You are on page 1of 12

International Journal of Heat and Mass Transfer 112 (2017) 850–861

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


journal homepage: www.elsevier.com/locate/ijhmt

Effect of initial cooling on heat and mass transfer at the cryogenic surface
under natural convective condition
Sangho Choi a,b, Sung Jin Kim a,⇑
a
Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology, 291 Daehak-ro, Daejeon 305-701, Republic of Korea
b
Department of Thermal & Aerodynamics, Korea Aerospace Research Institute, 115 Gwahangno, Daejeon 305-333, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: The effect of initial wall cooling from ambient temperature to cryogenic temperature on heat and mass
Received 26 December 2016 transfer at the cryogenic surface under natural convection was experimentally and numerically investi-
Received in revised form 7 April 2017 gated. The experimental study showed that the initial wall cooling had a strong effect on heat and mass
Accepted 9 May 2017
transfer at the cryogenic surface. The frost under initial wall cooling grew considerably thicker than the
case without initial wall cooling. The maximum heat flux under initial wall cooling was 40% of that with-
out initial wall cooling, and the minimum heat flux under the initial wall cooling was 52% of that without
Keywords:
initial wall cooling. In addition, a numerical model for the frost formation accounting for initial wall cool-
Frost formation
Cryogenic surface
ing was proposed. The proposed numerical model could explain the heat and mass transfer at the cryo-
Heat and mass transfer genic surface during the cooling process as well as the filling and holdup process. In order to validate the
Effect of initial wall cooling proposed numerical model, experiments were performed under various ambient air temperature and rel-
Natural convection ative humidity conditions: 10 °C  Ta  30 °C and 30%  RH  90%. The maximum and minimum heat
flux from the numerical model showed good agreement with experimental data within 10% and 25%
error, respectively. The final frost thickness from the numerical model showed good agreement with
experimental data within 13% error except for one case where mass transfer was reduced due to fog for-
mation near the cryogenic surface. Therefore, the numerical model will be useful for estimating the heat
flux in an uninsulated cryogenic system, such as a rocket oxygen tank.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction insulator. Because frost plays the role of an insulator, heat flux into
the cryogenic oxygen tank is reduced considerably compared to
Frosting phenomenon on cold surface occurs in a variety of that under no frost formation. Heat flux into the oxygen tank
applications, such as refrigeration, aerodynamics, propellant load- causes liquid oxygen to evaporate and to be stratified. Evaporation
ing in rockets, CO2 reduction, and many others. Under atmospheric of liquid oxygen requires additional filling during ground opera-
conditions, when the wall temperature drops below the dew point tion, and stratification renders some portion of liquid oxygen unus-
of air, water vapor in the air transfers to the wall and condenses at able during flight. Therefore, accurate knowledge of heat and mass
the surface because there is a pressure difference between the par- transfer at the cryogenic surface is important in optimizing the
tial pressure of water vapor in the air and the saturation pressure amount of oxygen loading in rockets.
at the wall temperature. If the wall temperature further decreases In the early stages of rocket development many researchers
below the freezing point of water, deposition of water vapor into investigated frost formation on uninsulated cryogenic surfaces.
ice particles occurs. The thickness of the frost layer then starts to Ruccia and Mohr [1] experimentally studied frost formation on
increase due to the growth and branching of the ice crystals. an oxygen-filled tank exposed to uncontrolled ambient air under
Heat and mass transfer in uninsulated cryogenic systems has natural convection and forced convection. They attempted to
been extensively investigated because of their application in rocket determine the applicability of certain existing heat transfer corre-
propulsion system. The liquid oxygen tank used in rockets has no lations at the cryogenic temperatures. However, these attempts
thermal insulation except in the case of the space shuttle as the were unsuccessful because the obscured heat flux measurement
frost formed on the cryogenic oxygen tank surface acts as a natural at the initial stage prevented them from verifying the calculated
heat flux. Nevertheless, they found that the experimentally mea-
⇑ Corresponding author.
sured mass transfer rate was one-sixth lower than the theoretically
E-mail address: sungjinkim@kaist.ac.kr (S.J. Kim).
expected results. Holten [2] experimentally studied heat and mass

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2017.05.033
0017-9310/Ó 2017 Elsevier Ltd. All rights reserved.
S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861 851

Nomenclature

B porosity Subscript
cp specific heat capacity at constant pressure (J/kgK) a air
Gr Grashof number cond conduction
h convective heat transfer coefficient (W/m2K) conv convection
hm mass transfer coefficient (m/s) cool cooling
H latent heat (J/kg) cryo cryogenic condition
k thermal conductivity (W/mK) d diffusion
Le Lewis number da dry air
m_ 00 mass flux (kg/m2s) D diameter
Nu Nusselt number f frost
Pr Prandtl number i ice
q00 heat flux (W/m2) ini initial condition
R gas constant (J/kgK) l liquid
Ra Rayleigh number L length
Re Reynolds number lat latent
RH relative humidity (%) LOX liquid oxygen
t time (s) m mass transfer
T temperature (K) rad radiation
w absolute humidity (kg/kgda) ref reference temperature
x spatial coordinate (m) s solid
SS stainless steel
Greek symbols sat saturation
a dimensionless wall temperature th thickness
d frost thickness (m) tp triple point of water
 emissivity v vapor
q density (kg/m3) wall wall condition
r Stefan-Boltzmann constant
; view factor

