R. B.

Holmberg
AB Svenska Flaklfabrlken,
Jbnkbping, Sweden
Heat Transfer in Liquid-Coupled
Indirect Heat Eichanger Systems
The theory of liquid-coupled indirect heat exchanger systems has been studied to ascer-
tain optimum criteria with respect to the coupling-liquid flow rate and the distribution
of total heat transfer area between the hot-side and cold-side exchanger units in the case
of counterflow arrangement. The optimum coupling-liquid capacity rate is derived and
given as a function of the over-all capacity rate ratio and the Ntu ratio between the two
exchanger units. For this optimum liquid capacity rate together with the proposed over-
all number of transfer units, it is shown that the over-all heat transfer effectiveness of
the liquid-coupled system can be expressed in the ordinary form for individual exchang-
er units in true counterflow.

Introduction locity is not treated. Furthermore, thermal optimum distribution

of total heat transfer area between the hot-side and cold-side ex-
This study of liquid-coupled indirect heat exchanger systems changer units is determined for two given conditions in the case of
concerns heat recovery for air conditioning. The results are, how- optimum liquid capacity rate. The first condition is constant total
ever, expected to be applicable to indirect exchanger systems in heat transfer area, which in a qualitative way corresponds to given
other fields. The system consists of two heat exchanger units to- total exchanger volume or weight. The second condition is con-
gether with a circulation system for the coupling-liquid (see Fig. stant total UA value perhaps of less interest in practice [2].
1). With this arrangement the circulating liquid, selected with re- The optimization rules in question can, for example, be used in
spect to the actual temperature level, transfers thermal energy sizing liquid-coupled heat recovery systems in air conditioning. In
from the hot fluid to the cold fluid. this case the capacity rate and the heat transfer coefficient of the
In such a system the coupling-liquid flow rate or capacity rate hot fluid can be replaced by "effective" values of these parameters
(mass flow times specific heat) is an independent variable and can in order to take water vapor condensation in the hot fluid into ac-
be chosen to maximize the over-all effectiveness for given heat count. The effective values includes the effect of latent heat trans-
transfer areas, given hot-fluid and cold-fluid capacities and given fer rate and are obtained by dividing the capacity rate and the
overall heat transfer coefficients. Here it is assumed that the free heat transfer coefficient by the "sensible heat factor" [4], which
flow area of the liquid in practical exchanger units can be chosen means the quotient between sensible and total heat transfer rate.
so that economical optimum liquid velocity is obtained. Thus, the In this particular case the procedure of maximizing the over-all ef-
optimum liquid capacity rate is determined for constant liquid ve- fectiveness becomes iterative, since the sensible heat factor is not
locity and variable free flow area, i.e., the overall heat transfer known from the beginning. The performance of heat recovering
coefficients are kept constant. Kays and London [1, 2] 1 derived in systems in air conditioning is described in reference [5].
connection with heat recovery in gas-turbine plants the optimum
liquid capacity rate in the special case of equal hot-fluid and cold- Effectiveness—Ntu Relations
fluid capacity rates and counterflow arrangements. They found The over-all effectiveness e0 of the liquid-coupled system is de-
that the optimum liquid capacity rate is equal to the capacity rates rived with the assumptions that the heat exchange with the sur-
of the hot and cold fluids, which themselves are equal. roundings of the system and the heat generation of the pump are
In the present paper this thermal optimum capacity rate is de- negligible. This is usually valid in practical cases.
rived without restrictions in fluid capacity rates, while the inde- The definitions of the over-all effectiveness t0, the effectiveness-
pendent problem to determine the economical optimum liquid ve- es (h and ec of the hot-side and cold-side units, respectively, are as
follows:
1
Numbers in brackets designate References at end of paper. Q Q
€„ = " « , = • (l)
Contributed by the Heat Transfer Division for publication in the JOUR- -'min,ouo •'minj/i ft - / niiB,c c
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer
Division March 20, 1975. Paper No. 76-HT-E. where the heat transfer rate is

Journal of Heat Transfer NOVEMBER 1975 / 499

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the units according to [1, 3] can be presented in the well-known
*hl '1,2
-#»» Ch form
Hot-side
1 - e x p [ - J V t u , „ ( l - R$\
exchanger
l
Ll >L2 for R„ *1,
l-i^expRv^d-iO]
(4)
Nf tu,n
"-" for Rn = 1,
Y+ Nf„
Circulating Coupling
C
WJ pump
L I liquid with the number of transfer units
N U
tu,n ~ nA,/CmiUl„
•LI '12

Cold-side
and the capacity rate ratio
exchanger
<c2 'cl

where n stands for c o r / i .

Fig. 1 Schematic view of llquld-eoupted heat exchanger system in coun-
The six equations for the over-all effectiveness can be expressed
terflow in the following equations, by inserting the relations for e>, and ec
(4) into the equation (3):

1
1 — Rr (I , RT
L
1 — a -b
Q = Cn^th = CAi, = CrAtr. (2) for RL = 1 and R0 * 1,
R„b
Cmin.o is the smaller of the hot-fluid and cold-fluid capacity rates, Nf. 1- b
Ci, and Cc. Cmi„?/, stands for the smaller of C;, and the liquid capac- (5)
1
ity rate CL in the hot-side unit, while Cmin,c is the smaller of the
capacity rates in the cold-side unit, Cc and CL. The temperature l_Z_i?afL + •"•
JV
tu
differences 6a, 8h, 0C, At/,, Atc and ML are clear from Fig. 2.
The relations (1) and (2) together with the temperature relation 1
e„ = •
1
= Bu + - Atr \i mill \j max + 1
obtained from Fig. 2, give the following expression for «„ as a func- where
tion of th and tc: a = e x p [ - A ' e u r a l " ( l - RL)}, b = exp[-Nt™(RL - R0)],

(3) Nt. = (M)miVCmln)0, Nta«"* = ( K 4 r * 7 C m l l n i 0 ,

Cjnin,o/^min,h _|_ ^ i r n , o ' min,c _ min,o
=
•"o ^min,o/^miH,o> "-L ~ CmiTil0/CL_
~cZ
Depending on the ratio between the three capacity rates this The superscript min indicates that exchanger unit, which has
equation results in six different expressions, which have been de- the smaller of the hot and cold fluid capacity rates, Ch and Cc,
rived in [2]. These equations require constant capacity rate values while max indicates that unit, which has the larger of Ch and Cc.
throughout the liquid-coupled system. In most practical cases this The subscripts min.o and max.o indicate the smaller and the larg-
requirement can be fulfilled by referring the capacity rates to rep- er of the hot-fluid and cold-fluid capacity rates, respectively, ex-
resentative temperatures in the system, for example, mean tem- cluding the coupling-liquid capacity rate.
peratures. Furthermore, the equations can be shown to be valid
without restrictions in flow arrangements. The hot-side and cold- Optimum Coupling-Liquid Flow Rate
side units can be arranged in counterflow, crossflow, etc., and need The coupling-liquid flow rate or capacity rate CL is chosen to
not be arranged in the same way. However, in liquid-coupled sys- maximize the over-all effectiveness e0 for given Ntu values. This
tems the exchanger units are usually arranged in counterflow, or optimum value of CL is obtained by partial differentiation of equa-
nearly in counterflow, and therefore only the counterflow arrange- tion (5) for RL ^ I and RL ^ R0 according to at0/aCL = 0 or
ment is discussed in the following. atJd(\IRL) = 0.
For the case of true counterflow, the effectivenesses th and tc of The derivation results for constant values of Ntumin and NtumBX

.Nomenclature.
A = heat transfer area R„ = over-all capacity rate ratio, C mln , 0l/ uid-coupled system
A* = heat transfer area ratio, Amin/Amax t> max.o
At = total heat transfer area, Amin + A m a x At = fluid-temperature change in passing Subscripts
At = dimensionless At, Um'mA,/Cmm,0 through exchanger c = cold-fluid or cold-side exchanger unit
C = heat capacity rate (mass flow times U = unit over-all heat transfer coefficient h = hot-fluid or hot-side exchanger unit
specific heat) U* = unit over-all transfer coefficient L ~ coupling-liquid
Ntu = number of heat transfer units, UA/ ratio, [/min/f/max o = over-all liquid-coupled system
t'min (UA), = total UA value, (UA), = (UA)min min.re max,n = the smaller and the larger
Ntu° - over-all number of transfer units, + (UA)ma% of the two fluids in exchanger re, respec-
1/(1/Ntumm + l/Ntuma*) e = unit heat transfer effectiveness, actual tively, where n stands for c, h, or o
Ntu* = Ntu ratio, N,umin/N,umM heat transfer rate divided by maximum
. NtUt„ = sum of Ntu, Ntumin + Ntumm possible heat transfer rate Superscripts
RL - liquid capacity rate ratio, Cmm.o/Cz. to = over-all heat transfer effectiveness min, max = exchanger unit with the small-
Rn = capacity rate ratio, C m j n] „/C maXin , 0 = temperature difference between inlet er and the larger of the hot and cold fluid
where n stands for c or h fluids to exchanger units or over-all liq- capacity rates, respectively

500 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Cold-side exchanger Hot-Side exchanger

Fig. 2 Schematic temperature conditions for counterflow In heat ex-

changer units

in (a-b) (1-a) (1-fa) = (1-a) 2 binb- (1-6) 2 a In a. where

This equation has only one solution a = b within the region of Nt„ N„ fj max
definition RL P* 1 and RL ^ R0, that is a ^ 1 and 6 ^ 1 . Thus the
relation a = b gives the maximum value of e0. This equality results is the over-all number of transfer units for the liquid-coupled sys-
tem. Note that JVto°, defined as this reciprocal sum, allows e„ to be
expressed in the ordinary form for individual exchanger units in
M min 4. D « max
Rr. true counterflow according to equation (4). The graphical repre-
Nf. + Ntu™ sentation of this equation is illustrated in Fig. 3 [1, 3]. The recipro-
(KA) n cal definition of Ntu° means that the coupling-liquid can be con-
where Nt \r max
(KA)" sidered as an ordinary thermal resistance in the case of optimum
liquid capacity rate or flow rate.
In order to investigate how a nonoptimum liquid capacity rate
{UAY + {UAT^h
affects the over-all effectiveness, t0 as a function of 1/RL is given in
CI>r opt "(tZ/4) min + («A) n C„ Fig. 4 and 5 for some values of the parameters R0, Ntu° and Ntu*
Consequently, the optimum liquid capacity rate in true counter- defined in the foregoing. It can be seen from the figures that the
flow is obtained as the reciprocal mean value of the hot-fluid and optimum becomes critical only for large Ntua values and that it be-
cold-fluid capacity rates in the case oiNtu* = 1. comes less critical when R0 goes to zero.
This optimum value RL,OPU that is o = 6, gives with (5) the fol-
lowing simple expressions for the over-all effectiveness Optimum Relation B e t w e e n Hot-Side and Cold-Side
Exchanger Units
1-a _ 1 - exp[-iy,„°(l - Rn)] In the case of optimum liquid capacity rate the area ratio A* =
for R0 * 1,
1-R0a 1-R0 expHV(l - jy ] A m , n M m a x for the two exchanger units is chosen to maximize the
N,in_ over-all effectiveness e0 for given total heat transfer area At = Amm
for /?„ = !,.
1 +Nt. + Amm. This optimum value of A* is obtained by partial differen-

— —r | 1

^^"0.25"
i r
—- 1 l 1 1 1

- Counterflow
- K0 = 1. A7„-1 -
- /o//\^-—17oo " - f 2 "~"~^^ ^
- A — IS \S>~~^ —
yOyyy -
Counter! ow 1/ '
0,4 «/. = « t opt f/*^ 0.75 —————— : _ _
- / - f 0.5
r. -
~1 -
-
-1
L i i i
0,5 1,0 1.5 2.0 2,5 3,0 3,5 4,0 1 1 1 l I
1,5 2.0 2.5

R-L opt I^L ~ Q / Q , opt

Fig. 3 Over-all effectiveness as a function of over-all number of transfer
units with over-all capacity rate ratio as a parameter for optimum coupling- Fig. 4 Optimum conditions for coupling-liquid capacity rate with Nlu° as a
liquid capacity rate parameter

Journal of Heat Transfer NOVEMBER 1975 / 501

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 1 o 1 I
0.5
——

J.'—-
- _ - B

- _ ''°
';/ 0
-// ^ ____0£~ —•
. /i/ //v-—
/^_______ -——»^
~«7=To —
— 1// Counterflow
Rl.optIRL ' >
-11/ — "lo.o = 4
-

_ N,u0 = 12
-
i i 1 1 I
2.5 1.0 t,5 2,0 3,0
= N
RL.op,/RL'CL/CL ,opt A' I A'opt iu
Fi 7
Fig. 5 Optimum conditions for coupling-liquid capacity rate with /?„ as a 9- °P«mum conditions for heat transfer area distribution with given
i o t a i UA
parameter value, In dimenslonless form N,Uj0

tiation of cp or of Ntu°. = 1 the curves are valid also for given At, in dimensionless form At
- Umi"Ai/Cmm;o. Fig. 8 shows how U* w* 1 influences the optimum
,4*
With Nt,° = character in the case of given At. It is seen from these figures that
C m i „, 0 ' ( 1 + U*A*){\+~A*)' the optimum is very flat for all values of the actual parameters.
{
equation
^~K* ° gives A*°>*= v^*) 0 - 5 .
= For example the overall effectiveness is decreased less than 4 per-
cent if the area ratio is varied in the range of 0.5-2.0 times the op-
This optimum relation is valid for arbitrary values of the over- timum value.
all capacity rate ratio Ra. It is also to be noted that the problem of
minimizing the total transfer area for a given overall effectiveness
is identical with the described problem. Conclusions
A sometimes used alternative, perhaps of less interest in prac-
tice [2], is the optimization of A* for given total UA value, (UA)t = The derived optimum coupling-liquid capacity rate Cz,j0pt (or
(UA)min + (UA)m™. The partial differentiation of Ntu° gives in RL,OPI) and the overall number of transfer units Ntu° (the recipro-
this case A*opl = l/U* for arbitrary values of R0 and optimum liq- cal sum of Ntumm and Af (u max ) are found to give the same e-Ntu
uid capacity rate. When U* = 1 this result A * o p t = 1 agrees with relation for the over-all effectiveness in counterflow as for the ef-
that of given At. fectiveness of the individual exchanger units in counterflow. The
In order to investigate how a nonoptimum distribution of the reciprocal expression of Ntu° means that the coupling-liquid can
total heat transfer area affects the overall effectiveness, t0 as a be considered as an ordinary thermal resistance. The optimum
function of A*/A*opt is given in Fig. 6 and 7 for RL = RL,OPI and value of the coupling-liquid capacity rate is critical only for large
some values of the parameter R0 in the case of given (UA)t, in di- Ntu" values. The optimum area ratio A*opt for the two exchanger
mensionless form Ntu.o = (UA)t/Cmin,0 = Ntum[n + N,u™\ For U* units is not critical for any value of the actual parameters.

1.0
1 i i ! 1

N = 16
tr_ -_£,5_
tu.o
0,8 — 1,0
—JJ —
__ .2,0 ~ - -

r
8

3 - __
2
Counterflow
«<, = '• « / , , o p r / « i = '
Cot nierflow ,4, = 4
=
'• R
L,opt RL-1 A , = 16

I 1 i 1
3,0 2.0

A'/A; opt
Fig. 6 Optimum conditions for heat transfer area distribution with given Fig. 8 Optimum conditions for heat transfer area distribution with given
total UA value, in dimenslonless form N,uo total transfer area At

502 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Acknowledgment 2 London, A. L., and Kays, W. M., "The Liquid-Coupled Indirect-
Transfer Regenerator for Gas-Turbine Plants," TRANS. ASME, Vol. 73
The author wishes to express his thanks to AB Svenska Flaktfa- 1951, pp. 529-542.
briken for permission to publish this paper and to Dr. 0 Strinde-
hag, the head of the Research and Development Laboratory, for 3 Bosnjakovic, P., Vilicic, M., and Slipcevic, B., "Einheitliche Berech-
nung von Rekuperatoren," VDI-Forschungsheft, No. 432,1951, pp. 5-26.
many valuable discussions.
4 Ashley, C. M., "Psychrometric Factors in the Air Conditioning Esti-
References mate," Trans. ASHVE, Vol. 55,1949, pp. 91-110.
1 Kays, W. M., and London, A. L., Compact Heat Exchangers, 2nd ed., 5 Strindehag, O., and Holmberg, R. B., "A Liquid-Coupled Heat System
McGraw-Hill, New-York, 1964. for Recovery From Exhaust Gases," to be published.

Journal of Heat Transfer NOVEMBER 1975 / 503

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A. P. Watkinson
O. Martinez Scaling of Heat Exchanger Tubes
Department of Chemical Engineering,
University ol British Columbia,
by Calcium Carbonate
Vancouver, B. C. Canada
Scaling of copper heat exchanger tubes has been studied under conditions that promote
rapid and severe scaling. Artificially hardened water of high dissolved and suspended
solids is recirculated through a heated test section operated at constant steam tempera-
ture. The effects of flow velocity, tube diameter, and bulk temperature on the. asymptotic
fouling resistance have been determined. Results are interpreted in terms of mathemati-
cal models of the scaling process.

Introduction ty and bulk temperature in tubes of different diameters under con-

stant wall temperature conditions. The velocity and diameter ef-
The fouling of heat transfer surfaces which has long been an op- fects were of particular interest. The choice of tube diameter in se-
erating problem in heat exchange equipment is beginning to re- vere fouling situations is not the same optimization problem be-
ceive the attention of research workers. Recent progress in the un- tween heat exchanger size and pressure drop as for the case of
derstanding of fouling processes has been summarized by Taborek, clean tubes, since the fouling resistance itself is a strong function
et al. [1, 2]. 1 Laboratory studies have been carried out on fouling of of velocity [1-4]. Some compromise is reached between the rule-
gas oils [3] and iron oxide suspensions [4]. The mechanisms of cal- of-thumb that large diameter tubes should be used in fouling sit-
cium carbonate and calcium sulphate scale formation have been uations to prevent tube blockage once fouling starts, and the posi-
investigated [5, 6]. Heat Transfer Research Inc. has carried out a tive effect of high velocity and hence smaller tube diameter in re-
wide ranged study of cooling tower water fouling [1, 2], the quanti- ducing both the fouling resistance and the tube side convective
tative aspects of which have not been published. The velocity and film resistance. One aim of the work was, therefore, to determine if
temperature dependence of the asymptotic fouling resistance, for there was a significant tube diameter effect on the asymptotic scal-
example, are not given. Although some generalizations can be ing resistance.
drawn from this work, as might be expected, the fouling process is
to a large extent specific to the chemical system involved. Thus, in
Experimental
the foregoing studies the rate of fouling, or the fouling resistance
The apparatus and method have been described in detail else-
has been found to depend somewhat differently on process param-
where [7]. The apparatus consisted of a steam condenser contain-
eters, and in most cases the rate controlling steps in the fouling
ing two copper tubes in parallel through which artificially harden-
processes are different.
ed water was circulated. The inlet liquid temperature was held
Even the relatively simple problem of scaling of CaC03 is not constant in most of the work at 329 ± I K . Tap water at a fixed
•yet completely understood. Hasson [5] has shown that under cer- flowrate was circulated through the tubes for long enough to estab-
tain conditions the deposition of calcium carbonate scale can be lish the heat transfer coefficient under clean nonfouling condi-
mass transfer controlled. In a series of experiments at uniform tions, then salt solutions were added to give an initial composition
heat flux, and at velocities below 0.9 m/s the amount of scale de- of 1.35 g/1 NaCl, 1.39 g/1 CaCl 2 and 1.23 g/1 NaHC0 3 . This solution
posited increased uniformly with time and the rate of scaling in- when heated above temperatures of 330 K gave very rapid scale
creased with velocity. A previous study in the author's laboratory formation on the tube walls. Five copper tubes from 5.3 to 20-mm
under constant wall temperature conditions where an asymptotic ID were tested. The film heat transfer coefficient was calculated
fouling resistance occurs indicated that at velocities above 0.8 m/s from the outside wall temperature measurements:
the asymptotic fouling resistance decreased with increasing veloci-
ty, in agreement with the qualitative results reported by Taborek, WCn
h=- - In ((Two ~ Tbi)/(TwQ - Tb2)) (1)
et al., for cooling tower water scaling.
The objects of the present work were to study the scaling of salt After addition of the scaling salts, the outlet fluid temperature de-
solutions due to CaCC>3 formation as a function of tube side veloci- creased as scale deposited in the tubes and the rate of heat transfer
dropped. The recirculating liquid was analysed for dissolved and
1
suspended solids, and pH. Suspended solids concentrations were
Numbers in brackets designate References at end of paper. about 400 ppm 15 min after salt addition, and increased by about
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer 100 ppm during the experimental run of 4 hr. Dissolved solids con-
Division June 12,1975. Paper No. 76-HT-H. centrations decreased slightly from the initial value of 3000 ppm,

504 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 7000 00010 1 1 1 1
D (m)
A aA 0020
00141
00005 1 \ a
V
0 - 0 101
00080 ~

! /?>%s °
00053

i i ' i <b $
i J ' i O
A® / i
O0002 - -
/ 1 ^
d ,' ^
- * \
376 K
O\
I..r
T
b,
330 K
i
000005 0-2
1 1
0-5 10 20 30
\ 50
i
m/s
10 2-0 30 40 50 6 0 Fig. 2 Effect of velocity on fouling resistance
Time hr
Fig. 1 Typical plot of heat transfer coefficient versus time
first increases with velocity, goes through a maximum and then de-
creases. Possible reasons for the velocity maximum were discussed
and the pH increased typically from 7.7 to 8.3 during each experi- elsewhere [7], As the tube diameter is increased, the maximum Rf*
ment. These operating conditions allowed a very rapid approach to increases, and occurs at lower velocity. Fig. 3 shows that Rf* gener-
the asymptotic fouling condition. Such accelerated tests while not ally decreases for all tubes at Reynolds numbers above 12,000.
necessarily being valid to establish design values of the asymptotic In Fig. 4, R/* is plotted versus tube diameter at three values of
fouling resistance, are of considerable use in evaluating the effects the bulk velocity. Although there is some scatter, the trend of the
of certain variables and geometries on the fouling tendency. data is clear. Increasing tube diameter from 8 to 20 mm causes a 33
percent drop in Rf* at V = 1 rn/s. This effect of diameter is con-
Heat Transfer Results founded by bulk temperature effects, since at fixed velocity and
A plot of typical heat transfer results with time as shown in Fig. inlet temperature increasing the tube diameter results in lower
1. The heat transfer coefficient h, remains constant for the first bulk liquid temperature and hence scale temperature along the
hour before the salts were added. As expected, clean coefficients tube.
increase with increasing velocity and decreasing tube diameter.
The effect of bulk temperature was determined in a separate se-
Immediately after salt addition a rapid decrease in h begins and a
ries of experiments in a 10-mm dia tube. The inlet temperature
steady-state value h* is reached after about 2 hr. Earlier work
was varied from 320 to 360 K. Results are shown in Fig. 5. The as-
using this technique showed that the asymptote was stable at least
ymptotic fouling resistance goes through a maximum with increas-
for 24 hr [7], and further addition of scaling salts had no effect on
ing Tb°. The reason for the maximum will be discussed later. Runs
the value of h*. The asymptotic fouling resistance Rf* is calculated
of different tube diameters and fixed Tjj also fall on the same line.
from the steady-state value, and the initial value of the heat trans-
This suggests that the observed diameter effect (Fig. 4) is a result
fer coefficient
of the average bulk temperature decrease as D increases.
Rf* = (llh* - l//i°) (2) For design purposes the effect of diameter at either fixed flow-
rate or fixed pressure drop is more important than at fixed veloci-
All experiments were run with the wall and steam chest tempera- ty. At equal flowrates as tube size is increased any decrease in Rf*
tures held constant with time so that as the deposit builds up, the due to the tube diameter effect is far outweighed by the increase in
temperature at the deposit/liquid interface drops. This drop de- Rf* due to the drop in tubeside velocity (Fig. 2). Thus, for exam-
creases the tendency for scale formation since the driving force for ple, at V = 1.5 X 1 0 - 4 m 3 /s tubes of 8-20 mm could be selected
further scaling, the supersaturation, is lowered. Deposits were with velocities varying from 3 to 0.48 m/s, respectively. Fig. 6
found [7] to consist essentially of CaCOa with loss on ignition at shows that the fouling resistance of the 20-mm tube is about five
1000°C of 42.7 percent and calcium content of 50.4 percent. Exam- times that of the 8-mm tube. The clean coefficient is also larger for
ination of the tubes after a run showed the deposit to be made up the small diameter tube.
of an easily removable sludge on top of a compact crystalline layer.
The effect of liquid velocity on the scaling of tubes of different Mathematical Models of Scaling
diameters is shown in Fig. 2. Fouling resistance for each tube at The predictions of current models of the fouling process were

-Nomenclature-
a = constant m = constant Subscripts
A = inside area of tube m 2 n = constant b = bulk liquid
b = slope of temperature solubility curve r = statistical correlation coefficient 1 = inlet of tube
mol/m 3 K Rf = fouling resistance (J/sm 2 K) _ 1 2 = outlet of tube
B = constant in equation (19) 1/s Rg = gas constant KJ/kg-mol K D = deposition
C = concentration mol/m 3 t — time hr R = release
Cp = heat capacity J/kg K T = temperature K n = constant
D = inside tube diameter m V = velocity m/s s = scale
E = activation energy KJ/kg-mol V = flowrate m 3 /s w = wall
/ = friction factor W = mass flowrate kg/s
h = inside heat transfer coefficient J/s x = deposit thickness m
m2K (3 = function Superscripts
k = thermal conductivity J/sm K T = shear stress at tube wall N/m 2 * = asymptotic value
kr = crystallization rate constant kg/m 2 s * = term in equation (6) m/s 0 = at time zero
(Mol/m 3 )" p = density kg/m 3 — = average value

Journal of Heat Transfer NOVEMBER 1975 / 505

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 examined for comparison with the experimental data. The original 00010 1 1 _
Kern-Seaton [8] model for dirt deposition predicts a strong diame- T w = 376 K
\»330 K
ter effect at fixed velocities:
; 0 0005-
(3)

whereas the models of Watkinson and Epstein developed for par-

? 0 0003 .-^s^ V=0~S5 m/s

~~V~~Y= 1-0 m / s
ticulate fouling [9] predict 00002
(4)
Rf* ^ V * 1-6 m/s

0 0001 1 " 1
where m = 1, 2, or 3 for deposition steps controlled by transfer of •005 010 •020
particles to the surface adhesion at the surface and both steps re-
spectively. If the tube is hydrodynamically smooth, equation (4) Fig. 4 Effect of lube diameter on fouling resistance
predicts Rf* increases with D 0 1 2 5 to D 0 - 375 depending on m
through the effect of Reynolds number on the friction factor. Oth-
erwise Rf* is expected to be independent of tube diameter. Apply-
ing equations (3) or (4) Rf* should correlate with V and possibly
and wrote the scale surface temperature in terms of the wall tem-
with D. Experimental data for NRe > 12000 were correlated within
±14 percent by the equation perature.

Rf* = 0.000101/V L33 O°- ! (5) Ts Tb = (9)

~ (llhxlks)
It is evident that equations (3) or (4) with integer values of m do Combining (6)-(9) we obtain:
not fit the experimental velocity dependence.
Expressions (3) and (4) are based on the Kern equation [8] d^ = kL MTw ~ Tb)\
-I — at (10)
where the rate of scale formation is given as the difference between dt ps\i + hxlks
a deposition term 3>D and a release or removal term <J>R This equation differs from those used in other models in that the
deposition term is not constant with time. The asymptotic fouling
- = *D - *# (6) resistance is obtained by setting dx/dt = 0 in equation (10) and
dt
putting Rf* = xlks and T = fV2p/2
The release term was given by Kern as being dependent on the
shear stress at the wall and the deposit thickness i.e., 4>R = arx. Rf*(l + Rfh)»-2bnkf"~Tb)n (11)
Tests showed that in the present study a fouling layer deposited at
a low velocity could be substantially removed by a short surge of Now the crystallization rate constant kr is a function of the scale
high velocity fluid. This implies that a removal or suppression surface temperature Ts, and would be expected to follow an Ar-
force does exist and argues for the inclusion of a release term in the rhenius equation:
mathematical description. The increase in suspended solids con-
tent over the course of the run suggests that particulate fouling is kr = kro exp -E/RgTs (12)
not the major deposition process occurring, but rather the deposit
is formed from solution. The deposition term thought appropriate Calculations using equation (9) show that over the course of the
in the present case was originally formulated by Reitzer [10] who experiment Ts drops typically from 375 K to about 350 K. Using
considered the rate of scale formation for inverse solubility salts in the activation energy of 25,150 kcal/g-mole given by Hasson [11]
evaporators to be dependent on the supersaturation raised to a the rate constant kr would be expected to drop by a factor of 13
power n. over the course of the run. This strong temperature dependence of
the crystallization process would tend to explain the very rapid ap-
dx proach to asymptotic conditions (Fig. 1) and the failure of at-
— =kr(Cb-Cs)«/ps (7) tempts to correlate Rf* with the wall temperature using an Ar-
dt rhenius type equation [7]. The latter could be more easily done if
No release or suppression term was given. He assumed that the in- experiments were carried out in a constant heat flux exchanger
verse temperature-solubility relationship could be approximated where Ts would remain constant with time.
by a linear equation Combining equation (9) and (12), equation (11) becomes:
• Cs = b(Ts - Tb) (8) 2krob" -E
Rf*(l + hRf*)"> -exp
p,ak. Tb*
Rs Tb* +
1 + hRf* /

00008
1
D Im) K IK) '
00006 - o 00053 330 T w = 376 K _
0-0006 a 00080 330 V • 1-0 m/s
A 00101 320 - 360
7 00141 330

00004 - _
00004

00002 \ -
00002

0 1 1 1 1
320 330 340 350 360 370
V
Fig. 5 Effect of average bulk liquid temperature at I = 0 on fouling resis-
Fig. 3 Fouling resistance versus Reynolds number tance

506 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 points fitted calculated predictions within 16 percent for re = 1 and
10.7 percent for re = 2. Equations (15) and (16) predict a specific
effect of V on Rf*. Using accepted dimensional equations for clean
film coefficients for water, the predicted slopes of the Rf* — V
plots were found to be —1.5 to —1.6 for n = 1, and - 1 . 3 to —1.4 for
re = 2 as shown in Fig. 7. Equation (16) then gives a prediction of
the velocity dependence that was confirmed experimentally (Fig.
2), whereas equation (15) does not.
The model also predicts the general shape of the Rf* — Tbo
curve of Fig. 5. In equation (13) it is evident that as Tb increases at
constant Tw the exponential term which describes the kinetics of
20
the crystallization process will increase, whereas the term (Tw —
10' V mVs
Tb)n which is a measure of the supersaturation or driving force will
Fig. 6 Fouling resistance versus volumetric flowrate
decrease. This results in a maximum value of the function Rf*(l +
hRf*)n. For the case of re = 2 , data of Fig. 5 and some other tests
run at different wall temperatures were plotted according to equa-
tion (13). Results shown in Fig. 8 are roughly linear, and lend sup-
port to the proposed model. A least squares fit of this data yielded:
(.TW - nr (13)
fV2 5.910 X 1019(T,„ - Tb*)2
R/*(l + hR/*)2 •• 2
exp-(17540/T s *) (19)
V
For the present purposes the functional dependence of the wall
and bulk temperature can be simplified to The temperature effect is somewhat larger than that found by
Hasson, who using a different analyses found E/Rg = +12,657 K.
Rf*(l + hRf*)n = ^ (14) It should be noted that while the form of equation (19) lends sup-
V2 port to the model,, it is not a recommended method to evaluate the
where ft, = f(Tw, Ts, Tb). From the literature on crystal growth activation energy. It is based on the average scale temperature at
[12] re is expected to lie between 1 and 2. For most sparingly solu- the end of experiment during which Ts has been dropping. A con-
stant heat flux exchanger would allow a more direct calculation of
ble salts re = 2 at high supersaturations. For n = 1 equation (14) is
the activation energy.
readily solved:
Equation (10) might be expected to give a good prediction of the

*~s[h3?r-'] (15) build-up of Rf* with time. Unlike Reitzer's cases for evaporator
scaling however, neither (Tw — Tb) nor the heat flux are constant
with time. The integration is further complicated by the presence
For re = 2 equation (14) becomes a cubic in Rf*
of the release term which Reitzer did not consider. Because of the
fl/3*+-fl/*2+^--^-=0 (16) complicated form of the integral it was decided rather to fit the ex-
' h ' re2 (hV)2 perimental data by the simple asymptotic equation of Kern [8]
which can be solved by algebraic or numerical techniques.
Rf = Rf*(l-exp(-Bt)) (20)
Agreement of data with the models was tested by calculating ft
and ft directly from equations (15) and (16) knowing Rf*, V and h
from the experiments, and testing to determine if the functions ft The fit is adequate (Fig. 9) but not too accurate especially near the
and ft were temperature dependent only. For data a t fixed inlet beginning of the scaling process. Equation (20) implies a constant
bulk temperature of 330 K, equations for ft and ft? were found, deposition term in equation (6) and is therefore inconsistent with
and the correlation coefficients determined. Bulk temperatures the Reitzer term. Initial rates were found to decrease with increas-
were below 350 K in all cases. ing V as expected from equation (10).

ft = 0.0000656 T 6 o - 0.02138 (r = 0.78) (17)

Conclusions
ft = 0.0002751 Tbo - 0.08489 (r = 0.80) (18) A study of the effects of velocity, bulk temperature and tube di-
ameter on the scaling from prepared hard water containing calci-
Inserting these functions into equations (15) and (16), respectively, um carbonate in dilute suspension was carried out in a. constant
allowed calculation of fitted results. Twenty experimental data wall temperature heat exchanger. At Reynolds numbers over 12000
the asymptotic fouling resistance varies with velocity to the power
—1.33. At constant inlet temperature, tube diameter was found to
have a weak effect on fouling resistance due to the difference in av-
erage liquid temperature as the diameter is changed. As the bulk
00005 temperature increases, the fouling resistance goes through a maxi-
mum and then decreases, because of opposing effects of scale sur-
face temperature and supersaturation. The results of this investi-
gation were consistent with a mathematical model of the scaling
00002 process where the deposition term was written in terms of Re-
itzer's model of scale formation in an inverse solubility salt, and
0-0001|- 0=001 m\
the release term due to Kern.

SLOPC = -l-55N
Acknowledgments
000005
This work was supported by a National Research Council of
Canada Grant. The apparatus was kindly loaned by Noranda Re-
I 2 3 4 search Centre, Pointe Claire, Quebec. This paper was presented at
V m/s the 79th National Meeting at the A.I.Ch.E., Houston, Texas in
Fig. 7 Predictions of scaling models March 1975.

Journal of Heat Transfer NOVEMBER 1975 / 507

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 1
00004
0-05 -
V=l m/s

002 - -
A
001 A "4A \ -

0005 A \ -
1 ! 1
2-76 2-80 284 2-88 292
10"
(K)
•hR f "
Fig. 9 Fit of typical experimental data to Kern's equation
Fig. 8 Fit of data to equation (13)

6 Hasson, D., and Zahari, J., "Mechanism of Calcium Sulfate Scale De-
References position on Heat-Transfer Surfaces," Ind. Eng. Chem. Fund., Vol. 9, 1970,
1 Taborek, J., Aoki, T., Ritter, R. B., Palen, J. W., and Knudsen, J. G., pp. 1-10.
"Fouling: The Major Unresolved Problem in Heat Transfer," Chem. Eng. 7 Watkinson, A. P., Louis, L., and Brent, R., "Scaling of Enhanced
Progr., Vol. 68, No. 2,1972, pp. 59-67. Heat Exchanger Tubes," Can. J. Chem. Eng., Vol. 52,1974, pp. 558-562.
2 Taborek, J., Aoki, T., Ritter, R. B., Palen, J. W., and Knudsen, J. G., 8 Kern, D. Q., and Seaton, R. E., "A Theoretical Analysis of Thermal
"Predictive Methods for Fouling Behavior," Chem. Eng. Progr., Vol. 68, No. Surface Fouling," Brit. Chem. Eng., Vol. 4, No. 5,1959, pp. 258-262.
7 1972, pp. 69-78. 9 Watkinson, A. P., and Epstein, N., "Particulate Fouling of Sensible
3 Watkinson, A. P., and Epstein, N., "Gas Oil Fouling in a Sensible Heat Exchangers," Proceedings 4th Int'l Heat Transfer Conference, Paris
Heat Exchanger," Chem. Eng. Progr. Sym. Ser., Vol. 65, No. 92, 1969, pp. 1970 Vol. 1, Paper HE 1-6 12 pp.
84-90. 10 Reitzer, B. J., "Rate of Scale Formation in Tubular Heat Exchang-
4 Hopkins, R. M., and Epstein, N., "Fouling of Heated Stainless Steel ers," Ind. Eng. Chem. Proc. Des. and Dev., Vol. 3, No. 4,1964, pp. 345.
Tubes by a Flowing Suspension of Ferric Oxide in Water," 5th International 11 Hasson, D., "Rate of Decrease of Heat Transfer Due to Scale Deposi-
Heat Transfer Conference, Tokyo, Japan, Paper HE 1-6, Sept. 1974, pp. tion," Sonderdruck aus Dechema—Monographien Band, Vol. 47, 1962, pp.
180-184. 233-252..
5 Hasson, D., Avriel, M., Resnick, W., Rozenman, T., and Windreich, 12 Konak, A. R., "A New Model for Surface Reaction-Controlled
S., "Mechanism of Calcium Carbonate Scale Deposition on Heat-Transfer Growth of Crystals From Solution," Chem. Eng. Sci., Vol. 29, 1974, pp.
Surfaces," Ind. Eng. Chem. Fund., Vol. 7,1968, pp. 59-65. 1537-1543.

508 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 S. KAO
Engineering Department,
Mobil Research and Development Corp.,
Dallas, Texas

fer Rate
This paper presents a matrix solution technique for multishell, multitube-pass heat ex-
changers with variable heat transfer film coefficients and heat capacity rates. Using this
method, a series of case studies has been made, whose results are used to check the valid-
ity of certain facets of the normal design practice. Some parameters have been examined
and discussed in details including the applicable range of Colburn's caloric temperature
factor, effect of shell nozzle orientation, local coefficient of the multipass exchangers
with tube side in laminar flow and flow instability phenomenon across the viscous flow
coolers.

Introduction rameters on heat exchanger performance and (iii) to re-examine

the behavior of tube side laminar heat transfer inside shell and
Shell and tube exchangers are commonly designed using the tube heat exchangers.
MTD chart by Bowman, et al. [2],1 which assumes constant heat
capacity rates and heat transfer coefficients. In many applications, Mathematical Formulation and Solution
these assumptions are not valid. For evaporators and condensers, I Single Shell, TEMA E Construction [7]. Let the ex-
both capacity rate and coefficient may undergo substantial changer have one shell pass and N tube passes with type I flow ar-
changes. In the case of viscous liquid exchangers, whose tube side rangement as shown in Fig. 1. Across the surface element dS of
flow pattern encompasses laminar, transition, and turbulent re- each tube pass, the rate of heat transfer and temperature changes
gions, a tenfold coefficient variation is possible. In fact, for some become
laminar flow cases, the heat transfer may reverse direction, de-
pending on the temperature driving force profile. Therefore, the dq = (WCt)df = (T' - t')UdS (la
MTD concept may sometimes lead to serious errors. To compen-
The heat balance requires that
sate for the coefficients that vary linearly against temperature,
Colburn [3] has formulated the caloric temperature concept for
single tube-pass exchangers. In a rigorous analysis, Gardner [4] de- dT' = TiRndK' (lb)
veloped an MTD chart, applicable to linear variation of the con-
trolling shell coefficient for a multipass case with infinite tube For large number of cross-baffles, it can be assumed that the
passes. Although both are satisfactory for most applications, nei- surface of all tube passes is exposed to the same shell stream.
ther can be generalized into a valid design method for both vari- Using an approach similar to reference [8], above differential equa-
able coefficient and capacity rate. tion systems can be solved by finite difference method via matrix
manipulation. Let the tube surface be subdivided into K incre-
The purposes of the present investigation are (i) to develop a
ments along the tube length as shown in Fig. 1, in vectorial form,
general solution for the design of multishell-pass,multitube-pass
the stream temperatures can be expressed as
exchangers with variable heat capacity rate and coefficient, (ii) to
explore, by means of case studies, the effects of various design pa- \T'\ = ( r „ T\, . . . T\, ... T'K)T
\t'\„ = (/'2, /'3, . . . . / ' * . . . . / V + 1 ) / n = 1,3,5,7...
= (/'„ /'2, . . . . ! ' „ . . . . / ' * ) / n = 2,4,6,8. . .
1
Numbers in bracket designates Reference at end of paper. Across surface hSkn of &th increment and rath pass, a difference
Contributed by the Heat Transfer Division for publication in the JOUR- approximation of equation (la) becomes
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer
Division March 13,1975. Paper No. 76-HT-N. 1- b ' -*• b (1 - Vjl'kn + (1 + Vjt'k.t.n

Journal of Heat Transfer NOVEMBER 1975 / 509

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 (/ + a ) | 7 " | + T',eK = A„\t'\„ + D„\t' | „.,,„,,,* (3)
in which | tl,, = 0, and An and Dn are defined as follows:
For odd passes, n = 1, 3, 5 . .

A„ = I + aT+ [??]„(/ -a1)

Dn=(l - vhn)En

For even passes, n = 2, 4, 6 , . .

A„ = I + a + [??]„(/ - a)

A , = (1 - vKn)EKK
in which the unity matrix / , vector ey and special unity matrix Eij
are defined in the Nomenclature [1]. The diagonal matrix opera-
tor, a, is defined as follows:

a = (0, eu e2, , e j„\)T for J x J m a t r i x (J = K)

Rearranging equation (3) and expressing |i'|„ in terms of | f | j

\t'\„= y„\t'\t, Yn = A n -'((A n ., -Dn)Yn_x

+ D„_iYn„2), n = 2, . . . N (4a)

Fig. 1 Temperature and flow diagram in She shell of multipass exchanger— Yt = I, Y0 = 0

shell flow arrangement, type I [
| r | i = Ai- ((I + a)\T'\ + T'teK) (4b)
Difference approximation of equation (lb) across feth increment
of all tube passes becomes:
-T\ + T'M = Rkl(t'M - /'„), + Ra(t'„ - t'M\
n = 1,3,5,
+ Rtoit'k ~ t\^t>, k = I, . . . K
= (1 + Vjt'kn + (1 ~Vjt'M,n
(2)
k =K, for n = 1,3,5,
in which
= 1, for n = 2,4,6,
7J*„ = 2(WCt/UAS)l!
Following matrix relation can then be established,
is evaluated at the local average conditions. For all N tube passes
and K increments, following relationships can be formed in terms (a -D\T'\ + T'fiK= E [R]nBn\t'\„ (5)
of K X K matrices and vectors:

JNTomenclature-
Cp = specific heat; Ca, at cup mixing tem- Ri, r = inside tube radius and radial posi- m; x = mAi
perature, J/kg°C, unless specified tion, respectively, m H = viscosity, mPa.s (=cp)
d0, di = tube OD and ID respectively, m, S, AS = total and incremental exchanger 5 = (T 0 - t0)l(Ti — ti), temperature ap-
unless specified surfaces, respectively, m 2 proach factor
ej = unit column vector, j t h element is 1, Re = Reynolds number, dimensionless
AS/AQ = incremental surface effective-
remaining elements are 0 (reference [1]) Additional matrices or vectors At Rn, Dn, F,
ness, percent of additional surface re-
Eij = special unity matrix, all elements are X, Xa, Xb • • • are as defined in the text
quired for each additional percent of
zero except element ij = 1 [ j , | |,( ) = diagonal matrix, column vec-
duty performed
Fc/Fcani = correction for Colburn's caloric tor and row vector, respectively
temperature factor; tav = tc + Fmn, (tt, — T,t= shell and tube stream temperatures, ( ) T = same as column vector | | (super-
tc): Fc, calculated equivalent tempera- respectively; T" = T — ti, t' = t — ti; ti, script T indicate transpose operation)
ture factor for multipass exchangers with tube flow inlet temperature to each shell,
variable U and WC °C Subscripts
FMTD = normal MTD for multipass ex- U, [/[, Uu = overall effective transfer coef- i, o = inlet and outlet temperatures respec-
changers with constant U and WC" C ficients based on FMTD; subscripts I tively for each shell
h = film coefficient of heat transfer based and II refer to types I and II nozzle ori- c, h = cold and hot terminal conditions re-
on tube outside surface; subscripts s and entations (see Figs. 2 and 3); Ux, average spectively, (temperatures, coefficients)
t refer to shell and tube sides, respective- local coefficient over surface AS, W/ k = &th surface increment, k = 1,K
ly, W/m 2 o C m2oC m = axial position. Also it refers to the
I = unity matrix, diagonal shell number starting from the shell
Ate/At/, = ratio of temperature differences,
k = thermal conductivity, W/m°C where tube flow enter the exchanger, m
cold terminal to hot terminal
L = overall tube length, m = 1,M
Ap = pressure drop on tube side, mPa un- v = axial velocities, m/s n = tube pass number for each shell, from
less specified Wt = tube flow rate, kg/s-tube the pass of tube inlet, n = 1, N (N; total
Q, q = total exchanger duty and rate of tube passes per shell)
WC = heat capacity rate or slope of the ex,in = exit and inlet conditions respective-
heat transferred, respectively, W/s
total heat curve, W/s°C ly for the multishell multipass heat ex-
R = WCt/WCs, varies with tube pass n and
tube length increment k x = axial position from the tube entrance, changers (Figs. 1, 2, and 3)

510 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 in which
n = 1,3,5.
n = 2,4,6.
B„ = I - aT
B„ = I - a BK
c XT
a, TEMA E TYPE I

Eliminating all \t'\n except |t'|i with equation (4a) and letting F Til
b, TEMA E TYPE II
N
x
•• {^2 [R\nBnYn)~ , equation (5) becomes
T
ib J ! Tip i
c, TEMA J TYPE III
\t'\i = F((a -I)\T'\
Solving for |T"| from equations (6) and (46), one obtains the
+ T,'eK) (6)
,GT Tot
shell temperature profile:
d, TEMA J TYPE IV
(7a) Q>'---\ (---'<>
TobjjTo' f tj
in which,
To
X = ((F + A," 1 )/ - (F - A^aY* (F - A," 1 )
Fig. 2 Various types of shell (low arrangements
This is really a system of K equations, relating all internal shell
temperatures directly to that of the inlet. The shell outlet temper-
ature can be readily found from the first equation, i.e., for type /
arrangement,
T0 = (1 - x,K)t{ + x,KTi (76) and error procedure. In order to compare the results of this study
with those of normal design practice, the usual "linear" FMTD
Exchanging Ta with T; and using negative WCS in computing
will be used as a basis for defining the equivalent effective overall
Rn and X, one obtains for type II arrangement shell outlet as:
film coefficient t/i or Uu for respective types:
T0 = (1 - l/xlK)ti + \/xiKTt (7c)
II Multishells in Series. Let us define the intershell tem- Q =EE((wc s k{TM - T„))„., Vu Uu = Q/(FMTD)(S)
perature vectors as,
III Divided Flow Arrangement, TEMA J Shell Construc-
\Tt\m = (t„ T0)mT, m = 1,2, . . . M, M + 1 tion. The divided flow type shown in Figs. 2(c) and 2(d) is used
where for M sections in series, to reduce shell side pressure drop. Shell flow enters at the center
and splits evenly toward the sides a and b (Fig. 2). The mathemat-
I Tt\l = (*la> Tex)J M f \ilti — \tex< ^ln) ical treatment, however, is a general one, and will not be restricted
A 2 X 2 matrix relationship can readily be formed based on equa- to an exact nozzle location and flow quantity. For types III and IV
tion (76), connected in series, there are effectively two parallel streams for
each side whose temperatures are Tib, Tia at the inlet; Ta, Tj, in
T,\ m = 1, M (8) between the shells and T0b, Toa at the outlet. Naturally, for com-
mon practice, Tin = Tib = Tia, and Tex = % (Tob + Toa). Let two K
where,
X K configuration matrices Bx and B/, be defined, which contain
1 - (1 - XiK)/xlKR, (1 - xiK)/x1KR for type I two nonzero elements in every column except the (/ — l)th and ;'th
1 - l/xiK, l/xlK m
columns:
for type II,
1 0
1 + (1 - xtl)/R, (xiK -D/R X based on
1 1.
1 " XiK, X
IK m negative
0 1. ,0
0 1 0
and,
Bv 110 0
R. = «T, - T0)/(t„ ~ tt))m
0 0 11.
Eliminating I T( I • • • \Tt\M by applying equation (8) repetitive- 0 1. 0
ly: 0
l l
\T,\ = Zk \T*\ • • • %%' Z\ %M' ZU-\
oi
In terms of matrix elements Z;y of ZMM, the exit temperatures of
a multishell exchanger can now be expressed as, 1 0
-1 1
z z z z
1 \v a, n' n (9) 0 -1 0
1 - l/z22, l/z22 o • 1 0
Bh = - 1 1 0 0
The temperature profiles are well defined by equations (3)-(9),
0 0-1 10
once the property matrices An, [R}„ are calculated, based on some
0 -1 1
reasonably assumed shell temperature profile. The intershell tem-
0 • • • o
peratures can then be calculated from equation (8). All internal
tube and shell flow temperatures can be obtained from equations -1 1
(7a), (46) and (4a)., bearing in mind that for type II, T0' must re- 0 -1
place Ti in equation (46) and negative WCS must be used for [R]„
And the new shell temperature vector | T/\ be defined as
in equation (5).
Note that the differential equation system is nonlinear, and T', T\ T',-\, T'j., T'}i. T'Ktly
therefore an iterative procedure between the assumed and the cal-
culated temperature profiles is required. In reality, for normal ex- in which Tj-', T/ are the dual shell temperatures at k = j where
changer design, An and [R]n are not overly sensitive to tempera- the shell flow splits or merges, i.e., Tj- = Tj = Ti for type III and
ture, and convergence can usually be obtained with the simple trial T0 = lk (Tj- + tj) for type IV. Previous matrix relationships,

Journal of Heat Transfer NOVEMBER 1975 / 511

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 equations (3), (4a), and (5) can still be applied, i.e., for heat bal- I SHELL PASS 2 TUBE-PASS

ance, equation (5) for type IV becomes U,/U c - I To = t o (S = 0)

B„\T'j\ - 7V, + T\eK= ( S W - B - ^ k ' l i = ^"'I'li h s CONTROLLING, h t = 0 0

(10)
N I

F = ( E [R]nBnYn)-\ F, = / .E
n=l •6h

in which, [R]n is evaluated with negative WCS for /e = 1, . . . j ,

where shell flow direction is contrary to that of Fig. 1. Similarly, h t CONTROLLING h s = = °
the heat transfer equation equation (3) becomes

BX\T'}\ + 7V, + T\eK = An\t'\„

+ Dn\t'\n.un= 1, N (11)

Based on equation (4a)

TYPE I ORIENTATION
/' _ v /' n . iV
-Hi
TYPE I I ORIENTATION
Hence, T0 t| T-C TUBE FLOW BEING C00LE0
T-H TUBE FLOW BEING HEATED
| * ' | , = F(Bh\T'j\ - T'aex + T'„eK) (12)

In almost identical fashion, the new shell temperature | T/\ can . RATIO OF TERMINAL TEMP DIFFERENCES

be solved from equations (10) and (12),

Fig. 4 Correction of Colburn's caloric temperature factor
T
\T', 'aXaei + T'bXbeK (13)

in which

Xa = (FBh - AfxBxYl(A{-1 + F) with T e x , is a simple one, hence needs no further elaboration here.
4 X
Xb = (FBh ~A^B,) (A{ ~ F) E f f e c t s of V a r i o u s D e s i g n P a r a m e t e r s o n E x c h a n g e r
Performance
And An and Yn are defined in equations (3) and (4a). Equation
Using the method developed in this paper, case studies of sever-
(13) is, in effect, an extension of equation (7a). Main exceptions
al heat exchangers are made and various design parameters ana-
are that WCS must reverse its sign when k = 1,. . j for type IV and
lyzed. The applicable region of Colburn's caloric temperature fac-
when k = j , . . K for type III (Fig. 2). In terms of matrix elements
tor, FCoib, is defined for cases with constant WC and linear varia-
Xaij, xUj, of Xa and Xb, respectively, the shell inlet and outlet tem-
tion of coefficient. The significance of the present matrix solution
peratures are related by the (j — l)th and / t h rows of equation
technique is demonstrated for exchangers with nonlinear variation
(13), i.e.,
of coefficient and heat capacity rate. The unique behavior of the
T multipass exchangers with tube side in laminar flow is explored,
T'i- = *».;-!, 1 'a + * b,i-i,K^ 6
(14) where the numerical integration procedures for laminar flow can
T'j = x
a,jl T a + x
tjKF b be readily adapted from the available literature [11, 12] with the
modification that the variable tube wall temperature be made de-
where by common practice, pendent upon the shell side temperature. The problem of flow in-
stability across the viscous flow coolers under rapid cooling is also
r u = T}. = T} for type III
evaluated. The results of the case studies are described as follows:
Tn = (T,. + Tj)/2 for type IV I Linear Variation of h Versus Temperature. In their sep-
arate analyses, the linear variation of h has been the basis by both
The procedure to eliminate Ta, Tj, so as to directly relate Tm Gardner and Colburn. Since many design methods [7, 9] require
the calculation of an average coefficient based on some average
temperature, caloric or otherwise, Colburn's temperature factor,
Fcoib, is used here as a basis for comparison, eventhough, for ex-
T..1 |»ex SHELL NOS. TYPES changers with a large tube pass and a controlling shell side coeffi-
M M I cient, hs, Gardner's chart can be directly applied. Several series of
( parametric studies are presented, including shell nozzle orienta-
TOM| 1 tion (flow types I and II), temperature approach, and number of
To4 1

3
i 3 11
tube passes. All of these parameters are critically influenced by
whether the shell side or tube side offers the controlling resistance.
( 1 Effects of Shell Nozzle Orientation or Flow Type. As was
pointed out by Gardner [4, 5], the performance of exchangers with
V. '«13
variable coefficient can be affected by the shell flow arrangement.
( 2 2 1 To develop a meaningful correlation, hundreds of cases have been
T
02 "»iZ calculated. The results indicate the effects are most pronounced
for exchangers with two tube-passes and a zero temperature ap-
( 1 1 11
proach. These results, expressed in terms of Fc/Fco\h versus Atcf
T
ex = T o| 1 ti,, - At/,, are plotted in Fig. 4 for both types I and II, with Un/Uc = 2
and S = 0. The effect of the variation range of coefficient is demon-
Fig. 3 Multishell In series, counter flow—TEMA E shell construction strated in Example I where a two tube-pass oil cooler is calculated

512 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Effect of shell flow type; [/,/£/„, EX I Table 2 Fc/Fcon, as affected by tube passes
At c /At„ UhIUc Ut/Un UhIUc Ui/Uu Uh/Uc = 3, Atc/Ath = 0.4
0.1 2 1.14 5 1.44 T u b e passes 2 4 6 8
1.0 1.16 1.47 Fc/Fcolb, type I 1.07 0.91 0.85 0.82
t y p e II 0.52 0.66 0.72 0.74

with <5 = 0, hs = 5680 while ht decreases linearly from 1200 at the design methods in the published literature, they can be readily
inlet. The results are as shown in Table 1: solved using equations (7o)-(9). The significance of such solution
The following observations are made: is illustrated in Example II. It is an oil cooler whose design require-
(a) Colburn's -Fcoib appears to be a fair approximation for flow ments are as follows:
type I exchangers. Shell flow, cooling water, 39690 kg/h. Cp = 418.6. Temperatures
(6) When hs controls, the use of Fcoib does not lead to exces- in/out; 26.7/48.9°C. Average hs = 5680.
sive errors for the majority of cases calculated. Tube flow, viscous oil, 15876 kg/h. C p = 209.3. Temperatures,
(c) Avoid flow type II when the tube flow is to be cooled, and in/out; 160/48.9°C. Re, in/out; 10000/1050. ht is assumed to vary
ht is controlling. Because, a 50 percent reduction in Fcoib may be linearly from 890 @ 160°C to 125 @ 78.3°C(Re = 2100) and then
required, indicating a 15 percent reduction in performance for the becomes constant at 125 till outlet.
cases illustrated. Exchanger configuration; 1 shell pass, 2 tube passes with 126-%
(d) Type II is preferrable to type I when heating the tube flow, in. 16 BWG tubes (d 0 /d;; 19 mm/15.8 mm). Tube length is to be
and ht is controlling, or when cooling the tube flow, and hs is con- calculated to meet the process requirements.
trolling. They are in general agreement with Gardner's observation Both shell flow types have been calculated. The computation re-
W. sults are summarized in Table 3.
(e) The effects of flow type increase with the range of coeffi- The performance of type I is 54 percent higher than type II,
cient variation. which is in line with previous observations, l i e , l i d . Needless to
2 Temperature Approach. In general, the performance of type say, the calculated Uj,Un, based on nominal FMTD, may not re-
II exchanger deteriorates rapidly as <5 approaches zero. This is flect the real coefficient, especially in the case of variable capacity
shown in Fig. 5 where the results of calculation of a few typical rate for condensers and vaporizers.
cases are plotted. To demonstrate the extent of heat transfer sur- Ill Laminar Flow in Tubes of Viscous Oil Exchangers.
face utilization, the incremental surface effectiveness, AS/AQ, is This is a special case where coefficient varies together with the
also shown. The following tentative conclusions are made: temperature driving force profile. Using the present solution pro-
(a) Fcoib factor is reasonable for either type of exchanger when cedure, the unique behavior of viscous oil coolers can be investi-
6 > 0.3. gated. Several series of case studies, EX III.l to EX III.3, have
(b) The value of the incremental surface effectiveness, AS/AQ, been made using one viscous oil as the tube fluid. The oil proper-
should be used as a guide in determining the required number of ties are as follows:
shell passes. For cases with large coefficient variation, the use of
zero temperature approach, <5 = 0, as a normal criterion, may no k = 0.14 - .000075 t, W / m C
longer be reasonable.
3 Tube Passes. As the number of tube passes increases, there C = 1.7637 + .003826 t, k J / k g - ° C
is a gradual decrease in performance for type I, and an improve-
ment in performance for type II exchangers. A typical case is li = ° 1 4 0 8 5 8 1 e x p ( 6 2 2 4 . 2 / ( i + 273)), m P a . s o r cp.
shown in Table 2, where ht is controlling and 8 = 0. Apparently
large difference between the two exchanger types exists only for 2 p = 902.4 - 0.6334 t, k g / m 3
or 4 tube-passes. For tube pass greater than 10, the value of Fcl
-Fcoib approaches 0.8 for either type. A constant Cs for the shell flow is used. The coefficient is delib-
II Variable Coefficient and Capacity Rate. For condens- erately set as 5680 for 1-tubepass and 2840 for two-tubepass ex-
ers, evaporators and viscous oil coolers, the film coefficient and/or changers. All tubes, unless otherwise noted, are % in. X 14 BWG
capacity rate may change substantially with the flow pattern as (djdi; 19/15.8 mm), 20 ft long (6.1 m).
well as the nature of the heat transfer mechanism. While these 1 Coefficient versus profile of shell temperature. Contrary to
problems are too complex to be handled by any of the conventional the popular belief, the laminar coefficient in a multipass exchanger
is somewhat dependent on the profile of the shell stream tempera-
ture. This can be seen from EX III l a and lb, where the 2-tube-
pass coolers of types I and II respectively are used to cool the oil
from 143 to 140°C, while heating the shell flow from 103.6 to
137.5°C. An additional demonstration is made in EX III l c , where
TUBE FLOW BEING COOLED
Atc/Ath=0.4 the same duty is performed in a cooler with one counterflow tube
TWO TUBE PASSES pass. The profiles of temperatures and local coefficients are shown
in Fig. 6. The following observation can be made:
(a) The coefficients in the counterflow pass are far greater
than in the parallel pass in Example l a and 16.
(6) In the parallel tube-passes, the local coefficients decrease
with the temperature driving forces. Furthermore, as the shell
temperature increases steeply, (being heated by all other tube

0 0.1 0.2 0,3 0.4 0.5 Table 3 Summary of calculation, viscous oil cooler, EX II
TEMP. APPROACH FACTOR S=(T 0 -1 0 )/(Tj-I,)
^I Uu Uh/Uc -Fcolb ^colb
Fig. 5 Effect of temperature approach on caloric temperature factor 202.7 131.7 6.3 0.245 122.0

Journal of Heat Transfer NOVEMBER 1975 / 513

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Ex I I I Ic COUNTERFLOW W, - 207 k g / h . TUBE

AXIAL POSITION ALONG TUBE LENGTH

Fig. 6 Profiles of local heat transfer rate, Ux, and temperatures—tube

flow laminar (EX III 1)
Fig. 7 Radial profiles of velocity and temperature, cold tube pass—EX III
1a

passes) a reversal in heat transfer is observed. Although this re-

sembles the phenomenon in the tail end of normal exchangers with
a close temperature approach, the cause is entirely different and is
unique only to the laminar process. A close examination of the dient, Mw/Min, three different oil inlet temperatures are assumed,
temperature profile, shown in Fig. 7, reveals that as tw or T in- namely, 127, 138, and 160°C. Both % and 1 in. tube sizes are used
creases at a rate too fast for the heat conduction process across the with a length of either 6.1 m or 3.2 m. As shown in Figs. 8 and 9,
liquid layers, a reversed radial temperature gradient, dt/dr, is the results are in general agreement with Mueller's prediction. The
created, resulting in a negative coefficient. Consequently, for the following conclusions can be made:
parallel pass of a multipass exchanger with a steep shell tempera- (a) Critical Flow Rate. For any viscosity gradient, there is a
ture profile, the application of correlations based on single tubep- critical flow rate above which the flow behavior is stable. This min-
ass analysis may be risky if not outright erroneous. imum flow rate, however, must be determined for each individual
(c) With the high coefficient of a counterflow pass, the one- cooling requirement and cooler configuration, unless the viscosity
pass cooler of EX III lc could, at least in theory, saves 25 percent gradient is very small.
in surface and 64 percent in pressure. (b) Viscosity Gradient, nwlnm- F ° r the sample cases, the criti-
2 Performance versus Shell Flow Type. For coolers with a cal viscosity gradient appears to be about 180, below which multi-
close temperature approach, the type II flow arrangement is slight- ple flow rates could coexist for different tubes across the same tube
ly better than type I. This is illustrated in EX III 2a and 26, whose bundle. This would cause the flow maldistribution hence perfor-
calculation results are summarized in Table 4, indicating that Un mance deterioration. Even at a critical gradient, the flow rate
is 9 percent better than U\. However, such an effect will disappear could vary substantially with Ap which could be easily influenced
quickly for cases with a large temperature approach factor. The by the velocity of the inlet nozzle. As a good design practice, the
local coefficient and temperature profiles are not shown since they actual gradient must be much smaller than this value. Needless to
are quite similar to Fig. 6. A similar reverse heat transfer phenom- say, this critical value of 180 applies to the sample oil only, and
enon can also be found in the parallel pass. would vary with the fluid viscosity index as Mueller indicated.
3 Flow Instability and Maldistribution. The problem of flow (c) Effect of Tube Size and Tube Length. From Figs. 8 and 9,
maldistribution across viscous flow coolers has been discussed by it appears that the critical viscosity gradient will not be signifi-
Mueller [10] based on a simple analogy using the available laminar cantly affected by either the tube size or tube length. However, the
solution in literature [11]. With the present computation proce- critical (minimum) flow rate can be substantially lowered by short-
dure, the region of flow instability can now be more precisely de- ening the tube length, or increasing the tube size, if the allowable
termined. In Example III 3, the sample oil is to be cooled on the pressure drop is limited.
tube side of a single tube-pass exchanger with 26.7°C cooling me-
dium (hs = 5680). The flow rate, Wt, is varied over a wide range to Conclusions
develop the Ap- Wt profiles. To explore the effects of viscosity gra- For multipass exchanger design with variable h and WC, the
matrix solution, equations (7a)-(9) and equations (13) and (14)
can be used for TEMA construction types E and J. Numerous case
studies have been made to investigate the significance of various
Table 4 Performance of 2-tubepass coolers, E X III.2 design parameters. One of the interesting conclusions is, for cases
Flow with linear coefficient variation and constant WC, the application
Cases types Wt Tin/Tout rin/tout U of the conventional MTD in conjunction with Colburn's caloric
III.2o I 31 26/59.2 93/60.8 27.7 temperature, is still reasonably accurate, if there are six or more
111.2b II 31 26/60.0 93/60.0 30.1 tube passes per shell and 5 5= 0.3.

514 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 I „ 26.7°C

140
100
I0| 60 '
8
-40
ft 6 30

1,
CASE 1 IN
Ic 127 180
2c 138 270

1 1 1 1 I I 1 1
6 8 10 20 40 60 80 100

W,, FLOW RATE, kfl/h. TU8E

6 8 10 20
FLOW RATE k g ' h TUBE Fig. 9 Instability of laminar flow with large viscosity gradient—EX III 3,
1-ln. tube
Fig. 8 Instability of laminar (low with large viscosity gradient—EX III 3,
%-in. tube

2 Bowman, R. A., Mueller, A. C, and Nagle, W. M., "Mean Tempera-

For laminar flow in tubes of multipass exchangers, the coeffi- ture Difference in Design", TRANS. ASME, Vol. 62, 1940, p. 283.
cient is sensitive to the profile of the temperature driving force. 3 Colburn, A. P., "Mean Temperature Difference and Heat Transfer
Coefficient in Liquid Heat Exchangers," Ind. Engg. Chem., Vol. 25, 1933,
With a close temperature approach and a steep shell temperature pp. 873-877.
gradient, the local coefficient in a parallel tube pass may become 4 Gardner, K. A., "Variable Heat Transfer Rate Correction in Multi-
negative. This indicates that theoretical analyses or empirical cor- pass Exchangers, Shell Side Film Controlling," TRANS. ASME Vol. 67,
relations for a single tube may not be adequate for designing mul- 1945, p. 31.
5 Gardner, K. A., "Mean Temperature Difference in Unbalanced-Pass
tipass exchangers. For stable operation of viscous flow coolers, Exchangers," Ind. Engg. Chem., Vol. 33,1941, p. 1215.
some critical (minimum) tube flow rate must be maintained, un- 6 Gardner, K. A. "Mean Temperature Difference in Multipass Ex-
less the viscosity gradient is well below a certain critical value (180 changers," Ind. Engg. Chem., Vol. 33,1941, p. 1495.
for the oil in sample illustrations). Both can be determined using 7 Standards of Tubular Exchanger Manufacturers Association,
TEMA, Fifth ed„ 1968.
the presently developed solution procedures. 8 Kao, S., "Design Analysis of Multistream Hampson Exchanger With
Paired Tubes," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Vol.
Acknowledgment 87,1965, p. 202.
9 Mcadams, W. H. Heat Transmission, Third ed., McGraw-Hill, N. Y.
Acknowledgment is made to Mobil Research & Development 1954, p. 191.
Corp. for permission to publish this paper. The writer also thanks 10 Mueller, A. C, "Criteria for Maldistribution in Viscous Flow Cool-
Messrs. E. S. Yu and J. M. Sirman of the same company for their ers," Proceedings of International Heat Transfer Conference, Heat Trans-
assistance in preparing the manuscript. fer, Vol. 5,1974, p. 170.
11 Kwant, P. B., Zwaneveld, A. , and Dukstra, F. C, "Non-Isothermal
Laminar Pipe," Chem. Engg. Science, Vol. 28,1973, p. 1303.
References 12 Kays, W. M. "Numerical Solution for Laminar-Flow Heat Transfer
1 Bodewig, E., Matrix Calculus, Second ed., North-Holland Publishing in Circular Tubes," TRANS. ASME, Vol. 77, 1955, p. 1265; also discussion
Co., Amsterdam, 1959, p. 5. by Lipkis, p. 1273.

Journal of Heat Transfer NOVEMBER 1975 / 515

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 J. K. Hagge 1
Research Asst.
Mechanical Augmentation of
G. H. Junkhan
Assoc. Professor.
Convective Heat Transfer in Air
An experimental investigation was conducted into augmentation of forced convection
Department of Mechanical Engineering,
Iowa State University,
heat transfer in air by mechanical removal of the boundary layer. A rotating blade ele-
Ames, Iowa ment passing in close proximity to a flat plate convective surface was found to increase
the rate of convective heat transfer by up to eleven times in certain situations. The blade
element effectively scrapes away the boundary layer, thus reducing the resistance to
heat flow. Parameters investigated include scraping frequency, scraper clearance, and
type of boundary layer. Increased coefficients were found for higher scraping frequen-
cies. Significant augmentation was obtained with clearance as large as 0.15 in. (0.0038
m) between the moving blade element and the convective surface. The technique ap-
pears most useful for laminar and transitional boundary layers, although some improve-
ment was obtained for the turbulent boundary layers investigated. The simple surface
renewal theory developed for scraped surface augmentation in liquids was found to ap-
proximately predict the coefficients obtained. A new relation is proposed which gives a
better prediction and includes the effect of scraper clearance.

Introduction the scraped-surface 2 technique had previously been used with gas-
eous convection. While there is apparently no report in the litera-
Applications where heat must be transferred across a solid/gas ture of the scraped-surface technique being applied to gaseous
interface are quite common. Unfortunately, the film coefficients convection, a considerable amount of work has been done using
associated with gaseous convection are quite low. In particular, the scraped-surface augmentation with liquids where up to an order of
most difficult part of any nonexpendable-coolant cooling system magnitude increase in film coefficients are obtained. The scraped-
which must reject heat to the atmosphere is usually the thermal surface technique with liquids has found application with falling
coupling between the solid convective surface and the cooling air. film evaporation apparatus for saline water conversion, chemical
It is well known that boundary layers provide the primary resis- distillation, and fractionation, and in the food processing and
tance to heat flow from the solid surface to the gas. It would seem chemical process industries. A discussion of the literature con-
logical then that atn approach which could remove these boundary cerning scraped-surface heat transfer with liquids from 1928 to the
layers would substantially increase the heat flow. Surface scraping present is given in reference [1] . 3
is proposed as a means to remove the gaseous boundary layers. A Two simple theories, or a combination thereof, have been ad-
scraping blade in close proximity to the convective surface is peri- vanced to model scraped-surface heat transfer with liquids. Kern
odically moved across the surface to remove the insulating bound- and Karakas [2], Kirschbaum and Dieter [3], and Lustenader, et al.
ary layers. [4] have utilized a stagnant-film theory which assumes all resis-
An extensive literature search was carried out to determine if tance to heat transfer occurs in a stagnant film of thickness 5 and
thermal conductivity k in contact with the convective surface. The
film thickness is generally taken as the scraper clearance or some
fraction thereof. The effective film coefficient is therefore
1
Presently with Collins Radio, Rockwell International, Cedar Rapids,
Iowa.
2
Contributed by the Heat Transfer Division of THE AMERICAN SOCI- Note that the term "scraped-surface" is widely used for both situations
ETY OF MECHANICAL ENGINEERS and presented at the 1975 ASME- where the scraper actually comes in physical contact with the surface, and
AIChE National Heat Transfer Conference, August 11-13, 1975. Revised for those situations where it is just in close proximity to the surface. This in-
manuscript received by The Heat Transfer Division September 19, 1975. vestigation is concerned only with the close proximity situation.
3
Paper No. 75-HT-42. Numbers in brackets designate References at end of paper.

