You are on page 1of 28

15

THERMODYNAMICS AND FLUID MECHANICS GROUP

ACCELERATION AND THE CRITICAL HEAT FLUX IN POOL


BOILING HEAT TRANSFER

By H. J. Ivey, B.Sc., PhD. (Student)*

Many theories predicting the critical (or burnout) heat flux in saturation pool boiling have been proposed
during the past decade. Although most are able to correlate existing experimental data, it is evident that the
theoretical models upon which they are based are widelp different in concept. Moreover, owing to lack of
experimental evidence it is not clear which, if any, are based upon the more correct models. Certain of the
theories implicitly contain the acceleration due to gravity raised to a different power; by performing a
relatively simple experiment it may be possible to accept or reject certain of the theories. An experiment of
this type is described in the present paper. On the basis of the results of this experiment it is possible to reject
several of the above theories.
The extent to which the acceleration of the boiling system is important as a parameter in determining the
critical heat flux was ascertained experimentally. A simple centrifugal apparatus is described which was used
for determining the critical heat flux in pool boiling to water at atmospheric pressure and accelerations in
the range 1 < (alg) < 160. A two-variable non-linear correlation involving a logarithmic transformation
was made upon the experimental results after allowing for the possible effect of pressure and potential sub-
cooling by extrapolation to zero subcooling. The resulting equation is of the form:
qcr, (aig) 0.27

THEORETICAL ASPECTS O F CRITICAL time. most are able to correlate existing sets of experimental
HEAT F L U X IN SATURATION POOL data, but their predictions for other conditions (for example,
BOILING accelerations greater than that of local gravity) are in many
Introduction cases dissimilar, while the models upon which they are
WITH THE ADVENT of many systems (nuclear reactors, based are often incompatible. It is the purpose of thc first
rockets) requiring heat removal with high heat fluxes, part of the present paper to attempt a rationalization of the
considerable effort has been spent on boiling heat transfer numerous theories proposed and to inspect critically the
in general, and on the critical heat flux in particular. As the initial assumptions of several of these theories. The extent
boiling mechanism is complex, much ingenuity and analysis to which the acceleration of the boiling system is important
have been directed towards solution of the simplest boiling as a parameter in determining critical heat flux is also
heat transfer condition : that of saturation pool boiling. indicated, this will be of use in the second part.
Many different theories and analytical models have been
proposed to predict the critical heat flux in saturation pool Notation
boiling; many of these invoke specific configurations of the Ar Dimcnsionless parameter = gZo3(pL- po)/v2p0.
liquid-vapour phase boundaries as functions of space and
a Area fraction of surface covered by vapour
The M S . of this paper was first received at the Institution on 9th bubbles.
January 1962, and in its revised form, as accepted by the Council a Acceleration, fi/sec2.
for publication, on 27th April 1962.
* Reactor Development Division, Atomic Energy Establishment, ak Dimensionless acceleration.
Winfrith. c Constant.
Proc Instn Mech Engrs Val 177 AT0 I 1963
1
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
H. J. IVEY

Drag coefficient, CD = 4 g ( ~ ~ - p , ) D b / 3 p ~ u , ~ . Description of the two-phase fluid in saturation


Specific heat at constant pressure, Btu/lb mass pool boring
OF.
Critical heat flux-system parameter relatiods may be
Ratio of total heat transfer to the heat transferred conveniently divided into two main groups :
by the production of vapour in bubbles
= qtotal/qbubble- (1) those derived by correlation of experimental
Vapour binding coefficient = n Db2. . results using dimensional analysis,
Diameter of vapour bubble at departure, ft. (2) those derived by analytical considerations.
Diameter of vapour bubble the instant it breaks The analytical group may be further divided into three
through the free liquid surface, ft. subsections depending upon the particular physical model
Frequency of bubble formation at a nucleating which is used. The use of physical models is necessitated
site, sec-1. by the inability in general to describe the liquid-vapour
Acceleration due to gravity, ft/sec2. interface as a function of time and space. The subsections
Conversion factor in Newton's law of motion, of the analytical group arise from three basic variants of the
lb mass ft/lb force sec2. model used to describe the system conditions occurring at
Enthalpy, Btu/lb mass. the critical heat flux with respect to the disposition of the
Dimensionless parameter = X/C,T,. liquid and vapour phases. The three basic variants of
Thermal conductivity, Btu/ftz sec "F/ft. the analytical group are:
Characteristic vapour length, proportional to
diameter of bubble on departure, (2a) vapour as the discontinuous phase, liquid as the
continuous phase;
lo" [gc&GL- P v F (2b) both vapour and liquid as discrete phases;
Number of bubbles per unit area, ft-2. (2c) vapour as the continuous phase, liquid as the
Absolute pressure, lb force/ftz. discontinuous phase.
Absolute critical pressure, lb force/ft2.
Heat flux, Btu/sec ft2. In variants (2a) and (2c) planar cross-sections of the two-
Critical heat flux, Btu/sec ft2. phase system would show the liquid and vapour phases
Radius, ft. respectively to be multiple-connected domains topologic-
Radial velocity, ft/sec. ally. In some examples of variant (2b) a planar cross-section
Radial acceleration, ft/sec2. whose normal is parallel to the accelerative field would
Reynolds number based on superficial vapour show that each packet of both the liquid and the vapour
velocity = qlo/Ap,v. phases are simply connected domains.
Absolute temperature of saturation, "R. The majority of relations determining the critical heat
Velocity of bubble, ft/sec. flux as a function of system variables in pool boiling heat
Average bubble growth rate, ft/sec. transfer implicitly contain the acceleration due to gravity.
Relative velocity of two fluids in parallel streams, In order to clarify the origin of the accelerative term for
ft/sec. each particular theory, a brief systematic survey of the
Velocity of vapour normal to heater surface, critical heat flux relations for saturation pool boiling is
ft/sec. made. Several of these theories have been derived using
Contact angle, radians. initial assumptions which subsequently have been shown
Degrees subcooling, "F, = saturation tempera- to be inaccurate according to latest experiments. Note is
ture-liquid bulk temperature. made of these cases where appropriate.
Angle, radians.
Angular velocity, radisec. Theories of critical heat j u x for saturation pool boiling
Angular acceleration, rad/sec2. (1) Correlation of experimental results using dimensional
Latent heat of vaporization, Btu/lb mass. analysis :
Absolute viscosity, lbjsec ft.
Kinematic viscosity = p/p, ft2/sec. Addoms (I)* derived a correlation by dimensional
Mass density, lb mass/ft3. analysis in which the average volumetric vapour dis-
Surface tension, lb force/ft. engaging rate per unit area of heater surface is a function
of a buoyancy term and of the product of the liquid
thermal diffusivity and acceleration due to gravity. It
may be expressed as :
Subscripts
L Liquid.
V Vapour.
cr Critical. Kutateladze (2) considered the hydrodynamic motion
sat Saturation. of a non-viscous two-phase fluid in terms of the equations
sub Subcooled. References are given ia the Appendix.
Proc Instn Mech Engrs Val 177 No I 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAT FLUX I N POOL BOILING HEAT TRANSFER 17

of motion for the two phases, a condition for equality of wrote an expression for the fraction a of the surface
turbulent tangential forces at the phase boundary, and a covered by bubbles :
condition for equality of pressures at the phase boundary.
A system of criteria was determined through dimensional 4
. . , 77
a = - Dbzn(8) .
analysis to give :
and assumed that the critical heat A ux condition occurred
4cr = Chpv’ [ ~ ~ ~ ~ P L - P J I ’ (2) - when successive bubbles leaving the surface touched and
Sterman (3) described the heat transfer between the coalesced :
heater surface and the boiling fluid with the equations of
motion for the fluid, heat transfer from the laminar sub-
ub=fDb . . . (9) . .
layer to the boiling bulk of the liquid by turbulent heat Equations (7)-(9) give :
transfer and molecular mass (thus heat) transfer, the qcr = hp,ub(+a) . . (10) .
equation of heat transfer at the solid-liquid interface for T o determine the bubble velocity 24, an expression
no change in phase, and the convective heat transfer equating buoyancy force to drag force on a freely rising
equation with an internal heat source. Dimensionless vapour bubble was used:
groups were obtained and compared with the experimental
data of Cichelli and Bonilla (4)to give:
(‘QevaJcr = 0.0365Ar@57KS-f . . (3)
g(pL-h)
.TI 17
5 Ob3 = cD 4D b 2 ( b o u b 2 ) - (11)

When the criterion K, is neglected and the above rela- It is stated that the vapour density is used in the expression
tion re-evaluated with the experimental data the following for drag on the vapour bubble, as the fluid near the wall
is obtained: is mostly vapour when film boiling begins. The Fritz
formula for bubble breakaway diameter was used to
(‘ZQevap)cr
= 0.168ArO.5 . . (4) complete the above expression (9), that is :
This may be written as :
4cr = 0.168+2 [glO(PL-Pa)lk * * (5)
which is the Kutateladze retation (equation (2)). Equations (lo)-( 12) give:
Borishanskii (5) considered the stability of a liquid jet
surrounded coaxially by the moving vapour phase. The
effect of viscosity was not neglected and two similarity
criteria were derived by dimensional analysis. On com-
4cr = const- a ($1 f
.hPv+[ggc4%-f,)l* (13)
It was assumed that a is a constant; the above expression
parison with experimental data the effect of viscosity then reduces to that reported by Cole (IO), although the
was found to be small; when it is neglccted the criterion additional term (pJpL)* multiplying the right-hand side
involving viscosity is removed and the Kutateladze is obtained when using the information supplied by
relation results (equation (2)). Cole. This is due to non-statement of the assumption
(2u) Vapour discontinuous, liquid continuous analytical made by Deissler of the use of vapour density rather than
model : liquid density in equation (1 I), which assumption is
suspect.
Jakob and Linke (6) attempt to derive a dimensionless Rohsenow and Griffith (11) (12)(13) first quote the
representation for the boiling heat transfer mechanism correlation of Addoms (equation (1)) and then state
by means of physical reasoning. The heat transfer is
‘since it was observed (14) that bubbles shoot away
postulated to be promoted by vapour bubble growth and
perpendicularly from vertical surfaces it appears that
quantities influencing this motion were introduced. gravity should perhaps not have a significant effect on
They write the total volume of vapour formed per unit
the peak heat flux values’. A model for the critical heat
time due to heat energy taken from unit area of heating
flux for saturation conditions is then proposed which
surface as: does not contain an accelerative term. The Jakob and
Linke expression (equation (6)) is written:
7T
It should be noted that this expression was not derived ‘?total = Cqqbubble = cq.hpv 6 Db3.fn (14)
for the condition of critical heat flux; however, their
work is mentioned as the above expression is similar to where the heat transfer to a single bubble was assumed
the starting point for most other theories employing a to be proportional to the total heat transfer. The critical
vapour discontinuous type of model for the critical heat heat flux condition was assumed to be given by a ‘vapour-
flux. binding’ condition at which bubbles on the surface just
Deissler (7) (S), starting with the Jakob and Linke touch and coalesce:
expression: n = cvb/Db2 . . . . (15)
This is similar to the assumption made in the analysis of
(7) Deissler, equation (8).
Proc Instn Mech Engrs Vol 177 No I 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


18 H. J. IVEY

Equations (14), (15) give : vapour bubbles are in a square array on the heater surface
4cr = CqCvbboCfDd -
(16) * - (Fig. I). 'The maximum number of bubbles from each
site follow the principle that when one bubble detaches
The final correlation, assuming (fD6)
is a constant, was from one site a new bubble just starts to form in the
obtained by comparison with experimental data: neighbouring site.' Using this same model for the
distribution of the two phases the peak heat flux is
estimated from two points of view. The first considers
The above observation made by the authors concerning the equation of motion of the bubbles :the mass-accelera-
the insignificance of gravity effect due to the 'shooting' tion of the bubble is equated to its buoyancy, friction and
away of bubbles is not surprising; they quote results induced mass effects are neglected, and this equation is
concerned solely with forced convection, highly sub- integrated to determine the time ( t 2 )for a bubble to rise
cooled, boiling experiments of McAdams et al. (15) a distance of 2Db from the heater surface. With the
whereas equation (17) should apply for saturation pool Deissler-type relation (equations (7) and (8)) :
boiling. 5T 1
Griffith (16)starts with equation (16) and the assump-
4cr = (6 Db3) . . (22)
tion (fDb) cc Vgrowthto give: the Fritz formula (equation (12)) and published data for
contact angles, the critical heat flux is given by:
4cr = hpuVgrowth * - (18)
4cr = (0.17 +-+ 0'23)h~~lgg,a(~L--p,>I~ (23)
An expression for Vgrowthat saturation conditions was The second Chang and Snyder (17) interpretation
determined by assuming the bubbles obey Stokes law: considers the capillary wave which is postulated to occur
- (l9)
ub = g ( f L - f ~ ) ~ b ~ / 1 8 ~ . L 3
at the interface of the bubble owing to relative motion
of the bubble. For two non-viscous fluids flowing in
for the rise velocity to give: parallel vertical streams the interface becomes unstable
when their relative velocity is :

Doubt is cast upon the use of Stokes law to obtain an


expression for the rise velocity at saturation conditions The maximum relative velocity is computed from the
since Cole (10) has shown that for saturation pool geometry of Fig. 1 and continuity: this gives vr = 1 . 8 ~ ~ .
boiling at the critical heat flux the buoyancy may be From equations (22) and (12), and the time ( t 2 )taken by a
equated to the drag force to give the following rise bubble to rise a distance of 2 0 , the following is obtained:
velocity : v
4cr = hJ% - - * (25)
Equations (24) and (25) give:
Chang and Snyder (17) postulate a model where the

Since this model is implicitly dependent upon the


postulated geometry of the vapour bubbles (Fig. l), and
since all existing evidence (photographic and otherwise)
indicates the chaotic behaviour of the two phases near
the critical heat flux, equations (23) and (26) bascd upon
the geometry of Fig. 1 are of dubious validity.
(2b) Both vapour and liquid as discrete phases :
Zuber has published a number of papers (18)-(24)in
which the following expression is derived for the critical
heat flux in saturation pool boiling:

It is extremely difficult to obtain a clear picture of the


model postulated in order to arrive at the above expres-
sion. The physical model has changed since the above
J equation was first presented. An outline is given:
\\\\\\' i\\i\\\\\\\\~\\\.,\\\\
Reference (18). Considered stability of capillary waves
Fig. I . Proposed model for mitical heatjux of Chang occurringin vapour flowaround perimeter of horizontal
and Snyder (17) tube (an illustrative diagram is given to this effect).
-
Proc Inrtn Mech Enzrs Vol I77 No 1 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAT FLUX I N POOL BOILING HEAT TRANSFER 19

