2424 SHORTER COMMUNICATIONS

-
-1-O 4.24 - 305 -0945 26.4 - 7.41 - 3.01
-5.0 490 -61.6 - 2.97 1520 - 6.30 - 3.9t
- 10.0 5.54 - 109.0 - 6.98 661.0 - 623 -4.16

The dimensionless velocity gradient at the wait is given by lD%FERENcEs

A. S. B-AH, Laminar flow in channels with porous
walls, J. A@. Phys. 24, 1232-1255 (1953).
R. M. TERWL and P. W. THOMAS,On laminar flow
The wall shear stress, which is proportional to the negative through a uniformty porous pipe, A&. Sci. Res. ZlA.
value of this quantity, is seen to be independent of e to the 3, 37-67 (1969).
first approximation and decreases slightly from R, = - 1.0 M. J. MANTON, Low Reynolds number flow in slowly
varying axisymmetric tubes, J. Fluid Me&. 49. 451-459
to -10-o (see the table). The effect of the second-order
(1971).
modification is generally small and diminishes with the 4. W. D. HAYES and R. E. FROIETEIN.Hypersur8icFh
increase of distance <. Thus, as in the hypersonic smaIl- Theory. Vol. I. Academic Press, New York (1966).
disturbance theory, the first-order approximation is expected 5. J. SWRIN, Asymptotic behavior of velocity profiles in.the
to describe rather accurately the flow conditions in a porous Prandtl boundary layer theory, Proc. R. Sot. 2% 491-507
paraboloidal pipe. 11967).

Im.1. HrorA&usTramffrr.Vol. 16,pp.2424-2428.Pergsmoa

Ross 1973. Priatedin GreatBritarn

THE HEAT l&ALAN- INTEGU METHOD*

DAVID LANGFORDt
Department of Mechanical Engineering, Drexct University, Philadelphia, Pennsylvania 19104, U.S.A.
(Received 1 Much 1971 and in revisedform 25 May 1973)

INTRODUCTXON The applicability to phase change problems is of special

i%PPROXlMATEsolutions to transient diffusion problems may importance [I, 2,163 because existing closed form solutions
be obtained relatively easily by the use of what is commonly to these signifkcant problems are highly restrictive as to
called “The Heat Balance Integral Method,” THEBIM. allowabk initial conditions and boundary conditions
THEBIM is applicable to one-dimensional linear and non- 13,12-153.
Iinear problems involving temperature dependent thermal Tk accuracy of an approximate solution is in general
properties [6,7,18], non-linear boundary conditions [7,9], unknown [2. S-8, 11, 161. Using THEBIM, attempts to
and phase change problems such as freezing [4,5,7-l& 17. increase the accuracy of an approximate solution have
sometimes actually caused a decrease in accuracy [6,7. 161.
There may therefore be some vaiuc in an accuracy criterion
which can be easily used even when the exact solution is
* This work was partially supported by a grant from the
unknown. The use of such a criterion is illustrated here for a
National Science Foundation. The computer work needed
for this study was ably done by Richard B. Walck, formedy classical problem.
+ Drexel student
t Formerly Associate Professor of Mechanical Engineer- A SAMPLE PROBLEM
ing at Drexel University, Phila. Pa. 19104. Now Regional
Radiation Representative, U.S. Environmentai Protection Let nx, r) be the temperature at position x at time t in a
Agency, Phila., Pa. 19106, U.S.A. semi-infinite slab having constant thermal conductivity k,
 SHORTER COMMUNICATIONS 2425

density p, specific heat c, and thermal ditfusivity a E kjpc. balanceintegral:

Also, let subscripts in x and t denote partial d&rent&ion,
SO T,,(x, t) = a2Z?axat. Finally, let 4(x, f) be delined as the [ - kU,.(O,t’) dt’ = $ pc[ Ufx, t) - ?;] dx. (11)
heat flow in the +x direction per unit time per unit of
transverse area. Then, from Fourier
Equation (11) will be used in place of (2). In place of the
q(x, t) = - kT’x, Q. 0 < x, 0 d t. (1) initial condition (3), the appropriate initial condition is
For no internal heat sources, (1) and conservation of p(O) = 0. (12)
energy yield the usual partial differential equation
Also, the boundary conditions (4.5) on nx, t) are now
kT,(x, t) = pc7’Jx. t). (2) replaced by the following boundary conditions on U(x, t).
Initial and boundary conditions to be used are UCO,t)==~ for t>0; (13)

T(x, 0) = ‘I;“*,**,=T=constant. x>0; (3) U(p(t), t) = T for t > 0. (14)