516 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 SHIMS
THRUST BEARINGS

— 8.0 -
(.203)

RPM GEAR

FLAT PLATE
JTm SECTION Y/ALL
— .25 RADIUS

Fig. 1(a)
LEADING EDGE VIEW

Scraper drive train

sf
mf'
•^L J
(.0063)

-*t 0.4 s H
(.010)
SECTION A - A
Fig. 1(6) Scraping blade

Fig. 1 Scraper mechanical and geometric arrangement—dimensions in

inches (meters)

h =- vertical in the wind tunnel. A motor driven shaft through the back
(1)
side of the wind tunnel and flat plate drove a surface scraper in
Harriot [5], Kool [6], and Latinen [7], have utilized a surface re- close proximity to the front heated surface of the plate as shown in
newal/penetration theory which assumes the convective surface is Fig. 1(a). A 0.625 in. (0.016 m) dia hollow shaft with a bearing sup-
periodically wiped clean, and fresh fluid at the bulk temperature port on the back of the flat plate and at the test section wall drove
comes instantaneously in contact with the convective surface. The the scraping blade. The 8 in. (0.203 m) blade used in these tests is
effective coefficient for this theory is given by shown in Fig. 1(b). The shaft was driven by a % hp (93 W) univer-
2 ^ /pcpk
sal motor operated on regulated d-c voltage. A series of shims al-
(2) lowed positioning of the thrust bearing housing for varying scraper
clearance. Four clearances, 0.15 in. (0.00371 m), 0.10 in. (0.00254
where p and c p are fluid density and specific heat and r is the time m), 0.050 in. (0.00127 m), and 0.030 in. (0.000381 m), were used.
between surface renewals. The front surface of the flat plate was a printed circuit board
For liquids, each of the theories is found to describe particular with a serpentine resistance pattern etched from 0.0013 in. (3.3 X
but different limit cases. The stagnant-film theory has been ap- 10~ 5 m) thick cupronickel foil on the surface to provide the heated
plied with relatively good success to highly viscous liquids with surface. A similar printed circuit board was located in the interior
laminar flows. The surface renewal/penetration theory, which by of the flat plate as a guard heater for losses through the back of the
its nature must include some local turbulence or small scale mix- plate. Thermocouples and pressure taps were mounted flush with
ing, has been relatively successful with liquids of low viscosity. the front surface of the plate from the interior in the locations
Combinations of these theories have been proposed by Penney and shown in Fig. 2. a-c power was supplied to the various heating ele-
Bell as cited by Penney [8], Thomas and Mohler [9], and Toor and ments with variable transformers. Construction of the flat plate is
Marchello [10]. A preliminary experimental investigation into sev- described in more detail in reference [l].
eral parameters involved in the use of surface scrapers to enhance Thermocouple E M F was measured on a multichannel recorder
convective heat transfer coefficients for gas boundary layers is de- referenced to an ambient thermocouple within the tunnel located
scribed in the following and comparison of results with the models upstream of the plate's leading edge. The ambient temperature in
referenced is discussed. the tunnel was in turn referenced to a 32°F (0°C) ice bath and
monitored with a potentiometer.
Experimental Investigation The rotational speed of the shaft was monitored using a magnet-
All experiments were performed in the subsonic wind tunnel of ic pickup in proximity to a 12 tooth steel gear on the shaft shown
the Mechanical Engineering Department at Iowa State University. in Fig. 1. The output of the magnetic pickup was monitored on an
The wind tunnel is an open circuit suction type, with a 60 in. (1.52 oscilloscope to determine rpm.
m) long test section that is 14 X 14 in. (0.36 X 0.36 m) in cross sec- Power to the heaters was monitored by measuring the voltage
tion. The maximum free stream turbulence intensity for the range across the heater elements together with previously determined re-
of tunnel velocities used was less than 0.6 percent. sistance values which were corrected in each test for temperature
An instrumented flat plate was installed with its leading edge changes in resistance. Separate current and voltage leads were run

-Nomenclature-
y = distance from wall
Cf = skin friction coefficient fective Reynolds number = ( [ / „ + UB) 8 = film or boundary layer thickness,
cp = specific heat x/v scraper clearance
h = convective film coefficient Rex = local Reynolds number = U«,x/v v = kinematic viscosity
k = thermal conductivity Re s = clearance Reynolds number = UBS/H <t> = dimensionless clearance = 5\/Re x /5x
N = revolutions per unit time s = unheated starting length p = density
Nuj = local Nusselt number = hx/k UB = blade velocity T = time between surface renewals
NUA = clearance Nusselt number = hS/K £/«, = free stream velocity Tw = wall shear stress
:
Pr = Prandtl number =• ncp/k (Reeff = ef- x - distance from leading edge M : absolute viscosity

Journal of Heat Transfer NOVEMBER 1975 / 517

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 NOSEPIECE-

XT

-QOGQOpOQ -

T SCRAPER
REMOVED

2GUAWT / ^GUARD ° PRESSURE TAP

HEATER / HEATER
^ T E S T SECTION HEATER o THERMOCOUPLE 0 . 0 5 0 I N C H (.00127 m) CLEARANCE

A GUARD HEATER THERMOCOUPLE

Pig. 2 Thermocouple and pressure tap locations on flat plate

10 iL

REYNOLDS NUMBER Re =

to the heater elements to eliminate measuring any power dissipa- Fig. 4 Scraped performance with 5 = 0.050 In. (0.00127 m)
tion losses in the lead wires.
The position of the flat plate was adjusted in the wind tunnel to
give a zero pressure gradient along the plate surface using the pres-
sure taps shown in Fig. 2. Prior to installing the scraper mecha- with the scraper removed and the shaft hole filled and smoothed.
nism, the shaft hole was filled and smoothed, and boundary layer • Heat transfer coefficient relations for a uniform surface heat flux
velocity profiles were measured with a small total pressure probe. are given by Kays [11, pp. 222, 244]; for Pr = 0.706 these relations
These tests indicated excellent agreement with a Blasius velocity reduce to
profile for Re* up to 1.79 X 105. The shaft bearing supports were
then adjusted to align the scraping blade parallel to the plate sur- Nu* = 0.403 Re*0-5 (3)
face, with shims installed at the thrust bearing to give the appro-
for laminar boundary layers, and to
priate clearance.
With the scraper speed and the free stream velocity at the de- Nux = 0.0249 Re* °-8 (4)
sired levels, power was supplied to the test section and the adja-
for turbulent boundary layers. Equations (3) and (4) are plotted in
cent heaters to provide a uniform heat flux over the surface of the
Fig. 3 along with the data for the tests with the scraper removed
plate. Corrections for losses from the front and back of the plate
showing reasonable agreement with theory from the laminar range
were made utilizing measured data and were each typically less
up to the turbulent range. The effect of the 2 in. (0.051 m) unheat-
than 10 percent of the total power.
ed starting length of the nosepiece was accounted for by applying
Convective coefficients were calculated using data obtained
the appropriate correction factors to the Nusselt number. For lam-
from the spanwise thermocouples shown as filled circles in Fig. 2.
inar flow Eckert's correlation [12] was used
An uncertainty analysis indicates the experimental determinations
of Nusselt number are of the order of ± 10 percent. 1/3

Results
NUco Nu,
['-©"]" (5)

For turbulent flow Rubesin's correlation [13] was used

For comparison with existing theory a series of tests were run
f / SS\\ 39/401-
39/401-7/9
NlWrected = N u * [1 - l-j (6)

Fig. 4 shows the experimental results for various rotational

speeds with one of the four scraping clearances tested. The indi-
cated Nusselt numbers have been corrected for the unheated start-
1000
E Q U A T I O N 4 —. JC ing length. This correction is somewhat arbitrary for the scraped-
surface situation but has been applied to maintain what is thought

/ /J
to be a fair comparison between the scraped and unscraped perfor-

z
J mance. Fig. 4 indicates up to eleven times improvement in heat
transfer in normally laminar flows, with the highest scraping
speeds providing the most augmentation. Above 100 rpm the aug-
NUMBER

mentation increases approximately linearly with scraping speed,

- J^ with the rate of increase being much higher in laminar flow than in
S

5
^ \ turbulent flow. Fig. 5 indicates that the effect of scraper clearance
i •^
r y ^ ^ ~ EQUATION 3

FREE STREAM VELOCITY

is small for the range of clearances tested, with the maximum aug-
mentation being at most inversely proportional to the fourth root
D
O
12.7 ft/sec {3.87 m/s)
3 9 . 0 ft/sec {11.88 m/s}
of clearance. However, much of the data indicates an even smaller
A 8 0 . 0 ft/sec (24.38 m/s) influence due to clearance. Fig. 5 is plotted as a function of the di-
mensionless scraper clearance ij>, where <t> is defined by
10 I I 1 1 1 1 ! t 1 1 1 1 i i 1 1 f t 1

REYNOLDS NUMBER Re

Fig. 3 Tests with scraper removed For laminar flow tj> is equivalent to the ratio of scraper clearance b

518 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 SYMBOL RPM Re
O 100 80,000
A 2000 80,000
B 100 255,000
A 2000 255,000
D 540,000
0 2000 540,000

-.— a - .

;fi— T 3

DIMENSIONLESS CLEARANCE 5

Fig. 5 Effect of clearance

CLEARANCE REYNOLDS NUMBER Re, = ^ B _

to the normal unscraped boundary layer thickness.
Fig. 7 Comparison with stagnant film theory and equation (9)
Discussion
The stagnant-film and surface renewal/penetration models were
proposed for conditions where free-stream Reynolds number ef-
fects are small relative to surface scraping effects, corresponding to
conditions where the effect of free-stream Reynolds number is
small. For the clearance of Fig. 4, the effect of free-stream Reyn-
olds number is small for scraper speeds of 1000 and 2000 rpm
below Re x = 105 and for scraper speeds of 500 and 100 rpm below
Nus = Re,
(?) p r l/3 (7)

Re* = 4 X 104. The boundary layer approaching the scraped sec- The skin friction coefficient can be expressed for turbulent flow as
tion at these Reynolds numbers is laminar according to the surveys [14]
made without the scraper. These data together with similar data 1/4

for other clearances are shown in Fig. 6, where a line representing -/. 0.0228 (8)
the surface renewal/penetration model is included. The time be-
tween renewals is taken as one-half the period of revolution for a Inserting equation (8) into (7) and multiplying by the ratio S/x,
two-bladed scraper. Agreement is good except at short time inter- Nu 5 = 0.0228 Rej 0 ' 75 Pr 1 / 3 (9)
vals where the model predicts lower heat transfer coefficients.
The same data are also shown on coordinates appropriate to the In equation (8), <5 is the boundary layer thickness, taken in this
stagnant-film model in Fig. 7, where the model is represented by case to be impressed on the test area as the scraper clearance by
the horizontal dashed line. The implication here appears to be that the blade moving with a velocity UB- The free-stream velocity is
the effective thickness of the stagnant film does not remain con- then considered as UB in equation (9). A line representing equa-
stant nor have a simple relationship to scraper rotational speed. tion (9) with the Prandtl number taken for air is also shown in Fig.
Noting that the purpose of the scraper is to disturb the boundary 7. The agreement is better, suggesting that a type of turbulent flow
layer, some improvement in correlation is obtained by use of the exists between the scraper blade and the plate surface. Note that
simple heat transfer-momentum transfer analogy and the assump- this occurs even for low scraper rotational speeds, corresponding to
tion of turbulent flow between the scraper and the plate. low blade velocities.
For the uniform incompressible flow past a flat plate at zero in- Assuming then, that the effect of surface scraping in air is relat-
cidence, the analogy is [14] ed to the turbulence generated by the blade, it is reasonable to ask
whether similar augmentation could be obtained by a free-stream
flow with high turbulence intensity and no scraper. For the condi-
tion of a flat plate with zero pressure gradient, it has been shown
by Junkhan and Serovy [15], among others, that the effect of free-
100 stream turbulence alone on the heat transfer coefficient is negligi-
(567.6 W/m 2 C)
ble; the major manifestation of the turbulence is simply an earlier
transition from a laminar to a turbulent boundary layer. The Nus-
selt numbers would be predicted by equation (4) at lower Reynolds
numbers for such conditions. The Nusselt numbers obtained with
the scraper are significantly higher than those predicted by equa-
tion (4), implying that the scraper is responsible for the augmenta-
tion observed.
The experiments performed in this study also included data
taken for conditions where free-stream Reynolds numbers had an
effect on the Nusselt number. Nusselt numbers for the overall
range of Reynolds numbers investigated have been correlated
within 25 percent by using an empirical relation formed from the
conventional turbulent flow equation by adjusting the Reynolds
number to reflect the effect of the surface scraping. The effective
Reynolds number was redefined by

Reeff = l ^ ± ^ (10)
T TIME BETWEEN SURFACE RENEWALS, SECONDS

Fig. 6 Comparison with surface renewal/penetration theory

Journal of Heat Transfer NOVEMBER 1975 / 519

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ;

: O 100 RPM
\
O 500 RPM
A 1000 RPM
9 2000 RPM
A
O

-
o, / \
~ &$y
A g(xs
n /n
_ a ^6
~ E Q U A T I O N 11 — ^
/&
_ / <"£>
O

i t 1. 1 ! I Mil ! 1 1 1 1 1
100 RPM
EFFECTIVE REYNOLDS NUMBER Re
0.050 INCH (.00127 m) CLEARANCE
Fig. 8 Comparison of all tests with equation (11)
0 50,000 100,000
U x
REYNOLDS NUMBER Re = —^-
x V
and the heat transfer relation is
Pig. 9 Augmentation In laminar flow
Nil* = 0.0249 Reeff0-8 (ID
A comparison oi
t\ comparison of an
all the
tne experimental
experiments, data for surface scraping
ehnwn in TTitT /£U Radio Company for their partial support of this work. We would
with equation (11) is shown in Fig. (8). like to thank Mr. R. W. Wright and Mr. V. J. Grondahl of Collins
Conclusions Radio for the use of Collins' facilities in fabricating the printed cir-
The test results for the particular geometry and flow conditions cuit heating elements of the apparatus, and the assistance of Mr.
studied lead to the following conclusions: D. D. Miller in this endeavor. Acknowledgment is made to a con-
1 Up to an order of magnitude increase in convective heat versation many years ago with Mr. H. J. Bronson and Mr. V. L.
transfer coefficients can be obtained with the surface scraping Carlson from which recently germinated the idea of augmenting
technique. gaseous convection with surface scraping.
2 Scraping speed is the dominant parameter, with augmenta-
References
tion approximately proportional to scraping speed for speeds
above 100 rpm. 1 Hagge, J. K., and Junkhan, G. H., "Experimental Study of a Method
of Mechanical Augmentation of Convection Heat Transfer Coefficients in
3 Clearance between the scraper blade and plate surface has a Air," Technical Report HTL-3/ISU-ERI-AMES-74158, Engineering Re-
relatively minor effect on convective coefficient augmentation for search Institute, Iowa State University, Ames, Iowa, Nov. 1974.
clearances from 0.030 to 0.150 in. (0.0007-0.0038 m). These rela- 2 Kern, D. Q., and Karakas, H. J., "Mechanically Aided Heat Trans-
tively large clearances still produced significant augmentation. fer," Chem. Eng. Prog. Symp. Ser. Vol. 55, No. 29,1958, pp. 141-148.
3 Kirschbaum, E., and Dieter, K., "Warmeiibergang und Teildestilla-
Note that the proposed relation given by equation (9) will predict tion in Dunnschichtverdampfern," Chem.-Ing.-Tech., Vol. 30 1958, pp.
significantly greater augmentation as the clearance is reduced. 715-720.
Such data has not yet been obtained and extrapolation to small 4 Lustenader, E. L., Richter, R., and Neugebauer, F. J., "The Use of
clearances should be done with caution. Thin Films for Increasing Evaporation and Condensation Rates in Process
Equipment," JOURNAL OF HEAT TRANSFER, Series C, Vol. 81, 1959,
4 Augmentation appears to be due to turbulent-type flow be- pp. 297-307.
tween the scraper and the plate surface, with the most significant 5 Harriot, P., "Heat Transfer in Scraped-Surface Exchangers," Chem.
increases occurring for normally laminar boundary layers. Based Eng. Prog. Symp. Ser., Vol. 55, No. 29,1958 pp. 137-139.
upon the data obtained for plate Reynolds numbers which are nor- 6 Kool, J., "Heat Transfer in Scraped Vessels and Pipes Handling Vis-
mally laminar, Fig. 9 shows the augmentation realized in this in- cous Materials," Trans. Instn. Chem. Engrs., Vol. 36,1958, pp. 153-158.
7 Latinen, G. A., "Discussion of the Paper 'Correlation of Scraped
vestigation. Film Heat Transfer in the Votator' by A. H. Skelland," Chem. Eng. Sci.,
5 Augmentation for laminar boundary layers is considerably Vol. 9,1959, pp. 263-266.
greater than would be obtained from an induced transition to tur- 8 Penney, W. R., "The Spiralator-Initial Tests and Correlations,"
Paper presented at the Eighth National Heat Transfer Conference of the
bulent flow at the same Reynolds number. AIChE-ASME, Los Angeles, Calif., Aug. 8-11, 1965.
6 Heat transfer performance for conditions where a laminar 9 Thomas, L. C, and Mohler, R. A., "Mathematical Model for Close
boundary layer would exist without the scraper is fairly well pre- Clearance Heat Exchangers," ASME Paper No. 71-HT-32, 1971.
dicted by the surface renewal/penetration theory or by equation 10 Toor, H. L., and Marchello, J. M., "Film-Penetration Model for Mass
(9). In situations where a turbulent boundary layer would exist and Heat Transfer," AlChE Journal, Vol. 4, 1958, pp. 97-101.
11 Kays, W. M., Convective Heat and Mass Transfer, McGraw-Hill,
without the scraper, the performance is approximately predicted New York, 1966.
by the relation given as equation (11). 12 Eckert, E. R. G., Introduction to the Transfer of Heat and Mass.,
The scraped-surface method offers potential augmentation for McGraw-Hill, New York, 1950.
13 Rubesin, M. W., "The Effect of an Arbitrary Surface Temperature
heat exchangers with normally laminar boundary layers. Examples Variation Along a Flat Plate on the Convective Heat Transfer in an Incom-
of such cases might be in electronics equipment or low face veloci- pressible Turbulent Boundary Layer," NACA TN 2345, 1951.
ty plate type heat exchangers. 14 Schlichting, H., Boundary Layer Theory, Fourth ed., McGraw-Hill,
New York, 1960.
Acknowledgments 15 Junkhan, G. H., and Serovy, G. K., "Effects of Free Stream Turbu-
lence and Pressure Gradient on Flat-Plate Boundary-Layer Velocity Pro-
The writers are indebted to the Engineering Research Institute files and on Heat Transfer," JOURNAL OF HEAT TRANSFER, Series C,
of Iowa State University, John Deere and Company, and Collins Vol. 89,1967, pp. 169-176.

520 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 M. C. Jones
R. G, Peterson A Study of Flow Stability in
Cryogenics Division,
Institute for Basic Standards,
National Bureau of Standards,
Helium Cooling Systems1
Boulder, Colo.
Methods have been developed for the computation of the frequency response and stabili-
ty of helium cooling systems in the frequency range of the density wave instability.
While more generally applicable, these methods were developed for a study of supercon-
ducting power transmission. Special features are the use of helium thermodynamic
properties derived at every point from the exact equation of state, and the full account-
ing of the effect of compressibility. Classical linear control theory is employed through-
out. By using a finite difference approach to the integration of the conservation equa-
tions over the space coordinate, the accuracy obtainable is limited only by the computa-
tion time available. Examples are given for representative transmission line parameters
with the helium at a variety of supercritical pressures. It is concluded that, while density
wave instability is a real potential problem in superconducting power transmission lines,
it is not difficult to select operating parameters in a safe stable regime.

Introduction cryogenic fluids by Zuber [3] and by Friedly, et al. [4] using a two-
Although the analysis and computations described in this paper zone equation of state that was originally intended for subcritical
are equally applicable to any helium cooling system, the work was systems. In the two-zone approximation the specific volume (or
prompted by the peculiar demands of superconducting power density) is assumed to be constant in the first zone representing
transmission line development. There are now several research and subcooled liquid. From the saturation point onward the specific
development projects being carried on in laboratories in the Unit- volume is assumed to be linear with the enthalpy, and the effect of
ed States and abroad and it appears certain that the next five pressure on specific volume is ignored. Fig. 1 plotted from [5]
years will see large scale experimental and prototype power trans- shows the specific volume plotted against enthalpy for helium at a
mission lines being built [l]. 2 This study has to do with the well variety of pressures, and if the effect of pressure is ignored, one
known, but not well documented, phenomenon of oscillations in can see that the two-zone approximation would indeed be reason-
helium flow systems. The purpose was to make analyses and com- able below the critical pressure (0.2275 MPa = 2.245 atm). How-
putations to determine the effect of system parameters on stability ever, above the critical pressure such an approximation is obvious-
and to define areas of stable operation. ly weak and the zone boundary somewhat arbitrary. The trans-
In particular, the paper deals with the type of flow system insta- posed critical temperature at a given pressure has been suggested
bility, well known in both subcritical and supercritical heat ex- as a suitable zone boundary, but Fig. 1 shows that this indicates
changers at high temperatures, in which disturbances are propa- nothing more than a weak knee in the isobars. In view of our con-
gated through the coolant channel at the fluid velocity. This is cern for accuracy and for realistic predictions, therefore, we decid-
commonly called a density wave instability or oscillation [2], the ed to discard the two-zone approximation and, instead, to work out
frequency being given by %T, where T is the residence time of the methods that would enable the fluid at all points to be represented
fluid in the channel. It is recognized at the outset that other modes by its true equation of state. Furthermore, we have not ignored the
of oscillation are possible, particularly acoustic modes. These are compressibility of the fluid, the neglect of which, as we will show,
the subject of continuing study. is not a conservative approximation.
The density wave instability has been studied for supercritical One of the system parameters which was found to be important
in higher temperature studies is the heat capacity of the coolant
channel wall and the rate of heat exchange between the wall and
coolant. We have also included this interaction, which leads to a
stabilizing effect, but we find that at helium temperatures this in-
1
Contribution of the National Bureau of Standards and not subject to teraction may be entirely ignored.
copyright. In the following sections we derive the working equations and
2
Numbers in brackets designate References at end of paper. describe two methods of solution. The first does make the assump-
Contributed by the Heat Transfer Division and presented at the Winter tion of zero compressibility, but this approximation proves useful
Annual Meeting, New York, N. Y., November 17-22, 1974, of THE AMERI- in understanding the phenomenon. It has also been useful in pro-
CAN SOCIETY OF MECHANICAL ENGINEERS. Revised Manuscript re-
ceived by the Heat Transfer Division January 13, 1975. Paper No. 74-WA/ viding a reference calculation in the development of the second
HT-24. method, which is quite general. Finally, we show some results of

Journal of Heat Transfer NOVEMBER 1975 / 521

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 _...M ~1~T~~
-r-pr-rc 1.0i 1 1/ ! i t 1 1
•J ' [ ! i
/l.5 2 3 / /Z.5

- A / 4
/1 _
_ / i -
Two - phase
Ertv elope -
/ i
-^e
/^H/ __ Fig. 2 Simple blow-down system
" / ; '-"Batm
-
_ / 1 /
-""lo
" -
__ / 1 // —
/ l// - SU

/ V/'/ / ^"i5 —
- INLET su.
CONSERVATION SH
OUTLET
/ ^y / -
— / \yr<^<^ -rx — RESTRICTION RESTRICTION

1 I
It
-s3=^Sg;

1 I 1 I 1 M i l ) i 1 1S
- T r a n posed
C n ical

f I 1 1 1 I 1 1 ! i 1 l 11 - i
-
-. 4- 8P|
EQUATIONS Sp
n+l
i- Sp ex ,, = 0

150 200 300 350 Fig. 3 Block diagram for simple blow-down system
H, J / m o l

Fig. 1 The v(H,p) representation of the thermodynamic states of helium I

Implicit in this approach is the need to integrate along the heat
load channel the linearized conservation equations for the pertur-
bations of the thermodynamic and hydrodynamic variables. As will
computations illustrating the effect of some of the system parame- be seen, this is the heart of the problem. In our approach we subdi-
ters for possible power transmission line configurations. vide the channel into sections and use average steady-state ther-
modynamic and hydrodynamic parameters over each.
Analytical Approach To be more concrete, we derive the working equations below for
In this study we have used the classical methods of linear con- the simplest cooling system possible: a blow-down system, Fig. 2.
trol theory [6] to investigate the local stability of various chosen In this system the refrigerant flows from a constant pressure reser-
operating conditions. The analysis is most conveniently carried out voir, via a flow restriction, through the channel of the heat load,
by taking the Laplace transform with respect to time of the linear- and via another flow restriction to a second constant pressure res-
ized time-dependent conservation equations. Setting initial condi- ervoir. As has been pointed out many times, this system is dynami-
tions for the perturbed variables to zero, the ratio (output/input) cally unchanged if the final flow restriction is choked. Anticipating
for chosen variables of the total system gives its unit impulse re- the derivation of the working equations below, the block diagram
sponse or transfer function F(s) for the system. As is well known, for this system is given in Pig. 3. In all the work described we have
instability is indicated by singularities of this function (or zeroes of selected SUi, the perturbation in velocity at the inlet of the heated
the denominator) for positive values of the real part of the Laplace channel, as output and <5po as input. The feedback is thus Spi, the
transform variable s. The system transfer function for any feed- perturbation in pressure at inlet.
back system can be reduced to the form
Derivation of the Working Equations
J(s) Basic Assumptions. Referring to Fig. 2, it is assumed for this
F{s) = - (1)
1 + J(s)H{s) analysis that dense supercritical helium flows from a reservoir at
where J(s) is the forward transfer function and H(s) is the feed- constant pressure po through a flow restriction into the channel of
back transfer function. (To avoid confusion with the enthalpy H the heat load. The transfer of heat from high temperature sur-
the feedback transfer function is always written here as H(s).) We roundings and the a-c electrical dissipation in conductor and di-
thus seek roots of the characteristic equation electric together produce a uniform mean heat flux q. However,
due to perturbations in helium conditions, the nonzero heat capac-
1 + J (s)H(s) = 0 (2) ity of the channel walls, and the temperature dependence of the
for positive values of the real part of s. This is done by making the losses, the heat flux to the helium will also, in general, experience
Nyquist plot of the function J(s)H(s) called the open loop transfer perturbations bq. The instantaneous transfer of heat from channel
function. The theory of the Nyquist criterion of stability is given in wall to helium is described by a steady-state heat transfer coeffi-
standard texts on control theory [6]. cient h. This is acceptable in all cases of interest in transmission

-Nomenclature-
(i = cross-sectional area, m 2 n = number of computational zones of p = density, kg/m 3
•P = h< heated channel T = time constant or residence time, s
kg-K p = pressure, Pa fl = reaction frequency, s _ 1 , defined in text
(P = channel perimeter, m
F = /(P/a, m" 1
q = heat flux into fluid, W/m 2
F(s) = system transfer function, m 2 -s/kg s = Laplace transform variable with re- Subscripts
/ = friction factor spect to time i = inlet of computational zone
G = mass flow per unit area, kg/m 2 -s T = temperature, K j = zone index
H = enthalpy of helium coolant, J/kg t = time, s n = final zone
U = fluid velocity, m/s n+l = conditions at outlet of final zone
H(s) = feedback transfer function, kg/m 2 -s
v = specific volume, m 3 /kg o = supply reservoir conditions
h = heat transfer coefficient, W/m 2 -K x = axial length coordinate of heated chan-
J(s) - forward transfer function, m 2 -s/kg nel, m
ki, ks = orifice coefficients for inlet and 0 = q(?/m, J/kg-m Superscripts
exit flow restrictions, respectively 5 = prefix denoting infinitesimal perturba- — = steady-state part of time dependent
m = mass flow rate, kg/s tion variable

522 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 line studies on account of the long periods of oscillations to be ex- linearized equations of conservation for the perturbed variables:
pected. equations (10), (11), and (12).
One-dimensional equations of conservation for mass, energy,
and momentum are used throughout; thus, fluid parameters such A^ + ^ + o p ^ + ^ + 5t/ |P=0 (10)
dt dx dx dX dx
as enthalpy, density, etc., refer to mean cross-sectional values, and
we limit this study at the outset to perturbations traveling along dii 1 95/)
+ U + o U — + — on — = - 6q = 0
the axis of the flow channels. In order not to limit the study to cir- dt dx BX p H
dx p zt pa
cular channels the geometry is described in terms of wetted perim- (11)
eter 0° and cross-sectional area fi.
Conservation Equations and the Equation of State. The
dx
general one-dimensional equations for conservation of mass, ener-
gy, and momentum for horizontal flow are well known and are + (2pUdU + SpU2)~ = 0 (12)
given without further comment.
The procedure of linearization limits the validity of these equa-
Dp dU_ tions, and hence any deduction therefrom, to an infinitesimal re-
(3)
Dt dx gion about the steady-state solution for the dependent variables.
We can, therefore, also use a local equation of state; i.e.,
D,
Pjft(H+ J/72) = q<P
a ' dt
(4)
- 6p .Qvs KLr , ,dv
(13)
DU _ -dp-pfU20> w. *PH
p (5)
Dt ~ dx 2a With this equation, Sp can be eliminated from (10), (11), and (12).
These equations relate the density p, velocity (7, enthalpy H, and We also take the Laplace transform with respect to time neglecting
the pressure p of helium within the channel, x is the length coordi- initial values and arrive at equations (14), (15), and (16).
nate along the axis of the channel, t is the time coordinate, / is the
friction factor and q is the uniform flux of heat entering the fluid {s-U' + UL)6H + | ( C 7 ' - UL)6U
stream from the wall. The operator D/Dt is (a/at + Ua/ax).
In addition, we assume that an equation of state for thermody- + r[s- U' + UL)6p = 0 (14)
namic equilibrium is always applicable and relates the density or
specific volume v to the enthalpy and pressure: ( s - fi + UL)6H + P5U- (rO + =)6f> •~6q = 0 (15)
P pa
v = v{H,p) o r p = p{H,p) (6) - FTJ
(s + U' + FU + UL)6U
Now the most comprehensive description of the thermodynamic
and volumetric properties of helium is given by the equation of - FTI
state of McCarty [7], This is in the more conventional form with + [r(U>+±f) •_AL] 5 / , = 0 (16)
temperature and density as independent variables; however, Arp
has recently produced explicit expressions of the form (6), [8]. In Here we have used the notation L for the operator a/ax, U' is writ-
the steady-state solutions given in the following, the older form ten for aU/ax and we have introduced the following constants
was used since a computer program was already available for this. ,dv.
{
Steady-State Solution. For the steady state, the x variation of
the helium thermodynamic properties and the velocity may be ob-
-dH]r (units: s -
tained from equations (3), (4), (5), and (6) without regard to the
conduction of heat in the channel walls. Thus, equations (3) and (This is the so-called reaction frequency introduced by Zuber and
(4) can be integrated immediately, without knowledge of the equa- used also by Priedly)
tion of state, when we neglect the variation of U2 compared to that
B = ~- ( u n i t s : S. cm" 1 g~l)
of H. This will always be true at the velocities encountered in this m
problem. The result is:
F ='• ( u n i t s : cm - 1 )
pU — c o n s t . = m/a (7) a
r
=0/fiK=-Q.(units: J-Pa"lg'
,dH,
q<?
(H - Ht) (x - xt) (8)
In general, in order to make this set soluble, a further equation is
where the subscript i denotes the value of a variable at some
required—the energy equation for the channel wall—embodying a
known inlet point (in this case either the entrance to the whole
description of the perturbed part of the wall heat flux Sq in terms
channel or one section of it) and m is the total mass flow rate of
of the other perturbations. However in all cases considered in this
helium.
study dq was found to be quite negligible on account of the low
Also, in terms of thermodynamic derivatives obtained from the
wall heat capacity at superconductor temperatures and the small
equation of state, the momentum equation (5) becomes
temperature dependence of superconductor losses.
^ * w , 9 v . <7<p fm2<pv Pressure-Drop Perturbations Across Flow Restrictions. If
dp I T W p f i 2a3 the steady-state pressure drop Ap may be related to the kinetic
(9)
dx energy of the upstream fluid by
1 + (-)
A£ = kpU2 (17)
This equation must be integrated numerically because of its non-
linear nature. We note that, when the variation of U2 is neglected, where k is the empirical orifice coefficient for the restriction, and
equations (8) and (9) are completely equivalent to the equations we assume that on the time scale of the anticipated flow oscilla-
for dp/dx and dT/dx given by Arp [9], which are written in terms tions hydrodynamic equilibrium always exists, then the perturba-
of more customary thermodynamic derivatives. tion in pressure drop is given in terms of dH and bU at inlet to the
Equations for the Perturbations. The variables U, H, p, q, restriction by
and p or v are now written as the sum of a steady part and a time
-kG' , o r ? ) off + 2kG6U
dependent perturbation, for example, U - 0 + bU(t). These sub- dti p
6<\p (18)
stitutions are made in (3), (4), and (5), the steady-state equations
subtracted, and only first order terms retained, leading to a set of
1 + ftrf)
dp' H

Journal of Heat Transfer NOVEMBER 1975 / 523

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 For the inlet flow restriction in the blow-down system of Fig. 2, SH negligible at frequencies of interest, and noting that for r = 0, U' =
= 0 and, typically, the denominator will approximate 1.0. Equation a, we arrive at (21) and (22).
(18) gives the pressure drop perturbation across the flow restric-
tion in response to the velocity fluctuation SU. Thus (s-n + ui)6H + (p/a)(n-UL)8u =o (21)
6A/J = 2kfi5U0 (19) ( s - a + UL)6H + 0617= 0 (22)
with ki referring to the orifice coefficient of the inlet restriction. Simultaneous solution of (21) and (22) for constant Q and /? yields
Cases of most interest to us here have involved quite small inlet re-
6U- c o n s t a n t = 5U{ (23)
strictions. In other words, we have been interested in low source
impedance. In this case SU s» 8Ui. Referring to Fig. 3, the forward and
transfer function J(s) is thus simply /3617
6H (e<n.S)T_ 1} +6H.e^-s)r (24)
1 n
J(s) = c o n s t a n t = The subscript i always refers to conditions at the zone inlet, and
2k fi
we have used the usual definition of residence time, T. Thus
For the downstream flow restriction, for either choked flow or
where the downstream pressure is regulated, equation (18) gives T= f*dx/U{s) (25)
the pressure fluctuation at the exit from the heat load channel.
Since this will be the nth zone we have Finally, we employ equation (16) with, again, r = 0, U' = Q and
LSU = 0 to get the pressure drop perturbation for the jth zone.
-kEG^)mu + 2kBG[6U]ntl
°Ai*i =
i + ^ [ ( g ) Hu
(20)
(^-H=r/'4[(o + T) w
u
Here we have usually been interested in rather large restrictions -~(s + a + FV)6U]dx (26)
permitting expansion almost to atmospheric pressure and hence,
By substitution from (24) this integration can be carried out.
no approximations have been made.
When the pressure drop perturbations for the n zones are summed
Methods of Solution and added to 6p„+i from (20) we arrive at the required Spi, or
Spi/SU, since SU is a common factor.
With reference to Fig. 3 we identify the feedback transfer func-
Equations (23) and (24) show how disturbances at the inlet to a
tion, H(s) as Spi/SUi. Then with the system of equations de-
zone are propagated: in the case of SU, with no change, but in the
scribed in the foregoing, the open loop transfer function J(s)H(s)
case of SH, in a complicated way. To understand physically the na-
of equation (2) can be calculated once we have, first, the steady-
ture of this propagation it is necessary to transform (24) back to
state solution, i.e.; U, H, and p, as a function of the x coordinate of
the time domain. Friedly [11] has considered this same solution as
the channel. Second, it is required to integrate equations (14), (15),
illustrating the nature of the response of hyperbolic systems. The
and (16), along the x coordinate, subject to the boundary condi-
first term is the response to the internal disturbance SU, while the
tions SH\ = 0 and bpn+\ given by (20), to get Spi. For any given
second is the response to the disturbance SHi at the boundary. The
value of s we have used two methods to do this. The first to be de-
inverse transforms, when SU is a unit impulse disturbance, are, re-
scribed is an approximate method in that the helium compressibil-
spectively,
ity is neglected, ie., r —• 0. The second method is exact in the sense
that a finite difference numerical approach is used to solve the 6Ht(t) = /3e n t [l - X(t - T ) ] (27)
complete unabridged set of equations.
Steady-State Solution. The steady-state solution is computed 5H2(t) = e^m^t - r) (28)
by numerical integration of the T, p equivalents of (8) and (9), The first shows an exponentially increasing contribution due to
using the Adams-Moulton predictor-corrector method, with a vari- the velocity disturbance which is cut off by the Heaviside function
able x increment to achieve any predetermined accuracy. At each 3C after a time T, the time taken for the furthermost fluid (i.e., at
point the equation of state of McCarty is called to produce the cor- the inlet) to reach the end of the zone. The second contribution
rect values of the thermodynamic variables and appropriate deriv- due to the inlet enthalpy disturbance makes itself felt after a time
atives. Transport properties are also supplied at each point from delay r. Thus, as is well known, the disturbances are propagated a t
correlations also given by McCarty [5]. These are used to evaluate the velocity of the fluid. In particular, we note that the specific
a Reynolds number and hence a friction factor by a standard for- volume disturbance is directly proportional to the enthalpy distur-
mula for turbulent flow. A heat transfer coefficient is also comput- bance and we can visualize a density or thermal expansion wave
ed at each point using the correlation of Giarratano, et al. [10]. travelling at the fluid velocity.
Approximate Solution with Neglect of Compressibility in We note also that SH; for the zone under consideration will be
the Perturbed Equations (Method 1). This simplified solution given by (27) and (28) applied to the preceding zone; so that, as-
follows the previous work of Zuber [3] and Friedly, et al. [4] for su- suming SHi was zero for the first zone, the net result for SH at the
percritical fluids. It is a very useful approximation in that analyti- end of the last zone is a piecewise smooth exponentially increasing
cal solutions may be obtained. These authors used a two-zone function of time cut off at r = XJTJ as sketched in Fig. 4.
model (a constant density zone and one of constant thermal ex- In contrast to the enthalpy disturbance propagation at finite ve-
pansion, (duldH)p) and defined the zone boundary in terms of an locity, equation (26) shows that the pressure drop disturbance re-
appropriate thermodynamic state. Thus, at the zone boundary, sulting from SH and SU appear across the zone without further
fluid properties were constant, but the space coordinate was per- delay, in this approximation. Equation (20) shows that this is true
turbed. In our solution we permit any number of zones and solve also for the pressure drop perturbation across the downstream flow
the equations for a general zone of constant (svldH)p. We define restriction which will be delayed exactly by the total residence
zone boundaries in terms of the space coordinate x at which, in time a from the initiation of the impulse SU.
steady-state solution, the appropriate thermodynamic state is E x a c t Solution (Method 2). For a given value of s, equations
found. Thus, in our solution the thermodynamic variables are per- (14), (15), and (16) are a set of first order, ordinary differential
turbed at zone boundaries, b u t the space coordinate is not. The equations for the dependent variables SU, SH, and Sp in the inde-
two approaches to defining a zone are equivalent, but we feel the pendent variable x. However, we find a complicated x dependence
one chosen here is more useful in a multizone computation. in the coefficients £/', £), r, p, and F. For an accurate solution,
Considering then, a general zone, we set r = 0 in equations (14) therefore, it appeared that only a numerical method would suffice.
and (15). We also neglect the term s Sp/p in (15), which is always The boundary conditions specify SUi and SHi at the inlet to the

524 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Sp ! ~ b
i
SH?
~e!
i, i>i Su2 - h.

b
Sp2 0
°2 2
f
2 dj H SH, 0

>1 ?£ h 2 8u 3 0
92 h 2

0 ,. : ,
-* "
S
PRH 0
SHn 0

___S ' ^di-A-^i.--.. SU„ 0

i I I ' °n »„ IS, < K S

Pn 0
SHnt1 0
i ! %\\<„<K SUn+l 0

By partitioning into block-tridiagonal form followed by triangular

decomposition [12], this set is solved without having to invert a
more than 3 X 3 matrix.
Setting bU\ - 1 the solution gives Spi/SUi for the particular
value of s chosen so that a Nyquist plot can be constructed. Typi-
cally, for 100 values of s the computation for n = 30 takes 2 min.
Fig. 4 Illustration of time development of enthalpy perturbation in re-
sponse to a unit impulse velocity disturbance at inlet R e s u l t s of C o m p u t a t i o n s
A c c u r a c y of t h e Computations. Confidence in the accuracy
of the methods described was established by computations on the
problem defined in Fig. 5(a). This problem was chosen for its
heated channel, but <5p„+i at the outlet. Therefore, an implicit fi- closeness to the critical point, and the fact that the helium crosses
nite difference scheme must be used. Thus, we select n zones over the transposed critical line in its course along the heated channel.
each of which we assume the dependent variables and the x depen- It was felt that this problem represented as severe a variation in
dent coefficients are linear. Then for the jth zone we have, for ex- helium properties as would be encountered in practice, and there-
ample fore, if it could be accurately computed, all problems of practical
interest could be computed with at least the same accuracy.
5H=(5Hj + 6HjA)/2 (29)
We first determined that at least 8 zones were required in Meth-
L6H = {5Hj,l- dH^/^Xj (30) od 1 for the quantity | J(s)H(s)\ to be within one percent of its as-
ymptotic value. Method 1, which uses closed form integration,
where Ax is the zone length. The subscriptj refers to the inlet of could then be used to check Method 2 in the incompressible ap-
the y'th zone and j + 1 the outlet. In this way we can write down 3ra proximation, i.e., using the same input data but setting r = 0, U' =
equations and we shall have 3(n + 1) — 3 = 3n unknowns for the ft and neglecting the term s Sp/p in equation (15). This would es-
dependent variables at the zone boundaries. For the jth zone we tablish the validity of the finite difference approach of Method 2.
have the equations For the Method 2 calculations we set n = 96. We found the agree-
ment most encouraging. The worst discrepancy, as might be ex-
afiHj + bfiUs + cfipj + a/6HJtl + 6/6tf i + 1 + c/5pjtl = 0 pected, is at the point of resonance indicated by a peak in the sys-
dfiH, + efiV, +fj6pj + rf/6ff.+1 + efiU^ + ffipjtl = 0 tem gain. Here there was a 2.6 percent discrepancy in \J(s)H(s)\.
gfiHj + hfiUj + ifip, + gfiHjtX + k/SU^ + i/5pjti = 0 Over most of the range of s the discrepancy was less than 0.5 per-
cent.
(31)
The problem was next recomputed using Method 2 in the full
The coefficients a,,fay,etc., may be complex on account of the La- compressible case with n = 96 again. The Nyquist plot of J(s)H(s)
place transform variable s. In matrix form we have equation (32). for this case is shown in Fig. 5(a). In Fig. 6 we show the system

200
< « IO~!

50- 50- 100

80
<«-\50 100 150 150

(t &' / ' ( REAL REAL (9 'j

soN
^CIlS -50

-INCOMPRESSIBLE
APPROXIMATION

Fig. 5 (a) Nyquist plot of the open loop transfer function for a channel of
length 240.0 m, dia 0.5 cm, heat load 0.15 W/m—m = 1.0 g/s, T = 4.5 K,
p = 2.5 atm, k, = 20.0, kE = 20.0, n = 96. (b) Nyquist plot of the open
1 2 4 6 10 20 40 60 100 200
loop transfer function for a superconducting power transmission line. Rep-
FREQUENCY ( H i ) x 10-"
resentative Mb3Sn conditions: length 5.0 km, dia 5.0 cm, heat load 0.40
W/m—m = 138 g/s, T = 6.5 K, p = 10.0 atm, k, = 20.0, kE = 2000, n = Fig. 6 System transfer function plotted versus frequency for conditions of
50. Fig. 5(a)

Journal of Heat Transfer NOVEMBER 1975 / 525

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ' ' ' ' ( i|
—-—!___ ' < < < ' |
T=S

100
80
100 200
1__ J
(O /REAL
-50f V_^
(V: -100

1 2 4 6 10 20
L
40
FREQUENCY (Hz)
60 100 200
x 10"4

Fig. 7 Open loop transfer function and phase angle versus frequency for
Fig. 8 (a) Nyquist plot of the open loop transfer function for a supercon-
ducting power transmission line. Representative Nb conditions: length 5000
m, dia 3.0 cm, heat load 0.120 W/m—rh = 52.7 g/s, T = 4.5 K, p = 3.0
atm, k, = 20.0, kE = 900, n = 50. (o) Off-design conditions for line of Fip.
8(a): heat load 0.250 W/m, m = 43.9 g/s (other conditions unchanged).

conditions of Fig. 5(a)

gain, F(s), plotted against frequency in Hz showing a sharp reso- In Fig. 8(a) we give the Nyquist plot for a Nb case with a 3 cm
nance at 0.0034 Hz. Fig. 7 shows the open loop gain | J(s)H(s)\ with dia channel again computed with n = 50 and hi = 20. Because of
the phase angle superimposed in the usual Bode plot. For compari- the lower pressure in this case the orifice coefficient at the exit is
son we have indicated in Fig. 5(a) the results of the aforemen- limited to about 1600 if the return pressure is set at no less than
tioned incompressible calculation. 1.5 atm. We arbitrarily chose 900 for this calculation. Again, many
variations were computed for the Nb case, but all showed that at 3
From these plots we see that neglect of compressibility is not
atm, with a temperature range 4.5-5.3 K the system is stable.
conservative as regards stability prediction. Whereas the incom-
The operating temperature range for the Nb case was purposely
pressible computation predicted (fortuitously) marginal stability,
chosen to stay below the transposed critical temperature (5.58 K at
the rigorous computation predicts a definite instability at the reso-
3.0 atm) and this indeed proved a wise choice. For suppose now
nant frequency. The error in \J{s)H(s)\ caused by this neglect
that through a partial vacuum insulation failure, for example, the
ranges from 17 percent at s = 0-82 percent in the region of reso-
heat load were doubled in the above 3 cm dia case, and, to avoid
nance.
excessive pressure drop, the flow rate were reduced slightly such
Although the accuracy obtained with n = 96 is certainly very ad-
that the temperature at outlet increased to 6.38 K. The Nyquist
equate, the computation time for 100 values of s was 7 min. With n
plot of Fig. 8(6) illustrates this case and it is apparent that this cir-
= 30 the same computation took 2 min. We therefore wished to
cumstance has driven the system into an unstable condition. The
know the error involved in relaxing the space grid this much. A
cause of this instability is to be found in the expansion which takes
comparison showed the worst error to be 6 percent in the frequen-
place across the 5000 m of transmission line. In the NbgSn case the
cy range examined. However, in the vicinity of the resonance it was
specific volume ratio at outlet to inlet is 1.5. In the Nb case for
only 0.4 percent. We, therefore, felt that for most purposes nothing
normal operation this ratio is 2.0 while in this off-design case it is
important was lost by so relaxing the grid.
6. We note that the NbaSn case under normal conditions operated
In conclusion, it appears that Method 2 is capable of very ac- in a temperature range encompassing the transposed critical tem-
ceptable accuracy for quite reasonable computation times, and by perature (8.15 K at 10 atm), so there is here no possibility of devel-
taking full account of compressibility in the calculation some con- oping a similar instability through an accidental increase in heat
siderable errors are avoided. load.
Illustrative Examples. In this section we give results comput-
ed by Method 2 for some selected cooling system parameters which
Acknowledgment
we believe to be representative of possible a-c superconducting
power transmission lines. Vincent Arp is to be thanked for his constant interest in this
work and in particular for the computer program for the computa-
Computations were performed for two different conductors:
tion of the steady-state conditions. The continued encouragement
pure niobium (Nb) and niobium tin (NbsSn). Each requires a dif-
of Eric Forsyth and Jack Jensen of the Brookhaven National Lab-
ferent set of operating conditions. Thus, for Nb we selected tem-
oratory is also highly appreciated.
peratures of 4.5 K at inlet and 5.3 K at outlet. For NbsSn we se-
lected 6.5-8.5 K. In order to achieve sufficient density, the NbsSn
cable must operate at 10 atm while the Nb cable may operate at 3
References
atm. Specifying the length at 5000 m leaves only the channel diam-
1 Bogner, G., "Transmission of Electrical Energy by Superconducting
eter and wall heat flux to be specified; the mass flow rate of helium Cables," Superconducting Machines and Devices, Plenum Press, New
then follows. York, 1974 pp. 401-562.
We selected total wall heat fluxes—a-c losses, dielectric losses, 2 Bour£, J. A., Bergles, A. E., and Tong, L. S., "Review of Two-Phase
Flow Instability," ASME Paper No. 71-HT-42.
and heat in-leak—of 2.55 W/m 2 for NbsSn and half of this for Nb, 3 Zuber, N., "An Analysis of Thermally Induced Flow Oscillations in
these being typical projected figures. the Near-Critical and Supercritical Thermodynamic Region," Final Report
As a representative example of the NbsSn case, we give in Fig. NAS8-11422, NASA Marshall Space Flight Center, Huntsville, Ala., May
5(6) the Nyquist plot of J(s)H(s) for a 5 cm dia helium channel 1966.
4 Friedly, J. C, Manganaro, J. L., and Kroeger, P. G., Final Report
computed by Method 2 with n = 50, kj = 20 and kE = 2000. The NAS8-21014, NASA Marshall Space Flight Center, Huntsville, Ala., Oct.
value chosen for kj represents a small restriction—an orifice diam- 1967.
eter 50 percent of the pipe diameter, ks was chosen for a nozzle 5 McCarty, R. D., "Thermophysical Properties of Helium-4 From 2 to
permitting expansion down to about 1.5 atm. Several variations of 1500 K With Pressures to 1000 Atmospheres," Nat. Bur. Stand. (U.S.) Tech-
this case were also computed with different ks and different diam- nical Note, Nov. 1972, p. 631.
6 Murphy, G. J., Basic Automatic Control Theory, Van Nostrand,
eters, but all show essentially the same thing: that this is a very Princeton, N. J., 1957.
stable operating condition. 7 McCarty, R. D., "Thermodynamic Properties of Helium 4 From 2 to

526 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1500 K at Pressures to 108 Pa.," J. Phys. Chem. Ref. Data, Vol. 2, No. 4, 10 Giarratano, P. J., Arp, V. D., and Smith, R. V., "Forced Convection
1973, pp. 923-1041. Heat Transfer to Supercritical Helium," Cryogenics, Vol. 11,1971, p. 385.
8 Arp, V. D., "New Forms of State Equations for Helium," Cryogenics, 11 Friedly, J. C , Dynamic Behavior of Processes, Prentice-Hall, N. J.,
Vol. 14, No. 11, Nov. 1974, pp. 593-598. 1971.
9 Arp, V. D., "Refrigeration of Superconducting Rotating Machinery," 12 Westlake, J. R., A Handbook of Numerical Matrix Inversion and So-
Nat. Bur. Stand. (U. S.) NBSIR, June 1973, pp. 73-331. lution of Linear Equations, Wiley, New York, 1968.

Journal of Heat Transfer NOVEMBER 1.975 / 527

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 E. M. Sparrow
R. J. Goldstein Effect of Nozzle—Surface
M. A. Rouf

Department of Mechanical Engineering,

Separation Distance on
University of Minnesota,
Minneapolis, Minn.
Impingement Heat Transfer for a
Jet in a Crossflow
Measurements were made of the local heat transfer coefficients resulting from the im-
pingement of a turbulent jet which interacts with a crossflow. The experiments were
performed for separation distances between the exit of the jet nozzle and the impinge-
ment surface ranging from 3 to 12 times the jet exit diameter. The ratio of jet to cross-
flow mass velocity was varied from 4 to 12. To evaluate the heat transfer coefficients,
local recovery temperatures were also measured. The results for the larger mass velocity
ratios indicated that as the separation distance is varied, the impingement point heat
transfer coefficient attains a maximum when the separation distance is 5-6 times the jet
diameter. On the other hand, at a smaller mass velocity ratio such as 4, the maximum is
attained at a substantially smaller separation distance. Experiments were also per-
formed to examine how the transfer coefficients are affected by the presence or absence
of a simulated duct wall at the exit of the jet nozzle. The differences between the results
for the two configurations were confined to dimensionless separation distances greater
than 4, with the largest effects at low mass velocity ratios.

Introduction research was undertaken with two main objectives. The first was to
determine how the local heat transfer coefficients on the impinge-
In recent years, cooling or heating of a surface by impinging jets ment surface are affected by variations of the separation distance
has become an established technique. Since relatively high local between the jet nozzle and the surface. In particular, it was of spe-
coefficients are obtainable compared with those for nonimpinging cial interest to explore whether the heat transfer coefficient at the
flows, the jet impingement technique provides a designer with a impingement point attained a maximum value as the separation
means for more effective control of the temperature of the system distance was varied.
being heated or cooled. In response to a need for basic heat trans- A second objective was to investigate the influence of the jet exit
fer information, a number of investigations of impinging jets have configuration on the local heat transfer results. To this end, two
been performed. In the main, these studies have been concerned exit configurations were employed. In one, the end of the tube
with jets impinging normal to a surface situated in an otherwise from which the jet issued was fitted with a plate aligned with the
quiescent environment. There are, however, practical applications crossflow direction. The intended function of the plate was to in-
where the jet interacts with a crossflow. Such a crossflow interac- hibit flow field phenomena induced by the presence of the jet tube
tion is encountered, for example, in the internal cooling of turbine from influencing the interaction of the jet with the main stream.
blades. Such flow field phenomena would, of course, be absent if the jet
The present investigation is concerned with impingement heat were to issue from an aperture in a wall. In the second case, the
transfer associated with a turbulent jet situated in a crossflow. The plate was omitted, thereby permitting a freer interaction.
The experiments were performed with air as the fluid for both
the jet flow and the crossflow. During the course of the work, the
ratio of separation distance to jet exit diameter was varied from 3
to 12. A second parameter that was varied was the ratio of the
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer mass velocities of the jet and the crossflow. This ratio ranged from
Division March 6,1975. Paper No. 76-HT-F. 4 to 12.

528 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 There have been very few publications dealing with turbulent
impinging jets in the presence of a crossflow, and those have been
purely experimental. Kercher and Tabakoff [l] 1 investigated the
impingement cooling characteristics of a two-dimensional array of
circular jets. A controlled crossflow was not employed in the exper-
ELF.CTRICALLY HEATED SEGMENTS
iments, but the spent air from the other jets in the array was, in ef-
fect, a crossflow. The separation distance to jet diameter ratio was
JET TEMPERATURE
confined to small values. In the range investigated, the heat trans-
fer coefficient increased monotonically with the distance-diameter
ratio and a maximum was not encountered.
Metzger and Korstad [2] measured average heat transfer coeffi-
cients on a surface on which there impinged a single line of circular MEASUREMENT OF
CROSS FLOW VELOCITY SIMULATED DUCT WALL
jets situated in a crossflow. Measurements were made for three AND TEMPERATURE (SPLITTER PLATE)
values of the separation distance—jet diameter ratio (2, 3.5, and Fig. 1 Schematic top and side views (upper and lower diagrams, respec-
6.7), but the results did not show a clear cut trend as the ratio was tively) of the experimental apparatus
varied. Recently, Bouchez and Goldstein [3] performed experi-
ments with a turbulent jet in a crossflow and determined local heat
transfer coefficients, impingement cooling effectiveness values
(analogous to film cooling effectiveness), and impingement point ulating and flow control valves, an orifice metering station, and
locations. Inasmuch as the experiments were carried out for only electric heaters to facilitate temperature control.
two values of the ratio of separation distance to jet diameter (6 and Schematic diagrams showing top and side views of the wind tun-
12), the detailed effect of spacing could not be definitely estab- nel are presented in Fig. 1. As seen in the side view (lower di-
lished. The impingement point transfer coefficients were higher at agram), the jet air passes from its delivery pipeline into a tube
the smaller separation distance. which protrudes downward into the wind tunnel from the upper
Experiments on impinging turbulent jets without crossflow have wall. Tubes of various lengths were employed to obtain the desired
demonstrated that the impingement point heat transfer coefficient range of separation distances between the jet exit plane and the
attains a maximum value as the ratio of separation distance to jet lower wall of the wind tunnel, which served as the impingement
diameter is varied. Gardon, et al. [4, 5] found the maximum to surface. All tubes had a common internal diameter of 1.27 cm.
occur when the ratio was in the range from 6 to 9, depending on As discussed earlier, some of the data runs were made with a
the jet Reynolds number and nozzle diameter. A literature survey plate affixed to the lower end of the jet tube. The plate, which will
by Gauntner, Livingood, and Hrycak [6] concluded that the maxi- hereafter be referred to as the splitter, was aligned parallel to the
mum heat transfer coefficient should occur when the impingement tunnel throughflow. Its purpose was to isolate the jet—crossflow
plate is separated from the nozzle exit by the length of the poten- interaction from the flow patterns induced by the presence of the
tial core. The survey also points out that various core lengths have jet tube. The plate dimensions (10.16 cm long by 6.35 cm wide) de-
been obtained by different investigators. A core length of 6.1 jet picted in Fig. 1 are to scale with the duct dimensions. The plan-
dia was recommended for applications in which the actual core form of the plate is shown in the upper diagram of the figure.
length is not known. As can be seen in the upper diagram, the impingement surface
(lower wall of wind tunnel) was made up of 21 electrically heated
Apparatus and Experimental Procedure segments. Each segment was a 0.00254-cm thick stainless steel foil,
Experimental A p p a r a t u s . The description of the apparatus electrically isolated from its neighbors; the foils were connected in
and procedure will be aimed at presenting a clear picture of the ex- series and carried a common current. This heating arrangement
perimental work, but without exhaustive details. These details are gave rise to a nearly uniform heat flux to the flowing air. The heat-
contained in the thesis from which this work is drawn [7] as well as er segments spanned the 20.3-cm width of the wind tunnel and had
that of Bouchez [8]. a length of 5.08 cm in the flow direction. For support and insula-
The experiments were performed in a wind tunnel having a 15.2 tion, the segments were bonded to a textolite plate backed with
by 20.3 cm rectangular cross section (height by width) and a length styrofoam (textolite is a phenolic and cotton fabric laminate).
of about 115 cm. Air from the laboratory room was drawn through Thermocouples were affixed to the rear surfaces of the heater
a contraction section into the wind tunnel by a blower situated at segments by copper oxide cement. A total of 55 thermocouples
the downstream end. The tunnel throughflow served as the cross- were longitudinally distributed along the center span of the seg-
flow for the impinging jet. The air for the jet flow was taken from a ments, that is, along the x axis. The thermocouples were heavily
regulated building air supply. The insulated pipeline that deliv- concentrated on the segments that encompassed the expected im-
ered the jet air to the upper wall of the wind tunnel contained reg- pingement region of the jet. There were eight thermocouples on
each such segment. All of the thermocouples were made from a cal-
ibrated roll of wire.
In addition to the instrumentation already mentioned, sensors
E
Numbers in brackets designate References at end of paper. were installed to measure the velocity and total temperature of the

•Nomenclature.
cp = specific heat at constant pressure M = ratio of mass velocities, equation (1)
D = jet exit diameter q = local heat transfer per unit time and T „ = static temperature of crossflow
h = local heat transfer coefficient, equation area T „ c = total temperature of crossflow
(3) r = recovery factor, equation (4) Uj = jet exit velocity
ho = local heat transfer coefficient due to Tj = static temperature at jet exit u „ = crossflow velocity
crossflow alone Tj° = total temperature of jet x = streamwise coordinate
L = separation distance between nozzle TV = recovery temperature Pj = density at jet exit
and surface Tw = wall temperature p» = density of crossflow

Journal of Heat Transfer NOVEMBER 1975 / 529

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 crossflow and the total temperature of the jet flow. For the latter, a ficients defined in accordance with equation (3) are independent of
thermocouple was placed in the jet delivery pipeline, where the ve- the level of the heat flux q.
locity is relatively low. As indicated in Fig. 1, the thermocouple is The effect of using (Tw — T„) in the definition of h instead of
about 10 jet dia upstream of the exit of the delivery pipeline. Cor- (Tw — Tr) will be illustrated later. For the present, it is relevant to
roborating thermocouples were also installed on the wall of the note that if the temperature difference between the wall and the
pipe. The velocity and total temperature of the crossflow were fluid were large, the recovery effect would be unimportant. Under
measured at half height and midspan in a cross section 19 cm up- these conditions, it is not necessary to know Tr to apply the re-
stream of the jet discharge. An impact probe and a wall static tap, sults. In the present experiments, the wall-fluid AT's were small at
in conjunction with a capacitance type pressure meter, were used some surface locations, so that the use of (Tw — Tr) was appropri-
for the velocity measurement, and the temperature was measured ate.
with a thermocouple. To present the results on a dimensionless basis and, at the same
The thermocouples that were installed to measure total temper- time, to provide a comparison that is of practical interest, the local
atures actually read recovery temperatures. Computations indicat- heat transfer coefficients are expressed in terms of the ratio h/ho-
ed that for the conditions of the experiments, the difference be- In this ratio, h is the local coefficient in the presence of the jet and
tween the stream total and probe recovery temperatures is about the crossflow, whereas ho is the local coefficient (at the same
0.02°C, which is negligible. streamwise location) due to the crossflow alone. Measurements of
Experimental Procedure. A given data run was defined by ho were made with the tunnel free of any obstructions, i.e., neither
setting the velocity of the crossflow, the velocity and total temper- the jet tube nor the splitter were present. The departure of h/ho
ature of the jet, and the separation distance between the jet exit from unity identifies the surface locations at which the jet aug-
and the impingement surface. These enabled evaluation of the ra- ments the transfer coefficients. The streamwise distributions of
tios h/ho will be plotted against the streamwise position coordinate
x/D. As illustrated in Fig. 1, the origin of the x coordinate is situat-
(a) M = (pu)j/(puU, (b) L/D (1) ed directly beneath the center of the tube from which the jet
where M is the mass velocity ratio, L is the separation distance, emerges.
and D is the jet exit diameter. The mass velocity ratio M was em- The streamwise distribution of h/ho will be parameterized by
ployed in lieu of other candidate groupings involving pj, Uj, p», and the mass velocity ratio M = (pu)jl(pu)*, and the dimensionless
u „ because it had been found to be an effective correlating param- separation distance L/D. The velocities uy and u», respectively,
eter by Bouchez and Goldstein. represent the mean jet velocity and the free stream crossflow ve-
There is virtually an unlimited number of choices that can be locity.
made in setting the temperature of the jet relative to that of the The measured recovery temperatures were also expressed in
crossflow. The effect of temperature level was explored by Bou- terms of a dimensionless group, the recovery factor r,
chez and Goldstein [3] and was not one of the research objectives
T — T- T — T°
of the present investigation. Consistent with the present objec-
tives, it was deemed appropriate to select a standard temperature uj2/2cp Uj2/2cp
operating condition, namely, a matching of the total temperatures
where Tj is the static temperature of the jet at exit. Inasmuch as
of the jet and the crossflow. If T° denotes the total temperature,
Uj 2 /2c p represents the difference between the stagnation and static
then the matching condition is
temperatures of the jet, the recovery factor affords a comparison
Tj° = T„° (2) between the recovery temperature rise and the stagnation temper-
ature rise. The streamwise distributions of r are parameterized by
A data run consisted of two distinct parts, with the velocity and M and L/D.
air temperature conditions maintained constant throughout. In To complete the parametric description of the experimental re-
the first part, the temperature distribution along the impingement sults, two additional specifications are relevant. The first has to do
surface (i.e., along the x axis) was measured under the condition with the relative temperatures of the jet and the crossflow and, in
that no electric power was supplied to the heater segments. These the present investigation, total temperatures were matched as in-
measurements yielded recovery temperatures Tr(x) associated dicated in equation (2). The second relates to the velocity range of
with the jet impingement (the crossflow velocities were too low to the crossflow. Owing to the trip wire situated at the upstream end
cause a recovery effect). The maximum value of Tr(x), relative to of the test section (see Fig. 1, upper diagram), a fully turbulent
Tj° or T„° was about 2°C. In the second part, electric power was boundary layer can readily be obtained along the lower wall of the
supplied to the heater segments, and measurements were made of wind tunnel [8]. In view of this, there is little motivation to para-
the wall temperature distribution Tw(x) and of the power input. metrically vary the crossflow velocity. Therefore, for all data runs,
The power level was selected so that variable property effects in the velocity was held fixed at 12 m/s, which corresponds to a Reyn-
the air would be negligible {Tw — T „ < 20°C). As will be discussed olds number of 720,000 per meter (220,000 per foot). The stream-
shortly, the data from both parts of a data run were needed in wise distribution of ho corresponding to this flow condition will be
evaluating local heat transfer coefficients. presented later. For the jet, the nozzle exit Reynolds numbers
ranged from 38,000 to 115,000.
Results and Discussion Local Heat Transfer Coefficients. To show how the local
Presentation Variables and Parameters. Local heat transfer heat transfer coefficients respond to changes in the dimensionless
coefficients were evaluated from the definition nozzle-to-plate separation distance L/D, Figs. 2-4 are presented.
Each figure contains two graphs, with each graph corresponding to
h = ql(Tw - Tr) (3)
a specific L/D. In this way, the three figures give results for six
The local heat flux q for each heater segment was determined from values of L/D: 12,10, 8, 6, 5, and 3. The case L/D = 12 corresponds
the measured electric current flow in conjunction with the resis- to the situation in which the nozzle exit is flush with the upper
tance of the segment. The determination of Tw and Tr has already wall of the wind tunnel. To conserve space, results for other L/D
been discussed in the prior section of the paper. The use of (Tw — values within this range are not presented here, but are available
Tr) as the thermal driving force for convective heat transfer is well in [7]. Figs. 2-4 are for the data runs with the splitter affixed to the
established for flows in which there is temperature recovery when lower end of the jet tube. The results for the data runs without the
the wall is adiabatic. Among all candidate thermal driving forces splitter show identical trends and are discussed later.
for such flows, only (Tw — Tr) goes to zero when q goes to zero. In each graph, h/ho is plotted as a function of the dimensionless
Furthermore, it was demonstrated in [3, 8] that heat transfer coef- streamwise coordinate x/D. The curves are parameterized by the

530 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 12

10
- J
L / D = 12
8
h/h„
6
M
o 12.1
« 9.8
1 L/D 8
1 L/D = 6

o 8.1
" 3.9
4
—
M IVk M
2 1\ o 12.0 LV? o 12.0
a 10.1
—
\V\
Uj 1 \ \ ° 8.1
\\i " 10-1
ao
Yw °
1 1 I
10
X/D
1
20
! !
Iff 4.1

if
\\V » 4.1

Fig. 2 Streamwise variation of the local heat transfer coefficient along

the impingement surface, L/D = 12 and 10, with splitter

1 1 1 1 1 1 1 1
10 20 10 20
X/D X/D
mass velocity ratio M = (pu)j/(pu)„, with nominal values of 4, 8, Fig. 3 Streamwise variation of the local heat transfer coefficient along
10, and 12. Each curve depicts the distribution of h/ho along the x the Impingement surface, L/D = 8 and 6, with splitter
axis.
For the higher jet velocities, M = 8, 10, and 12, the distribution
curves all have a common behavior. At locations that are substan-
tially upstream of the impinging jet, h is within a few percent of An exception to the trends cited in the last several paragraphs
ho- As the impingement region is approached, h/ho increases. This occurs when the jet does not impinge directly on the surface, but
increase in h/ho is manifested several jet diameters upstream of rather is swept downstream by the crossflow. This behavior is in
the actual zone of impingement. As identified in flow visualization evidence for the lowest mass velocity ratio M = 4 and at the largest
studies by Bouchez and Goldstein [3, 8] and others (e.g., [9]), there separation distances L/D = 12 and 10. As the separation distance
is a highly turbulent recirculation zone situated upstream of the is diminished, the distribution curves for the M = 4 case display a
impingement area which is responsible for the upstream augmen- steadily growing peak at the impingement point.
tation of h. The upstream penetration of the recirculation zone is
To provide further perspective about the effect of separation
increased with increases of the mass velocity ratio, and the heat
distance on the impingement point transfer coefficients, the peak
transfer coefficients also increase. A similar outcome is brought
values from Figs. 2-4, together with additional data from [7], have
about by decreases in the separation distance.
been brought together in Fig. 5. In this figure, (h/ho)peak is plotted
Within the impingement zone, the curves rise to a peak, the against L/D for parametric values of M. The curves in the left-
height of which is accentuated with increasing values of the mass hand graph correspond to data runs with the splitter plate present,
velocity ratio. For a fixed value of the mass velocity ratio, the whereas those in the right-hand graph are for the case of no split-
peaks grow higher as L/D diminishes from 12 to 5, but further de- ter.
creases in L/D bring about a lowering of the peak. This interesting With the splitter present, the curves for M = 8, 10, and 12 take
result will be further illuminated shortly. on their respective maximum values at L/D = 5. However, for M =
Downstream of the impingement zone, a wall jet develops along 4, it appears that the largest of the peak values will be encountered
the surface and the h values decrease, tending to approach ho- when L/D < 3. For the case of no splitter plate, the results display
Considerable augmentation still remains, however, even at x/D = their.largest peak values at approximately the same L/D as the
40 (e.g., h/ho ~ 2 for M — 12). At a fixed separation distance, the aforementioned; for M = 12 and no splitter, the maximum appears
distribution curves in the downstream region are consistently or- to be closer to L/D = 6. A graph similar to Fig. 5 was prepared [7]
dered according to increasing values of the mass velocity ratio. in which h (rather than h/ho) was plotted as a function of L/D.
From this graph, which is omitted to conserve space, it was ob-
served that the L/D corresponding to the largest of the peak values
of h is the same as the L/D at the largest of the peaks of h/ho.

I T L/D=3
WITHOUT SPLITTER

M
tl ° 12.1
—• m Ytt A 9.9
111 Q 8.1
' l\ v 4.1

ft'

1 t 1 1 1 1

Fig. 4 Streamwise variation of the local heat transfer coefficient along Fig. 5 Dependence of the Impingement point transfer coefficient on sepa-
the impingement surface, L/D = 5 and 3, with splitter ration distance

Journal of Heat Transfer NOVEMBER 1975 / 531

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 It is interesting to compare the just-determined L/D corre-
M
sponding to the largest impingement point transfer coefficient to
o 12.0
those obtained by prior investigators in the absence of crossflow.
Gardon and Cobonpue [4] found L/D in the range from 6 to 9, with
the spread primarily due to a variation in the nozzle diameter. For
their largest nozzle diameter (~0.9 cm), which is somewhat smaller
than the 1.27 cm dia nozzle of the-present experiments, their re-
S
sults indicated an L/D value of about 6. The Gauntner literature _
v •%& L/D
WITHOUT
5
SPLITTER
9
survey [6] recommended a value of L/D = 6.1, albeit without a firm V

basis. The present experiments give L/D ~ 5-6 for M = 8, 10, and
12, and this is in very close agreement with the findings for no J I I I L I I
-10 0 10 20 30 40
crossflow. However, for a low mass velocity ratio such as M = 4, X/D

the separation distance corresponding to the largest impingement Fig. 7 Illustrative recovery factor results
point transfer coefficient is substantially less than that for no
crossflow.
The existence of an optimal separation distance which affords a
maximum impingement point transfer coefficient has been well ex-
plained by Gardon and Akfirat [5], so that only a brief discussion stream regions. On a percentage basis, the largest effects are in evi-
will be included here. At small separation distances, when a poten- dence for the smallest mass velocity ratio M = 4 where, for in-
tial core exists within the jet, the impinging core velocity will be stance, for L/D = 8, the peak values for the two cases differ by
essentially unchanging as the separation distance increases. On the about 25 percent. In addition, in the presence of the splitter, the
other hand, measurements have shown that the turbulence level downstream deflection of the M = 4 jet is greater than in the ab-
within the core flow increases. Consequently, the impingement sence of the splitter. As a tentative explanation of these findings, it
point heat transfer coefficients should also increase. With further can be reasoned that since the splitter inhibits fluid from above
increases in separation distance, the core is ultimately engulfed by the jet exit plane from participating in the jet mixing process, the
the jet mixing processes and the impingement velocity decreases turbulence level of the jet is somewhat reduced. This should re-
and, subsequently, the turbulence also decreases. This brings duce the heat transfer coefficients of the impinging jet. Another ef-
about a decrease in the impingement heat transfer coefficient. fect of the blockage of fluid downflow by the splitter is to decrease
This pattern of an initial increase and a subsequent decrease of the the pressure downstream of the jet. This lowering of the pressure
transfer coefficient with increasing separation distance is consis- tends to pull the jet downstream, resulting in a greater deflection.
tent with the experimental results. At the smaller separation distances (L/D = 4 and 3), the h/ho
Attention may now be turned to a comparison of the local trans- curves for the two cases are very nearly coincident. It is, thus, seen
fer coefficients obtained with and without a splitter plate. Local that the influence of the splitter is confined to separation distances
coefficients for the no-splitter case were measured for six separa- greater than L/D = 4, with the greatest effects at low mass velocity
tion distances, and these are available in [7]. For establishing the ratios.
relationship between the results for the splitter and the no-splitter Before concluding this section of the paper, there are a few addi-
cases, observations obtained by overlaying corresponding figures tional observations that may be made about the distribution
for the two cases will be reported. In addition, for purposes of cor- curves of Figs. 2, 3, 4, and 6. The first relates to the shifting of the
roboration, results for the no-splitter case for two L/D values (8 location of the peak with variations of L/D and M. This shifting is
and 3) are presented in Fig. 6. related to the degree of deflection experienced by the jet owing to
From the aforementioned comparisons, it is found that at larger the action of the crossflow. The extent of the deflection and the
and intermediate separation distances (down to L/D = 5), the h/ho corresponding downstream movement of the peak is accentuated
results in the presence of the splitter are somewhat lower than at larger L/D and lower M. The second item of interest is the local
those for the no-splitter case. The differences between the two sets maxima that flank the main peak in the distribution curves when
of results are primarily confined to the impingement and down- the separation distance is small. These local maxima become more
prominent at larger M values. They are believed to indicate a tran-
sition from laminar to turbulent boundary layer flow. Similar local
maxima were encountered previously by Gardon and co-workers
[4, 5] in the absence of a crossflow.
O t h e r Results. A complete presentation of recovery factor re-
sults is available in [7], and only a single figure, Fig. 7, will be given
here for illustration. In Fig, 7, r is plotted versus x/D for the case
of L/D = 5 and no splitter. T h e trend in the data for M = 8, 10,
and 12 is typical of that encountered at the other separation dis-
tances, except for differences in detail. On the other hand, the
plotted points for M = 4 are, in a sense, random. This is because
the temperature difference (Tr — Tf) which enters the definition
of r (equation (4)) is typically in the range of 0.1-0.2°C for M = 4.
Thus, a measurement uncertainty of a few hundredths of a degree
(due, for example, to variations in the air temperature) can have a
significant effect on r. Fortunately, temperature recovery played
an inconsequential role in the evaluation of the heat transfer coef-
ficients for M = 4.
Perhaps the most noteworthy feature of Fig. 7 is the fact that r
> 1 in the impingement region. It is believed that this finding is re-
lated to the fact that the jet static temperature at the nozzle exit is
less than that of the surrounding crossflow and, as a consequence,
-10 0 10 20 30 40 -10 0 10 20 30 40
X/D X/D the jet gains enthalpy as it entrains surrounding fluid. In order
Fig. 6 Distributions of the local heat transfer coefficient for the no-splttter that the higher enthalpy of the entrained fluid have a significant
case effect on the impingement point recovery factor, it must penetrate

532 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 40

J L

Fig. 9 Local surface heat transfer coefficients for the crossflow without
Jet Impingement

al values of h may be desired rather than the h/h0 ratio. To facili-

tate the determination of the h values, a graph of h0 versus x/D is
presented in Fig. 9. It has been demonstrated by Bouchez [8] that
the ho distribution corresponds to a turbulent boundary layer with
a virtual origin situated about 35 cm upstream of the trip wire and
with heating being initiated at the virtual origin.

