Sei sulla pagina 1di 12

International Journal of Heat and Mass Transfer 88 (2015) 662673

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


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

Comparison of cryogenic ow boiling in liquid nitrogen and liquid


hydrogen chilldown experiments
Jason Hartwig a,, Hong Hu b, Jeremy Styborski c, J.N. Chung b
a

NASA Glenn Research Center, Propulsion and Propellants Branch, Cleveland, OH 44135, United States
Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611, United States
c
Turbine Durability, Hot Section Engineering, Pratt & Whitney, East Hartford, CT 06118, United States
b

a r t i c l e

i n f o

Article history:
Received 7 December 2014
Received in revised form 29 April 2015
Accepted 29 April 2015
Available online 20 May 2015
Keywords:
Nucleate boiling
Transition boiling
Chilldown
Liquid nitrogen
Liquid hydrogen
Cryogenic heat transfer coefcient

a b s t r a c t
This paper presents a comparison between experimental results from recent liquid hydrogen (LH2) transfer line chilldown experiments at high Reynolds (Re) numbers versus liquid nitrogen (LN2) experiments
conducted at low Re numbers. Parasitic heat leak, inner wall temperatures, inner wall heat uxes, and
heat transfer coefcients are computed to compare between the two systems. Analysis of temperature
traces and ow visualization indicates that the chilldown process evolves much more rapidly at higher
Re numbers due to a quick transition from vapor ow to annular liquid ow and near immediate liquid
contact along the pipe walls. The lower kinematic viscosity and surface tension of LH2, along with
reduced parasitic heat leak and higher Re numbers relative to the LN2 experiments, causes chilldown
to proceed almost immediately into the nucleate boiling regime, in comparison to low Re ows where
>75% of the chilldown is spent in vapor lm boiling.
Published by Elsevier Ltd.

1. Introduction
For decades, cryogenic uids have been used throughout the
chemical, aerospace, medical, and food industries. Cryogenic propellant technology development can be used to enable high performance in-space engines, in-space fuel storage depots, in-space
resource utilization systems, cooling, refrigeration, liquefaction,
thermal management, fuel cells, and life support systems.
However, there are challenging aspects when working with cryogenic liquids such as liquid hydrogen (LH2) due to the thermophysical properties. For example, the low normal boiling point makes
LH2 particularly susceptible to parasitic heat leak, resulting in
costly insulation systems and boil-off of precious propellant. The
low normal boiling point (NBP) and low surface tension make it
difcult to transfer single phase liquid from a storage tank, making
vapor ingestion downstream highly probable. Therefore, before
exible, robust systems can be conceived, fundamental understanding of the uid mechanics and heat transfer is rst required.
To enable all of the aforementioned systems, efcient methods
with which to transfer cryogenic propellant are required in order
to minimize consumption and loss. For example, before restart
of an in-space cryogenic engine [28,29,35], or before propellant

Corresponding author.
E-mail address: Jason.W.Hartwig@nasa.gov (J. Hartwig).
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.04.102
0017-9310/Published by Elsevier Ltd.

transfer from the in-space storage depot tank to the customer


receiver tank [38,21], the transfer line connecting the propellant
storage tank and engine or the line connecting depot storage and
customer receiver tanks must be rst be chilled down quickly
and efciently. Chilldown is dened as the process of cooling propellant tank and transfer line hardware down to cryogenic temperatures so that vapor-free liquid may ow between two points.
Chilldown therefore represents the rst stage of cryogenic uid
transfer. In-space engines require vapor-free propellant ow to
avoid combustion instability issues during restart, and in-space
depots will require vapor-free liquid ow in order to achieve very
high liquid ll fractions in the customer spacecraft receiver tank.
Before continuous, steady, vapor-free propellant liquid may ow,
the transfer line must rst be chilled down, therefore necessitating
the desire to understand the underlying ow boiling and heat
transfer associated with chilldown.
The purpose of this paper is to compare the fundamental
physics of two different sets of chilldown experiments conducted
at two different Reynolds (Re) regimes using two different uids
to understand how chilldown is affected by ow regime and
thermophysical properties. The paper is structured as follows: a
brief review of relevant literature is presented, along with background into cryogenic transfer line chilldown. Next, descriptions
of both low and high Re number experimental hardware are given.
Then experimental results are compared between systems.
Temperature traces and ow visualization, along with computed

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

663

Nomenclature
area [m2]
specic heat of transfer line [J/kg K]
pipe diameter [m]
time interval to steady state chilldown [s]
view factor, dimensionless
mass ux [kg/m2 s]
gravity [m/s2]
enthalpy [kJ/kg]
heat transfer coefcient [W/m2 K]
thermal conductivity [W/m K]
characteristic length scale [m]
mass of transfer line [kg]
mass ow rate [kg/s]
pressure [Pa]
ow energy [J]
stored line energy [J]
parasitic heat leak energy [J]
gas conduction heat leak [W]
radiation heat leak [W]
sensor heat leak [W]
solid conduction heat leak [W]
axial heat ux [W/m2]
inner wall convective heat ux [W/m2]
heat ux at the wall [W/m2]
radius [m]
Reynolds number

A
cP
d
dtchill
F
G
g
h
hi
k
L
m
_
m
P
Qow
Qline
Qparasitics
Q_ gascond
Q_ rad
Q_ sensor
Q_
solidcond

q00axial
q00i
q00w
r
Re

T
t
We

temperature [K]
time [s]
Weber number, dimensionless
thermal diffusivity [m2/s]
surface tension [N/m]
emissivity, dimensionless
efciency parameter, dimensionless
viscosity [Pa s]
density [kg/m3]
StefanBoltzmann constant [W/m2 K4]
initial state
cross-sectional
critical heat ux
exit
uid to gas
gas
inner surface
inlet
jth surface
liquid
Leidenfrost
outer surface
saturation
solid surface
steady state
vapor

a
c
e

g
l
q
r
0
c
CHF
e
fg
gas
i
in
j
l
Lei
o
SAT
solid
SS
v

parasitic heat leak, inner wall temperatures, inner wall heat uxes,
and heat transfer coefcients are used to compare chilldown characteristics for the two different ow regimes. Finally, both data sets
are compared with empirical correlations used to predict the critical heat ux and Leidenfrost temperature.

