t e r a t i o n p r o c e s s . H e r e the n u m b e r of i t e r a t t o n s plays the role of the r e g u l a r i z a t i o n p a r a m e t e r . If n e c e s -
s a r y the exit to the given discrepanc3; level is obtained by a suitable choice of the p a r a m e t r i c step ft. If
the input data fn a r e sufficiently a c c u r a t e , then f o r the above a l g o r i t h m it is g e n e r a l l y always possible to
specify beforehand the n u m b e r of i t e r a t i o n s (10-15) which would c o r r e s p o n d to a r e a s o n a b l e p r o x i m i t y of
the l a s t a p p r o x i m a t i o n to the actual heat flux function.
A p a r t f r o m its high a c c u r a c y this a l g o r i t h m is also economical in r e s p e c t of c.omputation t i m e and
c o m p a r e d to the a l g o r i t h m s of [6, 7] it p e r m i t s solution of p r o b l e m of c o n s i d e r a b l y l a r g e r dimensionality.

475
 The investigated approach can be extended to the solution of i n v e r s e p r o b l e m s with movable bounda-
r i e s [5, 8] also in nonlinear formulation.

NOTATION
is the t h e r m a l diffusivity;
b is the right end of the interval Ln x;
f(~) i s the given t e m p e r a t u r e dependence at the point x = b;
J(q) LS the funetLonal to be minimized;
k Ls the n u m b e r of iteration;
QO') i s the heat flux at the point x = b;
q(v) ts the desLred heat flux;
T(x, ~') ts the tempe rature;
T*(T) LS the computed t e m p e r a t u r e dependence at the point x = b;
X LS the coordinate;
Fo tS the F o u r i e r number;
k [S the t h e r m a l conductivity;
LS the d i m e n s i o n l e s s coordinate;
T LS the time;
Zm LS t h e end of the investigated tLme interval;
r T) iS the solution of the conjugate boundary value p r o b l e m ;
A iS the i n c r e m e n t of the correspondLng quantities.

LITERATURE CITED
1. B. M. Budak and F. P. Vasil'ev, Approximate Methods of Solution of Optimum Control Problems
[in Russian], II, MGU, Moscow (1969).
2. B. T. Polyak, in: Computational Methods and Programming [in Russian], XII, MGU, Moscow (1969),
p. 38.
3. O. M. Alifanov, in: Investigations on Convective Heat and M a s s T r a n s f e r [in Russian], Nanka i
Tekhnika, Minsk (1971), p. 322.
. V. A. Morozov, Z h u r n a l V y e h i s l i t e l ' n o i Matematiki i M a t e m a t i e h e s k o i F i z i k i , 8 , No. 2, 295 (1968).
5. O. M. Alifanov, Inzh. Fiz. Zh., 2.~3, No. 6 (1972).
6. O. M. Alifanov, Inzh. Fiz. Zh., 24_, No. 2 (1973).
7. O. M. Alifanov, Heat and M a s s T r a n s f e r [in Russian], Vol. 8, Minsk (1972), p. 89.
8. 0 . M. Alifanov, Inzh. Fiz. Zh., 26, No. 2 (1974).

476