Concluding Remarks
The deflection of the jet by the crossflow was found to be small
for mass velocity ratios M > 8, and for these mass velocities the
Fig. 8 Influence of the temperature difference used In the definition of the impingement point heat transfer coefficient achieves a maximum
heat transfer coefficient
value at separation distances L/D of 5-6. Prior studies without
crossflow have found a maximum at comparable L/D values. For a
mass velocity ratio M = 4, the maximum is attained at smaller sep-
to the center of the jet, that is, the jet mixing region must engulf
aration distances owing to the substantial deflections experienced
the potential core. The plate-to-nozzle separation distance at
by the jet at larger L/D. The influence of the presence or the ab-
which the potential core is engulfed by the mixing process is not
sence of the splitter plate was confined to cases where L/D > 4,
known in the presence of a crossflow. Consequently, the L/D range
with the largest effects at the lowest mass velocity ratio.
in which the aforementioned entrainment process is a major con-
tributor to r > 1 cannot be definitely established. In this connec-
Acknowledgment
tion, it is interesting to note that at the smallest L/D of these tests,
L/D = 3, the value of r at the impingement point is only slightly in This research was supported, in part, by the Power Branch of
excess of unity (~1.04). This suggests that at this separation dis- the Office of Naval Research.
tance, there is a potential core that is not much affected by the en-
References
trainment.
1 Kercher, D. M, and Tabakoff, W., "Heat Transfer by a Square Array
At locations away from the impingement region, the velocities of Round Air Jets Impinging Perpendicular to a Flat Surface Including the
are quite low so that Tr c^ T«, ca T „ ° . Furthermore, since T«,° = Effect of Spent Air," Journal of Engineering for Power, Series A, Vol. 92,
Tj° in these experiments, it follows that Tr - Tj = Ty° - Tj = ufl 1970, pp. 73-82.
2cp, which gives r = 1 in accordance with equation (4). 2 Metzger, D. E., and Korstad, R. J., "Effects of Cross Flow on Impinge-
ment Heat Transfer," Journal of Engineering for Power, Series A, Vol. 94,
The significance of employing (Tw — Tr) as the thermal driving 1972, pp. 35-41.
force in the definition of h is illustrated by the results shown in 3 Bouchez, J. P., and Goldstein, R. J., "Impingement Cooling From a
Fig. 8. In this figure, there are two curves, one based on (Tw — Tr) Circular Jet in a Cross Flow," International Journal of Heat and Mass
and the other based on (Tw — 7 \ J . The differences between the Transfer, Vol. 18,1975, pp. 719-730.
two curves are confined primarily to the impingement region. T h e 4 Gardon, R., and Cobonpue, J., "Heat Transfer Between a Flat Plate
and Jets of Air Impinging on It," Proceedings, 2nd International Heat
curve based on (Tw — T„) actually displays a local minimum at Transfer Conference, 1962, pp. 454-460.
the impingement point. When (Tw — Tr) is used as the thermal 5 Gardon, R., and Akfirat, J. C, "The Role of Turbulence in Determin-
driving force, the local minimum is eradicated and, instead, the ing the Heat Transfer Characteristics of Impinging Jets," International
distribution curve takes on a maximum at the impingement point. Journal of Heat and Mass Transfer, Vol. 8,1965, pp. 1261-1272.
6 Gauntner, J. W., Livingood, J. N. B., and Hrycak, P., "Survey of Liter-
As M decreases, the difference between the results based on (Tw — ature on Flow Characteristics of a Single Turbulent Jet Impinging on a Flat
Tr) and (Tw - T„) diminishes since T r approaches T„. At M = 4, Plate," NASA TN D-5652,1970.
the difference is essentially undetectable. 7 Rouf, M. A., "Effect of Separation Distance on the Local Impingement
Heat Transfer Coefficient in the Presence of Cross Flow," MS thesis, De-
The importance of the recovery effects in the present experi- partment of Mechanical Engineering, University of Minnesota, Minneapo-
ments is due to the fact that the wall-to-fluid temperature differ- lis, Minn., 1975.
ences were small. If the temperature differences had been larger, 8 Bouchez, J. P., "Heat Transfer to an Impinging Circular Jet in a Cross
then the h and h' curves would have been nearly coincident. Flow," PhD thesis, Department of Mechanical Engineering, University of
Minnesota, Minneapolis, Minn., 1973.
The local heat transfer results presented in this paper were 9 Colin, P. E., and Olivari, D., "The Impingement of a Circular Jet Nor-
made dimensionless by employing the local coefficients ho for the mal to a Flat Surface With and Without Cross Flow," Von Karman Institute
crossflow without jet impingement. For various purposes, the actu- for Fluid Dynamics, Report No. 1,1969.

Journal of Heat Transfer NOVEMBER 1975 / 533

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 R. E. Mayle
Pratt and Whitney Aircraft,
East Hartford, Conn.
Multihole Cooling Film
F- J- Camarata Effectiweness and Heat Transfer
United Technologies Research Center,
East Hartford, Conn. Adiabatic film effectiveness and heat transfer measurements with injection of secondary
air through arrays of holes in a flat plate into a turbulent boundary layer are presented.
Measurements were taken both within and downstream of the multihole pattern for dif-
ferent hole pitch-to-diameter ratios and blowing rates. The hole configurations consid-
ered were regular patterns of staggered holes. The holes were angled 30 deg to the plate's
surface and 45 deg to the mainstream. A comparison between the measured adiabatic
film effectiveness and a model based on the superposition of point heat sinks is made.

Introduction model for the film effectiveness in the neighborhood of three-di-

mensional injection.
The continuing trend toward higher turbine inlet temperatures The only investigations of multihole-cooling surface heat loads
in modern gas turbine engine development has resulted in ever in- are those of LeBrocq, et al. [6] and Metzger, et al. [7]. LeBrocq pre-
creasing heat loads to turbine components. Consequently, in order sents a study on the influence of hole stagger, coolant-to-main-
to maintain acceptable life requirements, greater demands are con- stream density ratio and blowing rate on the film effectiveness, b u t
tinually being made on component cooling schemes. One type of fails to give sufficient information to form an empirical or semiem-
cooling scheme which has been introduced in response to these de- pirical correlation. On the other hand, Metzger's work, the first to
mands develops a cooling film on the surface by discharging cool- include measurements of both film effectiveness and heat transfer,
ant into the mainstream through an array of numerous small di- treats injection in a direction normal to the surface at relatively
ameter holes in the surface. This scheme, called multihole cooling, low coolant-to-mainstream mass flux ratios and provides, in this
relies on the convective cooling within the holes as well as the pro- blowing regime, sufficient information to determine the surface
tective film where, in general, each contribute an equal share to heat loads for one in-line and one staggered multihole configura-
the overall cooling effectiveness of the scheme. This, which is the tion. Consequently, at this time, only a meager amount of informa-
attractive feature of multihole cooling, also presents difficulties in tion is available and many areas remain to be investigated before
experimentally determining the surface heat load. surface heat loads for different injection geometries and blowing
In contrast to most film cooling schemes where the heat load is rates can be reasonably determined. Accordingly, the present work
found some distance downstream from the point of injection, the was undertaken to furnish information in one of these areas, name-
heat load for a multihole cooling scheme must be determined with- ly, the influence of hole and row spacing.
in the injection region. Most film cooling work to date has been In the following, film effectiveness and heat transfer measure-
done for injection through slots or rows of closely spaced holes, as ments are presented for air discharged into a turbulent boundary
1
recently summarized by Goldstein [l], with much of the effort layer through regular arrays of staggered holes. The holes, each 2.5
concentrating on determining the adiabatic wall temperature, mm in diameter, are inclined 30 deg to the surface with an angle of
commonly expressed as a film cooling effectiveness, 45 deg between the projection of the hole center line on the surface
and the mainstream as measured from the downstream direction.
V : The hole pattern, shown in Pig. 1 and described by the hole pitch-
Te~Tc to-diameter ratio, P/D, covers approximately the same streamwise
at different coolant-to-mainstream mass flux ratios, M = pcUcl distance on the surface for each of the three P/D ratios, 8, 10, and
p<JJm. The most relevant information for multihole cooling is that 14.
concerned with, discrete hole injection [2-5], which also include a
D e s c r i p t i o n of E x p e r i m e n t
The experiments were performed in a closed-circuit wind tunnel
with the airflow supplied by a centrifugal blower. Downstream of
1
Numbers in brackets designate References at end of paper. the blower, after a honeycomb section and screens, the flow is ac-
Contributed by the Heat Transfer Division and presented at THE celerated through a nozzle into a rectangular test section where, as
AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Revised manu-
script received by the Heat Transfer Division June 5, 1975. Paper No. 74- shown in Fig. 2, the multihole test plate forms the lower wall. A
HT-9. 6-mm divergence in height is provided over the length of the test

534 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 REMOVABLE PLATE ALTERNATE
— • 0.666 P — 0.866 P - i POSITION SCANNER
RAOIOMETEP TRAVERSE PLATE
TRAVERSE ASSEMBLY

| -<£• - i f"
p .-,. 45s /h
I ! v ® T •
I / < P/2
FLOW DIRECTION
0 ' ^ •-•••-<?-

o o
Fig. 1 Hole pattern BOUNDARY LAYER
BLEED SLOT C00 |

DIMENSIONS IN CENTIMETERS

Fig. 2 Test section

section to partially compensate for boundary layer growth and the
displacement effect associated with coolant injection. As a result,
only a 3 percent variation in the mainstream velocity is found for
the highest coolant flowrate. At the test plate leading edge, the
mainstream velocity and temperature are uniform and nominally a mosaic of film heaters, each in turn covered by a 0.75 mm thick
maintained at 23 m/s and 35°C, respectively. On the test plate, 5.1 copper plate. All bonds are formed by a high temperature rubber
cm downstream from the leading edge, a 0.75-mm trip wire is cement. The heaters are about 0.25 mm thick and consist of a Ni-
placed. This wire induces a turbulent boundary layer which at the chrome filament embedded between two Kapton strips. Each cop-
first row of holes may be represented by a one-fifth velocity profile per plate is drilled to conform to the hole geometry and contains a
having a momentum thickness of 0.58 mm. Beneath the plate, a chromel-alumel thermocouple. Gaps of about 0.5 mm are left be-
plenum fitted with inlet baffles and a porous foam pad is provided tween adjacent heaters and copper plates and are filled with a low
for the coolant. The coolant flow distribution within the hole pat- conductivity foam adhesive. Finally, the entire surface is coated
tern was checked by measuring the plenum pressure at a number with a black paint.
of holes and found uniform. The air temperature in the plenum is The streamwise length of a copper plate in the hole region is
maintained about 27°C below the mainstream temperature for the equal to the distance between rows. Each such streamwise strip is
film effectiveness tests, while that for the heat transfer tests is divided roughly into thirds in the spanwise direction with the
maintained equal to the mainstream temperature. heater and copper plate in the center used for obtaining data,
For the film effectiveness tests, the upper wall of the test section while the flanking heaters and copper plates act as guards. The
is fitted with a radiometer (Barnes Engineering RM-2A) and tra- spanwise length of the center heaters is equal to an integer number
verse assembly as shown in Fig. 2. The radiometer is aligned to of hole spacings within a row. Also, the portion of the plate up-
look vertically downward and measure the average plate surface stream and downstream of the hole region is heated and, as that
temperature within a circular area 0.5 mm in diameter. The tra- within the hole pattern, divided into thirds spanwise. During the
verse assembly may be located in either of two positions which test, the plate upper surface is maintained at a uniform tempera-
permits scanning both within and downstream of the hole pattern. ture of 55°C and the power to each heater recorded. Again, as with
In general, traverses of at least two hole pitches were made midway the effectiveness plates, thermocouples are located on the lower
between each row of holes, except between the first two rows where surface to allow an evaluation of the heat losses and a trip wire is
additional traverses were made, and at intervals of between 2.5 used to induce a turbulent boundary layer at the first row of holes.
and 5.0 cm downstream of the last row. Approximately eight read- Heat loss corrections were mainly calculated by considering a
ings were recorded per centimeter of spanwise travel. one-dimensional heat balance for the test plates. The result for the
Three film effectiveness test plates, one for each P/D hole geom- film effectiveness plates, which accounts for the heat loss through
etry, were cast using urethane foam. Each plate is 2.5 cm thick and the plate and increase in coolant exit temperature, indicates the
contains the hole pattern as cast. The upper surface is coated with difference between the film effectiveness, ?/, and that obtained
black paint having a known emissivity except for a white line using the actual surface and plenum temperatures, -q,n, is about
which is used to orient the radiometer traverses. A thermocouple 0.07(1 — rim). Corrections for the heat transfer measurements in-
buried in the plate surface permits checking the radiometer, while clude those for power dissipated by the heater leads and radiation
others located at coolant hole exits and on the lower surface allow as well as for back losses attributed to conduction through the
an evaluation of the heat losses. foam and heater and thermocouple leads. Representative values of
Three heat transfer test plates were also made. The bulk of the these corrections are 8, 9, and 7 percent of the measured power, re-
plate consists of 2.5 cm thick urethane foam identical to the film spectively. In order to estimate the error involved by using the
effectiveness test plates. The upper surface, however, is covered by one-dimensional approach, a conduction analysis for a vertical

-Nomenclature.
D = hole diameter X = streamwise coordinate distance mea- fir = spanwise averaged adiabatic film ef-
M = coolant-to-mainstream mass flux ratio sured in number of hole diameters fectiveness downstream of a single row of
P = hole pitch or spacing in a single row Z = spanwise coordinate distance mea- holes
Pe = Peclet number, U^D/at sured in number of hole diameters p = density
Rex = Reynolds number based on distance at = turbulent thermal diffusivity
from effective origin of the thermal rj = adiabatic film effectiveness
boundary layer rjm = film effectiveness using measured Subscripts
S = row spacing surface temperature and plenum air <= = mainstream state
St = Stanton number temperature aw = evaluated at surface of an adiabatic
T = absolute temperature T] = spanwise averaged adiabatic film effec- wall
U = velocity tiveness c = coolant or injected fluid state

Journal of Heat Transfer NOVEMBER 1975 / 535

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 hole in a plate having a constant surface temperature was exam- The heat transfer results are presented in Fig. 5 in which the
ined. The result indicates that the maximum heat loss with blow- Stanton numbers are plotted against a length Reynolds number.
ing is about 2.5 times the correction used for back loss through the The length Reynolds number is based on the distance from an ef-
foam alone and probably represents a maximum inaccuracy in the fective origin of the thermal boundary layer which in turn was es-
reported heat transfer coefficients of about 8 percent. Although tablished by matching the Stanton numbers measured on the
this uncertainty appears somewhat high, it is, at least for the scale plates between the trip and the first row of holes with the flat plate
and hole geometry of the present experiments, one which is both correlation of Reynolds, et al. [8]. Measurements of Stanton num-
difficult to reduce and accurately calculate. ber with no injection were compared to the correlation using this
method and found to be in agreement within a scatter of about 12
Results percent except for a slight deviation in the hole region of the P/D
Two typical spanwise distributions of film effectiveness are = 8 and 10 plates. This deviation is attributed to roughness intro-
shown in Fig. 3. The streamwise locations selected are between the duced by the heater plate junctions and tape patches which were
first and second row of holes and about halfway through the hole used to seal the holes during the test. The effective origins used in
pattern. Here the main feature to notice is that the spanwise varia- Fig. 5 are 6.3 cm upstream of the first row of holes for the plates
tion of T\ remains highly nonuniform within the pattern. This sit- having P/D = 8 and 10 and 5.5 cm for the P/D = 14 plate which
uation is found even downstream of more than twenty rows of gives estimated enthalpy-thickness Reynolds numbers at the first
holes and points out the integrity of the individual jets. row of holes of 360 and 327, respectively.
Measurements of JJ, such as shown in Fig. 3, were area averaged The heat transfer results indicate significant increases of the
over a multiple of the hole pitch to obtain a spanwise-averaged heat transfer coefficient are associated with injection. Within the
film effectiveness, TJ. The streamwise variation of r\ for each P/D hole pattern, the Stanton number increases as P/D decreases and
geometry at different mass flux ratios are presented in Fig. 4. In all as the mass flux ratio increases. For the larger hole density geome-
cases, the film effectiveness rises quite rapidly near the beginning tries and mass flux ratios the Stanton number remains essentially
of the hole pattern and the rate of increase gradually diminishes constant within the hole pattern and is approximately given by St
0 28
with distance downstream, in the case of M = 0.5, the rate becomes = 0.038(D/P)M - . This indicates that the flow near the surface is
essentially zero for all hole pitch-to-diameter ratios at a distance of streamwise periodic and dominated by the coolant jets. Thus, it
about 100 dia downstream. This result indicates a balance is nearly appears that as the hole spacing is decreased or the coolant flow
reached between the jet-mainstream mixing, which reduces the increased, a transition is gradually made in which the usual
cooling effect, and the periodic coolant injection which, of course,
is intended to increase cooling. At higher mass flux ratios, the film
effectiveness is seen to be still increasing at the last row of holes,
however, the rate of increase is reduced from that in the first half
of the pattern. Besides being a consequence of the film approach- I I
P.'D S
ing the coolant temperature, with the result that each successive 8 A
injection is less effective when based on the original coolant-main- 10 ©
stream temperature difference, the reduced rate of increase is also o
-^^ ^ 8 * ^\*
a consequence of jet coalescence. Evidence of this could be seen ') j
from the spanwise r\ distributions and was particularly noticeable
i
—^^*- ,o°° lo@o<«y ®
M - 0 5

for the P/D = 8 geometry over the last half of the pattern. This sit- )°°" °o
n 0
uation was also detected by flow visualization tests using powder
dusted on the surface and appears to arise from a net spanwise
drift of fluid near the surface caused by the spanwise component
of injection. The interesting point is that the film effectiveness for <p*» 0
the P/D = 8 and 10 geometries are nearly equal. This result is «SxA
'" , 0
probably caused by the spanwise drift, which is particularly iarge
for the P/D = 8 case, and subsequent coalescence of jets which, as ~1 30°" 0 ®
®
» 0°
with the in-line schemes of LeBrocq and Metzger, proves to be L,J
a®5 % Oo O O
detrimental. Downstream of the hole pattern, the film effective-
ness gradually decreases as a result of the continual mixing be- »
tween the film and mainstream. In Fig. 4, this decay is only shown $
for the P/D = 10 and 14 geometries, since that for P/D = 8 was af-
fected by mainstream fluid swept into the region from the tunnel J*"**f @ _
sidewalls as a result of the spanwise motion near the surface. 4 £^i
@ ,., 1.5

«d
oOO°c 3 o ° q
o o 0
Pt^—

o
& o o
8
® »
O

ol &
M- 2 0

|#®®®.
s
@#®

si04
e
CKTO~ 9
>?
0 o • ®
o % 0
s
o
tf
"%
&>
od

Fig. 3 Spanwise variation of film effectiveness, P/D = 10, M = 1.0 Fig. 4 Streamwise variation of average film effectiveness

536 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 streamwise growth of the thermal boundary layer yields to a peri- infinite row of holes, spaced a distance P apart is found to be
odic growth governed by the jets.
/l.
Downstream of the hole pattern, the Stanton number decreases Vr(X) M p e e l / 2 P e X K |o ( - P e x ) (2)
rapidly to the flat plate correlation. Similar behavior for uniform 4VP/
suction and blowing heat transfer experiments have also been ob- where Ko is the modified Bessel function of the second kind of
served by Kays [9]. Again, as in the presentation of the film effec- order zero. This result is, of course, identical to that obtained for
tiveness data, and for the same reason, results for the P/D = 8 ge- an infinite line sink moving in an infinite medium and having the
ometry in this region have been omitted. same strength per unit length as that for the row of point sinks in
equation (1). For typical values of Pe, P e X is generally greater
Comparison with Point Sink Model than 10, so equation (2) is well represented by
An adiabatic film effectiveness model for discrete hole injection
has been presented by Ramsey, et al. [4] and a method for super- v (3)
posing the film effectiveness for individual jets proposed by Gold- 4 \P/ X
stein, et al. [2], In the following, the suggested model will be com- The streamwise variation of the film effectiveness, i), now may
pared to the present film effectiveness measurements. be found by superposing the effectiveness of the individual rows in
The model is based on the solution of the energy equation for a the multihole pattern. Linear superposing of the average effective-
point sink (Or source) of heat moving at a constant velocity in an ness was tried at first without success. The streamwise variation
infinite conducting medium. As a result, the adiabatic film effec- was then found using the "additive film" method of superposition
tiveness is given by first proposed by Sellers [10]. Using Sellers' argument, namely that
the gas temperature used to determine the film effectiveness
1 MPe
exp •-Ve[VX2 + Z2-X] (1) downstream of an injection location should be the adiabatic film
8 VX2 + Z2 temperature due to all injections upstream of that location, it is
where the Peclet number, Pe, is defined as U^D/at and must be not difficult to show that the effectiveness at the streamwise loca-
determined experimentally. tion X is given by
The variation of average film effectiveness may be found by first
determining the spanwise averaged film effectiveness for all jets in q = 1 - exp j £ In ( l - r,r [ x - (A - 1) ^
a single spanwise row and then superposing the result for each row.
Assuming linear superposition, the averaged -q downstream of an where n is the number of rows upstream of the position X and S,
the row spacing. However, for hole pitch-to-diameter ratios of the
order of ten, 7jr is usually less than 0.1, so to a good approximation

L c PATTERN -
i-w{-Zi,[x-(H-l)?\

mm Substituting equation (3) into the foregoing, one obtains for uni-
form blowing

1 — exp
V7T
T\PJ
/D\
My 'Pi E ( X-k
This expression was evaluated midway between each row of
D)
-1/21
(4)

holes and compared to measurements for each mass flux and P/D
ratio.
Typical results of the comparison within the hole pattern are
shown in Fig. 6 for the geometry of P/D = 10. The values of the
Peclet numbers used in the comparison are 55, 22, 12 for mass flux
ratios of 0.5, 1.0, and 1.5, respectively. Note that equation (4) dis-
plays a proper streamwise behavior only for the case having a mass
flux ratio of 1.0 and is less satisfactory for the other blowing rates.
Since similar discrepancies are also found for the other cases and
the Peclet numbers which give best agreement are different than
those quoted above, it is apparent that further analytical work in
this area is necessary.

0 05 0 10

Rex X 1 0 ~ D
Conclusions
Within a pattern of holes discharging coolant into a turbulent
Fig. 5 Streamwise variation of Stanion number
boundary layer, the film effectiveness was found to be highly non-
uniform and reflected the integrity of individual jets. This integri-
ty and the possibility of jet coalescence or interaction, which may
adversely affect the development of an effective cooling film, will
have to be considered in any film effectiveness analysis of a multi-
hole cooling scheme. In this respect, the model using a superposi-
tion of point sinks does not appear too promising.
The heat transfer coefficient within the hole pattern was found
to increase significantly as either the mass flux ratio or hole densi-
ty was increased. Since increases as high as 2.5 were found in the
average heat transfer coefficient, it is apparent that greater in-
creases in local values are to be found. Consequently, the variation
0 40 80 0 *0 30 0 40 B0 ! 30 160 200
of the local heat transfer coefficient within the hole pattern must
be considered together with the film effectiveness distribution be-
Fig. 6 Comparison of measured and calculated film effectiveness fore the multihole-cooling surface heat load may be determined.

Journal of Heat Transfer NOVEMBER 1975 / 537

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Acknowledgment 5 Eriksen, V. L., "Film Cooling Effectiveness and Heat Transfer With
Injection Through Holes," NASA CR-72991, Aug. 1971.
T h i s work was c o n d u c t e d u n d e r C o n t r a c t F 3 3 6 1 5 - 7 2 - C - 1 9 0 2
6 LeBrocq, P. V., Launder, B. E„ and Priddin, C. H., "Discrete Hole
g r a n t e d b y t h e Air F o r c e A e r o P r o p u l s i o n L a b o r a t o r y a n d is r e - Injection as a Means of Transpiration Cooling—An Experimental Study,"
p o r t e d in A F A P L - T R - 7 3 - 3 0 . Imperial College of Science and Technology, Report EHT/TN/A/26, June
1971.
7 Metzger, D. E., Takeuch, D. I., and Kuenstler, P. A., "Effectiveness
References and Heat Transfer With Full-Coverage Film Cooling," Journal of Engineer-
1 Goldstein, R. J., "Film Cooling," Advances in Heat Transfer, Aca- ing for Power, TRANS. ASME, Series A, Vol. 95, Apr. 1973.
demic Press, Vol. 7, 1971, pp. 321-379. 8 Reynolds, W. C , Kays, W. M., and Kline, S. J., "Heat Transfer in a
2 Goldstein, R. J., Eckert, E. R. G., Eriksen, V. L., and Ramsey, J. W., Turbulent Incompressible Boundary Layer, I-Constant Wall Temperature,"
"Film Cooling Following Injection Through Inclined Circular Tubes," NASA Memo 12-1-58W, 1975.
NASA CR-72612, Nov. 1969. 9 Kays, W. M., "Heat Transfer to the Transpired Turbulent Boundary
3 Ramsey, J. W., and Goldstein, R. J., "Interaction of a Heated Jet Layer," NASA Report NGL05-020-134, 1971, (Thermal Division Depart-
With a Deflecting Stream," NASA CR-72613, Apr. 1970. ment of Mechanical Engineering Stanford University Report HMT-14, June
4 Ramsey, J. W., Goldstein, R. J„ and Eckert, E. R. G., "A Model for 1971).
Analysis of the Temperature Distribution With Injection of a Heated Jet 10 Sellers, J. P., "Gaseous Film Cooling With Multiple Injection Sta-
Into an Isothermal Flow," Heat Transfer 1970, Elseuiar, Amsterdam, 1970. tions," AIAA Journal, Vol. 1, Sept. 1963, pp. 2154-2156.

538 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 M. F. Blair
Research Engineer,
United Technologies Research Center,
New Techniques for Measuring
East Hartford, Conn.
Mem. ASF.lii
Film Cooling Effectiweness
RB D. Lander Techniques to measure local film effectiveness distributions on large-scale models of tur-
Assistant Project Engineer, bine blade and vane sections were developed for use in subsonic, room-temperature
Pratt & Whitney Aircraft Division,
wind-tunnel environments. Highly detailed film cooling effectiveness distributions were
East Hartford, Conn.
obtained on near-adiabatic, cast rigid foam test models using infrared radiometer scan-
ning techniques. Sample results from both flat-plate film cooling studies and airfoil
leading-edge film cooling studies are presented.

Introduction because the castings are dimensionally stable and a thin skin re-
sults naturally on the surface of the foam. The casting skin is as
Turbine airfoil metal temperatures in many current high tem- smooth as the mold surface in which the model was made. The
perature gas turbines are maintained at acceptable levels by means foam models can be subjected to steady temperatures of 60° C
of film cooling. In order to predict the exterior wall temperature without distortion. The low thermal conductivity of the foam mod-
and convective heat flux distribution on turbine components els resulted in reduced conduction errors from the test surface to
cooled in this manner, both surface adiabatic wall effectiveness the coolant plenum on the model's back surface. The backloss er-
distributions and surface heat transfer coefficient distributions are rors for the present tests were significantly reduced compared to
required. The present paper deals with low temperature IR tech- those which occurred for previous tests using Micarta or fiberglass
niques developed to obtain detailed effectiveness distributions models. An analysis presented in the Appendix delineates the in-
near coolant injection sites where nonuniform cooling is commonly place backloss calibration procedure.
observed. The techniques described in this paper can be employed
for model surface temperatures as low as 0°C while IR sensitive Traversing-Spot-Radiometer Technique
film, which has been used by other workers for measuring temper- A spot radiometer was used to measure surface temperatures for
ature distributions in hot rig tests, requires a source temperature a film cooling investigation on a flat plate model with an array of
of approximately 480° C for practical exposure times. Thus, the coolant injection holes. The tests were conducted in a low-speed,
techniques presented herein are practical for basic engineering re- continuous recirculating flow wind tunnel in which the flow passes
search studies carried out in test facilities which operate at near through a rectangular test section, 20.3 cm high by 50.8 cm wide. A
room temperature. detailed description of the tunnel in the region of the test plate is
shown in Fig. 1. The test plate was 122 cm long and formed the
Test Models lower wall of the test section with the leading edge of the plate at
Near-adiabatic models, for the measurement of film effective- the duct entrance. A coolant plenum designed to produce uniform
ness distributions, were fabricated by casting the desired configu- flow from the coolant holes was mounted beneath the test plate.
rations from rigid urethane foam (Witco Chemical—ISOFOAM®) Spanwise variations of the coolant flow from hole to hole were im-
in accordance with the manufacturer's instructions. Urethane perceptible within the central region of the plate where effective-
foam was employed because the thermal conductivity of the result- ness measurements were obtained. The upstream boundary layer
ing bodies is very low (8-12 X 10~ 5 gm-cal/cm-s K) and because it was removed through a 0.6 cm wide slot, inclined at 30 deg to the
can be cast in the shapes required for injection configurations. Ex- test surface, between the test plate sharp leading edge and the
cellent aerodynamic bodies can be fabricated with this rigid foam lower wall of the nozzle. Velocity surveys at various distances above
the test plate indicated only slight nonuniformities in the duct
flow outside the wall boundary layers. A temperature difference
between main and coolant flows of approximately 28°C was ob-
Contributed by the Heat Transfer Divsion and presented at the — THE tained by refrigerating the coolant approximately 22°C below room
AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Revised manu- temperature and by operating the tunnel at approximately 6°C
script received by the Heat Transfer Division April 7, 1975. Paper No. 74-
HT-8. above room temperature. Film effectiveness data were obtained

Journal of Heat Transfer Copyright © 1975 by ASME NOVEMBER 1975 / 539

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 for a freestream velocity of 22.8 m/s and over a coolant Reynolds
number range (based on the coolant hole diameter of 1.27 cm) from
2 x 103 through 25 x 103, a span which encompasses typical cur-
rent engine conditions. The upper surface of the test section was
formed in part by a traversing mechanism which supports the
radiometer (Barnes Engineering Model RM-2A). This radiometer
measured the test plate surface temperature within a 0.05 cm dia
spot. The combination of translational and rotational motions of
the traverse and the removable alternate plate shown in Fig. 1,
permitted the radiometer to scan a region approximately 102
cm long in the streamwise direction. The upper surface of the
plate was coated with a high emissivity flat-black paint (3M
C-101 Black Velvet®) except for a white reference line which was
used to orient the radiometer traverses. The paint emissivity was
Fig. 1 Sketch ol test plate installation for effectiveness measurements
checked against a laboratory black body standard and was found
to be independent of the number of coats applied. The measured
emissivity of the paint ranged from « = 0.956-0.979. The radiome-
ter was calibrated in place periodically during each test to obtain
approximately ±1.0°C accuracy. The data from the radiometer spreading of the cooling effect, is in qualitative agreement with the
and a location indicating potentiometer were fed to an analog-to- data of references [1, 2] which were obtained downstream of a sin-
digital converter and punched on paper tape for processing on a gle slanted hole on a flat plate. A quantitative comparison of the
digital computer. The surface temperature, coolant temperature, present results with those from the single hole of references [1,2]
and external flow temperature data were processed to obtain the is difficult because the lateral separation of the holes for the
film effectiveness at each surface temperature measurement loca- present study was so small the hole cooling patterns overlap and
tion. The results from each spanwise traverse were linearly inter- influence each other. The coolant hole pattern of the model of Fig.
polated to construct a rectilinear array with 1600 points of effec- 2, with staggered rows of two-dimensionally spaced holes, is near
tiveness. This array in turn was used to construct an isoeffective- the minimum practical spacing for a drilled hole film cooling
ness contour map on a memory oscilloscope. scheme. The lateral variations of effectiveness for Fig. 2 indicate
For the Flat Plate Models, the uncertainties in the measured that even for this spacing the cooling pattern was not a uniform
free stream temperatures were approximately ±0.25°C while the two-dimensional film. Additional effectiveness results obtained for
uncertainties in the measured coolant plenum temperatures were these flat plate models at various coolant injection rates and with
approximately ±1.0°C. As previously discussed, the uncertainties various coolant hole patterns can be found in reference [3].
in the temperatures indicated by the spot radiometer were ap-
proximately ±1.0°C. Using the procedure described in reference Scanning Radiometer Technique
[4],1 with the foregoing uncertainties in the measured tempera- A second measurement technique was developed to obtain de-
tures, the effectiveness data obtained for the Flat Plate Models tailed film effectiveness data on large-scale models of turbine air-
were estimated to be accurate within ±0.05. foil leading-edge regions using a scanning infrared camera. Data
A sample "map" of effectiveness contours is presented in Fig. 2. for this investigation were obtained in an open circuit, low-speed
For the model sketched in Fig. 2, the coolant hole diameter was wind tunnel which draws from the room air into a low turbulence
1.27 cm, the lateral hole spacing/hole diameter was 2.0, and the convergence section, through a 30 X 50 cm test section and returns
holes were inclined at 30 deg to the surface. These results are pre- the air to the room. The test section velocity can be varied from 0
sented in terms of the standard film cooling parameter, -qmm = (Te to 36 m/s and the test section turbulence intensity increased from
— Taw)/(Te — Tc). Although coolant injection was present across a minimum value of 0.5-11.0 percent by installing various turbu-
the entire 50.8 cm width of the test surface, the isoeffectiveness lence generating grids. A sketch of the tunnel in the region of the
contours were only constructed for a representative 10 cm wide test model is presented in Fig. 3. Effectiveness distributions were
streamwise strip near the tunnel center line within which effects of again obtained on a nearly-adiabatic model fabricated of cast rigid
the tunnel sidewalk were considered negligible. urethane foam. The rear portion of the test model was designed to
The principal feature of Fig. 2 is the extreme spanwise variation eliminate unsteady flow in the leading edge region resulting from a
of the measured effectiveness indicating that very little lateral rear body wake and to prevent separation within the test region
spreading of the coolant fluid occurred. This result, little lateral which extends from the stagnation line to the point of maximum
thickness on the body. For these tests, the film cooling was simu-
lated by air which was heated to a temperature approximately
1
Numbers in brackets designate References at end of paper. 28°C hotter than the mainstream air.

—Nomenclature.

gm-cal/gm K St = Stanton number, h/(peUeCp), dimen- JJO = experimentally measured effective-

h = convective heat transfer coefficient, sionless ness (equation (A-6)), dimensionless
W/cm 2 -K t = wall thickness, cm TINF = no-flow effectiveness parameter,
hi, h„ = interior and exterior wall convec- T = temperature, K (equation (A-4)), dimensionless
tive heat transfer coefficent, respective- TCp, TCe = temperature of coolant at ple- v = kinematic viscosity, cm 2/s
ly, W/cm 2 -K num and hole exit, respectively, K p = density, gm/cm 3
k = thermal conductivity, gm-cal/cm-s K Tui, TWe = temperature of wall on inside
qi, qe = interior and exterior heat flux, and outside surface, respectively, K Subscripts
W/cm 2 x = distance from leading edge, cm aw = adiabatic wall
Re = Reynolds number, Uexh, dimension- i?fiim = effectiveness parameter, (Te — Taw)/ c = coolant
less (Te — T c ), dimensionless e = free-stream value

540 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The surface temperature of the curved surface models was de-
termined using a Bofors IR scanning camera. This camera had a COOLANT SUi'PL Y TUBE

depth of field of approximately 5 cm when focused on the test sec-

tion 35 cm from the camera. The IR camera viewed the test surface
through a sheet of Vinylidene Chloride-Vinyl Chloride co-polymer BOFORS SCANNING
INfRARED CAMERA
(Saran Wrap food wrap) which served as an infrared window in the
tunnel sidewall. The camera was calibrated with the Saran Wrap
in place to account for any "tailoring" of the detector sensitivity
resulting from slight infrared absorption. The test surface of the
model was coated with the same high emissivity flat-black paint as !
0.304",

previously described except for a 0.6 cm wide strip of aluminized j AIHFOIL LEADING EDGE
COOLING CONFIGURATION

mylar tape which was used to orient the infrared camera displays.
The thermocouple measurements provided accurate temperature
measurements at sixteen locations and the IR scanning camera
Fig. 3 Sketch of leading edge model Installation
provided isotherm contours at selected temperature levels which
would have been possible only with a multitude of thermocouples.
For each coolant flow rate approximately eight separate isother-
mal levels were "mapped." Three sample isotherm displays are
presented in Fig. 4. For these test conditions, the uncertainties in Bofors Scanning camera were approximately ±1.0°C. Using the
the surface temperature measurements were approximately procedure described in reference [4], with the foregoing uncertain-
±1.0°C. The individual isothermal contours were assembled into a ties in the measured temperatures, the effectiveness data obtained
composite temperature distribution from which isoeffectiveness
contour plots were constructed. A sample plot of local effectiveness
contours is presented in Fig. 5. For this leading edge model the
rows of hole were located at 17 and 47 deg from the nose stagna-
tion line and the holes were inclined at 30 deg to the surface. The
model leading edge radius was 3.2 em and the diameter of the cool-
ant holes was 0.674 em. The data are presented in a plane view of
the curved leading edge model with the coolant holes shown cross
hatched. For the Leading Edge Film Cooling Models, the uncer-
tainties in the measured free stream temperatures were approxi-
mately ±0.25°C while the uncertainties in the measured coolant
plenum temperatures were approximately ±1.0°C. As previously
discussed, the uncertainties in the temperatures indicated by the

32

5
5
5
28

24

I
E
3 3
u
3
I
w
<:J
0
w
7
""m'"
3~
<:J
Z
0
«
w
--'
:2'
0
0:
LL
w
u 12
z
«
I-
(J)

0
8
TRIP WIRE

--------- --(---
4

00 4 8 12 0
Tdi,ptay .34.4 C 11)- 0.38)
SPANWISE LOCATION -em
Fig. 4 Typical Bofors Infrared camera displays lor airfoil leading edge
Fig. 2 Isoeffectlveness distribution for flat plate film cooling cooling conflgurallon tests

Journal of Heat Transfer NOVEMBER 1975 / 541

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 where Te = free-stream adiabatic wall recovery temperature, Taw
= adiabatic wall temperature in the presence of film cooling, and
TCe * coolant temperature at the exit from the coolant hole. A sim-
ple one-dimensional backloss correction will be formulated in the
following paragraphs. Calculations indicate that the local conduc-
tion effects resulting from the flow through the coolant holes are
concentrated within approximately one coolant hole diameter of
an injection site and the results developed here should be applica-
ble to all other regions of the test surface. This correction was ap-
plied to the data presented for the flat plate model only.
For the film cooled plate with finite conductivity, shown in the
following,

0 1 2 3 4 COOLANT FILM
SUR F ACE D I S T A N C E / L E A D I N G EDGE R A D I U S , S/R
Fig. 5 Isoeffectlveness distribution for airfoil leading edge cooling config-
uration

for the Leading Edge Models were estimated to be accurate within

±0.05.
The principal features of these results are the large streamwise
and spanwise variation in film cooling effectiveness. Despite the
dissimilarities between the two flow situations the present results,
for a leading edge injection scheme, are qualitatively similar to the
results presented in reference [2] for a laterally slanted single hole
in a flat plate. The contours of constant effectiveness presented for
the single hole inclined at 35 deg to the flat plate surface and 90
deg to the flow direction are similar in shape to the present results. a one-dimensional heat flux analysis of the plate yields

Concluding Remarks qa = he (Taw - TWe) = <?cond = - (Tw; - Twi) = Qi = h{ (Tm - TCp)

The successful employment of cast urethane foam fabrication
techniques and infrared scanning has demonstrated the potential (A-2)
of these techniques for the acquisition of highly detailed film effec- where q is the local heat flux, h is the local external or internal
tiveness data in low temperature facilities. heat transfer coefficient, k is the material thermal conductivity,
and T is the local temperature. The equations of (A-2) can be com-
Acknowledgment bined to obtain
The authors are grateful to Messrs. F. J. Camarata and B. V. T —T
^ / i ^UK
Te-Tm
Johnson of the United Technologies Research Center and Dr. R. E. V _ \(A.3)
Te - TCe -TCe \l/hi + t/k/
Mayle of Pratt & Whitney for their suggestions during the devel-
opment of these techniques. Although the temperatures are measured, the overall heat transfer
parameter, (l/he)/(l/hi + t/k), is unknown and must be deter-
References mined before the adiabatic film effectiveness, rjf, can be evaluated.
1 Goldstein, R. J., Eckert, E. R. G., and Ramsey, J. W., "Film Cooling Although the thermal conductivity of the material can be mea-
With Injection Through a Circular Hole," NASA CR-54604, May 1968. sured, measurement of the internal heat transfer coefficient would
2 Goldstein, R. J., Eckert, E. R. G., Eriksen, V. L., and Ramsey, J. W., be difficult. Consequently, the heat transfer parameter was ob-
"Film Cooling Following Injection Through Circular Tubes," NASA CR-
72612, Nov. 1969. tained with an in-place calibration by passing flow through the
3 Muska, J. F., Fish, R. W., and Suo, M., "The Additive Nature of Film coolant plenum behind the flat plate configuration with the cool-
Cooling Prom Rows of Holes," ASME Paper No. 75-WA/GT-17. ant holes plugged and measuring the test plate, coolant, and exter-
4 Kline, S. J., and McClintock, F. A., "Describing Uncertainties in Sin- nal flow temperatures. These were combined to obtain a no-flow
gle Sample Experiments," Mechanical Engineering, Vol. 75, No. 1, 1953, pp. effectiveness parameter
3-8.
T —T
APPENDIX I n o flow - VNF — '
T - T
(A-4)

Film Cooling Effectiveness Conduction Corrections where TWtNF = exterior wall temperature with, no flow passing
through the coolant holes. The value of TJJVF was experimentally
The following analysis leads to an equation for determining the
determined to vary from 0.04 to 0.06 depending on the plenum
adiabatic wall film effectiveness from data obtained with the low-
flow rate—a constant value of 0.05 was used as a reasonable ap-
thermal-conductivity test plates. A direct measurement of the adi-
proximation for all coolant flow rates.
abatic wall recovery temperature cannot be achieved with finite
conductivity material. However, the use of low-thermal-conductiv- Using the one-dimensional heat flux balance shown previously,
ity material, such as the urethane foam, minimizes the heat con- the following equation was obtained to relate the measured tem-
duction and small heat flux corrections for backlosses can be incor- perature to the no-flow effectiveness parameter and heat transfer
porated into the data reduction procedure. coefficient parameter of equation (A-3)
For an adiabatic wall the standard film cooling effectiveness pa- Te UK VNF
(A-5)
rameter is defined TweNF ~ TCp VHi + t/k 1 - 7JNF
If we assume that he^p = he and IIMF = h;, the adiabatic film ef-
(A-l)
v- T V 7 ? ! fectiveness is obtained in the following form:

542 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 coolant holes indicated that the temperature rise along the hole
was negligible and (Te - TCp)/(Te - TCe) = 1. Thus, the adiabatic
wall film effectiveness is related to the experimentally measured
where 7/0 = (Te -TWe)/(Te - TCp) is the experimentally measured effectiveness with the following relationship:
effectiveness and the temperature ratio, (Te — TCp)/(Te — TCe), is a
measure of the heat absorbed by the coolant air in the injection
hole. A measurement of the coolant temperature at the exit of the

Journal of Heat Transfer NOVEMBER 1975 / 543

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A. J. Suo-Anttila
Graduate Research Asst. The Effect of a Stabilizing
I. Catton
Assoc. Professor.
Temperature Gradient on Heat
Assoc. Mem. ASME

School of Engineering and Applied Science,

Transfer From a Molten Fuel
University of California, Los Angeles, Calif.

Layer With Volumetric Heating

The Landau method is used to determine the heat transfer to the upper and lower sur-
faces of a volumetrically heated layer of molten fuel for various fixed values of the upper
and lower surface temperatures. To use the Landau method, the Stuart shape assump-
tion is imposed and solutions to the linearized stability problem are used. The results
are generalized in terms of two Nusselt numbers (upper and lower surface heat transfer)
which are presented as a function of the Rayleigh number based on the internal heat
generation rate for various values of an external Rayleigh number based on the surface
to surface temperature difference. Results are also presented for a molten UO2 layer for
various typical values of decay heat. Comparison is made with experiment, where possi-
ble, and found to be good.

Introduction Earlier work in this area has been restricted to equal upper and
lower surface temperatures. Kulacki and Goldstein [2] made ex-
In nuclear reactor design and safety, postulated reactor core perimental measurements of the downward heat flux and maxi-
meltdowns have led to the consideration of heat transfer from mol- mum temperature in the layer. Jahn and Reineke [4] made similar
ten fuel layers with internal heat generation. In one failure mode, measurements and numerical calculations as well. They report
the containment of the molten layer will depend upon the down- slightly different results from those of Kulacki and Goldstein. Cat-
ward heat flux. Earlier work [l] 1 in this area has been restricted to ton and Suo-Anttila [3] made numerical computations of the heat
a molten layer which is confined above and below by a crust of transfer under these same circumstances using the Landau method
solid fuel. However, the crust of solid fuel may not form if a large and reported good agreement with the experimental results of Ku-
upward thermal resistance exists above the molten layer. In this lacki and Goldstein.
situation, the molten fuel layer is confined below by solid fuel and Many investigators have calculated the heat transfer-Rayleigh
above by some other material whose temperature is greater than number relationship for the Benard problem. Malkus and Veronis
the melting point of the fuel. The situation described is stabilizing [5] used the Landau power-integral method (see Landau and Lifs-
and decreases the amount of convection in the layer. An example chitz [6]) to study the case without internal heat generation. The
of this could be a molten layer of UO2 covered by a molten layer of Landau method has also been used by Edwards and Catton [7] to
iron. The downward heat flux can be considerably enhanced when predict heat transfer in confined regions, by Catton [8] to predict
the fluid layer is stabilized. In this report an estimate is made of heat transfer across an infinite horizontal layer under the influence
the heat flux when a molten layer of fluid with internal heat gener- of rotation, by Catton and Ayyaswamy [9] for momentum transfer
ation is bounded by rigid surfaces with the upper surface tempera- between concentric cylinders with the inner cylinder rotating, and
ture such that a stabilizing temperature gradient is imposed. by Sharman, Catton, and Ayyaswamy [10] for heat transfer across
the gap between concentric cylinders with the inner cylinder rotat-
ing. In all cases, excellent agreement between theory and experi-
ment was obtained up to 10,000 times the critical value of the pa-
1
Numbers in brackets designate References at end of paper. rameter governing the onset of motion. For example, good results
Contributed by Heat Transfer Division and presented at the Winter An- were obtained for the Benard problem up to Rayleigh numbers of
nual Meeting, New York, N.Y., November 17-22, 1974, of THE AMERI- 107. In this work the Landau Power Integral method is used to
CAN SOCIETY OF MECHANICAL ENGINEERS. Revised manuscript re- study the case of internal heat generation with an imposed exter-
ceived by the Heat Transfer Division January 27, 1975. Paper No. 74-WA/
HT-45. nal stabilizing temperature gradient. The heat transfer-Rayleigh

544 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 where g is the gravitational body force, a is the fluid coefficient of
///////////-, thermal expansion, and v is the fluid thermal diffusivity.
Equations (3), (4), and (5) together with the boundary condi-
tions given by equation (2) are sufficient to solve the stability
problem. The onset of motion, the lowest mode, has been obtained
by Kulacki and Goldstein [2]. In this work the higher modes are
needed, thus requiring resolving the problem. The equations are
linearized by neglecting the inertial terms in the momentum equa-
tion and the second order convective terms in the energy equation.
Fig. 1 Configuration and coordinate system used In application of the An approximate solution (variational approach) to a similar prob-
power Integral method
lem has been obtained by Chandrasekhar [11]. His approach yields
a relationship between the characteristic Rayleigh number and the
wavelength of the disturbance. These results will be combined
number relationship obtained is compared with Kulacki and Gold-
with the Landau method to yield the heat transfer.
stein's [2] experiments.

Governing Equations The Onset of Motion

Initially a quasi-incompressible (Boussinesq) fluid element with The stability problem reduces to solving the following set of
internal heat generation (see Fig. 1) is heated from above to estab- equations (see Chandrasekhar [11, p. 90]),
lish a temperature gradient parallel to the gravitational body force
(along the z axis). The initial velocity and temperature distribu- (D2 - a2) 8 = -R,W{z - $ + REW (8)
tions are given by
{D2 2
2 2
)W = a2 8 (9)
where 6 and W, the temperature and vertical velocity component,
y (1) have been analyzed into normal modes to remove the x and y de-
dz kt ' 2> L
pendence, a is the wave number and
where Q is the internal heat generation rate per unit volume and kf
d_
is the fluid thermal conductivity. The fluid is confined between D = (10)
rigid, fixed-temperature walls. The corresponding boundary condi- dz
tions are The associated boundary conditions are
u = 0, T = c o n s t a n t (2) W = DW =8 = 0 (ID
on both surfaces. on the bounding horizontal surfaces.
The instability is assumed to appear via a marginal stationary
Briefly, the stability problem is solved by letting
state; hence the governing dimensionless disturbance equations
are F = '{D2 - a2)2W = a28 (12)

V«u 0 (3) and rewriting the differential equations in the forms

u (D2 - a2)2W = F (13)
Pr-\vm)xa = -V(/> xi —) + 0k - Wx(Vxu) (4)
{D2 - a2)F = -Rfl2W{z - h + REWa (14)
u» V<9 + u - k G - R / z - | ) + RE) V28 (5)
with boundary conditions on F being the same as 6 at the rigid hor-
where u, 0, and p are the velocity, temperature, and pressure mea- izontal surfaces.
sured in units of K/L, nK/agL3, and PVK/L2, respectively. All dimen- Since F will be even or odd, it is expanded in a sine series of the
sions are scaled by the layer thickness L. The Prandtl number, Pr form
= V/K appears as a multiplicative factor in the inertial terms and
the parameters R/ and R E are given by F = S Fm = TJ Am svnfjrmz) (15)

=
R/ ,7.„u— i n t e r n a l modified R a y l e i g h n u m b e r (6)
VKkf The corresponding Wm are solutions to equation (13) with the
boundary conditions given by equation (11). Solutions to equation
external Rayleigh number (7) (13) are

•Nomenclature.
a = wave number - TOI(VK) 0 = dimensionless temperature
D = derivative with respect to z R/ = modified internal Rayleigh number, K = thermal diffusivity k/(pc)
g = gravitational acceleration agQLs/(VKk) v = kinematic viscosity
k = thermal conductivity T = temperature p = density
k = vertical unit vector u = velocity
L = depth of fluid layer v = velocity Subscripts
Nu = Nusselt number = heat flux at w = vertical component of velocity 1 = downward or bottom of layer
boundary /conductive heat flux at bound- W = dimensionless vertical component of 2 = upward or top of layer
ary velocity which has been analyzed into 0 = initial
Pr = Prandtl number = VIK normal modes / = fluid
q = surface heat flux z = vertical coordinate m = mode or term in series expansion
Q = volumetric heat generation rate a = isobaric coefficient of thermal expan- n = term in series expansion
RE = external Rayleigh number, agL3(T2 sion z = vertical component

Journal of Heat Transfer NOVEMBER 1975 / 545

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A, Table 1 Critical Rayleigh numbers, wave numbers, and transport coeffi-
w = Z(W JT 2 2 {B, < m ) s i n h a z + A2imh cients
+ Cpf
+ B2{m)z s i n h az + s i n muz) (16)
where 0 1 3.73xl04 4.00 .1061 io4 1 1.07xl05 5.03 .0986

im) m 2 7.35xl03 8.36 .1018 2 8.87xl05 8.24 .0723

Bi = mn(a + ( - l ) ) / A
3 4.45xl06 11.8 .0491 3 4.72xl06 11.9 .0495
Aiim) _ _m7r(sinh2a + (-l)ma sinh a ) / A
4 1.59xl07 15.4 .0314 4 1.68xl07 15.7 .0330

B 2 < m ) = w7r{(sinha c o s h a - a) 5 4.37xl07 18.9 .0209 5 4.67xl07 19.4 .0229

m 1.04xlO 8
25.5 .0127 6 1.12xl08 25.6 .0149
+ (-l) (a cosha - sinha)}/A 6

2 2
A = sinh a - a (17) IO 6
1 3.37X10 6
9.94 .0636 io 8
1 2.76x10 23.7 .0741

2 1.06xl07 9.26 .0461 2 9.67xl08 24.1 .0429

Substituting for F and W in equation (14), multiplying by sin
3 i.83xl07 12.2 .0272 3 1.37xl09 20.9 .0208
(nirz) and integrating from z = 0 to one yields
4 3.36xl07 15.5 .0199 4 1.78xl09 18.5 .0124

(wzV + a 2 ) R 5 6.53xl07 18.9 .0156 5 2.20xl09 19.7 .00828

£ i m { """" 1? .-X ^ -) sm{tvnz)dz
6 1.26xl08 22.2 .0122 6 2.60xl09 23.4 .00595

+ R
el Wm sm(pnz)dz} = 0 (18)

which is a set of linear homogeneous equations for the coefficients Substituting for the horizontally averaged temperature gradient in
Am. The determinant of equation (18) must vanish if the Am are to equation (22) yields the convective contribution to the heat trans-
be nontrivial. Thus fer
(m2ir2 + a2)3 R w
(KT'V2T')
2a 2 j / m(z ~ 9) sin(nirz)dz (VtT')=^N{l +
Q/k(v,T' (z ~ - ) >

<»-r,x-^/*>
+ ^EL Wm sm(mrz)dz = 0 (19)
} (24)
is a characteristic equation for the R/. Q/k{v,T> (z - - ) >
Equation (19) was solved for a series of values of RE and the re-
where
sults are presented in Table 1. Equation (19) yields both positive
and negative values of R/. The negative values are solutions to the
problem when a heat sink is present and are not reported here.
(«,r'Xu.r'(| - f)>
AT = (25)
The smallest R/, resulting from a solution to equation (19), is iden- (v,T>2) - (v,T>)2
tical to that reported by Kulacki and Goldstein [2], The larger R/,
higher modes, have not heretofore been calculated. By defining the Nusselt number as the ratio of the heat flux at the
boundary to the heat flux at the boundary due to conduction alone
Heat Transfer
The method by which power integrals could be used for the hori- ,dT
zontal layer problem was first deduced by Malkus and Veronis [5] -& —
Nu, ?2 dz 1=1 (26)
and their approach will be closely followed. The method involves
<7cond QJk + k{-dT/dz)
averaging of various equations with the assumption that the am-
plitude but not the shape of the disturbances changes with increas-
ing Rayleigh number. This is sometimes known as the Stuart a relationship for the convective contribution given by equation
shape assumption [12]. In the following development, either the (24) is found to be
energy or momentum equation could be used. The energy equation
is chosen for convenience. Nu,2 - 1 = <^> (27)
* {KQL/2k) + K{-dT/dz)
The temperature is defined as a perturbation temperature T'
plus a horizontally averaged temperature T, Scaling velocity and temperature with the same factors used in the
stability problem allows equation (24) to be written as
T = T> + T (20)
with T" = 0. Using this expression for temperature, the energy ^ = *{1 + ^m±JA}^f- (28)
equation (5), in dimensional form, becomes

KV2T 9 ,ar. + v VT' + v • V T (21) If it is assumed that the shape of the disturbance is adequately
-"Vz^ described in terms of normal modes and that only the amplitude
Subtracting the horizontally averaged energy equation from equa- changes with increasing Rayleigh number [12], then multiplying
tion (21), multiplying by T' and volume averaging gives equation (8) by 8 and integrating over the entire layer yields

- Wfl2 - a'2) 9) + (W8)RE (29)

(KT'V2T') + <--gjw,T'> 0 (22) Rr =
{wB{z - ±)>
where the bar denotes a horizontal or area average and the < > de-
notes an average along a vertical line between the bounding sur- This allows (vzT') to be related to R/„ and R/. Using the relation-
faces. The horizontal averaged energy equation is integrated and ship given by equation (29) and the definitions of R/ and R^ in
the constant of integration eliminated by volume averaging and equation (27) yields
subtracting. The result is
N
Nu, 1 = • ( 1 - R , /R,) (30)
a r _ ,9T,
dz ~ ^dz'
L
K{<v,T>) ^ j - f ( z - f ) (23) — R^/Rj

546 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 for each individual mode. R/c, is the eigenvalue, N is a result of
prescribed integrations of the products of the eigenfunctions, RE is
the imposed external Rayleigh number, and R/ is the internal
modified Rayleigh number at which one desires the Nusselt num-
ber.
If it is assumed that the interactions between the individual
modes can be neglected and that the eigenvalues and eigenfunc-
tions of the linearized equations can be used, then

Nu, 1 = (i | V E Nm(l - ~^)U(R, - R, ) (31)

Nu, - 1 = - ( i + Zx^ENjl - R f )U(RI - R , ) (32)

10 10 10 io- ,o 9 io 10
where R;m is the mth characteristic Rayleigh number, U(Ri — R/„,) EXTERNAL RAYLEIGH NUMBER
is the unit step function, and Nm is the result of substituting the
mth eigenfunctions into equation (25). Fig. 3 The effect of stabilization on heat transfer
Results were obtained for R/ up to 109 and for 0 < R E < 109.
Some results are shown in Fig. 2 for comparison with the experi-
mental results of Kulacki and Goldstein [2]. In Fig. 3, a series of
constant Nusselt number results, Nui, are presented on an R/ - ternal temperature difference is large enough to cause a change in
RE coordinate graph. The effect of thermal stabilization on down- the convective flow pattern. This results in the upward curvature
ward heat transfer is made apparent in this type of presentation. of the lines of constant Nusselt number. The final region appears
at very large Rayleigh numbers. The Nusselt numbers approach
D i s c u s s i o n of R e s u l t s asymptotes with a slope of unity.
A comparison of the Nusselt number predicted by this method In applying these results to post-accident heat removal (PAHR)
with experiment can only be done for the case of R^ = 0 since no situations, only very large Rayleigh numbers (R/ > 109) are of in-
other measurements are known to the authors. Fig. 2 shows the ex- terest. If one assumes that a 10 cm deep molten layer of UO2 is
perimental data of Kulacki and Goldstein. The solid line is the covered by molten iron such that the upper surface is 200° C hotter
Nusselt number predicted in this work. As can be seen, there is than the lower surface, then Rjg = 4 X 108. The corresponding Ri
good agreement. The dashed line is the experimental correlation of for 3 cal/cm 3 s internal heating is 10 11 . By extrapolating the results
Jahn and Reineke. There is a small difference in slope and in the given in Fig. 3 it is found that 25 percent of the heat is transferred
value of the Nusselt number. The reason for this discrepancy is not to the lower surface rather than the 12-13 percent which would
known at the present time. Since the predictions obtained for RE occur with equal upper and lower surface temperatures.
= 0 appear to be good, one could assume that Nusselt number pre- The temperature rise of the upper surface required for complete
dictions for RE 9^ 0 will give reasonable approximations because stabilization of the molten layer can be found from the asymptotic
nonzero RE is stabilizing and amplitudes of w and 0 will be smaller. part of the Nu = 1 curve. The approximate relationship is
It was found that the temperature profile deviated strongly from
R£ = 0 . 3 R „ Rj > 10G (33)
experiment away from bounding walls. Finite amplitude effects on
the temperature profile for RE = 0 were addressed by Catton and which reduces to
Suo-Anttila [3],
AT 0.3 QL2/k (34)
Fig. 3 shows a plot of constant Nusselt numbers for downward
heat transfer with R/ and RE as the axes. There are three regions Substituting values for molten UO2 results in
of interest. In the first region the Nusselt number is constant and
independent of the external Rayleigh number. Here, the applied L AT
external temperature difference is not large enough to cause a
change in the convective flow pattern. In the second region the ex- 1 cm 127°C
10 cm 17,200°C
20 cm 68,800°C

1.6 1 1 |-n 1 1m r r for a decay heat of 3 cal/cm 3 s. As can be seen, the predicted tem-
• KULACKI 9 GOLDSTEIN DATA POINTS [2)
PREDICTIONS FROM THIS WORK peratures are far too high, hence complete stabilization would not
RE=EXTERNALRAYLEIGH NUMBER
CORRELATION OF JAHN S REINEKE 14] occur except for very thin layers.
CORRELATION OF KULAKI a GOLDSTEII
• 1.4
Summary and Conclusions
A solution has been obtained by the power integral method for
the heat transfer from a horizontal layer of fluid with internal heat
generation and a stabilizing temperature gradient confined by
!-, 1.2 rigid boundaries. It was found that a stabilizing temperature gradi-
ent, which can result from a high thermal resistance above the
molten layer, causes significant increases in the downward heat
transfer. Complete stabilization of a molten layer is not expected
to occur.
Future work in this area should include experimental verifica-
tion of the predictions presented here. Additional numerical and
10' 10" 10'
experimental work could include the effect of a free boundary at
INTERNAL MODIFIED RAYLEIGH NUMBER, R,
the upper surface since low viscosity fluids could be superposed
Fig. 2 Comparison of predictions with experiment above molten fuel layers.

Journal of Heat Transfer NOVEMBER 1975 / 547

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 tional Heat Transfer Conference, Vol. HI, Sept, 1974, pp. 69-73.
Acknowledgment 4 Jahn, M., and Reineke, H. H., Proceedings of the Fifth International
T h i s r e s e a r c h was s u p p o r t e d b y t h e R e a c t o r R e s e a r c h Division Heat Transfer Conference, Vol. Ill, Sept. 1974, pp. 74-78.
a n d t h e D i r e c t o r a t e of L i c e n s i n g of t h e A t o m i c E n e r g y C o m m i s - 5 Malkus, W. V. R., and Veronis, G., Journal of Fluid Mechanics, Vol.
sion a n d b y t h e N a t i o n a l Science F o u n d a t i o n . T h e conclusions 4,1958, pp. 225-260.
6 Landau, L. D., and Lifshiftz, E. M., Fluid Mechanics, Pergamon
r e a c h e d in t h i s s t u d y s h o u l d n o t b e i n t e r p r e t e d as r e p r e s e n t i n g of- Press, London, 1959.
ficial o p i n i o n or policy of t h e A t o m i c E n e r g y C o m m i s s i o n . F u r t h e r 7 Edwards, D. K., and Catton, I., The International Journal of Heat
a c k n o w l e d g m e n t is given t h e U C L A C o m p u t e r C e n t e r for c o m p u - Mass Transfer, Vol. 12,1969, pp. 23-30.
tational support. 8 Catton, I., Proceedings of the Fourth International Heat Transfer
Conference, Paris, France, Vol. IV, 1970.
9 Catton, I., and Ayyaswamy, P. S., Journal of Applied Mechanics,
References Vol. 39,1972, pp. 33-36.
1 Catton, I., and Erdmann, R. C , Post-Accident Core Retention for 10 Sharman, R. D., Catton, I., and Ayyaswamy, P. S., Chemical Engi-
LMFBR's, UCLA-ENG-7343, July 1973. neering Progress Symposium Series, 1972.
2 Kulacki, F. A., and Goldstein, R. J., Journal of Fluid Mechanics, Vol. 11 Chandrasekhar, S., Hydrodynamic and Hydromagnetic Stability,
55,1972, pp. 271-280. Oxford Press, England, 1961.
3 Catton, I., and Suo-Anttila, A. J., Proceedings of the Fifth Interna- 12 Stuart, J. T., Journal of Fluid Mechanics, Vol. 4,1958, pp. 1-13.

548 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 G. C. Vliet
Assoc. Professor.

D. C. Boss
urbulent Natural Confection on
Graduate Student.

Department of Mechanical Engineering,

Upward and Downward Facing
University of Texas at Austin,
Austin, Texas
Inclined Constant Heat Flux
Surfaces
Local heat transfer data were obtained for turbulent natural convection on vertical and
inclined upward and downward facing surfaces. The test surface consisted of a 1.83 m (6
ft) wide X 7.32 m (24 ft) high plate with a constant heat flux obtained by electrical resis-
tive heating of a metal foil on the surface. The tests were conducted in air for modified
Grashof numbers up to 10lb. Measurements were made of the local surface temperature
for this constant heat flux condition, for the 'plate inclined at angles from 30 deg to the
vertical (upward facing, unstable) through the vertical to 80 deg to the vertical (down-
ward facing, stable).
The results show the location of the transition to be a function of the plate angle. For
the unstable case, the transition length decreases as the plate angle Increases from the
vertical while for the stable case the position of transition increases with the angle from the
vertical. The laminar data for both orientations are correlated as:
Nux = 0.55 (Grx*Pr)0-20

in which the gravity is the component along the surface, g cos 0. The turbulent natural
convection data are correlated quite well by the relation:
Nux = 0.17(Grx*Pr)025

In the turbulent case the correlation is independent of angle for the unstable case,
whereas for the stable case the data correlate best when the gravity is modified by cos2 0,
where 0 is measured from the vertical. Thus, there is a significant influence of angle on
the convective heat transfer for the stable turbulent region.

Introduction of Vliet [l] 1 for water exhibit a strong effect of angle on transition;
show the laminar data are correctable at various angles using the
Several physical phenomena involve turbulent natural convec- gravity component along the surface; and show the turbulent data
tion on inclined surfaces: thermal convection during the thermal to be independent of angle (except for location of transition).
stratification of liquids in vessels (of primary interest in the cryo- Lloyd and Sparrow [2] have experimentally examined the nature
genic industry), natural convection mass transfer during cavity de- of transition on inclined surfaces. Recently, instability analyses for
velopment in solution mining, and thermal convection along slop- natural convection flow on inclined surfaces have been performed
ing terrain—a meteorological phenomenon. The problem of ther- by Pera and Gebhart [3], Haaland and Sparrow [15, 16], Iyer and
mal natural convection along inclined surfaces for the thermally Kelly [4], and Kahawita and Meroney [17]. Other inclined plate in-
unstable orientation (upward facing heated surfaces) has been ex- vestigations [5, 6] have also been made. A number of experimental
amined to some extent. The constant heat flux inclined plate data investigations of turbulent natural convection on vertical surfaces
have also been conducted [7-13]. In general these turbulent data
exhibit a very weak variation of the local heat transfer coefficient
with length, and it correlates with the inclined unstable turbulent
Contributed by the Heat Transfer Division and presented at the Winter
Annual Meeting, New York, N. Y„ November 17-22, 1974, of THE AMERI- data using the vertical gravity vector.
CAN SOCIETY OF MECHANICAL ENGINEERS. Revised manuscript re-
ceived by the Heat Transfer Division, August 14, 1975. Paper No. 74-WA/
1
HT-32. Numbers in brackets designate References at end of paper.

Journal of Heat Transfer NOVEMBER 1975 / 549

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Some of the foregoing naural convection phenomena involve
convection along inclined surfaces such that the density gradient
normal to the surface is negative (stable): natural convection heat Constant Heat Flux Surface

transfer from a surface facing downward, natural convection mass Thermocouples along Center-line

transfer near the bottom of a cavity in solution mining, and the

case of flow down sloping terrain when the ground is cooler than
the air (i.e., cold surface facing upward). It is expected that the
laminar flow for these cases can be correlated using the gravity
component along the surface as for the laminar flow region for the
unstable case. However, the effect of the surface angle on the tur-
bulent convective flow is not immediately evident. This paper pre-
sents the results of an experimental investigation [19] and an ex-
tension of this work dealing with the effect on the transition and
turbulent convection heat transfer for heated inclined surfaces
with primary emphasis on the thermally stable condition—heated
surface facing downward.

Experimental Apparatus
1),(2) Steel I-Beams
The experimental apparatus was designed with the goal of pro- 3) Plywood
viding at least a 2 orders of magnitude range of fully turbulent Urethane Foam
Grashof number, since one of the shortcomings of many previous Guard Heater
Phenolic Substrate
turbulent natural convection studies has been an inadequate range
S.S. Foil
of turbulent flow beyond transition. The measurements were con- Adhesive Bonds
ducted in air rather than water as had been done previously by one Double Backed Tape
Plate Thenrocouple
of the authors. To achieve the desired range of turbulent flow in
Guard Thermocouple
air, a rather large test surface is required; thus a constant heat flux
surface was selected, as it is considerably less expensive to build,
easier to control and instrument and more closely approximates
tXTX
19)
some physical processes of interest. To achieve the two-dimension- Fig. 1 Natural convection heat transfer apparatus
al turbulent natural convection boundary layer flow desired, it is
essential that the experimental test surface be:
(a) long enough in the flow direction to achieve Grashof num- urethane foam between the surface and guard heaters to monitor
bers at least two orders of magnitude greater than transition; guard heater temperature and permit it to be operated near the
(b) wide enough to minimize edge effects; surface temperature. 12 thermocouples were used to monitor the
(c) located in a fluid of uniform temperature; and ambient air temperature along the plate.
(d) isolated from disturbances present in the room. The test plate was installed with perpendicular side panels,
The experimental test surface used in this investigation consist- which extended the full length of the plate, to reduce edge effects.
ed of a 1.83 m (6 ft) wide X 7.32 m (24 ft) high (long) constant heat At the outer edge of these side panels, a layer of standard window
flux plate as shown in Pig. 1. The constant heat flux condition was screen was installed to minimize the effects of ambient room dis-
achieved by the electrical resistance heating of a thin (0.05 mm, turbances on the natural convection flow. Rather than locating the
0.002 in.) stainless steel foil mounted on an insulated and thermal- test apparatus (heated surface, panels and screen) in a separate
ly guarded structural surface as shown in Fig. 1. The structural room, it was located in a large bay and tests were conducted after
surface consisted of a steel frame with a plywood substrate. Laid or prior to normal working hours while the air in the room was
upon the plywood was a layer of urethane foam insulation, a screen thermally unstratified and when convective disturbances in the
guard heater, a second layer of urethane foam insulation, then a fi- room were minimal. For these tests the bottom of the test surface
berglass-phenolic sheet to which the stainless steel foil was bonded was located no closer than approximately 20 cm (8 in.) above the
with pressure sensitive adhesive. The stainless steel foil surface floor.
consisted of three 0.61 m (2 ft) wide X 7.32 m (24 ft) long pieces With this apparatus and heat fluxes of approximately 57 W/m 2
laid side by side, with a small gap (approximately 1 mm) between, (18 Btu/hrft 2 ) (i.e., approximately 20°C (35°F) temperature differ-
and connected in series to match the d-c power supply. The wire ence in the turbulent region) Gr*Pr ~ 4 X 10 14 at the top of the
screen guard heater, sandwiched between the two foam layers, was plate. This provides approximately 3 orders of magnitude in Gr*Pr
electrically heated and operated near the average surface foil tem- of turbulent flow for the vertical or unstable orientations. For the
perature to minimize conduction losses through the back of the stable orientations the extent of turbulent flow is less and de-
plate. 25 thermocouples for surface measurements were located creases with shallower angles. Boundary layer profile measure-
near the plate center line at approximately 0.305 m (1 ft) intervals ments have not been made at this date. However, using the veloci-
along the surface. Two similar spanwise thermocouples were locat- ty profile data of Cheesewright [8] for a constant temperature
ed across-the plate at 6.1 m (20 ft) from the leading edge to detect plate in air, the 0.95 boundary layer thickness for the present 7.32
spanwise variations in surface temperature. These thermocouples m long vertical surface and the above conditions is approximately
for measurement of surface temperature were 0.25 mm dia and in- 45 cm, or approximately % the plate width.
troduced through the insulated back of the plate, with the junction
Smoke tests were conducted with the test apparatus in the verti-
against the back of the phenolic substrate (Fig. 1). Five guard
cal orientation which indicated that the width of the surface, the
heater thermocouples were located near the guard heater in the
presence of the side panels and the screening produced an essen-

-Nomenclature.
g = standard gravity Nu.Gr, qu" = surface heat flux
&TP = plate to bulk temperature differ-
Or* = local Grashof number Nu x = local Nusselt number
ence
Grx* = modified local Grashof number, Pr = Prandtl number S = angle relative to the vertical

550 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Natural convection data summary
Angle t o Heat A v e r a g e * 2 ' Percent Actual^ Bulk« 4 > Te.p.«
Vertical . . (1) HeaJ Flux Temp Diff
Corrections
(degrees) w/m* (B/hf ) Cond. Rad. W/m ( B / h f ) °C (»F) °C (»F)

Unstable a 30 28,0 (8.9) 1.2 23 21.1 (6.7) 33.4 (92-1) 9.6 (17.2) 1.7
b 72.0 (22.8) 1.0 19 58.0 (18.4) 33.4 (92.2) 19.6 (35.3) 4.3

Vertical c 0 27.0 (8.6) 0.0 22 21.4 (6.8) 25.2 (77.3) 9,5 (17.1) 1.7
d 69.6 (22.1) -0.9 18 58.0 (18.4) 22.8 (72.9) 19.1 (34.4) 4.8

21.6 (70.9) 20.7 (37.3) 5.1

27.1 (a.6) (6.8) (77.7) (16.9)

70.3 (22.3) (18.5) (80.0) (36.0)

26.4 (8.4) (6.4) (75.6) (16.5)

71.5 (22.7) (18.5) (82.6) (3B.9)

26.8 (8.5) (6.8) (72.7) (19.2)

68.5 (21.7) (17.1) (74.5) (42.3)

26.5 (8.4) (5.6) 22.7 (72.8) 12.8 (23.0)

70.6 (22.4) (16.4) 23.2 (73.8) 27.6 (49.8)
96.8 (30.7) (23.0) 24.2 (75.5) 37.4 (67.4)

26.8 (8.5) (4.7) (90.7) (27.6)

54.8 (17.4) (11.6) (95.2) (45.2)

28.0 (8,9) (4.7) (89.1) (29.6)

52.0 (16.5) (9.2) (94.7) (51.0)

(1) e l e c t r i c a l heat d i s s i p a t i o n per u n i t area

(2) Average percent c o r r e c t i o n s f o r c o n d u c t i o n and r a d i a t i o n l o s s e s . p o s i t i v ! r e f e r s t o l o s s
(3) Average o f a c t u a l c o n v e c t i v e heat f l u x over upper 40 p e r c e n t o f p l a t e
(4) Average over upper 40% o f p l a t e
(5) v a l u e a t t o p o f p l a t e ( u s i n g standard g r a v i t y )

tially two-dimensional boundary layer flow in the central one-third percent porosity); the particle board side panels (of 0.95 emissivi-
to one-half of the plate over its length, with no noticeable effect of ty) and the open upper end of the enclosure. The side panels and
ambient disturbances on the flow, under the conditions of test. screen were both assumed to be at the ambient temperature. As
The test surface could be orientated at any angle for either the can be seen from Table 1 the necessary radiation corrections var-
thermally stable or unstable conditions, using hoists and cables; ied from 18 to 41 percent of the dissipated electrical energy with
however the primary interest herein is in the thermally stable ori- an average of 26 percent. Corrected convective heat fluxes were
entation and only limited unstable data are presented. computed at each of the plate thermocouple locations along the
The time constant to reach a steady-state condition for the plate plate. Local heat transfer coefficients were computed from the
was approximately 3 hr, due to the low heat transfer coefficients local corrected fluxes and the local plate to bulk temperature dif-
and the thermal capacitance of the surface. The tests were con- ferences. Fluid properties were evaluated at the average film tem-
ducted in the late evening or early morning when disturbances in perature.
the laboratory were minimal, there was little thermal stratifica- Fig. 2 presents the variation of surface to bulk temperature dif-
tion, and the air exhibited little temperature variation over a 3-4 ference AT P along the plate for seven orientations (30 deg unsta-
hr period. Preliminary tests were performed to determine the rela- ble, vertical and 30, 45, 60, 70, and 80 deg stable) all for the low
tive plate and guard power levels to permit operation of the guard heat flux condition. I t is seen from this figure that: (a) there is no
heater near the plate surface temperature at various operating pronounced maximum in plate temperature prior to transition as
conditions. For each test, plate and guard powers were set at ap- is the case for similar constant heat flux natural convection tests in
propriate levels after which plate, guard and bulk thermocouple water [1, 9]; (6) the plate-bulk temperature difference (and, there-
temperature data were recorded for the duration of the run. From fore, the heat transfer coefficient) exhibited little variation in the
this record the requirements of reaching steady state, maintaining turbulent region as has been evidenced in previous investigations
low bulk thermal stratification and maintaining guard and plate [1, 7, 8, 9, 10]; and (c) as the surface is inclined at larger angles to
temperatures. nearly equal could be insured. For the lower heat the vertical, there is an increase in ATP (decrease in the heat trans-
flux tests the average nonuniformity of the bulk temperature was fer coefficient) in the turbulent region and an extension of the ap-
±5 percent of the plate to bulk temperature difference while for parent laminar region.
the higher heat flux tests the average bulk temperature nonunifor- No specific method 3 for determining the initiation of transition
mity was ± 3 percent. (end of laminar flow) and approach to fully turbulent flow was
used, other than inferring it from the characteristics of the plate
Experimental Results temperature variation. Laminar flow was assumed to terminate
A summary of the natural convection tests discussed herein is where the plate temperature leveled out or reached a weak maxi-
presented in Table 1. The thermal energy generated in the foil is mum as it did in some cases. Fully turbulent flow was assumed to
dissipated by natural convection, conduction through the insulat- begin where the plate temperature became reasonably constant.
ed backing, and by radiation to the surroundings; thus the convec- The transition range was designated as between these two loca-
tive heat flux was obtained by correcting for the conductive and tions. These locations were not consistently distinct but did gener-
radiative losses. The conductive loss was computed from the tem- ally exhibit: earlier transition with increasing angle from the verti-
perature difference between the foil and guard, and the thickness cal for the unstable case, delayed transition with increasing angle
and thermal conductivity of the polyurethane foam insulation. from the vertical for the stable orientation, and earlier transition
The presence of the guard heater limited the conductive losses to with increasing heat flux. Fig. 3 shows these approximate transi-
an average of 2.6 percent of the dissipated electrical energy or a tion regions in terms of G r ^ P r as a function of angle. The general
maximum of 6.7 percent in all cases (see Table 1). The net radia- trend with angle is apparent; however, the erratic behavior is also
tion loss from the stainless surface was computed using a radiation obvious. The instability analysis of Iyer and Kelly [4] indicates a
network which included: the stainless steel foil surface (of 0.11 delay in transition with increasing angle from the vertical for the
emissivity 2 ); the galvanized wire screen (of 0.25 emissivity and 65 stable case, which is to be expected.