T(4 t) = T&hcr = T,=constan& t >O; (4) The approximation procedure is now reduced to finding a
function U(x, t) u T(x, t) such that (13) and (14) are sati&d,
E r = constant,
VW, t) = TL.LII~I t finite. (5) this function U then being used in (11) to fti tbe penetration
Equations (2)-(5) constitute a classical problem with depth p(t), subject to the initial condition (12). In fmding
solution [3, 161 U(x, t), it is convenient (but not necessary [I) to restrict
the search to polynomial functions of the form
(T- T3/(T, - ?J = [l - erf(,/(x’/4ut))]. where (6)
U”(X,t) = A&) + B”(t).x + C,(t). x2 + . . . x”, (15)
Q&) P q...a (0, t) = - kT’J0, t) = k(T, - 7&,&z). (7)
where the coefficients are functions of time, and where the
corresponding value of p(t) will be denoted by p&).
THE APPROXIMATION PROCJCDURE
Assume there exists a function U with REsuLT!3
The first possible approximation of the form (15) which
Ufx, t) = nx, t). (8)
can satisfy the ww conditions (13) and (14) is
The actud temperature distribution 7Yx,t) satis& the
partial differential equation (2) and also the following U,(x, t) = A*(t) + B,(@.x. (16)
integral quation : Use of (16) with (13, 14) yields

d 4(0, t’) dt’ = j - kT.0, 1’)dt’ = i pc[nx, t) W, - Q/K - 231 = I1 - X/P,(Ql. (17)
0 0
To lind p,(t), it is nccegPary to use (17) in (ll), yielding
- 770, t)] dx. (9a)
- kP, - , j [df/p,Wl = pdT, - ‘I;lp,W2. (18)
This is simply a “balance” of the heat energy input on the cl
left against its measurable effect on the right ; it may be called
the “heat balance integral”. Many approximation techniques DifIerentiation and rearrangement yields
(e.g. that of “lumped parameters”) do not satisfy (2) but do p1 .(dp,/dt) L 2u = constant. (19)
satisfy f9a) [l, 161. Here also it is not required that u
satisfy (2) but it is required that U satisfy (9a) so Integration and USCd the initial cxmditicm(12) yields

PA0 = JW). (20)

[ - kUd0, t’) dr’ = 9 pc[Ufx, t) - B dx. (9b)
The !imt approximation U,(x, t) is fully de!ined by (17)
and (20). Note that
Assume that, at any finite time t. the sign&ant (measur-
able) effects of the boundary disturbance (3) and (4) do not QA4 f - kU,_@,t) = MT, - a/p,
penetrate beyond some finite distance x - p(t). This
- MT, - T,uJw4; (21)
assumption can be stated in the following mathematical
form : M, 5 [Q#)/Q&) - 11 = - 0114 = - 11.4 per cent. (22)

U(p(t),t)=T for t>O; (loa)

HIGHER APPROXIMATIONS
U(x, t) = ‘1; for x > p(t). (lob) It is possible to increase the degree of the polynomial
Use of (lob) in f9b) yields the following form of the heat approximation U,(x, t). The degree n that can be used
2426 SHORTER COMMUNICATIONS

increases with the number of conditions such as (13) and (14) whence
avaiiabte to determine the coeffkients of C.J,,(x,r). Extra JW
conditions cannot be obtained by simple d~e~n~tion of &U~~~r~r) - fcUJ0, r) = J @Ax, r) dn. (28)
the boundary conditions ( 13)and (14) because such difkrenti- 0
ation yields two conditions already satisfied identically by But, from (14), Ufplt). t) = ‘I; = constant so (28) can be
U r(x, t), namely written
UJO, t) t 0, (23) d <P(t)
kU,W), t) - kU,& c) = z{ j pc[Uk r) - T] d-x-/. (29)
and
U&Q), tf. fdp/dr) i- U,&t). t )= 0. 124
Differentiation of t 11) with respect to time and use in (29)
It is however possible to require that U(x, t) satisfythe heat then yields the following as the s&&id form of the new

Table 1. Approximate temperawe distrjbutions U, in order of decreasing e,,*

- .