and Q_ sensor is the measured electrical heat input from the pressure
transducers.
Qow is the ow energy stored in the uid available to combat
the other two energy terms:

Q flow

2. Background and governing physics

where Qline is the thermal energy stored in the line:

Q line

t ss

mcP

t0

dT
dt
dt

Qparasitics is the sum of total unwanted heat leak into the system,
which includes radiation, solid and gaseous conduction (assuming
system is in a partial or full vacuum environment), and heat leak
due to instrumentation:

Q parasitics

t SS



Q_ rad Q_ solidcond Q_ gascond Q_ sensor dt

t0

where

Q_ rad

erAi F ij T 4i  T 4j

_ e  hin dt
mh

t0

The purpose of line chilldown is to cool hardware down to cryogenic temperatures so that single phase cryogenic liquid may ow
between two points in a system. An energy balance for any transfer
line system, which includes all piping, valves, pumps, etc. can be
expressed in the following form:

Q flow Q line Q parasitics

t SS

dT
Q_ solidcond ksolid Ac
dz

dT
Q_ gascond kgas
Asurface
dx

For a subcooled liquid in the storage tank, both the latent and
sensible energy is available for line chilldown:

Q flow

t SS



_ hfg h  hSAT P dt
m

t0

By denition, in order to achieve perfect vapor-free liquid at the end


of the transfer line, the subcooled portion (sensible energy in Eq.
(8)) of the ow energy must exceed the sum of line and parasitic
energies absorbed by a uid element.
The combination of cryogenic uid ow and warm hardware
implies that there will be two phase ow, vigorous boiling, and
heat transfer as the chilldown process evolves in time. The uid
mechanics and heat transfer of cryogenic transfer line chilldown
can be accurately mapped using a boiling curve, which is a
log/log plot of heat transfer as a function of the temperature
above saturation. Shown in Fig. 1 is a typical saturated boiling
curve for water. There are four distinct heat transfer regions separated by three characteristic points. Because the system is initially at a warm ambient temperature, which is signicantly
warmer than the cryogenic uid saturation temperature, chilldown generally begins in the vapor lm boiling regime, due to
violent ashing of the liquid. As the system chills, and as wall
temperature decreases, the system approaches the Leidenfrost
point, or point of minimum heat transfer between cold uid
and warm pipe. Heat transfer is a minimum at the Leidenfrost
point due to inefcient heat transfer between cold vapor and
wall. The next region is known as the transition boiling regime,
or regime between liquid dominated chilldown in nucleate

664

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

Fig. 1. Characteristic saturated chilldown curve for water.

boiling, and gas dominated chilldown in vapor lm boiling. As


the wall temperature decreases, the system quickly approaches
the critical heat ux, or point at which liquid begins to contact
the wall. The next region is characterized by liquid dominated
nucleate boiling regime. As the wall further cools, and as the
void fraction decreases, the system approaches the onset of
nucleate boiling (ONB) point, characterized as the point at which
the system evolves from nucleate two phase cooling to single
phase liquid convection. Heat transfer is a maximum at the
point of critical heat ux due to the highly efcient cooling process of boiling liquid through the use of sensible and latent heat
and due to the fact that the insulating vapor layer is absent
while the temperature difference between wall and uid is the
greatest. It is interesting to note that the evolution of chilldown
generally mimics a reverse boiling curve. The Leidenfrost point,
point of critical heat ux, and ONB are all characteristics points
for each uid, but the general shape of the curve is the same.
There are numerous studies on room temperature or storable
uid ow boiling and heat transfer such as water [1820] and carbon dioxide [23]. Previously reported relevant studies on cryogenic
boiling, heat transfer, and chilldown are listed as follows: Westbye
et al. [39] conducted laminar and turbulent horizontal 1-g chilldown tests using R113. Kawanami et al. [16,17], Yuan et al.
[41,42], Zhang and Fu [44], and Hu et al. [9,10], conducted various
horizontal, upward, and downward 0 g and 1 g laminar and turbulent tests with LN2. Heat transfer coefcients for LN2 are available
in the literature for pool boiling [2,11], forced convection with

subcooled liquid [36], forced convection with nitrogen at supercritical pressures [37], and lm boiling [6]. Additional ow boiling
characteristics are detailed in the recent papers, see for example
[9]. Flow boiling information for liquid oxygen (LOX) is available
in [30].
Meanwhile, the only known relevant chilldown experiment
using LH2 was the work from [3], for horizontal 1-g turbulent ow.
They did not however report information regarding ONB,
Leidenfrost point, or ow boiling heat uxes. Pool boiling heat
uxes are available in [32,7,24], as well as some ow boiling characteristics for LH2 are available in the literature [33,34]. However
the heat ux data does not match the well-known theory from
[27], which predicts that bulk uid velocity magnitude has little
effect on the critical or maximum heat ux value; data in [34]
show a dependence on velocity and do not converge.
Additionally, two separate methods were used to deduce the mass
ow rate, one using a weight scale, and one using the gas ow rate
into the liquid supply dewar. While the weight scale can be used as
an accurate measure, provided that tests are run for a very long
time (long enough to establish a steady state velocity), using the
gas ow rate into the liquid dewar introduces complications due
to condensation heat transfer across the tank liquid/vapor interface, due to compressibility effects, and due to potential solubility
of the pressurant gas into the liquid. Using the pressurant gas ow
rate is only accurate in the limit of an isentropic tank drain, where
heat transfer between pressurant gas and liquid cryogen is
minimized.