2 3
This was the same stainless steel used in 1] for which this emissivity was Neither detailed flow visualization nor hot wire measurements were
measured by a Gier-Dunkle radiometer. made to specifically detect the transition region.

Journal of Heat Transfer NOVEMBER 1975 / 551

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 to O o y o
A» o Q V G
- LOW HEAT FLUX TESTS
« a - 30° Unstable A « o 0 V o
c - 0° V e r t i c a l

A h - 30' Stable A O 0 O V o
- <•> j - 45° Stable
ABt O 0 V o
O 1 - 60° Stable

V o - 70° Stable 40 » O o V o -
o q - 80° Stable
EA a 0 Q V o
O V
QA» <J> Q V o
• «A<> Q V o
Q » AJ> O V o
« ^ O V o
• Ek£ Q o
O o
• O AD o -
• A Q o V O
«AQ 0 V O
1 i 4a • Ol w O 1 1 1
n 5 1 10 15 1 ?o ?s 1 30°F
0 5 10 15°C

TEMPERATURE DIFFERENCE AT p =• °C (°F)

Fig. 2 Temperature difference variation, various angles

Also presented for comparison are the transition ranges for the plotted in Fig. 5 in terms of Nu x versus G r ^ P r where the accelera-
inclined plate water and air data of [1]. The one data point for air tion in Gr x * is the vertical gravity vector g. The unstable data are
is consistent with the present data; however, the Grj*Pr range for seen to correlate very well with the vertical data as has been pre-
water in general falls above the air data. This is qualitatively con- viously reported [1]; however, there is seen to be a monotonic de-
sistent with the work of Godaux and Gebhart [18]. viation of the stable turbulent data. A cross plot of these data indi-
A correlation of the laminar portion of the data is presented in cated that the stable data could be correlated by replacing the ac-
Fig. 4, in which the gravity in the modified Grashof number is re- celeration in Gr** by g cos 2 6. All of the turbulent inclined surface
placed by the component along the plate, g cos 9, where 6 is the data are presented in Fig. 6 in the form of Nu x versus Gr I *Pr,
angle measured from the vertical for both the stable and unstable where for the unstable orientation the vertical gravity vector is
orientations. The criterion for the end of laminar flow is as stated used, but for the stable orientation g is replaced by g cos 2 6. Al-
in the foregoing. The work of Dotson [14] for air on a vertical plate: though this correlation of data also exhibits substantial scatter
Nu* = 0.55 (Gr I *Pr) 0 - 2 ; and that of Vliet [1] for water on inclined (±15 percent), the correlation is greatly improved. The relation as-
(unstable) surfaces: Nu* = 0.6 (Gr I *Pr) 0 - 2 are shown with the cribed to this data Nu* = 0.17 (Gr*Pr) 0 2 5 may be compared to the
present data for comparison. The spread in the present data is at turbulent inclined plate constant heat flux water data of Vliet [1]
least partially attributable to nonconstant heat flux in the laminar which was best described by Nil* = 0.302 (Gr x *Pr) 0 - 24 . The present
region where the heat flux corrections varied with length, and to data fall approximately 30 percent below this earlier data for
the bulk temperature variation which was generally greatest near water. It is to be noted that the vertical constant temperature tur-
the leading edge of the plate. bulent data for air of Warner [7] and Cheesewright [8] correlate
All of the turbulent high heat flux data (stable and unstable) are approximately as:

Nu x = C(Gr*Pr) 1 / 3
where the coefficients C are approximately 0.088 and 0.122 for the
two sets of data, respectively. If this relation is recast as

Nux = C'(Gix*Pr)lM
. Water Data (1)
then the coefficients C are approximately 0.16 and 0.21 for the
two sets of data, respectively, and thus the agreement of the
present data with [7, 8] is quite good.
Present Air Data
The reduction in the turbulent heat transfer coefficient with in-
creasing angle from the vertical for the stable case may be attrib-
uted principally to the damping of the turbulence due to the ther-
mally stable condition of the boundary layer flow for this orienta-
• High Flux Series tion, and the decrease in the gravity component along the surface.
& Air Data The implications of this effect are significant in some physical
phenomena such as prediction of cavity shape development in so-
lution mining (natural convection mass transfer), and warm air
above colder sloping terrain (meteorological phenomenon), since
the reduction in the convective transfer coefficient with angle is
substantial.
90 80 60 40 20 20 40 60 80 90

Angle to Vertical - Degrees Conclusions

Fig. 3 Transition range as interpreted from local temperature behavior There is a very pronounced effect of surface orientation on the

552 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 CDA,B 3 0 UNSTABLE
a c,o 0° VERTICAL
o E 10 STABLE
A F,G 2 0 °
x H,I 3 0 ° "
rn x J,K 45°
x L,W,N60° "
o _ + O,P 7 0 ° "
<7 Q,R 8 0 "

,0.2
NU X =0.6(GR*PR)
[VLIET(I)
CM
o

-NUX=0.55(GR'PR)
f00TS0N(!4)l

108 10- 1010 101l 10 12 10 13

x
GRx PR
Fig. 4 Laminar flow region

o _
^_, © A.a 30°UNSTABLE
a c,o 0°VERTICAL
O E 1 OSTABLE
A F,G 2 0° .,
X H,l 30° «
ro X J,K 4 5 °
x L,M,N60° "
o _ + o,p 70° ,,
V Q,R 80° «
«*t
"*f
CxJ
O

o
10 10 11 10 12 10 13 10 14 10

'x PR
Fig. 5 Turbulent flow region

turbulent natural convection transfer coefficient for the stable ori- 2 Lloyd, J. R., and Sparrow, E. M , "On the Stability of Natural Con-
entation. Whereas the turbulent data for the unstable orientation vection Flow on Inclined Plates," Journal of Fluid Mechanics, Vol. 42,1970,
is a function of the vertical gravity vector, turbulent natural con- pp. 465-470.
3 Pera, L., and Gebhart, B., "On the Stability of Natural Convection
vection transfer coefficients decrease significantly with angle from Boundary Layer Flow Over Horizontal and Slightly Inclined Surfaces," In-
the vertical for the stable orientation and the data are best corre- ternational Journal of Heat and Mass Transfer, Vol. 16, No. 6, 1973, pp.
lated with the acceleration replaced by g cos 2 8. 1147-1164.
4 Iyer, P . A., and Kelly, R. E., "The Stability of the Laminar Free Con-
vection Flow Induced by a Heated Inclined Plate," International Journal of
Acknowledgments Heat and Mass Transfer, Vol. 17, No. 4, Apr. 1974, pp. 517-526.
This research was sponsored by the Bureau of Engineering Re- 5 Tritton, D. J., "Transition to Turbulence in Free-Convection Bound-
search, The University of Texas at Austin. ary Layer on an Inclined Heated Plate," Journal of Fluid Mechanics, Vol.
16,1963, p. 417.
6 Rich, B. R., "An Investigation of Heat Transfer From an Inclined
References Flat Plate in Free Convection," TRANS. ASME, May 1953, p. 489.
1 Vliet, G. C , "Natural Convection Local Heat Transfer on Constant 7 Warner, C. Y., and Arpaci, V. S., "An Experimental Investigation of
Heat Flux Inclined Surfaces," JOURNAL OF HEAT TRANSFER, TRANS. Turbulent Natural Convection in Air Along a Vertical Heated Flat Plate,"
ASME, Series C, Vol. 91, No. 4, Nov. 1969, pp. 511-517. International Journal of Heat and Mass Transfer, Vol. 11,1968, p. 397.

Journal of Heat Transfer NOVEMBER 1975 / 553

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 O
- © A,B 30° UNSTABLE
0 c,0 0° VERTICAL
» E 10° STABLE
A F,G 20° "
X H,l 30° "
X J,K 45° "
CO * L,M,N60° "
O ~ + 0,P 70 "
V Q,R 80° " N 11^0302 (GR*PR)Q , - ^ ^ • 9 ^ - ^ '

CM
O
- *
• J ' 9«
-;-s-r a \. „* 0 25
NUX = 0.17 (GR*PR)
[PRESENT DATA]

, , , , n , | • i ••, i , i , , M : ' ' ' ' " 1 ' ' ' < • < - < |

10 10 10 11 12 13 10 14 10 15
GR,-* PR
Fig. 6 Turbulent flow region correlated

8 Cheesewright, R., "Turbulent Natural Convection from a Vertical Transfer, Vol. 15, No. 2, 1972, pp. 193-202.
Plane Surface," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Se- 14 Dotson, J. P., "Heat Transfer From a Vertical Plate by Free Convec-
ries C, Vol. 90, No. 1, Feb. 1968, pp. 1-8. tion," MS thesis, Purdue University, 1954.
9 Vliet, G. C, and Liu, C. K., "An Experimental Study of Turbulent 15 Haaland, S. E., and Sparrow, E. M., "Wave Instability of Natural
Natural Convection Boundary Layers," JOURNAL OF HEAT TRANS- Convection on Inclined Surfaces Accounting for Non-Parallelism of the
FER, TRANS, ASME, Series C, Vol. 91, No. 4, Nov. 1969, pp. 517-531. Basic Flow," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Series
10 Fugii, T., Takeuchi, M, Fujii, M., Suzaki, K„ and Uehara, H., "Ex- C, Vol. 95, Aug. 1973, pp. 405-407.
periments on Natural Convection Heat Transfer From the Outer Surface of 16 Haaland, S. E., and Sparrow, E. M., "Vortex Instability of Natural
a Vertical Cylinder to Liquids," International Journal of Heat and Mass Convection Flow on Inclined Surfaces," International Journal of Heat and
Transfer, Vol. 13, No. 5, May 1970, pp. 753-787. Mass Transfer, Vol. 16, Dec. 1973, pp. 2355-2367.
11 Pirovano, A., Viannay, S., and Jannbt, M., "Natural Convection in 17 Kahawita, R., and Meroney, R. N., "Vortex Mode of Instability in
the Turbulent Regime Along a Vertical Plate," Fourth International Heat Natural Convection Flow Along Inclined Plates," International Journal of
Transfer Conference (Paris-Versailles, France), Vol. 4, Paper No. NCI.8, Heat and Mass Transfer, Vol. 17, No. 5,1974, pp. 541-548.
1970. 18 Gebhart, B., and Godaux, F., "An Experimental Study of the Transi-
12 Lock, G. S. H., and Trotter, F. J. deB., "Observation of the Structure tion of Natural Convection Flow Adjacent to a Vertical Surface," Interna-
of a Turbulent Free Convection Boundary Layer," International Journal of tional Journal of Heat and Mass Transfer, Vol. 17, No. 1,1974, pp. 93-107.
Heat and Mass Transfer, Vol. 11, No. 8,1968, pp. 1225-1232. 19 Ross, D. C, "Experimental Investigation of Natural Convection Heat
13 Kutateladze, S. S., et al., "Turbulent Natural Convection on a Verti- Transfer on Inclined Surfaces," MS thesis, The University of Texas at Aus-
cal Plate and in a Vertical Layer," International Journal of Heat and Mass tin, May 1974.

-Discussion- rect variation, along the surface, of the heat transfer rate rather
4
than of the indirect effect on the temperature difference across the
Y. Jaluria boundary layer, which would be the case for a uniform flux surface.
The authors have made an extensive experimental study of tur- However, it is true that in order to achieve an isothermal surface, a
bulent natural convection heat transfer from inclined heated sur- much greater elaboration is needed in the experimental arrange-
faces and have presented a very impressive collection of important ment. A uniform flux surface condition would, nevertheless, be
heat transfer data. Since much of the interest in such natural con- quite satisfactory in a study of the fundamental mechanisms in-
vection flows is in the heat transfer, these results are very valuable. volved, see Gebhart. 5 This is a very important consideration and a
However, there are a few considerations which, if taken into ac- study of the velocity and temperature fields in the boundary re-
count, would make such a study of greater fundamental and prac- gion would add considerably to an understanding of the transport
tical significance. phenomenon. This would not only be of great importance in the
From the standpoint of heat transfer, the isothermal surface study of turbulence in natural convection flows but in the under-
condition is of greater interest due to its more frequent occurrence standing of the heat transfer data obtained as well. Given the elab-
in practical problems. Moreover, it would allow a study of the di- orate experimental arrangement the authors have, it would be very

4 5
Member of Research Staff, Western Electric Co., Engineering Research Gebhart, B., "Instability, Transition and Turbulence in Buoyancy-In-
Center, Princeton, N. J. duced Flows," Ann. Rev. Fluid Mech., Vol. 5, 1973, p. 213.

554 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 O
- © A,B 30° UNSTABLE
0 c,0 0° VERTICAL
» E 10° STABLE
A F,G 20° "
X H,l 30° "
X J,K 45° "
CO * L,M,N60° "
O ~ + 0,P 70 "
V Q,R 80° " N 11^0302 (GR*PR)Q , - ^ ^ • 9 ^ - ^ '

CM
O
- *
• J ' 9«
-;-s-r a \. „* 0 25
NUX = 0.17 (GR*PR)
[PRESENT DATA]

, , , , n , | • i ••, i , i , , M : ' ' ' ' " 1 ' ' ' < • < - < |

10 10 10 11 12 13 10 14 10 15
GR,-* PR
Fig. 6 Turbulent flow region correlated

8 Cheesewright, R., "Turbulent Natural Convection from a Vertical Transfer, Vol. 15, No. 2, 1972, pp. 193-202.
Plane Surface," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Se- 14 Dotson, J. P., "Heat Transfer From a Vertical Plate by Free Convec-
ries C, Vol. 90, No. 1, Feb. 1968, pp. 1-8. tion," MS thesis, Purdue University, 1954.
9 Vliet, G. C, and Liu, C. K., "An Experimental Study of Turbulent 15 Haaland, S. E., and Sparrow, E. M., "Wave Instability of Natural
Natural Convection Boundary Layers," JOURNAL OF HEAT TRANS- Convection on Inclined Surfaces Accounting for Non-Parallelism of the
FER, TRANS, ASME, Series C, Vol. 91, No. 4, Nov. 1969, pp. 517-531. Basic Flow," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Series
10 Fugii, T., Takeuchi, M, Fujii, M., Suzaki, K„ and Uehara, H., "Ex- C, Vol. 95, Aug. 1973, pp. 405-407.
periments on Natural Convection Heat Transfer From the Outer Surface of 16 Haaland, S. E., and Sparrow, E. M., "Vortex Instability of Natural
a Vertical Cylinder to Liquids," International Journal of Heat and Mass Convection Flow on Inclined Surfaces," International Journal of Heat and
Transfer, Vol. 13, No. 5, May 1970, pp. 753-787. Mass Transfer, Vol. 16, Dec. 1973, pp. 2355-2367.
11 Pirovano, A., Viannay, S., and Jannbt, M., "Natural Convection in 17 Kahawita, R., and Meroney, R. N., "Vortex Mode of Instability in
the Turbulent Regime Along a Vertical Plate," Fourth International Heat Natural Convection Flow Along Inclined Plates," International Journal of
Transfer Conference (Paris-Versailles, France), Vol. 4, Paper No. NCI.8, Heat and Mass Transfer, Vol. 17, No. 5,1974, pp. 541-548.
1970. 18 Gebhart, B., and Godaux, F., "An Experimental Study of the Transi-
12 Lock, G. S. H., and Trotter, F. J. deB., "Observation of the Structure tion of Natural Convection Flow Adjacent to a Vertical Surface," Interna-
of a Turbulent Free Convection Boundary Layer," International Journal of tional Journal of Heat and Mass Transfer, Vol. 17, No. 1,1974, pp. 93-107.
Heat and Mass Transfer, Vol. 11, No. 8,1968, pp. 1225-1232. 19 Ross, D. C, "Experimental Investigation of Natural Convection Heat
13 Kutateladze, S. S., et al., "Turbulent Natural Convection on a Verti- Transfer on Inclined Surfaces," MS thesis, The University of Texas at Aus-
cal Plate and in a Vertical Layer," International Journal of Heat and Mass tin, May 1974.

-Discussion- rect variation, along the surface, of the heat transfer rate rather
4
than of the indirect effect on the temperature difference across the
Y. Jaluria boundary layer, which would be the case for a uniform flux surface.
The authors have made an extensive experimental study of tur- However, it is true that in order to achieve an isothermal surface, a
bulent natural convection heat transfer from inclined heated sur- much greater elaboration is needed in the experimental arrange-
faces and have presented a very impressive collection of important ment. A uniform flux surface condition would, nevertheless, be
heat transfer data. Since much of the interest in such natural con- quite satisfactory in a study of the fundamental mechanisms in-
vection flows is in the heat transfer, these results are very valuable. volved, see Gebhart. 5 This is a very important consideration and a
However, there are a few considerations which, if taken into ac- study of the velocity and temperature fields in the boundary re-
count, would make such a study of greater fundamental and prac- gion would add considerably to an understanding of the transport
tical significance. phenomenon. This would not only be of great importance in the
From the standpoint of heat transfer, the isothermal surface study of turbulence in natural convection flows but in the under-
condition is of greater interest due to its more frequent occurrence standing of the heat transfer data obtained as well. Given the elab-
in practical problems. Moreover, it would allow a study of the di- orate experimental arrangement the authors have, it would be very

4 5
Member of Research Staff, Western Electric Co., Engineering Research Gebhart, B., "Instability, Transition and Turbulence in Buoyancy-In-
Center, Princeton, N. J. duced Flows," Ann. Rev. Fluid Mech., Vol. 5, 1973, p. 213.

554 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 worthwhile to pursue the problem and contribute significantly in of transition over inclined surfaces would be valuable in clarifying
this area. Not much work exists at the present time on this aspect this point for this flow configuration, and would probably indicate
of the problem, for these flows, and it will be a very valuable con- the proper criterion for determining the boundaries of the transi-
tribution to the field of natural convection turbulence if such a tion regime.
study is done. It will provide the basis for further work on similar
turbulent natural convection flows.
Another important consideration concerns the question of the
criterion for the determination of the onset of transition. This has Authors' Closure
been a vague and controversial point in natural convection studies.
Recently, for vertical natural convection flows, Jaluria and The authors agree that the constant wall temperature boundary
Gebhart 6 have made a detailed study of the transition phenome- condition is probably more common practically than constant heat
non and have considered this question at length. They found that flux; however, in the turbulent range to which the present work is
following the growth of disturbances, turbulent bursts arose which directed, the constant temperature and constant heat flux condi-
led rapidly to a thickening of the boundary region and a distortion tions are basically the same, since the heat transfer coefficient is
of the base velocity and temperature profiles. This further led to essentially independent of length. In mass transfer applications
an effect on the rate of heat transfer and consequently on the mea- the boundary condition is, of course, constant concentration,
sured wall temperature. Therefore, it was found that the wall tem- which has an analogy to the constant temperature heat transfer
perature distribution shows a drop much farther downstream of problem.
the appearance of several events that characterize transition, like
As to the use of the word transition, no boundary layer measure-
turbulent bursts and thickening of the boundary region. Taking
ments were made to observe the onset of instability. The range of
the wall temperature distribution for determining the boundaries
"transition" data in Fig. 3 refers to that as inferred from local sur-
of the transition regime, therefore, excludes much of the actual
face temperature behavior and, therefore, is probably above the
transitional phenomenon in the boundary region. A similar study
onset of instability; however, from a practical viewpoint the effect
on heat transfer is important and it is felt that the information in
Fig. 3 is of value.
The authors have attempted to obtain support to extend this
6
Jaluria, Y., and Gebhart, B., "On Transition Mechanisms in Vertical work to detailed measurements of the turbulent natural convec-
Natural Convection Flow," J. Fluid Mech., Vol. 66,1974, p. 309. tion boundary layer but to no avail.

Journal of Heat Transfer NOVEMBER 1975 / 555

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 N. Seki
Professor. iperimental Study on Natural
S. Fukusak©
Assl. Professor.
Confection Heat Transfer With
M. Nakaoka Density Inwersion of Water
Graduate Student.

Department of Mechanical Engineering,

Between Two Horizontal
Hokkaido University,
Sapporo, Japan
Concentric Cylinders
An experimental investigation concerning the effect of density inversion on steady natu-
ral convection heat transfer of water between two horizontal concentric cylinders with
diameter ratio ranging from 1.18 to 6.39 is carried out. Water, as a testing fluid, has the
maximum density at 4°C. Temperature of the inner cylinder is maintained at 0°C, while
temperatures of the outer cylinder are varied from 1 to 15°C, with Grashof number rang-
ing from 3.2 X 101 to 2.7 X 105. Photographs and qualitative description of the flow pat-
terns, temperature profiles, local and average Nusselt number are presented. From the
present experimental investigation, it is demonstrated that the effect of density inver-
sion is unexpectedly large and the average Nusselt number is a peculiar function of tem-
perature difference between outer and inner cylinder, unlike the previous results on
common fluids without density inversion.

Introduction is a function of temperature difference between hot and cold sur-

face temperature of opposing walls and furthermore it has a mini-
Many studies on natural convection heat transfer related to the mum value at 8° C where two counter standing eddies of approxi-
melting of ice have been performed. Yen [l] 1 investigated the mately the same size coexist in the vessel.
problem of the onset of natural convection in a melted layer of ice Many experimental and analytical studies on the natural con-
heated from below. The experimental results on the same subject vection heat transfer of common fluids with a monotonic relation
but heated from above were reported by Yen and Galea [2], Sun, et between density and temperature have been reported. For exam-
al. [3] studied both analytically and experimentally the effect of ple, Liu, et al. [7], using five sets of cylinders, determined flow pat-
density inversion on the onset of natural convection and on the terns, temperature profiles, and overall heat transfer relation on
melting rate of ice. Recently, Sugawara, et al. [4] presented an ex- air, water, and silicon fluid. Grigull and Hauf [8], using air as a
perimental investigation concerning the heat transfer phenomena testing fluid, categorized the flow patterns into three regimes and
in a melted layer of ice heated from above under various radiant evaluated experimentally the local and the mean Nusselt number
heat fluxes. They reported temperature distributions and melting on the surface of inner cylinder. And Bishop, et al. [9,10] also in-
rates for the cases in which the mode of heat transfer in the layer is vestigated the same problem both experimentally and analytically
either conductive or convective depending on a hydrodynamic in- and reported that one steady and three unsteady flow patterns
stability caused by the density inversion. Watson [5] and Seki, et were observed in their experiments.
al. [6] investigated analytically the natural convection heat trans- The purpose of this study is to evaluate the effect of density inver-
fer of water under the influence of density inversion in a confined sion on natural convection heat transfer of a melted water con-
rectangular vessel. They showed that the average Nusselt number tained in two horizontal concentric cylinders whose surface tem-
peratures are kept isothermal. Five inner and two outer cylinders
are used to change the diametric ratio and the curvature. Temper-
1
ature of the inner cylinder 71,- is uniformly maintained at 0°C,
Numbers in brackets designate References at end of paper. while that of the outer cylinder T0 can be varied from about 1 to
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer 15°C. Visualized flow patterns, temperature profiles, local Nusselt
Division January 27,1975. Paper No. 76-HT-G. number Nui<)c, and average Nusselt number Nu are also discussed.

556 / NOVEMBER 1975 Copyright © 1975 by ASME Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 B traversing Ihermocoupl

ft! O Thermocouples
Water .rtet. dra>n

I
Fig. 1 A schematic diagram of experimental apparatus

Experimental Apparatus and Procedure Table 1 Range of test variables, using water as testing fluid and T, = 0°C
A schematic diagram of the experimental apparatus is shown in
Fig. 1. Lucite outer cylinders of both 121.5 and 65.5 mm in diame- 121.5 19.0 1.5 9.5 6.39 51.2 4.1 x 10* 2.7 X 10 3
ter and 2 mm of thickness are used and values of diametric ratio
38.0 3.0 11.0 3.20 41.7 4.4 X 10 1.7 X 10
d0ldi ranging from 1.18 to 6.39 are obtained by using the inner cyl-
55.4 3.0 8.0 2.19 33.0 2.2 x 10* 6.3 X 10 4
inders with diameters of 19.0 and 38.0 mm (copper), 55.4, 69.6, and
99.7 mm (lucite). Surface of the test cylinder, except a part of ob- 69.6 3.0 7.4 1.75 25.9 1.1 x 10 4 2.7 x 10 4

serving slit of the outer cylinder, is finished with a black paint to 99.7 3.0 7.0 1.22 10.9 8.7 X 10 2
1.9 X 10 3
minimize the reflection of light beam. For observation of the flow 65.5 19.0 3.0 15.4 3.44 23.2 3
8.0 x 10 4.0 x 10 4
patterns, an end wall of the cylinder is installed with a pairglass.
38.0 3.0 9.5 1.72 17.7 1.7 x 10 3
5.1 x 10 3
Surface temperature of the inner cylinder is uniformly kept at
55.4 1.0 4.0 1.18 5.0 3.2 x 10 1
9.0 X 1 0 1
0°C by a temperature-controlled coolant delivered into the cylin-
der. Surface temperature of the outer cylinder is uniformly main-
tained at a constant temperature by using twelve independent and In each run, it takes about 8-15 hr to attain a steady state.
controllable main heaters respectively prepared by guard heaters Under this condition, visualization of flow patterns and measure-
outside. Rubber felt of 1.5 mm in thickness is embedded between ment of the temperature distributions are performed, while power
the main and the guard heaters. As will be shown in Fig. 1, outside input is recorded. T0 is varied every 1°C ranging from about 1 to
of these heaters two vinyl pipes of 10 mm in diameter are equipped 15°C. Table 1 summarizes the tests performed and the range of ex-
to provide a coolant and to avoid the heat flow from the surround- perimental parameters covered.
ings. Three brass tubes of 2.5 mm in diameter to let tracing parti-
cles into the testing fluid are attached on the top of the outer cyl- Discussion of the Results
inder. Eight thin probes (stainless steel pipe of 0.8 mm in diame- Flow Patterns. Typical flow patterns under several TD are
ter) with Cu-Co thermocouple of 0.1 mm in diameter are arranged shown in Figs. 2, 3, 4, 5, and 6 along with schematic drawings. In
at every 30 deg angular intervals from the top of the outer cylinder these experiments, the flow patterns are stable and symmetrical in
to measure the temperature distributions in the fluid. Uniformness respect to the vertical axis, and two-dimensional flow is attained.
of the surface temperature of outer cylinder is accomplished by ad-
Fig. 2 shows the flow patterns observed for 4°C and djdi = 3.2.
justing the input of electric power to each of the heaters.
The flow seems to be very stable. The movement of eddy is upward
For visualizing the flow patterns, nine 100 W-projection lamps along the inner cylinder but downward along the outer cylinder. In
are arranged concentrically around the outer cylinder. The radiant this flow patterns, the fluid near the upper part of the vertical axis
heat energy of light beam is eliminated by passing through a semi- flows at a higher speed than that on the other part in the gap. The
circular water jacket of 20 mm in thickness. The tracing particle is center of the eddy is clearly observed in the upper part of the an-
a mix of aluminum powder and distilled water boiled for at least 3 nuli, while in the lower part a counter secondary eddy can be seen.
hours. Photographs of flow pattern are taken by 36 mm camera However, the intensity of the eddy is so weak that tracing particles
with ASA 100 (DIN 21) film. The time exposure is 60 s. The slit of do not follow it exactly. But this secondary eddy flow disappears as
light beam is packed with an insulating material except the operat- the gap width L between the inner and the outer cylinder de-
ing period of visualization. A distilled water is used as a testing creases. When L is very small, namely the diameter ratio in the
fluid to minimize undesirable air bubble on the surface boundaries order of 1.18 or 1.22, the flow behaviors are different from the
of the inner and the outer cylinders. aforementioned ones.

-Nomenclature

di = inner cylinder diameter cylinder, defined in equation (2) AT = temperature difference, Ta - T; (=

d„ = outer cylinder diameter Nui oc = local Nusselt number, defined in T0)
dtJdi = diametric ratio [1 = coefficient of thermal expansion (abso-
equation (1)
g ~ gravitational acceleration lute value)
P r = Prandtl number
Gr = Grashof number, g /3L3 (AT)/v2 8 = dimensionless temperature, (T — Ti)/
h = heat transfer coefficient Qt = total heat transfer rate, per unit
(To - Ti)
k ~ thermal conductivity length
v = kinematic viscosity
ke = effective thermal conductivity, de- q = local heat flux
fined in equation (3) R = dimensionless radius, (r — r,)/(r 0 — r;) Subscript
L = gap width, (d„ — d;)/2 r = radius i = refers to inner cylinder
Nu = average Nusselt number for inner o = refers to outer cylinder
T = temperature

Journal of Heat Transfer NOVEMBER 1975 / 557

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 size, while another one flowing upward along the inner cylinder is
getting smaller and finally disappears as To increases. Flow pat-
terns for L = 10.9 mm (dold i = 1.22), To = 4°C are considerably
different from the aforementioned ones, as shown in Fig. 6. A very
sma]] eddy which flows downward along the vertical axis is ob-
served in the vicinity of the top of annuli for all To. This is be-
lieved due to the effects of L and cylinder curvature.
Temperature Distributions and Local Nusselt Numbers.
Distributions of temperature T versus radial position r are ob-
tained for all values of 6.T (= To) and also at selected angular posi-
tions for every aspect ratio dold i covered in this study.
It is well known that the temperature profiles of the fluid with-
out density inversion in either annuli or concentric sphere have
consistently a general trend except the case of extremely small L.
In case of the concentric sphere, Bishop, et al. [11] reported that
Fig. 2 A photograph and a schematic view of flow paUem, d Q = 121.5 the temperature profiles had a similar feature and they might be
mrn, d, = 3S.0 mm, and TQ = 4 Q C conveniently defined by five distinct regions. It should therefore
be noted that the dimensionless temperature profile at each select-
ed angular position is similar and independent of both 6.T and a
common test fluid.
The typical dimensionless temperature 0 as a function of dimen-
sionless radius R is shown in Figs. 7(a)-9(a). The corresponding
local Nusselt numbers based on inner and outer cylinder are shown
in Figs. 7(b)-9(b). The local Nusselt number in this study is de-
fined as
qL
NUl oc =-- (1)
6.Tk
where q is a local heat flux. The local heat flux for the outer cylin-
der is evaluated by electric input to each main heater. On the other
hand, the local heat flux for the inner cylinder is obtained by mea-

Fig. 3 A photograph and a schemallc view of flow paUem, dQ = 121.5

mm, d, = 55.4, mm and TQ = 5°C

FIg. 5 A photograph and a schemallc view of flow pauern, d Q = 65.5 mm,

=
d, 19.0 mm, and TQ =SoC

Fig. 4 A photograph and a schematic view of flow paUern, dQ = 121.5

mm, d, = 69.6 mm, and To = 6°C

The flow patterns obtained for To = 5°C and do/di = 2.19 are al-
most similar to those for 4°C; however, the counter secondary eddy
in the lower part can be seen more clearly, as shown in Fig. 3. The
appearance of counter secondary eddy may be due to the effect of
density inversion at 4°C. When To ranges from 6 to 7°C, the influ-
ence of density inversion becomes more evident. That is, the eddy
in the lower part of the annuli becomes large and finally two coun-
ter eddies of almost equal strength exist in the gap (see Fig. 4). For
L = 25.9 mm (dold j = 1.75), To = 6°C, aforementioned secondary
eddy occurs in the upper part of the gap, as can be seen in Fig. 4.
For the combinations of dold i with L ranging from 17.7 to 51.2
mm, such flow patterns as shown in Fig. 4 are always observed.
As can be seen from Fig. 5, in the case of higher To (= 8°C), the Fig. 6 A photograph and a schemallc view of flow paUern, d Q = 121.5
eddy flowing upward along the outer cylinder gradually extends its =
mm, d, 99.7 mm, and To = 4°C

558 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 7(a) Temperature profiles, d^/d, = 3.20 and T„ = 4°C Fig. 8(a) Temperature profiles, d„/di = 1.75 and T„ = 6°C

Pure conduct ion Pure conduction

.6 12 Nu„ Nu
l\ * « ioc,i

180°
Fig. 7(6) Local Nusselt numbers of the Inner and the outer cylinders, Fig. 8(6) Local Nusselt numbers of the inner and the outer cylinders,
do/d, = 3.20 and T„ = 4°C do/d, = 1.75 and T0 = 6°C

suring a temperature gradient adjacent to the surface. In these fig- temperature gradient in the vicinity of the separation region be-
ures, angular position is measured from the upper vertical axis and comes relatively small. Thus, the local Nusselt number for the
the broken lines in each temperature profile indicate the 4°C iso- outer cylinder near the separation region becomes minimum value,
therm. It should be pointed out that these temperature profiles, in while for the inner cylinder the local Nusselt number increases
general, are distinctly different from the ones for common fluid, with increasing angular position as shown in Fig. 8(6).
which are independent of AT. Fig. 9(a) shows 8 versus R for T0 = 8°C corresponding to the
Fig. 7(a) shows 6 versus R in the annuli with T0 = 4°C. In this flow pattern as shown in Fig. 5. The temperature at 0 deg angular
figure, it is understood that the region with a precipitous tempera- position is the highest one in the gap and it decreases as the angu-
ture drop is related to the high heat convection rates through the lar position increases. But it can be seen that the behavior of the
thin boundary layer near the cylinder. Magnitude of the precipi- temperature profiles at 90, 120, and 150 deg is a little different
tous drop near the outer cylinder decreases as the angular position from those at other positions. This could be understood from the
increases from 0 C C, but the trend is reversed near the inner cylin- fact that these results are obtained when the measurement is made
der. Temperature inversion is observed at every angular position across the boundary between two eddies shown in Fig. 5. And in
ranging from 0 to 90 deg. Such a temperature inversion may imply the vicinity of the angular positions between 0 and 30 deg, temper-
a possibility of high heat transport by angular conveetion flow ature is higher than the others. This implies that there is such a
' rather than radial one. It is also evident that magnitude of the stagnant region where the conduction heat transfer is dominant
temperature inversion increases with decreasing angular position. rather than the convection one especially in the vicinity of the
As will be seen in Pig. 2, little convective flow occurs in the region outer cylinder surface. Thus, as might be expected, local Nusselt
ranging from 160 to 180 deg. number for the outer cylinder decreases as the angular position de-
Fig. 7(6) shows that the local Nusselt number for T0 = 4°C creases and reaches a minimum value at 0 deg as shown in Fig.
reaches the minimum value at 0 deg of angular position on the 9(6). But the local Nusselt number for the inner cylinder gradually
inner cylinder and 180 deg on the outer cylinder. This characteris- decreases as the angular position increases from 90 to 180 deg,
tic feature may correspond to the aforementioned behavior of the while between 90 and 0 deg it increases significantly and reaches a
temperature profiles as illustrated in Fig. 7(a). maximum at 0 deg. This seems to be resulted from the fact that
Fig. 8(a) shows the result for T 0 = 6°C where two standing near the top of the inner cylinder the fluid heated by the outer cyl-
eddies of almost same size coexist. As will be estimated from the inder comes down directly near the top of the inner cylinder (see
flow pattern in Fig. 4, it should be noted that the temperature near Fig. 5).
the boundary between two standing eddies sharply drops at angu- For T0 = 12°C temperature profiles are inversion of those shown
lar positions between 90 and 150 deg as shown in Fig. 8(a) and the in Fig. 7(a) and local Nusselt number has an opposite tendency to
temperature near the outer cylinder at 180 deg where the cooled that shown in Fig. 7(6).
water comes down is lower than that at 90 deg. Furthermore, the Average Nusselt Number. The average Nusselt number Nu

Journal of Heat Transfer NOVEMBER 1975 / 559

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 for the inner cylinder is defined as, T.S 4°c
d./
T.i 9«c ttST.SS'c //
639
//
Nu = (2) 3 '/<> /?(.
irdiATk Z 19 Mueller.et af.
1 75 *' 7/ \ (air,water,silico
1 11
where Qt is the total heat transfer rate which is evaluated by sum- .1 M e JL* Kraussotd
ming up the heat flows of all the separated main heaters. More- 1 72 Y/ta (water oil)
1 .18
over, a convenient representation in terms of an effective thermal V,
>s
conductivity ke is defined as, •

j t In (djdj) 0.5
kjk> (3)
2-rrkAT °/4 "
where kjk represents the ratio of actual convection heat transfer
rate to pure conductive heat transfer rate.
Liu, et al. [7] summarized their experimental results on common
0.0 ~**'
fluids as air, water, and silicon fluid and presented a correlation in 0 8
terms of kelk and Pr 2 Gr/(1.36 + Pr). In their experimental study 1„„ 1 Pr Gr ,
Log ( — « . D , I )
T, is higher than Ta and the effect of density inversion on the heat (1.36*Pr)
transfer is not discussed. The present experimental results are
Fig. 10 Comparison of test data by previous investigators
shown in Fig. 10 along with the correlations by Liu, et al. [7] and
Kraussold [12]. The present data agree fairly well with those of
previous investigators in temperature ranges without the effect of
density inversion, that is, T„ < 4°C and Ta > 9°C. However, the
present results ranging approximately from T0 = 4 to 9°C shows
on the natural convection heat transfer under the influence of den-
a different tendency from the others. This appears to be caused by
sity inversion in a rectangular confined vessel suggested that a
the existence of two eddies which are resulted from the density in-
minimum of the average Nusselt number was obtained at 8°C of
version of water in the gap. Especially in such a case that the size
the temperature difference. This discrepancy must be due to the
of these eddies is nearly the same (6°C < T „ < 7°C), it may be ex-
fact that the heating area of the outer cylinder is larger than the
pected that Nu takes a minimum value.
cooling area of the inner cylinder in the present study. Further-
The analytical studies reported by Watson [5] and Seki, et al. [6] more, this is clearly indicated in Fig. 11 which shows the relation
between Nu and AT (= T0). In this figure, it is evident that Nu
does not increase monotonously with increasing AT as on common
fluid. That is, when a single large eddy occupies the major portion
of the gap as in the case of T0 < 4°C or T0 > 9°C, Nu increases
monotonously as T0 becomes higher. But, a peak of the average
Nusselt number can be recognized at about t0 = 4°C and the mini-
mum value appears between about 6 and 7°C. It should also be un-
derstood in Fig. 11 that the magnitude of average Nusselt numbers
increases in general as the gap width L increases.

Conclusion
From the aforementioned results, it can be concluded that the
flow patterns of the water in the gap are variously changed on ac-
count of the density inversion, that is, the density inversion has an
important effect on natural convection heat transfer in the annular
gap. Especially in the case when two counter eddies of approxi-
mately same size coexist in the gap, the average Nusselt number

Fig. 9(a) Temperature profiles, d„/d, = 3.44 and T0 = 8°C

"'A, l (m m) i
0 6.39 51. 2 1
O
« 3.20 <l. 7
Pure conduction 30
0 3.44 23.2
0 1 .72 17.7
— Pure c o n d u c t i o n °f
1
1

20
1
i /
0/

/ •„
6 Nu, /
° ^ 0
Nu.,
10
9. ^ '

_o_ ^ - - r / o
--

5 10 15-c
4T ( = T.)
Fig. 9(6) Local Nusselt numbers of the inner and the outer cylinders, Fig. 11 Correlation of average Nusselt numbers and temperature differ-
do/di = 3.44 and T0 = 8°C ences

560 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 indicates a minimum value and the minimum Nu increases as the the Melting of a Horizontal Ice Layer," Trans. JSME, Vol. 40, 1974, pp.
gap width increases. 3155-3165.
5 Watson, A., "The Effect of the Inversion Temperature on the Con-
The effect of the curvature of the outer cylinder on the flow pat- vection of Water in an Enclosed Rectangular Cavity," Quarterly Journal of
terns is observed only in small gap widths (L < 10.9 mm) in this Mechanics and Applied Mathematics, Vol. 15,1972, pp. 423-446.
study. 6 Seki, N., Fukusako, S., and Sugawara, M., "Natural Convection Heat
Transfer With Density Inversion of Water by Vertical Opposite Walls in a
Closed Rectangular Vessel," Preprint of JSME—Meeting of Heat Engineer-
Acknowledgments ing, Paper No. 740-17, 1974, pp. 87-90.
The authors are grateful to Prof. M. Arie, Hokkaido University, 7 Liu, C , Mueller, W. K., and Landis, F., "Natural Convection Heat
for a discussion of this subject and his encouragement, and to Mr. Transfer in Long Horizontal Cylindrical Annuli," International Develop-
ments in Heat Transfer, Paper No. 117, Part 5,1961, pp. 976-984.
K. Sawada for his assistance in the process of producing the exper- 8 Grigull, C , and Hauf, W., "Natural Convection in Horizontal Cylin-
imental apparatus. drical Annuli," Proceedings of the Third International Heat Transfer Con-
ference, Paper No. 60, Vol. 2, 1966, pp. 182-195.
9 Powe, R. E., Carley, C. T., and Bishop, E. H., "Free Convective Flow
References Patterns in Cylindrical Annuli," JOURNAL OF HEAT TRANSFER,
1 Yen, Y. C , "Onset of Convection in a Layer of Water Formed by TRANS. ASME, Series C, Vol. 91, 1969, pp. 310-314.
Melting Ice from Below," The Physics of Fluids, Vol. 11, 1968, pp. 1263- 10 Mack, L. R., and Bishop, E. H., "Natural Convection Between Hori-
1270. zontal Concentric Cylinder for Low Rayleigh Numbers," Quarterly Journal
2 Yen, Y. C , and Galea, F., "Onset of Convection in a Water Layer of Mechanics and Applied Mathematics, Vol. 21, Part 2,1968, pp. 223-241.
Formed Continuously by Melting Ice," The Physics of Fluids, Vol. 12, 1969, 11 Scanlan, J. A., Bishop, E. H., and Powe, R. E., "Natural Convection
pp. 509-516. Heat Transfer Between Concentric Spheres," International Journal of
3 Sun, Z. S., Tien, C , and Yen, Y. C , "Thermal Instability of a Hori- Heat and Mass Transfer, Vol. 13,1970, pp. 1857-1872.
zontal Layer of Liquid With Maximum Density," AIChE Journal, Vol. 15, 12 Kraussold, H., "Warmeabgabe von Zylindrishen Flussigkeitsshichten
1969, pp. 910-915. bei Naturlieher Konvektion," Forsch. a. d. Geb d. Ingenieurswesen, Vol. 5,
4 Sugawara, M., Fukusako, S., and Seki, N., "Experimental Studies on Series 4, 1934, pp. 186-188.

Journal of HeaJ Transfer NOVEMBER 1975 / 561

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 T. C. Chawla
G. Leaf The Application of the Collocation
W. L. Chen
M. A. Grolmes Method Using Hermite Cubic
Argonne National Laboratory,
Argonne, III. Splines to Nonlinear Transient
One-Dimensional Heat Conduction
Problems
A collocation method for the solution of one-dimensional parabolic partial differential
equations using Hermite splines as approximating functions and Gaussian quadrature
points as collocation points is described. The method consists of expanding dependent
variables in terms of piece-wise cubic Hermite splines in the space variable at each time
step. The unknown coefficients in the expansion are obtained at every time step by re-
quiring that the resultant differential equation be satisfied at a number of points (in
particular at the Gaussian quadrature points) in the field equal to the number of un-
known coefficients. This collocation procedure reduces the partial differential equation
to a system of ordinary differential equations which is solved as an initial value problem
using the steady-state solution as the initial condition. The method thus developed is
applied to a two-region nonlinear transient heat conduction problem and compared with
a finite-difference method. It is demonstrated that because of high-order accuracy only a
small number of equations are needed to produce desirable accuracy. The method has
the desirable characteristic of an analytical method in that it produces point values as
against nodal values in the finite-difference scheme.

Introduction ment methods based on the use of a Galerkin or a weighted residu-

al procedure can achieve high-order accuracy thereby substantially
Various numerical methods have been devised for solutions' to reducing the size of the numerical system. Hence, the data han-
the problems of transient nonlinear heat conduction. The most dling requirements are less severe and multistep time-differencing
common of these are finite-difference methods and finite-element methods can be used in order to increase the size of the time steps.
methods using Galerkin or weighted residual procedures. Finite On the other hand, for nonlinear problems, these methods require
difference methods are usually of low order and therefore require the computation of integrals at each time step. This in turn
the solution of large systems of equations in order to achieve a sat- implies considerably more arithmetic at each time step when com-
isfactory truncation error. A large system of equations implies pared to low-order finite-difference methods. Thus, in spite of
large storage requirements and consequently time differencing their greater accuracy, these higher-order Galerkin procedures
must be done with single time step methods whether explicit or when applied to nonlinear conduction problems may not be sub-
implicit. In an explicit method, due to stability conditions, the stantially faster than finite-difference methods.
largest time step that can be taken is limited. On the other hand, A third method is based on collocation combined with the use of
an implicit method will allow larger time steps, but only at a higher suitable approximating subspaces. For nonlinear problems, this
cost per time step. In either case, substantial machine times are approach has an advantage over Galerkin type procedure in that
encountered in solving such large systems of equations. there is much less arithmetic at each time step. On the other hand,
In contrast to low-order finite difference methods, finite-ele- it has the disadvantage, that unless the collocation points are suit-
ably chosen, this method produces low-order accuracy regardless
of the nature of the approximating subspaces. This implies large
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer systems of equations in order to obtain suitable accuracy, just as in
Division April 30, 1975. Paper No. 76-HT-B. the case of low-order finite-difference methods. However, as shown

562 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 9 7>
0 at r = Ra (2a)
dr
dTF
-Kr ~ff = he[TF(Rf,t) - Tc(RF\t)], at r = RF (26)
97
K,F 'E- = Kr ^ - at r = Rr. (2c)
dr dr
dT
-Kc^f = h(l)[Tc(Rc,l) - TNa(t)\ (2d)

Here, Tp = 7V(r, t) refers to the temperature distribution in the

fuel while Tc = Tc(r, t) refers to the temperature distribution in
COOLANT the clad. The temperature of the external medium (sodium) sur-
Fig. 1 Cross-sectional view of a fuel pin showing the knot sequences rounding the clad is denoted by 7V„(t). The definition of the other
symbols used is given in the Nomenclature.
The foregoing equations can be simplified somewhat by applica-
tion of Kirchoff s transformation
in references [1-4] , l the use of orthogonal collocation points such
e=
as the Gaussian quadrature points can provide as much accuracy
as a Galerkin procedure. Hence, in this case, the advantages of a
if CK(T)dT (3)
high-order Galerkin-type procedure are achieved, namely a small Upon the introduction of the variable 6 into equations (1) and (2),
system of equations and the use of multistep time differencing. In we obtain:
addition, like finite-difference methods, it requires only a small
amount of arithmetic at each time step. In this paper, we demon- at ~ " V 2 +
r ar} + Kn (4)
strate an application of this method to the nonlinear heat conduc-
98F
tion equation where nonlinearity is introduced by requiring the = 0 at r = R„ (5a)
transport properties to be a function of temperature.
90*
'KOF -jf = h^T^d^Rp^t) - Tc(ec(RF\/)}, SLtr = RF (56)
D e s c r i p t i o n of t h e P r o b l e m
For this application, consider a single fuel pin for a nuclear reac- 90g
K F K„ at r = Rf (5c)
tor. This pin consists of a hollow cylindrical solid of mixed oxide « dr dr
fuel encased in a cylindrical sheath of stainless steel cladding. 96,
-KB = h(t)[Tc(ec(Rc,t))-TNtt(i)} (5d)
There is a finite contact resistance between the fuel and the clad, dr
and the outer surface of the clad is cooled by convection. The fuel Here the diffusivity a = K(T)/p(T)CP(T) must be evaluated by
is headed by a uniform volume heat source which can vary with means of the transformation T = T(B) which is the inverse of the
time. A schematic cross-sectional view of such a system is shown in transformation given in equation (3).
Fig. 1. The figure contains dimensional nomenclature. If the vol-
ume heat source in the fuel is denoted by q(t) and if axial and cir- A M e t h o d of C o l l o c a t i o n
cumferential conduction terms are neglected, then the heat con- We shall seek an approximate solution of equations (4) and (5)
duction equation can be written in cylindrical coordinates as using the method of lines in conjunction with collocation. To this
9T end, we assume that at each fixed time, the transformed tempera-
p{T)Cp(T)
~9T l^wv^) + <?w (i) ture S(r, t) can be approximated by a function which is a piecewise
cubic polynominal in r and the function together with its first de-
When the foregoing equation is applied to the clad, the heat source rivative is continuous. More specifically, let [a, b] denote either the
term is set equal to zero. The boundary conditions which must be fuel region [Ro, RF] or the clad region [RF, RC]- Let this interval be
satisfied are: subdivided by the set of points

7r: a • xx < x2 < . . . < xmi = b, hj = Xj - Xj.x

Relative to this partition w, the approximating subspace Hs(w) will
1
Numbers in brackets designate References at end of paper. consist of all functions f(x) such that:

•Nomenclature.
A = coefficient matrix
a = inner boundary r = radial coordinate &ij = Kronecker delta function
b = outer boundary rj = knot points in fuel iliji = collocation points in general and also
Cifi = defined by equations (20) Rc = external radius of clad in fuel region
Cp = specific heat RF — external radius of fuel 0 = Kirchoffs transformed temperature
F;,t = defined by equations (18) Rj = knot points in clad defined by equation (3)
h = coolant heat transfer coefficient Ro = internal radius of fuel {,> = collocation points in clad region
hg = gap heat transfer coefficient T = temperature xc = mesh for the clad region
hj = step length in general and in fuel re- TN0 = coolant temperature TTF = mesh for the fuel region
gion t = time p = density
Hj = step length in clad region Xj = knot points in general •Mr) = Vj(r) + \lrVj(r)
K(T) = thermal conductivity Sj(x) = basis function defined by equation Wr) = S"j(r) + l/rS'j(r)
Ko = thermal conductivity at reference (66)
temperature To Vj{x) = basis function defined by equation
N = number of intervals in clad region (6a) Subscripts
n = number of intervals in fuel region a = thermal diffusivity C = clad region
q = rate of power generation per unit-vol- jS, y = constants defined by equations F = fuel region
ume in fuel pin (15a) i, j , k = indices

Journal of Heat Transfer NOVEMBER 1975 / 563

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 f(x) is equal to a cubic polynomial in each subinterval [ *,-, Now
i,+i] for 1 < i < n; n+i
2 f(x) and f'(x) are continuous at the points x; for 2 < i < n; | f = B[V/{r)8F(rpt) + SVtfMr,,/)]
and
3 /(x) satisfies the appropriate boundary conditions in equa- Since Vj(n) = 0, S[(ri) = 1, and S}(n) s= 0 for ;' > 1, we find that
tions (5) depending on whether the interval [a, b] is the fuel the boundary condition (5a) implies
or the clad.
Since this problem has two regions, we will generate two approxi- 6An,t) ^ o (9a)
mating sets of functions H f V f ) and Hic)(Trc), one relative to a Differentiating this expression with respect to time, we find also
partition irp of the fuel, the other relative to a partition TTC of the
clad. These two sets of functions are required to satisfy the com- 8F'(rut) = 0 (96)
mon interface conditions (5b) and (5c).
Similarly, we find the following relations at r = r„+i from the
A convenient basis for generating either set of approximating
boundary condition (56).
functions is the set {Vj(x), Sj(x)}"=i, where
2
[(1 + ^-—J) [ l - 2 ( ^ — i ) ] for xM £x£Xj •"•OF
h, hj
- Tc(8c(Rut))] = Ff (9c)
V,tx) = < ( 1 -JLZJLL)t [1 + 2 (•x^ - i x) tJ for Xj l
7+t (6a)
V(r» + ,,0 = - TA T M ^ X M ^ , , 0) - r c («c(«i.^))]
•• 0 e l s e w h e r e 0F

for Xj.-i
fc," «jr(»"nri»0 j f
flF(r„tl,t) 8c(J?1,t)
s,W = • * — *.) for
ify+it (1 JVJ =£ x £ xM (6b) x ec(Ritt)] = fF + AF8F(rn,ut) + Ac8c(Rvt) (9d)
ft.
* 0 elsewhere For the clad region (see Fig. 1), we find
It is assumed that Vi(x), S\(x) vanish to the left of Xi and N*t
Vn+i(x), S„ + 1 (x) vanish to the right of xn+i. From the defining
%'*> = S [viW&e(fli»0 + s,Me0'{R,,t)]
equations, we observe that each of the functions has the following
properties: N+l
1 Each Vj(x) and Sj(x) is continuous together with its first de- = ac(ec(r,t))Blec{Rpf)4>j{r)
rivative on the interval [a, b].
2 Each Vj(x) and Sj(x) is a cubic polynomial in each subinter- + flc'tR/f *)tfp,(r)] for Rt £ r < RKtt (10)
val and they vanish outside the subinterval [XJ-I, xy+i].

[3] V,{xj) = 6u, IVU,) = 0J 8C'(RU t) = %* BF>(rn,u t) = g*F, (11a)

i,J n + 1
Sitxj) =0, Sj'OO = 6,s\ A
oc
1
The set |V,(x), SKx))"^ will form a basis for the set of functions
Hs(ir). Thus we assume that the transformed temperature field has = Al?Uilfr + AFeF(rn,x,t) + Ac8c(Rut)] (116)
oc
the following form
n+l
e c ' U W ) = - ^ - [ r c ( 0 c t f W ) ) - W ) ] = FC (He)
e(x,t) =E [e(xJ,t)vJ(x) + e'(Xj,t)sj(x)] (7) A
oc
where \B)xj,t), 0'(xj,t)}"=i are the unknown coefficients and in view h'(RN,ut) =-~{Tc{ec(R^,t))- TNa(t)}
A
of property (3), these coefficients represent, respectively, the un- oc
known temperature and spatial derivative of temperature at each
of the knot points (XJ, 1 < j < n + 1). fr 0o l dec «C ( R JVtl' ( >
Let
or
T!F: R„ = r , < r 2 < . . . < r„ + , = RF
h'(R>M,t) =fc+ BjciRu^f) (lid)
be a partition of the fuel, and
Here, in equations (9d) and ( l i d ) , we have made the following
7TC: RF ~ R\ <C R% <C . . . <C RR*\ — RQ substitutions
be a partition of the clad. We seek an approximation to the tem-
perature field 8(r, t) of the form shown in equation (7) in the fuel fF =-£*-[Tr(8F(r„.um
K
6F
and the clad. Substituting equation (7) into equations (4) and (5),
we obtain the following relation in the fuel dT
Tc(8c(Rut))], AF=~p *
v 0 F aoF e W
F n*\'*'>
dT
et c
- K C
,fc
n+i " K0F dBc ec(Rvt)
= a/(eir(r,i)){S[ei,(rjJf)^(r) + eF'(rj,t)^,j(r)}}
= -~{Tc(8clRN.ut)) - TNa(t)} + ~~- fNa(t)
+ a^BjrirJVqW/Koriorri r„.i (8)
h dTc
where

0 . ( r ) = V/'(r) + lv/(r),4>j(r) = S/>(r) + -S/(r)

Substituting equations (9) into equation (8), we obtain

564 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 « _ ^ vj.i'(rljJ = -v/(iljil) = -6Al-y)/h]= -vr'/hj
+ [V„i (r) + Sml(r)AF]B^r^t) + SMAr)AcBc(Rt,t) SM'hiA) = S/fo > f 2 ) = 1 - 40 + 3/3* = V '

= -Swl(r)/F + " ^ ^ ^ + aF{BF{r,t)) V * " ' ^ = S / f o i . i ) = 1 " 4y + 3 y ' = S/

V,-,"foi.i) = V / ' d b > 2 ) = - 6 ( 1 - 2j3)/fe/ = - V / ' / V
x ( £ ^ ( r ^ r t * j ( r ) + E eF'(rj,t)ipjM Vj./'ivjJ = V y ' f a , ,) = - 6 ( 1 - 2y)//z/ = - K , " / l i /
i=l i=i ' (16c)
+ i ^ „ » ) for r , ^ rwl (12) < W % . . ) = -S/'fe?J>2) = - ( 4 - 60/ft, = - V ' A ,
Substituting equations (116), ( l i e ) , and ( l i d ) , into equation (10), S,-,"vl., j 2 ) = - S / ' ^ j ) = - (4 - 6y)/fy = Sr"/hj
e a m
° Evaluating equation (12) at the collocation points rntk in the fuel
f r „ , ,; ,„ A , „ M ; ,,„ ,M and noting that each Vj(x), Sj(x) has its support in [xj-r, xj+1]
L, [Vj(r)ec{R]tt) + Sj(r)ec (#,.,/)] and .n p a r t i c u l a r
F (r)
+ (Sl(r)Ac f*- + Vx{r))ec{Rut) ' = ° f ° r ^ [ r " r s ) ' V^W) = S"*<(r) = ° f ° r r
^»,r„+1]
oc
. we obtain the following system of equations:
For
+ (BcS^ir) + VNtl(r))ec(RNtUt) + Si(r)AF ^ e^r^t) ' = 2'fe = X
«- e - w >
F i f a z . i ) ^ , * ) + V2(rilt x)el.r2,t)

-/jrSiWft - / e S ^ W + ac(ec(r,f))
+ S2h]2A)6F'(r2,l) = F2A (11a)

x(S e c (fl i ( f) * , ( r ) + S e c ' («,,,) 0 , ( r ) Fo r i - = 2 , ^ 2 ( i . e . , , 2 , 2 )

J=1 J=2
w vx{n2i2)8Fb\,i) + F 2 (n 2 , 2 )Mr 2) /)
+ S tic'(Rj,l) ^-W + « , ( r ) ^ | j £ + QMMFC) + S2bl2f2)8F'(r2,t) = f2>2 (176)
In
for Rx < r < i?„ +1 (13) S e n e r a l for 3 < i < n, 1 < /e < 2,
The unknown coefficients contained in equations (12) and (13) are: F J . J ^ J ^ e / r , . ! , ^ ) + Vi{r]i ^Spir^t) + S^fy^^B/iri^,/)

{B^MJ-J, {eAriMrt, {OCIRJMM and {dc'(Rj,l)}^ + S ^ , , ) 0 / ( r , . , / ) = Fitk (17c)

which add up F to 2n + 2N unknowns. In order to obtain 2n + 2N — • • 1 1 ^ L .- <w \
.. . ., . .. ,10. , .. ,. , . „ F o r i = n+ 1,1 <k <2(i.e., Vn+i,k)
equations, we can require that equation (12) be satisfied at 2ra
points in the fuel region and equation (13) be satisfied at 2JV „ / w / A <., 1 \ , c 1 \A in 1 A
points in the clad region. Since there are n intervals in the fuel re- " " " ^ I . A W V + l> n + lW„ + 1 , s ) + ^ i W „ t l , ^ A r J M r „ + 1 , / )
gion and N intervals in the clad region, it therefore seems natural + S„(r]mltk)8F'(rn,l) + Sml(r)Mlik)Ac6c(Rx,l) = Fmlik (17c/)
to locate two points in each interval. As recommended in refer- wnprp
ences [1-4], the Gauss-Legendre quadrature points of order 2 are
chosen as the collocation points in each subinterval [x,, xi+i]. F2,k= a^F{y)Mtt))(6F(ru0 <^>M2J + BF{r2,t)^2(v2,l)
Thus, the collocation points are given by „ ,, ,. , . •/,%/ N, ^ , „ ,* „ N
+ BF'(r2,t) ip2(v2,k) + q(t)/K0F) for 1 < k £ 2 (18a)
1 , ^ . , ^k.hi

l1
-£ k k~<2 2 ((14)
14)
+ B/ir^t) iPi.fyiJ + 8F'{ritt) 0,Gj,>t) + g W / / ^ )
It is demonstrated by Shalev, Baruch, and Nissim [1, 2] that the
or
use of above collocation points yields a residual error for the differ- - — * — "> — « —- '••""'
ential equation which satisfies the "least squares" principle. Con- Fmi>k = -SmfymUk)fF + o ^ M ^ i , * / ) ) (Ff $n*Mm\J
sequently, the accuracy from a collocation method can be compa-
+
rable with that from a least square method. M ' r „ , ' ) * n ( ? ) n+ i,ft) + 8F(rmX,t) * n + i(7j n t l t f t )

! ! 1 1 , + ^ ' ^ n . ^ *nO?*l,ft) + ? W / ^ O F ) (18C)

/3 = — (1 — - 7 T ) and y = x (1 + 7 ? ) (15a)
6 ^ 0 £ VJ jt m a y jje noted that since the collocation points do not include the
in equation (14), we obtain end points, the value of r\ = Ro can be zero without incurring any
7),-, j - Xj T) ,--2 - Xj \ singularity in the system.
jl. ~ P> il. ~ Y I Similarly, evaluating equation (13) at the collocations points
_ _ > (156) which are denoted by £,-,* and given by equation (14) for the clad
J
'l •z=^- = (3, j ' 2 —1=^- = y \ region, we obtain the following system.
tij lij )
Using these relations in equations (6) give the following expres- [ s 1 ( | 2 ) f t ) A c —Q£ + VA%2ik)] 6 c(Rut) + V2{^iik) B c(R2,t)
sions for the values of Vj(x), Sj(x), and their derivatives at the col- oc
location p o i n t s , , , + ^ ^ ,{R^ + ^ ^ K^ ^ ^
Vj-iiVj,,) = VjimJ = (1 - /3)2 (1 + 2/3) = VA K c
«
1 2 C H H 2 (19tt)
^•-1^,2) = vMui) = (1 - r ) (1 + r ) = vrI = >>*
2
SJ-MJ.J = Sjbis,,) = hfid - j3) = hfr ( ^.i(««.*)ec(«,.i.fl + ^(«i,*)ec(Ri t) + sauje^iR^t)
2
S>i(T/i,2)=-Si(r)i,,) = V ( l - y ) =^S r / + S^^Sc'to,,/) = C l i t for 2 < * < JV, 1 < fe s 2 (196)

Journal of Heal Transfer NOVEMBER 1975 / 565

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 boundary conditions (5) and an initial uniform heat source q(t =
0). The effect of step size or the number of knots placed in fuel and
clad regions on the accuracy of the solution was investigated. A set
of five calculations were made. In the first calculation, 2 knots
= C / M . J . for 1 < k £ 2 (19c) were used in the fuel and 2 knots in the clad. Since 2 knots corre-
where spond to one subinterval, the temperature distribution in this cal-
culation was approximated with one cubic polynomial in the fuel
and another cubic polynomial in the clad. The second calculation
•"•oc
used 3 knots in the fuel, while the fourth calculation used 4 knots
+ ecOW) *2(^2,,) + ec'(Rut) q&tJ in the fuel with both calculations using 2 knots in the clad. The
fourth calculation used 5 knots in the fuel and 3 knots in the clad.
+ ^^jFf^-] l < i < 2 (20a) The last calculation was made with 20 knots in the fuel and 5
knots in the clad. The positions of knots chosen both in fuel and
c,,» = ac(M5J,*,f))(ec(ft,.1>fl *i-i U<,*) + e ^ . f ) *, (?,,») cladding for the foregoing five examples are nonuniformly distrib-
uted. In the choice of these positions, no attempt was made to ad-
+ 6fc'tRf-t,*) *,.,(«,,*) + ec'(Rt,t) *,(£,,»)) just these positions in order to improve the accuracy. It should be
for 2 < * £ iV, 1 < ft < 2 (206) noted that it is not the selection of the knot positions which pro-
vides high order accuracy for this collocation procedure, but rather
CJM,, = -/ c s M (^ +1 , fe ) + dclecf^M.slXec^ji.^^w,*) the choice of Gaussian collocation points relative to each subinter-
val determined by these knots.
For purposes of comparison, the solution to this problem was
+ 0Jir+i(€jf*i.*)-P'c) for 1 < ft s 2 (20c) also approximated by a finite difference type procedure as used in
We note that the transformed temperature 0(r, t) must be evalu- the T H T B program [6] which is a general purpose transient heat
ated at the collocation points in order to evaluate the property conduction program. In this method, the spatial interval is subdi-
functions such as the diffusivity a. For this purpose, we use equa- vided into subintervals (called nodes). For one-dimensional prob-
tion (7). For example, in the fuel at the collocation point i);,*f (x,-_i, lems, the spatial approximation to the conduction equation is es-
XJ) we have sentially the same as the usual three point approximation to the
second derivative and is derived by means of an integral heat bal-
eFbii,k,t) = eF{rt.ut) tVitoi.*) + M^.flVifai.*) ance over each node. The time derivative is approximated by im-
plicit single time step difference procedure. The implicit time dif-
+ eF,(rl.1,t)st.1fy,,J + e/lrt.fiStbu.k) (21) ferencing leads to a linear system which must be solved at every
time step. This system is solved iteratively by means of the point
Substituting for the values of the basis functions and their deriva-
Gauss-Seidel iterative method. With this program, 4 calculations
tives from equations (16) into equations (17) through (21) and rew-
were made. The first calculation used NF = 4 and Nc = 2, where
riting the resulting equations in matrix form, we obtain:
Np denotes the number of nodes in fuel and Nc denotes the num-
ber of nodes in clad; the second calculation used NF = 6 and Nc =
[A]{b}=G{U,t) (22)
2; the third calculation was made with Np = 35 and Nc = 6; the
fourth calculation was made with NF = 80 and Nc = 15. The
where U(t) is the In + IN dimensional vector
choice of nodal boundaries as used in the foregoing four calcula-
tions were nonuniformly distributed. In addition to the effect of
{eF(rj,tm u{eF'(r},t)}%, uie^RjMfJi1 u{ec'UV)}& spatial mesh size, the effect of the size of the time steps was also,
and G(U, t) is the vector of the entire right-hand side. The details investigated. For this purpose three time steps, namely, 0.01,
of above matrix equation are shown in Fig. 2. From an examination 0.005, and 0.001 s were used with this finite difference program.
of Fig. 2, it is clear that the coefficient matrix [A] is a band matrix. In Table 1 we have compared temperature distributions as cal-
This property is a consequence of the basis functions Vj(x) and culated by the two methods. In this table, the time is 4 s after the
Sj(x) having local support. It should be noted that in general, the initiation of the power transient. The six positions at which the
coefficient matrix [A] will depend on the temperature because of temperature is shown in this table are the node positions corre-
the convective interface and boundary conditions. This depen- sponding to the case of Np = 4 and N c = 2 for the T H T B calcula-
dence is explicitly shown in equations (9d) and ( l i d ) . Thus, the tion. In the case of the other three calculations, the subintervals
matrix [A] has to be inverted at each iteration of every time step. were selected such that the original nodal positions corresponding
The band structure is utilized in this inversion process. to the case of NF = 4 and Nc = 2 were included among these
With given initial values |£/(0)}, the system (22) represents a sys- nodes. Thus, the temperatures for the T H T B calculations shown
tem of nonlinear ordinary differential equations. This system is in this table did not have to be interpolated. For the method of col-
solved with the ordinary differential equation subroutine GEAR location, the temperature at these six positions is calculated by
[5]. This ordinary differential equation routine was selected be- means of equation (7) which provides an interpolation procedure
cause of the modest size and the nonlinearity of the system. Thus, entirely consistent with the approximation procedure. We have
when this routine is used the time steps are automatically selected taken the T H T B calculation with NF = 80 and Nc = 15 corre-
in accordance with the error criterion and the algebraic nonlinear sponding to time step 0.001 s as the benchmark calculation for this
system which arises at each time step is solved efficiently by this table. The relative percent deviation from this benchmark is
routine. shown under the corresponding temperatures in the table. The
number of equations solved for each case both with the collocation
Numerical Results method and the finite difference method (note that in case of
The method of collocation as described previously was used for T H T B , the number of equations is equal to the total number of
the solution of the transient conduction problem for the cylindrical nodes) are also shown in this table. From this table, we note that
fuel-clad pin configuration shown in Fig. 1. The transport proper- for the same number of equations solved, the method of colloca-
ties of the materials and the dimensions of the pin are given in the tion is substantially more accurate than the finite difference meth-
Appendix. The transient was initiated by permitting the volumet- od. In fact, we see that the accuracy obtained with only 12 equa-
ric heat source in the fuel to increase exponentially with time as tions in the collocation method is comparable with the accuracy
specified in the Appendix. The initial temperature distribution obtained with 41 equations in the finite difference method corre-
was obtained by a steady-state solution of equation (4) with sponding to time step of 0.01 s. This table also shows the corre-

566 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 -h,Sr
"9>„t)
-h 7 S« 0
9f('2."
hjS,, Vr - h , S r 9F(r2,t)
h,S y V„ -h 3 S e 0 4 (r,,l)
0 V<, h, S„ Vr - h , S r e;(fj,t)
V. h4 Sr Vfl - h 4 S .

h s
YD n-i% W i r

Vr h„.,S r V^-h n ,S s 0 9 F (r n .„t)

0 V, hn S„ V, - h n S, 9>M,t)
Vr hn Sr v„ - h n S, 0 9 F (r„,t)
0 V, h„4|S? V r -A F h M S r -hn„SrAc
9>„,t) F
n+i.i
v
r l>n,|Sr Vj-Afh^Sj -h n .|SjA c 0 F
n4i ,;

r W r f e ^ - f e ^

v
V
r -H 2 S r 9 c (R,,t] • = iC,

H
•^ft *S^ * -H z Sj 0 9'C(R2,I)

v9 HjS9 Vr -HjS,
9"c(R2,t)
V -HjS s
Vr HjS r * e c (Rj,t)
k.i V* H,SS 9c'(R,,t)
k-2 Vr H,S r -H«S. ec(R4,t)
§ 3

Vj H K .,S S Vr -HN.,S, 9C(RN.2,D

l.k.2 V, H N .,S, VS -HN.,Sj 9>W> CfH,2

lei V, HNS0 Vr -HNSr 9>N.„t)

k-2 Vr HNSr Vj -H S, 9C(RN,0
N CN,2

9>N,t)

I'"
0 V„ HNt,Se Vr-B,HKMSr CN+1,1

H
K+I Sr V9-BcHNt,Sj

k-3
Fig. 2 Display of matrix equation ( 2 2 )

5000

4500

4000

^-3500 -
UJ
cc
fe3000 -
UJ FINITE-DIFFERENCE METHOD:
i 2500 - 35 NODES IN FUEL, 6 NODES IN CLAD
UJ
r— 4 NODES IN FUEL, 2 NODES IN CLAD

2000- COLLOCATION METHOD:

2 KNOTS IN FUEL, 2 KNOTS IN CLAD
1500-
TIME = 8.0 sec
1000-
500
0.00 0.02 0.04 0.06 0.08
RADIUS, in.

Fig. 3 Illustrative comparison of a finite difference solution with the collo-

cation method at time = 8.0 s

Journal of H e a l Transfer NOVEMBER 1975 / 567

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 sponding Central Processing Unit (CPU) times on IBM 195/360
computer for the various time steps displayed in the table. It may
be seen that CPU time for the collocation method with 12 equa-
tions is an order of magnitude less than for the case of the T H T B
calculation employing 41 equations yielding the same order of ac-
curacy as the collocation method.
We may also note from this table that the collocation method
with 4 equations provides a relative error in the temperature less
than 1 percent which is more than adequate for many engineering
calculations. It may be noted that there exists a great deal of un-
certainty in some of the parameters such as gap conductance and
thermal conductivity of the fuel; thus these uncertainties can pro-
duce errors in the temperature which far exceed the errors due to
the collocation method. Therefore, the use of large number of col-
location points may be of doubtful value for this problem.
Fig. 3 shows the temperature distribution at time = 8 s illustrat-
ing the comparative accuracies for the two methods. In addition,
the figure shows the jump in the temperature at the fuel-clad in-
terface.