transfer in an oxygen-filled tank exposed to humidity-controlled based on a numerical analysis and were not validated. Liu et al. [8]
air under natural convection. He found that heat flux was the max- experimentally studied frost formation on a liquid nitrogen cooled
imum at the moment of exposure to ambient air and decreased surface under various conditions. They found that oxygen-rich
rapidly, and then approached a minimum as time passed. Further- water droplets were formed on a cryogenic surface, and they mea-
more, he found that the effective heat transfer modes were convec- sured the frost layer thickness at various wall temperatures, air
tion and radiation at cryogenic temperatures and phase change temperatures, and relative humidity conditions.
heat transfer had a minimal effect. Barron and Han [3] studied heat All the previous researchers experimentally or numerically
and mass transfer in a nitrogen-filled tank in a temperature and studied frost formation under a constant cryogenic temperature,
humidity controlled chamber. They found that heat transfer was although the main application, namely, rocket prelaunch opera-
in good agreement with the theoretical expectation, but mass tions, is not a constant temperature process. In the rocket pre-
transfer was much lower than the theoretical prediction. They launch operations, the oxygen tank should be initially cooled
explained that mass transfer was reduced because water droplets from the ambient temperature to the cryogenic temperature before
or ice particles in the boundary layer blocked diffusion of the water a constant cryogenic temperature is reached during the filling and
vapor. Shah [4] experimentally and numerically investigated frost holdup process, as shown in Fig. 1. During the initial wall cooling,
formation on a liquid nitrogen cooled surface under forced convec- which usually takes approximately 60 min, frost starts to form on
tion. He proposed a simple theoretical model and the experimental the wall and a considerable amount of frost is accumulated on the
results showed good agreement with the proposed model. He wall by the time the wall temperature reaches the cryogenic
found that the density distribution within the frost layer was
nearly uniform. Brian et al. [5] experimentally studied frost forma-
tion on a liquid nitrogen cooled plate from a humid air stream.
They suggested an empirical correlation for the frost thermal con-
ductivity with average frost density and temperature. Biguria and
Wenzel [6] conducted an experiment to obtain the thermal con-
ductivity and density of frost. They used Freon as the coolant which
was cooled by liquid nitrogen. They suggested complex correla-
tions for the frost thermal conductivity and density as a function
of the wall temperature, humidity, air velocity, time, and flow
regime.
Recently, Kim et al. [7] numerically investigated frost formation
on a cryogenic oxidizer tank wall for a liquid-propulsion rocket.
They numerically analyzed the effects of parameters and stated
that wind speed, air temperature, and relative humidity had signif-
icant effects on heat and mass transfer. However, their results were Fig. 1. Wall temperature variations with and without initial wall cooling.
852 S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861

temperature. Since accumulated frost acts as a thermal insulator, Table 1


the temperature at the interface between the porous frost layer Experimental conditions.

and the ambient air is much higher than the cryogenic tempera- Test Test Ambient air Relative Cooling
ture, and heat flux into the oxygen tank is much smaller than that number labeling temperature [°C] humidity [%] time [min]
of the constant cryogenic temperature. Therefore, it is important to 1 RH60T20NC 20 60 0
accurately estimate the heat flux into the oxygen tank wall in an 2 RH60T20 20 60 60
actual situation. In this study, the effect of initial wall cooling from 3 RH60T30 30 60 60
4 RH60T10 10 60 60
ambient temperature to cryogenic temperature on heat and mass 5 RH90T20 20 90 60
transfer at the cryogenic surface under a natural convective condi- 6 RH30T20 20 30 60
tion is experimentally investigated, and the results are compared
with those of the cryogenic constant wall temperature condition.
A numerical model for frost formation accounting for initial wall
cooling is then proposed. In order to validate the proposed model, measurement accuracy of ±0.033 mm, the frost thickness was mea-
experiments at various ambient air temperature and relative sured at 60-s intervals. In addition, the frost formation and the
humidity conditions are performed under initial wall cooling and growth pattern were monitored using another USB microscope
the results of the numerical model are compared with experimen- with a 60-Hz speed. The wall temperature was measured with
tal data. three T-type thermocouples, and they were placed 100 mm above
the bottom surface of the cylinder. The thermocouples were cali-
brated with an accuracy of ±0.3 °C. The wall heat flux was mea-
2. Experiment sured using a heat flux sensor (HFS-50, Captec) that was placed
on the outer surface of the cylindrical tank corresponding to the
The primary objective of this experiment was to observe the position of the frost thickness measurement. The heat flux sensor
effect of initial wall cooling on heat and mass transfer at the cryo- was calibrated with an accuracy of ±5% within the range of 0–
genic surface under natural convection. A secondary goal was to 2000 W/m2. The temperature and heat flux measurements were
validate the proposed numerical model with these measurements recorded by a data acquisition system (SCXI-1600, National
for frost formation under various ambient air temperature and rel- Instruments).
ative humidity conditions. Table 1 lists experimental conditions of The frost mass was measured at 30 min intervals, and an
ambient air temperature, relative humidity, and cooling time. axisymmetric mass distribution was assumed when measuring
Fig. 2 shows a schematic illustration of the experimental appa- the frost mass. To minimize disturbance, the frost mass was mea-
ratus. The experimental apparatus was placed inside an environ- sured from the opposite position to the measurement position of
mental chamber, which controlled the air temperature and the frost thickness. To measure the frost mass deposited on the
relative humidity. The size of the environmental chamber was test section, frost deposited on the wall was removed and placed
3 m (width)  3 m (depth)  2.5 m (height). The environmental into a paper cup using a plastic spoon. The frost mass was mea-
chamber controlled the air temperature from 10 °C to 80 °C and sured using a balance (GX-600, AND) and the measurement accu-
the relative humidity from 20% RH to 95% RH. During the test, racy of the frost mass was ±0.02 g. The frost density was
the air temperature was controlled with ±1.0 °C accuracy, and calculated using the measured mass, removal area, and frost
the relative humidity was controlled with ±2.0% accuracy. Near thickness.
the test section, the air humidity and temperature were measured When conducting an experiment for frost formation without
using a thermo-hygrometer (CENTER 342, CENTER). The measure- initial wall cooling, a cylindrical tank was insulated using two lay-
ment accuracy of the ambient air temperature and relative humid- ers of 20-mm Armacell insulator. After the cylindrical tank was
ity was ±0.4 °C and ±3%, respectively. When controlling the air cooled to a cryogenic temperature, liquid nitrogen was supplied
temperature and relative humidity, a fan, suspended in the ceiling, into the cylindrical tank. During the cooling (chill-down) process,
blew the air, and the air velocity near the test section was 0.5 m/s. no frost was formed at the tank wall owing to the thermal insula-
To reduce the forced convection effect, perforated Styrofoam was tor. After the tank was filled with liquid nitrogen, the insulator was
installed around the test section. Thus, the measured air velocity removed. The frost thickness and heat flux were then measured.
near the test section was less than 0.1 m/s, and the value of When conducting an experiment for frost formation with initial
Gr L =Re2L was approximately 100. Therefore, a natural convection wall cooling, an insulator was not attached to the cylindrical tank.
condition was established. Prior to the experiment, a cryogenic tank was placed in an environ-
The nitrogen from supply tank was supplied to test section. The mental chamber for approximately 4 h to obtain a steady-state
pressure and temperature gauges were installed at the inlet to temperature. After the wall temperature approached ambient air
monitor the status of supplied nitrogen. The pressure of supplied temperature, liquid nitrogen was supplied from the nitrogen sup-
nitrogen was regulated using a pressure control valve to maintain ply tank to the cylindrical tank. The frost thickness and heat flux
a constant mass flow rate. The supply pressure was maintained were measured at the moment the nitrogen was supplied. The time
from 3.5 bar to 4.0 bar, which is a typical supply pressure in a for the wall temperature to reach a cryogenic temperature was
rocket prelaunch process. selected as 60 min, which is typical cooling time in rocket pre-
Fig. 3 shows a schematic illustration of the test section and sen- launch operations.
sor locations. The test section was a cylindrical tank made of stain-
less steel with 5 mm thickness. The diameter of the cylindrical tank 3. Numerical model for frost formation with initial wall cooling
was 330 mm, and the height was 900 mm. The cylindrical tank was
supported by four aluminum supports. The cylindrical tank and 3.1. Physical assumptions
support were connected by carbon fiber material to reduce heat
transfer. A level gauge was installed inside the tank to monitor The following physical assumptions are employed in the
the level of liquid nitrogen, and the inlet and vent ports were numerical model for frost formation.
located at the top of the apparatus.
The frost thickness was measured using a USB microscope (BS- (1) The frost starts to form if the wall temperature is lower than
200T2, BOSSCOPE) with a 15 resolution. After calibration with a the freezing point of water.
S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861 853