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

3. Experimental setup
3.1. LN2 experiments
Low Re number liquid nitrogen line chilldown experiments
were performed at the University of Florida. Specic details regarding experimental hardware and procedure are available in [9].
Fig. 2(a) and (b) shows a photo and schematic of the test setup,
respectively. The system was designed to t inside an apparatus
frame to minimize line lengths and thus parasitic heat leak as
shown in Fig. 2(a). The test section was a 1 cm outer diameter,
0.1 cm thick, 25.4 cm long see-through Pyrex section. The inlet
condition was saturated ow, for quality between 0 and 0.122.
Re number range was 10004000. Gaseous nitrogen was used to
pressurize and drain the LN2 storage dewar. A three way valve

665

was used to prechill a portion of the line leading up to the test section. The ow was then routed into a vacuum section which contained the sight glass used for ow visualization. Flow could be
routed vertically upward or downward or horizontally. Flow was
then routed through a series of heat exchangers designed to completely vaporize the uid before entering the gas ow meter and
vent line. A vacuum pump was used to pump down the pressure
on the sight glass section to 23 kPa.
Temperature instrumentation is shown in Fig. 3. A set of three
thermocouples were mounted at three different stations along
the outer test section, spaced axially apart 120 to measure axial
and longitudinal temperature variations as the system chilled in.
Pressure was measured inside the liquid dewar and upstream
and downstream of the test section. A high speed camera was used
to visualize the ow to time correlate with the temperature and

Fig. 2. (a) Photo and (b) test schematic for the low Reynolds number liquid nitrogen line chilldown test setup.

666

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

Fig. 3. Location of temperature sensors for the low Reynolds number liquid nitrogen experiments.

Fig. 4. High Reynolds number liquid hydrogen line chilldown test setup.

pressure data. Vacuum pressure was measured with a low pressure


vacuum gauge. Mass ow rate was controlled by the pressurant gas
ow.

used to ensure liquid ow out of the bottom of the tank. Flow


was immediately routed through a Coreolis ow meter (FM),
which was close coupled to the tank to minimize two phase ow
readings. Liquid was then routed to a manifold, where a series of
valves and orices were used to control ow rate. Flow was routed
vertically upward through a dummy valve, which was used to
add mass to the transfer line. The ow control manifold could deliver LH2 ow rates over a wide range corresponding to a wide range
in liquid Re numbers (1.84  1044.33  105). The inlet condition
was saturated liquid. The vertical upward direction was chosen
to mimic axisymmetric heat transfer into the pipe as in the microgravity pipe ow case [40,43]. Liquid then passed through a sight
glass section for two phase ow visualization as the chilldown process evolved. Flow was nally routed up out of the top of the vacuum chamber, through approximately 30 m of piping through a
heat exchanger and eventually to the vent stack located outside
the facility. The outer diameter of the transfer line was 1.27 cm
(0.5 in) and the inner diameter was 1.02 cm (0.402 in).
The whole tank and line chill down assembly rested inside of a
vacuum chamber set to a background pressure of 1  106 torr and
250 K to simulate a cold solar inertial orbit of Low Earth Orbit
(LEO). The entire line assembly was constructed from 1.27 cm
(0.5 in) outer diameter (OD) stainless steel (SS), while the test section for ow visualization was made of Pyrex glass. The portion of
the line assembly between tank and valve manifold was always in
liquid, since there was no control valve between LAD and line
assembly. This is deemed the cold portion of the line, while all
piping and components downstream of the control valves is
deemed the warm portion of the line. The length and mass of
the cold portion of the assembly was approximately 2.34 m and
5.1 kg while the warm portion including the two valves, piping,
and sight glass was approximately 2.41 m and 2.3 kg.
Flow rate was measured using a Coreolis FM. Pressure was measured in the tank, upstream of the manifold, and upstream and
downstream of the sight glass. Stream temperatures were measured in the storage tank, upstream of the valve manifold (SD15),
and downstream of the sight glass (SD23), while skin temperatures
were measured on the SS pipe (SD17, SD18) and Pyrex sight glass
tube (SD19). A high speed camera was used to visualize ow to
correlate with the pressure and temperature data. Uncertainties
for all these measurements are listed in Hartwig et al. [8].

3.2. LH2 experiments


4. Results and comparison
High Re number liquid hydrogen line chilldown experiments
were performed at the NASA Glenn Research Center Small
Multi-Purpose Research Facility (SMiRF). Specic details regarding
experimental hardware and procedure are available in [26] only
general details required to understand the experimental results
in the current work are presented here. A schematic of the test
setup is shown in Fig. 4. A 1.84 m3 (487 gallon) storage vessel
was used to store the LH2 and condition the liquid to the desired
initial saturation temperature between 20.3 K < T < 24.2 K. A horizontal liquid acquisition device (LAD) located inside the tank was

4.1. Temperature traces and ow visualization


Fig. 5 plots a typical outer wall temperature versus time trace
from the LN2 chilldown tests from [9]. Still images from the high
speed camera are shown to correlate with the temperature/time
curve. Also shown in the gure is the quench front movement during the transition boiling regime. As shown, at lower Re numbers
and higher parasitics, most of chilldown is spent in the lm boiling
regime. Then, a dispersed droplet ow comprised of small

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

667

Fig. 5. Outer wall temperature trace and ow visualization for liquid nitrogen line chilldown test for TSAT = 79 K, 0.0044 kg/s, Re = 3085.

spherical liquid drops embedded in the vapor is observed. With a


further wall temperature decrease, the small droplets become denser and coalesce to form some larger and stretched droplets. After
that, a continuous liquid core occurs which ows in the center of
the tube and is surrounded by a thin layer of vapor lm. This phenomenon is called the inverted annular ow. At this time, the lm
is thick enough to separate the wall and liquid core efciently.
With a further decrease of the wall temperature, the vapor lm
becomes thinner while the liquid core occupies more area. When
the wall temperature is reduced to the rewetting temperature,
the liquid is able to come into direct contact with the tube wall.
Some short liquid laments can be seen. The transient boiling,
characterized by intermittent liquid-wall contact and violent bubble generation, is observed during a limited period. The quenching
front occurs at the point of maximum heat ux. The vaporliquid
interface is highly wavy because of the KelvinHelmholtz instability. At this time the heat transfer rate from the hot wall to the liquid is signicantly increased and a rapid drop of the wall

temperature slope can be observed. A rapid change from transition


ow to the dispersed bubbly ow will occur directly after the
quenching front. The prevailing boiling regime after the quenching
front passes is the nucleate boiling with clear small bubbles generated at the wall. Finally, the system achieves saturated LN2 temperatures and bubbly ow.
Fig. 6 plots outer wall temperature versus time traces from the
LH2 chilldown experiments for saturated 20.3 K liquid for a medium ow test (0.014 kg/s, Re = 1.22  105). Superimposed in the
plot is the inner wall stream temperature reading immediately
downstream of the sight glass as well as the liquid mass ow meter
reading as a function of time. Also plotted are time correlated video
clips from the sight glass section. As shown, at higher Re numbers
and lower overall parasitic heat leak, the outer wall skin temperatures immediately plummet within the rst 20 s of chilldown. Both
skin diodes bottom out in succession. Time correlated ow visualization reveals that this is due to liquid droplets forming along the
walls almost instantaneously from the start of chilldown. After the