Conclusions
The application of this collocation method for this type of non-
linear parabolic equations shows that this method can provide a
very accurate numerical solution with a very small number of
equations. The method is far more accurate and faster than the
usual finite difference methods.

Acknowledgments
The authors wish to express their sincere thanks to George
Hauser for his invalualbe help in programming and the data prep-
aration. We also deeply appreciate the skill and patience shown by
Debbie Lambert in typing this manuscript.
The work was performed under the auspices of the U.S. Atomic
Energy Commission.

References
1 Shalev, A., Baruch, M., and Nissim, E., "Buckling Analysis of Hydros-
tatically Loaded Conical Shells by the Collocation Method," AIAA Journal,
Vol. 11, No. 12, Dec. 1973, pp. 1603-1607.
2 Shalev, A., Baruch, M., and Nissim, E., "Buckling Analysis of Elasti-
cally Constrained Conical Shells Under Hydrostatic Pressure by the Collo-
cation Method," T.A.E. Report No. 137, TECHNION-Israel Institute of
Technology, Nov. 1972.
3 Villadsen, J. V., and Stewart, W. E., "Solution of Boundary-Value
Problems by Orthogonal Collocation," Chem. Eng. Sci., Vol. 22,1967.
4 Finlayson, B. A., The Method of Weighted Residuals and Variational
Principles With Application in Fluid Mechanics, Heat and Mass Transfer,
Academic Press, New York, 1972, pp. 96-144.
5 Hindmarsh, A. C, "GEAR: Ordinary Differential Equation System
Solver," UCID-30001, Rev. 3, Lawrence Livermore Laboratory, Dec. 1974.
6 Stephens, G. L., and Campbell, D. J., "Program THTB, for Analysis of.
General Transient Heat Transfer Systems," R60FPD647, General Electric
Co., Apr. 1961.

APPENDIX
Properties of Fuel.
(0.0160528/(14.17943 + 0.01183T F ) Btu/(ft s°F)
KF = } 932 < TFtF) < 2552
(0.0003612 Btu/(ft s ° F )
pF= 651.2 l b m / f t 3
CPF = [12.54 + 0.0170T F (°K) - 0.117 x 10" 4 T^CK)
+ 0.307 x 10" 8 r / ( ° K ) ] / 2 6 9 . 7 6 6 4 B t u / l b m ° F )

Properties of Cladding.
Kc = 2.22 x 10"3 + 1.25 x 10"6 T C (°F) Btu/(ft s ° F )
p c = 485.26 l b m / f t 3
Cpc = 0.1105 + 2.632 x 1 0 ' 5 T C (°F) B t u / l b m ° F

568 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Heat Transfer Coefficients. Power Transient.
q{t) = q0 exp(0.1 t)
he = 0.2778 Btu/(ft 2 s°F)
where Qo is the initial steady-state power set at 45266.32 Btu/(ft3s)
ft= 6.944 Btu/(ft 2 s°F) and t is the time in seconds.

Journal of Heat Transfer NOVEMBER 1975 / 569

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 W. Contreras
Caloritech, Inc.,
Manhasset, N. Y.
Transient Melting of a Solid
R. S. Thorsen Heated by a Condensing
Assoc. Professor and Head,
Department of Mechanical Engineering,
Polytechnic Institute of New York,
Brooklyn, N. Y.
Saturated Vapor—Case :!::
Mem. ASME
Negligible Interface Curvature
A transient analysis has been performed to analytically determine the instantaneous di-
mensionless thickness of a vertically suspended solid at its melting temperature upon
which a saturated vapor is condensing. The spatial variation and transient behavior of
the melt-condensate liquid film, which is continuously drained, has also been obtained
from the closed form analytical solution of the uncoupled equations for the solid-liquid
and liquid-vapor interface motions under the condition of assumed negligible interface
curvature. Classical boundary layer assumptions were applied to the convection process
in the subcooled liquid film which was analyzed using integral techniques. From the an-
alytic solution to the resultant interface motion equations in terms of characteristic
curves, the functional dependence of the melting solid and the liquid layer thicknesses
on the time and space variables was established. The liquid mass flux and heat transfer
coefficient were then obtained analytically.

Introduction Analysis
Consider Fig. 1 in which an initially vertical symmetric slab of
The analysis of phase-change processes resulting from the inter- infinite depth and finite width and length, at its melting tempera-
action between an arbitrarily shaped solid at its melting tempera- ture TSL, is suspended and in contact with a stagnant pure vapor
ture and a condensing gas of the same molecular species has re- of the same chemical composition as the solid. Upon contact be-
ceived little attention in the past with the notable exception of tween gas and solid the vapor condenses onto the surface of the
Tien and Yen [l] 1 who recently, using a steady-state similarity so- solid joining the melt from the simultaneous melting of the solid to
lution and restricting the interface to remain vertical at all times, form a combined liquid film. This insulating layer of liquid, grow-
found the condensate-melt liquid film to obey the same relation- ing increasingly in thickness until a steady-state configuration is
ships as those developed by Nusselt [2] for the steady film conden- achieved, is continuously drained at the bottom. The regression of
sation of a saturated vapor on a vertical wall, except for the ap- the solid continues until it is completely melted at time tm.
pearance of the additional heat of fusion. The effect of the vapor
In general the same classical assumptions used by Nusselt [2]
layer and shear stress at the interface upon the liquid film was as-
and Rohsenow [3] in film condensation analyses are postulated for
sumed unimportant and the solution limited to Prandtl numbers
the condensation-melting problem. Equivalent assumptions have
equal or larger than one.
been recently used in reference [1] for a simplified steady-state so-
This paper deals with the transient and steady-state vertically lution. The assumptions used here are:
symmetric melting of an initially rectangular, two-dimensional 1 Stagnant, saturated vapor region.
solid at its melting temperature with its interface, mathematically 2 Subcooled, nonlinear liquid film temperature profile.
considered to remain vertical, in contact with a vapor condensing 3 Inertia terms throughout the liquid region are neglected.
continuously upon it. The liquid layer, assumed to drain contin- 4 Shear stresses at the liquid-vapor (L-V) interface are ne-
uously, is analyzed using integral techniques. glected.
5 The liquid film is of the boundary-layer type.
6 Solid at its melting temperature.
1
7 Condensate flow is laminar.
Numbers in brackets designate References at end of paper. 8 Constant physical properties.
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer In addition to the foregoing assumptions the solid curvature is as-
Division April 7, 1975. Paper No. 76-HT-L. sumed to remain negligible throughout the entire melting period,

570 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 In equation (8) care has been taken to explicitly identify the in-
dependent variables to be employed in the subsequent analysis.
Saturated The variable £, defined by equation (2), is introduced in anticipa-
Vapor tion of equation (22) for the liquid temperature profile.
In addition, the present analysis requires that Os(x, i), t) = 0
throughout the solid (assumption 6) and consideration of the ener-
gy equation for the solid .is therefore unnecessary.
(b) Boundary and Interface Conditions. In addition, the
following initial, boundary and interface conditions are to be satis-
fied.
At the solid-liquid interface (y = A or { = 0):
Saturated « £ L„ = 0 (9)
Vapor

= _A_
PL
IK (10)

Liquid Film
J
s l„=i U=o
(11)

80^
_ ( P*) (12)
Fig. 1 Schematic representation of two-dimensional condensation-melt- PL a S
L sL
ing problem

(13)
dt 6 8|
i.e., Ax — 0. Equation (9) follows from the no-slip condition at the solid-liq-
Governing Equations, Boundary and Interface Conditions. uid boundary under the condition of negligible curvature, while
The following dimensionless space, temperature and property vari- equation (10) expresses conservation of mass during phase change.
ables are introduced, preparatory to presentation of the governing The solid-liquid interface temperature is at all times equal to the
field equations and interface conditions: melting temperature TSL as expressed by equation (11) and con-
y servation of energy with the solid at its melt temperature requires
r){x,t) = 0 £ T] < 1 (l)
A(x, t) that equation (12) be satisfied.
y - &(x,t). Equation (13) arises by considering the expression for the total
; 0 < | < 1 (2) differential of 6L a t y = A or f = 0, i.e.,
6(x, t)
T(x,l,t) SL deL = — A dyy + —£• dt (14)
9(x,!;,t) = (3) dy at
In the absence of curvature and noting that at y A, dBL = 0 and
1*L> (4) dy/dt = At yields

TSJSi)if', = -A,
aeL
(5) (15)
at By
from which equation (13) follows.
(a) Liquid Film Conservation Equations. For the boundary At the liquid-vapor interface (y = A + 5 or £ = 1):
layer liquid film the continuity, momentum, and energy equations
then take the form-
= 0, (16)
(=1
1 dVr
+ -=- 0 (6) = 1 (17)
dx 6 9|
d2Ur
(7) -[(*» + «t)6, (18)
H2 -{PL
2
+ UT £1 + ILL d®L\ __ "J,
aL " d"T,
eL> \ (o\ Equation (16) results from assumption 4 stated earlier and
dt '*'y dx 6 3£ ~5T~dlT equation (17) states that at the liquid-vapor interface the tempera-

J^omenclature-

g = gravitational acceleration v = y component of velocity { = dimensionless coordinate defined by

h = heat transfer coefficient x, y = coordinates defined in Fig. 1 equation (2)
k = thermal conductivity a = thermal diffusivity p = density
I = initial height of solid 5 = liquid film thickness measured in y-
Mc = condensate mass flux direction Subscripts
Mm = melt mass flux A = solid half-thickness L = pertains to liquid
Pr = Prandtl number, {^Cp/k) Ao = initial solid half-thickness LV = at liquid-vapor interface
SLV = defined by equation (5) rj = dimensionless coordinate defined by S = pertains to solid
SSL - defined by equation (4) equation (1) SL = at solid-liquid interface
t = time 6 = dimensionless temperature defined by SS = denotes steady state
tm - time required to entirely melt solid equation (3) t = a/at
T = temperature XLV — latent heat of vaporization x = a/dx
TLV = temperature of saturated vapor ^SL = latent heat of fusion
TSL = melting temperature of solid M = dynamic viscosity Superscripts
u = x component of velocity v = kinematic viscosity ( ) ' = denotes dimensionless quantity

Journal of Heat Transfer NOVEMBER 1975 / 571

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ture is at all times equal to the temperature of condensation, TLV-
a= {8 + ^ - [ ( 1 + 2SSLy»~ 1]} (31a)
Conservation of energy at the liquid-vapor interface results in •-> S i
equation (18) for the case of a saturated vapor.
!/2
Integration of equation (7) with respect to £ and use of equations b = | ( 1 - - ^ ) ^ { 1 8 + ^ - [ ( 1 + 2S S i l - 1]} (31*)
(9) and (16) results in
c = 48QI£{(1 + SLV)

v 5] (19)
PL h ^ - (1 + k s i )-^ f(l + 2S s t )V2 - 1]} (31c)
^ 'JSi
from which
Similarly from (24), (27), and (28)
1
~ - (1 - & ) 5 2 ^ (20)
l«»i ~- o u PL

Proceeding to integrate equation (8) between the limits £ = 0 and £

V
L
A, = ^ft* WT~zs^-
' S i
1
-} (32)

= 1, using equations (6), (11), (17), and (18), and applying Leibniz' Dimensionless Equations for t h e Solid-Liquid a n d Liquid-
rule it follows that Vapor I n t e r f a c e Motions. The following dimensionless variables
are defined:
_9_
dtL~[6/
J 0 V«] + dxL Yx^fSWft
J
o o
f = ^ f t; b' = 4-; A' = ~; and x' = -f- (33)
] (21)
3?
In terms of these variables equations (30) and (32) become, after
The functional form used in the integral procedure for establishing rearranging,
6L(x, £, t) is (34)
r , ' 6 ' 6 ' t , + r2'bnb'x, = 1
2
eL(x,£,t) =A(x,t) + B(x,M + C(x,t)£ (22)
AV = r 3 / 1.
y
(35)
Then from equations (8), (9), (10), and (22), for £ = 0, it follows
that The dimensionless constant coefficients are
2 + d'
8£t = [ ^ C + ( ^ - l ) ^ (23) r1 '
1 (364
3f Pi b 3£ 12 aL
(1 + £b) - (1 + fs i F )c/'
From (11), (13), (17), (22), and (23) then
T
18 + Id'
i = ^ ( l - ~pL) ^ aL
) * ' [ '
2ffl L (l - B) + (^)6AtB = 0 (24) (i + sLY
i r ;) -
- (i
\i +
+ ig-SjjyJd'
Pi
(366)
However, from equations (12) and (22)
r ,3' = 2 <fi*5fi.) ^ L l l l ; (36c)
A
^=-S]tM)B (25) psas d'

d' = ^ - [(1 + 2 S S I ) i ' 2 i] (36rf)

Then from equations (24) and (25) it follows that

B2 + K B (26) g*
f= «s^i
(36e)

where Equation (34) is recognized as an uncoupled, quasi-linear, parabol-

ic partial differential equation of first order, solvable by the stan-
R(x,t) = (27)
dard Method of Characteristics as described in Hildebrand [4], for
example.
in the present analysis with no solid subcooling. When restricted to the first quadrant of the t' — x' plane, the
The solution to (26) is given by single, real family of characteristic curves is described by:
R \/l df ry I
IR2 v (28)
S = ± (— + ) dx' ~ r y IP7 (37)
2 * V4 SSL'
The physical requirement that TL 5: TSL everywhere in the liq- db' 1 l
3
uid film dictates that the + sign in equation (28) is to be used. dx' iY 6' (38)
Thus, from (11), (22), and (28) we have that
db' l 1
ff2 1/2 R df = r,' 6' (39)
eLb,u) = [(£- + 3 %
Si
subject to the conditions
O 1/2 O
[(_ +
f-) ~% " I ] ? 2 (29) 0: t' 0 x' = x'
(40)
From (21), (27), (28), and (29), recalling (19) and (20), 6' = 0; x' = 0 V = t'

abt + b 5 2 5 x =-- g (30) From equations (37), (38), and (39) it follows that

where b'(x') = <£r x')Ui (41)

al)L\ b'(f) = 2 -Ml/2

(=^rf) (42)
2
It is to be stressed that at u = 0 is not evaluated from equation (22).
Recall that y and t are the independent variables of the problem and £ is in-
troduced for compactness in equation (22). The limiting characteristic curve, i.e., the curve passing through

572 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the origin of the t\ x' coordinates, gives rise to the steady-state so- Similarly
lution. Thus
&(x,t S tssj = A„ - a^'H (56)
1 / 2
tss.L' = r,'(^-) (43) where
a = \^L- (i _ Rx.)?]"1 fins
5 K
Then, from equations (41), (42), and (43) it follows that 10 p L '*> ps'

Sss' = (jW"*"" (44) JWTJ")^ - i - l l W + 7{a + 2 S „ ) " ' - l}]

2
Furthermore, from (42) ' t(l + S
^ + S
4 +
$sL) ~ « + ^ ( 1 + 2SSL)^]

s s i
' H e a t T r a n s f e r Coefficient. From the usual definition of the
li( uid film h e a t
From (43), (44), and (45), returning to dimensional variables, l transfer coefficient we have

(58)
where ^
/~Tfj 7, r Equations (29), (48), and (58) then give rise to
{2S + [(1 + 2S )1/2
J Tip -ate ^ ^ " 1 ] > P ^' / 2 ,
1/2
{ l 8 S s i + 7[(1 + 2 S „ ) i A - 1]} { 2 [ 1 + S ffi ) w h e r e

s 1/2 2
+ Si7d + &) - (1 + ^ d + 2SSL) ]}" r V(fa -Py)gCM1'/«
(47) 160 y r
Similarly x r{l8S s . r , + 7[(1 + 2 S S £ ) ' / 2 - !]}[(! + 2 S , „ , ) 1 / 2 - l ] V / 4

2 5
«<M 'ss.i) = ss = « 2 ^ / 4 (48) 2 S s i 4 [ ( l + SSI) + S L F ( | + SSL)
where , 1 ,. ,,,„.,

L
^ " (Pi-PvkJ (60)
2[(1 + S s i ) + S i r ( - + SSL) Similarly from (33), (45), and (58) we obtain
1/4
- (l + | s i r ) ( i + 2S„)1A] W4
}lL(x,t < /ss>i) = hsStL &*)*'* (61)
x {• fL } (49) t
2
18 S S i + 7[l + 2SSL) - 1] M a s s F l o w R a t e / U n i t Width. The mass flow rate per unit
and width is given by
b{x,t < tgs L) = <utUl (50) A
r = pL6JrL(X,U)^ (62)
t(l + SSL) + SLV(^ + Ssz) From (19), (48), and (62)
1/2 2
- (1 + | s i 7 ) ( l + 2 S S i ) } " T(x,t - tSSiL) = TSS= a,**'* (63)
- 51 w h e r e
«3 = V 2 4 a i { - —2S
^ ^+ fi
[1 _+L 2ocS s i )i/2
) " 2 _- i1]i 1 ( ) . . . , , „ ,
SL 1/4
1
_ r ( M
a" 7. ~~ L 3 v 3 ' ^ ^ ( ^ V
~ P v )g[
With a closed-form solution to equation (34), the solution to equa- Cpi? L
tion (35) can now be determined. 9[l -t- S )
SL
From equations (35) and (44) or (45) it follows that 3/4

r , , , ,/2 + H + SSL) - (1 + | s L F ) ( l + 2 S s i ) 1 / 2 ] ,
(64)
^•'bc>,t> < W ) = § 7 <*«*-) (52) ><{ i a s „ + 7 U l + 2S„)'*-l] }

^ , ' ( ^ ' £' > + 'N _ a |r„\ Similarly from (19), (50), and (62)
6
SS' £ 3
r < tss = r s s (
where F' 3 is given by equations (36c) and (36d). (*^ -^ 7^7^ ^
Proceeding to integrate equations (52) and (53) and returning to
dimensional variables results in Melt Mass-Flux. From equation (13) it follows that the melt-
mass flux is given by
A(x,t < taj = A0 - atti/2 (54)
d6
hS8L L (66)
where Mn 5CPL 85
= v x ai(^t) Then from (29), (27), (48) or (49) and (66)
>> Ps
[ ( i - # W r T - !] (2S«i + [d + 2S
S X ) 1 / 2 - 1])- Mj •= Sl^L_ [(1 + 2S Jl/2 _ l]x-l/4 (67)

2[d + SSL + S i ^ + S „ ) - U + l S i r ) t t + 2Ssi)p/2 ^ = p^ [ ( i + 2SSL)U2 _ i ] f l / 2 (68)

fl3
(55) I^SS.L

Journal of Heat Transfer NOVEMBER 1975 / 573

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Q533 1.066 £133 2.666 3.199 I.U
0.799 1333.10" 3 Z399 £933 3.466.1CP
0.2 0.4 0.6 Q8 1.0 1.2 1.4 1.6 1.8 2.0 2 2 2 4 2.6 \ X'=H20 Solid Height = | - , i a 2
0.9 - V 4=20 ft. (ft/ft)-
6 — Ao=lft
0.8 \ '

tl 0.7
V\\\ ^ s = 1.09 .
«i<r 0.6
"< : % -
£ I0"3

<ness
0.5
— Tien-Yen Similarity Solution (PL/PS=\0) • m •
0 0.4
v v \%
a Present Solution (pjps = 1.0)
: :

id Th
— Present Solution {pL/Ps
X ' = H 2 0 Solid Height ( f t / f t )
= 1.09)
0
co
0.3

0.2 4
WV-
Jl = 2 0 ft
A n = I ft.
* V\W "
0.1
10"'
N
, 1 , p , 1 , i ,\ S \ \i \
0
2.0 4.0 6.0 8.0 10.0.12.0 I4.0xl0 2
2.666.K)"5 5.333 7 3 9 9 10.666 033310 6.999 1&666.I0"5

0.1 0.2 0.3 0 4 0.5 0.6 0.7 0.8 0.9 1.0 I.I 1.2 1.3 Time t (sec) or '' = 'Tf"t
6
0.I33-I0' 0-399 0.666 0.933 1.199 .51.466 1.733
"o
0.266 0.533 0.799 1.066 1.333-K)0 1.599 Fig. 3 Solid region thickness variation with time for solid height V = %,
Time t (sec.) or t' = -^-f-1 1,2
"0
Fig. 2 Liquid film thickness variation with time for various solid height
locations

The liquid film transient is small compared to the solid melt-

ing rate.
Consensate Mass-Flux. From equation (18) The solid interface velocity becomes independent of time
after steady state has been achieved by the liquid film.
d6
MK = PlP&V L (69) The equations developed for the liquid film during the melt-
ing-condensation process parallel the relationships obtained
by others (e.g., references [3, 5, 6]) for transient and steady
It follows from (29), (27), (48) or (49) and (69) that
film condensation only, except for the appearance of the heat
of fusion in addition to the usual heat of vaporization.
M = PfL^v {2SsL - [(1 + 2 S f f l ) 1 / i - 1]}*"1/4 Close agreement with the steady-state similarity solution of
a
l SsL
SS,L Tien and Yen [1] is indicated. The small differences repre-
(70) sent a measure of error due to the integral solution for the
liquid temperature distribution used in the present work.
M„ = - a¥ i | £ rS { 2 S s i - [ ( l + 2S S L )'"-l]}f- 1 / 2
t<t ss.i * "
(71)

Condensate-Melt Mass-Flux Ratios. From equations (67),

(68), (70), and (71) we have

a
c
Ms _ r %SSL
(72)

Results
An uncoupled solution has been obtained by reducing an ex-
tremely complex problem to a configuration which permitted ar-
riving at a closed form analytical solution in terms of characteristic
curves. This solution assumes A* - • 0 (negligible solid curvature)
and #o = 0 (solid initially at its melting temperature) and is shown
graphically in Figs. 2-4 for the case of water at one atmosphere.
Results reported by Tien and Yen [1] for their steady-state simi-
larity solution of the condensate-melt liquid film thickness of a
similar problem in which the solid and liquid densities are as-
sumed to be the same are also shown in Fig. 2 for comparison pur-
poses. Their values agree closely with the results obtained in this
investigation.
0.1 0.2 0.3 0.4 0,5 0.6 0.7 0.8 0.9
From the results obtained for both the liquid and solid regions,
the following general observations can be made: Solid Thickness A' (ft/ft) = - g -
• The liquid film does indeed reach a steady-state configura-
tion. Fig. 4 Solid profile variation with time

574 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 This small difference justifies the use of the integral solution The authors have undertaken and completed a study of these
technique in a more general problem involving interface cur- two effects (curvature and solid subcooling). However, because the
vature and solid subcooling effects. problem does not admit an analytic solution under these more gen-
° The effect of the density ratio, PL/PS (1-00 or 1.09 in the eral conditions the results are not reported here. They will be pub-
present study), is confined primarily to the solid regression lished in a subsequent paper.
rate and is not a significant factor in the liquid film growth
or steady-state configuration.
References
Conclusions
1 Tien, C, and Yen, Y. C, "Condensation-Melting Heat Transfer,"
Analytic results for a special case of the extremely complex Chem. Eng. Prog. Symposium Series, Vol. 67, No. 113,1971, pp. 1-9.
problem of combined melting and condensation of a solid in the 2 Nusselt, W., "Die Oberfiachen Kondensation des Wasserdampfles,"
presence of its vapor have been obtained. The assumptions of zero zeitschrift des Vereines Deutscher Ingenieure, Vol. 60,1916, pp. 541-569.
solid subcooling and negligible interface curvature together permit 3 Rohsenow, W. M., "Heat Transfer and Temperature Distribution in
Laminar-Film Condensation," TRANS. ASME, Vol. 78, 1956, p. 1645.
uncoupling of the equations governing the motion of the two inter- 4 Hildebrand, F. B., Advanced Calculus for Engineers, Prentice Hall,
faces. The results indicate that the solid-liquid interface, as ex- New York, Vol. 11,1955, pp. 368-378.
pected, cannot remain vertical and curvature must eventually be- 5 Sparrow, E. M., and Gregg, J. L., "A Boundary-Layer Treatment of
come important. As this happens and the top of the solid recedes, Laminar-Film Condensation," JOURNAL OF HEAT TRANSFER,
TRANS. ASME, Series C, Vol. 81, 1959, p. 13.
the validity of the present analysis must be checked against a more 6 Sparrow, E. M., and Siegel, R., "Transient Film Condensation," Jour-
complete, though mathematically more complex, analysis. nal of Applied Mechanics, TRANS. ASME, Series E, Vol. 26,1959, p. 120.

Journal of Heat Transfer NOVEMBER 1975 / 575

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 M.Toren
Y. Zvirin Melting and Evaporation
Y. Winograd 1
Phenomena During Electrical
Faculty of Mechanical Engineering,
Technlon—Israel institute of Technology,
Haifa, Israel Erosion
A modeling method is presented for the evaluation of the eroded crater during electrical
discharge machining. The method is based on the propagation of the melting crater
when other effects such as evaporation can take place. Calculations are performed for a
point heat source on the surface of copper and steel electrodes. An experimental tech-
nique for the measurement of the amount of eroded material is described: the method
consists of extraction of the metal from the dielectric fluid by chemical reactions and
measurement of the resulting color intensity.

1 Introduction glected or it is assumed that the surface temperature does not ex-
Electrical discharge machining (EDM) of metals has become an ceed the evaporation temperature. While these assumptions can be
established method for machining, especially for complicated justified for approximate calculations, a more accurate approach
geometries and high accuracy. The practical techniques of EDM must take into account the latent heat of evaporation, which is
have developed considerably in recent years. However, in spite of much larger than that of melting. Moreover, other phenomena
the progress made in the theory of EDM, many aspects of the pro- such as ionization and excitation of the evaporated material can
cess are still unexplained. The process of the discharge between occur, and the existing mathematical models cannot describe
two electrodes (tool and workpiece) immersed in a dielectric fluid them.
is very complicated and involves numerous phenomena, e.g., heat Problems of heat transfer with a change of phase are also en-
conduction and phase changes (melting, evaporation, ionization, countered in other fields, such as solidification of castings, design
excitation) of the electrodes and the dielectric fluid, electrical forc- of shields for re-entry vehicles on the basis of aerodynamic abla-
es, formation and collapse of gas bubbles and energy distribution tion, freezing of foodstuffs and sinks for energy storage. Boley [15]
in the discharge channel. Several very simplified mathematical and Muehlbauer and Sunderland [16] presented a thorough review
models for the erosion mechanisms have been suggested, but a of the literature published until 1963, which was limited to a single
generally accepted theory does not yet exist. change of phase depending on a single space variable. Since 1963
several authors treated two- and three-dimensional problems, but
Electrical forces exerted by the electric field between the elec-
only a single change of phase. Several of these works are men-
trodes can account for the erosion only for very short pulses—up
tioned by Budhia and Kreith [17], who solved the problem of melt-
to 10~ 6 s, [1-4]. 2 For longer pulses, the basic mechanism for metal
ing in a wedge.
removal is that of melting and evaporation [5-14]. The discharge
channel is the energy source and the amount of eroded material is Sikarskie and Boley [18] solved the problem of heat transfer
determined by the isotherm of the melting temperature. The with melting caused by a heat source which varies spatially along
mathematical models suggested [5-14] for this mechanism treat the boundary surface. They have also suggested a method for
the melting process by modification of the specific heat of the ma- treating the initial propagation of the melting front, based on the
terial and not by sinks of latent heat. Evaporation is either ne- assumption that the problem can be regarded as one-dimensional
during the initial stages. Pedroso and Domoto [19] suggested a per-
1
This paper is dedicated to the memory of Prof. Y. Winograd who was turbation method for solidification problems. They simplified the
killed in action during the Yom-Kippur War, October 1973, while defending problem by transforming the variables such that the position of
his country. the solidification front replaces time. Lazaridis [20] presented an
2
Numbers in brackets designate References at end of paper. implicit numerical solution for the temperature field and the posi-
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OP HEAT TRANSFER. Manuscript received by the Heat Transfer tion of the interface for a multidimensional solidification. He
Division September 17,1974. Paper No. 76-HT-J. treated, again, only a single phase-change.

576 / NOVEMBER 1975 Copyright © 1975 by ASME Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 This present work is apparently the first attempt to develop a For the former, it was assumed that the surface temperature does
method for treating a heat transfer problem with multiple phase not rise above that of evaporation; for the latter, constant and
changes. The method can serve as a first approximation for evalu- equal properties are assumed for the solid and liquid phases, and
ating the propagation of the melting crater during electrical dis- the specific heat capacity was modified by the latent heat to ac-
charge machining. count for melting, Snoeys and Van Dyck [14]. Cobine and Burger
The model is applied here to calculate numerically the melting [23] even assumed that all the energy supplied is absorbed by
isotherm resulting from a single spark of a point source on the evaporation.
electrode surface when melting and evaporation take place. These Observation of the behavior of the relevant physical properties
phenomena are first included accurately, by sinks of latent heats for metals, c.f., Smithells [24], shows that they do not change dras-
distributed on the melting and evaporation isotherms (Section 2). tically at the melting point. Some properties vary continuously
In Section 3 an approximate method is developed in which the through this point while the discontinuous change in other proper-
sinks due to melting and evaporation are "transferred" to the site ties is small. Since the variations of the properties between the
of the point source. The magnitudes of these sinks remain as in the solid and liquid phases are quite moderate and due to the fact that
exact model, i.e., the heats required to melt and evaporate the the crater is tiny, the assumption of constant properties seems rea-
propagating melting and evaporation shells. The results for copper sonable as a first approximation.
and steel electrodes show that the approximate model is sufficient- The properties of the vapor phase still remain a difficulty and a
ly accurate for practical calculations. This method is clearly sim- limitation of the model. However, since the electric current pene-
pler than the exact one and enables a considerable reduction of nu- trates through the vapor phase, most of the heat is applied at the
merical computation and thus of computer time. liquid-vapor front so that heating of the vapor and conduction in it
In order to develop these simple models which represent the are not important. The dominant effects are melting and evapora-
complicated process of the discharge, several assumptions need be tion, and only when the temperatures are much higher, ionization
made. The first one is that of spherical symmetry. It is well known, and excitation in the gas phase become significant. Moreover, the
c.f., Soneys and Van Dyck [14] and Greene and Guerrero-Alvarez thermal resistance of the vapor phase is also compensated for by
[21], that the crater formed by a single spark is a symmetrical seg- solid particles in the crater which are torn from the electrode sur-
ment, which tends to a half-sphere with increasing pulse duration. face during the first stages of the discharge, c.f., Zolotych [8].
This is explained by the following reasons: the diameter of the dis- The experimental part of the work included measurements of
charge channel is much smaller than that of the workpiece elec- the quantities removed from the electrodes during single dis-
trode, heat conduction into the dielectric fluid is negligible com- charges. One of the main problems of measuring the amount of
pared with that in the metallic electrode, and convection in the material eroded during a single spark is that this amount is very
vapor phase is not important until the collapse of the bubble which small. One method is based on measurement of the dimensions of
contains the vapor. It is believed, e.g., Zolotych [8] and Crookall the crater [6, 12, 13], but the shape of the crater becomes irregular
and Heuvelman [22], that some evaporation of the dielectric fluid with increasing pulse duration. Another method is based on aver-
takes place at the beginning of the discharge. The electric arc is aging the weight loss of the workpiece resulting from many sparks
then established in this gap, while more vapor, mainly from the [25-26]. In this way, however, every spark has a new configuration.
electrodes, is accumulated in a bubble. The bubble collapses only Determination of the erosion by radioactive tracers [27] gives rea-
after termination of the discharge (for pulse durations up to 1000 sonable results but was found to be inconvenient.
us, investigated here). Convection in the vapor phase can be im- The experimental method reported here (Section 4) for mea-
portant only then; it is neglected in the present work. This as- surement of the amount of eroded metals consists of chemical ex-
sumption is further justified by the results of Greene and Guer- traction of the metal from the dielectric fluid and determination of
rero-Alvarez [21]. They compared machining in a dielectric fluid its quantity by the intensity of colored solutions in a spectro-pho-
and in a gas. In the former case, a single crater is created while in tometer.
the latter, secondary craters are formed around the central main
The experimental results for copper and steel electrodes are
crater. This shows that the dielectric fluid prevents the spreading
compared with those of the theoretical model. Good agreement is
of the energy and hinders convection.
found for intermediate pulse duration; for longer and shorter puls-
A further assumption is that the properties are constant and es the theory can provide orders of magnitudes of material remov-
equal in all three phases. In most of the reported works on melting al. It is thus concluded that the simple model can serve as a first
or solidification the problem of the properties in the different approximation in spite of the assumptions and simplifications
phases is avoided. For example, Sikarskie and Boley [18] consid- which were being made.
ered instantaneous removal of the melted material and heating ap-
plied to the melting front. The same authors, and also Budhia and 2 P r o p a g a t i o n of t h e M e l t i n g a n d E v a p o r a t i o n
Kreith [17], solved problems in which the whole liquid phase is at Surfaces
the melting temperature. In order to accommodate the change in The basic mechanism of metal removal during EDM is based on
the properties between the phases, sophisticated and complicated melting and evaporation. The discharge channel is the heat source,
numerical methods must be used, e.g., Lazaridis [20], In the exist- assumed here to be constant and concentrated at a point on the
ing models for electrical discharge machining, evaporation and surface of the electrode. The crater of eroded material is deter-
melting are not included directly, as mentioned in the foregoing. mined by the isotherm of the melting temperature, see Fig. 1.

•Nomenclature.

e = special heat 8 = temperature, dimensionless

k = heat conductivity t = time T = time, dimensionless
I = length scale t* = time scale p = density
L = latent heat T = temperature \p = temperature due to a point heat
M = melting parameter V = evaporation parameter source, dimensionless
QM, QV - heat sinks due to latent heats of WM, wv = temperatures due to heat sinks
melting and evaporation, respectively of melting and evaporation, dimension- Subscripts:
Q = point heat source less M = melting
r = radius, dimensionless a = heat diffusivity V = evaporation
R = radius 5() = Dirac delta function 0 = initial time

Journal of Heat Transfer NOVEMBER 1975 / 577

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 point heat source
\p(r, T) = - e r f c — = (9)
r v r
wtiir, r;rM) = I 1/3 d/dr' rju3 dr' - | r - r j l ( r ' ) ] ! / ( r - r ' ) 4.

4TcrM(T') (r - r'Yn
.e-[r+rM(r')JV(r-r')j (1Q)

.and a similar expression for wv, obtained by replacing the sub-

Fig. 1 The melting crater script M by V in equation (10).
From (5), (9), and (10) it is seen that the temperature at any
time T depends on the history of the melting and evaporation
As mentioned in the foregoing, it is assumed that the physical propagation from the start (T = 0) to this time T. In order to evalu-
properties are constant and equal for the phases. The temperature ate the increase with time of the radii rM and rv, it is necessary to
field in the metal can be described then by the following Fourier solve two equations which follow from (5) with S = BM and 8 - 8V:
equation:
OM ~ \^(ru, T) - Mw\i(rM, T; TM) - Vwv(rM, r; rv)] = 0
eT
pe- - fev2T = QS(R) ~ qM(nM; t) - qv(Rv; t) (1) (11a)
st
Ov ~ [<p(rv, T) - MwM(rv, T; rM) - Vwv(rv, r; rv)] = 0
where Q is the intensity of the heat source and QM and qv are the
(116)
instantaneous heat sinks due to melting and evaporation taking
place at RM and Rv-' Equations (11) were solved numerically for T M W and rv(r). The
4 d n , computer program performs the numerical quadrature of the inte-
QM(RM; t) = - irpLM — Rtit (2a) grals which appear in (10), when TM(T') and rv(r') for 0 < r' < T
are known from the previous steps and rjw(0) = ry(0) = 0. The sys-
4 d tem of two simultaneous nonlinear equations (11) for rM(r) and
qv(Rv; t) = — irpLv — Rv3 (26)
rvU) is solved by iteratively minimizing the sum of the squared
3 dt
left hand sides of equations (11a) and (116). The method of solu-
Other phenomena like ionization and excitation can be described
tion can be applied to any metal; the solution in the dimensionless
by adding appropriate terms on the right-hand side of equation
form depends on the metal properties (via M and V), thus, a com-
(1). plete universal solution cannot be obtained.
The initial condition is taken as a uniform temperature To, just
Results for the propagation of the melting and evaporation
before application of the source: fronts for copper and steel are shown in Figs. 2 and 3. Obviously,
T = To at t = 0 (3) RM and Rv tend asymptotically to constant values as time be-
comes very large. Then a steady state is reached where all the heat
It is assumed that the heat loss from the surface of the electrode is from the source is conducted into the solid.
negligible (see Fig. 1); thus, because of symmetry, the problem re- The asymptotic radius of the melting isotherm for copper is
duces to a spherical one-dimensional one. The other boundary smaller than that of steel, because the product &TM is larger for
condition is: copper (see equations (6), (7), and (9)). For short times, the domi-
T finite as R — » (4) nant phenomenon is that of evaporation which requires the largest
part of the energy. The latent heat of evaporation is bigger for
which states that the dimensions of the crater are much smaller steel than for copper; therefore, the melting radius of steel is
compared with the electrode. smaller for short times. Moreover, this phenomenon also causes
Equation (1) for the temperature is linear, therefore, the solu-
tion can be composed of three terms, written in the following di-
mensionless form (cf., Carslaw and Jaeger [28], p. 293):
6(r, T) = f(r, T) - MuiM(r, T; TM) - Vwv(r, T; rv) (5)
where the functions \p, WM, and wy denote the temperature fields
10s I
resulting from a point source at r <= 0 and from heat sinks (due to
melting radius - exact
melting and evaporation) distributed on the propagating fronts R/im melting radius-approximation
nvf(f) and rv(r), respectively. evaporation radius-exact
evaporation radius-approximation
The dimensionless variables 0, r, and r are defined by:
a
6=
T - T Q R
; r = —;
L (6)
t*
TM - To I
where the length and time scales are:
O
^ . t* = . (V)
I 47r/e ( T M ~ T 0 ) 4a
It is noted that the length scale is chosen as the radius of the
melting crater at the steady state.
The dimensionless parameters M and V which represent the rel-
ative' importance of melting and evaporation, are defined as fol-
lows:
LM Ly
M V= (8)
3rl/2 c(TM - To)' . 37T1/2 c(TM - To) 10 10* I0 3 f/isec

The functions \p, wu, and wy are given by Carslaw and Jaeger Fig. 2 Propagation of melting and evaporation radii rM and rv with time
[28]: for copper—continuous point source of 2000 W

578 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the melting and evaporation radii of steel to be closer to one an-
• approximate method of this work
other than for copper (see Fig. 2 and 3). 10
experiment) results 4000 W
50 A, 100 V, Copper Anode

3 A p p r o x i m a t e M e t h o d for Calculation of M e t a l
Removal
The method described in the foregoing for calculation of the
propagation of the melting isotherm is an exact method. Its appli-
cation to the problem of a point source requires a considerable
amount of numerical work and of computer time. For more com-
plex problems the exact method may become cumbersome and a 500 W
significant increase of numerical computation will be needed.
Therefore, an approximate model is suggested, which simplifies
the evaluation of the melting crater while taking into account all
the phenomena involved. The approximation is made by "transfer-
ring" the sinks of melting and evaporation to the site of the point
Q=200W
source. The magnitudes qM and qy of these sinks remain, however,
as in the exact model, equations (2a) and (26). Other effects such
as ionization and excitation can be included in the model by the
addition of similar terms.
The temperature field is, again, expressed by equation (5), but
10' 10s t/isec
in the approximate model WM is the temperature field resulting
Pulse Duration
from a time dependent heat source at the origin [28]:
Fig. 4 Total mass of copper eroded by a point source, as a function of the

•r
d
3 -rV(r-r') pulse duration
WM (12)
dt (T- ./)3/Z

and similarly for wy.

The results of the approximate model for copper and steel are
confirmed by the experimental results, see Fig. 6.
shown in Figs. 2 and 3 and compared with those of the exact meth-
od. It can be seen that the approximations are very good, justifying 4 Experimental Technique and Results
the adoption of this model.
A single spark is discharged between two electrodes in a cup
The total erosion of the electrode is found by calculating the vol-
filled with a dielectric fluid (kerosene), see Fig. 7. The eroded
ume of the melting crater; in the case of a point source the crater is
metal is extracted from the dielectric fluid by chemical reaction
a half-sphere. The total mass of copper and steel removed by melt-
and the amount is measured by the intensity of colored solutions
ing and evaporation are plotted in Figs. 4 and 5 versus pulse dura-
in a spectro-photometer.
tion for several source powers. The results show that during short
The experimental rig is described in detail in a previous article
pulses copper is eroded more than steel (for the same power sup-
[2% The upper electrode is a rod (diameter 6 mm) tapered to a
ply) and vice versa for long pulses. The reason is that at the begin-
cone and the lower one is a disk (diameter 12 mm, thickness 3
ning of the pulse melting and evaporation consume most of the en-
mm). The gap spacing between them is 10 /tm and is measured by
ergy. As time increases, a steady state is approached where melting
a micrometer (accuracy 1 /on). The point of contact between the
and evaporation tend to stop and the energy is conducted into the
electrodes is determined by the closing of an electrical circuit
solid. The latent heats of steel are higher than those of copper;
through an ohm-meter. The pulse generator is capable of produc-
therefore melting and evaporation are more rapid in the latter
ing a rectangular pulse of duration above 20 us. Experiments were
metal. Since the heat conductivity (or rather IITM) of copper is
higher than for steel, the asymptotic eroded crater is smaller in
copper, as explained in the foregoing. These conclusions are also

4000 W
10"
• merfing radius-exact
- meiting radius-approximation
- evaporation radius-exact
- evaporation radius-approximation
RM2 RMI

I0 2 10" t^isec
10 10* 10" t^sec Pulse Duration

Fig. 3 Propagation of the melting and evaporation radii rM and rv with Fig. 5 Total mass of steei eroded by a point source as a function of the
t i m e for steel—continuous point source of 2000 W pulse duration

Journal of Hea! Transfer NOVEMBER 1975 / 579

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Total that when the chemical procedure is complicated and takes time,
erosion
the color of the solution can change if it is exposed to the light. For
steel and copper, however, the chemical method is simple and
takes only minutes, so there was no such problem. The main rea-
son for the variation in the experimental results is probably the
complicated nature of the process of EDM. For example, tiny par-
St Cathode! ticles are torn from the electrode surface by electric forces at the
Cu Cathode"!
beginning of the pulse, e.g., Williams [1-4]. The size and location
St Anode J of such particles depend on the metallurgical condition of the elec-
trode surfaces; since these cannot be exactly reproduced, one can-
not expect all the discharges to be exactly the same.
As mentioned in the foregoing, the actual process occurring dur-
50 500 1000 ing the discharge is very complicated. The analytical model pre-
Pulse duration (^ sec) sented in this work is a simplified procedure to describe the domi-
nant phenomena of the process, namely melting and evaporation.
Fig. 6 Total erosion of copper and steel electrodes: 100 V, 50 A
A more exact model should take into account the energy distribu-
tion in the channel, reactions and phase changes in the dielectric
fluid, electrical forces, convection in the gas phase, and more com-
plex geometries. Therefore, the theoretical results of this simpli-
carried out with pulses of 20-1000 MS, pulse voltage of 100 V and fied model can hardly be expected to match the experimental find-
current of 50 A. The shape of the pulse is observed in a Tektronix ings exactly.
memory oscilloscope. The pulse duration is measured from the However, it can be seen from Figs. 4 and 5 that for intermediate
moment of breakdown, after the "ignition delay." pulse durations this simple model yields reasonable estimates of
The sparks reported here were made between copper and steel. total erosion. The agreement between the results indicates that
Experiments were carried out first with a conical copper electrode evaporation and melting of the metals are indeed the dominant ef-
and then with a conical steel electrode. Several experiments were fects for intermediate pulse durations. For very short pulses elec-
conducted for every pulse duration (20, 50, 100, 200, 400, and 800 trical forces are considered to be more important [1-4].
yxs) and then the polarity was changed for the same configuration. For long pulses, a steady state is approached whereby evapora-
After each pulse a solution of 10 cm 3 HNO3 (1:3) is added to the tion and melting tend to stop; in this range the experimental re-
kerosene and dissolves all the eroded metal. The solution is sepa- sults show that the asymptotic constant value of eroded mass is
rated and a biquinoline solution is added and reacts with the cop- much lower and is reached much sooner than the theoretical pre-
per to form a pink colored solution. This solution is separated and dictions. This can be explained by several reasons: (a) the dis-
the color intensity is measured in the Spectronic 20 spectro-pho- charge channel has a finite width causing a heat source distributed
tometer. The amount of copper is determined from a calibration on a finite area rather than a point source; (b) ionization and exci-
curve. Now, to the other phase which contains the steel, more tation of the evaporated metal and the dielectric fluid take place,
HNO3 is added (to pH = 1) and 10cm 3 of 1.5 potassium-thio-cya- requiring energy thus decreasing the energy available for melting;
nate. The latter reacts with the steel and forms a yellow solution. (c) a certain part of the eroded material fuses back in the crater
The color intensity is measured, again in the spectro-photometer, after completion of the pulse. The model suggested in this work
and the amount of steel is also determined from a calibration can be extended to describe the first two effects. In order to over-
curve. come the difficulty of the third one and to prevent fusion, flushing
The experimental results for total erosion of the disk-shaped of the dielectric fluid between the electrodes may help. It is noted
electrode during a single spark are shown in Fig. 6. It is seen that that flushing of the dielectric fluid is being considered for control
copper is eroded more than steel at short times (short pulses) and of the voltage across the gap between the electrodes [19].
steel is eroded more than copper for long times (long pulse dura-
tions), in agreement with the theoretical predictions. Fig. 6 also Acknowledgment
shows the effect of polarity: erosion is higher when the disk-shaped This research work has been sponsored by the Deutschen
electrode (the workpiece) is the cathode. This known phenomenon Forschungs-Gemeinschaft (D.F.G.).
cannot be explained by the melting and evaporation mechanism.
Figs. 4 and 5 include a comparison between the experimental References
and theoretical results for copper and steel. For each pulse dura-
1 Williams, E. M., "Theory of Spark Machining," AIEE Trans., Vol.
tion, several measurements of single sparks were conducted. The 71,1952, pp. 105-108.
experimental results presented in Figs. 4 and 5 include the aver- 2 Williams, E. M., Woodford, J. B., and Smith, K. E„ "Recent Develop-
ages of these measurements together with the range of variation ments in the Theory and Design of Electric Spark Machining Tools," AIEE
for each case. The spectro-photometer used for measuring the Trans., Vol. 73,1954, pp. 83-88.
3 Williams, E. M., and Smith, R. E., "Phenomena Accompanying Tran-
amounts of the removed metals, is very sensitive. The error in- sient Low-Voltage Discharges in Liquid Dielectric (1)," AIEE Trans., Vol.
volved in measuring amounts of steel and copper did not exceed 74,1955, pp. 164-169.
0.5 jigr, in the range of the experiments discussed here. It is noted 4 Williams, E. M., and Smith, R. E.," Phenomena Accompanying Tran-
sient Low-Voltage Discharges in Liquid Dielectric (2)," AIEE Trans., Vol.
76,1957, pp. 94-97.
5 Hockenberry, T. O., and Williams, E. M., "Dynamic Evolution of
Events Accompanying the Low-Voltage Discharges Employed in EDM,"
IEEE Trans., Vol. 1, G.A. No. 4,1967, pp. 301-309.
6 Zingerman, A. S., "The Effect of Thermal Conductivity Upon the
Electrical Erosion of Metals, "Soviet Physics Technical Physics, Vol. 1,
1957, pp. 1945-1958.

0&
pulse
generator

1
iOu
ft-
tra dielectric fluid

> electrodes
7 Zingerman, A. S., "Electro Erosional Properties of Metals," Physics
of Metals and Metallography, Vol. 5,1957, pp. 58-67.
8 Zolotych, B. N., "The Mechanism of Electrical Erosion of Metals in
Liquid Dielectric Media," Soviet Physics Technical Physics, Vol. 4, 1960,
pp. 1370-1373.
9 Nekrashevitch, I. G., and Bakuto, I. A., "Present State of Theoretical
Concepts of the Electrical Erosion of Metals," in Electrospark Machining of
Fig. 7 Schematic arrangement of the EDM circuit Metals, Krasijuk, B. A., ed., Consultant Bureau, New York, 1965, pp. 17-22.

580 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 10 Afanasiev, N. V., "Theory and Calculation of Electric Erosion Pro- Method for Planar Solidification of Saturated Liquid With Convection at
cess," in Electrospark Machining of Metals, Krasijuk, B. A., ed., Consultant the Wall," International Journal of Heat and Mass Transfer, Vol. 16,1973,
Bureau, New York, 1965, pp. 60-66. pp. 1816-1819.
11 Saito, N., "Mechanism of Electric Discharge Machining," Bulletin of 20 Lazaridis, A., "A Numerical Solution of the Multidimensional Solidi-
the Japanese Society of Proceedings Engineer, Vol. 1, No. 2, 1964, pp. 9 5 - fication (or Melting) Problem," International Journal of Heat and Trans-
101. fer, Vol. 13,1970, pp. 1459-1477.
12 Heuvelman C. J., "Some Aspects of the Research on Electro-Erosion 21 Greene, J. E., and Guerrero-Alvarez, J. L., "Electro Erosion of Metal
Machining," Annals of CIRP, Vol. XVII, 1969, pp. 195-199. Surfacesj' Metallurgical Transactions, Vol. 5, 1974, pp. 695-706.
13 Zolotych, B. N., "Theorie zum Phanomen der Funkenerosiven Bear- 22 Crookall, J. R., and Heuvelman, C. J., "Electro-Discharge Machining—
beitung," Fertigung, Vol. 2, 1971, pp. 185-191. the State of the Art," Annals of the CIRP, Vol. 20,1971, pp. 113-120.
14 Snoeys, R., and Van-Dyck, F., Investigation of Electrodischarge Ma- 23 Cobine, J. D., and Burger, E. E., "Analysis of Electrode Phenomena
chining Operations by Means of Thermo Mathematical Model, Annals of in the High Current Arc," J. Appl. Phys., Vol. 26,1955, pp. 895-900.
the CIRP, Vol. 20,1971, pp. 35-36. 24 Smithells, C. J., Metals Reference Book. Vol. 3, Butterworths, Lon-
15 Boley, B. A., "The Analysis of Problems of Heat Conduction and don, Fourth ed., 1967.
Melting," High Temperature Structure and Materials, Proceedings of 3rd 25 Nekrashvich, I. G., and Nitkerich, S. P., "Some Relationships of the
Symposium on Naval Structural Mechanics, Pergamon Press, Oxford, Phenomena of the Electrical Erosion of Metals in a Low Voltage Discharge
1963. in a Liquid," Soviet Physics Technical Physics, Vol. 1,1956, pp. 83-88.
16 Muehlbauer, J. C , and Sunderland, J. E., "Heat Conduction With 26 Konnerth, K. L., "A Systematic Experimental Study of Electrode
Freezing or Melting," Appl. Mech. Rev., Vol. 8,1965, pp. 951-959. Phenomena Accompanying Transient Low Voltage Arcs in Liquid Dielec-
17 Budhia, H., and Kreith, F., "Heat Transfer With Melting or Freezing trics," PhD dissertation, Department of Electrical Engineering, Carnegie In-
in a Wedge," International Journal of Heat and Mass Transfer, Vol. 16, stitute of Technology, 1961.
1973, pp.195-211. 27 Winograd, U., and Almagor, M., "Experimental Investigation of Elec-
18 Sikarskie, D. L., and Boley, B. A., "The Solution of a Class of Two- tric Discharge Machining by Radioactive Tracers," J. Eng. Materials and
Dimensional Melting and Solidification Problems," International Journal Technology, TRANS. ASME, 1973, pp. 99-102.
of Solids Structures, Vol. 1, 1965, pp. 207-234. 28 Carslaw, H. S., and Jaeger, J. C , Conduction of Heat in Solids. Ox-
19 Pedroso, R. I., and Domoto, G. A., "Exact Solution by Perturbation ford, 1959.

Journal ©f Heat Transfer NOVEMBER 1975 / 581

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 O. M. Silvares
Asst. Professor, Escola Politecnica da
The Thermodynamics of Water
Universidade de Sao Paulo, Sao Paulo, Brazil.
Assoc. Mem. ASME
Transport From Biological
E. G. Cravalho
Assoc. Professor Cryogenic Engineering
Laboratory. Mem. ASME
Cells During Freezing
W. M. Toscano 1 A thermodynamic model for the freezing of biological cells has been developed and has
been applied to human erythrocytes. Analytical expressions describing the dynamics of
Research Associate Cryogenic Engineering
Laboratory. Assoc. Mem. ASME
water loss during the several stages of the freezing process have been derived from a cell
modeled as an open system surrounded by a membrane permeable to water only. The
permeability of the membrane to water is the most significant cell parameter in this pro-
Department of Mechanical Engineering, cess and in the present model, and is assumed to be a function of the temperature and
Massachusetts Institute of Technology,
osmolality of the extracellular solution. The resulting set of differential equations de-
Cambridge, Mass.
scribing the cell freezing process is solved numerically for various cooling rates. For cool-
ing rates less than 3000 K/min, erythrocytes lose 95 percent of their intracellular water
C. E. Huggins before the eutectic temperature is reached. For cooling rates greater than 3000 K/min,
Assoc. Professor, Harvard Medical School, Chief, the fraction of intracellular water remaining at the eutectic temperature is a strong
Low Temperature Surgical Unit and Director, function of cooling rate. The effect of supercooling of the extracellular solution on the ki-
Blood Bank and Transfusion Service,
netics of the cell water loss is also analyzed. As a consequence of the supercooling, the
Massachusetts General Hospital, Boston, Mass.
volume of water present intracellularly at a given temperature is substantially greater
than when no supercooling occurs. This condition favors intracellular ice formation and
is consistent with experimental observations in this laboratory.

Introduction ume. During freezing and thawing, the amount of intracellular

water may be altered drastically by events occurring in the extra-
The preservation of biological materials by reversible freezing cellular medium. For instance, ice may form outside the cell during
has been a challenge for a large sector of the scientific community freezing, thereby increasing the concentration of solutes in the ex-
in the fields of medicine, biology, and more recently, engineering. tracellular medium. As a result, the chemical potential of water in-
Although the potential of the process has been successfully dem- side the cell will be higher than that outside, and cell water will be
onstrated for cells in suspension such as erythrocytes, yeast, and expressed through the cell membrane to maintain osmotic equilib-
spermatozoa and for some tissues such as skin and human cornea, rium. On the other hand, it is possible that the temperature of the
further research is necessary if this experience is to be extended to cell suspension is decreased more rapidly than the cell can respond
other tissues and whole organs. The complexity of the physico- osmotically, and water inside the cell is nucleated before signifi-
chemical processes and the lack of knowledge of the parameters in- cant mass transfer can occur. Obviously, the final intracellular
volved in these processes are primarily responsible for the slow de- water content is vastly different in these two cases.
velopment of the basic research in this field. Mazur [l] 2 has developed an analytical model to study the kinet-
Analysis of the composition of a large number of plant and ani- ics of water loss from cells at subfreezing temperatures from which
mal cells shows that in general a large fraction of the volume of the he inferred predictions about the onset of intracellular ice forma-
cell is occupied by water. For example, in human erythrocytes (red tion. This model has been tested experimentally for some simple
blood cells) this fraction is approximately 63.15 percent of cell vol- cells, and some of these experiments have confirmed the findings

1
Present address: Cryogenic Technology, Inc., Waltham, Mass.
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer
Division July 8,1975. Paper No. 76-HT-O. ' Numbers in brackets designate References at end of paper.

582 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 of the model [2-4] whereas others [5, 6] have found quantitative
data that contradict the model. Mazur's model predicts that intra- SEMI-PERMEABLE
cellular ice will form in essentially all red cells at a cooling ;ate of ' CELL MEMBRANE
3000 K/min, but Diller, et al. [6] have observed experimentally / ^ CELL ^ \ ^
that the frequency of intracellular ice increased from 0 percent at
fllN
—840 K/min to 100 percent at —850 K/min. These investigators / N X N°WT
have also noted that when the extracellular medium is allowed to
1H^O
supercool 12 K, the frequency of intracellular freezing increases N? U T
from 0 percent at —6 K/min to 100 percent at - 1 6 K/min. Diller, \ NIN /
H20
et al. [7] point out that other parameters not included in Mazur's
model may also affect its ability to predict the cooling rate that
marks the onset of intracellular ice.
v__^/ CONTROL
C- ^ - f H f t O T ^
Mazur [8] has also postulated that two mechanisms of freezing VOLUME - -
injury are responsible for the inverted U-shape of the survival sig- BOUNDARY
natures (survival versus cooling rate curve): "solution effects" and
\
"intracellular freezing." The solution effect is responsible for the Fig. 1 Physical model for the freezing of a biological cell
damage of cells cooled at rates below the optimum value which
gives maximum survival. During the freezing period, pure ice pre-
cipitates in the extracellular medium and the cells are trapped in
pools of concentrated salt solution in the liquid state. When the as a single cell immersed in a salt solution. As shown in Fig. 1, the
cells are cooled slowly, the time during which the cells are subject cell contains water, salts, and proteins enclosed in a semipermea-
to this increasingly concentrated solution is longer than when ble membrane which allows water, but not solutes, to flow across
cooled rapidly. During this period, the osmotic pressure difference it. The pressure is assumed uniform throughout the system and
builds up and the cell dehydrates causing the solute concentration constant. Thermal equilibrium is assumed to prevail even when
inside the cell to increase to lethal levels. the temperature is changing with time. Both the intracellular and
At higher cooling rates, there is less time available for the intra- extracellular media are homogeneous. Initially the internal and ex-
cellular solution to equilibrate osmotically with the extracellular ternal media are in chemical equilibrium, i.e., the chemical poten-
medium. Thus, at any particular temperature, there is more water tials of the various components are equal across the membrane in
inside the cell, and the larger the volume of supercooled water re- the initial state.
maining inside the cell, the greater the probability of ice nuclea- The extracellular medium is modeled as an aqueous solution of
tion inside the cell. Although there is experimental evidence that sodium chloride in which the salt is completely disassociated. In
the intracellular ice is a damaging factor for the cell [9, 10], the keeping with the experimental observations of McGrath [11] and
precise mechanism responsible for the damage is not yet under- others, it is assumed that during the freezing process the external
stood. medium nucleates first, followed by intracellular nucleation if the
In the present work, we shall discuss the thermodynamic pro- thermodynamic conditions are suitable. The mechanisms of ice
cesses involved during cell freezing with special attention devoted formation are too complex to be treated here but are discussed by
to the human erythrocyte as a cell model. We also shall develop a the present authors in a related paper [12]. Nucleation in the ex-
more complete thermodynamic model that describes these events tracellular medium upsets the chemical equilibrium that prevails
and includes factors neglected in Mazur's analysis. initially across the membrane. Since the membrane offers a finite
resistance to the flow of water through it, a difference in chemical
Analysis potential forms across the membrane. The present model assumes
The biological cell immersed in its natural medium is modeled that once ice forms extracellularly, chemical equilibrium prevails

-Nomenclature.
A = surface area, M2
A7 = affinity of the reaction, erg/mole Lp = permeability coefficient, cm 3 /dyne-s
O-T,P - partial derivative of affinity with re- n = number of moles 7 = activity coefficient
spect to £ at constant T and P [ = ( M / P = absolute pressure, dyne/cm 2 ATS = degree of supercooling, K
d£)T,p], erg/mole R = universal gas constant, erg/mole-K M = specific molar chemical potential, erg/
B = cooling rate, K/min T = absolute temperature, K mole
b = permeability temperature coefficient, Tg - permeability reference temperature, n* = specific molar chemical potential of
(K)- 1 K pure component, erg/mole
cpi = partial molar heat capacity of i at TM = melting temperature of pure water, v = osmolality, mOsm/1
constant P\ erg/mole-K K £ = extent of reaction
H = molar enthalpy, erg/mole Tjv = nucleation temperature of the extra- 0 = osmotic coefficient
hr,p = heat of reaction at constant T and cellular solution, K
P[es(dff/a£)r,p], erg/mole t = time, s Subscript
k = permeability of cell membrane, (mole) 2 / tjv = time of onset of nucleation, s s = solute
/i 5 -atm-min V = volume, M3 7 = solvent
k* = permeability of cell membrane, n3/ Vi = initial volume, M3
H 2 -min-atm vm = partial molar volume of water, cm 3 / Superscript
L 7 = latent heat of fusion at temperature mole in = intracellular
T, erg/mole VT.P = volume change of reaction at con- out = extracellular
Ly° = latent heat of fusion at temperature stant T and P [=(aVh^)T,p], cm 3 /mole a = liquid phase
TM, erg/mole x = molar fraction 0 = solid phase

Journal of Heat Transfer NOVEMBER 1975 / 583

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 between the solid and liquid phases in the extracellular medium as S t a g e II—Cooling After Extracellular Nucleation. When a
the uniform temperature of the cell suspension decreases with suspension in state 2 nucleates in the extracellular medium, it is
time. Thus, the flow resistance of the membrane causes the intra- assumed that the solution outside the cell will revert to an equilib-
cellular medium to be supercooled. rium state located somewhere on the locus of two-phase equilibri-
In a typical freezing protocol there are at least four distinct um states. The exact location will depend on the latent heat re-
stages: Stage I—Cooling prior to the onset of nucleation in the ex- leased relative to the refrigeration power producing the cooling
tracellular medium. During this period equilibrium conditions pre- process. In the general case of cells frozen in bulk, a limited
vail throughout the system as the temperature decreases. Stage II— amount of refrigeration power is available. Consequently, the nu-
Cooling after extracellular nucleation. During this period equilibri- cleation of the extracellular solution is a nonisothermal process in
um conditions prevail between the liquid and solid phases in the which the latent heat of fusion released during the nucleation in-
extracellular medium, and the intracellular solution follows a path creases the temperature of the system. The extracellular solution
in the supercooled region of the phase diagram. Stage III—Cooling reaches an equilibrium state somewhere on the locus of two-phase
below the eutectic temperature. The extracellular solution solidi- equilibrium states as represented by point 4 in Fig. 2. Because the
fies completely below the eutectic temperature. Stage IV—Hold- cell is not able to respond to the osmotic pressure difference that
ing period at minimum temperature. When the system reaches the can develop across its membrane during the time period required
desired minimum temperature it is kept a t a constant temperature for the nucleation of the extracellular solution, the molar fraction
for a period of time. of the intracellular water will remain constant, and the state of the
Stage I—Cooling Prior to the Onset of Extracellular Nu- intracellular solution may be represented by point 5 in Fig. 2. Thus
cleation. I t is possible for the extracellular medium to be super- for the general case the states corresponding to points 4 and 5 in
cooled as a consequence of the dynamics of extracellular nuclea- Fig. 2 represent the initial conditions for Stage II of the freezing
tion, but this supercooling terminates the instant ice forms. The process. In the present analysis it is assumed that the refrigeration
intracellular medium will continue to be supercooled throughout power is sufficient to counteract exactly the latent heat released
the freezing process even though ice is present extracellularly. The during the phase change. Thus, the nucleation of the extracellular
phase diagram of Fig. 2 shows a typical situation. A cell suspension medium will follow an isothermal path in the phase diagram and
its state just after nucleation occurs will correspond to point 3 in
with an initial water molar fraction of Xy is cooled at a con-
Fig. 2. This is approximately the case that prevails in the freezing
stant rate B. If no supercooling occurs in the extracellular medi-
of cell suspensions in thin layers like the ones used in cryomicros-
um, the temperature of the suspension decreases continuously
copic observations. Note that the present analysis can be applied
with uniform chemical potential of the water inside and outside
even in the more common case in which cooling follows states 1, 2,
the cell given by
and 4 in Fig. 2, and where ATS is made to approach zero.
fly" •• M T = ny* (T,P) + RTifalnXyL (1) For this new equilibrium state, the chemical potential of the
water in the external liquid solution is given by
The intracellular and extracellular solutions are assumed to have
equal osmotic coefficients so equation (1) implies fiy" o u t = MT* (T,P) + RTs4>slnx 7 3 (4)
A-yl Xy\ (2) If the degree of supercooling ATS is small compared with the
value of the temperature Tlt <f>3\nxy3aoM in equation (4) can be ob-
When state 1 is reached, extracellular nucleation occurs. Further tained by a Taylor's series expansion around the initial equilibri-
cooling of the cell suspension forces the extracellular medium to um state (state 1 in Fig. 2).
follow the locus of two-phase equilibrium states. If the thermody-
namics of the situation are such that the extracellular nucleation ifalnx Y3 PllliXy - — (0ilnx 7 ,l
) ATS . . . (5)
does not occur in state 1, the suspension supercools and the path of
the cooling process continues beyond the two-phase equilibrium In order to evaluate the first derivative in equation (5), it is neces-
line. In Fig. 2, state 2 would be such a state for which the chemical sary to determine the equation of the equilibrium curve in Fig. 2.
potential of the water would be given by It can be shown [13] that along the equilibrium curve t h e mole
fraction is related to the latent heat of fusion.
MT2oin = MT2"out = M7* (T,P) + RT24>Mxy (3)

which again implies equation (2). f (Ly/RT2) dT (6)

JTM

The temperature dependence of the latent heat of fusion is given

in [13] by

Ly(T)=Ly°+ P (cPya - cpySrkb) dT. (7)

M
- — P A T H OF COOLING
PROCESS

1 Substitution of equation (7) into equation (6) yields

T h
V 5 - r-^rv LOCUS OF TWO PHASE
\ v -EQUILIBRIUM STATES.
/
ln7 7 o r o u t x 7 " O T , t '
^-"s R \T TM>
cPy" — cpy&_ (",_ (TM\ _ TM
i
1
;
i
: x
^ v
R [-£) + 1 (8)

The relationship between the activity coefficient and the osmotic

E
coefficient is by definition
i ' i
i 1 i
I 1 !
i i i ln7 T (9)
i i t
i ' 1
i i f Substitution of equation (9) into equation (8) and differentiation
*rz *n v. with respect to T, yields
MOLAR FRACTION OF tYArER.X a{4>inXya " U t ) = Ly° Cpy" - CPyV /TM _ 1_\
(10)
Fig. 2 Initial states for the freezing process aT RT2 R \r 2 ?>

584 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Terms with derivatives of higher order than the first can be ne- freezes extracellularly in order to maintain the equilibrium be-
glected in the series expansion. Then equation (5) can be written tween solid and liquid phases outside the cell. In this case, the
equilibrium relation of equation (17) is replaced by
b-Anxy3 0ilnx T i
drc/QUt _ dnyai
L / A T , cpj. - cp. (18)
ATS (11) dT dT
RT!2 VTI2 TJ
and equation (16) and (18) are solved simultaneously with the ini-
Substitution of equation (11) into equation (4) together with equa- tial conditions corresponding to those at the eutectic temperature.
tions (1) and (3) and the assumptions that (a) the variation of the This system of differential equations describes Stage III of the
osmotic coefficient with temperature is negligible, and (b) the in- freezing protocol.
tracellular and extracellular solutions have the same osmotic coef- Stage IV—Water Transfer at Minimum Temperature.
ficient, yields the difference in the chemical potential between the Stage IV of the freezing protocol is characterized by a constant
supercooled and equilibrium states. temperature mass transfer. In order to maintain solid-liquid equi-
(12) librium in the extracellular medium at the minimum temperature,
M72 •• L^TJTsATJTr*
all the water transferred across the membrane during this period
must solidify immediately extracellularly. This equilibration pro-
If the extracellular medium assumes the equilibrium state at the
cess is described by equation (16) and the following equilibrium
instant of nucleation while the intracellular medium remains in
relation in the time domain
the supercooled state because of the membrane permeability,
equation (12) gives the relationship between the water molar frac- dnff out ^ dn^

tion of the solutions inside and outside the cell when nucleation (19)
dt dt
The initial conditions for this system of differential equations are
those corresponding to the final state of the cooling period.
cT3 = exp iL^TJATJiRTt*] (13) Conservation of Mass of Water and Solutes. Since the con-
where <j> is the average value of the osmotic coefficient between the trol volume in Fig. 1 is closed, water and all solutes must be con-
molar fractions xyi"m and xy3a"ul. Equation (13) then gives the served within its boundaries. Then the following conservation
initial conditions for the process of freezing subsequent to extra- equations must be satisfied at all times
cellular nucleation.
1
The loss of extracellular water due to the formation of the solid + ny" o u t + n/ o u t = constant (20)
phase results in the difference between the chemical potentials in- and
side and outside the cell described earlier. The cell contents at-
tempt to equilibrate with the external medium by water transport ns'n + nsout = constant (21)
across the semipermeable membrane. The flux of water across the
membrane is given by the phenomenological expression. and in addition nsm and n s ° u t are constant due to the assumed im-
permeability of the cell membrane to solutes.
Initial Conditions. The initial conditions for the integration of
Jy = - ^ ( M 7 " i n - M 7 a o u t ) - (14)
the system of differential equations describing Stage II of the
Then the rate of change of the water content inside the cell is
freezing protocol are the following:
dny« in 1 At T = Ti = Ti — ATS (nucleation temperature of the extra-
-kA(nyain- M T " out )- (15) cellular 'medium)
dt
Introduction of the definition of the cooling rate, i.e., B = dT/dt -"explLyiTJATJiRTfl. (22)
into equation (15) yields
If the supercooling is zero (i.e., ATS = 0), equation (22) reduces to
dnyam ain QUt
= -kA(ny - MT" )/£.' (16) (23)
dT
2 At T = T2
The succession of thermodynamic states through which the ex-
tracellular system passes during Stage II of the freezing protocol ny<f (24)
can be represented by a succession of equilibrium states which can a m
be described by the van't Hoff relation, viz. where nyo is the number of moles of water inside the cell under
isotonic conditions.
r
dny"in rfnT"ou'l The initial conditions for the remaining stages of the freezing
(17) protocol correspond to the final conditions of the previous stage,
dT dT T
i.e., if the minimum temperature is lower than the eutectic temper-
Equation (17) represents displacement of the state of the extracel- ature, the number of moles of water inside and outside the cell and
lular medium along the equilibrium line. It gives the variation of in the solid phase at the end of Stage II, represents the initial con-
the number of moles of water in the solid phase with respect to dition for the integration of the system of differential equations
temperature. describing Stage III of the freezing protocol. When the tempera-
Stage III—Cooling Below the Eutectic Temperature. ture reaches the pre-established minimum value, the final values
When the system reaches the eutectic temperature, the extracellu- of the variables constitute the set of initial conditions for Stage IV
lar medium is completely solidified. The thermodynamic state of of the freezing protocol. The intracellular water content a t any
the intracellular medium at this temperature depends on the cool- temperature during freezing is given by
ing rate. In the low range of cooling rates, quasi-equilibrium condi-
tions across the membrane prevail before the system has reached V1°in = nT«inum. (25)
the eutectic temperature. Thus, no significant volume of water is
transferred at subeutectic temperatures, and the solid phase is Results
present on both sides of the cell membrane. In the high range of The model just presented is general and may be applied to single
cooling rates, the system is far from equilibrium when the temper- cells satisfying the conditions stated. One type of cell for which
ature reaches the eutectic value. Supercooled intracellular water freezing is of great clinical importance is the human red blood cell.
then continues to flow across the membrane and immediately Human Erythrocyte Geometry and Composition. Measure-

Jourrtal of Heat Transfer NOVEMBER 1975 / 585

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ments of the diameter, area, and volume of the individual human
red blood cells have been reported by several investigators [14-16]
using different techniques, but the geometric parameters used in
the present work are those averaged values determined by Canham
and Burton [14], i.e., an average volume, surface area, and diame-
ter of 107.5 M3, 138.1 ii2, and 8 p., respectively. In the present model
it was assumed that the .surface area is constant and equal to the
average value regardless of any changes that may occur to the vol-
ume of the cell during freezing.
The human erythrocyte is an unnucleated cell containing a com-
plex aqueous solution of hemoglobin, nonhemoglobin proteins, lip-
ids, glucose, electrolytes, and other components (see references [17,
18]). However, approximately 97 percent of its volume is occupied
by water and hemoglobin with the volume of intracellular water
available for transport representing 63.15 percent of the total vol-
ume of cell. (The actual volume of water in the cell is 71.5 percent 250-
of the total volume but 11.6 percent of this water is water of hydra-
tion associated with the hemog'obin.) In osmotic experiments, Far-
rant and Woolgar [19] have shown that the mass of intracellular
water at 273.15 K decreases as the osmolality of the suspending
medium increases. However, beyond 2000 mOsm/1, when the
weight of the cell water is 0.33 of that in the cell under isotonic
1.0 0.9 0.8
conditions, very little additional loss of water occurs. In the appli-
MOLAR FRACTION OF WATER, X
cation of the present model to erythrocytes, we have assumed that
Fig. 3 Locus of states of Intracellular solutions cooled at different rates
all the water is transferable because the concentration of the extra- with no supercooling—locus of states of extracellular solution coincident
cellular solution equal to 2000 mOsm/1 (corresponding to a temper- with two-phase equilibrium states for all cooling rates
ature during freezing of 269.35 K) reported by Farrant, is still low
compared with the concentration that the cell will be exposed to at
temperatures as low as 252.0 K (eutectic point for NaCl-aqueous
solution).
Composition of the External Medium. The human red cell analysis [23].
"in vivo" is suspended in an aqueous solution containing primarily Osmotic Coefficient. In the integration of equation (17), the
proteins and salts. It is common clinical practice to suspend nonideality of the external solution is accounted for by the osmotic
packed red blood cells in normal saline solution with no significant coefficient which is a function of the concentration of solutes in
damage to the cell. In the present model we have assumed that the the solution. Values of this parameter for sodium chloride solu-
extracellular medium is a sodium chloride aqueous solution which tions have been determined by Robinson and Stokes [24] for the
initially is isotonic, i.e., 0.154 moles of NaCl in 11 of solution. equilibrium conditions that prevail in the extracellular medium at
Water Permeability of the Erythrocyte Membrane. Several any given temperature. The effect of the hemoglobin on the osmot-
studies on the permeability of human erythrocyte membranes to ic coefficient of the intracellular solution was neglected because of
water are available. The effect of temperature on the permeability its low molality [25].
to water was first measured by Jacobs, et al. [20] and these data Application of the Model. For the case of human erythrocytes
suggest that the temperature dependence of water permeability the set of differential equations governing the water flow from bio-
(for the temperature range of their experiments, 303.15-273.15 K) logical cells during Stages II, III, and IV of the freezing protocol
is given by were solved numerically. Fig. 3 shows the thermodynamic state of
the intracellular solution of a cell cooled at several rates with zero
k* kg exp[b(T • Te)l (26) supercooling. The locus of the thermodynamic states of the extra-
cellular solution in all cases follows the equilibrium curve. The
Experiments to determine the constants kg and b have been car- cooling rate of 3000 K/min distinguishes two different modes of
ried out in the temperature range above 273.15 K. One of the more behavior of the system below the eutectic temperature. For cooling
recent determinations of these constants was by Sidel and Solo- rates lower than 3000 K/min the system reaches equilibrium be-
mon [21] and their values are kg = byPln'1 min-atm and b = fore the eutectic temperature and no changes in volume occur
0.0325(Kr l . Recently Rich, et al. [22] studied the effect of the ex- below that temperature. For cooling rates greater than 3000 K/
ternal medium osmolality on the water permeability of human and min, Fig. 3 shows that supercooled intracellular water continues to
dog erythrocytes. They found that the permeability coefficient for be transferred at subeutectic temperatures as described previous-
water, Lp, (identical to the permeability, k*, except for the system
of units) of human red blood cells varies from 1.87 X 1 0 u cm 3 / ly.
dyne-s at 199 mOsm/1 to 0.76 x 10 11 cm 3 /dyne-s at 516 mOsm/1. In The effect of the membrane permeability on the cooling trajec-
addition they found that the change in Lp is rather rapid and it is tory is presented in Fig. 4 in which the locus of states of the intra-
completed within 50 ms after a finite change in the osmolality of cellular medium are shown for cooling cells of different permeabili-
the external medium. It can be shown that their experimental data ties at a fixed rate of 1000 K/min. The variation of the cell perme-
for human red blood cells obey the following relationship: ability was obtained by changing the reference temperature Tg in
equation (28). While this method of varying k* is not physically re-
lnL„ = 335.174/v - 26.2864. (27) alistic, it is convenient from a computational point of view; the net
result of this technique is the same as a simultaneous variation of
Since during freezing the concentration of the extracellular me- kg and b. Thus, from Figs. 3 and 4 it follows that the cooling tra-
dium varies as the temperature changes, the combined effect of jectory of a cell with a high water membrane permeability cooled
these two factors on the permeability is given by at a low rate is similar to the cooling trajectory of a cell with a low
water permeability cooled at a high rate.
k = 1.852 X 10" 15 exp [335.174/p - 26.2864 + b(T - Ts)]. (28)
During State IV of the freezing protocol, the cooling process is
Other factors may influence the permeability of the membrane at suppressed and the temperature is maintained constant and equal
subzero temperatures, but these are not included in the present to that value. Depending on the cooling rate and the minimum

586 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 P^
1.0
=5g„^ ' ' ' ' '

\ \ v ^ \ \ ^ ~ \ ATs = 0"K
-

- 1 \ \ \ \ \ ^ \ -lOOOO-K/MIN -
\ \ -I000"K/M]N \ \ ^ v .
>.5
_
\ \
\
\ \
\ "500°K/MIN
\
\
^ \
\--
-750B'K/MIN -
I
\ / \
\ ^V
-100-K/Mlh .-3000*K/MIN
< \
~ \
-5000-K/MlrJ
~

^ - _ _ \^ ^\_ -
1 1 1 I 1 I

TEMPERATURE,T (°K)
Fig. 6 Volume of intracellular water of erythrocytes cooled at constant
rates w i t h no supercooling

1.0 0.90 0.8

MOLAR FRACTION OF WATER ,X but in the present model the permeability was assumed to be de-
Fig. 4 Effect of m e m b r a n e permeability on locus of states of intracellular pendent upon the temperature and the osmolality of the external
solution c o o l e d at 1000 K / m i n
medium as discussed earlier. Note that the present model assumes
that the solutions behave as nonideal solutions and the latent heat
of fusion varies with temperature, whereas the model in reference
[1] assumed ideal solution behavior and constant heat of fusion.
temperature, equilibrium conditions across the membrane may not
Comparison of these two models for a given permeability function-
be established at that instant. The chemical potential difference
al dependence shows these refinements to have a negligible effect
between the two compartments forces water to flow across the
on the calculated volume of cell water during freezing. On the basis
membrane at constant temperature. This isothermal process con-
of recent experiments in this laboratory, the results of the present
tinues until equilibrium across the membrane is reached. A typical
model for the cell permeability are a better representation of the
example of this equilibration period is presented in the phase di-
actual conditions associated with the freezing of red blood cells.
agram in Fig. 5.
Fig. 6 presents the variation of the volume of the erythrocyte Fig. 8 presents the state of the intracellular solution of cell sus-
water content during freezing at several cooling rates. Fig. 7 com- pensions subjected to supercooling of the extracellular medium for
pares the results of the present model with those obtained by several cooling rates. Fig. 9 shows the temperature dependence of
Mazur [1] for some typical cooling rates. The quantitative differ- the intracellular water content of those cells. The comparison be-
ences between the two models are not due to any differences in tween Figs. 3 and 8 shows that for the same cooling rate, the initial
basic assumptions, but instead due primarily to the different mod- water flux across the membrane is significantly larger when super-
els of the water permeability and the different values assumed for cooling of the extracellular medium occurs, than the case when the
the cell surface area. Mazur assumed that the water permeability supercooling is zero. The same conclusion is apparent by compar-
of the erythrocyte membrane is a function of temperature only, ing the slopes of curves in Fig. 6 with those in Fig. 9. Thus, when
the extracellular medium supercools 10 K, the reduction in volume
after extracellular nucleation is much faster than when no super-
cooling occurs. This effect is more noticeable in the low range of
cooling rates. The effect of the supercooling is particularly impor-
tant for the study of the conditions for intracellular nucleation as

PRESENT MODEL
-I000°K/M[N
5000°K/MIN
MAZUR'S MODEL
-I000°K/MIN
<
rr 5000°K/M!N

0.0

MOLAR FRACTION OF WATER, X TEMPERATURE , T (°K)

Fig. 5 Equilibration of intracellular and extracellular solutions held at m i n - Fig. 7 Volume of intracellular water of erythrocytes cooled at constant
i m u m t e m p e r a t u r e after cooling at several rates rates w i t h no supercooling for present m o d e l and model in [ 1 ]

Journal of Heat Transfer NOVEMBER 1975 / 587

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 >
- K. ' ' ' -
- ATg'IO"* -
>
:>
- \ \-5000 D K/MIN :
° 0.5 - _
< - \-1000°K/MIN \ _
Q:

VOLUME
-100°K/MIN

-
1 , !
260 250
,
TEMPERATURE,T(°K)

Fig. 9 Volume of intracellular water of erythrocytes cooled at constant

rates with supercooling of 10 K

Institute of Technology, Department of Mechanical Engineering, 1973.