Fig. 2. The schematic illustration of the experimental apparatus.

Fig. 3. The schematic illustration of the test section and sensor locations.

(2) The frost layer grows in a direction normal to the cold analyze heat and mass transfer, the wall temperature was
surface. required. However, the physical phenomenon was quite complex
(3) Heat and mass transfer within the frost layer is one- because boiling and convection took place inside the tank and frost
dimensional and quasi-steady. formed on the outer surface of the tank. To simplify the analysis, it
(4) The frost density and the thermal conductivity are uniform was assumed that the wall temperature was given. The experimen-
over the frost thickness. tally measured wall temperature under a constant mass flow rate
(5) Water vapor on the frost surface is saturated. behaved as an exponential function during the cooling process.
(6) Dry air and water vapor behave like ideal gases. After reaching a cryogenic temperature, the wall temperature
(7) Natural convection and thermal radiation within the frost was nearly constant. Therefore, following equations can be used
layer are negligible. with time, t, in minutes.

3.2. Wall temperature T wall ðtÞ ¼ T ini expðC 1 tÞ if t < tcool ð1aÞ

The wall temperature changed from ambient temperature to


T wall ðtÞ ¼ T cryo if t P t cool ð1bÞ
cryogenic temperature during the initial wall cooling and
remained constant during the filling and holdup process. When Fig. 4 shows the experimentally measured wall temperature
the wall temperature was higher than the dew point, there was and numerical data using Eq. (1). The value of constant C 1 is
only heat transfer and no mass transfer. If the wall temperature 0.0185 min1 when the wall is cooled for 60 min. Examining the
was lower than the dew point but higher than the freezing point two data sets showed that the maximum difference between the
of water, heat transfer and mass transfer occurred simultaneously. experimental data and the numerical approximation was ±13 °C.
Condensation consequently occurred on the cold surface. If the A more precise approximation was possible if a different constant
wall temperature was lower than the freezing point of water, was used for each case; however, this was not implemented to
deposition of water vapor into ice occurred and frost formed. To maintain generality.
854 S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861

3.3. Heat and mass balance ture are nearly bounded, and changes in temperature or pressure
in the frost layer are slow. This approach may be extended to a
In the early stages of the cooling process, the wall temperature transient wall temperature condition because the inertia term is
was higher than the dew point. The wall heat flux was calculated much smaller than the conduction term in the energy equation.
using the following equation. At 30 min in test number 2, the orders of magnitude of the inertia
and conduction terms are calculated as follows:
q00wall ¼ q00conv þ q00rad ¼ hðT a  T wall Þ þ ;    rðT 4a  T 4wall Þ ð2Þ
 