WI U,(x, t) - 2 X Conditions used to

n e, b, %(Y) =
(%) T,-T ’ Y”p obtain U, (x, I)

- il.38 01667 4 ur = (1 - y) 11-15

:L - 11.38 @I667 4 UZ&= 1 - y t Oy” 1l-15,25
3LR -39.58 0@07 16 US&A= I - gy + oy2 + ;ys 11-15.25.26

3L -640 0.0417 8 ujf_ = 1 - fy + Oy’ + ;yl 1 l-15,25. 27

3R “I-8.54 @0233 24 QR = (1 - yY = 1 - 3y + 3yJ - y3 11-15, 26.27
2 i- 2.33 O-0192 12 u2 = fl - rV z 1 - 2y + yz 11-15.
2R + 1.13 0.0175 Jf33P 3 U2r’~+~~~~+~~j92 11-15, 26 27
4 - 2.92 QOO53 lid = 1 - 2y + oy2 + 2yJ - yJ 1i--15.25,26.27

* Note: M ss [ U,(O, t)/T.O, t) - 1] = dimensionless error in surface heat flux : p = ,/fb,at).

condu~on quation (2) at x = 0 or x = p(t) or both. Use of domoin condition (27) :

(2) in f23) and (24) yields
kUJp(t), t) = 0. (30)
UJO, 0 = 0, (25)
Let U#, t) be the second degree polynomial approxi-
and mation corresponding to the use of (27) in lieu of (25) or (26).
(26) Rsalts are in Table 1.
U&W. r).(dp/dr) f oU,&r), 0 = 0.
“Higher:’ ap~o~~tio~ can be obtained by requiring
Let Us&x, r) and U&x, t) be the second degree pofy- that Uk t) satis& two or three of the conditions (253-127).
nomiaf approximations correspondin respectively to the Results are in Table 1.
use of the extra conditions (25) and (26) at the left end
(X = 0) and right end (x = p) of the domain, with each ACCURACY
approximation also satisfying (11)-(15). Results are in
The following is a calculable measure of the error inherent
Table 1, and it may be noted that UJ,, does not seem to have
in the use of any approximation Utx, t) :
been published previousiy.
An alternate second degree approximation can be obtained
i?(t) 3 j [ku&, t) - pcu,k t)]’ dx > 0. (31)
by requiring that the approximation Ufx, t) satisfythe heat 0
conduction equation (2) on an algebraic average basis over
If U(x, t) were the exact solution Tfx, t), then E(t) would be
the domain 0 Q x $ fit). Mathematically, this domain
identically zero by (2). The effect of the square under the
condition can be stated thus :
integral is to prevent algebraic canceling of errors of opposite
sign and to ma&y the importance ofspaceregions wherein
j [kU&x. t) - pcU& 41 dx = 4 ml
U(X, t) does not closely satisfy the heat conduction equation
 SHORTER COMMUNICATIONS 2427

(2). Let E.(t) be the result corresponding to u&z, t), pats. the relative acYxmcy of sudvc approximations even
Then when the exact sofution is not known.
E&t) = e,. Uc[T, - B I2 . (at)-*, e, = constant. (32)
The lower the value of E&t) or of e, the more accurate the 1. V. S. ARPAC~,Conduction Heat Tram+, pp. 78-83,
solution is expected to be. This expectation appears to be 159-161, 474-475. Addison-Wesley, Reading, Mass.
borne out by a comparison of the V&, t) with ?Tx, t). (1966).
A graph cotdd be used to show the V&z, f) and ZTx,t) asa 2. B. A. BGLEY,A&y&r of Pr&i~ of Heat Conduction
function of the single independent variable Y=(~/~(~I)). and Me&&g, Proceedings of the 3rd Symposium on

Table 2. Error txrsw position when using an approximation U,

scaledermr=~~,-q/lT,-I;) X
-- -T-l;
Y"- &4ar)
T.-T n=l,2L n=3LR ?I - 3t n = 3R n==2 n=2R n==4

0.0 omwJ 0.0

01
02 07772 BO227 0.0867
tEiz
Ml119
40050
-0oo27
08003
oao70
0.0032 01
0.2
@3 06713 0.0286 01253 00152 ooo43 zzz @0145 03
@4 05716 GO283 QlS85 O.Ol54 0.0141 00198 0.0222 z O-4
05 04795 0.0205 0.1852 0.0122 0.0246 (30264 00285 t35
06 @3961 00038 Q2045 OiX)56 00443 00310 0.0323 z!; 06
0.7 0.3221 -0.0221 02161 -OQO4O W421 00328 09330 ooo21 0.7
08 02578 - 0.0578 02199 -Go159 0.0474 OQ316 00303 -06028 @8
u9 02030 -0.1030 02164 -00288 00580 00276 00245 -O@oo84 0.9
1-O 0.1572 -&1572 02063 -00411 0+@9 00213 00159 -00141 1.0
1.1 O-1197 * @1907 -0Q512 0@474 00133 OQO53 -00191 1.1
1.2 00896 * 01706 -00570 09430 -0oo63 -00229 1.2
1.3 0.0659 t 01475 -90564 00373 -Z -@OlEl -0.0251 1.3
1.4 @0477 + Q1224 -0.0475 0.0309 -Ml09 -W291 -00253 1.4
1.5 0.0338 * 00967 * 0.0243 -0.0159 1 -Ml235 15
1.6 a0236 * 00716 * 00180 -00178 L -0Q201 l-6
l-7 OQ162 * 00480 * 00124 -0.0158 * -00155 1.7
l-8 OQ109 * 0.0269 * 0.0077 * * -0.0109 1-8
1.9 ooO72 * 08091 * Wo40 * * * 1.9
2.0 OQD46 * -Oil046 * OOOlS * * * 2.0