3 Mazur, P., "Manifestations of Injury in Yeast Cells Exposed to
Subzero Temperatures. I. Morphological Changes in Freeze-Substituted and
in Frozen-Thawed Cells," J. Bacterial., Vol. 82,1961, pp. 662-672.
0.9 0.8 4 Nei, T., "Effects of Freezing and Freeze-drying on Microorganisms,"
MOLAR FRACTION OF WATER, X
in Recent Research in Freezing and Drying, A. S. Parkes and A. U. Smith,
eds., 1960, pp. 78-86.
Fig. 8 Locus of states of Intracellular solutions cooled at different rates 5 Diller, K. R., "A Microscopic Investigation of Intracellular Ice For-
with supercooling of 10 K—for ali cooling rates the locus of states of extra- mation in Frozen Erythrocytes," ScD thesis, Massachusetts Institute of
cellular solution coincident with two-phase equilibrium states for tempera- Technology, Department of Mechanical Engineering, 1972.
tures less than nucleation temperatures 6 Diller, K. R., Cravalho, E. G., and Huggins, C. E., "An Experimental
Study of Freezing in Erythrocytes," to be published in the Journal of Medi-
cal and Bio Engineering, 1975.
7 Diller, K. R., Cravalho, E. G., and Huggins, C. E., "Intracellular
Freezing in Biomaterials," Cryobiology, Vol. 9, 1972, pp. 429-444.
discussed in a related paper [12]. 8 Mazur, P., "Causes of Injury in Frozen and Thawed Ceils," Fedn.
Proc. Fedn. Am. Socs. Exp. Biol., Vol. 24, suppl. 15,1965, pp. S175-S182.
9 Mazur, P., and Schmidt, J. J., "Interactions of Cooling Velocity,
Conclusions Temperature, and Warming Velocity on the Survival of Frozen and Thawed
The present analysis quantifies the competition between heat Yeast," Cryobiology, Vol. 5,1968, pp. 1-17.
transfer and mass transfer during freezing of cell suspensions. At 10 Rapatz, G., Nath, J., and Luyet, B. J., "Electron Microscope Study of
Erythrocytes in Rapidly Frozen Mammalian Blood," Biodynamica, Vol. 9,
low cooling rates, mass transfer dominates and a considerable por- No. 177, 1963, pp. 83-94.
tion of the intracellular water leaves the cell to form extracellular 11 McGrath, J. J., "The Dynamics of Freezing and Thawing Mammalian
ice. At high cooling rates, heat transfer dominates as the tempera- Cells: The HeLa Cell," ScM thesis, Massachusetts Institute of Technology,
ture is reduced at a faster rate than water can flow through the Department of Mechanical Engineering, 1974.
12 Toscano, W. M., Cravalho, E. G., Silvares, 0. M., and Huggins, C. E.,
membrane. The resulting large volume of intracellular water favors "The Thermodynamics of Intracellular Ice Nucleation in the Freezing of
intracellular ice nucleation. Erythrocytes," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Series
The present analysis also shows that high extracellular osmolal- C, Vol. 97,1975, pp. 326-332.
ity attendant to freezing introduces an additional resistance to 13 Prigogine, I., and Defay, R., Chemical Thermodynamics, Wiley, New
York, 1954.
transmembrane water transport that must be superimposed on the 14 Canham, P. B., and Burton, A. C, "Distribution of Size and Shape in
temperature dependence of the permeability. The net result is a Populations of Normal Human Red Cells," Circulation Research, Vol. 22,
reduction in intracellular water loss during freezing. The analysis 1968, pp. 405-422.
demonstrates the major role of the water permeability of the mem- 15 Ponder, E., and Millar, W. G., "The Measurement of the Diameters
of Erythrocytes. I. The Mean Diameter of the Red Cells in Man," Quart. J.
brane and the necessity for further experimental measurements of Exp. Physiol., Vol. 14,1924, pp. 67-82.
cell permeability if the freezing process in biomaterials is to be 16 Chin, P. C. Y., and Fung, Y. C, "Extreme-Value Statistics of Human
better understood. The present results also show that nonideal so- Red Blood Cells," Microvascular Research, Vol. 6, 1973, pp. 32-43.
lution behavior and a temperature dependent heat of fusion do not 17 Wintrobe, M. M., Clinical Hematology, Sixth ed., Lea & Febiger,
Philadelphia, 1967, p. 105.
result in any significant improvement to models in which these ef-
18 Beutler, E., and Srivastava, S. K., "Composition of the Erythrocyte,"
fects are neglected. In the present analysis the consequence of su- Hematology, McGraw-Hill, New York, 1972, p. 94.
percooling the extracellular medium is that a larger volume of 19 Farrant, J., and Woolgar, A. E., "Human Red Cells Under Hyperton-
water remains inside the cell at lower temperatures than when no ic Conditions: A Model System for Investigating Freezing Damage. I. Sodi-
supercooling occurs. This condition also favors intracellular ice um Chloride," Cryobiology, Vol. 9,1972, pp. 9-15.
20 Jacobs, M. H., Glassman, H. N., and Parpart, A. K., "Osmotic Prop-
formation. erties of the Erythrocyte: The Temperature Coefficient of Certain Hemolyt-
ic Processes," J. of Cel. Comp. Physiol., Vol. 7,1935, pp. 197-225.
Acknowledgments 21 Sidel, V. W., and Solomon, A. K., "Entrance of Water Into Human
Red Cells Under an Osmotic Pressure Gradient," J. Gen. Physiol, Vol. 41,
This work was supported in part by USPHS Grant 1 P01 1957, pp. 243-257.
HL4322-03 from NHLI; and Fundacao de Amparo a Pesquisa do 22 Rich, G. T., Sha'afi, R. I., Romualdez, A., and Solomon, A. K., "Effect
Estado de Sao Paulo and Estola Politecnica de Universidade de of Osmolality on the Hydraulic Permeability Coefficient of Red Cells," J.
Sao Paulo, Brazil. Gen. Physiol., Vol. 52, 1968, pp. 941-954.
23 Levin, R. L., Cravalho, E. G., and Huggins, C. E., "A Membrane
Model Describing the Effect of Temperature on the Water Conductivity of
References Erythrocyte Membranes at Subzero Temperatures," submitted to Cryobiol-
1 Mazur, P., "Kinetics of Water Loss From Cells at Subzero Tempera- ogy, 1975.
tures and the Likelihood of Intracellular Freezing," Gen. Physiol., Vol. 47, 24 Robinson, R., and Stokes, R. H., Electrolyte Solutions, Academic
1963, pp. 347-369. Press, New York, 1955.
2 Ushiyama, M., "Volumetric Changes in Saccharomyces Cerivisiae 25 Gary-Bobo, C. M., and Solomon, A. K., "Properties of Hemoglobin
During Freezing at Constant Cooling Velocities," ScM thesis, Massachusetts Solutions in Red Cells," J. Gen. Physiol., Vol. 52,1968, pp. 825-853.

588 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 K. R. Perkins
Engineer,
Aerojet Nuclear Co., Idaho Falls, Idaho
lean Temperature Profiles in
D. M. McEIigot Heated Laminarizing Air Flows
Professor,
Aerospace and Mechanical Engineering Experiments for air flowing upward in a vertical circular tube are presented for heating
Department,
rates causing significant property variation. Entering Reynolds numbers vary from 3800
University of Arizona, Tucson, Ariz.
Assoc. Mem. ASME to 8500 with the dimensionless heating parameter, qm"' l(GcPiiTi), varying from 0.001 to
0.0055. Data include the first developing internal profiles for the apparent reverse tran-
sition from turbulent to laminar flow due to heating. Emphasis is placed on internal
temperature distributions, t(x, r), taken at axial stations from about 3-25 dia after the
start of heating. The flow regimes range from nearly constant property, turbulent flow to
strongly heated laminarizing flow. Results for strongly heated flows show a definite in-
crease in the thickness of both the linear sublayer and the viscous sublayer. For laminar-
ized flow, the linear sublayer extends about half the distance to the cener line of the tube
or more.

Introduction careful measurements of the thermal boundary conditions and

inlet conditions.
The phenomenon of laminarization has come under scrutiny re-
cently due to the resulting reduction in heat transfer parameters The Experiment
which is sometimes severe [1, 2, 3]. 1 Such reductions are not pre- The apparatus employed in this investigation is comparable to
dicted by the empirical correlations accounting for property varia- that of Perkins, Schade, and McEIigot [1] except insulation and
tion [4, 5] nor by numerical predictions [6]; under certain severe guard heating is used around the test section instead of vacujum,
heating conditions they agree fairly well with laminar predictions an air compressor replaced the gas bottles, and the tube exit ex-
in the downstream region even though the local bulk Reynolds hausted directly into the laboratory. The test section consisted of a
numbers often are still higher than the usual value for transition to seamless, extruded Inconel 600 circular tube with 27.4 mm (1.08
turbulent flow. We call this phenomenon laminarization. Refer- in.) ID suspended vertically and heated resistively. A 50-dia adia-
ences [1, 2] are recommended for more extensive reviews of the batic entry preceded the 32-dia heated section.
pertinent literature. Internal mean temperature profiles were measured with mini-
The experiments involved before have been conducted in tubes aturized boundary layer thermocouple probes of premium grade,
with diameters less than 13 mm (% in.). Consequently, internal Chromel-Alumel as illustrated in Fig. 1. The probe was introduced
profile measurements were not feasible so the data are not of much through the open exit of the tube and positioning was accom-
help in guiding analysts to better simulations of the turbulence plished with a fixture composed of three mutually perpendicular
phenomena in question. Steiner [7] and Carr, Connor, and Buhr [8] traversing mechanisms. Internal temperature profiles were cor-
obtained profile measurements in flows with strong, probably rected for the temperature increase due to radiation from the tube
dominant, buoyancy forces; unfortunately, neither investigation wall to the sensor. Experimental uncertainties in temperature
documented the inlet and boundary conditions in sufficient detail measurements are shown as brackets in the later figures with the
to use in verifying numerical predictions based on proposed turbu- uncertainty in the radiation correction usually dominating these
lence models. estimates.
The present investigation is aimed at obtaining mean axial and The experiment was designed to approximate a constant heating
radial distributions, for steady, low Mach number, turbulent and rate to air, entering with a fully developed turbulent profile at a
laminarizing flows with dominant forced convection and including uniform temperature. The reduced data show that for q+ < 0.0036
the heating rate remained constant to within ±5 percent for 3.0 <
x/D < 25.3. For higher heating rates there was an axial variation in
1
Numbers in brackets designate References at end of paper. heating rate up to ±12 percent. However, the numerical predic-
Contributed by the Heat Transfer Division and presented at THE tions used for comparison are based on the deduced heat flux dis-
AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Revised manu-
script received by the Heat Transfer Division June 2, 1975. Paper No. 74- tribution rather than on a constant heat flux. Adiabatic velocity
HT-24. profiles show that the flow entering the heated section agreed with

Journal of Heat Transfer NOVEMBER 1975 / 589

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 3.2 8.9 3.2 Table 1 Experimental runs
\ i ^ I 1.6 D Inconel
i ^ -1. -—janf "*-•.—f^-~ s Ente rine parameters Downstream parameters (x/D • 25.3}
-jtL.
Ubel Re 1 q Gr/Re£ Se y .
i m 'linear VISCOUS *c

1.2 D ceramic T91 8520 0.0010 0.064 7900 11 <v30 200 Turbulent
T61 6020 0.0011 0.094 5560 13 75 ISO Turbulent
SIDE VIEW
T62 6030 0.0018 0.16 5290 13 60 130 Turbulent
S64 6020 0.0035 0.31 4740 19 40 83 Subturbulent

S65 6040 0.0045 0.40 4510 25 45 72 Subturbulent

8.9 L46 3760 0.0055 0.78 2680 26 37 53 Laminarizing

Alumel i^is^Criromel

3.2
END VIEW The shapes and trends of these normalized temperature profiles
bear distinct similarities to velocity profiles, u + ( y + ) , observed in
THERMOCOUPLE PROBE external flows laminarizing due to acceleration [12, 13]. It is seen
Fig. 1 Typical thermocouple probe; dimensions in mm that yc+ decreases as the Reynolds number is reduced, as expected
in adiabatic flows, and it also decreases as the heating rate is
raised.
predictions for fully developed, low Reynolds number turbulent The term viscous sublayer is defined variously by different in-
flow within the experimental uncertainty. vestigators. In the present paper the terminology suggested by
Inlet Reynolds numbers range from 3800 to 8500 with dimen- Bradshaw [14] is adopted. Thus, we define the linear sublayer as
sionless heating rates, <j+, to 0.0055. The air inlet was at room tem- the region where molecular effects dominate the mean velocity
perature, about 299 K (538°R). Wall temperatures ranged to 870 K profile to give u+ « y + ; for high Reynolds number, adiabatic flow
(1560°R) while wall-to-bulk temperature ratios up to 2.0 were the linear sublayer typically extends to y+ » 5. The viscous sub-
achieved. Compressible effects have been neglected as the maxi- layer is the region where both molecular effects and turbulent ef-
mum Mach number was 0.017. The buoyancy parameter, Gr/Re; 2 , fects are significant, e.g., the range 5 < y+ < 30 for high Reynolds
varied from 0.06 to 0.78 so forced convection would be expected to number, adiabatic flow. The viscous sublayer then usually includes
dominate but the experiments of Bates, et al. [9] showed that nat- the region of maximum production of turbulence kinetic energy
ural convection effects may be significant in some runs. The profile and its behavior is expected to be important due to that observa-
data were taken at 2.9 < x/D < 25.3. tion alone.
For convenience, in the present study data beyond x/D = 10 and By plotting t+(y+) one may estimate the thickness of the linear
falling within 10 percent of the empirical correlation by Magee and sublayer and the viscous sublayer. With constant properties and a
McEligot [4] are labeled turbulent. Data which agree with the lam- constant wall heat flux, a fully established temperature profile for
inar correlation of McEligot and Swearingen [10] within 10 percent pure laminar flow can be written as
before the end of the test section are classified laminarizing. Data
falling between the two correlations are labeled "subturbulent." = Pr . +\1 + lL_(r)2+l/L)3] (1)
Results of this classification scheme agree with the categories es- Ty 2 rw \r,
l %r w +\rj 4\rJ
+
J
tablished earlier in small tube studies [1]. Tabulated data for all or, in the limit of small distances, t « Pr • y . For a turbulent
experimental runs and further details on the experiment are pre- core, Johnk and Hanratty [15] suggest
sented in Perkins' thesis [11].
t+ = 3.32 + 5.09 In y + (2)

Downstream Temperature Profiles The two limiting relations are presented in Fig. 2 as solid lines;
Rather than attempt a description of all experimental runs, a se- center-line markings indicate ye+ for each run and dashed lines
quence of typical runs has been selected to illustrate the effects of parallel to equation (2) provide guides to identify possible "log
varying both the inlet Reynolds number and the heating rate law" regions. The relation t + = Pr • y+ has been extended to aid in
(Table 1). The data at x/D = 25.3 are presented in Pig. 2 for this determining the edge of the linear sublayer but this point has been
sequence of runs. The local bulk Reynolds number at this position estimated mainly from the apparent inflection point in the data
varies from about 2700 to 7900 so all would be expected to be tur- for each run. The edge of the viscous sublayer is approximated by
bulent if they were adiabatic at the same flow conditions. the juncture with the dashed line. These estimated thicknesses are

.Nomenclature.
/4CS = cross-sectional area q+ = wall heat flux parameter, qw"/(Gcp i-
cp = specific heat at constant pressure y = distance from wall Ti)
D = tube diameter th = eddy diffusivity for heat Re = Reynolds number, GDIp.
G = average mass flux, m/Acs tm ~ eddy diffusivity for momentum t + = temperature, V g ^ p l cp(Tw - T)l
g = acceleration of gravity p. = viscosity a "
gc = dimensional constant c = kinematic viscosity, pip u+ = axial velocity, u/Vgcrw/pw
h = convective heat transfer coefficient, p = density yT = wall distance coordinate,
qw'HTu, - Tm) TW = apparent wall shear stress, —(D/4) y^gcTw/p^/uw
k = thermal conductivity Wdx)\p + GVpgc]
( = mixing length Subscripts
m = mass flow rate Nondimensional P a r a m e t e r s i = inlet
qw" = heat transfer to gas-per-unit area / = apparent friction factor, 2gcTwPm/G2 m = evaluated at local bulk (mixed mean)
rw = tube radius Gr = Grashof number, gg„,"£>4/U2fc;T;) temperature
T = absolute temperature Nu = Nusselt number, hD/k vD = van Driest
x = axial coordinate Pr = Prandtl number, p.cp/k w = wall, evaluated at wall temperature

590 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 In y+. The excessive temperatures in the viscous sublayer may be
Run Re, q+ explained by higher upstream wall heat fluxes for this particular
o T9I 8 520 0.0010 /-
/.^"''L.46 run. At the center line, equation (1) predicts tc+ = 0.75 Pr • y c + «
D T6I 6020 0.0011
27.8, which is less than 5 percent above the data; while this differ-
" ^> T62 6030 0.0018
A S64 6020 0.0036 f ' 1 ence is greater than the estimated experimental uncertainty in the
temperatures, the fact that the position would still be in the devel-
Q S65 6040 0.0045
7 1S oping thermal region prohibits discriminating whether the central
- fc, L4G 3760 0.0055 fc> % ' ! 5 IT62
r A ^A- : T6
L-1 region effectively has molecular or turbulent transport. The signif-
A
t
t = Pry+ ! - '' < kx^>. JK-- icant conclusion is that the mean temperature profile conforms ap-
proximately to laminar predictions over most of the cross section.
A
7 Numerical Predictions
Wo *•— ^Johnk a Hanratty By solving the governing equations numerically, one eliminates
~
<a. ° uncertainties about the analysis other than those concerned with
the turbulence model itself. The numerical method used was the
same as the program of Bates, et al. [9] which includes a body force
A O W ^ term in the x -momentum equation to account for buoyancy ef-
-
fects; otherwise, it corresponds to the approach of Bankston and
McEligot [6]. Reynolds analogy, «/, = em, is assumed.
fcfed
1 For this investigation three turbulent models are used to illus-
100 trate idealized flow behavior for comparison to the mean tempera-
ture distribution. The first is called the laminar model and ac-
Fig. 2 Downstream temperature profiles; conditions listed In Table 1 counts only for molecular transport by a Newtonian fluid but uses
the entering turbulent velocity profile as the initial condition; its
predictions represent a hypothetical limiting condition where lam-
also listed in Table 1. inarization is immediate and complete.
At low y+ each run converges on the linear sublayer prediction The second model is the "van Driest-wall properties" mixing
with the exception of a few points in the immediate vicinity of the length model,
wall. All show a possible turbulent core where t+ varies as In y + ; as
with pipe flow velocity profiles, this logarithmic variation appears ivB = Ky[l - exp(-y+/y(+)] (3)
at large y/rw despite the lack of a constant heat flux, analogous to a +
constant shear layer [14]. Two, or perhaps three, runs appear to with properties in y being evaluated at the wall temperature.
show a constant temperature near the center line implying that the This model has been used with success for fully turbulent flows [6]
thermal boundary layer may not fill the tube at this station; this and is used in the present investigation to illustrate turbulent be-
observation will be discussed further in a later section. Although havior.
this figure does not seem to lead to a better understanding of the A third model, proposed by McEligot and Bankston [16] ac-
core region, the data do confirm the numerical conclusion of McEl- counts for variable viscous sublayer thickness at low Reynolds
igot and Bankston [16] that thickening of the viscous sublayer pro- numbers, for gas property variation, and for delay in the response
vides a plausible explanation of laminarization in terms of wall pa- of turbulent quantities to local conditions. In the present paper we
rameters. refer to it as the modified van Driest model. Details are available
in references [11,16].
It is interesting that none of the data agree with the "universal
correlation" suggested by Johnk and Hanratty [15]. For this range
Thermal Development
of Reynolds numbers it seems apparent that there is a strong de-
pendency on heating rate. Even the fully turbulent, nearly con- Numerical predictions are compared to data from the three
stant property run (T91) shows a 10 percent increase in t+ over the types of runs—turbulent, subturbulent, and laminarizing—in Figs.
higher Reynolds number correlation suggested by Johnk and Han- 3-5.
ratty. It should be noted that this disagreement is not an entry Turbulent Flow. The developing temperature profiles of Run
length effect since a shorter entry would result in a lower dimen- T62 are shown on Fig. 3 as representative of all three turbulent
sionless temperature profile. For y+ > 13 the three turbulent runs runs. Downstream, the data demonstrate typical turbulent tem-
diverge from purely laminar behavior and show an extensive tur-
bulent core region, although the data are 10-20 percent higher
than predicted by Johnk and Hanratty.
Only when the subturbulent regime is reached does the linear
sublayer thicken substantially. For Run S64, the linear sublayer
extends to near y+ = 19. The conclusion here is that although Run
S64 no longer appears predominantly turbulent it just as surely
has a large region away from the wall which disagrees with a purely
laminar prediction. Run S65, which might be characterized as al-
most laminarized, shows further increase in the linear sublayer to T/Tj |.5
y+ =i 25, but there is still a large core (25 < y + < yc+ = 72 or two-
thirds of the radius) which demonstrates nonlaminar behavior.
Whether Run L46 has been laminarized to completely molecular
transport by x/D = 25.3 is difficult to discern with only measure-
ments of the mean temperature profile. To examine this question,
equation (1) has been plotted as a dotted curve for the conditions
of run L46. For this run estimation of the linear sublayer thickness
by identifying an inflection point is not conclusive since equation
(1) also shows an inflection in this region. In the region from y+ «
10-40 the normalized temperature data are higher than this purely Fig. 3 Thermal development of turbulent run T62; Re( = 6030, q,+
laminar prediction then beyond y + = 40 the data appear to vary as 0.0018, Re 2s = 5290

Journal of Heat Transfer NOVEMBER 1975 / 591

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 perature profiles with the temperature gradient increasing toward (Fig. 4). At the first station, x/D = 2.9, the data show a slightly
the wall, thereby showing that most of the thermal resistance is thicker thermal boundary layer than predicted and the profiles
concentrated near the wall. Since Tc > T; at the second station it still appear as a boundary layer at the exit. However, the center-
is concluded that the thermal boundary layer has penetrated to line temperature shows that the boundary layer has again filled
the center in less than 14 dia. Thereafter, the temperature across the cross section within 14 dia. Although the heating rate is more
most of the tube shows a flat, well-mixed core region. than twice that of the turbulent runs, the center-line temperature
In Fig. 3 it is also seen that the data are predicted well by both is approximately the same at the exit.
the van Driest-wall model and the modified van Driest model with For the subturbulent runs the modified van Driest model and
the modified version being slightly better, particularly in predict- the laminar model provide predictions which differ near the en-
ing the wall temperature. The van Driest-wall model underesti- trance but which converge on one another successively down-
mates the temperatures near the wall. The hypothesis of purely stream. The data correspond to the modified model at the entry
molecular transport, the laminar model, leads to distribution of then begin to diverge, beginning near the wall at about 8 dia. Near
the thermal resistance across a greater portion of the tube, but for the exit both models overpredict the temperatures in the impor-
all three runs the predicted thermal boundary layer only reaches tant wall region. In comparison to the laminar model, the numeri-
half the radius by the downstream location. The laminar predic- cal results from the modified van Driest model imply slightly fast-
tions differ from the data significantly with the discrepancy great- er thermal boundary layer growth due to turbulent transport but
er near the wall in the other two turbulent runs. neither model shows it penetrating to the center line while the
Subturbulent Flow. The two subturbulent runs show the data do. The modified model predicts essentially laminar trans-
same trends as each other so only Run S65 is presented graphically port for much of the tube while the data indicate that turbulent

25.3 Fig. 4 Thermal development of subturbulent run S65;

Re, = 6040, q,+ = 0.0045, Re25 = 4510

:
=~?253

Fig. 5 Thermal development of laminarizing run L46;

Re, = 3760, q,+ = 0.0055, Re25 = 2680 ~

592 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 transport persists to a greater extent. Conclusions
At 25 dia the van Driest-wall model, representing fully turbulent From these comparisons, it appears that the modifications to
flow, underpredicts the wall temperature more than the modified the van Driest mixing length model suggested by McEligot and
van Driest model overestimates it. Consequently, for applications Bankston [16] lead to substantially correct predictions of heat
where the heating rate is controlled as in gas cooled nuclear reac- transfer during laminarization of a gas due to heating. The viscous
tors, the modified van Driest model could be considered conserva- sublayer thickness is seen to depend on both the local Reynolds
tive and the van Driest-wall model dangerous in the subturbulent number and the heating rate. It is, however, evident that the modi-
flow regime. fied model needs further improvement or replacement for applica-
Laminarizing Flow. The thermal development of the lami- tion to the subturbulent regime.
narizing run, L46, is demonstrated in Pig. 5. The heating rate, q+, The data indicate that there is a substantial region of flow and
is about three times that of the turbulent run in Fig. 3. Compared heating rates where turbulent transport is suppressed though com-
to predictions, the measurements show a thicker thermal boundary plete laminarization does not occur. On the other hand, for condi-
layer which again fills the tube within 14 dia and possibly within 9 tions where local Nusselt numbers agree with laminar predictions
dia. The thermal resistance is spread across the tube downstream at Reynolds numbers above the accepted value for transition, the
as in laminar flow rather than being concentrated near the wall as viscous sublayer occupies the majority of the tube cross section.
in turbulent flow. The downstream temperature profile appears al-
most parabolic (equation (1) corresponds to a fourth-order parabo-
la). The thermal resistance is much greater than in the turbulent Acknowledgments
regime since the center-line temperature is actually less while the This study has been supported by the National Science Founda-
heating rate is three times as great; in contrast, constant properties tion. The University Computer Center provided the CDC-6400
correlations would suggest an increase of only about 7 percent computer for the numerical calculations. Prof. R. B. Kinney super-
based on the local Reynolds numbers and bulk temperatures. vised the early stages of the experimental program.
Application of the modified van Driest model to the laminarized
run yields results which are in close agreement with the measure-
ments throughout the tube. The laminar model also agrees closely References
though slightly diverging from the data near the wall proceeding 1 Perkins, K. R., Schade, K. W., and McEligot, D. M, "Heated Relami-
downstream; it predicts that the thermal boundary layer would not narizing Gas Flow in a Square Duct," International Journal of Heat and
quite reach the center by 25 dia. At the first station both models Mass Transfer, Vol. 16, 1973, pp. 897-916.
predict the same temperature distribution since the thermal 2 Hall, W. B., Jackson, J. D., and Watson, A., "A Review of Forced
boundary layer still appears to be within the viscous layer; then a Convection Heat Transfer to Fluids at Supercritical Pressures," Proc. Inst.
Mech. Eng., Vol. 182, Part 31,1967-1968, pp. 10-22.
slight difference appears due to the small turbulent transport still 3 Shiralkar, B., and Griffith, P., "The Effect of Swirl, Inlet Conditions,
implied in the modified van Driest model. It is apparent that, even Flow Direction and Tube Diameter on the Heat Transfer to Fluids at Super-
at the high heating rate of Run L46, laminarization does not occur critical Pressure," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Se-
instantaneously and the inherent delay in decay of the turbulent ries C, Vol. 92,1970, pp. 465-474.
4 Magee, P. M., and McEligot, D. M., "Effect of Property Variation on
behavior results in a slightly lower wall temperature than predict- the Turbulent Flow of Gases in Tubes: The Thermal Entry," Nuc. Sci. Eng.,
ed by the laminar model. Again predictions based on the fully tur- Vol. 31,1968, pp. 337-359.
bulent model are seen to be dangerously misleading. 5 Perkins, H. C, and Worsoe-Schmidt, P. M., "Turbulent Heat and
Momentum Transfer for Gases in a Circular Tube at Wall-to-Bulk Temper-
Buoyancy Effects. The turbulence models applied do not di- ature Ratios to Seven," International Journal of Heat and Mass Transfer,
rectly account for an effect of buoyancy although the x -momen- Vol. 8,1965, pp. 1011-1031.
tum equation includes a body force term. Bates, et al. [9] found 6 Bankston, C. A., and McEligot, D. M., "Turbulent and Laminar Heat
that buoyant body forces can cause significant modification of heat Transfer to Gases with Varying Properties in the Entry Region of Circular
Ducts," International Journal of Heat and Mass Transfer, Vol. 13, 1970,
transfer behavior for forced turbulent flow and significant diver- pp. 319-344.
gence from numerical predictions. Since they used the same nu- 7 Steiner, A., "On the Reverse Transition of a Turbulent Flow Under
merical program as the present study it appears that the turbu- the Action of Buoyancy Forces," J. Fluid Mech., Vol. 47,1971, pp. 503-512.
lence model should include provision for buoyancy effects when 8 Carr, A. D., Connor, M. A., and Buhr, H. O., "Velocity, Temperature
important. and Turbulence Measurements in Air for Pipe Flow With Combined Free
and Forced Convection," JOURNAL OF HEAT TRANSFER, TRANS.
Reviewing the data of Bates, et al., one finds only a slight change ASME, Series C, Vol. 95,1973, pp. 445-452.
in wall temperature for variation of |Gr/Re,-2| from 0.076 to 1.05 in 9 Bates, J. A., Schmall, R. A., Hasen, G. A., and McEligot, D. M., "Ef-
a rapidly laminarizing flow such as Run L46, and the trends were fects of Buoyant Body Forces on Forced Convection in Heated Laminarizing
Flows," Heat Transfer 1974, Vol. II, Proceedings, 5th International Heat
predicted well. This observation would correspond to a situation Transfer Conference, 1974, pp. 141-145.
where the thermal boundary layer remains well within the viscous 10 McEligot, D. M., and Swearingen, T. B., "Prediction of Wall Temper-
sublayer so that molecular transport dominates and, consequently, atures for Internal Laminar Heat Transfer," International Journal of Heat
further modification of the turbulence model is not necessary. The and Mass Transfer, Vol. 9,1966, pp. 1151-1152.
present turbulent data are at lower values of IGr/Re,-2! than those 11 Perkins, K. R., "Turbulence Structure in Gas Flows Relaminarizing
by Heating," PhD thesis, University of Arizona, 1975.
where the turbulent results of Bates, et al., were affected so the 12 Blackwelder, R. F., and Kovazsnay, L. S., "Large Scale Motion of
present agreement with the predictions is not surprising. The Turbulent Boundary Layers During Relaminarization," J. Fluid Mech., Vol.
subturbulent regime in the present study corresponds to the "slow 53, 1972, pp. 61-83.
laminarization" of Bates, et al., where moderate buoyancy effects 13 Badri Narayanan, M. A., and Ramjee, V., "On the Criteria for Re-
verse Transition in a Two-Dimensional Boundary Layer Flow," J. Fluid
were noticed at |Gr/Re,-2| « 0.23; thus, the discrepancy between Mech., Vol. 35,1969, pp. 225-241.
prediction and measurement in Run S65 (and S64) may be, in 14 Bradshaw, P., An Introduction to Turbulence and its Measurement,
part, due to lack of provision for buoyancy in the present turbu- Pergamon,-Oxford, 1971, p. 51.
lence models. Neither of these investigations has been exhaustive 15 Johnk, R. E., and Hanratty, T. J., "Temperature Profiles for Turbu-
lent Flow of Air in a Pipe," Chem. Eng. Science, Vol. 17,1962, pp. 867-892.
so resolution of this difficulty is an appropriate topic for further 16 McEligot, D. M., and Bankston, C. A., "Numerical Predictions for
study. Circular Tube Laminarization by Heating," ASME Paper No. 69-HT-52.

Journal of Heat Transfer NOVEMBER 1975 / 593

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 R. N. Smith 1
R. Greif Turbulent Transport to a Rotating
University of California,
Department of Mechanical Engineering,
Berkeley, Calif. Cylinder for Large Prandtl or
Schmidt Numbers
The heat or mass transport to a rotating cylinder is obtained by solving the conservation
equations for large Prandtl or Schmidt numbers. A modified mixing length, which in-
cludes the effect of centrifugal forces in terms of the Richardson number, is used. Explic-
it relations are presented for the heat or mass transport and a comparison is made with
the experimental data over a range of Prandtl and Schmidt numbers from 190 to 11,000.

Introduction od by Kappesser, Cornet, and Greif [9] for a value of the Schmidt
number of 400.
The rotating cylinder geometry has long been of interest in the All of these results have been found to correlate well with the
study of heat, mass, and momentum transport, and analytical and equation [7]
experimental results have been published for a wide variety of con-
Nu Sh /f\
ditions. The present paper deals with the prediction of the heat or
mass transfer to a rotating cylinder in turbulent flow for large
PJ.0.356 SC0356 © Re (1)

values of the Prandtl or Schmidt number. Under these conditions for 200 < Re < 300,000 and 0.7 < Pr < 11,000. The friction factor,
the significant variation of the temperature or concentration is /, is defined by
confined to a region very close to the wall, with respect to the vari- T0
ation of the velocity, yielding simplifications in the analysis. L- (2)
2 ~ p(r0a>)2
Experimental results at lower values of the Prandtl number (for
heat transfer in air, Pr ~ 0.7) have been reported by several inves- For smooth cylinders, Theodorsen and Regier [10] found that the
tigators which are in good agreement with one another [1-4]. 2 Heat relation
transfer results at higher Prandtl numbers were reported by Seban
and Johnson [5] for a horizontal cylinder rotating in oil (Pr = — = -1.825 + 4.07 log 10 (/ l/2 Re) (3)
190-660) and in water (Pr = 3-5) and by Becker [6] (Pr = 3).
Eisenberg, Tobias, and Wilke [7] used solid dissolution of benzoic was an accurate representation of their measured values of the
acid and cinnamic acid rotating in water, as well as an electro- drag on a rotating cylinder.
chemical technique, to obtain mass transfer rates for rotating cyl-
inders for Schmidt numbers varying from 800 to 11,000. The elec- Transport to a Rotating Cylinder
trochemical system- used was the ferro-ferricyanide red-ox couple To obtain analytical results for the heat or mass transfer to a ro-
in NaOH solution and measurements were made of the mass- tating cylinder, and provide a basis for modifications, we begin
transfer-limiting current associated with these reactions. The lim- with the basic conservation equations. The mean equation for the
iting current method was also used by Cornet and Kappesser [8], conservation of momentum in the circumferential direction, d, is
who measured limiting current rates of the oxygen reduction reac- given by [11, 12]
tion in NaCl solution in a rotating cylinder system. The Schmidt
number was 370 for these experiments. The mass transfer to a — (u'v'r V C4)
rough rotating cylinder was obtained by the limiting-current meth- ~"drl dr\r)\ dr
This equati on may be integrated to yield
1 3 d JU\ — 2
Presently, Asst. Professor, Rice University, Houston, Texas. - const = r0ro (5)
2
Numbers in brackets designate References at end of paper. dr \r/
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer where TO is the wall shear stress. The mean equation for the con-
Division March 31,1975. Paper No. 76-HT-D. servation of energy about a rotating cylinder is given by

§94 / NOVEMBER 1975 Copyright © 1975 by ASME Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 d_ > dT with K3 = 4.16 X 10-"1. A similar relation has been obtained by Pa-
0 =: kr pc„ ru
• dr v ar *) m terson and Greif [18, 19] for flow about a rotating disk and by Pa-
terson [18] for flow about a rotating cylinder. Since the analysis for
Introducing the turbulent diffusivity for heat, CH, according to u't' the rotating disk has been published we shall refrain from repeat-
= -etifdT/dr)3 and integrating yields ing that for the rotating cylinder. However, it is important to note
dT that the details for the rotating cylinder problem do prove to be
-qr = r (k + pcpm) = const = -Qoro (7) more complex than those for the rotating disk.
dr
Using equation (14) for the eddy diffusivity and carrying out the
We now specify « according to
corresponding integration yields the following result for the heat
p tH = i!±-^l d_ /U- transfer:
v v v dr (T)-«*^£) m 3K3 1/2 Re / P r \ V 3 / / \ i / 2
and obtain
T0-T
Nu =
2TT O &
The results obtained from equations (12) and (15) are plotted in
(15)

T+(r+) =
qo/pcpur Figs. 1 and 2 along with the heat and mass transfer data from ref-
erences [5, 7, 8]. The range of the values of the Prandtl and

f>r0* r+
r<T
i / p r + [(e+)ypTt]r+d/dr+
<JrQ (U+/r+)
dr+
(9) Schmidt numbers is from 190 to 11,000. It is seen that neither
equation is in agreement with the heat or mass transfer data. Re-
For large values of the Prandtl numbers the essential temperature call, however, that both of these models were successful in predict-
changes occur in the region near the wall. For this condition we fol- ing the mass transfer for other rotating systems; for example, for
low Deissler [13] and obtain disks [20, 19] and for large angle cones [21]. It thus appears that
there may be a fundamental difference in the turbulence transport
dr+ which makes the foregoing models inadequate for the rotating cyl-
Tb+ SL T+(r °) %)ro+ .l / P r + (^+) 2 /Pr t no) inder.
4
We now specify the mixing length by the relation [13]
Modifications
(+ = n2(r+ - r 0 + ) 2 , n = 0.124 (11) In the rotating disk studies, the primary direction of transport is
and integrate equation (10) to obtain perpendicular to the planar surface of the disk while the centrifu-
gal forces are parallel to the surface and hence may be neglected.
2r 0 + P 4n2l/2r0+ / Pr \ V* 2 n 2 ^ R e / J P £ \ i/4 / / y / 2 However, for the rotating cylinder these effects are in the same di-
Nu<
n* IT \Pr(/ IT VPr,/ (i)' rection; namely, the radial direction, and a contribution to the tur-
bulent mixing may be anticipated.
(12)
This result was obtained by Deissler [13] for tube flows and used In his discussion of the turbulent heat or mass transport to a ro-
by Seban and Johnson [5] in their study of heat transfer from a ro- tating cylinder Kappesser [22] suggested that the turbulent mixing
tating cylinder. It should be noted that the friction factor, /, speci- was enhanced by centrifugal forces. He proposed a "centrifugal
fied in equation (3), may be approximated by [7] eddy diffusivity" which he calculated by considering a fluid parti-
cle moving in a centrifugal force field. Another approach for in-
f/2 = 0.079 R e - (13) cluding the effect of rotation is to alter the mixing length by intro-
0 85 ducing the swirling flow Richardson number, Ri. The basis for this
which results in a Re - dependence for the Nusselt number. approach is presented by Bradshaw [23] who developed an analogy
Another approach for the determination of the transport was in- between the effects of curvature or rotation and buoyancy on tur-
troduced by Tien, Wasan, and Wilke [15-17]. They obtained an bulent mixing. He obtained the following relation for the modified
expression for the eddy diffusivity by expanding the velocities near or apparent mixing length, lc, for small buoyancy effects:
the surface in a Taylor series. The resulting expressions were then
required to satisfy the conservation equations and the boundary Cc = t(l - 0Ri) (16)
conditions at the surface. Their result for nonrotating flows was
given by where (3 is an empirical constant. Hughes and Horlock [24] have
discussed this result and it has also been utilized by Koosinlin and
- = K3(r+ - r 0 +J 3 + . (14) Lockwood [25] and Koosinlin, Launder, and Sharma [26], who rec-
ommend a value of /5 = 5. We shall follow this procedure in the
present study.
For a rotating cylinder, Ri is given by
(2U/r2)d/dr(Ur)
3 (17)
4
Note that u'v' = -imr d/dr(U/r).
Note that in the region near the wall this result is identical to that pre-
m- [rd/dr(U/r)f
sented by Van Driest [14]. The Richardson number represents the ratio of the energy avail-

-Nomenclature-
cp = constant pressure specific heat r+ = dimensionless radial coordinate, ruT/i> u',u' = turbulent fluctuating velocity com-
/ = friction factor, equation (2) Re = Reynolds number, 2rg2u>/v ponents
k = thermal conductivity Ri = Richardson number eu = eddy diffusivity for energy transfer
I = mixing length Sc = Schmidt number CM = eddy diffusivity for momentum trans-
l+ = dimensionless mixing length, iurlv Sh = Sherwood number fer
lc = modified mixing length T = temperature H = viscosity
n = constant, equation (11) t' = turbulent fluctuating temperature ;< = kinematic viscosity
Nu = Nusselt number T+ = dimensionless temperature, equation p = density
Pr = Prandtl number (9) T = shear stress
Prt = turbulent Prandtl number U = velocity
q = heat flux U+ = dimensionless velocity, U/uT Subscript
r = radial coordinate ur = friction velocity, (ro/p) 1 '' 2 0 = evaluated at wall

Journal of Heat Transfer NOVEMBER 1975 / 595

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 M l 1 1 1 1 | 1 I I I1 ' the Richardson number correction and the experimental data, are
o Cornel and Koppesser [8] (Sc = 370) plotted in Figs. 1 and 2, respectively. The constant /3 was set equal
- A Sebon and Johnson [5] (190 < Pr < 660) ><r o — to 5 as recommended by Koosinlin, Launder, and Sharma [26] and
- o Eisenberg, Tobias ond Wilke [7] (800<Sc<ll,000) r J& a „
a Pr ( was taken to be unity. In each case the agreement with the
&h
Eq. (12)
- data is substantially improved over the unmodified result.
A brief comment on the indicated dependence of the experimen-
tal data on the Prandtl or Schmidt number is in order at this
10' — point. As noted in the introduction, the experimental data have
__ '^F? —
a J?/ been correlated by Eisenberg, Tobias, and Wilke [7] with an equa-
- - tion using a Pr 0 - 366 (or Sc 0 3 5 6 ) dependence. Equation (22), which
CD D-> X /
has a Pr 1 / 3 (or Sc 1/3 ) dependence, is also in good agreement with
/ ~~ the experimental data and is shown plotted in Fig. 2. From Figure
10 —
/ 1, however, we see that the presentation of the data with a Pr 1 / 4 (or
~1 Sc 1/4 ) dependence is less successful and we, therefore, conclude
1 /
/
l l 1 1 1 1 1 11 1 I I I 1 i
_. that equation (22) is the more desirable result.
In general, Figs. 1 and 2 show that modifications to nonrotating
analyses to account for the effects of centrifugal forces may be
used with success when applied to rotating cylinder transport phe-
Fig. 1 Results for the heat and mass transfer to a rotating cylinder
nomena. Recall that the results presented here follow from an
analysis for large values of the Prandtl or Schmidt number, so that
the primary resistance to heat or mass transfer occurs in the sub-
able to a fluid particle due to stratification of body forces to that layers. The linear buoyancy modification of Bradshaw is derived
available in the form of turbulent kinetic energy. When the Rich- from studies of atmospheric boundary layers (cf. Lumley and Pa-
ardson number is negative, as in the case of a rotating cylinder, en- nofsky [28]). Its physical basis is in a region far enough from a sur-
ergy is added to the fluid particle as it moves in the stratified force face so that velocity fluctuations are sufficiently large to be af-
field and an enhancement of the turbulent mixing results. In this fected by the stratification of body forces. Application of the
study we consider large values of the Prandtl (or Schmidt) number model to the region within the viscous sublayer must therefore be
and therefore approximate the Richardson number by its value at viewed cautiously and the present results should not be viewed as
the surface of the cylinder, Rio, where a verification of the manner in which the centrifugal forces affect
the turbulent flow field in the sublayers. However, it is important
-4 /2\
Rio = ; (18) to emphasize that the present analysis is a direct extension of pre-
Re <7) vious work [23, 26] and no new empirical constants, etc., have been
The modified mixing length is then given by introduced.
In closing, we note that it would be desirable to extend the theo-
(19) ry developed here to lower values of Pr or Sc, or alternatively to
\ Re //
treat more completely the large Pr or Sc problem, by considering
Hence, the effect of the modification is to increase the mixing the local variation of the Richardson number. The momentum and
length, and the eddy diffusivity, and ultimately the predicted heat energy equations may then be integrated numerically to determine
or mass transport. Note that the use of equation (19) implies that the velocity and temperature profiles near the wall and to illus-
the linear buoyancy result of Bradshaw, as presented in equation trate more clearly the effect of the mixing length correction. The
(16), is applicable within the sublayers. This assumption will be work of Koosinlin and Lockwood [25] and Koosinlin, Launder and
discussed shortly. Sharma [26] on rotating cones and axisymmetrie flow over rotating
We may now model the effect of the centrifugal forces by speci- cylinders is cited in this regard along with the study by Kinney
fying the new mixing length, tc, and proceed as before to obtain [27].
the transport, the result for the Nusselt number is then given by
2rt2^2//\i/2 /Pry/" / 4£ 2\i/2
2\i Acknowledgment
NU: e + (20) The authors acknowledge with appreciation the support of this
TT \2/ \Pr(/ \ RrJ7/
research by the Sea Water Conversion Laboratory of the Universi-
which for /J = 0 reduces to equation (12). ty of California, Berkeley.
It is also of interest to utilize the method proposed by Tien,
Wasan, and Wilke, in conjunction with the Richardson number
modification of the mixing length, to obtain the transport. The
eddy diffusivity obtained from this method is not directly based on 1 M M 1 II 1 I I I
1 16
a specification of the mixing length, but near the wall we obtain ° Cornet ond Koppesser [8] (Sc = 370)
A Sebon ond Johnson [5] (190 < Pr < 660) -,
<=M,c __ u Eisenberg, Tobias and Wilke (7] (800<Sc<ll,000) ^
Kc3(r+ - r0+)3 = (tc+)2 = (i+)2 ( l + ~ " 7) (21)
tq. \i.L\ AFr
\ Re / /
with ! I0 2 — Eq. (15) ayy —
_
£ ^ 9 ^

K3 = K3
>(/i <#"/
\ Re / / o
- O0S"V
^ / -
™ £ # * /
With this relation for the eddy diffusivity the result for the Nus- & /
cwS?
selt number becomes >
z
,n
10 /
-
4/3 2\2/3 ' -a /
Nu: (22) / ~
•K \2/ \Prt/ V Re // /
which for /3 = 0 reduces to equation (15). 1 /
I i l l 1 1 1 11 1 I I I 1 i

Discussion of Results and Conclusions Re =

Equations (20) and (22), the transport predictions which include Fig. 2 Results for the heat and mass transfer to a rotating cylinder

5S6 / NOVEMBER 1975 Transactions of ihe ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 References 14 van Driest, E. R., "On Turbulent Flow Near a Wall," J. Aero. Sci.,
Vol. 23,1956, p. 1007.
1 Anderson, J. T., and Saunders, O. A., "Convection Prom an Isolated 15 Tien, C. L„ and Wasan, D. T., "Law of the Wall in Turbulent Chan-
Heated Horizontal Cylinder Rotating About Its Axis," Proceedings Royal nel Flow," Physics of Fluids, Vol. 6, Series 1,1963, p. 144.
Society London, England, Series A, Vol. 217,1953, p. 555. 16 Wasan, D. T., Tien, C. L., and Wilke, C. R„ "Theoretical Correlation
2 Etemad, G. A., "Free Convection From a Rotating Horizontal Cylin- of Velocity and Eddy Viscosity for Flow Close to a Pipe Wall," AIChE Jour-
der to Ambient Air With Interferometric Study of the Flow," TRANS. nal, Vol. 9, No. 4,1963,
ASME, Vol. 77,1955, p. 1283. 17 Tien, C. L., "A Note on the Distributions of Temperature and Eddy
3 Dropkin, D., and Carmi, A., "Natural Convection Heat Transfer Diffusivity for Heat in Turbulent Flow Near a Wall," J. App. Math, and
From a Horizontal Cylinder Rotating in Air," TRANS. ASME, Vol. 79, Physics (ZAMP), Vol. 15, No. 1,1964, p. 63.
1957, p. 741. 18 Paterson, J. A., "Heat, Mass and Momentum Transport in Rotating
4 Kays, W. M., and Bjorklund, I. S., "Heat Transfer From a Rotating Flows," PhD thesis, University of California, Berkeley, 1973.
Cylinder With and Without Crossflow," TRANS. ASME, Vol. 80, 1958, p. 19 Paterson, J. A., and Greif, R., "Transport to a Rotating Disc in Tur-
70. bulent Flow at High Pr or Sc," JOURNAL OF HEAT TRANSFER,
5 Seban, R. A., and Johnson, H. A., "Heat Transfer From a Horizontal TRANS. ASME, Series C, Vol. 95,1973.
Cylinder Rotating in Oil," NASA Memorandum 4-22-59W, 1959. 20 Ellison, B. T., and Cornet, I., "Mass Transfer to a Rotating Disc,"
6 Becker, K. M., "Measurements of Convective Heat Transfer From a Electrochemical Sci., Vol. 118,1971, p. 68.
Horizontal Cylinder Rotating in a Tank of Water," International Journal of 21 Kappesser, R., Greif, R., and Cornet, I., "Mass Transfer to Rotating
Heat and Mass Transfer, Vol. 6,1963, p. 1053. Cones," Appl. Sci. Res., Vol. 28,1973, p. 442.
7 Eisenberg, M., Tobias, C. W., and Wilke, C. R., "Mass Transfer at 22 Kappesser, R., "A Study of Heat and Mass Transport From a Rotat-
Rotating Cylinders," Chem. Eng. Prog. Symp. Series, Vol. 51, No. 16, 1953, ing Cylinder," PhD thesis, University of California, Berkeley, 1970.
p. 1; also Eisenberg, M., "Studies of Rates of Solid Dissolution and of Elec- 23 Bradshaw, P., "The Analogy Between Streamline Curvature and
trodes Reactions at Rotating Cylindrical Bodies," PhD dissertation, Univer- Buoyancy in Turbulent Sheat Flow," J. Fluid Mech., Vol. 36, Part 1, 1969,
sity of California, Berkeley, 1953. p. 177.
8 Cornet, I., and Kappesser, R., "Cathodic Protection of a Rotating 24 Hughes, D. W., and Horlock, J. H., "Effect of Rotation on the Devel-
Cylinder," Trans. Inst. Chem. Eng., Vol. 47,1969, p. 194. , opment of the Turbulent Boundary Layer," Paper 18, B78, 1971, Sympo-
9 Kappesser, R., Cornet, I., and Greif, R., "Mass Transfer to a Rough sium on Internal Flows, University of Salford, Salford, England.
Rotating Cylinder," J. Electrochem. Soc, Vol. 118, No. 12,1971, p. 1957. 25 Koosinlin, M. L., and Lockwood, F. C , "The Prediction of Axisym-
10 Theodorsen, T., and Regier, A., "Experiments on Drag of Revolving metric Turbulent Swirling Boundary Layers," AIAA Journal, Vol. 12, No. 4,
Disks, Cylinders, and Streamline Rods at High Speeds," NACA Rep., 1944, 1974, p. 547.
p. 793. 26 Koosinlin, M. L., Launder, B. E., and Sharma, B. I., "Prediction of
11 Modern Developments in Fluid Dynamics, S. Goldstein, ed., Oxford Momentum, Heat, and Mass Transfer in Swirling, Turbulent Boundary
University Press, Oxford, Volume I, 1952. Layers," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Series C,
12 Townsend, A. A., "The Structure of Turbulent Shear Flow," Cam- Vol. 96, May, 1974, p. 204.
bridge University Press, Cambridge, 1956. 27 Kinney, R. B., "Universal Velocity Similarily in Fully Turbulent Ro-
13 Deissler, R. G., "Analysis of Turbulent Heat Transfer, Mass Transfer tating Flows," J. Applied Mechanics, Vol. 89,1967, p. 437.
and Friction in Smooth Tubes at 'High Prandtl and Schmidt Numbers," 28 Lumley, J. L., and Panofsky, H. A., The Structure of Atmospheric
NACA Report 1210,1955. Turbulence, Interscience Publishers, New York, 1964.

Journal of Heat Transfer NOVEMBER 1975 / 597

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 M. S. Sahota
P. J. Pagni Temperature Fields in Structural
Mechanical Engineering Department
University of California,
Berkeley, Calif.
Elements Subject to Fires
The transient, two-dimensional temperature field in a rectangular structural element
subject to time dependent, nonlinear mixed boundary conditions has been obtained with
a minimum of numerics. General results are given for two extreme fire histories, the
American Society for Testing and Materials E-119 time-temperature curve and a short-
duration high-intensity time-temperature curve. Comparisons are made with finite ele-
ment and experimental temperature fields in a reinforced concrete column; good agree-
ment is obtained. The stress field generated from the analytic temperature field also
shows good agreement with finite element calculations. Applications to structural fires
are discussed.

Introduction boundary conditions. Bresler and co-workers [3-5] have ap-

proached this problem, along with the internal stress problem,
An important unsolved problem in fire safety is determining the using the finite element method. Comparisons with their numeri-
structural integrity of a building after a severe fire. Not all perti- cal solutions and experimental data are given here. In some cases,
nent damage is visible, and so one needs to predict the internal in particular at long times for simple geometries with constant
stresses which result from any given fire. The first step is to pre- properties and a small Stefan number, this quasi-analytic tech-
dict the temperature field a known fire produces in a structural el- nique is faster and more convenient.
ement. This temperature field then determines the thermal expan- Few exact solutions for heat conduction with nonlinear bound-
sions which in turn generate internal stresses. ary conditions have been found [6-7]. Approximate solutions for
The problem considered here is a rectangular column surround- nonlinear problems have been obtained [8-9]. Chambre [10] pre-
ed by a known fire with an external bounding surface as shown in sents an elegant solution for the surface temperature in a semi-in-
Fig. 1. Since most structural damage occurs after flashover, only finite solid subject to an unsteady, nonlinear boundary condition.
this time period will be considered. As in the standard test meth-
ods [1-2], l the fire is described by a specified gas phase tempera-
ture history, assumed uniform over the entire column surface. The
interaction between the fire and the column is described by a view
factor, total emissivities of the fire and the column surface, and a
convective heat transfer coefficient, all assumed constant and uni-
form over the column surface.
It is assumed that the column is homogeneous with constant and
known properties. These simplifications are introduced to facili-
tate a solution. Moisture and cracking in concrete columns con-
strain their applicability. However, the results obtained here indi-
cate that appropriate average thermal properties may be deter-
mined from subsequent comparison with experiment. The un-
known variable is the unsteady, two-dimensional temperature field
within the column, subject to nonlinear boundary conditions. The
solution is obtained in quasi-analytic form which reduces to the
standard analytical expressions for the limiting case of linear

1
Numbers in brackets designate References at end of paper.
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer Fig. 1 Cross section of the column showing the coordinates and compo-
Division May 2,1975. Paper No. 76-HT-A. nents of the system considered

598 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Several studies of nonlinear boundary conditions for one-dimen- The obvious basis for equation (7) is the known solution to the
sional solids are available in the literature [11-14], corresponding linear problem whose eigenvalues are given by
The following section describes the general model and problem equation (8). Even though the set of equations (l)-(6) is not a reg-
formulation. The analysis section delineates the solution method ular Sturm-Liouville problem, the trigonometric functions in the
and examines special cases. The remaining sections contain results assumed solution remain orthogonal, and application of orthogon-
and comparisons with experiment. ality to equation (7) yields

Analysis nCmn(t) = / / ®(x, V, /) COS pX COS f/_V dxdy (9)

The system considered is shown schematically in Fig. 1. The ' 0 -'o
analysis begins at flashover when the fire begins to significantly af- where
fect the structure and the approximations of a uniform fire with
constant interaction coefficients becomes reasonable. The effects •_ = 1<> + ^ m + *%.) no)
of the preflashover fire are accounted for by an arbitrary initial
temperature distribution in the column To(x, y). The spatial uni- Differentiating equation (9) and substituting for dO/dt from the
formity of the fire temperature, T/(t), introduces quadrant sym- governing equation (1), and integrating the results by parts, one
metry in a coordinate system at the center of the column with zero obtains
flux conditions at the two internal boundaries.
1 dC
With the foregoing assumptions, the transient temperature field an + e 4>mF(t) - GJO (11)
ym dt
in the solid is described by the following nondimensional equa-
tions: where
2 2 r™ = A,2 + <ij/c2 (12)
ae i_ a e _ ag_
2 1 (1) sin
(13)
9? c Jy ~ at 4>nm = P„ s i n 1mA<l>m,Pnqm)
and
(x,y,0) = 0Q(x,y) (2)
St
[cos pnf e4(l,y,0cos qmydy
^.(0,:v,/) = 0 (3)
ox c o s qm f'8'l(x, 1, t)cos p„xdx] (14)
(x,Q,t) = 0 (4)
9v It should be noted that Gmn(t) in equation (11) is unknown and
4
can be found from equation (14) only after the surface tempera-
8(h V,t) + Ijf^U.y,') = W) - §e (l,;y,fl (5) ture is known from equation (7). Equations (7) and (11) are thus a
simultaneous set. If Gmn(t) is negligible, which will be the case
0(x, 1,1) + ~ | ~ (x, 1, t) = F{t) - H 6\x, 1, t) (6) when radiation can be neglected compared to convection, equation
(11) can be integrated directly.
One advantage of obtaining this partially analytic solution is
The nondimensional parameters are the Biot number, Bi = hb/h,
that, whereas complete numerical solutions require the results
the Stefan number, St = ofTf^b/k, and a geometry factor c = b/a.
throughout the column at the previous time step, the analytic solu-
Symbol F(t) is used for the term 6f(t) + df4(t)St/Bi. A value for
tion for the entire column requires only previous results on the
the effective shape factor, /, is obtained by assuming that the fire
surface.
is a uniform, isothermal, gray gas as described in the Appendix.
Quasi-Analytic Solution. It is assumed that a solution of the Rather than putting equation (11) in finite difference form and
form solving equations (7) and (11) by iteration, a technique with more
rapid convergence is used. Consider the solution of equation (11)
for a small step in time from time ti to t
6(x, y, /) = £ E Cm(t) c o s p„x c o s q„y (7)
Cm(t) = Cmi exp[-7rm(t - /,)]
exists, where pn and qm axe the set of all positive roots of the tran- + 4>™rmn expi-y^t) ffF(t) exp(ymt)dt
scendental equations
- ym ex&ryJhfGjft exp(ymlt)dt (15)
Pn t a n Pn = B i / c qm t a n qn Bi (8)

^Nomenclature.
a, b = column half widths q = dimensional heat flux <t>mn, l / w given by equations (10) and
A = area St = modified Stefan number afTf^b/k (13)
Bi = Biot number hb/k t - dimensionless time la/a1
c = ratio b/a I = dimensional time
cp = specific heat T = absolute temperature Subscripts
Cm„, Cmn' = defined by equations (7) and x, y = dimensionless coordinate system x • / = fire
(18) x/a, y = y/b m, n = indices of the roots of equation (8)
/ = effective shape factor for radiation x, y = Cartesian coordinates s = column surface
F = fire temperature group 0/ + 0/4 St/Bi a = thermal diffusivity k/(pcp) w = walls (bounding surfaces)
Gmn = defined by equation (14) ymn = given by equation (12) x, y = components of a vector in x- and y-
h = convective heat transfer coefficient e = emissivity directions, respectively
k =. solid phase thermal conductivity 8 = dimensionless temperature T/Tf$ 0 = initial condition
Pn, Qm = roots of equation (8) p = density 1 = value of a variable at the previous time
q = dimensionless heat flux aq/(kTfto) a <= Stefan-Boltzmann constant step

Journal of Heat Transfer NOVEMBER 1975 / 599

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 If the time interval is chosen to be small, it may be assumed that
Cmit) = [Cm{0) + iP^Y^l exp(rmnt)ef(t)dt]exp(~rmt)
F(t) and Gmn(t) are both linear over this interval. With this as-
sumption and the known F(t), integration by parts yields (25)

cj,t) = !i>mF{t) + [cm,i - lpmFl + (^/rJKF - FO

The integral in equation (25) can be carried out analytically for a
/(t - ti) + G Ml ] exp[-Ym,(t - tj] - (<Pm/Yj(F - Ft) restricted set of 0/(t). The temperature field is then known at any
time using C mn (0) directly from the initial condition.
At ~ 'i) - Gw(() + (l/rm){l - exp[-ym(t - tt)]} Convective Boundary Condition. When radiation can be ne-
x (G„ - GmX)/{t - tt) (16) glected compared to convection, St « Bi, F c^ 6f, and Gmn(t) -* 0
as in the previous case. The only difference from the first case is
that since h is finite, p„ and qm are not explicit. Equation (25),
Equations (7) and (16) may be solved by iteration. The conver-
therefore, again applies for this case.
gence of the series in equation (7) is improved by rearranging
equations (7) and (16). From equation (7) Scaling Laws. An important application of this analysis is to
define the conditions under which laboratory scaled experiments
accurately describe full-scale tests. The governing nondimensional
parameters are c = b/a, x = x/a, y = y/b, 0 — T/T[Q, t = ia/a2, Bi
B(x,y,t) = 6f(t) + E S c j f ) c o s p„x c o s clmy (17>
m=l n=\
= hb/k, and St = ofbT/rf/k. The requirement for scaling is ob-
where viously that all these groups remain the same in the laboratory and
full scales. Geometric similarity maintains c constant. For dynamic
C„'{f) = CJ/) - ip^G/t) (18) similarity, the dimensionless time-temperature curve must remain
unchanged. Thus, the values of 7/ should be the same in the labo-
Equation (16) can now be written in terms of Cmn' as ratory and full scales. As far as the thermal field is concerned, the
material, i.e., a, may be the same in both cases. This may not be
true for the stress field where aggregate size would probably be
Cmn' = ll>m0f\f) S t / B i + [ C „ , ' - Ifl^e^ St/Bi scaled. As in all transient conduction problems to maintain the
+ i^/rJiF - Ft)/(t - t{) + Gml] exp[~ym,(i - /,)] Fourier number, the laboratory time decreases by the square of the
factor by which the length scale decreases. Thus, a half scale
- U'^/rJiF - Fi)/(t - ti) ~ Gm(l) + (1/Ynm) ASTM Standard E-119 fire test would traverse the same tempera-
tures four times faster than full-scale. Scaling would be cost effec-
x {1 - e x p [ - y m l ( / - ti)]]{G„ - G„,)/{t - lt) (19)
tive in both material and time. In order to maintain the Biot num-
ber constant for the same column material, h must vary inversely
In order to start the computation, initial values for 0f, F, Gmn, with the length scale. Thus, the laboratory h must exceed the full
and C m „' are required. From the initial condition specified by scale h, e.g., by artificially introducing additional forced convec-
equation (2) tion. Note that this is an average h and it is assumed the detailed
flow need not be scaled. One problem here is experimentally deter-
0,(0) = 1 (20) mining the appropriate full scale h.
Finally the Stefan number can be maintained by varying the ef-
-FX0) = 1 + S t / B i (21) fective shape factor, /, inversely with the length scale. Scaling can
St 1 be accomplished only over a restricted range of full scale /, e.g., 0 <
Gm(0) A „ [cos p„J 0 o 4 ( l , y ) c o s q^ydy / < 0.5. An increase in / could be achieved by seeding the test fur-
nace fuel with silicone monoxide particles to increase the emissivi-
COS f/„
/ e0\x, l ) c o s />n«fct] (22) ty of the flame. It would also be possible to coat the surface of the
specimen to alter its total emissivity. Finally changing the size of
the test enclosure could be used to vary / as indicated in the Ap-
Equations (9) and (18) give pendix.
Within the assumptions of this model, the temperature field in
1 ' ' structural elements can be obtained from appropriately scaled lab-
c
» ' ( ° ) = T ~ X /„ eo(*> :v)cos />„* c o s ^ tf*tfy - ^ (23) oratory tests. Experiments are required to determine whether free
water, internal cracking, or temperature-dependent properties
would vitiate those assumptions. It also remains to be determined
In the special case when the initial column temperature is assumed whether the more complex problem of the stress field can be
equal to the initial fire temperature, then 0o(x, y) = 1, so that scaled.
G m „(0) = ^ m „ St/Bi and C mn '(0) = 0.
Limiting Cases. When the nonlinear terms in the boundary Results
condition can be neglected, analytical results are obtained which The specific case of a square column with a uniform initial tem-
clarify the general effect of each parameter. The solution becomes perature equal to the initial temperature of the gas is considered to
analytic in the following cases. illustrate the results of the analysis. Two temperature histories are
Temperature Boundary Condition. In the limit h -* •», the utilized to represent the extremes expected for actual fire develop-
column surface temperature approaches the fire temperature, the ment within structures:
mixed boundary conditions reduce to temperature conditions, and 1 A long duration, moderate intensity fire as defined by the
equations (8) give American Society for Testing and Materials (ASTM), E-119 stan-
dard as shown in Fig. 2 [1],
2 A short-duration, high-intensity (SDHI) fire developed from
Ai= l^i)l q^^L^l m the work of Magnusson and Thelandersson [2] as shown in Figs. 3
and 5. The standard ASTM and SDHI time-temperature curves
can be extracted from the plots labeled Bi = <*> in Figs. 2 and 3
and the solution is simply equation (7) with Cmn(t) given by equa- using the thermal properties of concrete as k = 1.5 kcal/hr-m-°C, p
tion (11) with Gmn(t) = 0 and F(t) - 8f(t). Equation (11) can be = 2400 kg/m 3 , and cp = 0.27 kcal/kg-°C, and a column size of 0.3 X
integrated directly to obtain 0.3 m.

600 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A*'

St = 0.05 Bi=0.5
A. S. T. M. q = 10.0 IS r* 1
S.D.H.I, q =1.0 I S K
-58—

0.03 0.04 005 A A

t = 0.02 t = 0.04
NORMALIZED TIME, I =

Fig. 2 Temperature histories at two locations in the column for the ASTM
time-temperature curve, parameterized in the Blot number, Bi, for a fixed
St/Bi ratio NORMALIZED TIME, I = lc/az

Fig. 4 Transient dlmensionless heat flux vectors at the surface of a col-

umn subjected to the ASTM and SDHI fime-temperafure curves

Fig. 4 shows heat flux vectors for St = 0.05 and Bi = 0.5. The di-
mensionless heat fluxes are normalized on kTfo/a. Heat flux vec-
tors at points along the line A-A on the surface are indicated at
several times. Due to symmetry, i.e., c = 1, these results describe
the entire surface. As expected, the heat fluxes quickly increase to
a maximum and then decrease. Except at the corner, this peak oc-
curs at t ~ 0.04 for the ASTM curve and at t ~ 0.01 for the much
faster SDHI curve. In both the cases, heat fluxes approach zero as
£ -* co due to the fact that both ASTM and SDHI fire curves be-
come flat at large times, and therefore the temperature everywhere
in the element approaches the approximately steady asymptotic
fire temperature. The heat fluxes appear here to be larger for the
ASTM fire, but this is not true for times smaller than 0.01, where
the hot, fast SDHI fire produces fluxes of 0(10) and the ASTM fire
produces fluxes of 0(2). A reversal of the heat flux vector is first
observed for the corner point in a SDHI fire slightly before a non-
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
dimensional time of 0.03. Because of symmetry, the heat flux vec-
NORMALIZED T I M E , f = t o V a 2
tors in both cases, for the central surface point and for the corner
Fig. 3 Temperature histories for the SDHI time-temperature curve for point, are, respectively, along the column center line and along the
conditions identical to Fig. 2
line joining the origin with the corner. At other points the heat flux
vectors rotate in a counter clockwise direction with increasing
time.