@2T 273  170
The heat transfer coefficient was determined using the Church- kf  O 0:06  ¼ Oð105 Þ ð7aÞ
ill and Chu correlation [9]. @x2 0:0052
8 92
>
< >
= @T
hL
1=6
0:387RaL qf cp;f  Oð12  2000  0:05Þ ¼ Oð103 Þ ð7bÞ
NuL ¼ ¼ 0:825 þ h ð3Þ @t
ka >:  9=16 i8=27 >
;
1 þ 0:492
Pr Calculation results for other experimental cases or other time
indicate similar results. Therefore, the quasi-steady assumption
If the cooling process was continued, the wall temperature was during the cooling process is justified. After initial wall cooling is
lower than the dew point but higher than the freezing point of finished, the wall temperature remains constant. In this case, the
water. Water vapor in the air condensed on the cold wall and addi- situation is similar to frost formation at constant wall temperature
tional heat transfer by a phase change phenomenon subsequently and the quasi-steady assumption is justified [4].
occurred. The wall heat flux was calculated by the following The governing equation of frost formation is adapted from
equation. Kandula’s numerical model [10] which is developed for a moder-
q00wall ¼ q00conv þ q00rad þ q00lat ately low constant temperature condition. However, the wall
boundary condition is modified to reflect the transient wall tem-
¼ hðT a  T wall Þ þ ;    rðT 4a  T 4wall Þ þ m
_ 00 Hlv ð4Þ perature condition and the frost surface boundary condition is
The condensate mass flux was obtained from the following modified to include radiation heat transfer between the ambient
equation. air and the frost surface. Fig. 5 shows the heat and mass transfer
  in the frost layer. There are two mechanisms of energy transfer:
RHP sat;a Psat;wall conduction heat transfer and latent heat transfer due to the phase
_ 00 ¼ hm ðqv ;a  qv ;wall Þ ¼ hm
m  ð5Þ
Rv T a Rv T wall change of water within the frost layer. If the frost density is
assumed to be uniform within the frost layer [4], the amount
The mass transfer coefficient was determined using the Chilton-
of water vapor frozen must be the same at all locations [11].
Colburn analogy.
The energy balance within the frost layer can be expressed as
h follows:
hm ¼ ð6Þ
qcp Le2=3 2
d T dm_ 00x m_ 00
kf 2
¼ Hsv ¼ Hsv d ð8Þ
If the wall temperature was lower than the freezing point of dx dx xf
water, deposition of water occurred and frost formation began.
where the latent heat of phase change for water is given by
The wall heat flux could not be calculated directly because the
temperature of the frost surface was unknown. To obtain the wall Hsv ¼ 2:88  106  195:0T f ð9Þ
heat flux, the temperature distribution within the frost layer
should be solved. A general approach for modeling frost formation The boundary conditions for Eq. (8) are given below.
at a constant wall temperature is to assume a quasi-steady state T ¼ T wall ðtÞ at x¼0 ð10aÞ
because both the wall temperature and the frost surface tempera-
T ¼ Tf at x ¼ xf ð10bÞ

dT 
kf _ 00th þ ;    rðT 4a  T 4f Þ at
¼ ½hðT a  T f þ Hsv m x ¼ xf
dx
ð10cÞ
The emissivity of the frost is estimated using the following
equation, suggested by Dietenberger [12].

 ¼ ð1  Ba Þ½ðSS  i Þ expð105 xf Þ þ i  ð11Þ


The emissivity of stainless steel, SS , and the emissivity of ice, i ,
are 0.17 and 0.94, respectively [9,12].
The deposition mass flux is obtained from the following
equation.
 
RHP sat;a Psat;f
_ 00 ¼ hm ðqv ;a  qv ;f Þ ¼ hm
m  ð12Þ
Rv T a Rv T f
The mass transfer from the ambient air contributes partly to
increasing the frost density and partly to increasing the frost thick-
ness. This can be expressed as follows:

dxf dqf
m _ 00d ¼ qf
_ 00th þ m
_ 00 ¼ m þ xf ð13Þ
dt dt
Fig. 4. Experimentally measured wall temperature and numerical data from Eq. (1). The rate of change of the frost thickness can be expressed by
S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861 855

Fig. 5. Heat and mass balance in the frost layer.

 
dxf dqf between 10 and 50 kg/m3 [16]. Since there is no reliable correlation
¼ m_ 00  xf =qf ð14Þ
dt dt for natural convection applicable to a wide range of wall tempera-
ture, a correlation for the frost density based on experimental
The temperature distribution in the frost layer was obtained by results was proposed. The correlation was composed of absolute
integrating twice and applying boundary conditions. The result can humidity, non-dimensional time, air temperature, triple point of
be expressed as shown in Eq. (15). water, and wall temperature. This correlation is valid for 10 °-
1h  i H C  Ta  30 °C and 30%  RH  90% when the wall is cooled for
T ¼ T wall þ _ 00 þ ;    rðT 4a  T 4f Þ x  sv m
hðT a  T f þ Hsv m _ 00 x2
kf 2kf d 60 min.
ð15Þ  0:7362  0:0503
qf t T a  T tp
¼ 1:2209w0:8398 ð17Þ
Finally, the wall heat flux was determined using Eq. (16). qi a
t cool T a  T wall

dT   Fig. 6 shows a comparison of the suggested correlation and the
q00wall ¼ kf _ 00 þ ;    rðT 4a  T 4f Þ
¼ ½hðT a  T f þ Hsv m ð16Þ
dx x¼0 experimental data for the purpose of validation. The solid line rep-
resents a perfect match between the correlation and experimental
data. The average error of the frost density was approximately
3.4. Frost properties ±15% except for low density.