Reference (19). No clear statement of the model is made


except ‘if the thickness of the film is small compared
with the diameter of the tube we can use the plane
approximation’.
Reference (20). The model for the analysis is a hori-
zontal liquid-vapour interface. Fig. 2. Proposed model for critical heat flux of Berenson (27)
References (21)(22). A slightly different model is pre-
sented which adds the consideration of the stability of No analytical model exists to date using this configura-
a fluid jet surrounded by another fluid. tion. It is, however, important to note that Wallis (28)
Reference (23). The following statement appears, ‘The (29)has shown that the ‘fluidization’ point in a general
system analysedinreferences (19)(21)contains regularly fluid-fluid system (the transition from gas jet to dropwise
spaced vertical jets of vapour rising from a horizontal flow) is predicted by an equation similar to that of
heated surface in pool boiling. These jets are postulated Kutataladze, that is :
to be on a square array whose dimensions are deter- -
v v , cr = K~u-+[ggco(~~-~u)I’ (29)
mined by the wave lengths of waves which are unstable This transition point may be regarded either as the
under Taylor instability.’ upper limit of the gas jet regime or as the lower limit of
Reference (24). It is assumed that the vapour rises in the suspended drop regime (the ‘fluidization point’).
continuous circular columns through the liquid (i.e. Below this point the liquid drops rest on one another and,
as opposed to discrete bubbles of vapour as in the first depending upon surface properties, may join to form
two references (IS)(IS). The result is unchanged. liquid jets.
It should be pointed out that the entire analysis is based EXPERIMENTAL DETERMINATION OF THE
upon consideration of Taylor instability of a liquid- EFFECT OF ACCELERATION ON CRITICAL
vapour interface. This was made possible (though not HEAT FLUX IN POOL BOILING
necessarily permissible) by assuming no liquid-solid Introduction
contact in transition boiling. It has been shown (25) Many theories and correlations for the determination of the
that when an electric current was passed in series through critical heat flux in saturation pool boiling have been pro-
a bubbly spheroid and the heating surface, the circuit posed in the last decade. Many predict an increase in
was closed briefly at the instant the vapour bubble broke critical heat flux with increase in acceleration (2) (3) (5) (7)
through. It has also been shown that liquid-solid contact (17)(19) (27); several predict an even greater effect of
exists by a different method (26) for transition boiling acceleration (I) (16); and one predicts no effect of accelera-
through the necessary existence of a triple interface. tion (12).The principal objectives of the second part of the
This conclusion invalidates a major assumption made by present paper are the experimental determination of the
Zuber in his analysis of the peak heat flux as the upper effect of acceleration upon the critical heat flux in saturation
limit of transition boiling, that liquid-solid contact does pool boiling at atmospheric pressure, and the examination
not exist. The assumed non-existence of liquid-solid of the predictions of the above correlations and theories in
contact allowed the application of Taylor instability to the light of experimental results.
the liquid-vapour interface which in turn determined Although a theoretically derived prediction that critical
the geometrical spacing of the liquid and vapour columns heat flux for saturation pool boiling as a monotonically
through the critical wavelength. increasing function of acceleration has been available for
Berenson (27).The experimental work of Borishanskii over a decade (2),the first suggestion to test this conclusion
(25) and Berenson (26) conclusively demonstrate that experimentally has been made only recently (19).Present
liquid-solid contact exists in transition boiling. Pre- experimental information on the effect of acceleration on
sumably because the assumption of Taylor instability is boiling heat transfer is extremely limited. Qualitative
invalidated by this, Berenson postulates a different two- evidence for the importance of buoyancy forces in saturation
dimensional model of contra-flowing planes of vapour pool boiling under zero-gravity conditions has been
and liquid. This is shown in Fig. 2. The following demonstrated (30); some information on zero-gravity
expression is obtained (which still requires a constant) by boiling and condensing is available (31); and the effect of
minimizing the kinetic energy per unit time at any one accelerations to (a/g) = 21 on the natural convection-
horizontal plane for the assumed model: nucleate boiling region has been determined (32).
4cr = Cbu’[ggc~(~L-~v>I‘x The only data available concerning the effect of accelera-
tion on the critical heat flux in pool boiling are those of
Usiskin and Siege1 (33) and Costello and Adams (34). In
the former paper the results were obtained for accelerations
The comments made concerning the analytical model of in the range 0 < (u/g) < 1, and are subject to the error of
Chang and Snyder (17)apply here also. being obtained not under steady-state but under transient
(2c) Vapour continuous, liquid discontinuous analytical conditions owing to the limited time of fall of the experi-
model. mental apparatus. In the latter paper the results indicate a
Proc Instn Mech Engrs Vol177 No 1 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
20 H. J. IVEY

general increase of critical heat flux with increase of accelera-


tion, but the nature of this increase is at variance both with \
theoretical predictions and with the results of the present
investigation. This is discussed further in a later section,
and in the discussion to the paper of Costello and Adams
(35).

Experimental method
Concept of design
To produce a steady-state acceleration greater than
(a/g) = 1, it is necessary to produce either a uniform rate of
change of velocity under rectilinear motion or a uniform

n
rate of change of direction at constant speed. The former
method, though producing the simplest accelerative field,
CONDUCTORS
must of necessity be limited to transient or unsteady-state
experiments. Thus the latter method was chosen, although
the acceleration produced is complex. In general, the
acceleration of a particle in plane polar co-ordinates about STAINLESS STEEL
TEST VESSEL
the instantaneous centre of rotation consists of the terms
r and r82 acting parallel to the instantaneous radius, and
rb’ and 2i8 acting normal to the instantaneous radius. The
boiling test section was rotated in the horizontal plane at
constant angular velocity 8 and constant radius r to produce
an accelerative field to (alg) = 160. The only acceleration to
which the test section was subjected was centrifugal
I
(r82); the bubbles leaving the test section had a radial --------
velocity component and were subjected in addition to ,,,,,, ,,,,,,,,,,,,,,,, , p U X I L I A R Y HEATER
CorioIis acceleration (2iB).
Fig. 3. Schematic v i m of centrifugal test vzssel
Apparatus
The test section was placed tangential to the circle swept by 36 S.W.G. chrome1alurnel thermocouples immersed in the
the radius arm, so that the acceleration was effectively vessel at different levels. The counterbalance served as an
uniform along its length. The test vessel was not freely insulated temperature-reference vessel for the test vessel
pivoted as the acceleration due to gravity could be neglected thermocouples; this enabled the resultant thermocouple
for all but the lowest centrifugal acceleration, The majority e.m.f. to be obtained from the rotating assembly using only
(142) of the experimental determinations of the critical heat material-junctions of low thermal e.m.f.’s. The equivalent
flux were carried out with 18/8/Ti stainless steel tubes of thermocouple circuit is shown in Fig. 4. The test section
0.049 in. outside diameter, 0.040 in. nominal inside dia- voltage and thermocouple e.m.f.’s were obtained fiom the
meter and l$ in. in length. The remainder (18) were identical rotating vessel through an instrument slip ring assembly
except that the nominal inside diameter was 0.042 in. The consisting of eight electro-deposited silver rings and
ends of the tubes were soft-soldered to stainless steel blocks 85 per cent silver-I5 per cent graphite brushes on beryllium
and voltage taps were taken from the solder near the test copper springs.
section ends so that the voltage drop across the tube could Power was supplied by d.c. generator with 2 per cent a.c.
be accurately determined. The ends of the electrically heated ripple at 1 volt and less than 1 per cent a.c. ripple over
test section were clamped to two stainless steel conductors 3 volts. The power was controlled by varying field excitation
which led into the stainless steel vessel through Tufnol with two variacs placed in series to enable small step incre-
seals. The radius at which the central axis of the test section ments to be made at any particular operating condition.
was placed in the present series of experiments was 2553 in. The d.c. current was measured with a shunt. The d.c.
Owing to the large radial distance compared with the size of currents for test section and auxiliary heater were carried
the test section, a total variation of 4 per cent in centrifugal to the test tank through a large silver slip ring assembly.
acceleration occurred between the points on the test section It was possible to carry out a number of burnout runs on
nearest and farthest from the axis of rotation. This appears one specimen with the use of a burnout detector. The test
favourable when compared with the experiments of section formed one arm of a Wheatstone bridge, one of the
Costello and Adams (34), in which this variation amounts to electrical conductors entering the test vessel formed another.
almost 10 per cent. Details of the test vessel arc shown in Two variable external bridge arms were incorporated into
Fig. 3. the detector to enable a continuous balance to be made
The temperature of the water was measured with two over a wide range of conditions. When the bridge became
Proc Insrn Mech Engrs Vo1177 No 1 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
CRITICAL HEAT FLUX I N POOL BOILING HEAT TRANSFEK 21

ROTATING NON - ROTATING


a-
!
I

CH ROM EL COPPER
/------

t‘,%Y%L----
MEASURING
THERMOCOUPLE
JUNCTION
f
INSULATED REFERENCE
I

LSI
LVER-ORAPH ITE
COPPER

COPPER
TERMINAL
POSTS

THERMOCOUPLE JUNCTION BRUSHES

SILVER SLIP
RINGS
Fig. 4. Equivalent thermocouple Circuit

unbalanced, the following sequence of events occurred: an then set accurately to the desired value, the burnout
internal high-speed relay was tripped, releasing two other detector bridge balanced and switched into circuit, and the
relays in series, which in turn released a spring-loaded power raised slowly to provide ample time for the boiling
circuit breaker. Total time of operation from first out-of- to be considered as steady-state. During this time the
balance signal to open circuit varied between 25 to 30 msec. detector bridge circuit was manually balanced as the rate of
This was sufficiently fast to prevent physical burnout of the change of resistance of the test section was low.
test section at all values of acceleration. When a burnout was initiated in the test section the rate
The rotational speed was obtained with a photo-transistor of change of resistance became too great to allow manual
which generated six pulses per revolution. A digital rebalancing; the detector then tripped the d.c. circuit
electronic counter recorded the number of pulses over a breaker, saving the test section from physical burnout. The
time interval determined by a clockwork clock and actuated rotating assembly was shut down. The reference vessel
by an electronic timing unit. A continuously variable d.c. temperature was measured with a mercury-in-glass ther-
motor was used to drive the main shaft supporting the test mometer, the quantity of water in the test vessel checked,
vessel and counterbalance. the assembly was reset spinning, and a further test
commenced.
Experimental procedure The centrifugal acceleration of the test section was cal-
Experiments were carried out in two different apparatus, culated from the radius of the section and the angular
The first series employed a static stainless steel pool velocity. The hydrostatic pressure at the test section level,
boiler to determine the effect of liquid subcooling an the resulting from the high acceleration, was computed from
critical heat flux in boiling to water. The second series the depth of water and the system acceleration making an
employed the previously described centrifugal test vessel to allowance for the radial variation of acceleration. The
determine the critical heat flux as a function of acceleration. depth of water was changed within the limits of 1 in. and
The experimental procedures for the two series of experi- 3+ in. and its initial conductivity varied between 2 and
ments are identical except for the additional factor of 10 pmhos/cm.
determination of the angular velocity of the centrifuge in the
second series of experiments. Discussion of results
A brief description of a typical critical heat flux test I t is well known (36) (37) that a pool boiling experiment
carried out on the centrifugal apparatus is now given. The carried out with relatively small diameter or thin-walled
test section was produced in a special jig and the dimensions test sections may lead to reduced values of critical heat flux
measured; it was then attached between the electrical if the diameter is too small or the wall too thin. Ideally the
conductors, the voltage tappings connected and the speci- characteristic length of the boiling element should be large
men cleaned with acetone. The test vessel was weighed with and the wall thick. In the present experiment the slip ring
a specific amount of demineralized water to provide a was limited to a maximum current of 120 amp; a compro-
balanced rotating assembly. The amount added was varied mise was therefore made between test section diameter and
for different tests to allow the production of different wall thickness. The test sections chosen were 18/8/Ti
induced subcoolings at the test section level through steel tubes 0.049 in. diameter and approximately 0.004 in.
differing heights of liquid above the test section. The and 0.005 in. wall thicknesses. The above considerations led
assembly was set in rotation and a moderate current to a series of control experiments; these were carried out
passed through the test section for approximately an hour to upon stainless steel specimens of different diameters and
de-gas the test section surface. The speed of rotation was wall thicknesses in a separate static pool boiling apparatus
Proc Insrn Mech E t p s Vol177 No I 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
22 H. J. IVEY

160

120

80
0.010 0.100
HEATER DIAMETER-in

Fig. 5. Effect of heater diameter on critical heat flux in water

containing demineralized water at saturation conditions and addition a faired curve developed by Bernath (36) from the
atmospheric pressure. The variation of critical heat flux results of Bender (38) is given for comparison. As before it
with test section diameter is shown in Fig. 5, where the is apparent that the particular nominal wall thickness chosen
results of the present investigation are compared with the for the centrifugal experiments provides reasonable repre-
faired curve developed by Bernath (36) from a number of sentation of the critical heat flux when compared with much
experimental investigations. It may be seen that the par- greater thicknesses. The variations of critical heat flux with
ticular diameter chosen for the main series of experiments heater diameter and wall thickness are interdependent.
in the centrifugal apparatus provides a reasonable repre- The object of presenting Figs 5 and 6 is to demonstrate that
sentation of the critical heat flux when compared with test both the diameter and the thickness chosen for the present
sections of larger size. As this size of heater allows some experiments were satisfactory.
critical bubble or vapour packing condition to develop on In the centrifugal apparatus, owing to the height of
the heater surface (corresponding to the critical heat flux) liquid above the test section and the high accelerative field
at normal gravity condition, it is reasonable to assume this to which it was subjected, pressure was produced at the
condition develops at higher accelerations. This may be test section level which also induced a potential subcooling.
inferred from the Fritz relation (equation (12)); according Thus the variation of critical heat flux with liquid subcool-
to this an increase in acceleration reduces the size of the ing was determined under conditions of normal gravity in
bubble at departure and this occurs for constant heater size the static pool boiling apparatus. The results of the experi-
in the present experiments. The variation of critical heat ments for stainless steel tubes of various sizes are shown in
flux with test section wall thickness is shown in Fig. 6. In Fig. 7; for comparison the curves obtained by Ellion (39),

195

150
X

\
E,
3 125
I
s
-1
LL
kw 100 !-
i
I

s
F
ti 75
CURVE DUE TO EERNATH (36)
1.
u

50

25
I

HEATER WALL THICKNESS- in.