668

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

Fig. 6. Outer wall temperature trace, stream temperature, and ow visualization for liquid hydrogen line chilldown test for TSAT = 20.3 K, 0.014 kg/s, Re = 1.22  105.

skin diodes bottom out near the saturation temperature, the


majority of the rest of the chilldown is spent transitioning from
annular to slug to bubbly ow. The stream temperature reading
shown in red eventually reads saturated liquid hydrogen temperatures after 155 s. Slight, temporary increases in internal stream
temperature readings are attributed to transitions in ow proles
from vapor to droplet and from annular to churn/bubbly ow [26].
As shown in Fig. 6, most of the chilldown is spent in annular and
churn/bubbly ow as the system reaches an equilibrium between
parasitic heat leak and ow energy. Since parasitics for this experiment were slightly higher than maximum ow energy, perfect

single phase liquid was not achieved for every test [26].
Nonetheless, because the system experienced full vacuum conditions, and because parasitics were signicantly smaller here versus
the LN2 tests (see Section 4.2), liquid droplets form along the wall
almost immediately. This leads to signicantly faster chilldown
times relative to the lower Re LN2 case.
4.2. Inner wall temperature and inner wall heat ux
To generate reverse boiling curves, the method of Burgraff [4]
can be used to compute inner wall temperature from the measured

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

outer wall temperature. With knowledge of the parasitic heat leak,


inner wall temperature can then be used to compute the inner wall
heat ux. The inner wall temperature is determined from a truncated series solution using the assumption of adiabatic outer wall:

  
 
r 2o
ri
ri
dT o
 1  2 ln
2
4a
ro
ro
dt

 
 
 2
2
2
 r r
r2 r2 d T o
1  4
ri
r4
ri
 o 2 ln
o i2
r i  5r 4o  o 2i ln

2
2
64a
8a
ro
16a
ro
16a
dt

Ti To

9
The inner and outer radii are known, as are thermal properties
of the metal and Pyrex. Outer wall temperatures from the data are
fed into Eq. (9) to obtain inner wall temperature as a function of
time. Inner wall heat ux is then obtained from the Burgraff
method (1964). Total convective heat ux is then computed by
using the heat conduction equation:

q00i q00w q00axial


r o  00
q q00solidcond q00gascond
r i rad

10

where the axial term is assumed to be negligible [16,17]. Table 1


lists the estimated parasitic heat leaks incurred during the high
Re LH2 chilldown experiments from [26] and the low Re LN2 chilldown experiments from [9], along with associated uncertainties.
The cold section of the LN2 tests was the portion contained within
the vacuum section while the warm section was the portion of the
line in between liquid storage dewar and vacuum section. For LH2
tests, the cold section was the horizontal portion of the transfer
line assembly that always contained liquid; the warm section
was the portion of the line initially at 250 K. The total estimated
heat leak for the entire LN2 and LH2 line chill assemblies from storage dewar to a point downstream of the sight glass is estimated to
be 175 W and 19 W, respectively. Clearly, the total parasitic energy
into the system over the duration of the test is heavily dependent
on the total chill down time. For a typical test, at the point of steady
state, the average parasitic heat leak energy into the LN2 line chill
assembly was approximately 25.5 kJ while the average parasitic
heat leak into the LH2 line chill assembly was 6 kJ. Uncertainties
for the LN2 parasitics are listed in Table 1. Uncertainties for the
computed inner wall heat uxes for the LH2 tests are estimated in
[8].
Fig. 7 plots two reverse boiling curves for the low Re LN2 experiments (Re = 3500) from [9] (Fig. 7(a) and (b)) versus the high Re
LH2 experiments (Re = 3.11  105) from the current work
(Fig. 7(c) and (d)). Fig. 7(a) and (b) show that the LN2 tests traversed the reverse boiling curve through vapor lm to transition

Table 1
Estimated parasitic heat leaks at steady state for liquid nitrogen and liquid hydrogen
chilldown tests.
Liquid nitrogen tests

Warm line
radiation
Cold line
radiation
Sensors
Solid
conduction
Warm line gas
conduction
Cold line gas
conduction
Total

Liquid hydrogen tests

Uncertainty
(%)

Heat
transfer rate
[W]

Uncertainty
(%)

2.65

11

10.91

15

0.33

17

5.69

10

0
0.05

0
35

1.91
0.05

15
40

0.02

15

10

0.01

15

10

18.59

Heat
transfer rate
[W]

172
0.3
175.33

669

to nucleate boiling regimes. The maximum heat ux measured at


the sight glass was roughly 8000 W/m2, and inner wall temperature
dropped to near LN2 boiling temperatures after 2 min.
Fig. 7(c) and (d) show that the high Re LH2 tests enter immediately into transition boiling, and that maximum heat ux is
achieved within seconds. The immediate entrance into transition/nucleate boiling is shown for all three skin diode locations
on the transfer line. Flow visualization during testing substantiates
this claim, because liquid droplets were observed along the sight
glass wall almost immediately after the start of the test. Inner wall
temperatures follow outer wall temperatures by dropping rapidly
to LH2 saturation temperatures. At higher ow rates, the maximum
heat ux through the stainless steel pipe is almost 150,000 W/m2,
and the maximum inner wall heat ux at the sight glass is about
40,000 W/m2. Heat ux is higher at upstream locations due to
lower quality and a greater difference between stream and wall
temperature upstream. Comparing the heat ux at SD17 and
SD18 to the heat ux at SD19, heat ux at the stainless steel piping
is signicantly higher over the Pyrex test section.
The heat ux at SD19 suggests that ow at the sight glass may
be close to lm boiling at the start of the test. This correlates with
the visualization of vapor ow at the sight glass at the beginning of
each test. The high heat uxes at SD17 and SD18 at the beginning
of the test imply immediate liquid contact on the pipe upstream of
the sight glass. Therefore higher number Re ows using LH2 are
associated with an immediate entrance into transition boiling, faster chilldown, and higher maximum inner wall heat uxes relative
to low Re number LN2 ows.
4.3. Heat transfer coefcient
Heat transfer coefcients at the inner wall can now be computed from knowledge of the temperature difference between
inner wall and local saturation temperature, and knowledge of
the inner wall heat ux from Eq. (10):