3.4.1. Frost density 3.4.2. Frost thermal conductivity


To obtain the frost surface temperature by solving the above Another key property in frost formation modeling is the frost
equations, the thermo-physical properties, such as frost density thermal conductivity. A wide range of correlations for the frost
and frost thermal conductivity, are required. Many correlations thermal conductivity exist. Yonko and Sepsy [17] suggested an
have been suggested to predict frost density, and a simple correla- empirical correlation for the frost thermal conductivity as a
tion by Hayashi et al. [13] has been widely used. Hermes et al. [14] quadratic-polynomial function of the frost density. Later, Sturm
modified Hayashi’s correlation by taking into account the effect of et al. [18] followed the same approach and suggested another
the wall temperature as well as the frost surface temperature. empirical correlation as a linear function of frost density. Brian
Kandula [10] surveyed a wide range of experimental data and sug- et al. [5] studied frost formation on a cold plate, where the wall
gested a correlation as a function of non-dimensional temperature temperature varied from moderately low temperatures to cryo-
and Reynolds number. Recently, Hermes et al. [15] reviewed exist- genic temperatures, and suggested a correlation for the frost ther-
ing correlations for the frost density and suggested a semi- mal conductivity as a function of the average frost density and the
empirical correlation for the frost density. They included the time wall temperature as
effect explicitly and did not use the frost surface temperature
because knowledge of the frost surface temperature was limited. kf ¼ 1:56441  105 T 1:441
wall þ 4:25352  10
11
ðqf  25ÞT 3:055
wall
However, all of these correlations have been developed for lam-
ð18aÞ
inar forced convection and are only applicable to moderately low
temperatures ranging approximately from 40 °C to 5 °C. where qf > 25kg=m and T wall < 255:2K and as
3

Although a few researchers have conducted experimental studies


wall þ 3:93212  10 qf T wall
kf ¼ 2:42527  105 T 1:272 8 1:74
on frost formation under natural convection at cryogenic tempera- ð18bÞ
tures [1–3], they did not suggest any correlations for the frost den-
sity. The frost formed at moderately low temperatures under where qf < 25kg=m3 and T wall < 255:2K.
forced convection is characterized by densities ranging from 300 Dietenberger [19] developed a comprehensive semi-empirical
to 500 kg/m3, while the frost formed at cryogenic temperatures model to calculate the frost thermal conductivity. A theoretical
under natural convection is usually light and fluffy with densities model of a random mixture of ice cylinders and ice spheres at
856 S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861

Fig. 7. Comparison of wall heat flux using various frost thermal conductivity
Fig. 6. Comparison of measured and correlated frost densities. correlations.

kf ¼ akNegrelli;T ref þ ð1  aÞkBrian;T cryo if T wall > T cryo ð20aÞ


low densities or of air bubbles and ice planes at high frost densities
Z T ref
was suggested. Auracher [20] developed an effective thermal con- 1
ductivity model as a blend of serial and parallel thermal resis- kf ¼ kBrian dT if T wall ¼ T cryo ð20bÞ
T ref  T wall T wall
tances. Kandula [21] developed a fully theoretical model for the
effective thermal conductivity including the effect of mass diffu- where non-dimensional temperature is given in Eq. (21).
sion and eddy convection on stagnant conduction in the fluid.
T wall  T cryo
Recently, Negrelli and Hermes [22] reviewed existing correlations a¼ ð21Þ
and suggested a semi-empirical correlation as a function of
T ref  T cryo
porosity.
 c3 Ba 3.5. Numerical procedure
kf ka
¼ c2 ð19Þ
ki ki If the wall temperature is higher than the freezing point of
Negrelli and Hermes’s correlation is only applicable to moder- water, calculation of the heat and mass transfer rates is straightfor-
ately low temperatures. Coefficients for Negrelli and Hermes’s cor- ward. After calculation, time marches on the next time step. If the
relation are listed in Table 2. Fig. 7 shows a comparison between wall temperature is lower than the freezing point of water, the
the experimental and calculated wall heat fluxes when frost ther- frost starts to form. After frost formation, the frost surface temper-
mal conductivity correlations were incorporated into the numeri- ature and the wall temperature are not equal. To calculate the heat
cal model. The wall heat fluxes based on previous correlations and mass transfer rates, the frost surface temperature should be
were quite different from the experimentally measured heat flux. determined by solving Eqs. (12), (14), and (15). For this, the frost
Since there was no correlation for the frost thermal conductivity surface temperature is guessed and then the frost properties are
over a wide range of temperature, a combined model was used calculated from Eqs. (17) and (20). With the frost properties, the
in the numerical model. The thermal conductivity model devel- heat and mass transfer rates are calculated from Eqs. (12) and
oped by Negrelli and Hermes [22] was used at moderately low (16). Finally the frost thickness and the frost surface temperature
temperatures and the model reported by Brian et al. [5] was used are calculated from Eqs. (14) and (15). If the difference between
at cryogenic temperatures. The thermal conductivity was linearly the guessed frost surface temperature and the calculated frost sur-
interpolated using non-dimensional temperature at other temper- face temperature satisfies the convergence criteria, time marches
atures. The frost thermal conductivity by Brian et al. was strongly on the next time step until the final time is reached. If the conver-
affected by the wall temperature, as shown in Eq. (18). However, gence criterion is not satisfied, the frost surface temperature is
Brian’s model did not consider the transient wall temperature updated by using the guessed and calculated frost surface temper-
effect. To add the transient wall effect to Brian’s model, tempera- ature. A flowchart for the numerical procedure is shown in Fig. 8.
ture averaged thermal conductivity was used.
4. Results and discussion

Table 2
The effect of initial wall cooling on heat and mass transfer at the
Coefficients of Eq. (19). cryogenic surface was observed from test numbers 1 and 2. The
effect of the ambient air temperature on frost formation on the
Temperature range Coefficient c2 Coefficient c3
cryogenic surface with initial wall cooling was observed from test
10 < T wall < 4 °C 1.576 0.797 numbers 2, 3, and 4. The effect of the relative humidity on frost for-
19 < T wall < 10 °C 1.594 0.761
mation on the cryogenic surface with initial wall cooling was
30 < T wall < 19 °C 1.035 0.797
observed from test numbers 2, 5, and 6. In addition, results of
S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861 857