Fig. 6. Effect of heater wall thickness on critical heat flux in water


Proc Instn Mech Engrs Vol I77 No I I963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
CRITICAL HEAT FLUX I N POOL BOILING HEAT TRANSFER 23

800
UNCORRECTED CURVES DUE TO ELLION (39)

HEATER DIAMETERS
SUBCOOLED RESULTS : 0.105 in.- A
0064in.- +
0.0481n.- 0
(0.040 in. I D 1
0.048in.- 0
(0.042in I D )
-SATURATION RESULTS 0.105in- -.A
0.064in.- X
0~048in.- e
~ (0.040in. I.D.)
0.048in- I
(0.042 in. I.D.)

b 80 1 0
LlOUlO SUBCOOLING -*C

Fig. 7. Critical heat $ux in pool boiling as a function of liquid subcooling. Results for water

using a 0.0043-in. thick stainless steel strip boiling from one is clear from the figure that the results of Costello and
side only, are also shown. The uncorrected curves of Ellion Adams are at variance with the present results and with
were obtained by defining the liquid subcooling to be the the theoretical predictions. A possible explanation has
temperature difference between the saturation temperature been proposed in which, owing to the large diameter
of the liquid and the temperature of a thermocouple placed of cylindrical heater used, the vapour forming at the surface
half an inch directly above the boiling section. The corrected farthest away from the axis of rotation was unable to rise
curves of Ellion were obtained by correcting the tempera- freely at the lower accelerations (34) (35). It is apparent
tures obtained directly above the boiling section to the that the results of the present experiments are in qualitative
temperatures of the bulk liquid away from the section. The agreement with the theoretical prediction in which the
total gas content of the water in the present series of exponent of (a/g) is equal to 2 since the increase in critical
experiments was in the range 10-14 cm3/litre; the results heat flux due to the increased acceleration alone is certainly
fall between Ellion’s experimental curves. The critical heat lower than that shown in Fig. 8, in which the uncorrected
flux values in boiling to both aerated and degassed water experimental values include the possible effect of subcool-
become similar at fmite subcoolings lower than approxi- ing upon the values of critical heat flux. One point of
mately 30°C. This permitted the use of partially degassed similarity between the present results and those of Costello
water as the induced subcoolings in the centrifugal experi- and Adams is the similarity of the scatter limit of values
ments did not exceed 20°C. of critical heat flux about their means of approximately
The results of the present series of experiments with the f 1 2 per cent.
centrifugal apparatus to accelerations of (a/g) = 160 are Four correlations were made upon the experimental
shown in Fig. 8. Also shown are the results of Costello and results :
Adams (34) for accelerations from (a/g) = 1 to 44 and No. 1. Correlation of the uncorrected data.
Usiskin and Siege1 (33) for accelerations from (u/g) = 0 to 1. No. 2. Correlation of the corrected data to allow for
For qualitative comparison the function the calculated effects of pressure and subcooling assuming
the variation of dimensionless critical heat flux with pres-
deduced by theoretical considerations& alio shown- in sure and subcooling for normal gravity to apply to higher
the same figure where the exponent n is both f and 4.It accelerations.
Proc Instn Mech Engrs Voll77 No 1 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
24 €1. J. IVEY

ILI
x COSTELLO AND ADAMS ( 3 4 ) 1 - d a / g ) < 44

USISKIN AND SIEGEL (33) 0 <(a/g)<

60 - 80 100 120 140 3

DIMENSIONLESS CENTRIFUGAL ACCELERATION AT TEST SECTION RADIUS ( a / g l

Fig. 8. Comparison of uncorrected critical heat flux-acceleration variation with the results of Costello and Adams (34) ; Usiskin
and Siege1 (33). Results for water

No. 3. Correlation of the corrected data to allow for the Correlation No. 2
calculated effects of pressure and subcooling assuming The increase in critical heat flux with increase in accelera-
the variation of absolute critical heat flux with pressure tion is accompanied by an increase in pressure at the test
and subcooling for normal gravity to apply to higher section level; this in turn induces a potential subcooling.
accelerations. The effects of pressure and subcooling may be allowed
No. 4. Correlation of the corrected data to allow for for by calculation of their known effect on critical heat flux
the possible effects of pressure and subcooling by under normal gravity conditions, followed by their deduc-
extrapolation of the experimental results to conditions of tion from the experimental values at high accelerations.
atmospheric pressure and zero subcooling. The method of calculating the effect of pressure and sub-
All four correlations were of the two-variable non-linear cooling upon the critical heat flux under normal gravity
type in which use was made of a logarithmic transformation. is now presented. The effect of pressure on the critical heat
The data were first divided into a number of groups flux for water at saturation conditions was calculated by
according to the independent variable (acceleration) and equation ( 2 ) ;this is shown in Fig. 10. The effect of liquid
the mean values of critical heat flux and acceleration subcooling on the critical heat flux was complicated by the
calculated for each group. These mean values were then effect of pressure on the variation of critical heat flux with
weighted according to the number of experimental deter- liquid subcooling. Use was made of the experimental results
minations of critical heat flux for each group, and this was of Kutateladze and Schneiderman (40), who obtained values
followed by the correlation procedure. of the critical flux as a function of subcooling for water at
two pressures: 0 and 27 lb/in2 gauge. Their results are
Correlation No. 1 shown in Fig. 11. In order to obtain the critical heat flux
Correlation of uncorrected values of critical heat flux as a variation with subcooling between these two pressures,
function of acceleration is shown in Fig. 9. The equation is: interpolation was made using the Kutateladze relation for
the effect of not heating the liquid up to saturation
temperature. That is (2):
This over-estimates the effect of acceleration as the effects
of pressure and liquid subcooling are neglected.
Proc Instn Mech Engrs VoI I77 No 1 I963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAT FLUX I N POOL BOILING HEAT TRANSFER 25

MEAN C R l T l
AND MEAN

1oo 10' 10' 103


DIMENSIONLESS ACCELERATION ( a / g )

Fig.9. Non-linear correlation of the uncorrected data: correlation No. 1

The coefficient of recirculation (1 -n) was determined from variation of critical heat flux with actual subcooling under
the experimental curves and then used to predict curves for normal gravity conditions in a static pool boiler is also
the intermediate pressures. Correlation No. 2 assumes the shown. A linear mean-squares curve was fitted to the data
variation of dimensionless critical heat flux with pressure points for (a/g) values of both 72 and 126, these being the
and subcooling under normal gravity to apply to higher only two accelerations with sufficient data points and a large
accelerations ; this postulated variation is illustrated in enough dispersion on the horizontal axis to afford a reason-
Fig. 12. The second correlation (shown in Fig. 13) is: able variation of critical heat flux with potential subcool-
ing. A further value of the variation of critical heat flux
with actual subcooling was obtained at (a/g) = 1 in the
static boiler for low subcoolings. The variations are:
Correlation No. 3
(a/g) = 1, gradient 12: 5.5 w/cm* "C
This assumes the variation of absolute critical heat flux with
pressure and subcooling under normal gravity to apply to (alg) = 72, gradient 1: 4-1 w/cm2 "C
higher accelerations ;this postulated variation is illustrated (a/g) = 126, gradient N 5.7 w/cm2 "C
in Fig. 12. The third coi-relation (shown in Fig. 14) is:

(33)
Correlation No. 4
The variation of critical heat flux with potential induced THE COEFFICIENT-OF
RECIRCULATION
subcooling (with acceleration as parameter) is shown in (1-n) IS DETERMINED FROM THE
Fig. 15. No allowance is made for the slight variations in
accelerations of individual data points in each group. The
r

Fig. 10. Calculated variation of critical heat JEux with Fig. 11. Critical heat flux as a function of subcooling at
pressure for water at saturation conditions various pressures for water
Proc Znstn Mech Engrs Vol 177 No I I963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
26 H.J. IVEY

:::k
4

qcr, x lb/in2 gauge1.3 3


~ C P ,sub

qCr, 0 Ib/in2gaugo 1.2 9cr, sot


2
l.oo
1.1 5 10 15
1
0 20 40 60
PRESSURE -lb/in2 gaugo A TSub -‘C
CORRECTED CORRELATION No. 2 : ASSUMES VARIATION OF DIMENSIONLESS
CRITICAL HEAT FLUX WITH PRESSURE AND SUBCOOLING AT ( o / g ) = l TO OPERATE AT OTHER
ACCELERATIONS

1000

looo!
800

q c r , sub
-W/cm2 400

200
- 0
0 5 10 15 0 20 40 60
PRESSURE -lb/in2 gaugo AT,,b -‘C
CORRECTED CORRELATION No.3: ASSUMES VARIATION OF CRITICAL HEAT
FLUX WITH PRESSURE AND SUBCOOLING A T ( o / g ) = l TO OPERATE AT OTHER ACCELERATIONS

Fig. 12. Assumptions made in corrected correlations Nos 2 and 3

DIMENSIONLESS ACCELERATION b / g )

Fig. 13. Non-linear correlation of the corrected data: correlation No. 2

3 400
I

Y
9I 200

q
t
MEAN CRITICAL HEAT FLUX (CORRECTED FOR EFFECTS
OF PRESSURE AND SUBCOOLING) AND MEAN ACCELERATION

5 100 I
100 10’ 102 lo3
DIMENSIONLESS ACCELERATION -( o/g)

Fig. 14. Non-linear correlation of the corrected data: correlation No. 3


Proc Instn Mech Engrs Vvl177 N o I I963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
CRITICAL HEAT FLUX IN POOL BOILING HEAT TRANSFER 27

0
0

800

600

.
N

5
k
I
X
2
VARIATION OF CRITICAL HEAT
LL
-1 ,/---FLUX WITH LIQUID SUBCOOLING
AT ACCELERATION ( 0 f . g ) = 1

2I 4 0 C
-I

uk
u
/ 1 ACCELERATION AT TEST SECTION

/
/ RADIUS
b/g)

/
20c

c __ 1 - 1 . 1 40

LIQUID SUBCOOLING - OC
60 80

Fig. 15. Variation of critical heat flux with potential liquid subcooling, with accelerationas third parameter. Results for water

As these variations are similar, the mean value was towards the increase in critical heat flux as it predicts a
determined and used to predict the corrected values of variation with acceleration much less than that of correla-
critical heat flux for zero subcooling. The fourth correlation tion No. 4, in which the variation is derived directly from
(shown in Fig. 16) is: the experimental results. Correlation No. 3 predicts a
0-27 variation which is very close to that of correlation No. 4;
qcn ( a k > clearly this hypothesis is more correct.
qco (aid = 1 = *
(34)
- The exponent of (a/g)was found to be 0.25 according to
From the above four methods of correlation it is apparent correlation No. 3, although this agrees extremely well with
that correlation No. 2 is the least correct. This over- the majority of theories which predict 0.25; it is the themy,
estimates the contribution of pressure and subcooling not the hypothesis in this particular correlation, which is to

j
. T I
1
I

3
I
x 400
3
LL
I

P-
\

1 '
-1
qcr, ( a h )
__
0.273

'4-41
k-

I
200 ,
MEAN CRITICAL HEAT FLUX (CORRECTED BY EXTRAPOLATION
1 'cr' i
k
LL TO A T s u b = O ) AND MEAN ACCELERATION

100
100 10' 102
DIMENSIONLESS ACCELERATION ( a / g )

Fig. 16. Non-linear correlation of the corrected data: correlation No. 4


Proc Instn Mech Ergvc 1/01 177 N o 1 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