hi

q00i
T i  T sat

11

A more accurate value of heat transfer coefcient could be


obtained by using the stream temperature, however, the use of saturation temperature is widely accepted in the two phase ow community. Uncertainties in the computed heat transfer coefcients
are given in [8].
Fig. 8(a) and (b) compare the heat transfer coefcient obtained
from the low Re number (Re = 3500) LN2 tests conducted by Hu
et al. [9] with that obtained from the high Re number
(Re = 3.11  105) LH2 ows from the current experiment. Similar
to the boiling curve plots, the LN2 heat transfer coefcient plot is
characterized by two local inection points corresponding to the
Leidenfrost point and point of maximum heat ux. Heat transfer
coefcient reaches a maximum at about 200 W/m2 K for low Re
ows; for the LH2 experiment, heat transfer coefcient was calculated at SD17, SD18, and SD19. Similar to the boiling curve plots for
LH2, only one maximum occurs, associated with the point of maximum heat ux at the change from transition to nucleate boiling.
Heat transfer coefcient is again higher for upstream diodes. This
is due to the higher heat uxes associated with upstream locations
due to lower quality ow and larger temperature gradient between
wall and uid. This result is supported by Westbye et al. [39] and is
reportedly due to greater subcooling and liquid content upstream.
Heat transfer coefcient calculated at metal tube sections dropped
near the boiling point of LH2 while the heat transfer coefcient at
the sight glass dropped at roughly 50 K. This behavior is attributed
to the lower thermal conductivity of Pyrex versus SS, which is on
average 6% of the SS value across the full temperature range
incurred during chilldown experiments. Heat transfer coefcient

670

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

300

300

(b)
Inner Wall Temperature [K]

Inner Wall Temperature [K]

(a)
250

200

Film Boiling

150

Transition Boiling

100

Nucleate Boiling

50

50

100

250

200

150

100

50

150

1000

2000

5000

6000

7000

8000

300

(c)

250

Inner Wall Temperature [K]

Inner Wall Temperature [K]

4000

Inner Wall Heat Flux [W/m ]

300

Transition Boiling

200
Nucleate Boiling

150

SD17
SD18
SD19

100
50
0

3000

Time [s]

20

40

60

80

100

120

140

200
150
100
SD17
SD18
SD19

50
0

160

(d)

250

50000

100000

150000

200000

250000

Inner Wall Heat Flux [W/m ]

Time [s]

Fig. 7. (a) LN2 inner wall temperature prole, (b) LN2 reverse boiling curve, (c) LH2 inner wall temperature prole, and (d) LH2 reverse boiling curve. Re = 3500 for LN2 test and
Re = 3.11  105.

2000

Heat Transfer Coefficient [W/m K]

150

100

50

50

100

Wall

150

-T

Sat

200

(b)

(a)

Heat Transfer Coefficient [W/m K]

200

250

SD17
SD18
SD19

1500

1000

500

50

[K]

100

Wall

150

-T

Sat

200

250

[K]

Fig. 8. Heat transfer coefcients for (a) Low Re LN2 tests and (b) High Re LH2 tests. ReD = 3500 for LN2 test and Re = 3.11  105 for the LH2 test.

across the metal piping peaked at about 1000 W/m2 K while heat
transfer coefcient across the Pyrex sight glass reached roughly
700 W/m2 K.
5. Discussion of results

along the wall earlier in the chilldown process. At higher Re, liquid
droplets gain momentum and cause violent mixing with the vapor
lm, negating the traditional inverted annular ow regime seen in
many other uids. Therefore, a higher Leidenfrost temperature is
also expected due to the high ow rate conditions.

5.1. Effect of Reynolds number

5.2. Property effect

Obviously, higher Reynolds numbers will equate to higher heat


transfer between a warm tube and cold cryogenic uid. Due to the
high velocity, the LH2 test has higher Reynolds number, which
causes more vortices in the mean ow, decreasing the thickness
of the boundary layer and further increasing the boiling heat transfer coefcient. Therefore, a shorter chilldown time is expected. The
high velocity also increases the chance of liquid droplets forming

Comparison of the boiling curves in Fig. 7 between nitrogen and


hydrogen shows that hydrogen nearly instantaneously bypasses
the lm boiling regime at high Re, with only a small fraction of time
spent in transition boiling. Meanwhile, a lm boiling regime is present in all of the nitrogen data. The cause for the discrepancy in
behavior between uids is attributed to the special properties of
hydrogen. First, the liquid/vapor density ratio for hydrogen is

671

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

signicantly smaller than that of nitrogen. Second, the surface tension of hydrogen is nearly an order of magnitude smaller than that
of nitrogen. Third, the kinematic viscosity of hydrogen is less than
nitrogen. Whereas nitrogen can maintain a liquid core structure
due to large differences in properties between vapor and liquid,
ow visualization from LH2 chilldown tests showed that high Re
hydrogen ows did not support the inverted annular ow structure.
The higher surface tension of nitrogen may act to divide the interfaces whereas hydrogen may act like two gases mixing together
so that transition boiling is inevitable. Therefore, to some extent,
the ow boiling of hydrogen at higher Re can be regarded as
homogenous ow. Obviously, due to the limited sampling rate of
the data acquisition system, limited resolution of the Burggraf
method, and high Re ow, the Leidenfrost temperature cannot be
discerned in any of the LH2 chilldown data. Fourth, for the same
mass ux, the ratio of Weber numbers for hydrogen over nitrogen
is 53; the ratio of Froude numbers is 12. Both of these
non-dimensional numbers have a fundamental impact on the two
phase ow behavior and transition between different ow regimes.