e, = 0.1667 oo907 0.0417 0.0233 0.0192 00175 m53 e.

* Note: Blank entries in the table correspond to x > p&). For blank entries, U, = x and the scaled error is exactly
equal to the ne&ve of the value shown in the second column above.

The various curves on the graph would however lie very Naval Structural Mechanics, pp. 274-278, 295-296.
close to each other and would cross, making the graph Pergamon, Oxford (1963).
difficult to interpret. In lieu of the graph, Table 2 presents the 3. H. S. C-w and J. C. JA~G~JLConduction of Heat ia
scaled error [(U, - Tlft T, - QJ as a function of Y So&&. 2nd Ed., Chapter 11. Oxford University Press,
The relatively high accuracy of Ujl was not anticipattd, London (1959).
but was easily recognized by virtue of the low value of ess. 4. SUNG HWAN Cno and J. E. S-ND, Melting or
[Those involved in boundary layer analysis may wish to freezing of fmite slabs, ASME Paper 68-WA/HT-37
keep in mind t&t UIR is a highly accurate solution which (Dec. 1968).
5. T. R. GooDMAN, The Heat &lance integral and Its
does nor satisfy the domain condition (27) or, equivalently, Appiic&on to Problmu hwohing o Cw of Phase,
the smoothing condition (30)-3 1957 Heat Transfer and Fluids Mechanics Institute,
In conclusion, use of the numerical error criterion E(t) 19-21 June 1957. Stanford University Press. (Compare
appears to make possible the straightforward evaluation of the discussion of (30-33) with that of (IW) of [71.)
2428 SHORTER COMMUNICATIONS

6. T. R. CbODMAN, The heat balance integral-further 13. D. LANGFORD, The freezing of spheres. Int. 1. Heat
considerations and refmements, f. Heat Trmsfkr 83C, Mass Tram&r 9(S), 827-828 (1966).
83-86 (1961). (The discussion of accuracy is very 14. D. LANGFORD,New analytic solutions of the one-
instructive.) dimensional heat equation for temperature and heat
7. T. R. GOODMAN,Advances in Heat Transfer, Vol. I. flow rate both prescribed at the same fixed boundary
Academic Press (1964). (with applications to the phase change problem), Q.
8. T. R. G0oDIuAN and J. J. SHEA,The melting of linite Appl. Math. 24(4), 3 15-322 (1967).
slabs, J. Appl. Mech. 27(l), 16-24 (1960). 15. D. LANCWRD, Pseudo-similarity solutions of the one-
9. A. W. D. HELLS.A generalizat integral-profik method dimensional diffusion equation with applications to the
for analysis of unid~e~iona1 heat flow during sotidili- phase change problem, Q. Rppf. Mark 25(I), 45-52
cation, Trans. Met& Sot. A.I.M.E. 245, 1471-1479 (1967).
(1969). (Compare Poots.) 16. M. N. ~ZISIIC, Bow&y Value Problem of Heat
10. A. W. D. Hr~rzz and M. R. Mooan, The solidtficatton Conduction, pp. 301-311. International Textbook Com-
of pure metals under unidirectional heat flow conditions pany, Scranton, Pemrsylvania (1968).
---I. Solidification with zero superheat, Trans. Metaii. 17. 0. Peon, On the application of integral methods to the
Sot. A.Z.M.E. Us, 1481-1492 (1969). solution of problems involving the solidiAcation, of
Il. P. HEYCAK, Problem of ~li~~~tion with Newton’s liquids initially at fusion temperature, Int. J. Heat
cooling at the surface, A .J. CXE. n9(5), 585-589 (1963). hfuss Tram&r 5, 525-531 (1962).
12. D. LAN~~BD, A closed form solution for the constant 18. V. N. VOLKOVand V. K. Lt-Onr~v, A refinement of the
velocity solidification of spheres initially at the fusion integral method in solving the heat conduction equation,
temperature, Er. /. Appl. Phys.17(2), 286 (1966). Heat Transfer, Sov. Rcs. 2(2), 41-47 (1970).