28 H. J. IVEY

be tested for validity. However, this does not preclude the APPENDIX
conclusion that this evidence does support the hypothesis REFERENCES
governing correlation No. 3. (I) ADDOMS, J. N. ‘Heat transfer at high rates to water boiling
With the above comments in mind, the exponent (0.27) outside cylinders’, Sc.D. thesis Chem. Eng. Mass.
determined by direct inspection of the experimental results Inst. Tcch. 1948; also MCADAMS, W. H. Heat trans-
mission 3rd ed. 1954, 384.
(correlation No. 4) may be used to confirm or reject the (2) KUTATELADZE, S. S. ‘Hydrodynamic theory of changes in
predictions of the existing theories. the boiling process under free convection conditions’, Isv.
Akad. Nauk. SSSR, Otd. Tekh. Nauk. 1951 (NO.4), 529.
CONCLUSIONS (3) STERMAN, L. S. ‘On the theory of the heat transfer from a
An appraisal of the numerous theories for critical heat flux boiling fluid’, Zhurnal Tekh. Fiziki 1953 23 (No. 2), 341 ;
D.S.I.R. Translation C.T.S. No. 62.
in saturation pool boiling has shown that a basic parameter (4)CICHELLI,M. and BONILLA,C. F. Trans. Amer. Inst.
(equation (2)) which occurs in the majority of the predic- Chem. Engrs 1945 41,755.
tions may be derived from any of three different types of (5) BORISHANSKII, V. M, Zhurnal Tekh. Fiziki 1956 26, 452.
model for the analysis. These are: dimensional analysis, (6) JAKOB,M. and LINKE,W. Physik. Zeitschr. 1935 36, 267;
vapour discontinuous/liquidcontinuous, and the liquid and also JAKOB, M. Heat transfer 1949 1, 642.
(7) DEISSLER,R. G. Columbia University Heat Transfer
vapour as discrete phases. Symposium, New York, N.Y. 1954 (see reference (10)).
The extent to which the acceleration of the boiling (8) DEISSLER,R. G. Private communication from R. Cole in
system is important as a parameter in determining the respect of reference (7).
critical heat flux in water was ascertained experimentally (9) JAKOB,M. Heat transfer 1949 1, 630 (Chapman and Hall,
with a centrifuge. A two-variable non-linear correlation London).
(10) COLE, R. ‘Photographic study of pool boiling in the region
involving a logarithmic transformation was made upon the of the critical heat flux’, J. dmer. Insf. chm. E9zgrs 1960
results after allowing for the possible effects of pressure 6 (No. a), 533.
and subcooling by extrapolation to zero subcooling. The (11) ROHSENOW, W. TID 7529 1956 (pt. 1, bk. 2), 603.
correlation is : (12)ROHSENOW, W.and GRIFFITH,P. ‘Correlation of maximum-
heat-flux data for boiling of saturated liquids’, C.E.P.
qcr, (49 = (30.27 Symp. Series 1956 52 (NO. 18), 47.
4cry (./a =1 (13) ROHSENOW, W. ‘Heat transfer with boiling’, Medd. Var-
mestromsgruppen 1956 2 (NO.4), 2.
Although the results of Costello and Adams (34) indicate (14)ROHSENOW, W. and CLARKE,J. A. ‘Study of thc mech-
an increase of critical heat flux with increase of acceleration, anism of boiling heat transfer’, Truns. Amer. SOC.mech.
the character of this increase is at variance with the Engrs 1951 (July), 609.
theoretical predictions and the present experimental results. (15) MCADAMS, W. H., KENNEL, W. E., MINDEN, G. S., RUDOLF, C.,
This is shown clearly in Fig. 8 and is discussed elsewhere PICORNELL, P. M. and DEW, J. E. ‘Heat transfer at
high rates to water with surface boiling’, Ind. Engg
(35). Chem. 1949 41, 1945.
The present experimental results are approximately in (16) GRIFFITH,P. ‘Correlation of nucleate boiling burnout data’,
accord with the theoretically predicted variation of critical Amer. Soc. mech. Engrs 1957 paper no. 57-HT-21.
heat flux with acceleration at saturation conditions in pool (17) CHANG, Y. P. and SNYDER, N. W. ‘Heat transfer in saturated
boiling as derived by Kutateladze (z), Sterman (3), boiling’, C.E.P. Symp. Series 1960 56 (No. 30), 25.
(IS] ZUBER,N. Reactor Heat Transfer Conference U.S.A.E.C.
Borishanskii (s), Deissler (7), Chang and Snyder (17)~ TID-7529 1956 (pt. 1, bk. 2), 586.
Zuber (IS), and Berenson (27), which state the exponent (IF)ZUBER,N. ‘On the stability of boiling heat transfer’, Trans.
of (u/g) to be 0.25. The results are not in accord with Amer. SOC.mech. Engrs 1958 80, 711.
the theoretical variation predicted by Addoms (I) and (20) ZUBER, N. Author’s closure Trans. Amer. SOC.mech. Engrs
1958 80,719.
Griffith (16) which state the exponent of (alg) to be 0.33. (21) ZUBER, N. and TRIBUS, M. ‘Further remarks on the stability
The results are directly at variance with the theory of of boiling heat transfer’, Dept. of Engng U.C.L.A. 1958
Rohsenow and Griffith (12) which predicts no effect at all Report no. 58-5.
of acceleration upon the critical heat flux. (22) ZIJBER, N. ‘Hydrodynamic aspects of boiling heat transfer’,
Together with experimental results of the second part A.E.C. U . 4 4 3 9 1959.
(23) ZIJBER,N., and TRIBUS, M. Written discussion C.E.P.
of the present paper, and observations (made in the f i s t Symp. Series 1960 56 (NO. 30), 35.
part) concerning limitations of several of the anzlytical (24) ZUBER,N. TRIBUS, M. and WESTWATER, J. W. ‘Hydro-
theories, it is possible to question the validity of all theories dynamic crisis in pool boiling of saturated and subcooled
predicting the critical heat flux in saturation pool boiling liquids’, Int. Heat Transfer Conf., Boulder 1961 paper no.
27, 230.
other than those due to Kutateladze (2), Sterman (3), and (25) BORISHANSKII, V. M. ‘Heat transfer to a liquid freely flowing
Borishanskii (5). over a surface heated to a temperature above the boiling
point’, Problems of Heat Transfer during a Change of
ACKNOWLEDGEMENTS State: A.E.C.-tr-3405 1953 109.
The author wishes to acknowledge gratefully the assistance (26) BERENSON, .’1 J. ‘Transition boiling heat transfer’, Amer.
given by Mr D. J. Morris and Mr R. Wright. This work SOC.mech. EngrsJ. Heat Trumfer Series C, 1961 83 (No.
3, August), 351.
has been entirely supported by the United Kingdom Atomic (27) BERENSON, P. J. ‘Transition boiling heat transfer from a
Energy Authority, with whose permission it is being horizontal surface’, Mass. Inst. Tech. 1960 Heat Transfer
published. Lab. Tech. Report 17.
Proc Instn Mech Engrs Vol 177 hi’o 1 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAI‘ FLUX I N POOL BOILING HEAT TRANSFER 29

(28) WALLIS,G. B. ‘Two-phase flow aspects of pool boiling (35) IVEY,H.J. Communication to paper no. 30, ‘Burnout heat
from a horizontal surface’, U.K.A.E.A. 1961 Report fluxes in pool boiling at high accelerations’, (34); to be
A.E.E.W.-R103. submitted, 1961 Int. Heat Transfer Conf. Discussion,
(29) WALLIS,G. B. ‘Flooding velocities for air and water in Boulder.
vertical tubes’, U.K.A.E.A. 1961 Report A.E.E.W.- (36) BERNATH,L. ‘Theory of local boiling burnout and its
K123. application to existing data’, Chem. Engng Progr. Symp.
(30) SIEGEL,R. and USISKIN,C. M. ‘Photographic study of Series 1960 56 (NO.30), 95.
boiling in the absence of gravity’, Trans. Amer. SOC.
mech. Engrs 1959 series C 81, 230. (37) MCADAMS, W.H., ADDOMS,J. N., RINALDO,P. M. and
(31) HEDGEPATH, L.M. ‘Zero gravity boiling and condensation’, DAY,R. S. ‘Heat transfer from single horizontal wires
A R S NO.1322-60 N.Y. 1960. to boiling water’, C.E.P. 1948 44 (No. 8),639.
(32) MERTE, H.and CLARK,J. A. ‘Pool boiling in an accelerating (38) BENDLER,A. J. Columbia University Task I X Reports
system’, Trans. Amer. SOC.mech. Engrs 1960 paper no. U.S.A.E.C. contract 1957, 1958, 1959.
60-HT-22.
(39) ELLION,M. E. ‘Study of the mechanism of boiling heat
(33) USISKIN, C. M. and SIEGEL,R. ‘Experimental study of transfer’, JPL-Memo-20-88 1954.
boiling in reduced and zero gravity fields’, Trans. Amer.
Soc. mech. Engrs 1960 paper no. 60-HT-10. (40) KUTATELADZE, S. S. and SCHNEIDERMAN, L. L. ‘Experi-
(34) COSTELLO, C. P. and ADAMS,J. M. ‘Burnout heat fluxes in mental study of the influence of the temperature of a
pool boiling at high accelerations’, Inr. Heat Transfer liquid on a change in the rate of boiling’, Translation
Conf., Boulder 1961 paper no. 30, 255. AEC-t~-3405 1953, 95.

Communications
Dr J. M. Adams (Sacramento, California)-The effect of surface, has been observed by the present writer”. A
heater length on the critical heat flux has not been discussed discussion by the author on his treatment of the above
by the author although Bernath (36) has shown there is a effects was omitted in the paper and would be most helpful
definite effect on the crisis for heaters less than 2 in. in in judging the credibility of his data.
length. For the experimental apparatus described with a The definite effects on the critical heat flux of an ex-
1%in. heater the critical heat flux measured could be 10 per tremely short, thin-walled, small-diameter heater, coupled
cent higher (36) if the voltage measurements are taken at with the uncertain effects of subcooling and pressure
the ends. This, of course, is due to the high relative thermal render the comparison of the data presented to any analytical
loss through the ends of a short heater. This loss could be treatment rather tenuous.
accounted for by temperature measurements across the
heater or some mathematical approximation, neither of Mr R. D. Bundy (Oak Ridge, Tennessee)-This paper
which is mentioned by the author. It would seem that presents a thorough and enlightening review of previous
because of the large relative loss of energy, and the use of theories of saturated pool-boiling burnout as well as the
absolute values of heat flux, such discussion would be results of a careful study of one mechanism affecting it.
pertinent. Experimental studies of the effect of acceleration using
It cannot be determined from the paper how the effects round tubes appear to be open to the criticism that the
of heater diameter and wall thickness were incorporated. acceleration vector is not oriented the same relative to all
Apparently, from Figs 5 and 6, and the data presented by areas of the heater surface. Burnout might be expected to
Bernath (36), such effects cannot be ignored. In addition, occur on the reverse side of the tube where the negative
it has been observed by the present writer that the critical acceleration flattens bubbles against the heater rather than
heat flux in thin-wall heaters is affected by previous high- carrying them away. While this may not be a serious criti-
temperature excursions at the burnout condition. The 25 cism of experiments determining the effects of other quan-
to 30 msec indicated by the author to pass before power tities on burnout, a study of the effect of acceleration in such
cut-off would not prevent these high-temperature excur- a system might be quite misleading. However, Adams“ has
sions. This could be the reason for the wide scatter found that such is not the case (using half cylinders and flat
(f35 per cent) in data taken with the 0-005 in. heater wall surfaces of varying orientation). Apparently, bubbles are
at u/g = 1 (Fig. 6). Usiskin and Siege1 (33) noted that after swept around the reverse sides of cylinders, and burnout
each ‘burnout’ condition, a waiting period during nucleate
boiling was necessary to insure against premature burnout. * ADAMS, J. M. ‘A study of the critical heat flux in an accelerating
pool boiling system’, 1962 NSF G-19697 (University of
This effect, due to the loss of nucleating sites on the heater Washington).
Proc Znstn Mech Engrs Vol 177 No 1 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


30 COMMUNICATIONS

occurs where bubbles are accelerated away from the heater. (2) Curvature of the heated surface.
If this is true, the use of tubular heaters can be criticized ( 3 ) Dimensions of the heated surface.
only if bubbles swept around from the reverse side compli- (4) The condition of the heated surface.
cate the bubble dynamics ‘above’ the heater. (5) The number, if any, of the heated surfaces bound-
Scatter in the data of Fig. 15 for high accelerations is ing the coolant channel.
sufficiently great that its use to correct for the effect of (6) Vibration of the heated element.
pressure and subcooling does not appear to be justified. ( 7 ) The number of heated elements.
Variation in the data is only of the same order as experi- (8) The shape of the heat flux distribution on the
mental scatter evident in all the data (see Figs 5 to 8). element.
Similarly, the third significant figure in the exponent of (9) The time-heat flux relation.
the relative acceleration terms in correlations 1 to 4 is not (10) Local variation of coolant flow across the element.
meaningful (even the second is doubtful). The principal None of the equations mentioned in the paper considers
conclusion of the study appears to be rhat qc,(u[g)/ all the factors relevant to critical heat determination.
q,,(cz/g = 1) I(u/g)’/4rather than (u/g)@2’. Perhaps the most significant factor that has been neglected
is the detail geometry of the heater surface. This means that
Mr J. C. Chicken (Associate Member)-I would first a considerable amount of caution must be exercised if the
like to congratulate Dr Ivey on the ingenious design of his results are applied to conditions other than those of the
experiment and the amount of information he was able to original experiment.
derive from it. The last feature I would like to comment on is the
Next I would like to comment on three aspects of the practical implication of burnout having been determined for
paper, first certain features of the experimental procedure, the steady state conditions. In a number of possible applica-
secondly a number of factors that should perhaps be allowed tions of the work such as in marine reactors or nuclear
for in the analysis and lastly some considerations in the rockets both the limiting acceleration and heat flux may be
practical application of the work. In other relative work (41) only transient in duration. For this reason it is worth
(42) it has been shown that the heat transfer is signi- speculating on the possible effect of transient conditions.
ficantly affected by the surface roughness and the cleanliness Cole (46)showed that for power transients of the order of
of the heat transfer surface. From this it seems logical to 30 msec duration a metal ribbon can withstand transient
expect surface conditions to have some effect on the critical heat fluxes approximately twice the steady state burnout
heat flux. Was any measurement of surface roughness made ? flux. Taking this result as indicative of the behaviour
Further it does seem from the information in the paper that expected in a short duration transient it seems reasonable
there was no cleaning of the test specimen surface between to suggest that for short transients when u/g < 1 but > 0
tests. If this is the case, the scatter of the results shown in the expected reduction in the burnout flux due to the low
Figs 5, 6 and 7 may be partly accounted for by changes to u/g condition may be offset by the increase of the burnout
the specimen surface that took place during the tests. flux due to the transient nature of the phenomenon.
Another contributory factor to the scatter of the results
may well have been the changes that took place in the REFERENCES

coolant owing to boiling off the coolant in the region of the (41) CORTY, A, and FOUST,A. S. ‘Surface variables in nucleate
boiling’, Chsm. Engng Symp. Series 1955,51 (No. 17).
specimen. The ratio of coolant volume to specimen volume (42) JAKOB, M. Heat transfer vol. 1, 1956 (Wiley, New York).
was relatively small so the effect of local boiling on the gas (43) COLLIER, J. G. ‘The problem of burnout in liquid-cooled
content of the liquid might have been of some importance. reactors’, A.E.R.E. Report R.3698.
This brings me to the last point I wish to make about the (44) CHICKEN,
J. C. ‘Burnout in liquid-cooled systems’, 2nd
experimental procedures, that is, was complete analysis of edition 1961, A.H.S.B. Memo. (S)M.29.
(45) PEXTON, A. F. ‘A review of data on burnout heat flux for
the water used in the tests made ? My reason for asking this steam-water mixtures in uniformly heated channels’,
is that it is easier to relate the experimental data to other Report 203(R)1961.
work if the precise composition of the coolant used in the (46) COLE,R. ‘Investigation of transient pool boiling due to
experiment is known. sudden large power surges’, Tech. Notes nat. adv. Comm.
Aero., Wash. 3885, December 1956.
Looking at the analysis that has been carried out on the
results it is clear that an important part of the problem is
Professor R. Cole (Potsdam, New York)-The author
to find a completely satisfactory method of determining the
is to be congratulated on his fine experimental work and
critical heat flux; this particular question has been reviewed
analysis of the experimental results. On the basis of these
in several places, in particular (43) (44)and (45).For any
results and the author’s observations concerning limitations
method of determining critical heat flux to be satisfactory
of the analytical theories, it is pointed out that all theories
for a wide range of conditions the following factors should
predicting the critical heat flux in saturation pool boiling
be taken into account:
other than those obtained by means of dimensional analysis
(1) The physical properties of the coolant including are of questionable validity.
surface tension, enthalpy, temperature, pressure velocity, Considering the chaotic situation which exists in the
and gaseous and solid contaminants. region of the critical heat flux, it seems extremely unlikely
I’roc Instn Mech Engrs Vol177 No I I963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAT FLUX IN POOL BOILING HEAT TRANSFER 31