the numerical model for frost formation accounting for initial wall tial wall cooling. The heat flux was the maximum at the moment
cooling were compared with the experimental results. the thermal insulator was removed and then decreased rapidly,
and thereafter approached a minimum as time passed. The maxi-
4.1. Effect of initial wall cooling mum heat flux was 1900 W/m2 and the minimum heat flux was
800 W/m2. According to Barron [23], the maximum heat flux was
The results with and without initial wall cooling were com- 1800 W/m2 and the minimum heat flux was 1000 W/m2. There
pared when the ambient air temperature was 20 °C and the rela- was a slight difference between the two results because liquid oxy-
tive humidity was 60%. Without initial wall cooling, the surface gen is used in [23] instead of liquid nitrogen. Fig. 11 shows the
temperature of the cylindrical tank maintained a constant cryo- frost layer thickness without initial wall cooling. Frost started to
genic temperature. As soon as the thermal insulator was removed, form on the cryogenic wall at the moment the thermal insulator
water vapor in the air rapidly changed to water droplets or ice par- was removed, and the thickness of the frost increased with time.
ticles in the boundary layer, and fog formed in the boundary layer. However, the frost fell, owing to its own weight and the shear force
Fig. 9 shows fog formation in the boundary layer near the cryo- from natural convective flow. New crystals then grew in the same
genic surface. Some water droplets or ice particles in the boundary location. The maximum frost thickness was approximately 3 mm.
layer moved into the cryogenic surface and the remaining droplets The experimentally measured frost thickness was compared with
and particles were removed from the boundary layer with the nat- that of Ruccia and Mohr’s [1] experiment. The trend of frost growth
ural convective flow. Fig. 10 shows the wall heat flux without ini- was very similar; a strict comparison, however, was not possible
because the ambient conditions for Ruccia and Mohr’s experiment
were not specified.
With initial wall cooling, the surface temperature of cylindrical
tank changed from ambient temperature to cryogenic temperature.
After initial wall cooling, a constant cryogenic temperature was
maintained during the filling and holdup process. Fig. 12 shows
the wall heat flux with initial wall cooling. At the early stage, the
wall heat flux increased because the wall temperature exponen-
tially decreased (see Eq. (2)). At 6 min after the experiment began,
frost formation was first observed. As the frost grew on the wall,
the thermal resistance of the frost layer increased. The frost surface
temperature is affected by two factors. A decrease in wall temper-
ature tends to decrease the frost surface temperature. An increase
in thermal resistance as the frost grows tends to increase the frost
surface temperature. Therefore, there should be a maximum heat
flux when the frost surface temperature is a minimum. In the
experiment, the maximum heat flux occurred at 18 min after the
experiment began. The wall heat flux then decreased gradually
and finally reached a minimum. The maximum heat flux was
780 W/m2 and the minimum heat flux was 420 W/m2. Fig. 13
shows the frost layer thickness with initial wall cooling. At 6 min
after the experiment began, frost started to form on the wall and
it grew almost linearly during the cooling process. The frost thick-
ness accumulated on the wall at the end of the cooling process was
approximately 8 mm and the maximum frost thickness was
11.5 mm.
The maximum frost thickness with initial wall cooling was
11.5 mm, which was much larger than the maximum frost thick-

Fig. 8. A flowchart for the numerical procedure. Fig. 9. Fog formation near the cryogenic constant temperature surface.
858 S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861

Fig. 10. Wall heat flux without initial wall cooling. Fig. 12. Wall heat flux with initial wall cooling.

Fig. 11. Frost thickness without initial wall cooling. Fig. 13. Frost thickness with initial wall cooling.

air temperature on the wall heat flux. Increasing the air tempera-
ness, 3 mm, without initial wall cooling. The maximum heat flux ture caused the driving potential for heat transfer between the
under initial wall cooling was 40% of that without initial wall cool- ambient air and frost surface to become larger, and the wall heat
ing, and the minimum heat flux under initial wall cooling was 52% flux hence increased. Results of the numerical model showed a
of that without initial wall cooling. It could be concluded that pre- similar trend. The numerical results slightly overpredicted the
vious studies based on a constant cryogenic temperature condition wall heat flux at a high air temperature and underpredicted at a
overpredicted the wall heat flux significantly compared to that of low air temperature. The relative error, defined as jexperimental
actual rocket prelaunch operations. value  numerical valuej/ experimental value  100, of the maxi-
mum wall heat flux was less than 10% for all cases, and the relative
4.2. Effect of the ambient air temperature with initial wall cooling error of the minimum heat flux was less than 20%.
Fig. 16 shows the effect of the ambient air temperature on the
The effects of the ambient air temperature on frost formation frost thickness. The effect of the ambient air temperature on the
with initial wall cooling were examined through experiments experimentally measured frost thickness was not significant dur-
and numerical analyses for three air temperatures, i.e. 10, 20, ing the initial wall cooling. However, the final frost thickness
and 30 °C. Fig. 14 presents the frost density from the experiment decreased as the ambient air temperature increased. Although this
and calculation using the correlation to see the effect of the ambi- was somewhat unexpected, the same phenomenon was observed
ent air temperature on the frost density. The frost density in experiments by Lee and Ro [24,25] and Tahavvor and Yaghoubi
increased monotonically with time, and it increased as the ambient [26]. This can be explained as follows: the frost density increased
air temperature increased. Fig. 15 shows the effect of the ambient as the ambient air temperature increased for a given relative
S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861 859

Fig. 16. Effect of the ambient air temperature on the frost thickness.
Fig. 14. Effect of the ambient air temperature on the frost density.

for the supersaturation degree at the frost surface, but its applica-
ble range is very limited. Another reason could be that the numer-
ical model used an interpolated value of the frost thermal
conductivity between moderately low temperatures and cryogenic
temperatures.