Based on the energy balance from Eqs. 18, Shaeffer et al. [31]
introduced an efciency parameter to gauge the performance of
different chilldown methods:

Q line
Q
R tSS line
Q flow
_ fg dt

mh
t0

0:25

clv gql  qv
q2v

5.4. Critical heat ux


The critical heat ux (CHF) can also be used to compare the two
experiments. The CHF is a point of maximum heat ux between
wall and uid between transition and nucleate boiling. Over the
past half century, numerous correlations have been proposed for
calculating the CHF. Zuber [45] proposed the following model for
stationary uid, pool boiling, which holds for when the vaporliquid interface of the escaping vapor passage becomes unstable due
to the Helmholtz instability:

clv gql  qv
q2v

0:25

14

For ow boiling, the ground work was produced by Kattos


group (19801987). The correlation format proposed by Kattos
group is also based on the theory of the Helmholtz instability
on the interface between vapor and liquid. They also assumed
that the CHF appears when the heat from the heated surface
exactly balances the latent heat of the total evaporation of the
liquid owing into the liquid lm layer, and the density ratio
between vapor and liquid is sufciently smaller than unity.
Katto and Kurata [12] and Katto and Ohno [13] proposed the following functional relationship for the ow boiling CHF with no
inlet subcooling:

q00CHF 0:186Gl hfg

qv
ql

0:559

We0:264
l

15

where We is the Weber number dened as:

16

This correlation is backed by a large set of experimental data for


multiple uids (water, refrigerants, liquid helium) as well as
geometries (tube, at plate) [14,15]. Mudawar and Maddox [25]
further modied the CHF correlation to explicitly handle geometrical effects:

12

The calculated efciency for the low Re trickle ow LN2 experiments from [9] is less than 2%, which is comparably less than the
calculated efciencies for the pulse LN2 chilldown experiments
from [31]. Differences in efciencies between the two LN2 experiments is attributed to the different range of Re numbers, due to
higher efciencies in pulse over trickle ows, and due to improvements in mitigating parasitic heat leak.
Based on a similar energy balance, an efciency parameter
based on actual uid inlet and exit enthalpies was created for the
LH2 tests to compare performance between the different chilldown
methods, inlet liquid temperatures, and mass ow rates, and those
results are presented in [8]. For the 20.3 K saturated LH2 experiments, efciencies ranged from 0.1 to 0.56. This is a full order of
magnitude higher efciency over the low Re number trickle ow
tests from [9] and mid-range Re pulse ow tests from [31].
Higher efciencies for the LH2 tests over the LN2 tests are attributed to the lower parasitic heat leak and higher ow rates, which
causes turbulent mixing of the uid and faster chill down.

q00CHF 0:131qv hfg

Wel G2l L=ql clv

5.3. Energy analysis

ghfg

q00CHF 0:149qv hfg

q00CHF 0:161Gl hfg

qv
ql

15
23

8
23

Wel

 231
L
d

17

The above four correlations were applied to the low Re number


LN2 tests from [9] and the high Re LH2 tests from [8]. Table 2 lists
results. For the LN2 tests, correlations were applied to the test section inside the vacuum section; for the LH2 tests, correlations were
applied to the point just upstream of the sight glass, at P3 in Fig. 4
for the 24.2 K saturated high ow case.
Regarding the LN2 tests, all four correlations over-predict CHF
values. One reason for this is attributed to the low level of vacuum leading to relatively large heat leakage. The heat leakage
will introduce some vapor into the bulk ow which leads to a
relatively high quality saturated ow boiling rather than subcooled ow boiling. Furthermore, the increasing quality accelerates the bulk ow due to the buoyancy force, which increases
the ability to sustain a continuous liquid ow on the tube wall.
The phenomenon is also reinforced by the pressure oscillation
during the intensive boiling. Discrepancies between measured
and computed CHF may also be attributed to thermal hysteresis;
the CHF is lower during quenching than heating on a given
surface.
Compared to the measured heat ux for LH2 tests, the Zuber
[45] and Lienhard and Dhir [22] correlations underpredict and
the Katto group over predicts the data. (19801987). The Zuber
and Lienhard and Dhir correlations are for pool boiling whereas
experiments here are for ow boiling where it is expected to have
higher heat transfer rates with the strong vortex convection.
Kattos correlation was derived assuming external ow over a
heated plate whereas experiments here were for internal pipe ow.
Meanwhile the Mudawar and Maddox [25] correlation was validated for 0:0095 < qqv < 0:0102; density ratio for LH2 tests was
l

13

This pioneering correlation is based on the two-phase hydrodynamic instability for the boiling on an innite, upward-facing, horizontal at plate. Lienhard and Dhir [22] rened Zubers model by
assuming that the Helmholtz instability wavelength is equal to the
Taylor instability wavelength:

0.04, much higher than the validation range. Another cause for discrepancy between Mudawar and Maddox and the LH2 data is
attributed to the assumption that the two-phase ow must have
a steady vapor-blanket for the correlation to be valid. Flow visualization from Fig. 6 shows that annular ow is achieved relatively
quickly, and that the system spends little time in vapor lm
cooling.

672

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673

Table 2
Comparison of measured vs. calculated critical heat ux (CHF) for the LN2 and LH2
chilldown experiments.
LN2 experiment
Heat ux [kW/m2]
Measured maximum heat ux
Model
Zuber [45]
Lienhard and Dhir [22]
Kattos group (19801987)
Mudawar and Maddox [25]

LH2 experiment
Heat ux [kW/m2]

13.8

233.4

162
182
92
24.3

87.2
99.2
377.4
133.9

Table 3
Comparison of measured vs. calculated Leidenfrost temperatures for the LN2 and LH2
chilldown experiments.

Model
Berenson [1]
Carey [5]

LN2 experiment
Leidenfrost T [K]

LH2 experiment
Leidenfrost T [K]

170.2

143.8
144.5

60.8
287

the two uids. Analysis and discussions in the current work show
that the chilldown process evolves much more rapidly in hydrogen
over nitrogen due to lower density ratio, surface tension, kinematic
viscosity, and due to the quick transition from vapor to annular
ow. Meanwhile lm boiling is much more pronounced in nitrogen
chilldown tests due to the higher density ratio, higher surface tension, and higher kinematic viscosity. Second, higher ow rate LH2
chilldown tests yield higher efciencies than the corresponding
lower ow rate LN2 experiments; discrepancies are primarily
attributed to differences in parasitic heat leak. Third, the critical
heat ux correlations used to compute the maximum heat ux
are shown to over-predict the LN2 data, but generally
under-predict the LH2 data. Lastly, correlations for predicting
Leidenfrost temperatures agree reasonably well with LN2 data
but cannot be used to compare with the LH2 data. Results here
have direct implications on the design of all future cryogenic propellant transfer systems.