that any physical model could adequately simulate such a


condition. On this basis alone, all analyses which rely on
some physical model of the boiling action may be considered
questionable. Yet the models of Deissler (7), Chang and
Snyder (17)~Zuber (18)-(24), and Berenson (27) do yield
results which are essentially identical to those obtained by
Kutateladze (2).
It would appear that any physical model which satisfies
the following criteria will successfully correlate saturated
critical heat flux data under normal gravity conditions :
(1) The greater portion of the heat taken from the (Data for a/g > 1 corrected to standard vapour density by multi-
heating surface is utilized in producing vapour. (Hence plying by pv/p0 at a/g = 1. Correction is no greater than 9 per cent,
however.)
the use of a form of the Jakob-Linke expression (6).)
(1 Data of Usiskin and Siegel (33).
(2) As observed from the nucleate side of the boiling A Data of Costello and Adams (34).
curve, the critical heat flux is a result of vapour binding Data from photographic study Costello, Adams and Clinton (47).
or coalcscence. (Hence the variety of bubble configura- + Data of Costello and Adams (48).
tions.) Fig. 17. Critical heat jlux versus dimensionless
(3) As observed from the transition side of the boiling accelerationfor cylindrical heaters
curve, the critical heat flux is a result of the breakdown
of vapour binding. (Hence the various stability models.) (47) and (48) are averages of over 200 burnout tests in
Whether or not a model will correctly predict the varia- which the scatter was less than plus or minus 10 per cent.
tion with gravity seems to be dependent upon the correct- (In fact, 95 per cent ofthe data from (48), about 144 points,
ness of the expressions chosen to represent such quantities agreed within 6 per cent.) It was found (48) that no appreci-
as bubble velocities and diameters. Thus while the models able effect of heater diameter on the critical heat flux existed
themselves may be of questionable validity the criteria and, as shown in Fig. 17, that the data agreed very well with
upon which they are based appear to be valid. those of Usiskin and Siegel (33).
An interesting feature of the Deissler analysis is that the Gambill (49) has compared data from reference (34)
condition necessary to obtain an equation similar to that of with results of studies of the critical heat flux in flow
Kutateladze is that the vapour density be used in the channels. His extensive work shows that the data of (34)
expression for the drag force on a freely rising vapour agree extremely well with those obtained in comparable
bubble. In reality, the force balance (Ivey’s equation (11)) experiments in which accelerations were induced by swirl
is that for a liquid droplet falling through a vapour medium. flow.
Does this imply that the vapour should be treated as the Fig. 18 shows a comparison of data from reference (34)
continuous phase? As the author points out, no analytical with those of Choi (50). Choi’s data were obtained by
model exists to date using this configuration. inducing accelerations magnetically in Freon- 113. The
ordinate and abscissa of Fig. 18 are plotted to eliminate the
effects of fluid properties; however, even if all fluid property
Professor C. P. Costello, Ph.D. (Seattle, Washington) effects are not correctly appraised, the comparison would
-The following comments are offered ty furnish more be fairly valid in view of the fact that the properties them-
information on the variation of the critical heat flux with selves do not vary with acceleration. It is noteworthy that
acceleration in pool boiling as well as to correct some false two distinct variations of qcr with a/g are observable,
impressions given by the author’s work. depending on the range of a/g, and that the slopes of the
data of (34) and (50) agree quite well.
Variation of qcr with. a/g All of the data taken by the writer (34) (47) (48) and the
The significance of the data of reference (34) has been data of (33), (49) and (51) indicate that the exponent on
obscured by the coarse scale used in Fig. 8. In Fig. 17, the a/g in equation (34) varies with the range of a/g. For u/g
data are replotted to a more reasonable scale along with values from somewhat less than one to about 10, the peak
those of Usiskin and Siegel (33) and data recently obtained heat flux is proportional to a/g to about the 0.15 power,
by the writer (47) (48). The scatter and sparseness of the while for a/g greater than 10, the proportionality indicated
data obtained by the author for the range 1 Q u/g Q 10 by equation (34) is about right. This is discussed further
are so severe as to prevent fair comparison. below.
The data of (33) were obtained with a platinum wire in
drop tests. Those from references (34), (47) and (48) were Geometry efsects
obtained with graphite and carbon cylinders, ranging from The writer has presented an analysis based on approxima-
# in. to 6’3 in. diameter, installed in a centrifuge so that tion theory to attempt to explain the different variations of
centripetal accelerations were directed normally with qcr with a/g (48). Although the geometry of the system was
respect to their axes. The data points from references (34), once thought to be the cause (34), as the author notes, later
Proc Instn Mech Engrs Vol I77 No 1 I963
2 Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
32 COMMUNICATIONS

It is not surprising that data for flat plate and cylindrical


heaters agree so well in view of the fact, reported by the
writer (47), that the critical heat flux is reached on the
detaching side of a cylindrical heater (that is, the side from
which the bubbles depart from the heater). Surprisingly,
the half of a cylindrical heater towards which acceleration
is directed (and upon which bubbles are therefore momen-
tarily held) has a higher critical heat flux than the detaching
side at the same a/g (47).
Since the detaching side of a cylindrical heater is in
virtually the same hydrodynamic position as a flat plate
oriented as shown in Fig. 19, it is clear that the data should
agree. Note that this argument also invalidates the author's
contention that the heater size caused lower critical heat
flux values in (34).
Fig. 19 also shows the theoretical predictions of Zuber
(24) and Borishanskii (5). The predictions have been
corrected for the density variations caused by the centri-
fugal acceleration employed in (48). The slopes of the cor-
rected curves representing the theoretical equations show
that qcr is proportional to (a/g)0.30,very nearly. Note that
theory the data are parallel to the theoretical curves for a/g above
10, but show a lower slope at lower a/g. The slopes of curves
evidence indicates that this is definitely not true. Data in of qcr versus a/g are given in Table 1 along with the 95
Fig. 19 were obtained with flat-plate graphite heaters per cent uncertainty interval and average deviation of data
positioned as shown in the figure. With a flat plate at a given points. The uncertainties and deviations appear highly
radius, there will be virtually no variation of acceleration favourable compared to the data of the author.
effect over the active portion of the heater. Note that the Fig. 20 shows the various orientations of flat-plate heaters
two slopes of qcr versus aig are again clearly detectable. used in collecting data for (48). No appreciable differences
in behaviour could be noted in data from configurations IV
I 1

BUBBLES
and VII nor in those from configurations I1 and 111. This
1'0 shows clearly that Coriolis acceleration does not exert an
important effect on critical heat flux, Coriolis accelerations
were oriented in an entirely different manner with respect
to the heaters in the cases cited above, yet there was no
effect of the reorientation on critical heat flux (see Fig. 21).
Some results of the flat-plate tests are shown in Fig. 21.

Table 1
Fig. 1 Slope of curve
with 95 per cent
j confidence
I interval*
1 1 Q aig Q 10
each heater,
per cent
19 0.156f0.043 0,308f0.031 5.2
21a 0.156&0.043 0.308f0.031 5.2
b 0.148 f0.046 0.329 +0.022 4.4
C 0.157f0.047 0.30910.017 3.3
d 0.155f0.041 0.284*0.018 3.7
Fluid distilled water. Data is compared with Kutateladze and -0.100 f0,034 4.7
Zuber theories, for which correction has been made for variation 0.308*0.031 5.2
of fluid properties with a/g. (See Table 1 for slopes and uncer- 0.34010.016 2.9
tainties.) 0.355 A0.016 3.2
0.424 f0.037 7.1
o Runs with heater No. 1 .
! I Runs with heater No. 2.
A Runs with heater No. 3. * All points contributing to curves uncorrected for density varia-
Fig. 19. Data for three graphite flat plate heaters with tion. Effect of variation is to make individual data points fall
higher than curves by 9 per cent maximum, but effect is
acceleration normally away from surfaces negligible for most of a/g range.
Proc Instn Mech Engrs Vol177 No I 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
CRITICAL HEAT FLUX IN POOL BOILING HEAT TRANSFER 33

DIRECTION
OF ROTATION
CONFIGURATION I: CONFIGURATION P:

ROTATION HEATER

CONFIGURATION II: CONFIGURATION Yl :

Data uncorrected for density variation. (See Table 1 for slopes and
CONFIGURATION J
D: CONFIGURATION PII. uncertainties.)
Curve Configuration (Fig. 20)
I

CONFlGURATfON IT. Fig. 2J. Critical heat Jux versus dimensionless acceleration
for g a t plate heaters shown in Fig, 20

The higher values of (qc,/q,, at a/g = 1) obtained by the


Graphite heaters employed in each case. author are entirely attributable to a more rapid increase of
Fig. 20. Flat plate orientations employed in centrifuge (48) this quantity in the range 1 Q u/g Q 10. It remains to
establish why in this range the author’s data fell higher than
Again, the data shown are averages; some 50 tests were those of (34), (47) and (48) and seem to disagree with those
taken with each conjiiguration shown in Fig. 20 and 95 per of (33) and (51).
cent of the data were within 6 per cent of the lines shown First, the container size effects and secondary flows might
(see Table 1 for slopes and uncertainty limits). Note that if be a problem. Note the favourable comparison between
any component of the acceleration was directed normally (34), (47) and (48) and tests where centrifuges were not
towards the flat-plate heater, the critical heat flux was re- employed (33) (51). Moreover, tests are reported in (48)
duced (curves (d) and (e), Fig. 21). This was not the case wherein baffles were used to eliminate any potential sec-
with circular heaters, on which the bubbles had a relatively ondary flows with no effect on the data. Thus, unaccounted
easy path of escape. In fact, with semi-cylindrical heaters flows are not factors in the tests conducted by the present
qcr was considerably higher if accelerations were directed writer. The author does not assure us on this point, however.
normally towards the heater rather than normally away Secondly, operating with as large a radius of rotation as
from it (47). the author employed presents problems. The pressure at the
The results shown are part of the more than 2000 tests heater radius, r, above that at the free surface, rfryis pro-
of critical heat flux in accelerating systems conducted at the portional to (r2-rfs2). Assuming a one inch depth of cover-
University of Washington. In all of the tests using flat age in both the author’s apparatus and that used in (48),
plates facing the axis of rotation or cylindrical heaters we the pressure buildup in the author’s apparatus is 23 per
have observed that with a/g < 10, the critical heat flux cent more than that of (48) at any a/g*. The buildup in the
varies with a/g to about the 0.15 power; while at higher a/g, devices used to obtain data for (34) and (47) was even less
equation (34) holds approximately (34) (47) (48). The latter than that of (48) at any a/g. The author contends he has
fact is clearly illustrated by Fig. 8 which shows that the corrected for the pressure buildup assuming that the effect
author’s data and those of (34) are parallel at high values of subcooling is the same as that at a/g = 1. This cannot be
of a/g. This might well have been noted by the author in reconciled with the findings of Gambill et al. that subcooling
view of the fact that the statement to the effect that an exerts different effects on qcr in different ranges of alg (51).
equation such as (34) held at high a/g was made by the Moreover, the interrelated effects of subcooling and pres-
writer a year and a half ago (34). sure in pool boiling have not been established and since the
variation of these quantities in the author’s system is
Possible sources of disagreement * The pressure buildup is even more severe by comparison with (48)
It is clear that the experimental data of the writer and those ifgreater depths of coverage are used, If the depth of coverage were
2 in., the author’s pressure buildup would be 61 per cent higher
of the author have parallel slopes at higher values of a/g. than that experienced in (48) at any value of ale.
Proc Instn Mech Engrs Voll77 No I 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
34 COMMUNICATIONS

possibly substantial any correction to account for these which indicate that the analytic correlations do not ade-
parameters is a guess, at best. quately represent the critical heat flux problem, the follow-
Carrying the water at different levels to alter subcooling ing data are provided. Fig. 22 shows data obtained on a
is not in itself an absolute test of subcooling effects because flat-plate heater facing the axis of rotation as shown sche-
while this changes the potential subcooling in the system, it matically in Fig. 19. The surface tension was varied by
does not establish how the subcooling at the heater location adding no more than 0.1 per cent by weight of a wetting
is changed. I n reference (34) it is stated that pool water agent to distilled water (48). Fluid properties (vapour and
temperature as a function of r was determined by thermo- liquid densities, latent heat) were found to be unchanged
couple measurements. It was found that since the water to within 1 per cent by the addition of the wetting agent.
supply the heater flows countercurrent to the vapour stream, There is a great change in the exponent of a/g in the
the liquid is warmed to the saturation temperature at the equation qcrOc.(a/g)ndue to the change in surface tension.
heater location, very nearly. (The use of large heaters which Analytic correlations predict neither this effect nor the
produced substantial vapour was helpful in this respect.) pronounced effect of surface tension on qcr at a/g = 1. (It
Thus, in the devices used in (34) (47) (48) no appreciable was found that the latter effect was a function of the nature
effects of subcooling and pressure were present. This is of the heater surface.)
also demonstrated by Fig. 17; because the heaters employed
in (34) (47) and (48) were all at different radii, the potential
subcooling at a/g = 44 was 3.78"F for data from (34), Conclusions
6.2"F for data from (48)*. If these subcoolings actually ex-
isted at the heater, data obtained in the different containers
(1) In the range 0.06 a/g < <
10 it appears that
q c r ~ ( a / g ) @ 1(33)
5 (34) (47) (48) for cylindrical or flat-plate
should be quite different. Yet, note the good agreement heaters in distilled water with acceleration vectors directed
of data from (34), (47) and (48) shown in Fig. 17; the slight nearly normally away from them. It is possible that the
departures can be attributed to use of heaters with slightly nature of the surface may alter the exponent somewhat.
different surface conditions. (2) For a/g > 10, equation (34) holds approximately
It is not certain whether the pressure and subcooling
effects in the author's system can be as easily dismissed, (present author) (34) (47) (48).
(3) The opinion that qcr is differently affected by a/g
and it is probable that they cannot be corrected for in the in different ranges of a/g has been substantiated by choi
manner he suggests. However, it is not the purpose of the (50) in a system using Freon-I 13. Steady-state accelerations
writer to disparage the author's work. Indeed, the author were obtained magnetically.
has cast enough disparagement in this paper to last most (4) Analytic correlations do not correctly represent the
readers a lifetime. Possibly the points already raised can be behaviour of qcr as it is affected by a/g or by surface tension
answered and we must look for a third possible reason that
<
the data disagree in the range 1 a/g 10. < (present author) (34) (47) (48).
The heater surface offers a possible source of disagree-
ment. Experiments conducted at the University of Washing-
ton indicate that the surface is a most important factor in
determining critical heat flux, contrary to some analytic
studies. The surface effect might actually be enhanced by
increasing a/g, at least for the low range of a/g, since the
nature of the surface will dictate in large part the number of
bubble sites. The fact that platinum heaters (33), (50) yield
data which fall in line with those of carbon and graphite
heaters (34), (47) and (48) rather refutes this argument,
but does not do so completely. The disagreement between
the author's data and those of the references cited in the
previous sentence might be due to a surface effect.
In this connection, the use of acetone, as used by the
author to clean the test section, is a possible factor; it is Graphite heater, oriented as shown in Fig. 19. Data uncorrected
for density variations (see Table I for slopes and uncertainties).
indicated in (48) and below that fluids tending to reduce
the local surface tension have the effect of making the Surface tension, dynelcm
71.45
exponents on a/g, equation (34), higher than those obtained 65.02
with pure water. 51.55
44.08
39.21
32.08
Evidence on inadequacy of analytic correlations 27.55
To supplement the arguments presented by the author -
* These potential subcoolings were obtained with the CentrifUge full so Fig. 22. Critical heat flux versus a/g with surface tension
that the free surface was about at the axis of rotation. as parameter
Proc Instn Mech Engrs Vol177 No I 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAT FLUX IN POOL BOILING HEAT TRANSFER 35