4.3. Effect of the relative humidity with initial wall cooling

The effect of the relative humidity on frost formation with ini-


tial wall cooling was examined through experiments and numeri-
cal analyses for three relative humidity conditions, i.e. 30, 60, and
90%. Fig. 17 presents frost density from the experiment and calcu-
lation using the correlation to see the effect of the relative humid-
ity on the frost density. The frost density increased as the relative
humidity increased. Fig. 18 shows the effect of the relative humid-
ity on the wall heat flux. As the relative humidity increased, heat
transfer into the wall increased slightly because the portion of
the phase change heat transfer was relatively small compared with

Fig. 15. Effect of the ambient air temperature on the wall heat flux.

humidity. As the frost density became higher, more water vapor


was accordingly required to maintain the frost density. Therefore,
the mass flux contributed to increasing the frost thickness was
reduced at higher temperature because most of mass transfer
was used to increase the frost density.
The numerical results overpredicted the growth rate of the frost
thickness during the cooling process and underpredicted it during
the filling and holdup process. The relative error of the final frost
thickness was less than 13%. The maximum relative error of the
frost thickness was 53% and occurred at 40 min after the experi-
ment began. Even though the final frost thicknesses in the experi-
ments were almost the same in the numerical analyses, the trend
of the experimental and the numerical result was different. The
reason for the difference between the experimental and the
numerical results could be that saturated condition at the frost sur-
face was used in the numerical model. Na and Webb [27,28] con-
tended that numerical model with saturated conditions at the
frost surface overpredicted the mass transfer and supersaturated
conditions should be used. They suggested an empirical equation Fig. 17. Effect of the relative humidity on the frost density.
860 S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861

5. Conclusions

The effects of initial cooling on heat and mass transfer at the


cryogenic surface under natural convection condition were exper-
imentally and numerically investigated. From the experimental
and numerical results, the following conclusions are drawn.

(a) Under natural convection condition, an initial wall cooling


caused frost on the cryogenic surface to grow thicker and
the wall heat flux to be greatly reduced. The experimental
results with and without initial wall cooling were compared
when the relative humidity was 60% and the ambient air
temperature was 20 °C. Without initial wall cooling, the
maximum frost thickness was about 3 mm and the growth
of frost was unstable because frost fell owing to its own
weight and the shear force from natural convective flow.
With initial wall cooling, the maximum frost thickness was
approximately 11.5 mm and the growth of frost was stable.
The maximum heat flux under initial wall cooling was 40% of
that without initial wall cooling, and the minimum heat flux
under initial wall cooling was 52% of that without initial wall
Fig. 18. Effect of the relative humidity on the wall heat flux.
cooling.
(b) Under initial wall cooling, experimental results showed that
other modes of heat transfer. The numerical results showed good
heat transfer into the wall increased as the ambient air tem-
agreement during the cooling process. But they overpredicted the
perature increased because the driving potential for heat
minimum heat flux. The relative error of the maximum wall heat
transfer between the ambient air and the frost surface
flux was less than 6% for all cases, and the relative error of the min-
became larger. The frost thickness decreased slightly as the
imum heat flux was less than 26%.
ambient air temperature increased because most of mass
Fig. 19 shows the effect of the relative humidity on the frost
transfer was used to increase the frost density at high ambi-
thickness. Increasing the relative humidity caused the driving
ent temperature. As the relative humidity increased, heat
potential for mass transfer to become large, and hence the frost transfer into the wall increased slightly because the portion
thickness increased. The numerical results overpredicted the
of the latent heat transfer was relatively small compared
growth rate of frost thickness during the cooling process and with other modes of heat transfer. As the relative humidity
underpredicted it during the filling and holdup process. In test
increased, the frost thickness increased considerably.
number 6, continuous fog formation near the cryogenic surface (c) A numerical model for frost formation accounting for initial
was observed at 40 min after the experiment began. Since then,
cooling was proposed. The transient wall temperature and
the rate of change of the frost thickness decreased. And the mea- modified boundary conditions were considered in the
sured wall heat flux started to deviate from the predicted wall heat
numerical model for the frost formation. Also, a new corre-
flux because the numerical model for the frost formation did not lation for the frost density and a combined model for the
consider the fog formation. Except for test number 6, relative error
frost thermal conductivity were suggested. The proposed
of the final frost thickness was less than 10% and the maximum rel- numerical model could explain the heat and mass transfer
ative error was 20%.
at the cryogenic surface during the cooling process as well
as filling and holdup process. The maximum and minimum
heat flux from the numerical model showed good agreement
with experimental data, within 10% and 25% error, respec-
tively. The final frost thickness from the numerical model
showed good agreement with experimental data, within
13% error, except for one case where mass transfer was
reduced due to continuous fog formation near the cryogenic
surface. Therefore, the numerical model will be very useful
in estimating the heat flux in an uninsulated cryogenic sys-
tem, such as a rocket oxygen tank.

Acknowledgement

This work was supported by the National Research Foundation


of Korea (NRF) grant funded by the Korean government (MSIP) (No.
2012R1A3A2026427).

References

[1] F.E. Ruccia, C.M. Mohr, Atmospheric heat transfer to vertical tanks filled with
liquid oxygen, Adv. Cryog. Eng. 4 (1960) 307–318.
[2] D.C. Holten, A study of heat and mass transfer to unisolated liquid oxygen
containers, Adv. Cryog. Eng. 6 (1961) 499–508.
[3] R.F. Barron, L.S. Han, Heat and mass transfer to a cryosurface in free convection,
Fig. 19. Effect of the relative humidity on the frost thickness. J. Heat Trans. 87 (1965) 499–506.
S. Choi, S.J. Kim / International Journal of Heat and Mass Transfer 112 (2017) 850–861 861

[4] Yatish T. Shah, I. Theory of Frost Formation, Ph.D. Thesis, Massachusetts [21] Max. Kandula, Effective thermal conductivity of frost considering mass
Institute of Technology, 1968, pp. 20–127. diffusion and eddy convection, Spec. Top. Rev. Porous Media 1 (2010) 321–
[5] P.L. Thibaut Brian, Robert C. Reid, Yatish T. Shah, Frost deposition on cold 336.
surfaces, Ind. Eng. Chem. Fundam. 9 (1970) 375–380. [22] Silvia Negrelli, Christian J.L. Hermes, A semi-empirical correlation for the
[6] Biguria Gabriel, Leonard A. Wenzel, Measurement and correlation of water thermal conductivity of frost, Int. J. Refrig. 58 (2015) 243–252.
frost thermal conductivity and density, Ind. Eng. Chem. Fundam. 9 (1970) 129– [23] Randall F. Barron, Cryogenic Systems, second ed., Oxford University Press, New
138. York, 1985 (Chapter 7).
[7] Kyoung-Hoon Kim, Hyung-Jong Ko, Kyoungjin Kim, Yong-Wook Kim, Kie-Joo [24] Y.B. Lee, S.T. Ro, Analysis of the frost growth on a flat plate by simple models of
Cho, Analysis of heat transfer and frost layer formation on a cryogenic tank saturation and supersaturation, Exp. Therm. Fluid Sci. 29 (2005) 685–696.
wall exposed to the humid atmospheric air, Appl. Therm. Eng. 29 (2009) 2072– [25] Y.B. Lee, S.T. Ro, An experimental study of frost formation on a horizontal
2079. cylinder under cross flow, Int. J. Refrig. 24 (2001) 468–474.
[8] Zhongliang Liu, Yuwan Dong, Yanxia Li, An experimental study of frost [26] Ali R. Tahavvor, Mahmood Yaghoubi, Experimental and numerical study of
formation on cryogenic surfaces under natural convection conditions, Int. J. frost formation by natural convection over a cold horizontal circular cylinder,
Heat Mass Transf. 97 (2016) 569–577. Int. J. Refrig. 33 (2010) 1444–1458.
[9] Frank P. Incropera, David P. DeWitt, Fundamentals of Heat and Mass Transfer, [27] Byeongchul Na, Ralph L. Webb, A fundamental understanding of factors
fourth ed., John Wiley & Sons, New York, 1996 (Chapters 9 and Appendix A). affecting frost nucleation, Int. J. Heat Mass Transf. 46 (2003) 3797–3808.
[10] Max. Kandula, Frost growth and densification in laminar flow over flat surface, [28] Byeongchul Na, Ralph L. Webb, Mass transfer on and within a frost layer, Int. J.
Int. J. Heat Mass Transf. 54 (2011) 3719–3731. Heat Mass Transf. 47 (2004) 899–911.
[11] B.W. Jones, J.D. Parker, Frost formation with varying environmental
parameters, J. Heat Transfer (1975) 255–259.
Sangho Choi received the B.S. degree in Mechanical Engineering from the Pohang
[12] Mark A. Dietenberger, A Model for Nocturnal Frost Formation on a Wing
University of Science & Technology (POSTECH), Pohang, Korea. He received the M.S.
Section, NASA Contractor Report 3733, NASA Scientific and Technical
Information Branch, 1983. degrees in mechanical engineering from the POSTECH, Pohang, Korea. He is a Ph.D.
[13] Y. Hayashi, A. Aoki, S. Adachi, K. Hori, Study of frost properties correlating with Candidate in KAIST. He also works as a thermal engineer in Korea Aerospace
frost formation types, ASME J. Heat Transfer 99 (1977) 239–245. Research Institute (KARI).
[14] Christian J.L. Hermes, Robson O. Piucco, Jader R. Barbosa Jr., Claudio Melo, A
study of frost growth and densification on flat surfaces, Exp. Therm. Fluid Sci. Sung Jin Kim received the Ph.D. degree in Mechanical Engineering from the Ohio
33 (2009) 371–379. State University, Columbus, OH, USA, in 1989. He is a Professor with the Depart-
[15] Christian J.L. Hermes, Felipe R. Loyola, Valter S. Jascimento Jr., A semi- ment of Mechanical Engineering, Korea Advanced Institute of Science and Tech-
empirical correlation for the frost density, Int. J. Refrig. 46 (2014) 100–104. nology (KAIST), Daejeon, Korea. Before joining KAIST in 1997, he was been a Group
[16] Max Kandula, Correlation of water frost porosity in laminar flow over flat Leader of the Thermal Engineering Center, IBM, Tucson, AZ, USA, for seven years. His
surfaces, Spec. Top. Rev. Porous Media 3 (2012) 79–87. research group at KAIST held the National Research Laboratory status from 2006 to
[17] J.D. Yonko, C.F. Sepsy, An investigation of the thermal conductivity of frost 2011. He edited a book entitled Air Cooling Technology for Electronic Equipment.
while forming on a flat horizontal plate, ASHRAE Trans. 2043 (1967) I1.1–I1.9. Dr. Kim is a member of the Korean Academy of Science and Technology and an
[18] Matthew Sturm, Jon Holmgren, Max Konig, Kim Morris, The thermal
ASME Fellow. He was a recipient of the Scientific Achievement Award from the
conductivity of seasonal snow, J. Glaciol. 43 (1997) 26–41.
Korea Society of Mathematical Education, the Excellent Teaching Awards from
[19] Mark A. Dietenberger, Generalized correlation of the water frost thermal
KAIST, two Invention Achievement Awards and five Author Recognition Awards
conductivity, Int. J. Heat Mass Transfer 26 (1983) 607–619.
[20] H. Auracher, Effective Thermal Conductivity of Frost, International Symposium from IBM. In 2012, he was awarded a prestigious nine-year grant by Korea’s
on Heat and Mass Transfer in Refrigeration and Cryogenics, Dubrovnik, 1986, Creative Research Initiative to develop flexible and thin thermal superconductors.
pp. 285–302.

You might also like