Conict of interest
None declared.

References

5.5. Leidenfrost point


Another critical parameter for the chilldown process is the
Leidenfrost temperature, or rewet temperature. Based on the
Taylor wave action, Berenson [1] derived the following formula
to calculate the Leidenfrost temperature:

T lei;B  T sat DT

q hfg gql  qv lv
0:127 v
kv
ql qv 2

!0:5 

g 0 clv
gql  qv

0:33
18

Carey [5] developed a model of Leidenfrost phenomenon based


on liquid microlayer evaporation:

"s
#0:6
kl ql cpl
hfg
T lei;H  T lei;B
0:42
T lei;B  T l
kw qw cpw cpw T lei;B  T sat

19

Leidenfrost temperatures are computed using these two models


and compared with the actual experimental value; results are presented in Table 3. Note from Fig. 7(c) and (d) that the LH2 tests
never passed through the Leidenfrost point, so there is no experimental data to compare with the theory. Both correlations predict
reasonably well with the LN2 measured value. Meanwhile,
Berensons [1] correlation predicts a low Leidenfrost temperature
for LH2 tests. Differences between the correlations are attributed
to the fact that Berensons correlation was determined for transition pool boiling over a horizontal surface whereas the data was
taken for internal, forced two-phase ow. Meanwhile, Careys [5]
correlation interestingly predicts a Leidenfrost temperature above
the warmest initial temperature of 250 K for the LH2 tests. Because
the vacuum chamber wall for the LH2 tests was set to 250 K ambient, there is no way to substantiate or negate this correlation.
6. Conclusion
A systematic comparison has been made between two different
chilldown experiments conducted in two different uids having
different ow regimes. Analysis of the temperature traces, ow
visualization, heat ux, parasitic heat leaks, heat transfer coefcients, differences in properties, critical heat ux, and Leidenfrost
points indicates stark differences in chilldown behavior between

[1] P.J. Berenson, Transition Boiling Heat Transfer from a Horizontal Surface,
Technical Report 17, Massachusetts Institute of Technology, March 1, 1960.
[2] L. Bewilogua, R. Knoner, H. Vinzelberg, Heat transfer in cryogenic liquids under
pressure, Cryogenics 15 (1975) 121125.
[3] J.A. Brennan, E.G. Brentari, R.V. Smith, W.G. Steward, An Experimental Report
on Cooldown of Cryogenic Transfer Lines, NBS Report 9264, Boulder CO,
November 7, 1966.
[4] O.R. Burgraff, An exact solution of the inverse problem in heat conduction
theory and applications, ASME J. Heat Transfer 86 (1964) 373380.
[5] V.P. Carey, Liquid Vapor Phase Change Phenomena: An Introduction to the
Thermophysics of Vaporization and Condensation Processed in Heat Transfer
Equipment, second ed., Taylor and Francis, New York, 2008.
[6] U.H. Glahn, A Correlation of Film-Boiling Heat-Transfer Coefcients Obtained
with Hydrogen, Nitrogen, and Freon 113 in Forced Flow, NASA-TN-D-2294,
May 1, 1964.
[7] RW. Graham, R.C. Hendricks, R.C. Ehlers, An experimental study of the pool
heating of liquid hydrogen in the subcritical and supercritical pressure regimes
over a range of accelerations, Adv. Cryog. Eng. 10 (1965) 342352.
[8] J.W. Hartwig, J. Styborski, B. Stiegemeier, H. Hu, E. Rame, J.N. Chung, Liquid
hydrogen transfer line chilldown experiments. II. Analysis, Int. J. Multiphase
Flow, (in press).
[9] H. Hu, J.N. Chung, S.H. Amber, An experimental study on ow patterns and
heat transfer characteristics during cryogenic chilldown in a vertical pipe,
Cryogenics 52 (2012) 268277.
[10] H. Hu, T.K. Wijeratne, J.N. Chung, Two-phase ow and heat transfer during
chilldown of a simulated exible metal hose using liquid nitrogen, J. Low
Temp. Phys. 174 (2014) 247268.
[11] T. Jin, J. Hong, H. Zheng, K. Tang, Z. Gan, Measurement of boiling heat transfer
coefcient in liquid nitrogen by inverse heat conduction method, J. Zhejiang
Univ. Sci. A 10 (2009) 691696.
[12] Y. Katto, C. Kurata, Critical heat ux of saturated convective boiling on
uniformly heated plates in a parallel ow, Int. J. Multiphase Flow 6 (1980)
575582.
[13] Y. Katto, H. Ohno, An improved version of the generalized correlation of critical
heat ux for the forced convective boiling in uniformly heated vertical tubes,
Int. J. Heat Mass Transfer 27 (1984) 16411648.
[14] Y. Katto, S. Yokoya, Critical heat ux of liquid helium (I) in forced convective
boiling, Int. J. Multiphase Flow 10 (1984) 401413.
[15] Y. Katto, S. Yokoya, Critical heat ux of forced convective boiling in uniformly
heated vertical tubes with special reference to very large length-to-diameter
ratios, Int. J. Heat Mass Transfer 30 (1987) 22612269.
[16] O. Kawanami, H. Azuma, H. Ohta, Effect of gravity on cryogenic boiling heat
transfer during tube quenching, Int. J. Heat Mass Transfer 50 (2007) 3490
3497.
[17] O. Kawanami, T. Nishida, I. Honda, Y. Kawashima, H. Ohta, Flow and heat
transfer on cryogenic ow boiling during tube quenching under upward and
downward ow, Microgr. Sci. Technol. 19 (34) (2007) 137138.
[18] S. Kim, I. Mudawar, Universal approach to predicting heat transfer coefcient
for condensing mini/micro-channel ow, Int. J. Heat Mass Transfer 56 (2013)
238250.
[19] S. Kim, I. Mudawar, Universal approach to predicting saturated ow boiling
heat transfer in mini/micro-channels part I. Dryout incipience quality, Int. J.
Heat Mass Transfer 64 (2013) 12261238.