REFERENCES in an effort to determine the scatter inherent in saturated


C. P.,ADAMS,J. M. and CLINTON,
(47) COSTELLO, W. W. ‘Im- pool boiling burnout, we have sometimes observed a
provement of burnout heat fluxes by orientation of semi- directed upward or downward trend of the critical fluxes,
circular heaters’, J . Amer. Znst. chem. Engrs 1962 8
(No. 4, September), 569. apparently related to slight changes in the surface condition
(48) COSTELLO, C. P. and ADAMS,J. M. ‘Interrelation of geo- of the heater tube, which in turn indicates, of course, that
metry, orientation and acceleration in the peak heat flux the burnout phenomenon is not determined by solely
problem’ (submitted to Amer. Znst. cketn. Engrs, August, hydrodynamic criteria.
1962).
(49) GAMBILL, W. R. and BUNDY,R. D. ‘Evaluation of the pre- I n Fig. 7, the dependence of qcr on ATsubis not linear
sent status of swirl-flow heat transfer’, Amer. Soc. mech. as in most other studies, or as predicted by the subcooling-
Engrs paper no. 62-HT42, Fifth Nut. Heat Transfer factor correlation of Ivey and Morris (57). Would the author
Conf., Houston, Texas, August, 1962. care to comment on this rather marked non-linearity of his
(50) CHOI, N. Y. Ph.D. Thesis, Massachusetts Institute of
Technology, 1962. subcooled burnout data ?
(51) GAMBILL, W. R., BUNDY,R. D. and WANSBOROUGH, R. W. Finally I wish to point out that there are several un-
‘Heat transfer, burnout and pressure drop for water in certainties involved in the development of the author’s
swirl flow through tubes with internal twisted tapes’, correlation No. 4 (mainly assessment of the qcr-dTsub
Oak Ridge Nut. Lab. Report 2911, 28th March 1960;
also Chem. Engng Progr. Symp., Series no. 32 1961 57,
gradients) and that the dependence of qcr on the fourth
€27. root of a/g is probably as valid as the author’s exponent of
0.27, especially in view of semi-theoretical basis.
Mr W. R. Gambill (Oak Ridge, Tennessee)-The REFERENCES
variation of the critical heat flux in boiling heat transfer (52) NOYES,R. C. Experimental study of sodium pool boiling heat
with system acceleration is of considerable importance not transfer NAA-SR-6769, 30th March 1962.
only in a fundamental sense but also in space-vehicle (53) BRAGG,S. L. and SMITH, I. E. ‘Dimensional analysis of
burnout heat transfer’, Znt. J. Heat Mass Transfer 1961
nuclear reactors operating in a reduced-gravity environ- 3 (No. 3, October), 252.
ment. This interesting paper describes a carefully con- (54) MARKELS, M. jun. et al. U.S.A.E.C. Report NYO-9500,
ducted experimental study whereby the fourth-root depen- 30th September 1960.
dence predicted by the Kutateladze-Zuber equations is (55)ROHSENOW, W. M.and CHOI, H. Y. Heat, mass and rnomen-
tum transfer 1961, chapter 9, 228 (Prentice-Hall).
substantially confirmed, and thereby makes a considerable (56) STOCK, B. J. Observations on transition boiling heat transfer
contribution. phenomena ANL-6175, June 1960.
I should like first to mention briefly several uncited (57) IVEY,H.J. and MORRIS, D. J. On the relevance of the vapour-
studies which are pertinent to the author’s description of liquid exchange mechanismfor subcooled boiling heat transfer
present theories of critical heat flux for saturated pool at high pressure AEEW-R 137, January 1962.
boiling. Addoms’s correlation (equation (1)) has been re-
cently modified by Noyes (52) so as to correlate his low- Mr R. W. Graham and Mr Y. Y. Hsu (Cleveland,
pressure critical-heat-flux data for sodium with other data Ohio)-The author is to be congratulated for his efforts in
for water and four organic liquids. Noyes introduces both assembling in one paper the principle contributions to the
surface tension and viscosity into Addoms’s equation and critical heat flux problem. He has done a good job in point-
changes the exponent of the group containing g from 3 to a. ing out the salient features of each analysis and likewise for
Deissler’s criterion for the critical-flux condition (follow- the experimental work in the literature, the limitations of
ing equation (8) of the paper) has been used more recently the experimental approaches.
by Bragg and Smith (53), who obtained equation (2). However, it is difficult to understand how the author
It may also be mentioned that at the critical heat flux, the gained enough confidence through his own experiments to
fraction a of the surface covered by bubbles appears to be be so conclusive in estimating the value of the contributions
-0.5, as indicated by the experimental studies of both appearing in the literature. In order to be truly conclusive,
Ellion (39) and of Markels and co-workers (54). the author would have to develop an experiment which was
Equation (17) of the paper has since been modified by more definite in terms of the analytical models he is
Rohsenow and Choi (55) by the essentially empirical criticizing. Actually, with the exception of two of the refer-
introduction of the term (a/g)U4.Finally, the extensivestudy ences, the acceleration effect on burnout flux is predicted
by Stock (56) substantiates the conclusion of Borishanskii to be essentially the same. This, despite the differences in
and of Berenson that liquid-solid contact exists in transition the models, is a remarkable result. If the objective of Dr
boiling, which presumably invalidates, as mentioned by the Ivey’s experiment was to evaluate these models in terms of
author, the assumption of Taylor instability. their mechanisms and assumption, the test section would
The author states that use of a burnout detector made it have to correspond more closely to the flat-plate geometry
possible to carry out a number of burnout runs on one test assumed in most of the analyses. Starting with a geometry
specimen. Would he state the largest number of tests that is more consonant with the analytical models, he could
conducted with a single specimen and indicate whether devise his experiment such that particular assumptions
any non-random change of critical flux with number of included in the analyses could be tested. He did a good job
tests ever occurred ? In similar tests which we are running of summarizing the analyses appearing in the literature but
Proc Znstn Mech Engrs Vol I77 No I I963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
36 COMMUNICATIONS

he failed to develop an experiment which tested these two-phase flow regime transition and as such one would
models in a definitive fashion. expect geometry to play a role so that one should not expect
The test section geometry he selected can also be criticized the burnout heat flux or the power on gravity to be well
for certain undesirable characteristics that exist at high predicted from a plane theory for a ‘horizontal’tube.
accelerations. With a tube geometry, the magnitude and As the author mentions, the very high gravity fields give
direction of the acceleration vector is a function of the cir- rise to other effects too. One of these is the direct conduc-
cumferential location. In fact, sections of the tube surface tion into body of the fluid arising from the strong variation
experience an acceleration vector that points towards the in hydrostatic pressure over a small distance. All the
surface, while other sections experience a vector that points burnout correlations are based on replacing the heat flux
away from the surface. These extreme conditions greatly by two counter-current streams of liquid in and vapour out.
influence the ebullition phenomenon. The assumption implicit in this is, the heat goes entirely
It should be observed also that there will be an appreci- into the formation of vapour in the immediate vicinity of
able convection flow in the tank as a result of the pressure the surface. If the hydrostatic equilibrium temperature
gradient induced by the acceleration. This flow would pro- gradient is sufficiently large, this direct conduction into the
ceed over and around the test section and mix with the body of fluid can be substantial. The same factor is ap-
vapour coming off the tube. This effect might be envisioned parently important with boiling liquid metals too at one
as a ‘chimney’ that sweeps the vapour towards the centre gravity, as their conductivity is so high.
of rotation of the centrifuge. Undoubtedly this convection
also influences the ebullition process. Mr P. M. C. Lacey (Harwel1)-The author’s survey of
These two aforementioned effects also make it difficult the variables affecting burnout under high acceleration does
to reconcile Dr Ivey’s experimental results with the analy- not appear to take account of the probably large influence
tical models of the literature. of u/g on the flow pattern at low values of subcooling. At
The author’s use of Fig. 15 is to be questioned. He arrives high values of u/g there must be severe stratification of the
at the conclusion that the slope of critical heat flux with liquid, with the hottest at the ‘top’, nearer the axis; this
liquid subcooling for various accelerations is similar. will be disturbed by the high convection effects above and
The scatter of the data make a slope estimation around the heater. But when the subcooling near the heater
very difficult. The data he ignores for accelerations other becomes small or zero, the pressure gradient in theliquid
than 126 and 72 g could be interpreted as showing that will lead to a reversal of the temperature gradienr, since the
acceleration greatly affects the slope. The effect of accelera- temperature will tend towards the local saturation value, at
tion on the burnout at various subcoolings warrants further any rate in the vicinity of the ascending bubbles. Three
study. What is more, the author used his conclusion from inches of water at a/g = 160 represents a pressure difference
Fig. 15 to judge the correlation approaches. of the order of 15 lb/inZ, and a temperature difference of
Perhaps the whole question of subcooling could have over 20°C. Such a condition will be completely unstable
been obviated by controlling the bulk temperature or and lead to much more severe convection effects which
pressure in the tank to compensate for the ‘head’ effects at must influence burnout in a way that is not matched in pool
the test section surface. In this way, regardless of the boiling at low values of a/g, and it must be highly dependent
acceleration, saturation or near saturation, conditions at on the depth of the water and the general geometry of the
the heater surface could have been maintained for all the system.
tests. Perhaps, therefore, one might in part attribute to such
On p. 27, there is a sentence that reads, ‘CorrelationNo. 3 an effect the excess of the experimental exponent of a/g
predicts a variation which is very close to that of correlation above the theoretical value of 0-25, especially in the fourth
No. 4; clearly, this hypothesis is more correct.’ What is correlation which depends upon extrapolation to zero
the antecedent of ‘this hypothesis’ ? subcooling. It would be interesting to learn whether the
In the abstract and conclusions, it should be made clear burnout figures were sensitive to liquid depth at low sub-
that the data refer to water. It might be well to indicate cooling. It would be valuable, though obviously very
this on all of the figures. Some of the figures do indicate difficult, to obtain flow pattern data in future experimenta-
the fluid. tion of this kind.

Professor I?. W. McFadden, Ph.D. (Lafayette,


Professor P. Griffith (Cambridge, Massachusetts)- Indiana)-I would like to make a comment on the vapour
I think this is a fine paper and the author is to be commended discontinuous, liquid continuous analytical models
on the care and thoughtfulness that went into the experi- discussed in this paper. Recent measurements and analysis*
ments that he reports. I should, however, like to comment have indicated f2Db is proportional to the acceleration
on the comparison of these experiments with existing burn-
out correlations. * MCFADDEN P. W. and GRASSMANN, P. ‘Relation between bubble
frequency and diameter during nucleate pool boiling’, Int. J .
All the burnout theories which have a tangible physical Heat Mass Transfer 5, 169; and COLE,R. ‘A photographic
basis, (z), (24) and (2a), treat the case of a plane surface in study of pool boiling in the region of the critical heat flux’,
Preprint 21, Fourth Nat. Heat Transfer Conf., Bufalo, New
pool boiling. The burnout process is a manifestation of York, 1960.
Proc Znstn Mech Engrs Val 177 No 1 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
CRITICAL HEAT FLUX IN POOL BOILING HEAT TRANSFER 37

acting on the individual bubbles at the instant of departure I note first that the derivations of Addoms (I), Deisler
from the heating surface, and therefore that (using equation (7), Rohsenow (11), Rohsenow and Griffith (12) and
(12)) f o b is proportional to g* under standard gravity. Griffith (16)were done apparently without awareness of
Thus fDa is not a constant if the gravity (i.e. body force) the existence of Kutateladze’s very successful empirical
field is changed. This allows one to-start with the equation equation. The end results are relations significantly dif-
ferent from Kutateladze’s equation and from each other
q = Xp, Db3fn and therefore easily tested for how well they fit the existing
experimental data. The work of Chang and Snyder (17),
and show that p g * (or in general qccat) Zuber (18)-(24), and Berenson (27)are attempts to arrive
without the necessity of introducing a terminal velocity at Kutateladze’s equation analytically and thereby develop
(ub)to describe what is happening to the bubbles at the a theory. The end results are relations which are not,
surface, a doubtful growth velocity (Vgrowth), or some of course, significantly different from Kutateladze’s
dubious model. equation qr from each other-except in regions near the
critical pressure for the fluid where extensive experimental
Dr T.D. Patten (Associate Member)-The first part of data are lacking. Tests of the theories by experiment must:
this paper presents a good summary of major existing then focus on areas where the equations differ significantly
theories on critical heat flux in saturated pool boiling. The or concentrate on measurements of specific details of the
new experimental information is welcome and presented in theories.
this way it proves very useful for examining a particular Tests of this latter type are reported by Hosler and West-
aspect of these theories. water* as well as Berenson (26), and suggest that current
I have considerable dficulty in assessing the relative theories are not consistently accurate with respect to many
merits of correlation No. 3 and correlation No. 4. The details. Such findings are in accord with Dr Ivey’s inter-
validity of the correction to zero subcooling for correlation pretations of his critical examination of these current
No. 4 presumably results from an experimental condition theories. Most of these interpretations are based, for
(not stated) of constant bulk liquid temperature in all example, on noting that the neat geometrical physical
acceleration tests. If this is confirmed, will the author models postulated as bases for theory are not in accord with
please state if the value of 5-5 w/cm2 “C for the variation of the chaotic behaviour actually observed experimentally. Un-
critical heat flux with subcooling at u/g = I was also derived fortunately, such critical interpretation cannot be definitive.
from tests at constant bulk liquid temperature and varying The very chaotic behaviour of boiling allows for alternate
liquid level above the heating surface ? The assumption of interpretations that regular patterns of bubble formation,
a mean value, for all values of aig, of the variation of critical lack of liquid-heated surface contact, mostly vapour at
heat flux with subcooling appears to be open to question, heated surface, etc., are abstractions of the essential
and if this assumption lacks experimental support then mechanisms which actually control behaviour, but are
correlation No. 4 has no obvious advantage over correlation difficult or even impossible to observe experimentally.
No. 3. There is, however, another area which requires examina-
In agreeing with Dr Ivey that it is possible to question tion and is capable of more definitive criticism. We can
the validity of most models used in predicting critical heat accept, at least momentarily, any reasonable physical
flux, I nevertheless remember the important part played model, recognizing that because of the complicated nature
by many of the resulting theories in helping us towards an of boiling in general a great deal of idealization will be
understanding of the complex processes of nucleate boiling. necessary. We then must satisfy ourselves that well-founded
The author’s contribution is a farther step forward. applicable physical principles have been applied to the
physical model correctly and consistently. It is in this area
of critical examination that I should have preferred the
Mr R. I?. Stein (Argonne, Illinois)-I have read Dr author to delve into in more detail, for I believe that there
Ivey’s paper with enjoyment, and commend him for pre- are several incorrect applications in the analyses of refer-
senting an interesting research effort. The first part of the ences (17)-(24) and (27) that make arguments about the
paper, which compares and evaluates the bases for the reasonableness of the physical models upon which they are
development of various saturated pool boiling critical heat based of secondary importance at this time.
flux equations, represents a kind of effort badly needed in As an illustration I mention a particular result of an
this field. The second part of the paper, which describes analysis of a problem described in the literature of classical
carefully performed experiments aimed specifically at hydrodynamics which I believe has been incorrectly applied
determining the effect of uniform acceleration on the in all of the above references, although in each somewhat
critical heat flux for comparison with predictions of the differently. The problem of classical hydrodynamics to
equations, contributes a useful set of experimental data. I which I refer concerns the infinitesimal sinusoidal oscilla-
have comments concerning what is perhaps only my own tions of a horizontal interface between two inviscid im-
personal preference of a method of approach towards the miscible fluids of different density, each fluid having a
substance of the first part of the paper, as well as some
comments on data interpretation for the second part.
* HOSLER,A. E. R. and WESTWATER, J. W. Film boiling on a
horizontal plate, ARS 3. 1962 32 (No.4), 553.
Proc Insrn Much Engrs V o l I77 No I 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
38 COMMUNICATIONS

parallel uniform and horizontal steady velocity*. I n refer- heated surface than at the surface of the pool. As I under-
ence (17)~ a result of the classical solution to this problem stand the calculation procedure used, induced subcooling
is first cited for predicting the relative velocities of the two was computed by subtracting the normal boiling tempera-
streams at which the interface becomes unstable (see ture from the boiling temperature based on the local pres-
equation (31) of reference (17)). The fact that the equation sure at the test-section heated surface. This does not seem
cited is specifically for the case of the heavier fluid under- to me to be a valid measure of subcooling for comparison
neath the lighter one is not mentioned in reference (17). with subcooled pool boiling experiments at normal gravity.
The equation is then applied to fluids flowing vertically by With the latter, vapour bubbles that rise into subcooled
dividing by 1 / 2 without explanation or reference thereby regions will tend to condense by exchanging heat with sur-
obtaining equation (24) of Dr Ivey’s paper. I n references roundings at temperatures significantly below the local
(22) and (27) the same result of the classical solution to boiling point. Also, the pressure within a vapour bubble
the problem is used, but application to vertical flow is will not be too different from that of the surroundings.
made by simply equating the acceleration due to gravity in With the high-acceleration saturated pool boiling experi-
the classical result to zero. Not only do I suggest that some ments described in this paper and the resulting large pressure
demonstration of the validity of these manipulations is gradients within the liquid, the situation would seem to be
required, but I also believe that the relations obtained as a significantly different.
result of their use incorrectly represent the idealized physical I also suggest that further explanation is needed of how
model described. it was decided which of the four attempted correlations of
I am concerned about Dr Ivey’s interpretation of induced the effect of acceleration was best. Incidentally, I note from
subcooling for his experimental data. He notes that at high Fig. 15 of the paper that the effect of acceleration on the
accelerationsthe effectivelylarger specificweight of the liquid critical heat flux could possibly be accounted for solely as a
causes a significantlyhigher local pressure at the test-section function of the calculated induced subcooling, and that such
* LAMB,B. H. Hydrodynamics 1932 sixth edition (Art. 268), 461 a correlation could be well within the normal scatter of the
(Dover, New York). experimental data.

Author’s Reply
Dr H.J. Ivey-I wish to thank sincerely all the contribu- be 0.5 according to Ellion (39) for forced convection
N

tors to the discussion. Their comments have all been most boiling and also to Markels (54). A similar figure has been
helpful, stimulating and informative. It is clear that many determined experimentally by Donald and Haslam (58)
aspects of boiling heat transfer are still controversial, and for pool boiling at the critical heat flux.
that the particular case of pool boiling heat transfer is no I agree with Mr Stein’s comments concerning the
less controversial than others. As the principal objective question of the application of well-founded physical
of the paper was an attempt to appraise critically and principles to the various physical models of the critical
experimentally current analytical theories and corrclations heat flux. There does seem to be little value in proposing
for the critical heat flux in saturation pool boiling, the a specific model of the critical heat flux condition if well-
theoretical aspects will be dealt with first, followed by the founded physical principles have to be incorrectly applied
experimental aspects. in order to obtain a ‘reasonable’ answer. I have not dealt
Mr Gambill has added the studies of Noyes (52)and of with this aspect of the theories, as an excellent example
Bragg and Smith (53) to the list of theories for critical of this type of criticism has already been published
heat flux in saturation pool boiling. The disposition in (59).
space of vapour bubbles in the analytical model of Bragg The ‘remarkable result’ noted by Mr Graham and Mr
and Smith is remarkable in that the bubbles have to Hsu that the acceleration effect on qcr is predicted to be
satisfy simultaneously a horizontal critical bubble packing the same for most theories, in spite of wide differences
configuration similar to that of Rohsenow and Griffith (12) in the analytical models, is partially explained in the
and a vertical critical bubble touching configuration contribution of Professor Cole. Professor Cole has made
similar to that of Deissler (7). The plausibility of such an the interesting observation that provided ‘valid’ bubble
exact model is open to doubt when applied to the chaotic velocities and diameters are chosen, any given model will
behaviour of steam and water at the critical heat flux. probably predict the quarter-power law variation of qCr
There are some indications of specific behaviour at the with acceleration. This suggests that the role of exact
critical heat flux; Mr Gambill has mentioned that the physical model is of questionable importance in the
fraction a of the surface covered by bubbles appears to predictions of qcr. From this, it is a short step to ask
Proc Instn Mech Engrs Vol177 No I 1963
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
CRITICAL HEAT FLUX IN POOL BOILING HEAT TRANSFER 39

whether there is any merit in developing the exact Mr Hsu mentioned the possibility of designing experiments
physical model (7)(12)(17) (19)(27), as opposed to a less which could test the particular assumptions in each separate
rigorous general physical picture of boiling at qcr (2)(3) analysis. While some useful information may be derived
( 5 ) (48). from studies of this type, it would appear to be more
Professor McFadden has found in nucleate pool boiling rewarding to investigate thc effects of various parameters
(60)thatf2D, cc g, and by using equation (12) has derived (acceleration, surface tension, liquid, etc.) on qcn and
the result that fDbcc g*. By using equation (7) he has compare these results with the predictions of the
also shown that q oc g*. However, it should be noted that theories.
a term (pV/pl): additional to the Kutateladze expression Several contributors have mentioned other factors
(equation (2)) is obtained when using this procedure. Care which may or may not have an effect on qcn and which
must be taken in extrapolating to the critical heat flux a should therefore be taken into consideration in any analyti-
correlation obtained under nucleate boiling conditions cal description. Mr Chicken has kindly listed many of
where individual bubbles are distinguishable, compared these factors; others have been given by Bernath (36),
with the chaotic mixture of vapour and water which Hawkins (63) and Mueller (64). Most of the points in
occurred at the critical heat flux. Mr Chicken’s list are relevant to natural convection pool
Dr Adams, Mr Graham and Mr Hsu, and Professor boiling, but several could only apply to the case of forced
Griffith have all raised the question of experimental design convection boiling in channels. The references (41)(42)
in relation to the analytical models. In particular they which Mr Chicken quoted dealt with surface roughness
state that to simulate basic assumptions of the theories effects on nucleate boiling, not qcr as in the present paper.
more accurately a flat-plate heater geometry should be Surface roughness has an effect on nucleate boiling for
used in the experiments, I agree with the vicw that ideally pentane (26) and for water (41) (42) (65). However,
the test section should correspond as closely as possible references (26)(65) dealt with qcrin addition (for pentane
with the assumptions made in the theories, and for this and water) and no effect of surface roughness on qcr was
reason feel that the entirely new experimental results evident. Because of this no measurement of surface
derived from flat-plate heaters presented by Professor roughness is made in the present experiments. It must be
Costello in the communication represent a step forward mentioned that both Professor Costello and Mr Gambill
in this direction. Other advantages as well are obtained in their communications have noted some effect of surface
using flat-plate heaters, for example Professor Costello condition on qcr, but do not state to what extent. Possibly
has managed to isolate the potential effect of Coriolis the effects noted are of small magnitude.
acceleration (which effect appears to be negligible) by Many contributors have raised queries concerning
different orientations of the flat surface with respect to various aspects of the present experiments.
the axis of rotation. I n my own experiments a ‘horizontal’ Dr Adams states that a heater of 1: in. length could
tube had been chosen primarily for the sake of experimental give a measured qcr which is 10 per cent higher than that
simplicity. Several flat strip heaters have been tested with for ‘long’ heaters, according to the data presented by
the underneath and sides insulated by silicone rubber and Bernath (36) for vertical heaters. However, the present
araldite, but have been rejected. The adhesive expanded experiments refer to horizontal heaters which are tan-
slightly on being overheated in a qcrtemperature excursion gential to and in the same plane as the circle swept by the
which left a small gap between metal and adhesive. This rotating heater. Perhaps Dr Adams should have applied
acted as a preferential nucleation site, rendering the heater this correction to the experiments described by Professor
unusable. The size of the present cylindrical heater is Costello in which vertical heaters were used (34) (48).
limited by the current carrying capacity of the silver slip Since Bernath has shown that long vertical heaters give a
ring unit. However, as Mr Bundy points out, qcTappears qcr which is only 0.76 that of an equivalent horizontal
to occur on that side of the cylinder where bubbles are heater, this may be a far greater source of inaccuracy in
accelerated away from the heater. In this respect cylinders Dr Adams’s experiments. The purpose of Figs 5 and 6
behave as flat plates. This is not too surprising a result, as was to demonstrate that the heater diameter and wall
this has been known for some time at alg = 1. Under thickness chosen gave qcr values at a/g = 1 which were
normal gravity conditions the majority of qcr experiments satisfactory when compared with heaters of larger diameter
in boiling from horizontal wires and cylinders give values and wall thickness. Thus no allowance has been made for
of the constant in the Kutateladze equation (2) very close effects due to heater diameter and wall thiskness in the
to the values of the constant obtained in boiling from %at present experiments.
plates and discs. This applies for heaters boiling both to Mr Gambill has inquired into the number of burnout
water and to organic liquids; tables of comparative values runs attained on one test specimen using a burnout
of this constant for various systems are available (61)(62). detector in the centrifugal experiments. The largest
The above comments with regard to horizontal wires are number of tests conducted with one specimen was 21,
subject, of course, to the size limitations noted by Bernath while the average was lower at 5.5 per specimen. Each
(36). The use of the present size of cylinder minimized specimen was used for several different values of accelera-
the effect of variation of acceleration from inner to outer tion, thus it was not possible to furnish any conclusive
surfaces, compared with larger cylinders. Mr Graham and evidence on the effect observed by Mr Gambill of the
R o c Instn Mech Engrs Vol177 No I 1963
3
Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016
40 AUTHOR’S REPLY

slight variation of qcl with number of tests on one speci- (for a pool boiling system), this has been assumed to be
men. Before each burnout test a waiting period of approxi- similar to that for a/g = 1. In reply to Dr Patten’s query,
mately 15 minutes was allowed during which the specimen the qcr variation with subcooling at a/g = 1 was obtained
was vigorously boiled. As mentioned by Dr Adams, in a separate static pool with constant liquid level above
this allowed re-establishment of any loss of nucleating the heating surface and different bulk liquid temperatures
sites due to a previous burnout excursion. Mr Chicken to obtain different bulk subcoolings at the heating
mentions the scatter of the results shown in Figs 5, 6 and surface level. As Professor Costello has pointed out, dif-
7, and notes that this may be accounted for partly by ferent water levels above the test section in the centrifuge
changes in the specimen surface during the tests. The may alter the potential subcooling at the heater surface
results shown in these figures were obtained in a separate level, but will not necessarily alter the actual subcooling
static pool, and from a number of test specimens. Although at that level. I n view of the above remarks I agree with
each specimen had been cleaned carefully once before the Mr Graham and Mr Hsu that the effect of subcooling on
experiments, no subsequent cleaning was thought necessary qcrat various accelerations warrants further detailed study.
in the light of the results of Averin (65) and Berenson (26). On p. 33 Professor Costello has stated that Gambill et al.
This is partially substantiated in that the scatter of results (51)have found that subcooling exerts different effects on
for any one specimen is less than the scatter for all speci- qcr in different ranges of a/g. I would not disagree with
mens. As previously noted by Mr Gambill, however, small this, but it should be pointed out that Gambill found ‘the
effects of surface condition on qcr may exist. I n reply to influence of pressure and subcooling on qcr is negligible’
Mr Chicken, no complete chemical analysis of the water in the range 750 < at/g < 11 500; but for at/g = 35 ‘the
was made. It was supplied as demineralized water, the subcooling effect begins to appear’. The a, denotes tan-
electrical conductivity and gas content of which were gential acceleration in the swirl flow.
checked before use. Professor Costello asserts that the interrelated effects
Almost all the contributors have raised queries concern- of subcooling and pressure in pool boiling have not been
ing the interpretation of the results and the four suggested established. I would suggest consulting the excellent
different methods of correlation. It should be emphasized paper by Kutateladze and Schneiderman (4Q), which gives
that for the present experiments correlation no. 1 (the the variations of qcr with subcooling for three different
uncorrected data) may be considered as an upper limit of liquids at five different pressures for the case of a/g = 1.
qcr as a function of a/g. Similarly correlation no. 4 (data Mr Gambill has noted that in Fig. 7 the dependence of
corrected by extrapolation to zero potential s ubcooling) qcr on AT,,, is not linear as in ‘most other studies’. For
may be considered as a possible lower limit of qcr as a the case of static subcooled pool boiling the qcr on AT,,,
function of a/g. The two other correlations nos 2 and 3 curve is linear according to results of Kutateladze and
were included merely as demonstrations of possible Schneiderman (40) and Bendler (36) (38) (upon which
methods of calculating secondary effects of subcooling the correlation of reference (57) was based), and of
and pressure on qcr. Thus correlations nos 2 and 3 were Mirshak (66). The first work used horizontal graphite
added as possible guides for future studies of the interrela- heaters of unstated diameter, the second used vertical
tion of subcooling, pressure and acceleration on qcr, and nickel heaters 0-187 in. diameter, and the third used
should not be used as strict interpretations of the present vertical stainless steel heaters 0-75in. and 1.00in. diameter.
results. I agree with Mr Bundy and Mr Gambill that the Contrary to this, it would appear that when the heater is
third significant figure on the exponent of relative accelera- comparatively small the qcr on AT,,, curve is not linear
tion terms in the correlations has little meaning, and it has but concave upwards. This is demonstrated by the results
therefore been dropped. on small heaters of Ellion (39), Gunther and Kreith (67)>
Mr Bundy, Mr Graham and Mr Hsu have queried the Kutateladze (61, p. ZOS), and the present results shown in
use of Fig. 15 in which a correction for the possible effect Fig. 7. Since the potential induced subcooling did not
of potential subcooling on qcr is obtained. The values of exceed 20°C at the heater level in the present experiments,
potential subcooling plotted in Fig. 15 were calculated by the qcr on d Tsubcurve for a/g = 1 in Fig. 7 was approxi-
subtracting the normal boiling temperature at the free mated by a straight line to 20°C subcooling.
liquid surface from the boiling temperature derived from Mr Stein is correct in his statement that according to
the local pressure at the heater surface. I would agree with Fig. 15 the effect of acceleration on qcr could possibly be
Dr Patten and Mr Stein that the situation in the large accounted for solely as a function of calculated potential
pressure gradients of the centrifugal experiment described subcooling, in the sense that the Calculated value was a
here are probably significantly different from the situation monotonic function of acceleration for a given liquid
arising in a static (a/g = 1) subcooled pool boiling experi- depth and radius of rotation. Fig. 15 shows that the re-
ment. Thus correlation no. 1 assumes that at any given corded values of qcrfor a/g > 1lie above the curve obtained
acceleration qcl is a constant for increasing subcooling. for a/g = 1. Therefore the difference at least may be
Correlation no. 4 assumes that at any given acceleration attributable to an effect produced by the increased accelera-
qcrincreases linearly with increase in subcooling, and owing tion in this particular system, and not the calculated
to lack of experimental information of the variation of qc,. potential subcooling.
with subcooling for any acceleration other than a/g = 1 Several other factors which may be of importance have
Proc Instn Mech Engrs Vo1177 No 1 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


CRITICAL HEAT FLUX I N POOL BOILING HEAT TRANSFER 41

been mentioned by the contributors. Mr Chicken com- section and container used in the present simple experi-
ments that short duration power transients may produce ments may have been insufficiently sensitive to detect a
results different from those obtained in the steady state possible change in mechanism at a/g = 10. Use of the
conditions of the present experiments. Evidence of flat-plate geometry also enabled isolation of the effect of
Johnson et al. (68) has demonstrated marked non-linear Coriolis acceleration in the rotating system, and it was
behaviour, which suggests that the relation between steady shown to have no effect on qcr. Their results with varying
state and transient conditions is not too clear. It has already surface tension are of value in checking the experimental
been noted that Mr Stein believes that the large pressure dependency of qcr on surface tension, as compared with
gradients in the high acceleration experiments would theoretical predictions. At a/g = 1 it was found (48) that
produce a significantly different situation than under qcr cc U, as opposed to theoretical predictions which give
normal gravity conditions. This has been expanded upon qcr K d. This would seem to indicate limitations in
from different aspects by Mr Graham and Mr Hsu, present theories of far greater extent than the acceleration
Professor Griffith, and Mr Lacey. Professor Griffith is effect. Why was the Kutateladze equation (2) able to
concerned about the direct conduction of heat into the correlate qcr sufficiently accurately at a/g = 1 for many
body of the fluid (the non-boiling component of heat different liquids (and thus different u), but unable to
transfer). This was not considered in the present qcr correlate qer at all for a given liquid with different small
experiments for accelerations to a/g = 160 for the follow- amounts of surface active agent (and thus different u)?
ing reason. From the natural convection-nucleate pool I myself believe the difference could be caused by the
boiling work of Merte and Clark (32) for accelerations hydrodynamic behaviour of a foam in the latter case,
<
1 Q ajg 21, the heat flux at transition from natural compared with the normal bubble sizes and behaviour
convection to nucleate boiling may be approximately associated with the former case. Probably the hydrodyna-
estimated to vary as (aig)".". This is not too dissimilar from mic factors of importance in a foam are much different
the variation of qcr with acceleration, and as the natural from those in a simple boiling liquid with no additives.
convection component at qcr for a/g = 1 is of the order There are, however, a number of points which may be
0.1 qcr or less, the natural convection component is con- raised concerning the data presented by Professor
sidered as small. It would seem that at higher accelerations Costello. It should be noted that the experiments (shown
the natural convection component would become more in Fig. 19) at a/g = 1 were made in a separate pool (48);
important. I agree with Mr Lacey that the obtaining of spurious effects may have arisen owing to a difference
flow pattern data in any future experiments of this kind between the tests for a/g = 1 and a/g > 1. This is noted
would be of value. Unfortunately the flow patterns in because some weight is placed upon the a/g = 1 results
pool boiling are not entirely understood even for normal to demonstrate the discontinuity in the qcr on a/g curve.
gravity conditions, although recently much high-speed I n addition, the flat plate used was a flat strip 2 in. long
cine-camera work on boiling bubbles has been completed. x 0.30 in. wide.
In reply to the query of Professor Costello, no check was On p. 31 the data of reference (34) are stated to agree
made in the present experiment of the flow pattern or the extremely well with those obtained in comparable experi-
possible influences of a container size effect or of secondary ments in forced convection swirl flow inside tubes (49).
flows. In this respect I agree with the view expressed by Inspection of Fig. 7 of reference (49) shows that in swirl
Mr Chicken, that caution must be exercised in applying flow no additional effect on qcr due to the acceleration is
the present results (or indeed experimental results from discernible until a/g > 100. Two data points from
any other single boiling system) to conditions other than reference (34) are shown in Fig. 7 of reference (49), at
those of the original experiment. a/g = 10; one point is shown as having zero incremental
I wish to thank Professor Costello for his valuable and heat flux over that at a/g = 1, while at a/g = 44 the other
interesting contribution, in which are presented a large point is shown as having an incremental heat flux only
quantity of entirely new data for several different test 0.8 per cent of the maximum swirl flow incremental heat
sections. The work amplifies in much greater detail the flux. This is so although the qcrat a/g = 44 represents an
present results obtained from a simple apparatus. increase of 100 per cent over the qcr at a/g = 1 for the
Kutateladze and Styrikovich (61, p. 193) have noted results of reference (34). From this it does not appear
that an investigation by Styrikovich and Polyakov 'con- feasible on the basis of the above directly to compare
firmed the conclusion arrived at theoretically that a results of centrifugal pool boiling with those of swirl
horizontal plate with its heating surface turned upward flow, unless one is attempting to demonstrate from Fig. 7
should be used as a standard for comparison'. Because of that qcr is not affected by acceleration until a/g > 100 for
this, it may be considered that the recent flat-plate pool boiling
experiments conducted by Professor Costello (48) are The comparison shown in Fig. 18 between the data of
probably a distinct improvement on the present experi- reference (34) and those of Choi ( 5 0 ) may be misleading.
ments, as the flat-plate geometry corresponds closely to Choi obtained only two values of qcr at two different
theoretical idealizations, also there is less likelihood of values of electric field, and one value ofqcr with no electric
secondary flows influencing the boiling action on the heater field (normal gravity), all using a 0.020 in. diameter
surface. For example, the comparatively small size of test platinum wire. The equivalent accelerations (induced
Proe Instn Mech Engrs Vol 177 No I 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016


42 AUTHOR’S REPLY

electrically) at the heater surface in Choi’s experiments (33) in Fig. 17 as more than a general indication of the
were 110 times and 440 times normal gravity respectively. behaviour in steady state experiments may be misleading.
In the two experiments the electric field was non-uniform As the tests were transient the values of qcr obtained were
and directed radially away from the cylindrical heater. probably higher than would be obtained under steady
This is a very different situation to the single experiment state conditions, as noted previously by Mr Chicken.
with no electric field, in which the gravitational field was Finally, I wish to emphasize the need for more accurate
uniform and directed vertically. Yet the justification for statements of boiling problems in the various system
showing two distinct gradients for Choi’s three values of conditions encountered. Kutateladze has given what must
qct depends upon the one value of per obtained at normal be the simplest and clearest plausible statement of the
gravity, which was obtained in conditions much different pool boiling critical heat flux condition (2) (61). As the
from the two values of 4cr obtained with acceleration ranges of important variables are increased (acceleration,
induced electrically. Two quotations from Choi’s work surface tension, etc.) so new factors and mechanisms may
suffice to illustrate this point (so) : ‘Actually, this transition become of importance. Thus a re-defining of the problem
in Fig. 18 may be more apparent than real‘, and the con- is required, as the early definitions of a boiling problem
clusion that ‘the electrical field increases the peak heat may not always be sufficiently flexible or general to account
flux, the increase being approximately linear with E,’ for effects obtained under new ranges of variables. An
(electric field at wall surface). In Fig. 18 it should be example of this is the work of Breen and Westwater (69)
recollected that although the upper curve may give on film boiling performance in natural convection as a
exactly the prediction of the Zuber theory, the Kutateladze function of diameter of horizontal cylinders, where
theory contains a constant which is determined by experi- ‘classical’ film boiling theory is demonstrated to apply
ment. The Kutateladze theory may only be examined only over a limited range of cylinder diameters.
from the gradients of the experimental curves in Fig. 18,
whereas the Zuber theory must also be examined for REFERENCES
absolute values. (58) DONALD,M. B. and HASLAM, F. ‘The mechanism of the
On p. 33 Professor Costello has stated that pressure transition from nucleate to film boiling’, Chem. Engng Sci.
build-up in the present apparatus was 23 per cent more 1958 8,287.
than that of reference (48) at any a/g (for 1 in. liquid (59) STEIN, R. P. ‘Critical review of the Zuber and Zuber-
depth). The pressure build-up is greater for a given liquid Tribus theories of transition boiling’, Columbia Univ.
Dept. Chem. Engng Tech. Note IX-TN-3-58 1958.
depth and acceleration, because the change in acceleration (60) MCFADDEN, P.W. and GRASSMAN, P. ‘The relation between
with respect to radius is considerably smaller for the bubble frequency and diameter during nucleate pool
present apparatus with 25.53 in. radius of rotation, than boiling’, Int. J . Heat Mass Transfer 1962 5, 169.
for that of reference (48) with 2.5 in. radius of rotation. (61) KUTATELADZE, S. S. and STYRIKOVICH, M. A. ‘Hydraulics
The acceleration change with radius in the latter case of gas-liquid systems’ (Moscow 1958). Translation,
September 1960: Technical Information Center, Wright-
was more than 10 times that in the former. In view of this, Patterson Air Force Base, Ohio, U.S.A. F-TS-9814/V.
and the observation by Jakob (9, p. 623) and Cole (10) (62) KAZAKOVA, E. A. ‘Influence of pressure on the production
that steam bubbles achieved terminal rise velocity within of the first crisis in the boiling of water on a horizontal
one bubble diameter, has Professor Costello considered plate’, AEC-tr-3405, 86 (1953).
(63) HAWKINS, G. A. ‘A brief review of the literature on boiling
the rapid deceleration of the vapour with decreasing radius, heat transfer’, U.C.L.A. Bept. Engng Rep., June 1950.
and the vapour build-up, in his apparatus? On p. 34 (64) MUELLER, G. 0. ‘A review and assessment of boiling heat
Professor Costello has stated that ‘good agreement’ transfer and the departure from nucleate boiling’,
between data from references (34) and (48) showed KAPL-M-GOM-2 (1958).
no effect due to a difference in potential subcooling. I n (65) AVERIN,E. K. ‘Effect of material and machining of surface
on heat transfer to boiling water’, Isv. Akad. Nauk.
fact the difference in 4cTbetween the two sets of data at S.S.S.R., Odel. Teckli. Nauk. 1954 (no. 3), 116;
a/g = 44 due to potential subcoolings (3~78°Fand 6~2°F U.K.A.E.A., A.E.R.E. LIB/TRANS 562.
respectively) would be 3 per cent according to equation (66) MIRSHAK,S. ‘Heat flux at burnout in stagnant water’,
(31). Inspection of Fig. 17 at a/g = 44 shows an experi- DP-262 (1957).
mental difference of 20 per cent between the data of (67) GUNTHER,F. C. and KREITH,F. ‘Photographic study of
bubble formation in heat transfer to subcooled water’,
references (34) and (48), the former lower than the latter; Ht. Trans. and Fluid Mech. Irist. 1949, 113.
this is in the same direction as would be expected from the (68) JOHNSON, H.A., SCHROCK, V. E., SELPH,F. B., LIENHARD,
relative subcooling magnitudes. In addition Fig. 17 shows J. H. and ROSZTOCZY, Z. R. ‘Transient pool boiling of
that all data of reference (34) fall systematically below water at atmospheric pressure’, Proc. Int. Ht. Trans.
those of (48). Therefore some difference is evident be- Conf.,Boulder 1961 (pt. 2), 244.
(69) BREEN,B. P. and WESTWATER, J. W. ‘Effect of diameter of
tween the two sets of data. Consideration of the data horizontal tubes on film boiling heat transfer’, Chem.
obtained under transient conditions of Usiskin and Siege1 Engng Prog. 1962 58 (no. 7), 67.

Proc Instn Mech Engrs Vo1177 No I 1963

Downloaded from pme.sagepub.com at The University of Auckland Library on June 5, 2016

You might also like