Figs. 2 and 3 show the temperature histories at an interior point

(0.5, 0), and a surface point (1, 1} exposed to the foregoing fires.
The ratio of the Stefan number St = afbTf:o3/k to the Biot number
is held constant to maintain a constant relative effect of radiation
to convection. The specific value of St/Bi = 0.1 corresponds to a
maximum / = 0.9 and a minimum h = 4.9 kcaI/hr-m 2 -°C. It repre-
sents a limiting typical case where the effect of radiation is large k = 1.50 kcal/hr-m-°C
compared to convection. The curve for Bi = 0.01 in Figs. 2 and 3 P- 2 4 0 0 kg/m 3
for point (0.5, 0) is coincident with the t axis. Cp = 0.27 kcal/kg*c
h =4.9 kcal/hr-m^'C
Increasing Bi corresponds to increasing the rate of heat transfer f = 0.9
until the limit Bi -» <» is reached where the temperature history on QUASI-ANALYTIC
FINITE ELEMENT
the surface is the same as the fire temperature history as shown in
S.D.H.I. CURVE
Figs. 2 and 3 for the corner. Once Bi is 0(1), further increases do
not strongly influence the corner temperature. Since the corner
point is much more sensitive to the fire temperature than other
points on the surface, round columns might be preferred in terms
of fire endurance. Other calculations involving all locations on the
surface indicate that when Bi 5; 10, a temperature boundary condi-
tion may be assumed at the surface. Since the surface temperature
always lags the fire temperature, the former exceeds the latter dur- 0.0 0.2 0.4 0.6 0.8 1.0'
ing the cooling of SDHI curve. Heat is conducted from a tempera- TIME,T(HRS.)
ture maximum inside the column to the surface and from the sur- Fig. 5 Comparison of this solution with a finite element solution for a 0.3
face to the surroundings. X 0.3 m concrete column in a short duration high intensify fire

Journal of Heat Transfer NOVEMBER 1975 / 601

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 column contained eight reinforcing bars, 16 mm in diameter,
SURFACE TEMPERATURE placed on the perimeter of a square with sides of 0.2 m as shown in
the column cross section in Figs. 6 and 7. The column surfaces
were heated by radiation from electric lamps. The surface temper-
ature was monitored and the power to the lamps controlled so that
the surface temperature followed a prescribed curve. The compari-
son was made by letting Bi -* <= to simulate a temperature bound-
ary condition at the surface. The experimental time-temperature
histories were available for various points on the surface and inside
the column.
Due to the placement of the lamps, the temperature varied along
the surface in the experiment. The quasi-analytical solution does
not simulate this mode of heating (though it does account for vari-
ations in surface temperatures under a uniform fire with mixed
3.0 4.0 boundary conditions). The temperature given by the thermocouple
TIME, f (HRS.)
placed on the center line was used as an approximate boundary
Fig. 6 Comparison of experimental and theoretical temperature histories condition. This temperature will have the most influence along the
for the special case of a temperature boundary condition on a 0.3 X 0.3 m center line of the column where the comparison in these figures is
concrete column—the material properties are as indicated in Fig. 5 (the
x's indicate thermocouple locations and the circles are reinforcing bars) made. The maximum experimental variation in the temperature
for any two points on the surface was 20° C at 300° C.
Fig. 6 shows the time-temperature curves for thermocouples 54
and 67 imbedded in the column. Thermocouple 67 is 0.13 m from
the center (0.02 m from the surface) on the center line of the col-
Comparisons umn. This places it 19 mm from the outside surface of a reinforc-
Finite Element Method. Fig. 5 shows a comparison of the ing bar. Thermocouple 54 is located in the center of the column.
temperature histories for a 0.3 X 0.3 m concrete column with a fi- The agreement between analytic and experimental results is good
nite element solution [3]. The SDHI curve was chosen to give a in spite of the many simplifying assumptions, i.e., temperature in-
comparison under the more stringent conditions of faster changes dependent thermal properties, neglecting reinforcement effects,
in the fire temperature. The concrete material properties shown moisture diffusion, and specifying constant temperature at the
were suggested by Bizri [5]. The agreement between the two meth- surface. The results given by the finite element method, which
ods is excellent. takes into account most of these effects, are only slightly better
than those obtained by the analytic method. Thermocouple 67 in-
The computation time was of the same order for both methods;
dicates lower temperatures than the analytic and numeric results
however, the quasi-analytic solution included heat fluxes omitted
during the temperature rise portion. This discrepancy is probably
in the finite element calculation. For the quasi-analytic solution,
caused by the proximity of the thermocouple to the reinforcing bar
computer time reduces rapidly for smaller values of /, which was
which acts as a heat sink due to its high thermal conductivity rela-
chosen here to be almost maximum, while computation time for
tive to concrete. The slight discrepancies for thermocouple 54
the finite element solution is comparatively constant.
could be due to the fact that this point is farthest from the surface
When the results inside the element are required for fewer and therefore strongly dependent on the concrete thermal proper-
points, or the Stefan number is small, or if heat fluxes are desired, ties which are not precisely known. Moisture may not have played
it is more convenient to use the quasi-analytic solution. On the a role here since the column had been heated previously several
other hand, if the thermal properties are temperature dependent, times.
or the material is anisotropic, or the geometry is other than rectan-
gular, circular, or spherical, no analytical solution can be easily ob- Fig. 7 shows temperature profiles along a center line at different
tained and the finite element method would be the best choice. times. Good agreement between analytic, experimental, and nu-
E x p e r i m e n t a l Data. Experiments on a 0.3 X 0.3 X 1.5 m high meric results is obtained. The quasi-analytic results neglected the
reinforced concrete column were carried out by Bresler [15]. The reinforcement bars. The finite element results which included that
effect still differ slightly from the experimental data, possibly be-
cause of cracking in the concrete near the reinforcing bars due to
the different coefficients of thermal expansion for the steel and
concrete. Such material behavior is obviously difficult to incorpo-
rate in any thermal analysis. It can be seen from the results, how-
ever, that the overall effects of these phenomena may not be of
critical significance.
Internal Stresses. To assess the impact of using an approxi-
mate temperature field on the resulting stress field, a comparison
was made between the stress fields obtained using the finite ele-
ment and quasi-analytic temperature fields as inputs. The results
are shown in Fig. 8 as the stress at two points in a column cross
section as a function of time. Point 1 is in the steel reinforcing bar
near a corner, chosen to represent a maximum difference in local
temperature gradient. Point 2 is in the computational element ad-
jacent to the surface center line. The temperature fields used are
those shown in Figs. 6 and 7. The agreement is good.

Conclusions
0.02 0.04 0.06 0.08 0.10 0.12

DISTANCE FROM CENTER, y ( m )

In summary, the unsteady, two-dimensional temperature field
in a structural element subject to unsteady, nonlinear, mixed
Fig. 7 Comparison of experimental and theoretical temperature profiles
at various times for the same column, boundary conditions, and notation as boundary conditions has been obtained with a minimum of numer-
in Fig. 6 ics. This solution has been applied to a rectangular column unifor-

602 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ,|i| V v W

Fig. 9 Radiation analysis network describing the Interaction between the

column surface, the fire, and the bounding surfaces of the surrounding en-
closure
1.0 2.0 3.0 4.0
TIME,t{HRS)
Fig. 8 Comparison of stress histories obtained from the temperature field 9 Lach, Jan, "Variational Approach to the Unsteady Heat Conduction
shown in Figs. 6 and 7 With Nonlinear Heat Transfer Coefficient," Heat Transfer—Souiet Re-
search, Vol. 6, No. 1, Jan.-Feb. 1974, pp. 145-150.
10 Chambre, Paul L., "Nonlinear Heat Transfer Problem," Journal
Applied Physics, Vol. 30, No. 11, Nov. 1959, pp. 1683-1688.
11 Clauson, W. W., "Nonlinear Problem in Heat Conduction," Chemical
m a l l y e x p o s e d t o fire. T h e g o v e r n i n g n o n d i m e n s i o n a l g r o u p s h a v e Engineering Progress Symposium Series, Vol. 61, No. 57,1965, pp. 16-25.
12 Crosbie, A. L., and Viskanta, R., "Transient Heating or Cooling of
b e e n identified as t h e B i o t n u m b e r Bi = hb/k, t h e S t e f a n n u m b e r One-Dimensional Solids by Thermal Radiation," Third International Heat
S t = afbT/:oa/k, and a geometric ratio c = b/a. Transfer Conference—Proceedings, Vol. 5,1966, pp. 146-153.
G e n e r a l r e s u l t s h a v e b e e n o b t a i n e d for a wide r a n g e of B i o t 13 Dicker, D., and Asnani, M., "Perturbation Solution for Nonlinear Ra-
diation Heat Transfer Problem," Third International Heat Transfer Con-
n u m b e r s . T e m p e r a t u r e fields a n d h e a t fluxes w i t h i n t h e c o l u m n
ference—Proceedings, Vol. 5, 1966, pp.164-173.
w e r e o b t a i n e d for t w o e x t r e m e fire h i s t o r i e s , t h e A m e r i c a n Society 14 Kartashow, E. M., and Moryakov, V. E., "Class of Heat Conduction
for T e s t i n g a n d M a t e r i a l s ( A S T M ) E-119 t i m e - t e m p e r a t u r e c u r v e Problems With Time-Variable Heat Transfer Coefficient," Souiet Aeronau-
and a short-duration high-intensity (SDHI) time-temperature tics, Vol. 16, No. 1,1973, pp. 54-59.
curve. T h e m o s t severe c o n d i t i o n is t h e A S T M c u r v e w i t h s t r o n g 15 Iding, R., Lee, J., and Bresler, B., "Behavior of Reinforced Concrete
Under Variable Elevated Temperatures," Report No. UCB-FRG 75-8, Fire
r a d i a t i v e i n t e r a c t i o n . C o m p a r i s o n w i t h e x p e r i m e n t a l d a t a [15] a n d Research Group, University of California, Berkeley, Apr. 1975.
finite e l e m e n t c a l c u l a t i o n s [3] p r o v i d e good a g r e e m e n t , a n d p e r m i t 16 Min, Kun, and Emmons, Howard W., "The Drying of Porous
t h e u s e of t h e s e t e m p e r a t u r e fields as i n p u t t o t h e s t r u c t u r a l s t r e s s Media," Proceedings of the 1972 Heat Transfer and Fluid Mechanics Insti-
problem. tute, Stanford University Press, 1972, pp. 1-18.
N e x t , t h e effects of m a s s t r a n s p o r t [16] of free a n d b o u n d w a t e r
in c o n c r e t e m i g h t b e i n t r o d u c e d t o d e t e r m i n e w h e t h e r t h e a b r u p t
c h a n g e s in t h e t h e r m a l field o c c u r r i n g a t e v a p o r a t i o n f r o n t s h a v e a
significant i m p a c t on t h e s t r e s s field. I t w o u l d also b e i n t e r e s t i n g
APPENDIX
Surface Boundary Conditions
t o i n c l u d e m o r e realistic n o n u n i f o r m b o u n d a r y c o n d i t i o n s b y
m o d e l i n g t h e fire in a n e n c l o s u r e . A p r i n c i p a l difficulty is t h a t t h e T h e t o t a l h e a t t r a n s f e r t o t h e c o l u m n is
t e m p e r a t u r e in t h e e n c l o s u r e w o u l d n o longer be c o m p l e t e l y d e -
c o u p l e d from t h e t e m p e r a t u r e field w i t h i n t h e c o l u m n . qs = h(Tf - Ts) + aes x

Acknowledgments
€,€„,(! - €f) -+ A
:](r/ - r s V - - ( r / - TJ)
T h e a u t h o r s a r e grateful for s u p p o r t from N a t i o n a l S c i e n c e w ^-w

Foundation G r a n t s GI-43 and ERT70-01080-A05 and to B. Bresler

for his a s s i s t a n c e . efe„ [1 + £ M 1 - €,)(! - €„)%. €/€*
€/€„(! - €,) C, + *tt €,C
References (Al)
1 "Standard Methods of Fire Tests of Building Construction and Ma-
terials E 119-73," Book of ASTM Standards, American Society for Testing w h e r e t h e n e t w o r k a n a l y s i s s h o w n in Fig. 9 p r o v i d e s t h e r a d i a t i v e
and Materials, Part 14, Philadelphia, Nov. 1970, p. 462. component.
2 Magnusson, S. E., and Thelandersson, S., "Temperature Time
Curves of Complete Process of Fire Development," ACTA, Polytechnica E q u a t i o n ( A l ) c a n be u s e d as t h e b o u n d a r y c o n d i t i o n a t t h e col-
Scandanavia, Civil Engineering and Building Construction, Stockholm, Se- u m n surface once t h e u n k n o w n t e m p e r a t u r e of t h e b o u n d i n g s u r -
ries No. 65, Table A6,1970, p. 136. faces, Tw, is d e t e r m i n e d . In e a c h of t h r e e l i m i t i n g cases t h e t o t a l
3 Becker, J., Bizri, H., and Bresler, B., "FIRES-T A computer Program h e a t t r a n s f e r c a n be w r i t t e n in t h e form
for the Fire Response of Structures—Thermal," Report No. UCB FRG 74-1,
Fire Research Group, University of California, Berkeley, Jan. 1974.
4 Becker, J. M., and Bresler, B., "FIRES-RC A Computer Program for qs = h(Tf - Ts) + afiTf* - Ts4) (A2)
the Fire Response of Structures—Reinforced Concrete," Report No. UCB
FRG 74-3, Fire Research Group, University of California, Berkeley, July Case (i) Tw = Tf.
1974.
5 Bizri, H., "Structural Capacity of Reinforced Concrete Columns
Subjected to Fire Induced Thermal Gradients," PhD dissertation, Universi- + £ e,e„ 4 . 1 - €,
ty of California, Berkeley, Jan. 1973. ^ €,£„,(! - e,) A €
6 Mann, W., and Wolf, F, "Heat Transfer Between Solids and Gases
Under Nonlinear Boundary Conditions," Quarterly of Applied Mathemat- / =
ics, Vol. 9, No. 2, July 1951, pp. 163-184. N-[1 + T5-(1 ~ £/)(! - «,<
7 Jaeger, J. C , "Conduction of Heat in Solids With Power Law of Heat €/€»(!
€
Transfer at Its Surface," Cambridge Philosophical Society Proceedings, <V + s e f e«
Vol. 46, Part 4, Oct. 1950, pp. 634-641. e
<V + *
8 Goldstein, S., Modern Developments in Fluid Mechanics, Clarendon
Press, Oxford, England, Vol. 1, Chap. IV, 1938. (A3)

Journal of Heat Transfer NOVEMBER 1975 / 603

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 For realistic systems Aw » As, and the foregoing expression Case (Hi) Adiabatic Walls.
simplifies to Returning to the network in Fig. 9 and treating the bounding
surfaces as adiabatic gives
/ = e. (A4)

Case (ii) Tw « Ts. A,

ca[l + (1 - e , ) ^ - ]
Taking the temperature history of the walls equal to an average f = (A7)
column surface temperature history, ~~A~r e
1 + — («/ + s - £/es)

A.. (1 " O ]
Again in the limit Aw » As, this reduces to
/ = + e,„
(1 - € / )(l - e„)
€/d
e / = e. (A8)

(A5) Emissivity of a concrete surface is approximately 0.9. The value

of / is, therefore, close to 0.9 in the first and third cases. However,
For ^ » As, this becomes in the second case, / depends upon if as well. Choosing t s = ew =
0.9, and if = 0.3, equation (A6) gives / = 0.29. For e/ = 0.9, / = 0.82
e,e„ is obtained. The value of / applicable to real fires could therefore
(A6)
+ €„. e,e„ be expected to lie in the range of 0.3-0.9.

ERRATUM
Erratum: W. J. Minkowycz and E. M. Sparrow, "Local Non-Similar Solutions for Natural Convection on a Vertical Cylinder,'
published in the May 1974 issue of the JOURNAL OF HEAT TRANSFER, pp. 178-183

Owing to an inadvertency, the terms <t> and x were, respectively, Heat Transfer Conference, Vol. Ill, Sept. 1974, pp. 15-19). Inas-
omitted from the left-hand sides of equations (21) and (28). Fortu- much as neither the £ values nor the Prandtl numbers of Cebeci's
nately, the heat transfer results are insensitive to the omitted calculations coincide with those of the paper, interpolations were
terms. Calculations for £ = 0.5 and 5.0 have shown that when these necessary to carry out the comparison. The findings of the compar-
terms are included, the #'(£,0) values are only 0.4-0.5 percent less ison indicated agreement to within about one-quarter percent.
than those given in the paper. By any practical measure, these dif- The skin friction results, which are proportional to F"(£,0), are
ferences are negligible. slightly more sensitive to the omitted terms than is the heat trans-
Another affirmation of the good accuracy of the results given in fer, with deviations being in the range 0.5-1.25 percent.
the paper can be cited on the basis of a comparison with the finite- The authors are indebted to C. C. Chen of the University of
difference solutions of Cebeci (Proceedings, Fifth International Kentucky, who first called their attention to the omitted terms.

604 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 B. J. Jody
P. C. Jain 1 Determination of Thermal
S. C. Saxena

Department of Energy Engineering,

Properties From Steady-
University of Illinois,
Chicago, III.
State Heat Transfer
Measurements on a Heated
Tungsten Wire in Vacuum and
Helium Gas
An experimental facility in its developed form is described for the determination of a
number of thermal properties as a function of temperature in the range 400-2500 K.
These are the resistivity, tensile breaking stress, the total hemispherical emittance of
the tungsten wire, and the thermal conductivity of helium. These properties are comput-
ed from the knowledge of steady-state heat transfer data from tungsten wire in vacuum
and in the presence of helium at pressures covering the temperature-jump and contin-
uum regimes. These measured properties are discussed in the light of literature values
wherever available.

Introduction Experimentation
The details of the experimental facility, the experimental proce-
The knowledge of the thermal properties of gases, metal sur-
dures, the data processing techniques, and the underlying theories
faces, and metal-gas interface over a wide temperature range is of
are given in earlier publications [1-3]. The lower end of the column
great significance to proper engineering design of breeder reactors,
is modified so that known variable tensions on the axial tungsten
combustors, space vehicles, etc. We have been developing a hot-
wire can be applied either in vacuum or in a gaseous environment.
wire column type instrument for the accurate measurement of
The column design is described by Chen and Saxena [2] and the
such thermal properties over a temperature range of about 300-
modified lower end is sketched in Pig. 1 which is self-explanatory.
2500 K. In particular, we have reported the preliminary results on
It is to be noted that the design of the loading device permits the
the hemispherical emittance of tungsten and platinum wires [l], 2
application of variable stress on the wire. Its range is from 1.34 to
thermal conductivity of nitrogen [2], and the thermal accommoda-
5.50 kg/mm 2 for a 0.3048 mm wire. Six specially slotted weights of
tion coefficient for nitrogen covered tungsten-nitrogen system [3].
cold^rolled steel each weighing 43 g can sit on an O-ring when not
This heat transfer column facility is further improved and here we
employed to stretch the wire. These weights are suitably collimat-
present data on several thermal properties determined from mea-
ed for suspension on the wire in the column externally, by a spe-
surements of heat transfer as the tungsten wire is heated from am-
cially designed magnet.
bient (400 K) to the highest possible temperature (2500 K) in vac-
uum and in the presence of helium at different pressures. Two glass columns of lengths 165 and 132 cm and internal diam-
eter 8 mm, and each containing tungsten wire of diameter 0.3048
mm, are employed in both the modes of single and differential ar-
rangements [2], The glass surface is kept at a constant tempera-
1
ture by circulating water at 316.6 K. The tungsten wires, supplied
On Leave from National Physical Laboratory, New Delhi, India. by General Electric and stated to be chemically cleaned and
2
Numbers in brackets designate References at end of paper. straightened (type 218 CS), are used both as a heater as well as a
Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer resistance thermometer. The electric potentials developed across
Division February 18, 1975. Paper No. 76-HT-C. the tungsten wires are monitored on a multifunction meter and are

Journal of Heat Transfer Copyright © 1975 by ASME NOVEMBER 1975 / 605

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A

> 2600

b> Fig. 2 Comparison of the present resistivity values based on equation (1)
with the experimental data (curve 1), with the data measured at five fixed
points and extrapolated to high temperatures on the basis of a quadratic
relation in temperature (curve 2), with the recommended correlations due
to Toulouklan [4] (curve 3), and Smithells [5] (curve 4). Percent Dev. =
|p(equation (1)) - p(exptl. or reference)j100/{p(equation (1))|. Experi-
mental points: © and O.

tance data are converted to resistivity (p = RS/2L) values and

these are found to be well correlated by the following quartic poly-
QUU nomial in reduced temperature (6), within an average absolute de-
viation of 0.6 percent:

p(0) = -0.2025 + 20.000 + 1.54902 + 4.32103 - 1.2O804 (1)

Here p is in pil cm and 0.3 < 0 < 2.5. The temperatures obtained
on the basis of the given relation differ from the directly measured
temperatures by about 1.3, 1.0, 0.7, and 0.4 percent at 505, 1338,
2090, and 2455 K, respectively.
The resistivity values based on equation (1) are compared in Fig.
2 with the experimental data points, with the resistivity data mea-
sured at five fixed points and extrapolated to high temperatures
on the basis of a quadratic relation in temperature, and with the
recommended correlations due to Touloukian [4] and Smithells
[5]. Several conclusions can be drawn from these comparisons.
First, the maximum deviation of the directly measured data points
Fig. 1(a) The schematic of the lower end of the heat transfer column, not from the smooth plot is about one percent and this may be regard-
to scale—front view: 1-glass enclosure (diameter 28 mm), 2-glass-metal
(kovar) seals, 3-silver solder joints, 4-coupllng brass adapter, S-standard
ed as the measure of the overall precision of our resistivity deter-
glass joints, 6-brass rod (length 45 cm, dia 6 mm), 7-brass disk, 8-brass mination. T h e absolute percentage uncertainty is judged as 1.8 and
pins (six) for holding weights, 9-steel weight (six), and 10-O-ring, (to) top 1.4 at 500 and 2500 K, respectively. Second, as known, the practice
and front views of steel weight of extrapolating the low temperature resistivity data to high tem-
peratures is found to be unreliable. Third, the two sets of recom-
mended values [4, 5] available in the literature differ from each
measured on a Leeds and Northrup six-dial guarded precision po- other and also from the present measurements. This stresses the
tentiometer. The columns are evacuated from a high vacuum importance of always considering the resistivity data in reference
pumping system capable of creating a vacuum of about 2.7 X 10~ 5 to the sample for which it is originally taken. Fourth, the proce-
Pa. To ensure reliable and reproducible temperature characteris- dure adopted by us for resistivity data to use as the basis for tem-
tics of the tungsten wire, it is annealed for over six hours at about perature determination is reliable as well as accurate for most en-
2450 K and calibrated by determining its resistance at known tem- gineering needs and is in accordance with the recommendations of
peratures, which is explained in the following. the International Practical Temperature Scale of 1968 [6].

Resistivity of Tungsten Tensile Breaking Stress

The resistance of a known length of wire is determined at the The tension on the metal wire mounted in the column should be
triple point of water (273.16 K) in a triple point cell supplied by such that it remains taut without getting unduly stretched as its
Jarrett Instrument Company, and at the freezing points of tin temperature is raised by electrical heating. With the modified de-
(505.1181 K), cadmium (594.258 K), zinc (692.73 K), and alumi- sign of the column, it is possible to measure the tensile breaking
num (933.52 K) in a 8411 fixed temperature facility supplied by stress as a function of temperature. Known tensions were applied
Leeds and Northrup. The resistance of the tungsten wire is also on wires of about 0.127, 0.203, and 0.305 mm dia and their temper-
determined in the column in situ in the temperature range 1300- atures were slowly raised till breaking occurred. The results of
2500 K according to the earlier procedure [2] in which we have cor- such investigations are shown in Fig. 3 and are especially useful in
rected for the heat losses at the ends of the wire. All these resis- estimating the temperature to which a tungsten wire under a

-Nomenclature-
e = hemispherical total emittance, dimen- S = cross-sectional area of the wire, cm 2
sionless P = gas pressure, Pa T = temperature, K
k = thermal conductivity, mW/cm-K QH = power conducted through the gas, TH = temperature of the hot wire, K
Kn = Knudsen number, dimensionless W/cm B = reduced temperature = T/1000, K
L = one-half of the length of the wire, cm R = resistance per unit length, n9.lcm p = resistivity, ju^-cm

606 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 average absolute and maximum deviations of 1.5 and 4.5 percent,
respectively:
e(6) = -0.04259 + 0.25500 - 0.103002 + O.O315503 + O.O2O2104
-O.OO773O05 for 0.4 < 0 < 2.5 (2)

The upper part of Fig. 4 shows the temperature dependence of e

and the lower part a comparison of the actual data points with the
estimates based on equation (2). The estimated percentage error in
these e(0) values is always less than 4.8.

T h e r m a l C o n d u c t i v i t y of H e l i u m
The additional measurement of the electrical power required to
heat the tungsten wire to a known temperature in the presence of
1500
helium is enough to determine its conductivity as explained earlier
[2]. Helium gas is 99.9999 percent pure and is supplied by Mathe-
T,K son Gas Products. 0.3048-mm tungsten wires of length 46.24 and
Fig. 3 The tensile breaking stress, in kg/mm 2 , of tungsten wire as a func- 86.02 cm are used. The thermal conductivity values are displayed
tion of Its temperature, T, In K in Fig. 5 and these are correlated by the following polynomial in re-
duced temperature according to the method of least-squares with-
in average absolute and maximum deviations of 0.8 and 2.1 per-
cent, respectively:
known tension can be safely raised without deforming it. These ex-
periments also confirmed that resistivity of the tungsten wires is k(6) = 0.5504 + 3.35250 - 0.2116602 • O.O662603 for 0.4 < 0 < 2.3
independent of tension in the range shown in Fig. 3, and thus it is a (3)
unique parameter of its temperature only. The accuracy of the plot The values generated from equation (3) are shown by a continu-
of Fig. 3 though judged to be conservative and consequently a lin- ous curve in Fig. 5(a) and the scatter of the actual data points from
ear interpolation between successive data points is considered sat- these smooth values in Fig. 5(6). The three sets of k values ob-
isfactory. The usefulness of Fig. 3 is to be emphasized. tained by the single column (short and long) and differential ar-
rangements are found to be in good agreement with each other.
Hemispherical Total Emittance The measurements are taken at a gas pressure of 62660.4 Pa
The surface condition of metal wires plays a dominant role in where the temperature-jump correction is negligibly small and
controlling the magnitude of thermal radiation from it. Wire sur- continuum heat flow conditions exists. This conclusion is based on
faces are very sensitive to the methods of preparation, thermal his- the analysis of the Knudsen number which varies between the lim-
tory, and environmental conditions. The column instrument pro- its 0.001-0.0131. Previous investigations made in this laboratory
vides a viable basis for the accurate determination of hemispheri- suggest that continuum conditions exist in the column for Kn <
cal total emittance of metal wires as a function of temperature. 0.0149 for argon, Kn < 0.0119 for nitrogen, and Kn < 0.0137 for
The knowledge of the electrical power required to heat the cen- neon. The calculations further revealed that at this pressure the
tral section of the wire to different temperatures and the tempera- maximum energy convected is only 0.1 percent of the energy fed to
ture profiles at the ends enable the determination of the hemi- the wire. The Reynolds number is always less than 0.1, justifying
spherical total emittance according to a procedure outlined by thereby the assumption of laminar convection in the column im-
Chen and Saxena [1]. The results are given in Fig. 4 and the fol- plicit in our calculations. The maximum Rayleigh number for our
lowing polynomial in reduced temperature, obtained by the meth- experimental operating conditions is 1 and this is much smaller
od of least-squares analysis, correlates the observed data within than the critical Rayleigh numbers characterizing the end effect
and the onset of the secondary multicellular motion in the column
[2]. The apparent conductivity values are corrected according to
the procedures outlined by Chen and Saxena [2]. The accumulated
0.6 magnitude of the three corrections is nearly 0.3-0.8 percent. The
estimated random errors in the experimental data are 3.7, 2.6, 2.2,

300 800 1300 1800

300 800 1300 1800 2300 T,K
T,K
Fig. 5(a) Thermal conductivity of helium as a function of temperature,
Fig. 4 (Upper) Hemispherical total emittance, e, of tungsten as a function based on equation (3). (b) Comparison of three sets of experimental k
of temperature, T, in K. ® experimental points. (Lower) Comparison of the data with the values based on equation (3). ® long column, O short col-
experimental data with the predictions based on equation (2)—percent umn, D differential arrangement. Percent dev. = \k (equation (3)) -
dev. = [e (equation (2)) — e (exptl.)j100/e (equation (2)) fe(exptl.)j100/fe (equation (3)).

Journal of Heat Transfer NOVEMBER 1975 / 607

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 2300-

300 800 1300 1800 2300

T,K

Fig. 6 Comparison of the present k data with those of Faubert and

Springer [7] (curve 1), Saxena and Saxena [8] (curve 2), Timrot and
Umanskii [9] (curve 3), Vargaftik, et al. [13] (curve 4), Ho, et al. [11]
(curve 5), Kestin, et al. [13] direct measurements (curve 6), and Kestin, et
al. [13] principle of corresponding states (curve 7)—percent dev. = {k
(equation (3)) - k (reference)]lOO/jfc (equation (3))]

16

12

0.2 0.4
1/P, 1333.2/Pa

Fig. 8 Dependence of the hot-wire temperature on the gas pressure for

O constant power per unit length, QH/2L (W/cm)

and Davis [12]).

We have also measured QH values as a function of temperature
and helium gas pressures of 2146.5, 2639.7, 4186.2, 5732.8, 6919.3,
Fig. 7 Dependence of thermal energy conducted by helium on its pres- 8559.1, 11892.1, 17278.3, 22664.4, 30397.0, and 62660.4 Pa and
sure for fixed values of the hot-wire temperature these are shown in Fig. 7. It is to be noted that QH values system-
atically increase with pressure in this range (at least up to a pres-
sure of about 30397.0 Pa) at a fixed temperature. In Fig. 8 are
shown the plots of TH versus 1/P for constant values of power,
and 2.0 percent at 500, 1300, 2000, and 2300 K, respectively. QH/2L, and their linear nature in the pressure range of 2146.5-
The present thermal conductivity data as represented by equa- 11892.1 Pa provides the experimental confirmation to the temper-
tion (3) are compared in Fig. 6 with similar other data reported by ature-jump theory of heat transfer [15, 16]. The theory of heat
Faubert and Springer [7], Saxena and Saxena [8], Timrot and transfer for the temperature-jump region finally leads to the con-
Umanskii [9], and with the correlations of available conductivity clusion that 1/QH varies linearly with 1/P, [14]. In Fig. 9, values of
data developed by Ho, et al. [10], and Vargaftik, et al. [11]. The re- 2L/QH are plotted against 1/P for different values of TH and the
ported percentage experimental errors in the various works are linear nature of these curves confirms the conceived mechanistic
2.3-2.9 [7], 2.[8], and 5.5 [9]. If recognition is given to the reported model of heat transfer [16]. The intercepts with the 2L/QH axis of
uncertainties associated with each data set, the agreement between the linear plots enable the determination of k(T) as explained by
the present work and other column measurements appears to be Jody and Saxena [14]. These values at round temperatures are re-
satisfactory in each case. The correlated values are claimed to be ported in column 3 of Table 1. The linear plots of Fig. 9 are also
accurate within percentage uncertainties of 2-10 [10], and 2 [11] extrapolated to 1/P = 0 and the corresponding values of QH(TH)
and hence their agreement with present measurements is consid- are determined and these are employed to determine k(TH) as in
ered satisfactory. We also compared our conductivity values with the previous section and according to the procedure outlined by
the viscosity data, which have been measured more accurately. Chen and Saxena [2]. These fe(T//) values are given in column 4 of
Curve 6 of Fig. 6 is based on the viscosity data of Kestin, et al. [13] Table 1 and are in good agreement with the other two sets. This is
and the procedure outlined by Jody and Saxena [14]. Curve 7 o( a stringent test of the heat transfer model for the temperature-
this figure is based on their viscosity values generated by the appli- jump regime and must be regarded as a sufficient quantitative test
cation of the principle of corresponding states to the data for noble of the model.
gases [13]. It is to be noted that the majority of the earlier mea- We also determined k(Ttf) values from QH(TH) data taken at
surements of thermal conductivity are in appreciable disagreement pressures of 22664.4 and 30397.0 Pa and these values are smaller
with the reliable viscosity data of Kestin, et al. [13] and the present than k(Tn) values obtained for a pressure of 62660.4 Pa as well as
conductivity measurements agree within the range of mutual er- from data of QH(TH) for P = <*>. From these k{Tn) values two con-
rors of the two data sets over most of the temperature range. The clusions are clear. First, the k(Tn) values obtained from data
current data while substantially agreeing with other measurements taken in the pressure range above the temperature-jump region
provide a consistent set in the temperature range 400-2300 K and but below the continuum conditions are smaller than the actual
at a pressure of 101325.024 Pa as the measurements of Timrot and Fourier pressure independent conductivity values. Second, the
Umanskii [9] are suspected to be unreliable and particularly at pressure of 62660.4 Pa is high enough to establish the continuum
high temperatures due to their inadequate column design (Saxena heat flow conditions in our heat transfer column.

608 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Thermal conductivity of helium, In mW/cm-K, by different meth-
ods

Temp., K Equation (3) Intercepts of Fig. 9 in

conjunction with procedure
of
ref. 14 ref. 2

500 2.16

700 2.77

900 3.35

1100 3.89

1300 4.40

1500 4.88

1900 5.70

2300 6.33

Touloukian, Y. S., ed., Macmillan, New York, Vol. 1,1967, p. 1021.

5 Smithells, C. J., Tungsten, Chapman and Hall Ltd., London, 1952, p.
180.
1/P, 1333.2/Pa 6 "International Practical Temperature Scale of 1968," Metrologia,
Vol. 5,1969, pp. 35-44.
Fig. 9 Dependence of the rate of thermal energy conducted by helium on 7 Faubert, F. M., and Springer, G. S., "Measurement of the Thermal
its pressure for fixed values of TH Conductivity of Helium up to 2100 K by the Column Method," Journal
Chemical Physics, Vol. 58,1973, pp. 4080-4083.
8 Saxena, V. K., and Saxena, S. C , "Measurement of the Thermal Con-
ductivity of Helium Using a Hot-Wire Type of Thermal Diffusion Column,"
Brit. Journal Applied Physics (Journal Physics D), Vol. 1, 1968, pp. 1341-
1351.
9 Timrot, D. L., and Umanskii, A. C , "Investigation of the Thermal
Acknowledgments Conductivity of Helium in the Temperature Range 400-2400 K," High Tem-
This work is supported by the National Science Foundation perature, Vol. 3,1965, pp. 345-351.
under Grant No. GK-12519. Computing services used in this re- 10 Ho, C. Y., Powell, R. W., and Liley, P. E., "Thermal Conductivity of
search were provided by the computer center of the University of the Elements," Journal of Physical and Chemical Reference Data, Vol. 1,
1972, pp. 279-421.
Illinois at Chicago Circle. Their assistance is gratefully acknowl- 11 Vargaftik, N. B., Filippov, L. P., Tarzimanov, A. A., and Yurchak, R.
edged. Thanks are also due to the General Electric Company for P., "Heat Conductivity of Gases and Liquids," Foreign Technology Division,
their support and encouragement for this work. Wright-Patterson Air Force Base, Ohio, Report No. AD-736 963, 1971, pp.
45-51.
12 Saxena, S. C , and Davis, F. E., "Determination of Thermal Conduc-
References tivity of Gases at High Temperatures: The Column Method," Journal of
1 Chen, S. H. P., and Saxena, S. C , "Experimental Determination of Physics E: Scientific Instruments, Vol. 4,1971, pp. 681-688.
Hemispherical Total Emittance of Metals as a Function of Temperature," 13 Kestin, J., Ro, T. S., and Wakeham, W. A., "Viscosity of Noble Gases
Industrial and Engineering Chemistry Fundamentals, Vol. 12, 1973, pp. in the Temperature Range 25-700°C," Journal Chemical Physics, Vol. 56,
220-224. 1972, pp. 4119-4124.
2 Chen, S. H. P., and Saxena, S. C , "Experimental Determination of 14 Jody, B. J., and Saxena, S. C , "Thermal Conductivity of Neon From
Thermal Conductivity of Nitrogen in the Temperature Range, 100-2200°C," Heat Transfer Measurements in the Continuum and Temperature-Jump
High Temperature Science, Vol. 5,1973, pp. 206-233. Regimes," Phys. Fluids, Vol. 18,1975, pp. 20-27.
3 Chen, S. H. P., and Saxena, S. C , "Interface Heat Transfer and 15 KennardE. H., Kinetic Theory of Gases, McGraw-Hill, New York,
Thermal Accommodation Coefficients: Heated Tungsten Wire in Nitrogen 1938, pp. 311-315.
Environment," International Journal of Heat and Mass Transfer, Vol. 17, 16 Harris, R. L., "On the Determination of Thermal Accommodation
1974, pp. 185-196. Coefficients in the Temperature-Jump Region," Journal Chemical Physics,
4 Thermophysical Properties of High Temperature Solid Materials, Vol. 46,1967, pp. 3217-3220.

Journal of Heat Transfer NOVEMBER 1975 / 609

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 T. L. Brosseau
Mechanical Engineer. Reduction of Heat Transfer to Gun
J. R. Ward
Research Chemlsl.
Barrels by Wear-Reducing
u. S. Army Balllsllc Research Laboratories,
Aberdeen Pro.lng Ground, Md.
Additives
Heat transfer measurements were made in a 37-mm gun equipped with thermocouples at
a series of axial and radial positions. The heat transferred to the gun barrel was marked-
ly decreased when wear-reducing additives such as polyurethane foam, titanium diox-
ide/wax, and talc/wax were used as liners wrapped around the gun propellant. The re-
duction in heat transfer was greatest when the additives were folded over the forward
end of the propellant at the base of the projectile. For a given configuration, all three ad-
ditives tested reduced heat transfer to the gun barrel equally. This contradicts conclu-
sions reached in full-scale tank-cannon firing tests that TiOz/wax and talc/wax are su-
perior to polyurethane foam as erosion reducing additives. In these firings only the
metal-oxide/wax liners had flaps folded over the forward end of the propelling charge.

1 Introduction surface appears to be the critical factor in determining erosion

rates of metals by propellant combustion gases [6]. this article con-
Erosion rates of high-velocity guns such as tank cannons have cerns temperature measurements made with polyurethane foam,
been significantly reduced by the use of liners that are wrapped TiOz/wax, and talc/wax liners. In these measurements the location
around the propelling charges [1].1 The first additive was devel-
oped in Canada and consisted of high-density polyurethane foam
[2]. Soon after, two Swedish inventors [3] claimed that a mixture of
titanium oxide (TiO z) and wax coated onto a rayon liner would
provide a superior wear-reducing additive.
The Swedish inventors also incorporated flaps on the end of the
liner that could be folded over the forward end of propelling
charge at the base of the projectile. A photograph of the Swedish
additive is shown in Fig. 1 depicting the serrations on the end of
the wear-reducing liner. A summary of the erosion tests [4] made
in the M68 tank cannon is presented in Table 1. This table demon-
strates the efficacy of the wear-reducing liners in extending the
useful life of the M68 tank cannon. No flaps were incorporated on
the liner in the HEAT round because fins on the rear of the projec-
tile extend back into the propelling charge. More recently Picard
[5] has claimed that substitution of talc for the Ti0 2 improves the
erosion-reducing capability of Swedish additive. No erosion was
observed in fifty-round firing tests with both APDS and HEAT
rounds equipped with talc/wax liners.
Although the aforementioned additives successfully increase the
useful life of gun barrels, the mechanism by which the additives re-
duce erosion is not clear. Since the temperature of the gun barrel

1 Numbers in brackets designate References at end of paper.

Contributed by the Heat Transfer Division for publication in the JOUR-
NAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer Fig. 1 Tl02!wax addillve used In the 105 mm APDS round lIIuslrallng the
Division June 16, 1975. Paper No. 76-HT-K. lIaps

610 / NOVEMBER 1975 Copyright © 1975 by ASME Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Erosion fates of 105 mm j$68 tank cannon firing rounds equipped soo
with wear-reducing additives*a)

Roun-i Add it

M392 Ai'US none

M392 Ai'lS polyurethane (no flaps)

M592 Al'iS Ti07/wax (flaps)

M456 HUM none

M456 lli-AT Ti02/wax (no flaps)

Table 2 Location of thermocouples in the 37 mm gun barrel

Fig. 2 Pressure versus time in 37 mm gun tube—no additive (minihat

11.94 0.406 0.864 2.51 5.56 gage located 10.9 cm from base of case)
104.14 .483 .914 2.34 5.13

180.34 .178 .940 2.40 5.74

face are given in Table 2. The thermocouples consist of single-wire

constantan (0.076-mm dia) welded onto the gun steel by capacitor
discharge. Care was taken to place each thermocouple above a
groove. The inside diameter of the gun barrel ( n ) was 37.04 mm at
each axial location; the thickness of the gun steel (ro) decreased
and weight of the liner in the cartridge case were the same, as op-
down the barrel with values of 28.6, 22.2, and 7.85 mm at each of
posed to the M68 tank cannon erosion tests where flaps were only
the three axial distances in Table 2, respectively.
used in the TiC>2/wax liner in the APDS round. The potential im-
Wear-Reducing Additives. The three wear-reducing liners in-
portance of this may be seen from the results in Table 1. Without
vestigated are standard military issue. Their compositions are list-
additives the APDS round is slightly more erosive than the HEAT
ed in Table 3. The polyurethane foam liner is a high-density (480
round; with additives the wear life of the APDS round is ten times
kg/m 3 ) semirigid plastic 3 mm thick. The metal-oxide/wax liners
longer than the HEAT round.
consist of a blend of the metal oxide and wax coated on one side of
a rayon liner. The metal-oxide/wax blend is made by mixing the
2 Experimental Procedures
metal oxide with molten paraffin wax; this mixture is poured onto
The measurements, test setup, and procedures used in these ex-
the rayon liner and scraped to a 1-2-mm thickness. The coated
periments are the same as those used for the 37-mm gun tube fir-
side of the liner faces the propellant grains. The liners were each
ings in reference [7].
cut to a length such that they weighed 0.15 kg.
Temperature Distribution. Temperatures at four radial posi-
Chamber Pressure and Muzzle Velocity. Pressure in the
tions were measured at three axial locations along the length of the
chamber was measured with a Minihat Pressure Gage, The center
tube. The temperatures at each position were measured as contin-
of the gage hole was located 10.9 cm from the base of the cartridge
uous functions of time, and from these measurements temperature
case. Muzzle velocity was measured by placing three Lumiline
distributions were obtained for several time intervals. The axial
Screens downrange, 3.7 m apart. The chamber pressure and
locations and the distance from each thermocouple to the bore sur-
muzzle velocity were unaffected by the presence of the additives.
An example of the chamber pressure versus time measured for this
gun is shown in Fig. 2.
Table 3 Composition of wear-reducing additives Test Setup and Procedure. The weapon was constrained in a
rigid mount. It was fired manually to eliminate noise spikes pro-
Polyurethane Foajn Percent by Weight
duced on the temperature measurements when an electrical sole-
A. Resin Prepolymer Ingredients
noid is used to fire the weapon. The temperatures and pressure
Polyethylene glycol (200) 10.5 were recorded simultaneously on a Honeywell Magnetic Tape Re-
Polypropylene glycol (1200) 6.5 cording System.
Castor oil 36.5 Rounds fired throughout the test consisted of 0.871 kg T21 pro-
2, 4 Toluene dissocyanate 46.5 jectiles, M38B2 primers, and 0.165 kg of M2 propellant from Lot
B. Catalyst Ingredients No. RAD 35683. In the firings with the additives 0.015 kg of poly-
Polypropylene glycol 10.0
urethane foam, Ti02/wax, and talc/wax were used with and with-
Glycerine
out folded flaps at the front of the additive. The polyurethane
7.5
foam had to be heated slightly with a heat gun to soften it before
Polyethylene glycol 3.75
the flaps could be formed. In the firings with the TiCWwax and the
Tris (pentanedionato) iron (III) 0.15
talc/wax, a clean out round with no additive was fired between
Nigrosine black 0.25 rounds to remove the visible film along the bore left by these addi-
Oibutyl tin dilaurate 0.30 tives. The polyurethane foam did not leave a film.
Metal Oxide Additives
3 Results and Discussion
Titanium dioxide (or talc) 46.0
Figs. 3-5 are examples of the temperature versus time profiles at
Paraffin wax 53.5
four radial locations for rounds fired with no additive, with poly-
Dacron 0.5 urethane foam without flaps, and polyurethane foam with flaps. A
Stearyl alcohol (1.0 mi glance at the temperature profiles shows that the temperature in
cloth backing: viscose rayon the gun barrel is reduced by the polyurethane foam, and the addi-

Journal of Heat Transfer NOVEMBER 1975 / 611

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 I = .406mm FROM BORE SURFACE
2" .864
3 = 2,51
4=5.56

Fig. 3 Temperature versus time at 27.94 cm location—no additive

1= .4 0 6mm FROM BORE SURFACE

1- .864
3= 2.51
4=5.56

400 600
TIME.(ms)

Fig. 4 Temperature versus time at 27.94 cm location—potyurethane

foam, no flaps

tion of the flaps further assists in reducing the temperature. as follows. The rate of heat input into a unit axial length is given
Table 4 summarizes the mean temperatures measured at 27.9 by
cm from the base of the cartridge case at a depth of 0.406 mm from dQ = 27rpcATrdr (1)
the bore surface for all three additives equipped with flaps. Sur-
prisingly, the polyurethane foam was as effective as the TiC>2/wax where
and the talc/wax liners in reducing the barrel temperature profile dQ = rate of heat input into a unit length of the barrel, J/mm
suggesting that the positioning of these three additives is more p = density of gun steel, 7.85 X 10 3 kg/m 3
critical than the chemical composition. c = specific heat of gun steel, 419 J/kg-K
An attempt to estimate the total heat transferred to the gun bar- r = distance into the gun barrel, mm
rel at a unit axial position was made from the temperature profiles AT = temperature rise at depth r at 100 ms
at 100 ms. This time was chosen because the thermocouples near- The total heat input at a given distance along the gun barrel may
est to the bore surface had reached their maximum temperature be determined by integrating equation (1) between rx and ro as
indicating that no further heat was being transferred to these sec- shown in the following
tions of the gun barrel by the propellant gases. The temperature
distribution in the gun steel at this time was also nearly linear
which allowed an estimate of the gun barrel temperature profile to
be made by extrapolation to the bore surface. Fig. 6 depicts the
Q = 2irpc
f rATdr (2)

This integral was evaluated in the following manner. Values of

temperature versus distance in the gun steel for the firings with no r AT at various distances r were determined from plots of tempera-
additive present. ture versus distance into the steel such as Fig. 6. This results in
The estimate of the total heat input was made from these data plots of r AT versus r as shown in Fig. 7, for each of the three dis-

612 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 * .406mm FROM BORE SURFACE
2= . 8 6 4
3 = 2.51
4 = 5.56

200 400 600 800

TIME,(ms)

Fig. 5 Temperature versus time at 27.94 cm location—polyurethane

foam with flaps

Table 4 Results of barrel temperature measurements In 37 mm gun tube Table 5. Clearly the heat input at each axial location is reduced by
with wear-reducing Additives1"'6' the wear-reducing additives, and the positioning of the additive is
Time a f t e r prijiier No Additi P c . no f l a p s P. f l a p s Ti0 7 /wax, flaps Talc/Wa critical in determining the amount of heat reduction, as was sug-
rmpact, las T,K T,K _ T,K "T,K T,K gested from the firing tests in Table 1. Since the maximum cham-
25 494 475 467 471 467 ber pressure and muzzle velocity of the projectile are unaffected by
50 476 456 451 449 442 the wear-reducing additive, it appears that the additive does not
100 423 40G 392 394 395 reduce the temperature of the propellant gas. It would seem that
250 360 353 347 351 347 the rate of heat transferred through the propellant gas boundary
500 337 333 330 332 330 layer at bore surface was reduced in some manner.
1000 323 321 320 320 319 The result that the heat reduction for polyurethane foam with
flaps is equivalent to the heat reduction for the Ti02/wax and talc/
a t u r e s r e p r e s e n t average v a l u e s for t e n rounds Standard d e v i a t i i
wax with flaps implies that the addition of flaps or a disk to the
s l e s s than 0.6°K i n every c a s e . forward end of the propelling charge will improve the wear-reduc-
"Temperature measured a t 27.9cm from base of c a r t r i d g e case a t a depth of 0.406mm ing capability of the polyurethane foam and possibly eliminate the
from bore s u r f a c e .
c
need to use the more expensive TiC>2/wax additive. These results
P o l y u r e t h a n e foam.
contradict the conclusion that the metal oxide/wax additives are
superior to polyurethane foam, since this conclusion was reached
from full-scale firing tests in which only the TiC^/wax or talc/wax
tances along the gun barrel at which thermocouples were placed. liners had flaps [4].
The integral in equation (2) was then determined at each of these No data are shown for the TiC^/wax or the talc/wax without
axial positions from the area under the rAT versus r curve. This flaps, since the additive liners were not totally consumed, and the
procedure was followed for firings with no additive, with polyure- resulting temperatures were the same as the temperatures record-
thane foam (with and without flaps), and with TiOa/wax and talc/ ed with no additive. Similar results were obtained inadvertently
wax, each with flaps. A summary of these results is presented in [8] when the Ti02/wax additive was improperly placed in the rear

1= 27.94cm
2=104.14
3 = 180.34

Fig. 6 Temperature distribution at 100 ms at three

longitudinal locations along the barrel—no additive

4 6
DISTANCE FROM BORE SURFACE, (mm)

Journal of Heat Transfer NOVEMBER 1975 / 613

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 5 Estimated total heat transferred into the gun barrel at three axial
locations for various additives
Additivi and
Configu] i t i o n Heat T r a n s f e r r e d , J/eijn, a t Axial Location Indii

Polyurethane (no flaps}

Polyurethane (flaps)

Ti0 2 /wax (flaps)

Talc/wax (flaps)

Measured a t 27.9cm from the base of the c a r t r i d g e c a s e .

"Measured a t 104.1cia from the base of the c a r t r i d g e c a s e .

c
Measured a t 180.3cm from the base of the c a r t r i d g e c a s e .

20 21 22
References
1 Ward, J. R., "Survey of Wear-Reducing Additives," BRL Memoran-
Fig. 7 Curve of rA T versus r for data from Fig. 6 dum Report No. 2172, Apr. 1972.
2 Dickinson, L. A., and McLennan, D. E., "Improvement of the Firing
Accuracy and Cost Effectiveness of Guns Through the Use of Urethane
Foams," J. Cellular Plastics, 1968, pp. 184-187.
of a 105-mm cartridge case and some unburned residue was also 3 Ek, S.*Y., and Jacobsen, D. E., "Engineering Study of Barrel Wear Re-
ducing Additives," Wegematic Corporation Report, Part A, May 1962.
left after firing. It was noted that the rates of erosion were the 4 Wolff, R. O., "Reduction of Gun Barrel Erosion Part II. Barrel Wear
same as the rates obtained with no additive present. Reducing Additives," Picatinny Arsenal Technical Report No. 3096, Aug.
1963.
4 Conclusions 5 Picard, J., and Trask, R. L., "A New Gun Barrel Erosion Reducer," J.
Spacecraft and Rockets, Vol. 5,1968, pp. 1487-1488.
1 Heat transfer to gun barrels is reduced by wear-reducing ad- 6 Plett, E. G., Alkidas, A. C, Schrader, R. E„ and Summerfield, M„
ditives. "Erosion of Metals by High Pressure Combustion Gases: Inert and Reactive
2 The reduction in heat transfer is strongly a function of the Erosion," ASME Paper No. 75-HT-M.
location of the additive with respect to the propellant and the pro- 7 Brosseau, T., "An Experimental Method for Accurately Determining
the Temperature Distribution and the Heat Transfer in Gun Barrels," BRL
jectile. For polyurethane foam, Ti02/wax additive, and talc/wax Report No. 1740, Sept. 1974.
additive, the reduction of heat transfer to the barrel was the same 8 Frankle, J. M., "Interior Ballistics of High Velocity Guns Experimen-
when each additive was equipped with flaps. tal Program—Phase 1," BRL Memorandium Report No. 1879, Nov. 1967.

814 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 [I'llllllff \
]
u ul\{ j < "in J < i m t B
/

This section consists of contributions of 1500 words or equivalent. In computing equivalence, a typical one-column figure
or table is equal to 250 words. A one-line equation is equal to 30 words. The use of a built-up fraction or an Integral sign
or summation sign in a sentence will require additional space equal to 10 words. Technical notes will be reviewed and
approved by the specific division's reviewing committee prior to publication. After approval such contributions will be
published as soon as possible, normally In the next Issue of the journal.

On the Interface Temperature 1 !>T

= V2T (1)
of Two Suddenly Contacting a at
Materials 1 in the two contacting bodies. For most practical conduction prob-
lems Fourier's formulation for the heat conduction is acceptable.
However, as has been discussed by several investigators, Fourier's
heat conduction equation leads to erroneous results when very
M. S. Kazimi 2 and C. A. Erdman 2 short times are considered. The transient behavior of two contact-
ing materials at short times is of interest in many applications.
The interface temperature of two suddenly contacting semi-infi- Boley [l] 3 and Brazel and Nolan [2] discussed the significance of
nite bodies is developed using the hyperbolic heat-conduction non-Fourier effects in assessment of the thermal stresses that may
equation. The times for which the non-Fourier effects are impor- be generated when the surface temperature of a structure is sud-
tant are assessed. A numerical example for a system of uranium denly changed. Fauske [3] investigated the correlation between the
dioxide and sodium is considered. instantaneous interface temperature of two suddenly contacting
liquids and the potential for rapid development of spontaneous
Nomenclature nucleation in the cooler liquid. It is this latter concern that moti-
c = specific heat vated the current study.
k = thermal conductivity The purpose of this paper is to determine the interface tempera-
q = heat flux ture of two suddenly contacting semi-infinite bodies using the hy-
s = the Laplace transform parameter perbolic heat-conduction equation. The times for which the non-
t = time Fourier effects on the interface temperature are important are also
T = temperature
x = linear distance coordinate
a = thermal diffusivity 2 The Hyperbolic Formulation
0 = pc\ Consider a semi-infinite body of a uniform initial temperature
7 = X2/a T i suddenly contacting, at x = 0, a semi-infinite body of a uniform
X = speed of heat propagation initial temperature T% where T\ < T 2 .
p = density By applying the hyperbolic heat conduction equation to the
0 = temperature deviation from the temperature at t = 0 cooler semi-infinite medium we obtain:
Subscripts a26(x, t) 1 aB(x, t) 1 a28(x, t)
x>0 (2)
1 = for positions where x > 0 ax 2
a\ at \\z at*
2 = for positions where x < 0
where 8{x, t) = T(x, t) - Ti
in = at the interface (at x = 0)
Equation (2) describes the behavior of a damped temperature
1 Introduction wave with a finite speed of propagation, Xi. When Xx -* •», equa-
The instantaneous interface temperature of two suddenly con- tion (2) reduces to the more familiar Fourier heat conduction
tacting bodies has been commonly determined from the solution of equation.
the Fourier heat equations: The initial and boundary conditions to be applied to the wave
equation are:
a8(x, t)
(x, t) = Oand •0 at t = 0 and x > 0 (3)
1
Work performed under the auspices of the United States Nuclear Regu- at
latory Commission. (4)
2
Fast Reactor Safety Division, Department of Applied Science, Brook- B(x,t flin(t) = Tin(t) • • Ti as x
haven National Laboratory, Upton, N. Y.
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat !
Transfer Division April 10, 1975. Numbers in brackets designate References at end of technical note.

Journal of Heat Transfer NOVEMBER 1975 / 615

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 B(x,t) = 0 as 2 — °° (5) Tin ~ T t
= Vp2C2k2/plCikl (16)
Physically, the conditions of equation (3) imply that it takes the T2 - T h
temperature wave a finite amount of time to propagate into the The Fourier equation expression for the interface temperature
domain where x > 0. can be obtained from equation (13) by applying the condition:
By applying the conditions of equations (3), (4), and (5), the La-
place transform of equation (1) takes the form: L i m s - o s £ | / ( 0 ! = L i m ^ „ \f(t)} (17)
2
£|0(x, t)l = Hx, s) = ( U s ) exp | - x V s A i + s / ^ | 2
(6) In reality, only a very short time will elapse before the interface
temperature approaches the prediction of the Fourier equations.
The heat flux equation that is consistent with the wave equation This will be discussed in the following sections.
is of the form [4, 5]
4 Short-Time Behavior of the Interface Temperature
a(Hx, t) «i eqi(x, t)
t/iU', 0 = -ki (7) The exact solution for T i n ( 0 obtained from equation (13) using
at the convolution theorem contains complicated integrals which
which reduces to the Fourier heat flux equation as Xi -* =>. When must be evaluated numerically. However, if one is interested in
equation (7) for the heat flux is substituted in the differential very short times after contact, equation (13) may be simplified to
equation for conservation of energy in an elemental volume, the obtain an analytical solution. By applying the conditions
wave equation for heat conduction is produced.
* » Xi 2 /«i and s » X 2 2 /« 2 (18)
Taking the Laplace transform of equation (7) and applying the
condition and

qi(x, 0) = 0 (8) Ti = 7Vs and T 2 = T2/s (19)

to equation (13) and rearranging the terms we obtain:
within the semi-finite body results in the following equation:
1 Tx/3i(s + 72/2) + T2p2(s + 71/2)
<>hx,s) =-qi(x,s)\
r1 + «is]
— —
i (9)
Ti„(s)=-
ax L Xi^Jfti ?i(s + 72/2) + ft (s + 7i/2)
Tift + T 2 ft
By differentiating equation (6), we obtain:
S (ft + ft) + (ft72 + ft7i)/2
dh(x, s)
• = - v V A i 2 + s/ai 0 in (s) exp \-x Vs 2 /Xi 2 + s/c*i| ( T i f t 7 2 + T 2 ft7i)/2
dX (20)
•sMft + ft) + ((?m + ftm)/2]
(10)
where
By equating the right-hand side of equations (9) and (10) and ;
considering the resultant equation for x = 0, we get p,c; X;; i = 1 or 2 (21)

«i Vs + X x /ai . 2 and
3in(s) = TM - fr qi(0,s) (11)
ki\ Vs 7; = Xi a /a ; ; i = 1 or 2 (22)
Similar analysis can be applied to the region x < 0 (i.e., the rela- The inverse transformation of equation (20) can be readily deter-
tively hot medium) to correlate the instantaneous temperature mined to be:
drop at the interface (when approached from x < 0) to the initial
interface heat flux. The result may be written in the form: „ ,„ T t ft + T2ft / P172 + P271 \
Tin(0 = — ^ — ; exp
ft + ft * V 2(ft + ft) /
Vs + X 2 2 /a 2
r2-Tin(S); §2(0, S) (12) T i f t 7 2 + T202yi [" / ft72 + fe7i\]
k2\2 V! exp
(23)
ft72" + ftm L V 2 (ft + ft) ) \
From considerations of heat flux continuity at the interface
In the limit t —• 0, equation (23) yields the instantaneous interface
equations (11) and (12) may be used to obtain:
temperature as given in equation (15).
Tj n (s) - f\ p2c2\2 Vs + Xi 2 /ai
(13) 5 L o n g - T i m e B e h a v i o r of t h e I n t e r f a c e T e m p e r a t u r e
T 2 - Tin(s) Plci\i Vs + X 2 2 /a 2
An analytic solution can also be obtained for long-time behavior
of the interface temperature if the conditions
3 The Instantaneous Interface Temperature
The instantaneous temperature at the interface can be obtained s « Xi 2 /ai and s « X22/a (24)
from equation (13) by applying the condition
were applied to equation (13). The expression for long-time behav-
L i m s „ „ s £ | / ( 0 | = Limt-.o/W (14) ior thus obtained is given by:

to both sides of the equation. This condition yields the following Tift V T 7 + T202V^2 2(ft72V 71 + ft7i^72
Ti n = — ; — , — , „ — ; = — exp 2
relation: ftV7l + ft '72 (^ J P1V71 + 72 >)
7'i„(0) - T, P2 e 2 X 2 Tift V ^ + T 2 ftV'^i
(15)
T2-Tin(0) piCiXi ftV72 + 02Vy[
Physically, equation (15) implies that the interface temperature 2(^172^71 + ft7l V~y.
is determined by equalization of the amount of heat that can prop-
agate within the hot medium per unit length-per-unit time if the
I 1 - exp
exp (- -
3 l V 7 ! + ft 72
(25)
*)]
temperature is to drop suddenly from T 2 to Ti n , p2c2\2 (T 2 — Tin), In the limit t — <=, equation (25) yields the instantaneous interface
to the amount of heat that can propagate into the cold medium- temperature as given in equation (16).
per-unit length per unit time inducing sudden change in tempera- From the conditions (24), it is clear that the interface tempera-
ture from T\ to l\n, p\C\\\ (T\„ - Tj). ture approaches the long term value if
Equation (15) is equivalent to the more familiar expression for t » ai/Xx2 or t » a 2 /X 2 2
the instantaneous inu-rfi'ice temperature predicted by the Fourier
equations for heat transfer: which ever is larger.

616 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 1 1 1
U0 2 INITIAL TEMPERATURE - TJo7 INITTArfEM'pE'WATuRE
3000 2800 -
o
°
tu
2 600 2400 _ ~~—^_^
~l
H
^**s \
<
IE \\ \
" 2200
a. 2000
\ \ \ \ \\ LONG TERM
> \ \ \ SOLUTION
I- SHORT TERM \ A
1600 SOLUTION \ \.
u
It
IE
1400

INTE
1200

-FROM FOURIER S EQUATION

SODIUM INITIAL TEMPERATURE
800
1 1 1 I 1
10" IO-1 10" 10" 10"
300 500 700 900 TIME , sec
SODIUM INITIAL TEMPERATURE, °C
Fig. 2 Short term and long term behavior of the interface temperature
Fig. 1 Instantaneous Interface temperature of U0 2 , at 3000°C, and sodi-
um
2 Brazel, J. P., and Nolan, E. J., "Non-Fourier Effects in the Transmis-
sion of Heat," Proceedings 6th Conference on Thermal Conductivity, Day-
ton, Ohio, 1966.
Table 1 Values of properties used in the calculations*
3 Fauske, H. K., "On the Mechanism of Uranium Dioxide Sodium Ex-
plosive Interactions," Nucl. Sci. Eng., Vol. 51,1973, pp. 95-101.
Property UO, Na 4 Chester, M., "Second Sounds in Solids," the Physical Review, Vol. 131
No. 5, Sept. 1963, pp. 2013-2015.
3
p (gm/cm ) 8.52 0.83 5 Maurer, M. J., "Relaxation Model for Heat Conduction in Metals,"
C p (cal/gm°C) 0.12 0.31 Journal of Applied Physics Vol. 40,1969, p. 5123.
k(cal/cm°Csec) 0.005 0.0915 6 Chasanov, M. G., Leibowitz, L., and Gabelnick, S. D., "High Tempera-
X** (cm/sec) 1.7 X 10 5 2.3 X 10 s ture Physical Properties of Fast Reactor Materials," J. Nucl. Materials,
Vol. 49, 1973-1974, pp. 129-135.
7 Nettleton, R. E., "Relaxation Theory of Thermal Conduction in Liq-
* All values are based on the properties recommended uids," Phy. of Fluids, Vol. 3,1960, p. 216.
by Ref. 6. 8 Weymann, H. D., "Finite Speed of Propagation in Heat Conduction,
** The speed of heat propagation is assumed to equal Diffusion and Viscous Shear Motion," American Journal of Physics, Vol.
the speed of sound propagation. 35,1967, p. 488.

6 A Numerical Example
Consider the case of sudden contact between molten UO2 and
molten sodium, as may be hypothesized under accident conditions
Thermal Bond Conductance
in a sodium-cooled nuclear reactor. The instantaneous interface Between a Cylinder and a Plane
temperatures that are predicted by the wave heat-conduction Wall
equation and the Fourier equation are shown in Fig. 1, for initial
UO2 temperature of 3000° C and various sodium temperatures.
F. W. Ahrens
The values of the physical properties used in this analysis are
given in Table 1. It is clear that substantial difference exists be-
tween the two predictions. The short-time and long-time solutions
for the case of UO2 at 3000° C contacting suddenly with sodium at
800°C are shown in Fig. 2. From these results, it is seen that the
Fourier-predicted interface temperature is approached in approxi-
Nomenclature
mately 5 X 1 0 ~ u seconds. For most applications involving UO2 Y2/2d, dimensionless geometrical parameter
and sodium, therefore, the transient time in which the interface V l + 2R/d, dimensionless geometrical parameter
temperature of two suddenly contacting bodies deviates from the C = constant defined following equation (1)
values predicted by Fourier's equations is expected to be so small d = spacing between cylinder and wall (>0)
as to be inconsequential. / = dimensionless heat transfer factor, equation (4)
In this numerical example, the thermal speed is assumed to
equal the speed of sound. While this assumption seems to be ac- k = thermal conductivity of bond material
ceptable if applied to ordinary liquids at room temperature [7], it Kc = thermal conductance of bond, per unit length
has not been verified for either UO2 or Na. The prediction of ther- q" = heat flux
mal speed continues to be a subject of interest in thermodynamics q' = heat transfer rate per unit cylinder length
literature. Values of thermal speed that range between higher and R - radius of cylinder
lower than the speed of sound have been predicted for solids and T = temperature
liquids [5, 7, 8]. For gases, the thermal speed is predicted to be al-
ways less than the sound speed [8].
1
Research Engineer, Whirlpool Corp., Elisha Gray II Research and Engi-
References neering Center, Benton Harbor, Mich. Mem. ASME.
1 Boley, B. A., The Analysis of Problems of Heat Conduction and Melt- Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ing: High Temperature Structures and Materials Pergamon Press, New ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat
York, 1964, pp. 260-315. Transfer Division February 27,1975.

Journal of Heat Transfer NOVEMBER 1975 / 017

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 1 1 1
U0 2 INITIAL TEMPERATURE - TJo7 INITTArfEM'pE'WATuRE
3000 2800 -
o
°
tu
2 600 2400 _ ~~—^_^
~l
H
^**s \
<
IE \\ \
" 2200
a. 2000
\ \ \ \ \\ LONG TERM
> \ \ \ SOLUTION
I- SHORT TERM \ A
1600 SOLUTION \ \.
u
It
IE
1400

INTE
1200

-FROM FOURIER S EQUATION

SODIUM INITIAL TEMPERATURE
800
1 1 1 I 1
10" IO-1 10" 10" 10"
300 500 700 900 TIME , sec
SODIUM INITIAL TEMPERATURE, °C
Fig. 2 Short term and long term behavior of the interface temperature
Fig. 1 Instantaneous Interface temperature of U0 2 , at 3000°C, and sodi-
um
2 Brazel, J. P., and Nolan, E. J., "Non-Fourier Effects in the Transmis-
sion of Heat," Proceedings 6th Conference on Thermal Conductivity, Day-
ton, Ohio, 1966.
Table 1 Values of properties used in the calculations*
3 Fauske, H. K., "On the Mechanism of Uranium Dioxide Sodium Ex-
plosive Interactions," Nucl. Sci. Eng., Vol. 51,1973, pp. 95-101.
Property UO, Na 4 Chester, M., "Second Sounds in Solids," the Physical Review, Vol. 131
No. 5, Sept. 1963, pp. 2013-2015.
3
p (gm/cm ) 8.52 0.83 5 Maurer, M. J., "Relaxation Model for Heat Conduction in Metals,"
C p (cal/gm°C) 0.12 0.31 Journal of Applied Physics Vol. 40,1969, p. 5123.
k(cal/cm°Csec) 0.005 0.0915 6 Chasanov, M. G., Leibowitz, L., and Gabelnick, S. D., "High Tempera-
X** (cm/sec) 1.7 X 10 5 2.3 X 10 s ture Physical Properties of Fast Reactor Materials," J. Nucl. Materials,
Vol. 49, 1973-1974, pp. 129-135.
7 Nettleton, R. E., "Relaxation Theory of Thermal Conduction in Liq-
* All values are based on the properties recommended uids," Phy. of Fluids, Vol. 3,1960, p. 216.
by Ref. 6. 8 Weymann, H. D., "Finite Speed of Propagation in Heat Conduction,
** The speed of heat propagation is assumed to equal Diffusion and Viscous Shear Motion," American Journal of Physics, Vol.
the speed of sound propagation. 35,1967, p. 488.

6 A Numerical Example
Consider the case of sudden contact between molten UO2 and
molten sodium, as may be hypothesized under accident conditions
Thermal Bond Conductance
in a sodium-cooled nuclear reactor. The instantaneous interface Between a Cylinder and a Plane
temperatures that are predicted by the wave heat-conduction Wall
equation and the Fourier equation are shown in Fig. 1, for initial
UO2 temperature of 3000° C and various sodium temperatures.
F. W. Ahrens
The values of the physical properties used in this analysis are
given in Table 1. It is clear that substantial difference exists be-
tween the two predictions. The short-time and long-time solutions
for the case of UO2 at 3000° C contacting suddenly with sodium at
800°C are shown in Fig. 2. From these results, it is seen that the
Fourier-predicted interface temperature is approached in approxi-
Nomenclature
mately 5 X 1 0 ~ u seconds. For most applications involving UO2 Y2/2d, dimensionless geometrical parameter
and sodium, therefore, the transient time in which the interface V l + 2R/d, dimensionless geometrical parameter
temperature of two suddenly contacting bodies deviates from the C = constant defined following equation (1)
values predicted by Fourier's equations is expected to be so small d = spacing between cylinder and wall (>0)
as to be inconsequential. / = dimensionless heat transfer factor, equation (4)
In this numerical example, the thermal speed is assumed to
equal the speed of sound. While this assumption seems to be ac- k = thermal conductivity of bond material
ceptable if applied to ordinary liquids at room temperature [7], it Kc = thermal conductance of bond, per unit length
has not been verified for either UO2 or Na. The prediction of ther- q" = heat flux
mal speed continues to be a subject of interest in thermodynamics q' = heat transfer rate per unit cylinder length
literature. Values of thermal speed that range between higher and R - radius of cylinder
lower than the speed of sound have been predicted for solids and T = temperature
liquids [5, 7, 8]. For gases, the thermal speed is predicted to be al-
ways less than the sound speed [8].
1
Research Engineer, Whirlpool Corp., Elisha Gray II Research and Engi-
References neering Center, Benton Harbor, Mich. Mem. ASME.
1 Boley, B. A., The Analysis of Problems of Heat Conduction and Melt- Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ing: High Temperature Structures and Materials Pergamon Press, New ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat
York, 1964, pp. 260-315. Transfer Division February 27,1975.