J. Hartwig et al. / International Journal of Heat and Mass Transfer 88 (2015) 662673
[20] S. Kim, I. Mudawar, Universal approach to predicting saturated ow boiling
heat transfer in mini/micro-channels part II. Two-phase heat transfer
coefcient, Int. J. Heat Mass Transfer 64 (2013) 12391256.
[21] B. Kutter, F. Zegler, G. ONeil, B. Pitchford, A Practical, Affordable Cryogenic
Propellant Depot Based on ULAs Flight Experience, AIAA-2008-7644, 2008
SPACE Conference, San Diego, CA, September 911, 2008.
[22] J.H. Lienhard, V.K. Dhir, Hydrodynamic prediction of peak pool boiling heat
uxes from nite bodies, J. Heat Transfer 95 (1973) 152158.
[23] S.M. Liao, T.S. Zhao, An experimental investigation of convection heat transfer
of supercritical carbon dioxide in miniature tubes, Int. J. Heat Mass Transfer 45
(2002) 50255034.
[24] B. Louie, W.G. Steward, Onset of nucleate and lm boiling resulting from
transient heat transfer to liquid hydrogen, Adv. Cryog. Eng. 35 (1990) 403412.
[25] I. Mudawar, D.E. Maddox, Enhancement of critical heat ux from high power
microelectric heat sources in a ow channel, J. Electron. Packag. 112 (1990)
241248.
[26] E. Rame, J.W. Hartwig, J.B. McQuillen, Flow Visualization of Liquid Hydrogen
Line Chill Down Tests, AIAA-2014-1074, 2014 AIAA SciTech Conference,
National Harbor, MD, January 1317, 2014.
[27] W.M. Rohsenow, A Method of Correlating Heat Transfer Data for Surface
Boiling of Liquids, Technical Report #5, Ofce of Naval Research, July 1, 1951.
[28] J.R. Schuster, D.J. Howell, S.L. Lucas Jr., M.S. Haberbusch, J.D. Gaby, N.T. Van
Dresar, M.F. Wadel, Cold Flow Testing of Revised Engine Chilldown Methods
for the Atlas Centaur, 32nd AIAA-96-3014, 32nd Joint Propulsion Conference,
Lake Buena Vista, FL, July 13, 1996.
[29] J.R. Schuster, C.E. Bassett, E.H. Christensen, S.C. Honkonen, F. Merino, D.S.
Munko, J.R. Pietrzyk, M.A. Wollen, Centaur-Based Liquid Hydrogen Fluid
Management Flight Experiments, AIAA-91-3539, AIAA/NASA/OAI Conference
on Advanced SEI Technologies, Cleveland, OH, September 46, 1997.
[30] J.D. Seader, W.S. Miller, L.A. Kalvinskas, Boiling Heat Transfer for Cryogenics,
NASA-CR-243, June 1965.
[31] R. Shaeffer, H. Hu, J.N. Chung, An experimental study on liquid nitrogen pipe
chilldown and heat transfer with pulse ows, Int. J. Heat Mass Transfer 67
(2013) 955966.
[32] J.E. Sherley, Nucleate boiling heat-transfer data for liquid hydrogen at standard
and zero gravity, Adv. Cryog. Eng. 8 (1963) 495500.

673

[33] Y. Shirai, H. Tatsumoto, K. Hata, M. Shiotsu, Preliminary study on heat transfer


characteristics of liquid hydrogen for coolant of HTC superconductors, Adv.
Cryog. Eng. 55 (2010) 337344.
[34] Y. Shirai, H. Tatsumoto, M. Shiotsu, K. Hata, H. Kobayashi, Y. Naruo, Y. Inatani,
K. Kinoshita, Forced ow boiling heat transfer of liquid hydrogen for
superconductor cooling, Cryogenics 51 (2011) 295299.
[35] J.L. Sloop, Liquid Hydrogen as a Propulsion Fuel, 19451959, NASA-SP-4404,
1978.
[36] H. Tatsumoto, Y. Shirai, K. Hata, T. Kato, M. Shiotsu, Forced convection heat
transfer of subcooled liquid nitrogen in a horizontal tube, Adv. Cryog. Eng. 47B
(2008) 665672.
[37] H. Tatsumoto, Y. Shirai, K. Hata, T. Kato, M. Shiotsu, Forced convection heat
transfer of nitrogen at supercritical pressure, in: 2008 International Cryogenic
Engineering Conference Seoul, Korea, 2008, pp. 383388.
[38] C. Tanner, J. Young, R. Thompson, A. Wilhite, On-Orbit Propellant Resupply
Options for Mars Exploration Architectures, AIAA-2006-261, IAC-06-D1.1.01,
2006.
[39] C.J. Westbye, M. Kawaji, B.N. Antar, Boiling heat transfer in the quenching of a
hot tube under microgravity, AIAA J. Thermophys. Heat Transfer 9 (1995) 302
307.
[40] K. Yuan, Cryogenic Boiling and Two-Phase Chilldown Process under
Terrestrial and Microgravity Conditions (Ph.D. thesis), University of Florida,
2006.
[41] K. Yuan, Y. Ji, J.N. Chung, Cryogenic chilldown process under low ow rates, Int.
J. Heat Mass Transfer 50 (2007) 40114022.
[42] K. Yuan, Y. Ji, J.N. Chung, W. Shyy, Cryogenic boiling and two-phase ow
during pipe chilldown in earth and reduced gravity, J. Low Temp. Phys. 150
(2008) 101122.
[43] K. Yuan, Y. Ji, J.N. Chung, Numerical modeling of cryogenic chilldown
process in terrestrial gravity and microgravity, Int. J. Heat Fluid Flow 30
(2009) 4453.
[44] P. Zhang, X. Fu, Two-phase ow characteristics of liquid nitrogen in vertically
upward 0.5 and 1.0 mm micro-tubes: visualization studies, Cryogenics 49
(2009) 565575.
[45] N. Zuber, Hydrodynamic aspects of boiling heat transfer, AEC Report No. AECU4439, 1959.

Potrebbero piacerti anche