Journal of Heat Transfer NOVEMBER 1975 / 017

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Tt = temperature of cylinder surface The expression for the local value of (au=ax)x=o is determined
Tw = temperature of plane wall surface from complex variable theory to be:
u = T - Tw _ 1CXr
du
x,y = spatial coordinates; see Fig. 1
dx = 0 Xc2 + y1
xc = location of "source"
Yi = bond half-width at plane wall Use of this result in equation (2) gives:
Y2 = 2Y 1
z — x + iy q> = -fkY^T, - Tj/d (3)
where:
Introduction Y, = 2Y<
When a cylindrical tube or rod is bonded to a plane wall by a
material of low thermal conductivity relative to that of the tube 2 tan' • 1 /
and wall, the thermal conductance of the bond is important in pre- (4)
dicting the rate of heat transfer between the cylinder and wall. ,b + 1,
a In
Such configurations are typically found in applications where low
cost is paramount (e.g., solar collectors, "warm wall" condensers in = Y 2 /2d = Yi/d
small refrigeration devices, etc.). In this technical brief, an expres-
sion for the thermal conductance of such a bond is developed, and
based on an idealized model. -/I + ZR/d
For the purpose of achieving a simple analytical expression for Thus, the thermal conductance per unit cylinder length, defined
the bond conductance, the following idealizations are made: by9' = K c m - T J , i s :
1 Heat flow is two-dimensional in a plane perpendicular to the
cylinder axis. Kc = kfY2/d (5)
2 Heat losses from the exposed bond surfaces are negligible.
It can be shown that for ft » d and Yt « ft the above result reduc-
3 The portions of the plane wall and cylinder in contact with
es to that for heat conduction between two infinite parallel planes
the bond material have uniform, but different, temperatures.
(i.e., / —>• 1). Also, for the caseYj —«• <», the result becomes equiva-
4 The exposed surfaces of the bond have shapes defined by
lent to that for the buried cable problem [2]. In all cases, / < 1.
heat flow lines of the "buried cable" [1, 2] 2 or symmetrical source-
sink problem [3, 4]. A typical bond geometry resulting from this Finally, to more explicitly define the bond geometry for which
idealization is shown in Fig. 1. equation (5) rigorously applies, the following information about
the exposed bond surface has been derived. Considering the por-
Using these conditions, it is clear that the temperature distribu-
tion lying in the first quadrant (see Fig. 1), the shape is a circle
tion in the bond will be the same as that in the corresponding spa-
whose center is at (0, (Yi 2 — x c 2 ^/2Yi) and whose radius is (Yi 2 +
tial subdomain of the symmetrical source-sink problem.
xc2)/2Yi. This bond surface intersects the cylinder at the point (x,
y) defined by:
Analysis3
Consider the cylinder and wall to be located in an x-y coordinate (2ft + d)[d(2R + d) + Y t 2 ]/[(2ft + df + Yj 2 ]
system as shown in Fig. 1. The temperature distribution is [4]: (6)
2
2ftY!(2R + d)/[(2R + df + Yt ]
u = C In \z + xr_ (1) Discussion
The bond conductance derived here is exact for the bond geome-
where:
try assumed. The only geometrical parameters on which it depends
u = T(x, y) — Tw, for the present application
are YJd and ft/d. The primary usefulness of the results, however,
z = x + iy
may be for approximating the conductance of bonds having a re-
C = proportionality constant
semblance, but not exact similarity, to the idealized geometry. In
xc = source location
this regard, it should be noted that more general bond geometries
The relation between the source location and more pertinent
cause the point of contact between bond surface and cylinder and
parameters of the problem—the minimum distance between wall
the shape function of the exposed surface to become additional in-
and cylinder, d, and the cylinder radius, ft—may be shown to be:
dependent parameters. The following procedure should bracket
xc = Vd(d + 2ft). The constant C in equation (1) may be deter-
mined from the fact that u = Tt — Tw at z = d, giving:

C
ln(-i
- d
The heat transfer rate per unit cylinder length is determined by
integration of the local heat flux at the plane wall over the contact TUBE OR CYLINDER
width, 2 Yi:

2 / n q"iy)dy (2)

where:
3M,
q"ly) = -k dx'~)
THERMAL BOND MATERIAL
2
Numbers in brackets designate References at end of technical note.
3
A reviewer has noted that this problem could have been analyzed by a
direct application of a generalized theory of conduction shape factors [5].
With that method, a more natural (bicylindrical) coordinate system repre- PLANE WALL
sentation may be employed to achieve a final expression equivalent to equa-
tion (4). Fig. 1 Thermal bond configuration

818 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the conductance of such bonds. For the actual dimensions Yi, R,
and d, calculate the point of intersection of the idealized bond
with the cylinder, using equation (6). Then find, by trial and error,
a second value of Y\ which causes the ideal intersection point to
coincide with that of the actual bond. Finally, calculate the ideal r
conductances corresponding to the two values of Y\. One would ex- b
pect the actual conductance to fall in between these values.
r
References a ak-
1 Carslaw, H. S., and Jaeger, J. C, Conduction of Heat in Solids, Sec-
ond ed., Oxford University Press, London, 1959, pp. 449-451.
2 Ecken, E. R. G., and Drake, R. M, Jr., Heat and Mass Transfer,
1 z
McGraw-Hill, New York, 1959, pp. 60-64. k
) n
!
3 Lamb, Sir Horace, Hydrodynamics, Sixth ed., Dover Publications,
First American Edition, New York, 1945, p. 70.
4 Kreyszig, Erwin, Advanced Engineering Mathematics, Wiley, New ho
York, 1962, pp. 745-746. Fig. 1 Cylindrical geometry with circumferential channels (configuration
5 Yovanovich, M. Michael, "A General Expression for Predicting Con-
duction Shape Factors," Progress in Astronautics and Aeronautics, Ther- for Ftci)
mophysics and Spacecraft Thermal Control, Vol. 35, MIT Press, 1973, pp.
265-291.
Introduction
The work in reference [l] 3 treats the problem of flat plate-chan-
nel heat exchangers (for deep ocean applications) whose external
surface was cooled by convection, while the internal surface under
The Effect of Thermal the channels was heated by the channel fluid. That flat plate con-
figuration is similar to the one shown in Fig. 1, except that the
Constriction Resistance in the radii, ra and rt, are infinite. In reference [1] a correction factor
Design of Channel-Plate Heat CUHF was determined, which could be applied to the uniform re-
Exchangers: Cylindrical sistance of the flat plate (t/bk) in order to determine the actual re-
Geometry sistance in the plate caused by the constriction of the flux lines due
to the smaller channel width,
R. P. Forelund 1 and H. Q. Oliveira 2 itc(UHF) = (t/bk)CvliF (1)

The correction factor CUHF was based on a uniform heat flux

Nomenclature (UHF) boundary condition under the channel. The study [1] also
considers the effect of a convective heat flux (CHF) boundary con-
a = half channel width
dition at the channel surface, and a second correction factor, Cc,
b = half distance between tubes
for it was determined:
C = constriction coefficient
Cc = correction factor for the convective heat flux case /t c (CHF) = (t/bk)CvwCc (2)
h = heat transfer coefficient
Io, 11, K0, Ki = Bessel functions The intention of the work in reference [2] (the first part of which
k = thermal conductivity is reported in [1]), was to determine correction factors that could
Re = thermal constriction resistance of the plate be applied to the uniform flat plate resistance (t/bk) in order to
r = radius analyze either cylindrical or spherical configurations. For example,
t = plate thickness the constriction resistance for the two cylindrical configurations,
z = coordinate parallel to plate and normal to channel shown in Figs. 1 and 2 would be given by
(p = coordinate parallel to plate and normal to channel r t d ( U H F ) = (t/bk)CvwGx (3)
0o = hot Ik
£o, li> fe, = defined by equations (5), (6), and (7) ic C2 (UHF) = (t/bk)CVHFG2 (4)
^o, ^ i . 4>% = defined by equations (5), (6), and (7)
<Po', 'Pi, fa' = defined by equations (5), (6), and (7)
A = eigen values 3
7 = eigen values Numbers in brackets designate References at end of technical note.

Subscripts
a = plate surface in contact with channel fluid
6 = plate surface in contact with external fluid
0 = external or flat plate
1 = cylindrical, axial channels
2 = cylindrical, circumferential channels
CHF = convective heat flux
UHF = uniform heat flux

1
Assoc. Professor, Institute Militar de Engenharia, Rio de Janeiro, Brazil.
Assoc. Mem. ASME.
2
Asst. Professor, Escola de Engenharia, Universidade Federal do Espirito
Santo, Vitoria, Brazil.
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat
Transfer Division May 12,1975. Fig. 2 Cylindrical geometry with axial channels (configuration for RC\)

Journal of Heat Transfer NOVEMBER 1975 / 619

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the conductance of such bonds. For the actual dimensions Yi, R,
and d, calculate the point of intersection of the idealized bond
with the cylinder, using equation (6). Then find, by trial and error,
a second value of Y\ which causes the ideal intersection point to
coincide with that of the actual bond. Finally, calculate the ideal r
conductances corresponding to the two values of Y\. One would ex- b
pect the actual conductance to fall in between these values.
r
References a ak-
1 Carslaw, H. S., and Jaeger, J. C, Conduction of Heat in Solids, Sec-
ond ed., Oxford University Press, London, 1959, pp. 449-451.
2 Ecken, E. R. G., and Drake, R. M, Jr., Heat and Mass Transfer,
1 z
McGraw-Hill, New York, 1959, pp. 60-64. k
) n
!
3 Lamb, Sir Horace, Hydrodynamics, Sixth ed., Dover Publications,
First American Edition, New York, 1945, p. 70.
4 Kreyszig, Erwin, Advanced Engineering Mathematics, Wiley, New ho
York, 1962, pp. 745-746. Fig. 1 Cylindrical geometry with circumferential channels (configuration
5 Yovanovich, M. Michael, "A General Expression for Predicting Con-
duction Shape Factors," Progress in Astronautics and Aeronautics, Ther- for Ftci)
mophysics and Spacecraft Thermal Control, Vol. 35, MIT Press, 1973, pp.
265-291.
Introduction
The work in reference [l] 3 treats the problem of flat plate-chan-
nel heat exchangers (for deep ocean applications) whose external
surface was cooled by convection, while the internal surface under
The Effect of Thermal the channels was heated by the channel fluid. That flat plate con-
figuration is similar to the one shown in Fig. 1, except that the
Constriction Resistance in the radii, ra and rt, are infinite. In reference [1] a correction factor
Design of Channel-Plate Heat CUHF was determined, which could be applied to the uniform re-
Exchangers: Cylindrical sistance of the flat plate (t/bk) in order to determine the actual re-
Geometry sistance in the plate caused by the constriction of the flux lines due
to the smaller channel width,
R. P. Forelund 1 and H. Q. Oliveira 2 itc(UHF) = (t/bk)CvliF (1)

The correction factor CUHF was based on a uniform heat flux

Nomenclature (UHF) boundary condition under the channel. The study [1] also
considers the effect of a convective heat flux (CHF) boundary con-
a = half channel width
dition at the channel surface, and a second correction factor, Cc,
b = half distance between tubes
for it was determined:
C = constriction coefficient
Cc = correction factor for the convective heat flux case /t c (CHF) = (t/bk)CvwCc (2)
h = heat transfer coefficient
Io, 11, K0, Ki = Bessel functions The intention of the work in reference [2] (the first part of which
k = thermal conductivity is reported in [1]), was to determine correction factors that could
Re = thermal constriction resistance of the plate be applied to the uniform flat plate resistance (t/bk) in order to
r = radius analyze either cylindrical or spherical configurations. For example,
t = plate thickness the constriction resistance for the two cylindrical configurations,
z = coordinate parallel to plate and normal to channel shown in Figs. 1 and 2 would be given by
(p = coordinate parallel to plate and normal to channel r t d ( U H F ) = (t/bk)CvwGx (3)
0o = hot Ik
£o, li> fe, = defined by equations (5), (6), and (7) ic C2 (UHF) = (t/bk)CVHFG2 (4)
^o, ^ i . 4>% = defined by equations (5), (6), and (7)
<Po', 'Pi, fa' = defined by equations (5), (6), and (7)
A = eigen values 3
7 = eigen values Numbers in brackets designate References at end of technical note.

Subscripts
a = plate surface in contact with channel fluid
6 = plate surface in contact with external fluid
0 = external or flat plate
1 = cylindrical, axial channels
2 = cylindrical, circumferential channels
CHF = convective heat flux
UHF = uniform heat flux

1
Assoc. Professor, Institute Militar de Engenharia, Rio de Janeiro, Brazil.
Assoc. Mem. ASME.
2
Asst. Professor, Escola de Engenharia, Universidade Federal do Espirito
Santo, Vitoria, Brazil.
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat
Transfer Division May 12,1975. Fig. 2 Cylindrical geometry with axial channels (configuration for RC\)

Journal of Heat Transfer NOVEMBER 1975 / 619

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 UHF.1 r
i/t=5

100
I I i i i I 1
!
_1_
h o t/k = 1.0 6 as IJBO 53 Wo os to a 0.5 1.0
CUHF,1 - - "-UHF.I
r a /t= 5 a/b
;\ :
Fig. 4 Geometry factor for axial channels for rblrB = 1.2
10.0 -t/b
-0A\ -

oA
5.0 -
1
4>i' = yjttira-t'-1 - ra-V-1]
'0.5V
fi = hj - <t>orb)rb~2y'/(yj + 4>o>-b)
" as
i i I i i i
Jj = farjb, j = 1, 2, 3 , . . .
0,5 1.0 0
a/b and
Fig. 3 Constriction coefficient for axial channels, CUHF.I \j = jir/b,j= 1,2,3,...
Cylinder—Circumferential Channels.

Rcz = Va ln(rb/ra)/bk] (l - [1/lnWrJ]

for the UHF boundary condition. Here Gi and G^ are the geomet-
rical correction factors for the axial and circumferential channel • E W t o ' r o W s i n X,-a)A;a) 2 (7).
arrangements, respectively, as shown in Figs. 1 and 2. The spheri- j=i J
cal configuration is not mentioned in this article. Nor is the effect where
of the CHF boundary condition on the constriction resistance of
the cylindrical geometry treated here. i>i = fofo(X/„) + K0(\fa)

H = \j[hhihra) - K^Xfa)]
Analysis
The development of flci(UHF) and i?c2(UHF) for the cylindri- h = [XjtK^Xjrt) - 0 o tKo(Xyr 6 )]/[X;i/ 1 (\yr 6 )
cal configurations follows in much the same manner as that of + 4>ot Io(^Fb)]
flc(UHF) for the flat plate geometry given in [1]. Therefore, only
the resulting equations are presented here. and
Flat Plate. \j = jWb,j= 1,2,3,...

Re = (t/bk) [ l - E (2,fc,AWt)((sin Xj-aVXya)2! (5) Equations (6) and (7) can be divided by the uniform cylindrical

where

&> = to + 1
1.00
<Po' = Xy(& - 1)
G2 h o t / k = 0.01
£o = [(X;t - 0o£)/(Xji + <t>0t)] exp(-2Xyt) r
a/t = 5
and 0,96

\j = jir/b, j = 1, 2, 3 , .
Cylinder—Axial Channels. 092

Rci = [ra Mrb/ra)/bk] \l - [l/ln(rb/ra))

0B8,
• E (2<h/tyi'r 0 )[(sin Xya)Ajo]2[ (6) 0 OS 1.0 0 0.5 1.0 0 05 1.0 0 05 1.0
a/b

where Fig. 5 Geometry factor for circumferential channels for rtlr, = 1.2

620 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 plate resistance (r„ ln(n,/ra)/fefc) to obtain correction factors tances, and the increase in flux area with radius which diminishes
CUHP.I and CUHF,2 which are similar to CUHP for the flat plate the resistance. At low values of h0t/k, the flux lines are spread out
given in [1]. As an example, CUHF.I for the axial channel arrange- more evenly along the curved plate, and in turn suffer a greater
ment, is shown in Fig. 3. However, less importance is given to length of resistance. Therefore, values of Gi can be greater than
CUHF.I and CUHF,2 in this article, since it is the behavior of G i and unity, however over most of the range of parameters, Gi is less
G2 of equations (3) and (4) that are of interest. The geometry fac- than unity.
tors Gi and G2 are obtained by dividing equations (6) and (7) by
equation (5). Conclusion
In both cases, Gx and G2 vary by less than approximately ±10
Results percent about unity for 1.0 < rb/ra < 1.2; therefore, the designer of
Typical results for Gi and G 2 are shown in Figs. 4 and 5 for the deep ocean cylindrical channel-plate heat exchangers could initial-
case of Ti,lra - 1.2 (or ra/t = 5.0). This case was chosen as being a ly base his design on a flat plate model.
reasonable upper limit for rtlra for deep ocean applications, that For other applications of heat exchanger design for rtlra greater
is, the submarine hull thickness is 20 percent of the inside radius. than 1.2, equations (6) and (7) can be employed.
The other paraments used in the study fall within the range of Should the designer care to estimate the effect of the CHF
values for deep ocean design. boundary condition on the constriction resistance, Figs. 4, 5, and 6
In the case of both axial and circumferential channels, the for Cc of reference [1] could be applied, so that
values of G\ and G 2 approach the asymptote (0.913 in the case for
ri,/ra = 1.2) as a/6 goes to unity. The asymptote is simply the ratio flci(CHF) ss (t/6/j)C U H pGiC c
of the uniform cylindrical resistance to that of the flat plate of RC2(CHF) =* (t/bk)CuHFG2Cc
equal thickness, and is less than unity since the surface area for
heat flux increases with radius for the cylindrical configuration. Acknowledgments
As a/b goes to zero, Gi and G 2 approach unity, since the con- The numerical results presented in this article were obtained
striction resistance due to a point heat source becomes indepen- using the IBM 1130 at the Instituto Militar de Engenharia.
dent of the geometry. The values for G2 in Fig. 5 are always less
than unity since, although the length for the heat flux lines are the References
same as those for the flat plate, the area for heat flux increases 1 Oiiveira, H. Q., and R. P. Forslund, JOURNAL OP HEAT TRANS-
with radius for the cylindrical geometry, hence lower resistance. In FER, TRANS. ASME, Series C, Vol. 96, 1974, p. 292.
2 Oiiveira, H. Q., "Resistencia Termica em Trocadores do Tipo Tubo-
the case of axial channels, a trade-off exists between the increase Placa com Estrangulamento do Fluxo de Calor," MS thesis, Pontificia Univ-
in the length for the flux lines which contribute to larger resis- ersidade Catolica, Rio de Janeiro, 1973.

Turbulent Heat Transfer by a = heat transfer coefficient

Natural Convection Along an /3 = coefficient of cubical expansion
Mam = thickness of viscous sublayer
Isothermal Vertical Flat tH = eddy diffusivity of heat
Surface (M = eddy diffusivity of momentum
i\ = dimensionless coordinate
K. Noto 1 and R. Patsumoto 1 8 = dimensionless temperature
H = coefficient of viscosity
v = kinematic viscosity
p = density
Nomenclature
T = viscous stress
C\, C% C3 = function of Prandtl number 4> = stream function
/ = dimensionless stream function
g = acceleration of gravity Subscript
Gr = Grashof number (x3g(3AT/i>2)
x = at x
Nu = Nusselt number (ax/X)
m = average from x = 0 to x = x
Pr = Prandtl number
w = at wall
Pr ( = turbulent Prandtl number («M/W)
"> = at infinity
q = heat flux
Ra = Raleigh number Introduction
T = temperature Hitherto theoretical investigations in turbulent natural convec-
AT = Tw-T„ tion heat transfer along an isothermal vertical surface have been
u — velocity component in x -direction carried out by means of the integral method of boundary layer, the
un = representative velocity of natural convection method of exchange law between r and q, the successive integral
v = velocity component in y -direction method toward x -direction, the method which considers the trans-
x = coordinate along wall from leading edge port equation of Reynolds stress, and so on. Several experimental
y = coordinate perpendicular to %vall data have been obtained. But only one paper has reported about
y+ = y^rwx/p/v velocity profiles.
yfam = dimensionless thickness of viscous sublayer In this paper, the local similarity concept, which has not been
applied up to now, is adopted to turbulent heat transfer by natural
convection along an isothermal vertical flat surface.
1
Kobe University, Rokko, Nada, Kobe, Japan. P r o b l e m F o r m u l a t i o n a n d M e t h o d of S o l u t i o n
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat The basic equations used in the field of natural convection are
Transfer Division, May 19,1975. as follows:

Journal of Heat Transfer NOVEMBER 1975 / 621

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 plate resistance (r„ ln(n,/ra)/fefc) to obtain correction factors tances, and the increase in flux area with radius which diminishes
CUHP.I and CUHF,2 which are similar to CUHP for the flat plate the resistance. At low values of h0t/k, the flux lines are spread out
given in [1]. As an example, CUHF.I for the axial channel arrange- more evenly along the curved plate, and in turn suffer a greater
ment, is shown in Fig. 3. However, less importance is given to length of resistance. Therefore, values of Gi can be greater than
CUHF.I and CUHF,2 in this article, since it is the behavior of G i and unity, however over most of the range of parameters, Gi is less
G2 of equations (3) and (4) that are of interest. The geometry fac- than unity.
tors Gi and G2 are obtained by dividing equations (6) and (7) by
equation (5). Conclusion
In both cases, Gx and G2 vary by less than approximately ±10
Results percent about unity for 1.0 < rb/ra < 1.2; therefore, the designer of
Typical results for Gi and G 2 are shown in Figs. 4 and 5 for the deep ocean cylindrical channel-plate heat exchangers could initial-
case of Ti,lra - 1.2 (or ra/t = 5.0). This case was chosen as being a ly base his design on a flat plate model.
reasonable upper limit for rtlra for deep ocean applications, that For other applications of heat exchanger design for rtlra greater
is, the submarine hull thickness is 20 percent of the inside radius. than 1.2, equations (6) and (7) can be employed.
The other paraments used in the study fall within the range of Should the designer care to estimate the effect of the CHF
values for deep ocean design. boundary condition on the constriction resistance, Figs. 4, 5, and 6
In the case of both axial and circumferential channels, the for Cc of reference [1] could be applied, so that
values of G\ and G 2 approach the asymptote (0.913 in the case for
ri,/ra = 1.2) as a/6 goes to unity. The asymptote is simply the ratio flci(CHF) ss (t/6/j)C U H pGiC c
of the uniform cylindrical resistance to that of the flat plate of RC2(CHF) =* (t/bk)CuHFG2Cc
equal thickness, and is less than unity since the surface area for
heat flux increases with radius for the cylindrical configuration. Acknowledgments
As a/b goes to zero, Gi and G 2 approach unity, since the con- The numerical results presented in this article were obtained
striction resistance due to a point heat source becomes indepen- using the IBM 1130 at the Instituto Militar de Engenharia.
dent of the geometry. The values for G2 in Fig. 5 are always less
than unity since, although the length for the heat flux lines are the References
same as those for the flat plate, the area for heat flux increases 1 Oiiveira, H. Q., and R. P. Forslund, JOURNAL OP HEAT TRANS-
with radius for the cylindrical geometry, hence lower resistance. In FER, TRANS. ASME, Series C, Vol. 96, 1974, p. 292.
2 Oiiveira, H. Q., "Resistencia Termica em Trocadores do Tipo Tubo-
the case of axial channels, a trade-off exists between the increase Placa com Estrangulamento do Fluxo de Calor," MS thesis, Pontificia Univ-
in the length for the flux lines which contribute to larger resis- ersidade Catolica, Rio de Janeiro, 1973.

Turbulent Heat Transfer by a = heat transfer coefficient

Natural Convection Along an /3 = coefficient of cubical expansion
Mam = thickness of viscous sublayer
Isothermal Vertical Flat tH = eddy diffusivity of heat
Surface (M = eddy diffusivity of momentum
i\ = dimensionless coordinate
K. Noto 1 and R. Patsumoto 1 8 = dimensionless temperature
H = coefficient of viscosity
v = kinematic viscosity
p = density
Nomenclature
T = viscous stress
C\, C% C3 = function of Prandtl number 4> = stream function
/ = dimensionless stream function
g = acceleration of gravity Subscript
Gr = Grashof number (x3g(3AT/i>2)
x = at x
Nu = Nusselt number (ax/X)
m = average from x = 0 to x = x
Pr = Prandtl number
w = at wall
Pr ( = turbulent Prandtl number («M/W)
"> = at infinity
q = heat flux
Ra = Raleigh number Introduction
T = temperature Hitherto theoretical investigations in turbulent natural convec-
AT = Tw-T„ tion heat transfer along an isothermal vertical surface have been
u — velocity component in x -direction carried out by means of the integral method of boundary layer, the
un = representative velocity of natural convection method of exchange law between r and q, the successive integral
v = velocity component in y -direction method toward x -direction, the method which considers the trans-
x = coordinate along wall from leading edge port equation of Reynolds stress, and so on. Several experimental
y = coordinate perpendicular to %vall data have been obtained. But only one paper has reported about
y+ = y^rwx/p/v velocity profiles.
yfam = dimensionless thickness of viscous sublayer In this paper, the local similarity concept, which has not been
applied up to now, is adopted to turbulent heat transfer by natural
convection along an isothermal vertical flat surface.
1
Kobe University, Rokko, Nada, Kobe, Japan. P r o b l e m F o r m u l a t i o n a n d M e t h o d of S o l u t i o n
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat The basic equations used in the field of natural convection are
Transfer Division, May 19,1975. as follows:

Journal of Heat Transfer Copyright © 1975 by ASME NOVEMBER 1975 / 621

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1.01
(1)
dx Sy
9
8M 9W 9 r, \ «1 , aim m
OS thisarnksistPpO^.GotfCf0)
(2)
9A: Oo*ui2=n(Pr=0.72,GeFlC!0)

97/ 9T 3 , ^ , etf . 3 T , Eckertetal(PP=0.72Grx=lCf)

r
(3) ovOon CheesaMighfeGrx^J-a-IC?0)

Introducing the following stream function

dib dip
(4)

the representative velocity of natural convection

(5)

and dimensionless variables

Fig. 2 Temperature profiles

• (6)
77 = 0 ; / =
87)
(12)
Tw-T.
n= '977 = o,
equations (2) and (3) can be transformed as follows:
Pr< and IMIV in equations (13) and (14) given in the following
( l + _)_ r + {_f + _(_)}___(_) have been selected [l], 2 as these values are not yet clear for natural
convection,
B
+ 6-.x l°f/HJ) (7) Pr,= 1 (13)
€„/v = 0 . 4 / { 1 - exp (-0.0017y* 2 )} (14)
{J_ +J_£if}^+ / L + J_I(ii()i!6_J.8(y_
x f
Equations (10) and (11) can be easily integrated using equations
Pr Prf v 9?72 *47 P r t 8t) V j / ; / 9 7 ) ~ x
d(x,v) (12), (13), and (14). The use of local similarity concept given us the
(8)
following advantages.
The local similarity concept, viz. 1 f and 0 in equations (10) and (11) do not include the deriva-
9/ _ 99_ tives of x, but are apparently the functions of ri only. Consequent-
= 0 (9) ly, the value of x (or Gr*) at any point in x-direction can be an as-
dx ~~ dx
signable constant parameter. The value at any arbitrary point xa
can be applied as the variations of / and 0 in x -direction are very on the physical plane is denoted by Gr.ra at this point. Equations
small when compared with those in ^-direction and this can be ne- (10) and (11) can thus be solved in ^-direction only in respect of
glected. Hence equations (7) and (8) can be transformed as follows: parameter Gr l a , viz., the successive integration toward the trailing
edge is not required.
1 v /J 2 The a priori decision of a transition point is not necessary in
v drf 4 9rj ^ 97^ 2 9T/
this analysis. On the contrary, in the ordinary successive integral
do) method, it is necessary to obtain the exact location of the transi-
92e tion point before integration. Further, the location of transition
Pr( }^f +' {i^
v ' drf 4/ + r ^ ry ^ ^ ^ - 0
' P^r s 8TJ
l
(
V point affects the quantity of heat transfer.

(n) Results and Discussion

with the boundary conditions are as follows: The results obtained in this analysis are shown by solid lines in
Figs. 1-5.
Fig. 1 shows the dimensionless velocity profile. The experimen-
tal results by Cheesewright [2] are plotted in the same figure for
This analysis (Pr=Q72,Grx*ld°) comparison. Further, the theoretical values by Oosthuizen [3],
Eckert [4], and Mason, et al. [5] are also shown by a dotted line, a
Mason & Seban chain line and two dotted chain lines, respectively.
The velocity profile obtained does not strictly agree with the ex-
OosthuizentP^O.ya.Grx^O10)
perimental results of Cheesewright.
6*ertet.at.(Pr=0.72,Grx=l610) The results of Mason, et al., are similar to this analysis, but the
results of the others deviate further from the experimental values.
onV Cheesewright(air,Grx=<3~9>lrf0) The authors wish to point out the following considerations in re-
spect of these differences.
1 The geometrical conditions and the effect of enclosed space
in the apparatus can be considered to play an important role in af-
fecting the experimental results. The authors decided to compare
with Cheesewright's data in the absence of other experimental re-
30 40 ^ 50 60

2
Flg. 1 Velocity profiles Numbers in brackets designate References at end of technical note.

622 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 10s
1 i 1 8 —
thisaralysis(Pr=!t.O) 6
tatoetal(PKO) 4
Eckert et.ol.(Pr=lD) Grx=ld° 2? •Z-
— Oosthui7en(Pr=0.7) 2 1
j
a) ^s
10' o | Jakob 103
1
8 i
( Saunders(ait-) 6
o )
o Wjmer(air) 4

2 ~~"jf'T
103 °J* ;
J ^ in 2 /
8

o j « 7 ~4 zt
Jl
10 2
c
0^
#'*
0
A This analysis

f/
0 >
rt^v
YF7''S
/ in' /
8
6
<y
/
/
Kato etal.

y
fxkert etal.
4
/
id. 2
itfib»io9 D io» ib11 W io°
in 0
10"3 TO"2 10^' 10° 10'p^ 102 103 104 105
Fig. 3 Nusselt number
Fig. 5 Nusseli number

be necessary, if this method is to predict accurate results in the

suits. It is felt strongly that further developments experimentally
turbulent field.
would be very useful in the future for more accurate comparisons.
Fig. 2 shows the dimensionless temperature profile. The above
2 The equation (14) used in this analysis is found to be accu-
mentioned, other theoretical and experimental results are also
rate in the field of forced convection. T h e authors have applied
shown in the same figure. The analytical values in this analysis
this formula in the analysis and this fact may be one of the causes
show good agreement with the experimental results.
for the resulting deviations.
Fig. 3 shows the relation between mean Nusselt number and
3 Lastly, the authors wish to comment on the accuracy of this
local Rayleigh number for Prandtl numbers 0.7-1. The experimen-
analysis. T h e local nonsimilarity method [6] can be used to evalu-
tal results plotted in this figure are by Jakob [7], Saunders [8],
ate any level of truncation, but this method is thought to be a kind
Warner [9], Toulokian, et al. [10], and Fujii [11]. Moreover, the
of perturbation method. It is felt that further considerations may
theoretical values by Eckert, Oosthuizen, and Kato, et al., are also
shown. The analytical values are found to be in good agreement
with the experimental results.
Fig. 4 shows the relation between mean Nusselt number and
1 I 1 1 1
local Grashof number for Prandtl number about 40. In this figure,
THsanalysis(Pr=40) this analysis is found to be in closer agreement with the experi-
, mental data of Toulokian, et al., and Fujii than that of the theoret-
Katoetal.(Pr-40) ical analysis of Kato, et al.
Fig. 5 shows the relation between local Nusselt number and
Prandtl number [10~3 < Pr < 104] for Gr = 10 10 .
% o Toulokianetal.(Pr240) Thickness of viscous sublayer 5iam, dimensionless thickness of
6 this y + m and local shear stress at wall TWX in turbulent natural
• Fujii (Pr=30~60) convection were able to be expressed by following formulas in the
4
range of Gr* = 10 6 -10 15 , P r = lQ- 3 -10 3 .
'
2
6lam/^=C1(Pr).Gr/»-23
103 , #/
8
£6 P /,am=C2(Pr)-G«0-12 (15)
ff Tmxl/nv = C$Pr)-Gr*Jl'0
mjS r
£y In conclusion, authors feel that this analysis is a simple way to
obtain the velocity and temperature profiles and offers an engi-
© jS,* neering applicability in the field of turbulent natural convection.
^ / <
References
/
1 Kato, H., Nishiwaki, N., and Hirata, M., "On the Heat Transfer by
Free Convection From a Vertical Plate," International Journal of Heat and
Mass Transfer, Vol. 11,1968, pp. 1117-1125.
2 Cheesewright, R., "Turbulent Natural Convection From a Vertical
~W- io5 io* G r x 10s W W Plane Surface," JOURNAL OF HEAT TRANSFER, TRANS. ASME, Se-
ries C, Vol. 90,1968, pp. 1-8.
Fig. 4 Nusselt number 3 Oosthuizen, P. H., "A Numerical Study of Turbulent Free Convec-

Journal of Heat Transfer NOVEMBER 1975 / 623

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 tion From an Isothermal Vertical Plate," Rept. No. 1170 of the Thermal and ever, but they must be determined from a mass balance on the dis-
Fluid Sciences Group, Department of Mechanical Engineering, Queen's solved species at the melt line, y = 0. This balance is, in general
University, Kingston, Ontario, Canada, 1970.
4 Eckert, E. R. G., and Jackson, T. W., "Analysis of Turbulent Free [61,
Convection Boundary Layer on a Flat Plate," NACA Tech, Note 2207,1950. aC\
5 Mason, H. B.,"and Seban, R. A., "Numerical Predictions for Turbu- fhc = ~py — I + Cp(mc + m) (1)
lent Free Convection From Vertical Surfaces," International Journal of ay/j=o
Heat and Mass Transfer, Vol. 17,1974, pp. 1329-1336.
6 Minkowycz, W. J., and Sparrow, E. M., "Local Nonsimilar Solutions where A c (kg/m 2 s) is the flux of dissolved species at the interface
for Natural Convection on a Vertical Cylinder," JOURNAL OF HEAT and ?R(kg/m2 s) is the melting rate of pure solid. The density of the
TRANSFER, TRANS. ASME, Series C, Vol. 96, No. 2, 1974, pp. 178-183. mixture is denoted by p and y is the mass diffusion coefficient.
7 Jakob, M., Heat Transfer, Vol. 1, Wiley, New York, 1949, p. 526. Since the solid phase is impervious to the dissolved species (rhc =
8 Saunders, O. A., "Effect of Pressure Upon Natural Convection of 0) during the melting of glacial ice, this equation reduces to
Air," Proc. Roy. Soc, Series A, Vol. 157,1936, pp. 278-291.
9 Warner, C. Y., and Arpaci, V. S., "An Experimental Investigation of
Turbulent Natural Convection in Air Along a Vertical Heated Flat Plate," -CFITI = —p7 - (2)
International Journal of Heat and Mass Transfer, Vol. 11, 1968, pp. 397- ay/ y=o
406. At the ice-salt water interface the temperature is determined by
10 Toulokian, Y. S., Hawkins, G. A., and Jakob, M., "Heat Transfer by
Free Convection From Heated Vertical Surfaces to Liquids," JOURNAL OF the equilibrium condition
HEAT TRANSFER, TRANS. ASME, Series C, Vol. 70, No. 1, 1948, pp. fF = - e C F (3)
13-23. where c = 0.053 when C> is given in g/kg (see Martin and Kauff-
11 Fujii, T., "Experimental Studies of Free Convection Heat Transfer,"
Bull. Japan Soc. of Mech. Eng., Vol. 2,1959, pp. 555-558. man [3] or Pounder [6]), and the governing transport equations [1,
2] are coupled together by these interrelated balance conditions.

Results and Discussion

When the equilibrium condition (3) is introduced into the sys-
tem, the heat, momentum, and mass transfer solutions, already in-
A Note Concerning the terdependent because of the appearance of the melting rate in
Transport Processes Near each of the governing transport equations, become more closely
coupled. The energy-balance integral method used previously [1,
Melting Glacial Ice in Seawater 2] remains a convenient method of solution, and some sample re-
sults have been recomputed for water temperatures between 5 and
O. M. Griffin 1 20°C and for ice temperatures from —4 to —15°C. Two values of
free-stream salinity have been investigated, at C„ = 26 and 34
g/kg, since these are representative, respectively, of arctic and ant-
Introduction arctic sea water [3].
The heat, mass, and momentum transfer processes near melting The melting temperature TF is plotted in Fig. 1 as a function of
glacial ice in seawater were recently studied [1, 2] 2 for the case of the water temperature, for a free-stream salinity of 34 g/kg. The
forced fluid flow. It was assumed that the temperature at the melting point is depressed under these conditions from the pure
boundary between phases was equal to the melting temperature of water value of 0 to —0.35°C when the water temperature is 5°C
a pure ice-water system, or 0°C. However, such a system is correct- and the ice is at —4.8°C. When the ice temperature and salinity re-
ly described by an equilibrium condition at the interface whereby main constant but the water temperature is increased to 15°C, the
the melting temperature is depressed in direct relation to the con- depression of the melting point is reduced to —0.07°C. The second
centration of dissolved species at the boundary. This condition has curve in the figure corresponds to an ice temperature of —14.4°C,
been employed recently by Martin and Kauffman [3] in a study of and it is seen that the ice subcooling has little effect on the equilib-
the evolution of Arctic under-ice fresh water melt ponds and ear- rium melting temperature at the interface.
lier by Frank [4] in an analysis of combined heat and mass transfer There is a relatively simple, but interesting, physical reason for
during transient solidification. The transport equations for the this change in melting point depression with fluid temperature,
melt are complicated by this condition in that the heat and mass and the interfacial salinity CF and the ratio oc/bt of the boundary
transfer processes are more closely coupled and because the melt-
ing temperature is not known a priori as in the melting of a pure
solid. Also, the Stefan numbers—or the ratios of the sensible and
latent heats for the melt and solid—are no longer independent pa-
rameters since both depend on the resulting temperature at the in-
terface.

Analysis
The purpose of this technical brief is to extend the analysis just
mentioned [1, 2] to include an appropriate equilibrium condition „ -0.3
for the melting point depression. The melt problem is again formu-
lated by considering the boundary layer equations for combined
heat, mass, and momentum transfer which are described in the
previously cited references. The concentration CV3 and the tem-
perature TF at the phase boundary are not known a priori, how-

1
Ocean Technology Division, Naval Research Laboratory, Washington,
D.C. Mem. ASME.
2
Numbers in brackets designate References at end of technical note.
3
The nomenclature here is the same as [1].
Contributed by the Heat Transfer Division of THE AMERICAN SOCI- Fig. 1 The depression of the melting temperature TF as a function of sea
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat water and ice temperatures T„ and 7"0, respectively—the salinity C„ of
Transfer Division May 19,1975. sea water outside the boundary layer was held constant at 34 g/kg

624 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 tion From an Isothermal Vertical Plate," Rept. No. 1170 of the Thermal and ever, but they must be determined from a mass balance on the dis-
Fluid Sciences Group, Department of Mechanical Engineering, Queen's solved species at the melt line, y = 0. This balance is, in general
University, Kingston, Ontario, Canada, 1970.
4 Eckert, E. R. G., and Jackson, T. W., "Analysis of Turbulent Free [61,
Convection Boundary Layer on a Flat Plate," NACA Tech, Note 2207,1950. aC\
5 Mason, H. B.,"and Seban, R. A., "Numerical Predictions for Turbu- fhc = ~py — I + Cp(mc + m) (1)
lent Free Convection From Vertical Surfaces," International Journal of ay/j=o
Heat and Mass Transfer, Vol. 17,1974, pp. 1329-1336.
6 Minkowycz, W. J., and Sparrow, E. M., "Local Nonsimilar Solutions where A c (kg/m 2 s) is the flux of dissolved species at the interface
for Natural Convection on a Vertical Cylinder," JOURNAL OF HEAT and ?R(kg/m2 s) is the melting rate of pure solid. The density of the
TRANSFER, TRANS. ASME, Series C, Vol. 96, No. 2, 1974, pp. 178-183. mixture is denoted by p and y is the mass diffusion coefficient.
7 Jakob, M., Heat Transfer, Vol. 1, Wiley, New York, 1949, p. 526. Since the solid phase is impervious to the dissolved species (rhc =
8 Saunders, O. A., "Effect of Pressure Upon Natural Convection of 0) during the melting of glacial ice, this equation reduces to
Air," Proc. Roy. Soc, Series A, Vol. 157,1936, pp. 278-291.
9 Warner, C. Y., and Arpaci, V. S., "An Experimental Investigation of
Turbulent Natural Convection in Air Along a Vertical Heated Flat Plate," -CFITI = —p7 - (2)
International Journal of Heat and Mass Transfer, Vol. 11, 1968, pp. 397- ay/ y=o
406. At the ice-salt water interface the temperature is determined by
10 Toulokian, Y. S., Hawkins, G. A., and Jakob, M., "Heat Transfer by
Free Convection From Heated Vertical Surfaces to Liquids," JOURNAL OF the equilibrium condition
HEAT TRANSFER, TRANS. ASME, Series C, Vol. 70, No. 1, 1948, pp. fF = - e C F (3)
13-23. where c = 0.053 when C> is given in g/kg (see Martin and Kauff-
11 Fujii, T., "Experimental Studies of Free Convection Heat Transfer,"
Bull. Japan Soc. of Mech. Eng., Vol. 2,1959, pp. 555-558. man [3] or Pounder [6]), and the governing transport equations [1,
2] are coupled together by these interrelated balance conditions.

Results and Discussion

When the equilibrium condition (3) is introduced into the sys-
tem, the heat, momentum, and mass transfer solutions, already in-
A Note Concerning the terdependent because of the appearance of the melting rate in
Transport Processes Near each of the governing transport equations, become more closely
coupled. The energy-balance integral method used previously [1,
Melting Glacial Ice in Seawater 2] remains a convenient method of solution, and some sample re-
sults have been recomputed for water temperatures between 5 and
O. M. Griffin 1 20°C and for ice temperatures from —4 to —15°C. Two values of
free-stream salinity have been investigated, at C„ = 26 and 34
g/kg, since these are representative, respectively, of arctic and ant-
Introduction arctic sea water [3].
The heat, mass, and momentum transfer processes near melting The melting temperature TF is plotted in Fig. 1 as a function of
glacial ice in seawater were recently studied [1, 2] 2 for the case of the water temperature, for a free-stream salinity of 34 g/kg. The
forced fluid flow. It was assumed that the temperature at the melting point is depressed under these conditions from the pure
boundary between phases was equal to the melting temperature of water value of 0 to —0.35°C when the water temperature is 5°C
a pure ice-water system, or 0°C. However, such a system is correct- and the ice is at —4.8°C. When the ice temperature and salinity re-
ly described by an equilibrium condition at the interface whereby main constant but the water temperature is increased to 15°C, the
the melting temperature is depressed in direct relation to the con- depression of the melting point is reduced to —0.07°C. The second
centration of dissolved species at the boundary. This condition has curve in the figure corresponds to an ice temperature of —14.4°C,
been employed recently by Martin and Kauffman [3] in a study of and it is seen that the ice subcooling has little effect on the equilib-
the evolution of Arctic under-ice fresh water melt ponds and ear- rium melting temperature at the interface.
lier by Frank [4] in an analysis of combined heat and mass transfer There is a relatively simple, but interesting, physical reason for
during transient solidification. The transport equations for the this change in melting point depression with fluid temperature,
melt are complicated by this condition in that the heat and mass and the interfacial salinity CF and the ratio oc/bt of the boundary
transfer processes are more closely coupled and because the melt-
ing temperature is not known a priori as in the melting of a pure
solid. Also, the Stefan numbers—or the ratios of the sensible and
latent heats for the melt and solid—are no longer independent pa-
rameters since both depend on the resulting temperature at the in-
terface.

Analysis
The purpose of this technical brief is to extend the analysis just
mentioned [1, 2] to include an appropriate equilibrium condition „ -0.3
for the melting point depression. The melt problem is again formu-
lated by considering the boundary layer equations for combined
heat, mass, and momentum transfer which are described in the
previously cited references. The concentration CV3 and the tem-
perature TF at the phase boundary are not known a priori, how-

1
Ocean Technology Division, Naval Research Laboratory, Washington,
D.C. Mem. ASME.
2
Numbers in brackets designate References at end of technical note.
3
The nomenclature here is the same as [1].
Contributed by the Heat Transfer Division of THE AMERICAN SOCI- Fig. 1 The depression of the melting temperature TF as a function of sea
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat water and ice temperatures T„ and 7"0, respectively—the salinity C„ of
Transfer Division May 19,1975. sea water outside the boundary layer was held constant at 34 g/kg

624 / NOVEMBER 1975 Copyright © 1975 by ASME Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.3 melting rate from the pure water value. When the water tempera-
ture is increased to 15°C, the melting point depression is only
—0.07°C. This increases the Stefan number by less than 1 percent
-0.35 and results in a correspondingly small increase in the melting rate
from the pure water value.
Some typical results for the heat and mass transfer are listed in
< Table 1. The free-stream salinity in this case is C„ = 26 g/kg. Both
CM the energy and concentration thicknesses 6t** and <5C** increase as
0.30 < the fluid temperature is increased. The thermal boundary layer
thickness St increases slightly in relation to <5m, the momentum
boundary layer thickness, with increasing fluid temperature, while
the relative mass diffusion boundary layer thickness <5c/<5m doubles
as the fluid temperature increases from 5 to 20°C. The kinematic
viscosity of the water decreases sharply with rising temperature
0.25
and results in the increased relative boundary layer thicknesses.
An increase in the water temperature results in an increased ther-
mal driving force (T„ — Tp) and, consequently, results in the in-
creased melting rate Vp shown in the right-hand column of the
Fig. 2 Interfacial salinity C F and the railo bc/b, of ihe salinity and temper- table. The subcooling of the solid phase has little effect in the
ature boundary layer thicknesses as a function of water temperature dur- range of ice temperatures employed in the calculations.
ing the melting of glacial ice in sea water—the ice temperature was held
constant at T0 = —4.8°C and the sea water salinity was equal to G» = 34
g/kg. Summary, and Concluding Remarks
The energy-balance integral method has been employed in a
study of the steady melting of ice in sea water. It is necessary to
layer thicknesses are plotted in Fig. 2 as functions of fluid temper- satisfy simultaneously an energy balance, a species conservation
ature. The concentration Cp is determined from the balance, equa- equation and an equilibrium temperature condition at the inter-
tion (2), between the rate of diffusion and the bulk motion of the face between phases in addition to the transport equations for
fluid at the interface, and it is inversely proportional to the bound- heat, mass and momentum transfer. For a sea water temperature
ary layer thickness <5C and to the temperature difference (T„ — of 5°C, the melting point depression which results from the com-
Tp). Since bc increases with fluid temperature as the rate of bined heat and mass transfer processes increases the melting rate
species diffusion approaches the rate of diffusion for the solute by 9 percent from the pure water value. When the water tempera-
(water), the concentration Cp of dissolved species at the interface ture is 15°C, the melting rate is increased by less than 1 percent
decreases accordingly. For example, when the glacial ice at —4.8°C and the heat and mass transfer are independent of the melting
melts in sea water of 34 g/kg salinity, the interfacial salinity de- point depression.
creases from 6.5 g/kg to 1.4 g/kg as the water temperature in- Particular attention has been paid to the melting of glacial ice in
creases from 5 to 15° C. This decrease in salinity results in a change sea water, but the method is also applicable in general to the
in the melting point depression from Tp = —0.34 to — 0.07°C. steady melting of a pure solid in the presence of a small concentra-
The Stefan number for the water is no longer an independent tion of dissolved species in the liquid.
parameter when the melting point depression depends on the in-
terfacial salinity. As an example, for a water temperature of 5°C Acknowledgment
and a salinity of 34 g/kg, a melting point depression of — 0.35°C re- The author wishes to thank the Naval Research Laboratory for
sults in a seven percent increase in the thermal driving force (Ste- support of this work, and also to thank Prof. Seelye Martin of the
fan number) and a corresponding nine percent increase in the University of Washington, Seattle for several helpful comments.

Table 1 Heat and mass transfer results

( f ^ - 5'C, C_ = 26 g/kg)

'"J
t'6m " " V 6 n ' A2
- 4.8»C 0.20°C 0.0572 0.0174 0.436 0.108 0.0148

- 9.6 0.21 0.0572 0.0172 0.436 0.107 0.0144

-14.4 0.22 0.0572 0.0170 0.436 0.106 0.0141

4.8 0.13 0.0609 0.0240 0.463 0.140 0.0281

9.6 0.14 0.0608 0.0237 0.463 0.138 0.0274

0.15 0.0608 0.0234 0.463 0.137 0.0267

- 4.8 0.07 0.0648 0.0301 0.492 0.171 0.0417

- 9.6 0.08 0.0647 0.0298 0.492 0.169 0.0406

-14.4 0.08 0.0646 0.0294 0.492 0.168 0.0396

0.05 0.0682 0.0356 0.518 0.200 0.0549

- 9.6 0.06 0.0681 0.0351 0.518 0.198 0.0536

-14.4 0.06 0.0680 0.0347 0.518 0.196 0.0523

Journal of Heat Transfer NOVEMBER 1975 / 625

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 References the work of Alkidas and Durbetaki [1, 2]. 3 The effects of injecting a
1 Griffin, O. M., "Heat, Mass and Momentum Transfer During the hot inert gas through the porous wall surface of an axisymmetric
Melting of Glacial Ice in Seawater," JOURNAL OF HEAT TRANSFER, blunt body near the stagnation region, have been studied with re-
TRANS. ASME, Series C, Vol. 95,1973, pp. 317-323. gards to heat interactions between the gases and the wall, and the
2 Griffin, 0. M., "An Integral Method of Solution for Combined Heat ignition of the cold combustible mixture. The fluid has been con-
and Mass Transfer Problems With Phase Transformation," HEAT
TRANSFER 1974, Proceedings Fifth International Heat Transfer Confer- sidered compressible and the injection rate of the hot inert gas has
ence, Vol. 1,1974, pp. 211-215. been varied.
3. Martin, S. and Kauffman, P., "The Evolution of Under-Ice Melt The boundary-layer conservation equations of a multicompo-
Ponds, or Double Diffusion at the Freezing Point," Journal of Fluid Me-
chanics Vol. 64,1974, pp. 507-527. nent reactive mixture interacting with a constant temperature
4 Frank, F. C, "Radially Symmetric Phase Growth Controlled by Diffu- solid boundary have been derived by Alkidas and Durbetaki [1, 2].
sion," Proceedings Royal Society London, Series A, Vol. 201, 1950, pp. These equations remain unaltered for the problem under consider-
586-599. ation in this paper, and they are
5 Bird, R. B., Stewart, W. E., and Lighfoot, E. N., Transport Phenome-
na, Wiley, New York, Chapter 19,1960.
6 Pounder, E. R., The Physics of Ice, Pergamon, Oxford, Chapters 1 and r + / / " = a/2)[(/')2-0] (1)
2,1965. * / ' + ScfaF' = (ne + E
aF)ScDi(l/6)aFe- * (2)
X
N •+ Sc/< = 0 (3)
E /,,
0" + Pr/fl' = - ( n „ + ai,)PrDID^(l/0)a>e-• * (4)

The boundary conditions for the governing equations presented

in the foregoing describing the behavior of the problem under con-
Stagnation Point H e a t Transfer: sideration are
Ignition of a Combustible at n = 0:
f = iw /' = o e = ew (5a)
M i x t u r e With I n e r t Gas ti' = -Scfwcti fc,Ti = 0,F,P
Injection T h r o u g h a P o r o u s at t] - 0: 07/ = Scfw(ai — 1/Y^^) (56)
Wall f —1 0—1
as J; - (5c)
1 2 a/ — 1 for i = O, F, P, N
P. Durbetaki and H. W. Hsu
The above set of differential equations are coupled and they
must be solved simultaneously together with the boundary condi-
tions presented. The numerical computations were carried out
Nomenclature using Gill's modified Runge-Kutta fourth order method. The
Di = Damkohler's first dimensionless similarity group equations were first integrated with a set of estimated values for
Dn = Damkohler's second dimensionless similarity group the unknown boundary conditions at the wall, f"(0), 6'(0), and
Di,,- = critical first Damkohler number for ignition ctF (0). The calculated values of f, ap, and 0 at the free stream were
E = activation energy then compared with unity, the asymptotic condition. An improved
E* = E/RTe, dimensionless activation energy set of values of the initial conditions were found by using the
fw = injection parameter Gauss-Jordan Elimination Method. These iterations were contin-
f'(i]) = dimensionless velocity ued until the resulting values of /', CXF, and 8 at the free stream
ne = dimensionless stoichiometry measure were within 0.0001 of unity. The step-size during the integration
Nu = Nusselt number was varied using 0.01 increments for r/ < 3.0, the region where the
Pr = Prandtl number most drastic changes occur, and the step size was increased to 0.1
R = universal gas constant for r\ > 3.0, the region where there are no'significant changes. The
Re = Reynolds number integration was carried out until a value of r) = 12.5 where the pro-
Sc = Schmidt number files of /', 6, and ap became constant. The procedure was repeated
T = temperature for each of the Di values and with /,„ as the injection parameter.
Y = mass fraction
di = YJYi.e Results and Discussion
7] = dimensionless coordinate In order to establish some numerical values, with the computa-
8 = T/Te tional approach discussed in the foregoing, the case selected was a
stoichiometric mixture of methane and air at a free stream temper-
Subscripts
ature of Te = 300 K. The activation energy is found to be 1.675 X
e = free stream 105 J/mole. These conditions were used to produce the values of
F = fuel the dimensionless parameters and the boundary conditions needed
i = tth species for the computations from the governing equations. Together with
N = inert gas the values of the inert gas injection parameter used they are:
O = oxidizer
P = product Pr = 0.74 Sc = 0.74 E* = 67.1366
Several studies have been made during the last two decades on
On = 6.44518 ne = 0 0W = 3.0
the heat interaction between reactive mixtures of gases and solid
surfaces. The work reported in the present paper is an extension of /„ = 0, -0.15, -0.30, -0.50
The calculated results of Nu/(Re) 1 / 2 , for the range of.values for
1
the first Damkohler number that encompasses the transition from
Assoc. Professor, School of Mechanical Engineering, Georgia Institute of frozen flow through ignition are given in Fig. 1. It is apparent from
Technology, Atlanta, Ga. Mem. ASME.
2 this figure that the effect of inert gas injection through the porous
Graduate Assistant, School of Mechanical Engineering, Georgia Insti-
tute of Technology, Atlanta, Ga.
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat
3
Transfer Division August 19,1974. Numbers in brackets designate References at end of technical note.

626 / NOVEMBER 1975 Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 References the work of Alkidas and Durbetaki [1, 2]. 3 The effects of injecting a
1 Griffin, O. M., "Heat, Mass and Momentum Transfer During the hot inert gas through the porous wall surface of an axisymmetric
Melting of Glacial Ice in Seawater," JOURNAL OF HEAT TRANSFER, blunt body near the stagnation region, have been studied with re-
TRANS. ASME, Series C, Vol. 95,1973, pp. 317-323. gards to heat interactions between the gases and the wall, and the
2 Griffin, 0. M., "An Integral Method of Solution for Combined Heat ignition of the cold combustible mixture. The fluid has been con-
and Mass Transfer Problems With Phase Transformation," HEAT
TRANSFER 1974, Proceedings Fifth International Heat Transfer Confer- sidered compressible and the injection rate of the hot inert gas has
ence, Vol. 1,1974, pp. 211-215. been varied.
3. Martin, S. and Kauffman, P., "The Evolution of Under-Ice Melt The boundary-layer conservation equations of a multicompo-
Ponds, or Double Diffusion at the Freezing Point," Journal of Fluid Me-
chanics Vol. 64,1974, pp. 507-527. nent reactive mixture interacting with a constant temperature
4 Frank, F. C, "Radially Symmetric Phase Growth Controlled by Diffu- solid boundary have been derived by Alkidas and Durbetaki [1, 2].
sion," Proceedings Royal Society London, Series A, Vol. 201, 1950, pp. These equations remain unaltered for the problem under consider-
586-599. ation in this paper, and they are
5 Bird, R. B., Stewart, W. E., and Lighfoot, E. N., Transport Phenome-
na, Wiley, New York, Chapter 19,1960.
6 Pounder, E. R., The Physics of Ice, Pergamon, Oxford, Chapters 1 and r + / / " = a/2)[(/')2-0] (1)
2,1965. * / ' + ScfaF' = (ne + E
aF)ScDi(l/6)aFe- * (2)
X
N •+ Sc/< = 0 (3)
E /,,
0" + Pr/fl' = - ( n „ + ai,)PrDID^(l/0)a>e-• * (4)

The boundary conditions for the governing equations presented

in the foregoing describing the behavior of the problem under con-
Stagnation Point H e a t Transfer: sideration are
Ignition of a Combustible at n = 0:
f = iw /' = o e = ew (5a)
M i x t u r e With I n e r t Gas ti' = -Scfwcti fc,Ti = 0,F,P
Injection T h r o u g h a P o r o u s at t] - 0: 07/ = Scfw(ai — 1/Y^^) (56)
Wall f —1 0—1
as J; - (5c)
1 2 a/ — 1 for i = O, F, P, N
P. Durbetaki and H. W. Hsu
The above set of differential equations are coupled and they
must be solved simultaneously together with the boundary condi-
tions presented. The numerical computations were carried out
Nomenclature using Gill's modified Runge-Kutta fourth order method. The
Di = Damkohler's first dimensionless similarity group equations were first integrated with a set of estimated values for
Dn = Damkohler's second dimensionless similarity group the unknown boundary conditions at the wall, f"(0), 6'(0), and
Di,,- = critical first Damkohler number for ignition ctF (0). The calculated values of f, ap, and 0 at the free stream were
E = activation energy then compared with unity, the asymptotic condition. An improved
E* = E/RTe, dimensionless activation energy set of values of the initial conditions were found by using the
fw = injection parameter Gauss-Jordan Elimination Method. These iterations were contin-
f'(i]) = dimensionless velocity ued until the resulting values of /', CXF, and 8 at the free stream
ne = dimensionless stoichiometry measure were within 0.0001 of unity. The step-size during the integration
Nu = Nusselt number was varied using 0.01 increments for r/ < 3.0, the region where the
Pr = Prandtl number most drastic changes occur, and the step size was increased to 0.1
R = universal gas constant for r\ > 3.0, the region where there are no'significant changes. The
Re = Reynolds number integration was carried out until a value of r) = 12.5 where the pro-
Sc = Schmidt number files of /', 6, and ap became constant. The procedure was repeated
T = temperature for each of the Di values and with /,„ as the injection parameter.
Y = mass fraction
di = YJYi.e Results and Discussion
7] = dimensionless coordinate In order to establish some numerical values, with the computa-
8 = T/Te tional approach discussed in the foregoing, the case selected was a
stoichiometric mixture of methane and air at a free stream temper-
Subscripts
ature of Te = 300 K. The activation energy is found to be 1.675 X
e = free stream 105 J/mole. These conditions were used to produce the values of
F = fuel the dimensionless parameters and the boundary conditions needed
i = tth species for the computations from the governing equations. Together with
N = inert gas the values of the inert gas injection parameter used they are:
O = oxidizer
P = product Pr = 0.74 Sc = 0.74 E* = 67.1366
Several studies have been made during the last two decades on
On = 6.44518 ne = 0 0W = 3.0
the heat interaction between reactive mixtures of gases and solid
surfaces. The work reported in the present paper is an extension of /„ = 0, -0.15, -0.30, -0.50
The calculated results of Nu/(Re) 1 / 2 , for the range of.values for
1
the first Damkohler number that encompasses the transition from
Assoc. Professor, School of Mechanical Engineering, Georgia Institute of frozen flow through ignition are given in Fig. 1. It is apparent from
Technology, Atlanta, Ga. Mem. ASME.
2 this figure that the effect of inert gas injection through the porous
Graduate Assistant, School of Mechanical Engineering, Georgia Insti-
tute of Technology, Atlanta, Ga.
Contributed by the Heat Transfer Division of THE AMERICAN SOCI-
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat
3
Transfer Division August 19,1974. Numbers in brackets designate References at end of technical note.

626 / NOVEMBER 1975 Copyright © 1975 by ASME Transactions of the ASME

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 2.0 1.0
l__J-—— 0

1.6 0 .8 -.15 —
A
-.15 /
^ -
" """"l -.30
1.2
X
Q f. = - . 5 0 '/^"^ 1
if _ _ ^ — . — — — - lw- -.50

2 /V*"""^»»»'"""" B
\&
-

0 —
— \ 0 1 —"™'*™™^S-.30
-.50
™ 1_ . - .

10 10' 10'
Di
-.4 A
-.50 Fig. 3 The fuel concentration at the interface as a function of the first
- .30 Damkbhier number with inert gas injection rate as a parameter, 0W = 3.0
?'- - .15
0

10' 10" 10' 10°

Fig. 1 Dependence of surface heat transfer rate on the first Damkbhier

number with inert gas injection rate as a parameter, 8W = 3.0

'<• \ \t
wall is to increase the critical ignition first Damkohler number Bi,;. V \°* V
Another effect that can be observed from Fig. 1 is that at the Y \° V
ignition point the surface heat transfer rate decreases with increas- \ \ V
ing injection rate. The effect of the inert gas injection rate on in- ^\ \ \ \°
terface heat transfer is better illustrated in Fig. 2 where the curves
have been drawn for fixed values of the first Damkohler number. A
decrease of the heat added to the wall with increasing inert gas in-
jection rates, is due to the absorption of a larger fraction of the lib-
erated energy from the reacting mixture, by the inert gas.
The fuel concentration at the interface, as a function of the first
Damkohler number, at specified inert gas injection rates is shown ^J
in Fig. 3. The branches A and B on these curves have been labeled
to correspond with the previous figures. The decrease in the fuel
mass fraction at the wall is due to the dilution of the combustible Fig. 4 Comparison of temperature profiles at specified Damkbhier num-
bers and inert gas injection rates
mixture by the inert gas injected through the porous wall. Table 1

Table 1 Fuel concentrations at the interface at specified inert gas injec-

tion rates

a t D. a t Oj = 10 branch A

0 0.490 0.9999

-0.15 0.415 0.8093

-0.30 0.340 0.6479

-0.50 0.250 0.4733

presents a summary of the fuel concentration at the wall both

under frozen flow condition and the critical ignition point at four
specified inert gas injection rates.
Finally, Fig. 4 presents the temperature profiles at fixed Dam-
kohler numbers for different inert gas injection rates. The curves
are marked with the letter A or B to indicate the appropriate
branch they correspond to in Fig. 1. Inert gas injection, appears to
increase the maximum temperature near the critical ignition point
and decreases it near the equilibrium flow. Also, with injection the
flame is moved further away from the wall.

References
1 Alkidas, A., and Durbetaki, P., "Stagnation-Point Heat Transfer: The
Effect of the First Damkohler Similarity Parameter," JOURNAL OF HEAT
TRANSFER, TRANS. ASME, Series C, Vol. 94,1972, p. 410.
2 Alkidas, A., and Durbetaki, P., "Ignition of a Gaseous Mixture of a
Fig. 2 The effect of inert gas injection on the surface heat transfer Heated Surface," Combustion Science and Technology, Vol. 7,1973, p. 135.

Journal of Heat Transfer NOVEMBER 1975 / 627

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Laminar Film Condensation ^y
Heat Transfer in the Presence of
Electric and Magnetic Fields
Plate
>
1
A. S. El-Ariny, J. A. Sabbagh, and M. A. Obeid 1 2 E^
©
> B
Introduction —.—>
The mechanism of heat transfer with laminar film condensation
-*y \
on an isothermal plate was investigated by Sparrow and Gregg [l] 3
l
applying boundary layer treatment. Laminar film condensation in w—- —Boundary
the presence of electric and magnetic fields was analy2ed by Singer Layer
[2] using the integral method of boundary layer theory with the ^u
negligence of the inertia terms in the momentum equation. Very
recently, Murty, et al. [3] studied the influence of a weak magnetic
X
field on the condensation phenomena using perturbation tech-
nique. They found that the ponderomotive force developed due to Fig. 1 Sketch of the condensation boundary layer with electric and mag-
the application of magnetic field causes the condensation heat netic fields
transfer to decrease. In the present brief, the effect of external
electric and magnetic fields on heat transfer coefficients was inves-
tigated for laminar film type condensation along an isothermal
semi-infinite vertical plate, as shown in Fig. 1. Small magnetic
Reynold's numbers were considered so that the induced magnetic •to) (4)
field could be neglected. Techniques of similarity and asymptotic \ay/y=i \cDAt/ Jo
expansion were applied. Effects of electric and magnetic fields on The last two terms on the right-hand side of the momentum equa-
the heat transfer is presented in terms of electromagnetic parame- tion (2) represent the external forces due to gravity and electro-
ters for Prandtl numbers 1.0 and 0.008. magnetic fields, respectively. The last term of the energy equation
(3) represents Joule's heat supplied by the external electric and
Analysis magnetic fields.
The fluid was assumed to be electrically conductive and incom- The boundary conditions for the foregoing equations are,
pressible with constant specific heat, viscosity, and thermal condu-
ctivity. The flow is governed by the following nondimensional at y = 0; u = v = 0, t = tw = constant,
boundary layer equations: and
su dv
—+ —= 0 (1) at y = <5; — = 0 and t = tu = constant (5)
ax ay ay
dll dlt 2„.
1^ „a u Introducing a stream function \p(x,y), (u = a\p/ay, v = —af/ax),
u—+v—= + (1 - pJp)JFr ~ M(u + RE) (2)
ax ay R e ayz then using the following expansions for the stream function ^(^,TJ)
rif) rift 1 l^d and the dimensionless temperature 0(^,ij);
u — + v— = + MEC (u2 + 2REu + R E 2 ) (3)
ax ay Re ay2
*(f,u) = (4C/Pe)£3/4 £ (M0J/2fj(v) (6)
where
u = u/Ua, v = D/f/„, x = x/L, y = y/L m,v) = £ mmjM (7)
J j'=o
6=(t- tv)/(tw - tv), Re = pU^L/ix, RB = E/BUJ, 4- V» J-
equations (2) and (3) were reduced to two sets of ordinary differen-
ncp E = tial equations with fj and 8j as dependent variables and -q as an in-
Fr = U„2/gL, M = <rB2L/pU„, Pr c U„2/cp{tw-tv),
n dependent variables. These equations are omitted here for space
limitations. In equations (6) and (7) r\ — Cy/x 1/4 , £ = x are the sim-
ilarity variables, and
ufi are the physical velocity components; 1/4
[peRe[(l--)/4Fr-Mi?E]] (8)
x,y are the physical coordinates;

B is the magnetic field flux density;

Results and Conclusions
and E is the electric field intensity. The ordinary differential equations for fj and 6j were solved nu-
merically using Runge-Kutta method up to j = 2. The relationship
The mechanism of heat exchange at the liquid-vapor interface
between film thickness, TJJ, as a function of the condensation pa-
gives:
rameter, cpAt/hfg, was obtained from equation (4), in (£,rj) coordi-
nates, in the form

1
CpM/h,g=[i(Mi)i%(m)y
Assoc. Professors, Mechanical Engineering Department, University of
Riyad, Riyad, Saudi Arabia.
2
Assoc. Professor, Electrical Engineering Department, University of [£(M^/2fl;(,j)/(2;+3)J (9)
Riyad, Riyad, Saudi Arabia.
3
Numbers in brackets designate References at end of technical note.
Contributed by the Heat Transfer Division of THE AMERICAN SOCI- For each value of the condensation parameter there corresponds a
ETY OF MECHANICAL ENGINEERS. Manuscript received by the Heat value of TJJ and hence the results expressed in the form of Nusselt
Transfer Division. ratio, Nui.m/Nub.o, are plotted in Figs. 2 and 3, where

628 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Pr = 0 . 0 0 8

ncl I I I I

CpAt/h+g

Fig. 3 Nusselt ratio versus condensation parameter for Prandtl number

0.008

In the special case when both electric and magnetic fields were ab-
sent, the plots coincide with those obtained by Sparrow and Gregg
[1]. However, the presence of an electric field in a direction inward
CP At / h f g to the plane of paper, (RE = 2.0), and normal to the flow, tends to
decrease the heat transfer rates, whereas an electric field in the op-
Fig. Nusselt ratio versus condensation parameter for Prandtl number posite direction tends to increase the heat transfer rates. These ef-
1.0 fects are due to the ponderomotive forces developed by the electric
field current as it interacts with the magnetic field in either a di-
rection opposite to the flow or in a direction along the flow, respec-
U/4 tively.
N u ^ / N u , , , ) = [{1.0 - MR E F r /(l - Po/p)]lcpAt/hfg[
The presence of the magnetic field normal to the flow and in the
•Z (MxVl26f(Q) (10) plane of the paper tends to decrease the heat transfer rates, which
j-o is in agreement with results obtained in reference [3]. The last case
Nu*,o = [P c R e X 3 (l - P„/p)/4F r ] 1 / 4 • [hfg/CpAt]1'* (11) is shown in Figs. 2 and 3 for a typical value of RE = 2.0 when the
magnetic expansion parameter Mx ^ 0, for P r = 1.0 and 0.008.
and
References
Nuv 1 Sparrow, E. M , and Gregg, J. L., "A Boundary-Layer Treatment of
-»•© \ay/o Laminar Film Condensation," JOURNAL OF HEAT TRANSFER,
"11/4 TRANS. ASME, Series C, Vol. 81,1959, pp. 13-18.
-PeRex3{(l.-p„/p)/Fr 2 Singer, R. M., "Laminar Film Condensation in the Presence of an
Electromagnetic Field," ASME Paper No. 64-WA/HT-47, 1964.
• MRE) 3 Murty, K. N., Sarma, C. K., and Sarma, P. K., "Laminar Film Conden- .
(12) sation on a Vertical Plate-Effect of Magnetic Field," JOURNAL OF HEAT
TRANSFER, Series C, Vol. 97,1975, pp. 139-141.
X £ (MxV%'(0)

Journal of Heat Transfer NOVEMBER 1975 / 629

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 2{7V AX + 1 + ^- 1 AX(fl A r !S + eu")[($Afs)2 + (0M") 2 ]|
As \[/ —• <*>, (6) becomes identical with the authors' result as
shown in their Fig. 2. For further comparison, use N = 0, \p — 0.2,
X = 0.1, 6R = 1.0, and 00 = 0. Thus when T = 0, 6M" = B0 = 0 and
(6) yields, noting that BMSS = 1.0,

AT g (3.33) IO- 3

The authors' method at T = 0 yields

• \ AT § (5)10" 3
\
By reference to the authors' previous work [5] it is seen that the
:
" .) liiiiisiei time increment size given by Arpaci's method is a stable one, even
at low values of \p, for the problem considered.
As T —• oo, 6M" ~* <?M',S and the stability criterion above is the
same as the one given by the authors.
These results indicate that the authors' method in this problem
provides a "tighter" sufficient condition for stability than does
that of Arpaci.
In some work just completed by the discusser, an explicit finite
difference algorithm was used to solve a nonlinear transient con-

I
On Solving the Transient duction problem in which the thermal conductivity depends lin-
early upon temperature. The stability criterion derived using the
Conducting Slab With matrix method, Arpaci's method, and the derivative method of the
Radiating and Convecting authors was, at least at long times, the same for all methods.
Surfaces 1 Compared to the authors' method, Arpaci's stability technique
possesses some definite advantages. Arpaci's method can be ap-
J . Sucec. 2 The authors are to be commended for their contri- plied to both explicit and implicit finite difference schemes. It
butions to stability analysis of finite difference equations, for cer- seems more general in that the single statement forbidding the ov-
tain nonlinear problems, by use of a physical reasoning approach ershooting of the steady state is rather easily applied to different
which supplements the "derivative method" presented in their transient phenomena. Reference [3] gives some results for finite
previous work [5]. The value of these two methods can really be difference equations simulating two mutually coupled partial dif-
appreciated when one encounters a situation, such as the one the ferential equations in a transient conjugate heat transfer problem.
authors discuss, where, because of the nonlinear nature of the Finally, on the face of it, it appears reasonable to hope that Arpa-
problem, the matrix method of stability analysis [7] fails for at ci's method would be valid for both linear and nonlinear transient
least some of the possible values of the parameters involved. problems since the basic statement, about the finite difference so-
Another method, which the authors may be unaware of, also ex- lution not overshooting the steady state, does not seem to rest
ists which utilizes physical reasoning to establish stability criteria upon the partial differential equation being linear. Hence, the
for finite difference algorithms applied to unsteady-state prob- basic statement seems to be at least a necessary condition, even for
lems. This method is given by Arpaci 3 and can be summed up as nonlinear equations.
follows. These statements in favor of Arpaci's method are not intended
The time increment AT must be chosen in such a way that the as a refutation of the authors' method, indeed it was the authors'
value of the dependent variables at any node cannot change to a article, focusing attention on this nonlinear problem, which caused
value beyond the steady-state value in a single time increment. renewed interest in Arpaci's method on the part of the discusser
To illustrate, consider the authors' finite difference equation, who would be happy to learn about generalizations of their method
to implicit algorithms and to other types of transients.
0M n+1 - 0M" = [A - B6M" - C u V ) 4 ] A r (3)
Using the method of [3], the steady-state finite difference equation
for this node can be written as,

0 = A - B6MSS ~ C«VS)4 (4)

Authors' Closure
The superscript ss means steady-state conditions. Next (4), multi- The authors would like to thank Sucec for his interest shown
plied by AT, is subtracted from (3) and 6MSS ~ QM" is factored out through his discussion of our technical note. It should be noted
yielding, that the material presented therein was completed some three
years ago. At that time stability criteria for finite difference solu-
0 M " + 1 - 9M" = « V S - AM")AT
tions to the conduction problem with nonlinear (radiative) bound-
X \B + C(0M"< + eMnWM's)2 + (V)2]l (5) ary conditions were an area of current interest to the authors. Lit-
In order that 0M"+1 not go beyond the steady state, it is required tle information on the topic appeared to be readily available in the
that 0M n+1 — SM" S 0AISS — &Mn which leads to the following condi- literature.
tion for stability, after replacing B and C by their equivalents. It may be noted that mention of our consideration of the "deriv-
ative method" for a nonlinear partial differential equation was
given in [5, p. 59]. A limited number of other nonlinear partial dif-
ferential equations have been studied successfully (using both im-
1
By J. L. Milton and W. P. Goss, published in the November 1974 issue of plicit and explicit schemes) with the derivative method but not re-
the JOURNAL OF HEAT TRANSFER, TRANS. ASME, Series C, Vol. 96, ported. A discussion of either the applicability of the method to
pp. 547-549. implicit techniques for radiative boundary conditions or nonlinear
2
Assoc. Professor, University of Maine, Orono, Me.
3
Arpaci, V. S., Conduction Heat Transfer, Addison-Wesley, Reading, partial differential equations was not the objective of the afore-
Mass., 1966. mentioned articles. The objective was to discuss two sets of stabili-

630 / NOVEMBER 1975 Transactions of the ASME

Copyright © 1975 by ASME
Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ty criteria ("derivative method" and the "explicit method") that counts as indicated in his comparison problem (at time T = 0 ) :
had been found to be useful for solutions to a certain class of prob- (i) 8«"mas is given as ~0.474625 (Fig. 1) from the solution of
lem discussed in the article. The matrix of coefficients of the un- the quartic equation. Hence ( ^ BM" - 1.0). Moreover, #Aj"|max ' S
known solution vector will have (some) nonlinear terms in general not constant, but is time dependent.
requiring an iterative solution unless the terms are linearized; (see (ii) ATDiscuaser 5 ATExplicit ^ ATDerivative i.e., 3.33 X 1 0 - 3 < 4.74
Section 8-6). 4 X 1 0 - 3 < 5.00 X 10" 3
Keep in mind the fact that a number of different stability condi- As can be seen from these results, the derivative method gives the
tions with varying degrees of restrictions will be reported by vari- largest value of AT at time equal zero. It has been shown in [5, Fig.
ous authors before a complete understanding of the problem is 10] that the method is stable. As pointed out in our paper the de-
presented. Perhaps both necessary and sufficient conditions are rivative method was recommended because of its numerical sim-
even farther away. But in the meanwhile, many practical problems plicity. This recommendation is unchanged.
will be solved with existing stability criteria. (iii) Graphical interpretation is possible for both methods. In
The discusser was correct in stating that the authors may not the derivative method of stability criteria, the slope of the plot of
have been aware of the applicability of the method presented in [8, BM"+1 versus BM" has been employed to predict AT. In the explicit
Section 9-8] to the present problem with nonlinear boundary con- method the slope of the plot of AC?A//AT versus T (or 0Ai"|max) has
ditions. There were several reasons for our not considering that been employed to predict flwimax and hence AT. It should be
statement applicable: The reported work discussed examples with noted that for general problems with a nonlinearity: T " + 1 ?* (n +
linear boundary conditions. With the present problem, a nonlinear 1)AT since AT is itself variable.
relation exists between the surface nodal point and the nodal point Short of actual numerical computations, it is difficult to evalu-
in the "extended" material. The similar triangle scheme would not ate AT of the aforementioned example except a s r - > <». There it
work. A quartic polynomial would be needed there with a linear would be expected that all results would yield similar conditions.
polynomial in the nonextended material. If the extended material When (>Mnmax is used in place of the discussers' BMSS the results
were assumed to have temperature dependent thermal conductivi- from [8, Sec. 9-8] appear to yield conditions somewhat similar to
ty the partial differential equation would become nonlinear and those from the explicit method for AT. However, the authors point
beyond the scope of our work. out the fact that BM"\ m a x varies since it is obtained from the solu-
The discusser appear to have misinterpreted our work on several tion of the quartic equation, equation (2), that is time dependent.
Further numerical studies as well as theoretical investigations
will be needed before a complete understanding of stability condi-
tions will be available for finite difference solutions to problems
4
R. D. Richtmyer and K. W. Morton, "Difference Methods for Initial- with nonlinearities. It is hoped that the discussed article has
Value Problems," Second ed., Interscience, New York, 1967. helped to spark that aforementioned interest.

Journal of Heat Transfer NOVEMBER 1975 / 631

Downloaded 24 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm