VENTLESS PRESSURE CONTROL OF CRYOGENIC STORAGE TANKS

by

STEPHEN BARSI

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Dissertation Adviser: Dr. Iwan Alexander

Department of Mechanical and Aerospace Engineering

CASE WESTERN RESERVE UNIVERSITY

January, 2011
 CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of

Stephen Barsi

candidate for the Doctor of Philosophy degree *

(signed) Dr. J. Iwan D. Alexander

(chair of the committee)

Dr. Mohammad Kassemi

Dr. Yasuhiro Kamotani

Dr. Donald Feke

(date) August 2, 2010

*We also certify that written approval has been obtained for any
proprietary material contained therein.
 Contents

Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

1 Introduction 1

1.1 Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.2 Dissertation Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 34

1.3 Dissertation Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2 Experimental Investigation 36

2.1 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.1.1 Test Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.1.2 Test Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.1.3 Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.1.4 Subcooled Mixing . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.1.5 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.2 Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.2.1 Preconditioning . . . . . . . . . . . . . . . . . . . . . . . . . . 49

ii
 2.2.2 Filling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

2.2.3 Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.2.4 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.3 Test Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

2.4 Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

2.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

2.5.1 Self-Pressurization . . . . . . . . . . . . . . . . . . . . . . . . 58

2.5.2 Pressure Control . . . . . . . . . . . . . . . . . . . . . . . . . 66

2.5.3 Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3 Theory 92

3.1 What is an Interface? . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

3.2 Mathematical Description . . . . . . . . . . . . . . . . . . . . . . . . 94

3.3 Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

3.3.1 Mass Conservation . . . . . . . . . . . . . . . . . . . . . . . . 99

3.3.2 Balance of Linear Momentum . . . . . . . . . . . . . . . . . . 102

3.3.3 Conservation of Energy . . . . . . . . . . . . . . . . . . . . . . 105

3.4 Equation Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.5 Evaporation Condition . . . . . . . . . . . . . . . . . . . . . . . . . . 110

4 Model Development 124

4.1 Thermodynamic Model . . . . . . . . . . . . . . . . . . . . . . . . . . 130

4.2 Multizone Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

iii
5 Implementation 138

5.1 Homogeneous Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

5.2 Zonal Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

6 Results 158

7 Conclusions 177

A Properties of HFE-7000 182

B Heat Exchanger Model 192

C Non-Condensable Gases 196

C.1 Homogeneous Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

C.2 Zonal Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

iv
 List of Tables

2.1 Heat exchanger specifications . . . . . . . . . . . . . . . . . . . . . . 43

2.2 Variation of ambient temperature during the experiment . . . . . . . 44

2.3 Test matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.4 Jet speed calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.5 Liquid fill level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.6 Volume of various components in the tank . . . . . . . . . . . . . . . 54

2.7 Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

2.8 Relevant Dimensional Parameters . . . . . . . . . . . . . . . . . . . . 57

5.1 Quadrature points for a triangular element . . . . . . . . . . . . . . . 147

5.2 Quadrature points for a boundary element . . . . . . . . . . . . . . . 151

6.1 Sensitivity of the zonal solution . . . . . . . . . . . . . . . . . . . . . 162

6.2 Contributions to pressure rise. . . . . . . . . . . . . . . . . . . . . . . 172

A.1 Saturation densities for HFE-7000. . . . . . . . . . . . . . . . . . . . 183

A.2 Saturation internal energies for HFE-7000. . . . . . . . . . . . . . . . 183

A.3 Saturation enthalpies for HFE-7000. . . . . . . . . . . . . . . . . . . . 184

v
A.6 Comparison of different EOS . . . . . . . . . . . . . . . . . . . . . . . 189

vi
 List of Figures

2.1 Schematic of test tank . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.2 Detailed view of jet nozzle . . . . . . . . . . . . . . . . . . . . . . . . 41

2.3 Detailed view of the bottom endcap . . . . . . . . . . . . . . . . . . . 42

2.4 Time history of ambient lab temperature . . . . . . . . . . . . . . . . 45

2.5 Thermistor locations in the test tank . . . . . . . . . . . . . . . . . . 46

2.6 Schematic of the experimental apparatus . . . . . . . . . . . . . . . . 48

2.7 Reproducibility of the experimental data . . . . . . . . . . . . . . . . 59

2.8 Effect of heat input of pressure . . . . . . . . . . . . . . . . . . . . . 60

2.9 Effect of liquid fill level on pressure, Bottom heating . . . . . . . . . . 62

2.10 Effect of liquid fill level on temperature, Bottom Heating . . . . . . . 63

2.11 Tank wall temperature profiles . . . . . . . . . . . . . . . . . . . . . . 65

2.12 Effect of axial jet mixing on pressure . . . . . . . . . . . . . . . . . . 67

2.13 Effect of jet speed on pressure . . . . . . . . . . . . . . . . . . . . . . 70

2.14 Effect of Richardson number on the onset of rapid pressure collapse. . 72

2.15 Effect of jet subcooling on pressure . . . . . . . . . . . . . . . . . . . 75

2.16 Wall temperature profiles during mixing . . . . . . . . . . . . . . . . 76

vii
2.17 Effect of heat input on pressure during mixing . . . . . . . . . . . . . 79

2.18 Effect of fill level on pressure during mixing . . . . . . . . . . . . . . 81

2.19 Triangulation of experimental domain . . . . . . . . . . . . . . . . . . 83

2.20 Pressure and temperature histories . . . . . . . . . . . . . . . . . . . 85

2.21 Experimental temperature map . . . . . . . . . . . . . . . . . . . . . 87

2.22 Temperature histories . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

3.1 Sketch of liquid-vapor system . . . . . . . . . . . . . . . . . . . . . . 93

3.2 Geometric description of two-phase system . . . . . . . . . . . . . . . 94

3.3 Knuden layer above an interface . . . . . . . . . . . . . . . . . . . . . 119

4.1 Heat and mass transfer interaction in a two-phase system . . . . . . . 125

5.1 Finite element mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

5.2 2D linear triangular element . . . . . . . . . . . . . . . . . . . . . . . 145

5.3 1D linear element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

6.1 Effect of tank wall in homogeneous model . . . . . . . . . . . . . . . 159

6.2 Effect of heat exchange with ambient environment . . . . . . . . . . . 160

6.3 Wall temperature contours at time = 1800 s . . . . . . . . . . . . . . 163

6.4 Wall temperature contours at time = 7200 s . . . . . . . . . . . . . . 165

6.5 Wall temperature contours at time = 9900 s . . . . . . . . . . . . . . 166

6.6 Wall temperature contours at time = 12600 s . . . . . . . . . . . . . 167

6.7 Comparison of zonal pressure prediction with experimental data . . . 168

viii
6.8 Comparison of zonal temperature prediction with experimental data . 170

6.9 Model comparisons for different bath temperatures. . . . . . . . . . . 173

6.10 Model comparisons for a 1 W heat load. . . . . . . . . . . . . . . . . 174

6.11 Model comparisons for 73.5% fill level. . . . . . . . . . . . . . . . . . 175

6.12 Model comparisons for a jet speed of 0.114 cm/s. . . . . . . . . . . . 176

B.1 Sketch of shell and tube heat exchanger . . . . . . . . . . . . . . . . . 193

ix
 ACKNOWLEDGEMENTS

First, I would like to acknowledge the financial support from the National Defense

Science and Engineering Fellowship for the first year of this undertaking and the

support of the Ohio Space Grant Consortium for the following three years. I would

like to thank Dr. Mohammad Kassemi for giving me the opportunity to work on

this project and allowing me the latitude to explore various aspects of this problem

that I thought were interesting. I would also like to thank Dr. Iwan Alexander. Our

meetings, where we ground through some of the finer points of this problem, were

helpful. The gratitude of Dr. Charles Panzarella cannot be understated. His rigorous

approach to problem solving is something I’ve tried to emulate during the coarse of

this research effort.

My thanks are also extended to Dr. Feke and Dr. Kamotani for their thoughtful

comments and questions.

With regard to the ground-based experiment, several individuals deserve men-

tion. First, the engineering team of Mr. Scott Meyer, Mr. Frank Kmicik, and Mr.

Robert Butcher did an excellent job of transforming experimental concepts and re-

quirements into hardware. The efforts of Mr. John Juhas in the initial build up and

Mr. Frank Lam in subsequent modifications to the test rig is also greatly appreciated.

I would also like to acknowledge the help of Ms. Amy Wu in calibrating some of the

instrumentation. Thanks are also due to Dr. Nasser Rashidnia for helping me to
x
diagnose and troubleshoot various aspects of the experiment that weren’t behaving

as expected. Our lunchtime chats definitely made the time I spent in the lab more

enjoyable.

Computationally, the resources provided by the National Center for Space Explo-

ration Research, Case Western Reserve University, NASA Glenn Research Center,

and the Ohio Supercomputing Center are all greatly appreciated. The IT support

provided by Mr. David Thompson has also been helpful.

Finally, I would like to extend my appreciation to all my colleagues and friends

at Case Western Reserve University, the National Center for Space Exploration Re-

search, and the NASA Glenn Research Center. My years in Cleveland have been

some of my favorite.

xi
 Nomenclature

A Area

Ao Heat transfer area

B Baffle spacing

C Tube clearance

cp Specific heat at constant pressure

cop Ideal gas specific heat capacity

cv Specific heat at constant volume

D Diameter

De Hydraulic diameter

e Specific total energy

F Force vector

g Gravity vector

g Gibbs energy, Magnitude of gravity vector

h Heat transfer coefficient, enthalpy

H Heaviside function

Hjet Jet submergence depth

I Identity matrix

Imeas Measured current

j Condensation mass flux

J Jacobian matrix
xii
Je Total energy flux

Jq Energy flux due to heat transfer

K Stiffness matrix

k Thermal conductivity

l Interface thickness

L Characteristic length, latent heat

(L1 , L2 ) Area coordinates

m Mass

ṁ Mass transfer rate

M Evaporation rate

MLI Multilayer insulation

Mw Molecular weight

(n̂, t̂) Normal and tangential vectors

N Shape functions

Npts Number of integration points

Nt Number of heat exchanger tubes

Nu Nusselt number

p pressure

P Power

PEEK Polyether ether ketone

PT Tube pitch

xiii
Pr Prandtl number
′′
q Heat flux

Q̇ Heat power

Q̇w Total heat power

R Gas constant, Radius

Ra Rayleigh number

Re Reynolds number

S Viscous stress tensor

s Specific entropy

S Entropy

T Temperature

T Stress tensor

T Nodal temperatures

u Specific internal energy

U Internal energy, total heat transfer coefficient

v Velocity

V Volume, velocity

Vsource Source voltage

Wi Weighting coefficients

X Mole fraction

(Z, R) Nodal coordinates

xiv
(z, r) Coordinate directions

Greek

β Coefficient of thermal expansion

ξ Coordinate along 1D finite element

δ Dirac function

∆t Time increment

∂Ω Interfacial region

κ Bulk viscosity

ρ Density

σ Accommodation coefficient, surface tension

ω Mass fraction, accentric factor

Ω Bulk phase region

µ Dynamic viscosity

ν Specific volume

Subscripts and Superscripts

avr Average

bulk Bulk liquid or vapor phase

c Critical

E Elemental

xv
g Gas

I Interface

il Liquid side of the interface

in Inlet

iv Vapor side of the interface

jet Jet

l Liquid

n Time step

nozz Nozzle

o Initial, reference

out Outlet

s Surface quantity, shell

sat Saturation

t Tank, tube

T Transpose

v Vapor

wl Wetted wall

wv Non-wetted wall

+ Liquid region

− Vapor region

xvi
 Ventless Pressure Control of Cryogenic Storage Tanks

Abstract

by

STEPHEN BARSI

Future operations in space exploration will require the ability to store cryogenic

liquids for long durations. During storage, the tanks may self-pressurize due to heat

leaks from the ambient environment. When heat leaks into the tank, the cryogenic

liquid vaporizes causing the ullage pressure to rise. Being able to effectively control

tank pressure will make these long duration storage concepts feasible. One way to

control tank pressure involves the use of a subcooled axial liquid jet to both ther-

mally destratify the bulk liquid and remove energy from the tank. In this dissertation,

the effectiveness of using subcooled jet mixing as a pressure control scheme is ana-

lyzed by performing a small-scale experiment in a normal gravity environment with

a refrigerant. Following a period of self-pressurization, the jet’s speed and degree of

subcooling are parametrically varied so that relevant trends can be identified. Exper-

imental results show that mixing the bulk liquid is not sufficient to control pressure.

To sustain any pressure reduction, subcooling the mixing jet is necessary. The rate of

pressure reduction is greater for increased jet speeds and subcooling. Analytical and

computational models were developed in order to predict the pressurization behav-

ior. Model comparisons reveal that generally a thermodynamic model underpredicts

the self-pressurization and depressurization rates. The lack of agreement is primarily

attributed to the homogeneity assumption inherent in the model. To improve model

predictions, a zonal model is developed which relaxes the global homogeneity assump-

tion. Comparisons between the experimental data and the zonal model predictions

are excellent for moderate to high jet flow rates. For slower jet speeds, buoyant flow in

the bulk liquid adversely affects the effectiveness of a subcooled mixing jet and a more

detailed computational model is required to capture this intraphase phenomena.

xviii
 Chapter 1

Introduction

Future operations of many space propulsion, environmental control and life sup-

port, thermal and power systems will all depend on the ability to store, process, and

control a variety of single or multiphase fluids in reduced gravity environments.1 For

many of these systems, cryogens are the fluids of choice and indeed, in future missions

to the Moon or Mars, cryogens will play an integral role.

In a recent NASA workshop,2 the long term preservation of cryogens for life sup-

port systems was identified as a critical issue for the development of advanced human

life support technologies. For the in-situ resource utilization program, the preserva-

tion of cryogens is essential, as large quantities of liquid oxygen (LOX) and liquid

methane (LCH4) will have to be stored on the surfaces of the Moon or Mars.3, 4 To

realize a significant reduction in mission costs, NASA had envisioned placing large

cryogenic propellant depots in low earth orbit (LEO) or deep space for periods of

up to five years.5, 6 For these large storage tanks, which should be able to contain

100,000 lb7, 8 of LOX or liquid hydrogen (LH2), zero loss storage of these propellants

1
 INTRODUCTION

is essential. Even in NASA’s Constellation program, the efficient storage of cryogens

has been identified as a core technology.9 In the Exploration Systems Architecture

Study,9 cryogenic fuels are currently envisioned for the reaction control system on the

crew exploration vehicle, on the Earth Departure Stage (EDS) and on the Lunar Sur-

face Access Module’s (LSAM) descent and possibly ascent stages. For both the EDS

and LSAM, cryogenic fuel tanks will be loitering in LEO for several weeks. Addition-

ally, for the LSAM ascent stage, these cryogenic fuel tanks will be required to stay on

the lunar surface for up to six months.10 As reported by Motil and Meyer,10 based

on preliminary design estimates, the LOX and LCH4 tanks on the lunar surface may

lose up to 3% of usable propellant per month. Obviously reducing these losses would

translate into cost savings for the exploration program. This brief survey highlights

the many technologies that would benefit from efficient storage of cryogenic liquids.

Storage of cryogens though presents a significant challenge. Cryogens are stored

at very low temperatures and may be subjected to large heat loads while the storage

tanks are loitering in LEO, in transit, or sitting on the surface of the Moon or Mars.

The heat load can come from a variety of internal or external sources. The external

sources can include incident solar radiation, planetary albedo, aerodynamic heating,

or conduction loads from the tank’s support structure. Internally, the exothermic

reaction of ortho/para conversion of LH2 and the kinetic energy associated with

liquid sloshing, which eventually dissipates as heat,11 can all be factors. When heat

leaks into the tank, it will be carried to the liquid-vapor interface by conduction and

natural convection. Once this thermal energy reaches the surface, the liquid may
2
 INTRODUCTION

start vaporizing. Since vaporization occurs in a closed tank, the tank pressure will

increase. Design constraints regarding the tank’s maximum operating pressure and

requirements regarding tolerable liquid losses make controlling both the phase change

process and the tank pressure a necessity.

Various strategies have been identified as possible mechanisms to control tank

pressure. The simplest strategy involves periodically venting the vapor overboard to

reduce the pressure. Venting is a straightforward operation in a normal or partial

gravity environment where the lighter vapor sits atop the heavier liquid. In a reduced

gravitational environment, where the position of the vapor is less certain, venting

only vapor becomes more challenging. While venting does reduce tank pressure, it

does nothing to mitigate boil-off losses. That is, venting over time will still lead to

considerable propellant losses.

Other strategies involve storing the liquid as a bulk subcooled phase10 or using

combinations of MLI blankets to insulate the tank and sunshades to shield the tank

from solar radiation or planetary albedo.10, 12–14 Bulk liquid subcooling can increase

the amount of energy the liquid can hold before vaporization and boiling takes hold

thus delaying the tank’s pressure rise. Sunshades serve to reduce the heat load on

the storage tank. While both of these strategies offer some benefits, both lack the

robustness of an active pressure control system.

When heat leaks into the tank from the surroundings, the temperature field inside

often stratifies. Mixing the fluid inside the tank can destratify the thermal field;

reducing temperature gradients near the liquid-vapor interface and bringing the cooler
3
 INTRODUCTION

fluid that had settled out closer to the surface - both of which promote condensation

and a reduction in tank pressure. Mixing strategies alone have been the subject of

much research. Passive mixing has been shown to have a significant effect on fluid

behavior in low gravity environments. Gebhart showed that random disturbances can

result in transport rates much greater than would be expected in the absence of all

disturbances.15 Grodzka reports that experiments conducted during Apollo 14 and

17 and data taken from the oxygen tanks aboard Apollo 15 revealed that natural

convection caused by g-jitter can be significant.16 Passive mixing however lacks the

robustness necessary for active pressure control. Active mixers such as axial or radial

liquid jets or spray bars have also been studied. Axial jets exhibit a definite gravity

dependence. Oftentimes, the liquid jet flows opposite the direction of buoyancy which

tends to reduce its effectiveness. Spray bars, on the other hand, are typically gravity

independent since liquid is ejected radially into both phases from a bar running the

entire length of the tank. As will be discussed shortly though, which active mixing

strategy is more effective remains uncertain. Regardless, active mixing alone offers

only a temporary reduction in tank pressure. Because the mixing device itself adds

energy to the system, the net heat load post-destratification can be larger than if no

mixer were present.

Consequently, active mixers are often coupled to a refrigeration system. One

of the earliest refrigeration concepts considered for cryogenic storage systems was a

thermodynamic vent system (TVS).17 In a TVS, some sacrificial liquid passes through

a throttling device to reduce its temperature. The colder two-phase fluid then passes
4
 INTRODUCTION

through a heat exchanger which can be attached to the outer surface of the tank

to intercept the incident heat load.18–22 Due to heat transfer into the two-phase

fluid, vaporization occurs and the resulting vapor is eventually vented overboard.

Alternatively, after passing through the throttling valve, the fluid can enter a heat

exchanger which can be used to subcool a liquid jet or be placed internal to the

tank to remove energy from the bulk liquid and vapor. A spray bar/heat exchanger

TVS system designed by Rockwell Aerospace23 and extensively tested at the Marshall

Space Flight Center has demonstrated the ability to control tank pressure within a

very tight control band for a variety of cryogens and mission scenarios.24–27 Although

less common, a TVS system consisting of an axial jet mixer and a heat exchanger28, 29

has also been tested.30 In a recent review,31 Hastings reported TVS performance

comparisons between an axial jet mixer and a spray bar. Although tank pressure

decayed more rapidly during the spray bar operation, the axial liquid jet resulted in

better bulk mixing. The comparisons appear to be inconclusive especially since the

liquid flow rates were different in the two test cases. Regardless of the combination

of mixer or heat exchanger employed, a TVS is designed to sacrifice some liquid to

reduce the bulk energy of the system. Hence, operation over extensive periods of time

will lead to loss of usable propellant.

Recent advances in zero boil-off (ZBO) technologies have improved the prospects

of a truly zero loss storage system.32 The main refrigeration system used in a ZBO

system is a cryocooler. The cooler can be mounted outside of the tank and mated to

heat exchangers, internal condensing surfaces, heat pipes, or wall-mounted thermally

5
 INTRODUCTION

conducting sheets.33, 34 While Plachta has achieved ZBO conditions during ground

testing34 using passive mixing, because the conduction and natural convection time

scales are much slower in low gravity, a ZBO cryocooler is often combined with an

active mixer to enhance condensation and pressure control. Preliminary testing has

been performed with both axial jet mixers33 and spray bars.35

These preliminary proof-of-concept tests are very promising. Even a modest re-

duction in boil-off losses (and thus launch mass) can translate into significant mass

savings.36 Comparing a ZBO system with a passive storage and insulation system,

Plachta37 showed that for short duration missions a ZBO system is not as attractive

as a passive storage tank due to the increased launch mass of the mixer and cry-

ocooler. However, after a relatively short time (1 week for LOX, 2 weeks for LCH4,

and 2 months for LH2) the mass and cost benefits of a ZBO system are quickly real-

ized. Besides the cost and mass savings, ZBO adds to mission flexibility as delays in

rendezvous or docking would no longer jeopardize propellant mass margins.32 While

the benefits of a ZBO system may be substantial, more work is required before its full

potential is realized. From a power consumption and reliability standpoint, it may

not be practical to continuously operate an active mixer. Optimization is necessary

to tune the system to maximize performance. Optimizing a ZBO system requires a

better understanding of the complicated and coupled transport phenomena inside the

tank which can affect thermal stratification, self-pressurization, and pressure control.

These same issues were identified in past reviews38 as critical to cryogenic storage

and remain as relevant today.

6
1.1 LITERATURE SURVEY INTRODUCTION

Due in part to these reviews, extensive research dating back to the Apollo program

has been performed but unanswered questions still remain. To answer some of these

questions, large-scale flight demonstrations are generally preferred. Large-scale flight

experiments, however, can be costly and time consuming. As pointed out by Chato,39

computational tools can offer development cost savings and improved designs but

these tools must be quantitatively validated. As such, the design approach currently

favored by much of the cryogenic community is to develop numerical models, validate

them against small-scale experiments in both normal and reduced gravity, and then

use both the sub-scale experimental data and the computational models to extrapolate

the ZBO design to an actual flight system. Indeed, this is the approach adopted in

this dissertation.

Having established the relevance of the current research problem, before the disser-

tation objectives are outlined, a survey of prior research in the field will be presented.

1.1 Literature Survey

NASA has a rich heritage of flight testing and flight qualifying cryogenic fluid

management technologies in support of past exploration programs. Beginning in the

early 1960s, and continuing for several years, several experiments were conducted

aboard Aerobee sounding rockets which provided approximately 4 12 minutes of low

gravity. Knoll et al.40 performed LH2 experiments in a 9” diameter partially full

Dewar subjected to radiant heating. The measured self-pressurization rate was larger

7
1.1 LITERATURE SURVEY INTRODUCTION

than a simple thermodynamic analysis predicted. Thermal measurements indicated

that during the experiment an initially wetted-wall eventually formed dry spots during

the flight. The experiment sat on a de-spin platform to counter the effects of the spin-

stabilized rocket. Misalignment of the de-spin platform resulted in accelerations of

±0.02g. In a subsequent flight,41 a similar experiment was performed and this time,

nucleate boiling was observed. Unfortunately for this flight, the de-spin platform

malfunctioned. In a similar experiment reported by Nunamaker et al.,42 temperature

measurements indicated wall dry out and fluid sloshing in the tank. Once again, the

self-pressurization rate was underpredicted by thermodynamics. Later in the Aerobee

program, Aydelott43 conducted similar experiments but with a higher liquid fill level.

Aydelott reported that the measured self-pressurization rate was approximately twice

the rate predicted by thermodynamics, that ullage motion during the flight resulted

in some fluid mixing, and that it took approximately two minutes for the radiant

heaters to reach their set point temperature. Thus for a significant portion of the

experiment, the incident heat load was not relatively constant.

Abdalla et al.44 reported on the pressurization experiments conducted on an Atlas

rocket which provided for 21 minutes of low gravity. During the flight, the experimen-

tal pod began tumbling which resulted in an acceleration of 0.001g. Consequently, the

pressure rise characteristics were similar to testing done in 1g and temperature mea-

surements exhibited cyclic behavior in phase with the external perturbations. During

the Atlas/Centaur AC-8 flight, Lacovic et al.45 studied propellant behavior during an

orbital coast. Temperature measurements indicated significant thermal stratification

8
1.1 LITERATURE SURVEY INTRODUCTION

in the ullage.

Bradshaw46 and Navickas et al.47 described the self-pressurization results on the

Saturn IVB-AS203 flight. Continuous venting of the O2 tanks during most of the

flight provided a settling acceleration to the instrumented LH2 tank. Data was taken

during an orbital coast, however a loss of signal resulted in 50% of the data missing

during the middle of the test. Allgeier48 reported on a small-scale LN2 experiment

conducted on the AS-203 flight. During this experiment, the tank was allowed to

self-pressurize after which a small amount of liquid was withdrawn from the tank and

passed through a heat exchanger brazed to the outer wall of the test cell. This system

exhibited good pressure control.

Several relevant experiments were also conducted on the LOX/LH2 Titan/Centaur

upper stage.49, 50 While these experiments were primarily investigating liquid orien-

tation and engine restart capability after an orbital coast, temperature and pressure

measurements were also made to study stratification and pressurization. From tem-

perature measurements taken during these flights, Lacovic50 inferred that sections of

the tank’s wall dried out during the coast.

While many of these proof-of-concept flight experiments yielded important data,

unknown or uncontrolled boundary or initial conditions rendered them less useful for

the purposes of validating numerical models. As Chato noted in a recent review,39

piggy-backing on the Saturn and Centaur upper stages made many of the these exper-

iments cost effective but unfortunately prevented them from carrying more extensive

instrumentation to provide detailed data for future model validations.

9
1.1 LITERATURE SURVEY INTRODUCTION

Besides these orbital and sounding rocket tests, cryogenic fluid management ex-

periments have also been performed aboard aircraft. Ordin et al.51 mounted a 450

gal LH2 tank to the wing tip of a jet aircraft to investigate the effects of atmospheric

turbulence on thermal stratification. As a result of significant agitation during the

flight, the degree of stratification in the tank was diminished when compared to sim-

ilar ground tests. Bentz52 conducted a small-scale Freon 113 self-pressurization/axial

jet mixing experiment aboard a Lear jet flying parabolic profiles. The value of these

particular parabolic flight experiments was limited due to their short duration expo-

sure to low gravity (± 0.01g).53 In the experiment, Bentz reported that there was

insufficient time between parabolas for the liquid to reach a quiescent state.

To obtain more long duration periods of low gravity, Bentz and colleagues52, 54, 55

performed tank pressure control experiments (TPCE) on three shuttle flights in the

early 1990s. During the first flight, a small tank partially full of Freon 113 self-

pressurized due to heaters submerged in the liquid. Heating was maintained for

several minutes, after which, the heaters were turned off, liquid was withdrawn from

the tank, and pumped back in through an axial liquid jet. The pressure collapse

as a result of axial jet mixing was studied. On the second flight, during heating,

higher local superheats were observed which apparently resulted in flashing of the

liquid and a spike in tank pressure. The third experiment was similar to the first two,

except now the test was performed at a lower fill level. These TPCE experiments

provided useful data but unfortunately due to several shortcomings, the data is not

suitable for numerical validation. First, the liquid jet was not thermally controlled
10
1.1 LITERATURE SURVEY INTRODUCTION

and unfortunately no jet temperature measurements were made. It’s uncertain how

much heat leaked from the liquid jet to the surroundings. Any subcooling of the jet

would have had a profound effect on the pressure reduction times. Second, no thermal

controls existed between the tank and the ambient environment. Yet no attempts were

made to quantify the amount of heat lost from the tank. Third, there was only a 20

minute wait period between experiments and it’s unclear whether 20 minutes were

long enough for subsequent tests to begin from the same thermal and dynamic state.

Finally, for all three TPCE experiments, contaminant species leaked into the test

cell. The amount of non-condensable species was estimated from the overpressure

above saturation. During the three experiments, the partial pressure of the non-

condensable gases ranged between 1 kPa and 6.2 kPa. While the investigators claimed

the non-condensable contaminants had no effect on the results, other theoretical and

experimental studies would suggest otherwise. (see for example Rose,56 Minkowycz

and Sparrow,57 and Hastings et al.27 ).

In addition to the flight experiments, surveyed so far, there have also been exten-

sive ground tests both in support of and independent of these flight projects. While

investigating thermal stratification in a ground liquid nitrogen experiment, Fan et

al.58 observed a thin thermal boundary layer near the tank walls and noted that

convection heat transfer was significant. Beduz et al.59 performed wall heating strat-

ification studies of LOX and LN2 in small Dewars. Using temperature measurements

in the liquid, they were able to map the morphology of the temperature fields. Below

the interface, they observed a thin thermally conducting layer a few hundred microns
11
1.1 LITERATURE SURVEY INTRODUCTION

thick with a steep temperature gradient which resided on top of a convective layer

with a shallow temperature gradient. Both layers sat atop a near uniform bulk liquid.

Swim60 conducted stratification studies using Dewars partially full of liquid helium.

He observed steep temperature gradients on either side of the interface in line with

Beduz’s morphology studies. Tatom and colleagues,61 performing stratification ex-

periments in a 500 gal LH2 tank, noted that a considerable amount of thermal energy

went into raising the temperature of the bulk fluid which suggests the absence of well-

defined boundary layers. Neff and Chiang62 reported on stratification experiments

using both water and cryogenic fluids. They noted that an increase in the bulk liq-

uid temperature indicated quasi-steady flow and temperature conditions in the tank.

Moreover, they attempted to describe the stratification process semi-empirically by

approximating the temperature profile in the liquid as a polynomial. These stratifi-

cation tests were not limited to only recording temperature. Lovrich et al.,63 using

a Schlieren system, performed flow visualization experiments using Freon and water

with local side wall heating. They found that most of the heat remained above the

heater with a sharp drop off in the temperature profile below. Anderson and Kolar64

also used a Schlieren setup to compare a side wall heating configuration with a bot-

tom heating one. For side wall heating, they observed a stable temperature gradient

below the interface. The bottom heating configuration led to better mixing of the

bulk liquid which resulted in a more uniform temperature profile. More recently, Das

et al.65 used a dye injection system to map out the temperature field in a side heated

cavity containing water.

12
1.1 LITERATURE SURVEY INTRODUCTION

Nearly all of these stratification experiments were primarily concerned with ther-

mal behavior and not the coupling between the temperature field and the vapor

pressure. Ji et al.66 studied stratification and pressure rise and based on a scal-

ing analysis, identified three dimensionless parameters in an attempt to describe and

characterize the underlying physics. Several small-scale tests were performed to verify

the validity of these dimensionless groups. Results indicated that the vapor pressure

histories agreed reasonably well between the scaled pairs of tests but point-to-point

temperature matches were not obtained indicating that the three dimensionless pa-

rameters were not sufficient at characterizing the entire system behavior. Manson67

also identified several dimensionless parameters to provide geometric, dynamic, and

thermal similitude. In practice though, matching all of these parameters is difficult

to achieve. Neff 68 performed scaling experiments with LOX and LN2 to verify that

the dimensionless parameters that he had identified were able to completely charac-

terize the underlying physics. Comparing scaled pairs of tests, he observed similar

temperature profiles although agreement between the temperature values was again

lacking. Bourgarel et al.69 performed a scaling analysis without including conduction

and interfacial phenomena. They also chose not to match Grashof numbers in their

experimental tests. The similarity between their scaled pairs of tests was initially fair

but deviations developed with time.

Blatt70 posited a function of the self-pressurization rate in terms of heat in-

put, tank volume, and fill fraction. He used available data from ground and flight

LOX/LH2 self-pressurization tests in the literature to fit the constants in his func-
13
1.1 LITERATURE SURVEY INTRODUCTION

tion. The correlation however was based on only a limited number of data points

and resulted in deviations as high as 89%. Scott et al.71 investigated stratification

and pressurization in a partially full LHe Dewar. They initially observed a thermally

stratified liquid but after placing thermally conducting copper rods in the Dewar, the

temperature gradients were reduced and the recorded pressure rise agreed better with

a thermodynamic analysis.

In the early 1990s, a series of self-pressurization experiments were performed with

a 4.95 m3 partially full LH2 tank at NASA Glenn’s Plumbrook station. In these

tests, as with many experiments involving cryogens, the incident heat load was not

an independent parameter but was instead computed from measured boil-off rates. A

detailed description of this calculation as well as the tank’s thermal boundary condi-

tions are outlined in Stochl and Knoll.72 It’s unclear whether the heat load estimated

in this fashion was applicable to all the test points in the experiment. Regardless,

Hasan et al.73 still found that the self-pressurization rate increased with increas-

ing heat load. For the lower heat flux cases, there was less deviation between the

measured and thermodynamically predicted pressurization rates than for the higher

heat flux cases. Van Dresar et al.74 re-ran these experiments at lower fill levels.

In all cases, thermodynamics underpredicted the self-pressurization rate. The effect

of fill level on the pressurization rate was difficult to discern from the experimental

data. The expected trend, as predicted by thermodynamics, was not reflected by the

experimental data.

In support of the Aerobee sounding rocket tests, Aydelott75, 76 performed a series

14
1.1 LITERATURE SURVEY INTRODUCTION

of ground self-pressurization tests in a 9” diameter spherical tank partially filled

with LH2. He found the pressurization rate was affected mostly by the heating

configuration (top, bottom, or uniform heating) with only a slight effect from varying

the fill level. Comparisons were also made with a homogeneous thermodynamic model

and a surface evaporation model which assumed all the incident energy was used to

vaporize the liquid and keep the vapor in a saturated state. The experimental data

was bounded by the two models with the thermodynamic model underpredicting the

pressurization rate. Some tests were performed while shaking the tank. In this case,

the pressurization rate approached the homogeneous model’s prediction. Aydelott

and Spuckler77 also investigated the effects of tank size by comparing the previous 9”

tank tests with tests in a 22” diameter spherical tank. They found similar pressure

rises for equal values of the heating rate to volume ratio. Summarizing his results

from both the ground and flight 9” diameter tank tests, Aydelott78 noted a reduced

pressurization rate in low gravity. He attributed this effect to the increased wetted-

wall area and the fact that direct heating of the vapor would result in a larger pressure

rise rate.

While most of the these ground self-pressurization experiments were conducted

using small-scale tanks, Liebenberg and Edeskuty79 performed tests in a 55,000 gal

Dewar 94.7% full of LH2. The observed pressure rise rate was greater than the

prediction of a homogeneous model by nearly a factor of 10.

Ground testing was not limited to self-pressurization studies. Several pressure

control tests were also performed. Huntley80 experimented with a closed LN2 De-
15
1.1 LITERATURE SURVEY INTRODUCTION

war and found that after mechanically stirring the liquid, the pressure temporarily

decayed. However, he found that mixing the ullage resulted in an increase in the

pressurization rate.

In evaluating TVS designs, Bullard81 noted that a bottom-mounted axial jet was

more effective at collapsing the pressure than a side-mounted horizontal jet. In his

mixing-only tests, the pressure reduction was temporary and it was noted that the

minimum pressure was reached after only 20-30% of the liquid circulated through

the jet nozzle. He also observed that introducing a non-condensable gas into the

system significantly increased the pressure collapse time. Dominick82 also examined

the effects of jet orientation on the condensation rate in a small-scale tank partially

filled with Freon. He observed higher rates of condensation and liquid destratification

when the jet nozzle was oriented perpendicular to the liquid vapor interface. In

Dominick’s experiment however, the pressure in the tank was constant since vapor

was being supplied to the ullage at the same rate mass was condensing at the interface.

Moreover, there were no thermal controls between the tank and the surroundings and

it’s uncertain how much heat was being lost to the outside environment.

Lin et al.83 conducted a series of axial jet mixing experiments in a partially full

LH2 tank. However it’s unclear whether the same level of stratification was attained

between the different test cases. Nonetheless, they found the pressure collapsed faster

for faster jet flow rates. For the lowest flow rate considered, the pressure continued

to increase during mixing as the jet flow was not strong enough to counter the effects

of buoyancy. Lin et al.84 later showed that buoyancy effects could be neglected for
16
1.1 LITERATURE SURVEY INTRODUCTION

Richardson numbers less than 0.5. Here also, two mixing times were identified. One

was used to describe how fast the liquid destratified and the second was used to

describe how fast the pressure collapsed.

Finally, Jones and colleagues85, 86 conducted several ground-based pressure con-

trol tests in a small-scale tank partially filled with Freon 11. Self-pressurization was

initiated by activating a heating coil submerged slightly below the surface. After

pressurizing for a time, the heater was deactivated and an axial liquid jet was used to

destratify the liquid. They found buoyancy dominated the flow for jet Reynolds num-

bers (Re) below 2000. The data did not correlate well with steady-state condensation

or dye-mixing correlations. Later, Jones et al.87 developed a closed form pressure col-

lapse equation by assuming that the pressure drop was due to condensation and not

vapor cooling. Here, it was also noted that the dimensionless mixing time correlated

with Re for low jet Reynolds numbers. For larger values, the dimensionless mixing

time was constant.

While most of the ground and flight experiments described thus far were applied

in nature, there have been several investigations that have focused on more funda-

mental aspects of the problem. McNaughton and Sinclair88 studied the stability of

liquid jets in terms of Re. Four flow regimes were characterized: dissipated lami-

nar, fully laminar, semi-turbulent, and fully turbulent. Mollendorf and Gebhart89, 90

performed stability and perturbation analyses investigating how buoyancy effects liq-

uid jet behavior. Though primarily concerned with positive buoyancy, they noted

that the effect on the temperature and velocity fields could be particularly strong if
17
1.1 LITERATURE SURVEY INTRODUCTION

buoyancy opposed the jet motion.

Symons et al.91 and Labus92 performed a set of ground tests to characterize the jet

flow behavior for a GHe jet into a GHe medium. Symons and Staskus93 later studied

the interaction of a liquid jet and a free surface and developed a critical Weber number

(We) criterion that described the stability of the surface. The stability of the interface

is important for the current tank problem since a jet geysering into the ullage would

increase the interfacial area through which mass transfer can occur.

Berenyi et al.94 performed drop tower tests on a spherical tank with a radial jet

and tried to characterize the different observed flow patterns using the jet’s velocity.

Aydelott performed similar tests with an axial jet in both cylindrical95 and spherical96

containers. Several flow patterns, including jet geysering, were observed. Correlations

were developed to describe the flow patterns in terms of We and Bond numbers (Bo).

In Aydelott’s tests, only 70% of the incoming liquid was being withdrawn from the

tank. It’s unclear what effect the accumulation of liquid in the tanks had on his

conclusions.

In addition to these circulation and jet flow characterization tests, many exper-

iments were conducted to better understand jet mixing. Fox and Gex97 used an

acid/base neutralization technique to develop mixing time correlations in both lami-

nar and turbulent jet regimes. Both correlations exhibited a dependence on Re and

included a gravitational effect. Fossett and Prosser98 were able to correlate their jet

mixing data independently of Re. Okita and Oyama99 used their data, obtained us-

ing a density-matched dye technique, along with the Fox and Gex data to develop
18
1.1 LITERATURE SURVEY INTRODUCTION

a mixing time correlation that was dependent on Re for low Reynolds numbers but

independent of it for higher Re values. The mixing tests by Lane and Rice100 also

observed a stronger Reynolds number dependence for laminar jets than for turbulent

ones. Poth et al.101, 102 evaluated several different mixing devices and found jet mix-

ers to be superior. It was observed that the time required for the jet to reach the

interface was approximately twice as long as a simple kinematic analysis would pre-

dict which highlights the retarding effects of negative buoyancy. Aydelott95, 96 used

his drop tower experiments to develop a mixing time correlation in terms of We and

Bo. Lehrer103 attempted to derive a correlation independent of any empirical data

but the results were inconclusive. Comparing against the Fox & Gex and Fossett &

Prosser data, his correlation sometimes underpredicted and sometimes overpredicted

the measurements. Grenville and Tilton,104 reasoned that liquid jet entrainment con-

trolled the mixing of the entire vessel. They therefore included the path length of the

jet as one of the characteristic length scales in their mixing time correlation. Pat-

wardhan and Gaikwad105 performed several mixing tests and found their data agreed

best with the Grenville and Tilton correlation. It seems unlikely though that any

one correlation will fit all of the data. In a more detailed review, Revill106 notes that

mixing is highly dependent on a number of factors including the relative size of the

tank and jet, the protrusion of the jet into the tank, the fill level, and the shape of

the tank. In addition, a comparison between the correlations is complicated by the

fact that many of these investigators define mixing time differently from one another.

Most of these studies were only concerned with jet mixing and not the phenomenon
19
1.1 LITERATURE SURVEY INTRODUCTION

of a jet interacting with a condensing/evaporating interface.

Helmick et al.107 performed a set of steady-state steam-on-water condensation

experiments that employed an axial liquid jet. They noted that a turbulent liquid

enhances condensation and that the condensation rate can be quantified in terms

of fluid properties and the liquid-side turbulence. Thomas108 conducted similar ex-

periments and observed as the liquid jet penetrated the interface, the condensation

rate, which he characterized in terms of a heat transfer coefficient, increased. Chun

et al.109 observed that if the jet subcooling was significant (30-80 K) mass trans-

port can become unstable resulting in condensation bursts and rapid pressure decay.

Sonin et al.110 developed a turbulent condensation correlation assuming steady state

mass transfer, isotropic turbulence, small surface waviness, and negligible buoyancy.

His correlation required the r.m.s. value of the turbulent velocity at the free surface

which he estimated from a simple k-ǫ analysis. Brown et al.111 extended this work by

including buoyancy effects through a Richardson number dependency. Brown112 also

extended Sonin’s correlation by developing another expression for the r.m.s. turbu-

lent velocity at the free surface which was valid for smaller jet submergence depths.

It remains to be seen, however, whether these steady state condensation correlations

are applicable to the transient situation that prevails during the pressure control of

cryogenic storage tanks.

This comprehensive experimental review pertaining to thermal stratification, self-

pressurization, and pressure control underscores the degree of uncertainty in many of

these experimental investigations. Additionally, the experiments themselves provide

20
1.1 LITERATURE SURVEY INTRODUCTION

little detailed data that is conducive to a comprehensive model validation and ver-

ification effort. Detailed measurements of the flow field are typically not made and

oftentimes the bulk of the data is compressed into a simple engineering correlation.113

Unfortunately, this level of uncertainty and lack of more detailed data can cloud the

comparisons between model and experiments making validation efforts more difficult

and less systematic.

In parallel to the experimental efforts outlined above, numerous theoretical and

computational models of varying levels of sophistication have also been developed

to both interpret and predict experimental behavior. Historically, a homogeneous

thermodynamic model was one of the earliest analytical means to predict the self-

pressurization rate in a cryogenic tank partially full of liquid. Here, a First Law energy

balance is performed over the entire liquid-vapor system. To close the problem, the

homogeneity assumption is invoked. That is, the temperature of both the liquid and

vapor phases are equal and at saturation. There have been many versions of the

homogeneous thermodynamic model over the years. If the functional form of the in-

ternal energy of the two-phase fluid is known, then the energy derivative appearing in

the energy balance can be computed explicitly and the vapor pressure can be evolved

in time.76, 114–116 Assuming constant specific heats, others117–119 have represented the

time derivative of energy as the product of specific heat and the time derivative of

temperature. Still others,120 avoid the rate form of the First Law, and use a thermo-

dynamic balance to determine the final thermodynamic state of the two phase system

after knowing the net energy input. Naturally the thermodynamic analysis assumes
21
1.1 LITERATURE SURVEY INTRODUCTION

that the average energy of the liquid and vapor phases changes at the same rate as the

energy of the two phase mixture defined at the saturation temperature. Because this

condition is not met during the initial phases of self-pressurization experiments, when

thermal boundary layers are developing and temperature gradients in the liquid and

vapor are not stationary, the agreement between thermodynamics and experiments

has generally been poor. Moreover, because the homogeneous thermodynamic model

forces the energy of each phase to change at the same rate, and the thermal inertia

of the liquid is typically the largest contribution to the energy balance, thermody-

namics usually underpredicts measured self-pressurization rates. Recognizing this,

Aydelott76 developed a surface evaporation model which assumed the sensible energy

of the liquid is constant and all of the incident energy is used for vaporization or

maintaining the ullage at a saturation state. Since this model neglects the contribu-

tion of the liquid’s sensible energy to the net energy balance, it generally overpredicts

measured self-pressurization rates.

In order to obtain better agreement with experimental data, transport effects

should be included. A number of investigators have developed models which include

energy and mass transport. These models are most easily classified as zonal methods

whereby the liquid-vapor system is divided into zones at constant temperature and

engineering correlations are used to model the energy and mass transport between

the zones. The number of zones is completely arbitrary. Riemer developed a two

zone model, one for each phase.116 Estey et al.121 included a separate zone bound-

ing the interface. Epstein and Georgius122 divided the tank wall, liquid phase, and
22
1.1 LITERATURE SURVEY INTRODUCTION

vapor phase into many axial zones. Schallhorn et al.123 partitioned the liquid into

annular boundary layer zones and axial zones in the bulk. Riemer, comparing his two

zone model to the homogeneous thermodynamic model, noted that the zonal model

did a better job of reproducing experimental data. Hedayat et al.,124 compared the

zonal model of Nguyen125 against a self-pressurization experiment and found that

the model overpredicted the self-pressurization rate. The results of these models are

unfortunately not unique and can vary depending on the correlations used to model

heat and mass transport between the different zones.

Instead of relying on correlations, many investigators have tried to explicitly ac-

count for transport effects. In an early stratification analysis, Knuth126 modeled only

the liquid phase and treated it as a semi-infinite solid. For a change in the interfacial

temperature, he was determining the response of the temperature field in the liquid.

Knuth127 and Thomas et al.108 extended this analysis by modeling the liquid and

vapor phases as two semi-infinite media coupled at the interface. Schmidt et al.128

performed a subcooled stratification experiment and compared his results with the

semi-infinite conduction solution. He found there was some agreement initially, but as

time progressed, deviations increased. Segel129 performed his own pressurized strat-

ification tests and found for low heat fluxes, there was good agreement between the

measured temperature profiles in the liquid and the semi-infinite conduction solution.

For higher heat fluxes, however, deviations were observed and attributed to increased

convection in the liquid.

To improve on the conductive analyses, a number of investigators have used ap-

23
1.1 LITERATURE SURVEY INTRODUCTION

proximate integral methods to resolve natural convection in the liquid. Bailey et

al.130 developed a model to study stratification in the liquid. He assumed that all of

the incident energy appears as sensible heat to the free convection boundary layer,

that the energy in the boundary layer is carried to an upper stratum layer where it

remains, and that there is no mixing between the upper stratum and the bulk liquid.

With these severe restrictions, and assumed boundary layer temperature and veloc-

ity profiles, integral heat and mass balances were performed. Comparisons between

the model’s predictions and empirical stratification data was poor. In the stratifi-

cation experiment, Bailey et al. observed mixing between the upper stratum and

the bulk which was not accounted for in the model. Bailey and Fearn131 compared

their approximate integral model to a 70 ft3 LH2 stratification experiment but were

again unable to accurately predict the temperature profiles in the liquid. Tellep and

Harper132 performed a similar analysis but included an interfacial energy contribution

in the integral balance and assumed a time-invariant temperature profile in the upper

stratum. This resulted in a better agreement between the predicted and measured

temperature profiles. Ruder133 and Robbins et al.134 developed similar models but

noted modifications were necessary to include the effects of bottom heating and phase

change. Vliet135 extended the analysis by including the effects of bottom heating but

neglected the interfacial energy contribution in the integral balance. Barnett et al.136

included a parameter that allowed for energy exchange with the bulk liquid but still

observed poor agreement between the measured and predicted surface temperature

rise. More recently, Kirk et al.137 used these approximate methods to study a rotat-
24
1.1 LITERATURE SURVEY INTRODUCTION

ing upper stage. They assumed the liquid to be in solid-body rotation with a static

paraboloid interface. They noted that the rotation had an effect on the heat transfer

due to the increase in wetted-wall area, but no comparisons with experiments were

made.

Eventually these approximate integral methods grew into more sophisticated bound-

ary layer type analyses. Barnet et al.138 applied correlations for the boundary layer

thickness, the growth of the upper stratum, and the natural convection speed at the

edge of the boundary layer. He obtained reasonable agreement with measured tem-

perature profiles after including a term that accounted for the heat of compression in

the liquid. Arnett and Millhiser139 included both the liquid and vapor phases in their

analysis and accounted for inter-phase energy and mass transport. They assumed

turbulent free convection boundary layer profiles for velocity and temperature but

posited the functional form of the boundary layer thickness and velocity at the edge

of the boundary layer. Arnett and Voth140 used integral balances to compute these

parameters at every location along the boundary layer. Comparing their results with

Atlas/Centaur pressurization data45 they observed poor agreement between measured

and predicted temperature profiles and underpredicted the pressurization rate by as

much as 15%. Venkat and Sherif 141 extended the Arnett and Voth model by including

variable fluid properties, bottom heating, and an ortho/para conversion routine. Most

of these changes resulted in only marginal differences with Arnett and Voth’s predic-

tions. For the ortho/para concentrations analyzed, there was no discernible effect on

the results. Gursu et al.142, 143 tested several free convection boundary layer profiles
25
1.1 LITERATURE SURVEY INTRODUCTION

and found no significant effect on the results. Their ortho/para conversion routine

showed clearly, and quite intuitively, that the boil-off rate increases with increasing

ortho concentration. Most of these boundary layer analyses assume a stationary well-

developed free convection boundary layer and a relatively simple tank geometry both

of which may not be realistic in a real storage tank setting.

To obtain more meaningful predictions, these approximate techniques have given

way to more sophisticated computational models. Before accounting for the more

complicated interfacial heat and mass transfer, several numerical studies were con-

ducted which investigated thermal stratification in the liquid phase due to some ex-

ternal heating. Nikitin and Polezhaev144 considered a partially filled sphere in mi-

crogravity. Assuming a static, insulated, and spherically-shaped free surface, they

studied the interaction between buoyancy and Marangoni convection for different

ullage locations inside the tank. Cherkasov145 assumed a static, flat, and shear-free

surface and computed the time evolution of the temperature field for different wall and

interface heating configurations. Lin and Hasan146 numerically studied the steady-

state flow and temperature fields that developed as a result of buoyancy in a tank

with a flat shear-free interface. They tried to characterize the thermal behavior in

terms of the liquid subcooling - the interfacial temperature was assumed fixed at the

saturation temperature and the temperature at the bottom of the tank was fixed at

some subcooled value. Sengupta147 performed similar numerical stratification studies

by enforcing an adiabatic temperature condition along the interface. His computed

temperature profiles below the interface deviated from the experimental values but
26
1.1 LITERATURE SURVEY INTRODUCTION

the deviations decreased in the bulk liquid. Tanyun et al.148 also forced the sur-

face to be adiabatic and noted that the computed surface temperature was higher

than experimental measurements. Barakat and Clark149 set the interfacial temper-

ature equal to the saturation temperature and were able to obtain some agreement

with experimental temperature profiles. Navickas150 numerically studied the effect of

baffles on the thermal stratification in a partially full rectangular cavity with a flat

interface. His results indicated that through judicious placement of baffles, thermal

stratification can be suppressed.

All of these preceding computational studies lacked a gas-phase model and treated

the interface as a static surface. Grayson et al.151, 152 removed the second limitation

by allowing the interface to freely evolve and deform. They studied how the interface

and the thermal field respond to different gravitational accelerations. They, too,

however neglected gas-phase transport and assumed an adiabatic interface.

In addition to the numerical stratification studies, a number of numerical studies

on jet mixing has been performed. Hasan and Lin153 performed isothermal steady-

state computations of axial jet mixing and its interaction with a flat interface. They

compared the computed turbulent r.m.s. velocity profile with Sonin’s expression110

and noted an increased deviation close to the interface. They attributed the discrep-

ancy to the Neumann boundary conditions placed on the turbulent kinetic energy

and dissipation rate at the free surface. Later, Lin and Hasan154 performed a similar

computational analysis and obtained good agreement in comparison with Brown’s

expression112 for the turbulent r.m.s velocity profile for low jet submergences.
27
1.1 LITERATURE SURVEY INTRODUCTION

Hochstein et al.155 also investigated isothermal jet mixing. Instead of Neumann

conditions on the turbulent kinetic energy, they prescribed it to be zero at the inter-

face. Unfortunately, no comparisons to experiments were made. Hochstein et al.,156

again employing a k-ǫ turbulence model, neglecting buoyancy, and allowing the free

surface to deform, were not able to obtain a solution for high jet flow rates. For lower

jet speeds, they were able to model the interface as a static boundary. Wendl et al.157

were able to predict the flow patterns (including geysering) Aydelott observed95 in his

drop tower experiments but the results were only qualitatively in agreement. To bet-

ter predict geyser height, Thornton and Hochstein158 performed a sensitivity study on

how various parameters affect the computational solution. For all the parameters con-

sidered they found only marginal differences in the CFD solution. Critically reviewing

the Aydelott experiment,95 Simmons et al.159 inferred different boundary conditions

and fluid properties than had been used in previous simulations. Though, after im-

plementing these modifications deviations between measured and predicted geyser

heights persisted. Marchetta et al.160 tried to improve the geysering simulations by

using a k-ω turbulence model. However, the results indicated that the original k-ǫ for-

mulation yielded better agreement with the experiment. Marchetta and Bendetti161

performed 3D geysering simulations and tested a suite of two equation turbulence

models. They noted that while the k-ǫ model more accurately predicted the geyser

height when geysering occurred, Menter’s SST k-ω formulation162 was superior at

predicting the different flow regimes that Aydelott had observed.

These previously mentioned deformable free surface simulations employed the vol-
28
1.1 LITERATURE SURVEY INTRODUCTION

ume of fluid method. Chato et al.,163 however, used Jacqmin’s phase field model164

to simulate interfacial deformation during geysering. Their simulations overpredicted

the geyser height at faster jet flow rates. They attributed this discrepancy to a lack of

a turbulence model. Chato165 later added a simple turbulence model which assumed

a constant turbulent viscosity. This approach over corrected the results which led

to an underprediction of the geyser height. Chato166 also performed a parametric

study investigating the effects of contact angle, geometry, and surface tension on the

predicted geyser height. He found that by increasing the contact angle the geyser

height increased.

Mukka and Rahman167 conducted a steady-state finite element analysis of a liquid

jet entering a completely full tank. They evaluated the effectiveness of mixing by

comparing the circulation patterns produced by different jet configurations. Ho and

Rahman168 performed 3D steady-state finite element simulations of a horizontal liquid

jet impinging on a heat pipe in a full cryogenic storage vessel. They primarily studied

the different circulation patterns for varying jet speeds and did not include any two-

phase or thermodynamic effects.

A number of investigators have built on this body of computational work involving

thermal stratification and jet mixing by including the effects of self-pressurization. Lin

and Hasan169 developed a simple conduction model in the liquid and coupled it to

the thermodynamic pressurization model of Brown.112 They also neglected gas-phase

transport but allowed the interface to expand and contract radially. In the liquid,

they included a lumped compressibility term in their energy balance and found that
29
1.1 LITERATURE SURVEY INTRODUCTION

initially, the pressurization rate was greater for higher fill levels because of liquid

expansion. After some time had elapsed however, due to the large thermal inertia of

the liquid, the final pressurization rate decreased with increasing fill level. Hochstein

et al.,170, 171 neglecting gas-phase transport, employed an effective conductivity model

to account for transport in the liquid and performed a cell-by-cell mass balance along

the interface to account for evaporation. Comparisons with experiments44, 76 yielded

reasonable agreement for the bottom heating and uniform heating cases in 1g, but

discrepancies were noted for the top heating test case in 1g and for the uniform

heating case in low g. Although Hochstein172 noted that more work was required to

improve the heat transport modeling in the tank, they used this model to study the

effects of bulk liquid subcooling and found the self-pressurization rate decreased with

increased subcooling.

Grayson et al.,173 including transport in the ullage and using the pressurization

model of Hirt,174 attempted to simulate the AS-203 flight experiment.46, 47 Grayson’s

model resulted in a 3.5% deviation between measured and predicted pressurization

rates. The reported discrepancy is however a little misleading since only the initial

and final pressures were used to compute the pressurization rate. A more careful

inspection of the comparison between the predicted pressure and data extrapolated

from the experiment results in a larger deviation.

Merte et al.175 also developed a pressurization model which included the effects

of gas-phase transport. The interface was assumed flat and the pressure for the

incompressible/incompressible system was updated using a First Law energy balance.

30
1.1 LITERATURE SURVEY INTRODUCTION

Merte et al.176 later compared their predictions to data from the AS-203 flight but

the agreement was not good. They attributed the errors to inadequately modeling

the tank geometry. Val’tsiferov and Polezhaev,177 included the effects of transport in

the ullage and used an integrated form of the ideal gas law to update the pressure but

were not able to obtain agreement with Aydelott’s self-pressurization experiments.76

Tunc et al.178 also developed a gas-phase model to study thermal stratification and

pressure behavior when a LOX tank is pressurized with GHe. While no experimental

comparisons were provided, the computational results seem unreliable as the non-

condensable contribution to the ullage pressure appears to be incorrectly modeled.

Given the difficulties associated with including transport effects in the ullage, it’s

no surprise that several investigators continue to couple lumped thermodynamic bal-

ances in the ullage to numerical solutions in the liquid. Amirkhanyan and Cherkasov179

coupled an effective conductivity analysis in the liquid to a lumped model of the ullage.

The results appear to be inconclusive. When comparing with measured experimental

data, in some cases they overpredicted the pressurization rate, in other cases they

underpredicted it, and in no cases did their results agree with a homogeneous ther-

modynamic analysis. Panzarella and Kassemi117 rigorously coupled a lumped energy

and lumped mass model of the ullage to the detailed transport equations in the liq-

uid. While experimental comparisons were not made, their simulations for a given

heat load and fill level in a tank showed that the long-term self-pressurization rate is

independent of the heating configuration. Additionally, the computed pressurization

rates agreed with a thermodynamic analysis of the system. Extending this model to
31
1.1 LITERATURE SURVEY INTRODUCTION

larger LH2 tanks in microgravity, Panzarella and Kassemi180 showed that the thermal

stratification in a reduced gravity field can be significant and even in a 10-6 g field,

ullage migration from the center to the top of the tank occurs on a much faster time

scale than heat or mass transport inside the tank.

Numerical investigations have not only been limited to thermal stratification, jet

mixing, and self-pressurization but have also included pressure control with a sub-

cooled axial mixing jet. Albayyari181 developed a simple analytic pressure control

model and seemingly validated the model. The validation claim is surprising as the

nearly identical model of Bentz52 yielded better agreement with experimental data.

The major difference between the two models is that different correlations were used

to model the heat transfer between the impinging jet and the free surface.

In order to obtain better agreement with experimental data, more sophisticated

computational models have been developed. Lin182 numerically studied the effects of

jet Reynolds number (Re) and Prandtl number (Pr) on the condensation rate which

he quantified in terms of an average Stanton number. His steady-state computations

in a rectangular domain with a flat free surface at saturation revealed that for the

range of parameters considered, the average Stanton number is independent of Re and

only weakly dependent on Pr. Lin and Hasan183 continued this analysis and showed

how geometry, heat load, and jet subcooling can affect the condensation rate. In

the above numerical studies by Lin and colleagues, buoyancy was neglected. Hasan

and Lin184 included the effects of buoyancy and showed that natural convection can

reduce the condensation rate.

32
1.1 LITERATURE SURVEY INTRODUCTION

Panzarella et al.185 performed a numerical study of subcooled axial jet mixing

in microgravity. Their lumped vapor active liquid model showed that for most jet

speeds, subcooled jet mixing was effective at reducing tank pressure. For the lowest

jet speed however, buoyancy, due to residual gravity, prevented the jet from reaching

the interface and no pressure reduction was observed.

Van Overbeke30 developed a 1g pressure control model that neglected gas-phase

transport. In this two-point vapor model, he used the temperature at the upper

bulkhead of the tank and the saturation temperature at the interface to approximate

the temperature gradient on the vapor-side of the interface. This approach seems

questionable because in the experiments Van Overbeke was comparing against,83, 84

the vapor was superheated and it’s unlikely this global energy difference accurately

represented the local temperature gradient on the vapor-side of the interface. Not

surprisingly, the validation results were inconclusive. In some cases the pressure

collapse was overpredicted while in other cases it was underpredicted.

This rather comprehensive experimental and numerical review highlights several

limitations in previous research efforts. Experimentally, the degree of uncertainty

and lack of more detailed measurements makes these experiments unsuitable for vali-

dating and verifying self-pressurization and pressure control models. Moreover, prior

numerical-experimental validation efforts have yielded results less than satisfactory

with the sources of the discrepancies often left unexplained or misunderstood.

33
1.2 DISSERTATION OBJECTIVES INTRODUCTION

1.2 Dissertation Objectives

The objectives of this research are to address these shortcomings by rigorously

developing self-pressurization and pressure control models, discussing their strengths

and weaknesses, conducting a series of ground-based experiments with well defined

boundary and initial conditions, and then comparing the model’s predictions with

data from the experiment. Any discrepancies between model predictions and experi-

mental data will be honestly assessed with possible sources of the deviation discussed.

1.3 Dissertation Outline

The body of this dissertation is organized into seven chapters. In the first chapter,

the relevance of this research was discussed. A literature survey was presented which

highlighted the prior work in the fields of thermal stratification, self-pressurization,

and pressure control. The literature review also established that several shortcom-

ings exist in the current body of knowledge. How the current research effort addresses

this knowledge gap was outlined in the objectives section. The details of the experi-

mental apparatus and procedures are presented and both the self-pressurization and

pressure control experimental results are discussed in Chapter 2. In Chapter 3, the

instantaneous bulk transport equations and interfacial conditions are derived from

first principles. These equations are implemented into various models of varying de-

grees of complexity. The model development is presented in Chapter 4. Numerical

34
1.3 DISSERTATION OUTLINE INTRODUCTION

implementation of the models is described in Chapter 5. Model predictions are com-

pared to the experimental data in Chapter 6 and results are discussed. Finally in

Chapter 7, the major conclusions of this dissertation are discussed and possible areas

of future work are outlined.

35
 Chapter 2

Experimental Investigation

A series of small-scale model fluid experiments were conducted at the NASA

Glenn Research Center as part of the ground-based portion of the Zero Boil-Off Tank

(ZBOT) experiment. The objective of this series of experiments was to investigate

the effects of various parameters on self-pressurization. Additionally, the effectiveness

of using an axial jet mixer as a means of controlling tank pressure was studied. In

what follows, the experimental apparatus will be described in detail, the procedures

outlined, and the results discussed.

2.1 Apparatus

2.1.1 Test Cell

A schematic of the test tank is shown in Fig. 2.1. The tank, made of R-Cast

acrylic, had an internal diameter of 8” and an outer diameter of 10.125”. The cylin-

drical midsection of the test cell was 16” long. The tank was capped on both ends

with acrylic plates 16” x 16” x 2”. A section of a sphere (R = 8.5”) was machined
36
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

(a)

16

2
1

2.5
1
0.06 R = 8.5

16
8

10.125

2.375

(b)

Figure 2.1: Schematic of test tank (a), Relevant dimensions in inches (b).

37
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

1” into each plate. The top and bottom end caps were bolted to the cylindrical mid-

section with a single O-ring seal at the joining faces. Additionally, the bottom plate

was bolted to a vibration isolation table. For the heat loads imposed and pressures

observed, the tank was assumed rigid with negligible expansion. The density of the

acrylic was given by the manufacturer as 1190 kg/m3 . The thermal conductivity and

specific heat, as reported in the literature,186 are:

a0 = −0.02222
  X  T i
W a1 = 1.67196
k = 0.18987 ai
m·K 298.15
a2 = −0.64984

a3 = 0

  X  T i
kJ
ρ cp = 1614.1 bi b0 = −0.05640
m3 · K 298.15
b1 = 1.05638

To insulate the tank and associated plumbing from the environment, all external sur-

faces were blanketed with Volara low density cross-linked polyethylene foam. The

thickness of the insulation was 0.5”. The density of the foam was determined to be

32 kg/m3 . In the literature,187 there is some uncertainty regarding the thermal prop-

erties of the insulation. At room temperature,

38
2.1 APPARATUS EXPERIMENTAL INVESTIGATION
 
W
0.035 < k m·K
< 0.05

 
J
2611.8 < cp kg·K
< 2924.5

Prior to beginning the experimental tests, the test tank was leak checked by filling the

tank with air, pumping it down to the nominal operating pressure (500 Torr), sealing

it, and observing how fast the pressure rose in the system over a 24 hour period.

Leak paths include gas diffusion through the acrylic, which is generally negligible,

and leakage through the multitude of O-ring seals. When the thermistors entering

the tank through the side wall were not present, and their respective ports were

capped with O-ring sealed plugs, the leak rate was 0.22 Torr/hour, which over a

two hour pressurization period amounts to less than 0.5 Torr overpressure. When

all of the thermistors were present, the leak rate was higher at approximately 0.69

Torr/hour. Consequently, for most of the tests performed as part of this experimental

effort, the test tank was not fully populated with thermistors. Only in the last test

were all the thermistors present so that a qualitative map of the temperature field

could be obtained.

2.1.2 Test Fluid

The fluid used in the experiment was HFE-7000, a transparent low boiling point

refrigerant manufactured by 3M. Thermophysical properties of the test fluid are pro-

vided in Appendix A.

39
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

2.1.3 Heating

To simulate heat leaking into the tank, two etched foil Kapton heaters were affixed

to the inside surface of the test tank. The 1” wide heating bands were positioned

approximately 2” from either end cap. The relative positions are indicated in Fig. 2.1.

The bands were 24.4” long which resulted in a slight gap in the circumference to allow

for the silver-plated copper leads to exit the test cell. The leads were teflon coated

Type E 24 AWG. Adhesive was initially used to affix the heating bands to the inside

surface of the test cell. However, after preliminary testing revealed parts of the band

separating from the wall, a flat copper spring (1” wide, 0.06” thick) was placed over

the Kapton foil to keep the bands in place. The copper spring had a 0.5” gap in the

circumference to allow the heater leads to exit the test cell. Power to the heaters was

supplied by Keithley 2425 source meters. Prior to data acquisition, the resistances

of the heaters were measured by the source meter (± 0.07% rdg + 0.3 Ω) and the

voltage source (± 0.02% rdg + 12 mV) was programmed to produce the desired

power output from the heaters. During the experiment, the heater’s power output

was monitored by the source meter. Power output was determined by monitoring the

current (0.055% rdg + 6µA) and then applying the power relationship:

P = Vsource Imeas (2.1)

40
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

0.25

0.2

0.3

D = 0.125 0.5

o
50
0.5

3.75

Figure 2.2: Detailed view of jet nozzle. (Dimensions in inches)

For a typical test, the uncertainty in the measured power was approximately 1.7 mW.

During the experiment, the resistance of the heaters changed due to changing temper-

atures within the tank. To account for this changing resistance, the applied voltage

was modified periodically to keep the measured heater power to within 0.2 mW of

the set point.

2.1.4 Subcooled Mixing

Subcooled axial jet mixing was used to control tank pressure. Fluid entered the

tank from an AISI 304 stainless steel nozzle aligned along the longitudinal axis of the

tank. The nozzle projected 1” into the tank and converged to a 0.25” ID 0.2” from

41
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

Outlet Ports

Nozzle

Figure 2.3: Detailed view of the bottom endcap.

the nozzle exit plane as indicated in Fig. 2.2. Fluid was withdrawn from the tank

through 12 ports (0.125” ID) circumferentially located about the nozzle. The outlet

ports, depicted in Figs. 2.2 and 2.3, were tapped through the bottom plate at an

angle 50o from the horizontal. The fluid from the outlet ports collected in a common

manifold before being pumped through a heat exchanger and re-entering the tank

through the nozzle. The recirculation pump was a Series 120 magnetically driven

gear pump from MicroPump.

The heat exchanger was a single pass counter flow shell and tube heat exchanger

from Exergy (model 00256-2). The specifications of the heat exchanger are listed

in Table 2.1. In the heat exchanger, the test fluid exchanged heat with distilled wa-

ter. The water was contained in a Haake F8-C25 temperature-controlled recirculation

42
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

Table 2.1: Heat exchanger specifications

Number of baffles 7
Number of tubes 55
Tube length 10”
Tube OD 0.125”
Tube wall 0.0125”
Tube Clearance 0.162”
Shell ID 1.37”
Shell OD 1.5”
Baffle spacing 1.44”
Heat Transfer Area 1.43 ft2
Pitch Triangular

bath. The water was pumped through the shell-side of the heat exchanger at approx-

imately 12 lpm. The temperature of the bath was an independent variable of the

experiment and was controlled to within ± 0.01 o C.

2.1.5 Instrumentation

Ullage pressure was measured using an MKS model 120AA-01000 RCJ transducer.

The transducer had a resolution of 0.1 Torr and an accuracy of 0.05% rdg. Calibration

was performed independently at the calibration lab on-site at NASA Glenn. The

transducer was mounted to the vibration isolation table and was piped through the

upper acrylic plate of the tank using approximately 54” of 1 /4 ” OD insulated stainless

steel tubing.

Mounted axially along the cylindrical midsection of the tank were nine 4-wire 100

Ω Platinum RTD elements (Omega model SRTD-2). The RTDs were mounted to a foil

43
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

Table 2.2: Variation of ambient temperature during the experiment

Case No. Tmin (o C) Tmax (o C) Tavr (o C)
1 22.8 22.2 22.5
2 23.0 22.0 22.6
3 23.1 22.1 22.6
4 23.5 22.0 22.5
5 23.2 22.1 22.6
6 23.2 22.0 22.6
7 23.1 22.0 22.5
8 23.2 21.9 22.5
9 23.1 22.1 22.5
10 23.1 22.1 22.6
11 23.3 22.0 22.5
12 23.3 22.0 22.6
13 23.1 22.0 22.5
14 23.5 22.1 22.6

element 1” x 0.62” and were cemented to the outer acrylic wall using OMEGABOND

101 epoxy adhesive. RTD resistances were recorded and temperatures were deter-

mined using temperature-resistance calibration curves. For a typical test case, the

self-heating for a single RTD element was 0.11 mW.

The ambient temperature of the lab near the tank was measured using a 2.2 kΩ YSI

thermistor (± 0.1 o C). Room temperature was kept relatively constant throughout

the duration of the experiment. The average room temperature for each test case,

along with the minimum and maximum fluctuation is listed in Table 2.2. A typical

time history of the ambient lab temperature is shown in Fig. 2.4. The oscillations in

the response are due to the lab’s on-off air conditioning cycle.

Temperature inside the tank was measured with custom-made thermistors from

YSI. The thermistors were bead-in-glass types potted in a PEEK tube. The diameter

44
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

23.0

Temperature [ oC] 22.8

22.6

22.4

22.2

22.0
0 2000 4000 6000 8000
Time [s]

Figure 2.4: Typical time history of the ambient lab temperature.

of the glass bead was 0.06”. The outer diameter of the PEEK tube was 0.125”

and the wall thickness was approximately 0.03”. PEEK was chosen to house the

thermistor because of its stiffness and low thermal conductivity (k ≈ 0.25 W/m K.

The self-heating for a single thermistor was approximately 1 µW. The thermistors’

power supply and signal conditioning unit were custom built and the entire system

was calibrated independently at NASA Glenn’s calibration lab at 25 o C, 30 o C, and

35 o C. The reported uncertainty was ± 0.04 o C. The thermistors internal to the tank

were all positioned in the longitudinal midplane of the test cell. Fourteen thermistors

entered the tank through the side wall and a single thermistor entered the tank

through the center of the upper end cap. The thermistors were positioned at various

depths inside the tanks as shown in Fig 2.5.

45
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

T1 = ( 0.0000, 11.50)
T2 = (-3.0625, 15.00)
T3 = ( 3.3125, 13.00)
T4 = ( 2.1250, 10.25)
T5 = ( 1.5000, 8.75)
T2 T6 = ( 0.9375, 7.25)
R1 T7 = (-1.6875, 7.75)
R2 T3 T8 = (-1.0000, 6.25)
R3 T9 = ( 3.7500, 5.75)
T1 T10= ( 1.0000, 4.25)
T4 T11= (-0.9375, 4.75)
T12= (-0.1250, 3.25)
R4 T5 T13= ( 3.7500, 2.75)
R5 T14= ( 0.0625, 1.00)
T7 T15= (-2.1250, 1.00)
T8 T6 T9
R6 R1 (y = 15.00)
T11 R2 (y = 14.00)
T10 R3 (y = 12.75)
R7 R4 (y = 9.125)
T12 T13 R5 (y = 8.125)
R8 R6 (y = 5.125)
y
R9 T15 T14 R7 (y = 3.250)
R8 (y = 2.250)
R9 (y = 1.250)
x

Figure 2.5: Thermistor and RTD locations in the test tank. (Coordinates are
in inches. Blue dashed lines denote approximate locations of the 26.5%, 50%,
and 73.5% fill levels.)

46
2.1 APPARATUS EXPERIMENTAL INVESTIGATION

To minimize the interference of the thermistors on the flow field (and consequently

the temperature field and pressurization behavior) for most of the test cases, only the

thermistor entering from the top plate (T1 in Fig. 2.5) was present. For these

cases, the thermistor ports tapped through the side wall were capped with O-ring

sealed plugs. For the final case of the experimental test program, the tank was fully

populated with thermistors. Though interfering with the flow field somewhat, this

fully populated case provides a qualitative map of the temperature field inside the

tank.

The flow rate of the mixing jet was measured using a McMillan G112-7 flow meter.

The flow meter had a range of 60-1000 mL/min and an accuracy of ± 0.5% FS.

Flow measurements, pressure measurements, and temperature measurements (from

the RTD elements and thermistors) were recorded using an Agilent 34970A data ac-

quisition unit with three 20 channel multiplexers. Uncertainty from the DAQ unit,

which includes measurement and switching errors, were reported by the manufacturer

as ±0.005% rdg + 0.4 mV for voltages and ± 0.01% rdg + 0.04Ω for resistances. Dur-

ing the first 13 tests, measurements were made at 1 Hz and subsequently stored on a

laptop. For the last test, where the tank was fully populated with thermistors, mea-

surements were recorded once every ten seconds to accommodate the larger amount

of data generated during these runs. In all cases, the DAQ unit visited channels

sequentially. The time stamp between successive channels was 50 - 60 ms.

47
2.2 PROCEDURES EXPERIMENTAL INVESTIGATION

Vacuum LN2
Pump Cold
Trap

Reservoir Test Cell

Transfer
Pump

Pump
F

Temperature
Controlled
Bath

Figure 2.6: Schematic of the experimental apparatus.

2.2 Procedures

A schematic of the entire apparatus is shown in Fig. 2.6. Throughout the exper-

imental program, the test fluid was stored in a 5 gal reservoir tank. The test fluid

was preconditioned in the reservoir tank and then transferred to the test tank using a

no-vent fill technique described below. Once filled to the desired fill fraction, the fluid

was again conditioned to initialize each test case, the experiment was performed, and

at the end of the test, the fluid was pumped back into the reservoir tank. Each exper-

imental test case took approximately two days to complete with the preconditioning

in the reservoir tank occurring one day prior to the actual test.

48
2.2 PROCEDURES EXPERIMENTAL INVESTIGATION

2.2.1 Preconditioning

Preconditioning of the test fluid in the reservoir was necessary to degas any non-

condensable species dissolved in the fluid. To degas the fluid, a vacuum pump was

used to rapidly collapse the pressure in the reservoir. As the partial pressure of the

non-condensables in the ullage decayed, the dissolved species came out of solution

and were subsequently vented outside. As the ullage was pumped down, the vapor

pressure of the test fluid was also reduced. Consequently, bulk boiling of the test

fluid occurred. Bulk liquid boiling was beneficial to the degassing process as it served

to mix the fluid and carry the dissolved species to the liquid-vapor interface. Before

being vented outside, the evacuated gases and vapor passed through a liquid nitrogen

cold trap and the test fluid that condensed out was collected and saved for future use.

Degassing was performed multiple times prior to a test case. Each degassing process

only lasted several minutes and was limited by the amount of LN2 in the cold trap.

After a degassing process, the reservoir tank was undisturbed for several hours in

order to equilibrate to a new thermodynamic state. Subsequently, the pressure and

temperature in the ullage were measured and compared with the test fluid’s saturation

curve. If the measured temperature and pressure deviated significantly from the

saturation curve the degassing procedure would be repeated. For all the test cases,

one-to-one agreement with the saturation curve was never attained. The measured

pressure was typically 20 Torr higher than the saturation pressure corresponding to

the measured ullage temperature. The deviation is attributed to the presence of

49
2.2 PROCEDURES EXPERIMENTAL INVESTIGATION

a non-condensable gas. As will be described shortly, this resulted in a worst-case

non-condensable mole fraction of 3.6%. This non-condensable mole fraction is worst-

case because the saturation condition only applies on the two-phase boundary for an

inhomogeneous system. Both the temperature and pressure were recorded a distance

away from the interface. Hence, these measurements do not account for any thermal

stratification or hydrostatic head that may have been present in the system.

2.2.2 Filling

Once the fluid was sufficiently degassed, the test tank was filled by pumping the

fluid from the reservoir to the test cell using a two-way transfer pump. Prior to

fluid transfer, the test tank was pumped down using the same vacuum pump that

was employed during the degassing process. During the tank pump down, pressure

was monitored using both the MKS pressure transducer and an analog Varian 801

thermocouple vacuum gage. Once the pressure in the tank reached 10 mTorr, the

valve between the tank and the vacuum pump was closed and the valve between the

tank and the reservoir was opened. Since the reservoir was now open to an evacuated

tank, some fluid in the transfer line flashed and the tank pressure rapidly rose to near

the fluid’s saturation pressure. The transfer pump was activated and fluid continued

to enter the ventless test tank until the desired fill fraction was reached. The liquid

fill level was determined using a ruler (± 1 /16 ”) mounted along the outer wall of the

acrylic midsection. Once the fill level was recorded, the foam insulation was pulled

taut around the tank and then secured.

50
2.2 PROCEDURES EXPERIMENTAL INVESTIGATION

2.2.3 Initialization

As fluid was pumped into the tank during the no-vent fill, the ullage was com-

pressed, the tank pressure had risen above its saturation value, and consequently

condensation was occurring at the liquid-vapor interface. Thus, once the fluid trans-

fer ceased, the liquid-vapor system was not in a global equilibrium state. To accelerate

the equilibration between the two phases, the fluid in the test tank was conditioned.

Conditioning proceeded by running the axial jet mixing loop (with a throughput of

approximately 460 mL/min) and setting the recirculation bath temperature to the

desired initial temperature. Conditioning occurred for 4 hours prior to the start of

a test. Approximately 10 minutes before beginning a test, the mixer was turned off

and the valves between the test tank and the recirculation loop were closed. The 10

minute delay was included to allow for the jet-induced velocities to sufficiently de-

cay. During this 10 minute mixer-off period, the tank pressure was monitored. If the

tank pressure changed ± 0.1 Torr the conditioning process was repeated for an addi-

tional hour after which another check of the pressure was made. Fluid conditioning

continued until this pressure criterion was met.

2.2.4 Testing

During the 10 minute mixer-off period, the source meters for the heaters were

programmed and the measurement channels and recording frequency were set on the

DAQ unit. At time=0, power was supplied to the heaters and the tank was allowed

51
2.3 TEST MATRIX EXPERIMENTAL INVESTIGATION

to self-pressurize for 2 hours. During these 2 hours, the recirculation bath remained

active with the temperature set point now programmed to the desired subcooled jet

temperature. Approximately 30 s prior to activating the jet mixer, the jet speed was

set and the valves separating the test tank from the recirculation loop were opened.

Subcooled axial jet mixing began at time = 2 hours and continued for an additional

1.5 hours. The heaters remained active during this time.

After mixing for 1.5 hours, both the heaters and the mixing pump were deacti-

vated. The test fluid was then drained back into the reservoir using the reversible

transfer pump. All valves were then closed and the experiment was shut down.

One day after each test, the vacuum pump/LN2 cold trap system was used to purge

the test tank and fluid lines (up to the recirculation loop valves) of any remaining

HFE-7000 liquid or vapor. Once again, any test fluid that condensed in the cold trap

was reclaimed for later use.

2.3 Test Matrix

Fourteen test runs will be discussed in this dissertation. First, self-pressurization

will be examined. Here, the effects of liquid fill level and applied heat load on the

pressurization behavior of the test fluid will be discussed. Next the effectiveness

of using an axial liquid jet to control tank pressure will be analyzed. The effects

of average jet speed and degree of subcooling on the depressurization rate will be

studied.

52
2.3 TEST MATRIX EXPERIMENTAL INVESTIGATION

Table 2.3: Test matrix

Case No. Liquid Heat Heating Tsub Pump
height Load Configuration voltage
1 4” 2W B 22.5 o C 3V
2 4” 2W B - 3V
3 4” 2W B 17.5 o C 3V
4 4” 2W B 20.0 o C 3V
5 4” 2W B 20.0 o C 2V
6 4” 2W B 20.0 o C 1.75 V
7 4” 2W B 20.0 o C 1.5 V
8 4” 2W B 20.0 o C 1.25 V
9 4” 2W B 20.0 o C 1V
10 4” 1W B 20.0 o C 3V
11 4” 3W B 20.0 o C 3V
12 8” 2W B 20.0 o C 3V
13 12” 2W B 20.0 o C 3V
14 4” 2W B 20.0 o C 3V

Table 2.4: Jet speed calibration

Pump voltage Flow rate (mL/min) Jet speed (cm/s)
1V 136.7 7.2
1.25 V 176.9 9.3
1.5 V 217.2 11.4
1.75 V 257.5 13.5
2V 297.7 15.7
3V 458.8 24.1

Table 2.5: Liquid fill level

Height of Liquid in Tank Percentage of nominal
interior tank volume
4” 26.5%
8” 50.0%
12” 73.5%

53
2.3 TEST MATRIX EXPERIMENTAL INVESTIGATION

Table 2.6: Volume of various components in the tank

Region Volume
Interior test tank (nominal) 855.56 in3
Copper ring(s) 1.44 in3
Thermistor T1 0.0674 in3
Outlet ports 0.25 in3
Nozzle inside tank 0.09755 in3
Nozzle inside wall 0.069 in3
Ports in top plate 0.4 in3
Ports in side wall 0.09 in3
Tubing (ullage to pressure transducer) 1.374 in3
Tubing (ullage to relief valve) 2.655 in3
Acrylic side (nominal) 484.00 in3
Acrylic top (nominal) 486.34 in3
Acrylic bot (nominal) 486.34 in3

The test matrix is given in Table 2.3. A malfunction in the flow meter required

re-calibration after the test program was complete. The relationship between applied

voltage to the pump and the resulting flow rate is provided in Table 2.4 along with the

corresponding average velocity of the fluid exiting the nozzle. In the test matrix, Tsub ,

refers to the set point of the fluid circulating through the heat exchanger. Analysis

in subsequent chapters will show how this is related to the temperature of the jet

entering the test tank. Under “heat location” in Table 2.3, B indicates that only the

bottom heater was active during the experiment. The liquid fill level, given in inches,

is the distance between the liquid-vapor interface and the bottom of the cylindrical

side wall of the test tank. The liquid fill level expressed as a percentage of the nominal

interior tank volume is given in Table 2.5. The nominal interior tank volume is the

region enclosed by the acrylic side, top, and bottom walls. It neglects the contribution

of any ports tapped through the acrylic side wall, as well as the volume of the nozzle
54
2.3 TEST MATRIX EXPERIMENTAL INVESTIGATION

and thermistor inside the tank. To get a sense of the off-nominal value, the volumes

of some of these neglected components are listed in Table 2.6. Since these off-nominal

contributions are small, in subsequent calculations, only the nominal interior tank

volume is used.

When processing the experimental data, in order to filter out small variations in

the initial conditions, the initial pressure and temperature (Po and To , respectively)

are subtracted from later measurements. In this manner, the pressure and temper-

ature responses in subsequent plots begin at zero and may be positive or negative

depending on their value relative to the initial state. The initial pressure and tem-

perature for the different test cases are given in Table 2.7. All tests began within

+1.99 +0.33 o
492 −0.61 Torr and 22.27 −0.18 C. The largest variation occurred when the tank was

fully populated with thermistors which is again why this case was only used to obtain

qualitative temperature maps rather than to infer any quantitative trends.

Also included in Table 2.7 is an estimate of the mole fraction, Xg , of the non-

condensable gas present in the system during testing. In the present experiment,

both the temperature and pressure were measured away from the interface. Due to

a lack of more detailed spatial measurements, these measurements were used to infer

a deviation from saturation. Here, the non-condensable (assumed to be air) and the

vapor are assumed to be ideal gases. The mole fraction of non-condensable is then:

Xg % Pmeas − Psat (22.5 oC)

=
100 Pmeas

55
2.3 TEST MATRIX EXPERIMENTAL INVESTIGATION

Table 2.7: Initial conditions

Case No. Po (Torr) To (o C) Xg
1 492.04 22.20 3.2%
2 492.02 22.20 3.2%
3 491.62 22.18 3.1%
4 492.44 22.23 3.3%
5 491.95 22.23 3.2%
6 491.97 22.37 3.2%
7 491.89 22.28 3.2%
8 492.31 22.32 3.3%
9 491.95 22.31 3.2%
10 491.85 22.29 3.2%
11 492.02 22.26 3.2%
12 492.41 22.23 3.3%
13 493.00 22.10 3.4%
14 491.78 22.33 3.2%

where Pmeas is the measured total ullage pressure. The saturation pressure was eval-

uated at 22.5 o C, the set-point of the fluid initially circulating through the heat

exchanger. It was felt this temperature was more representative of the interfacial tem-

perature rather than the temperatures listed in Table 2.7, which were measured in the

vapor phase away from the interface. While the contaminant gas is non-condensable,

it may dissolve into the liquid phase. The dissolution of the non-condensable has

been neglected in the preceding analysis.

Future experimental tests should employ a more robust degassing scheme to fur-

ther reduce the amount of non-condensables in the system. Moreover, post-test char-

acterization of the different species inside the test tank would yield a more quantitative

measure of the amount of non-condensables present.

56
2.4 DIMENSIONAL ANALYSIS EXPERIMENTAL INVESTIGATION

Table 2.8: Relevant Dimensional Parameters

Parameter Description Definition Range
ρl Vjet Djet
Re Reynolds Number µl
1343 - 4646

2
gρl Djet
Bo Bond Number 4σ
10.07 - 10.34
2 D2
ρl Vjet jet
We Weber Number 8σRt
0.0267 - 0.296

gβl q ′′ R4t
Gr Grashof Number kl νl2
1.6·1010 - 5.2·1010

Ra Rayleigh Number Gr·Pr 1.3·1011 - 4.3·1011

gβl (Ts −Tjet )Hjet

Ri Richardson Number 2
Vjet
0.02 - 1.39

2.4 Dimensional Analysis

The approach of performing small scale experimental testing is to better under-

stand the underlying transport phenomena occurring in the tank and to gather data

that can be used to validate models. While the approach was not meant to achieve

complete static and dynamic similitude with a flight-weight cryogenic system, it is

nevertheless instructive to identify and evaluate the relevant dimensionless param-

eters (Table 2.8). The Reynolds number of the jet suggests that the liquid stream

entering the tank is turbulent. The magnitude of the Bond number indicates that

the static shape of the interface should be flat in this ground-based apparatus. The

magnitude of the Weber number suggests the jet’s momentum is small enough that jet

geysering into the ullage will not occur. The Rayleigh number range indicates that

natural convection in the liquid is not quite fully turbulent but not quite laminar

57
2.5 RESULTS EXPERIMENTAL INVESTIGATION

either. The flow is best characterized as being transitionary. The nearly two orders

of magnitude variation in the Richardson number suggests that for some cases, the

interaction between forced jet mixing and buoyancy will be weak but in other cases

the interaction will be stronger.

2.5 Results

First the effects of heat load and fill level on the self-pressurization behavior

will be studied. Next the effectiveness of using an axial liquid jet to control tank

pressure will be examined. Finally, in the last case, the tank was fully populated

with thermistors. This was done in order to obtain qualitative temperature maps

to help explain the behavior of the local temperature and pressure histories. In the

pressure and temperature histories, the points on the curves are instants in time

from the raw data and were not averaged in any way. All the data points obtained

during each run are not plotted. Sufficient points are included in order to identify

the different curves on the graphs.

2.5.1 Self-Pressurization

Before examining the different case studies, it is useful to gauge the reproducibility

of the data. In Fig. 2.7, the pressure and temperature time histories are shown for

the first nine cases. In these first nine cases, 2 W was applied to the bottom heater

and the liquid fill level was 26.5%. After self-pressurizing for 2 hours, the variation in

pressure and temperature are nearly within the error bars of the measurement. For
58
2.5 RESULTS EXPERIMENTAL INVESTIGATION

20
Case 1
Case 2
Case 3
Case 4
Case 5
Case 6
Case 7
15 Case 8
Case 9
P-P o [Torr]

10

0
0 2000 4000 6000 8000
Time [s]

(a)

1.0
Case 1
Case 2
Case 3
Case 4
Case 5
0.8 Case 6
Case 7
Case 8
Case 9

0.6
T-To [K]

0.4

0.2

0.0
0 2000 4000 6000 8000
Time [s]

(b)

Figure 2.7: Reproducibility of the experimental data. Pressure (a), Tempera-

ture (b).

59
2.5 RESULTS EXPERIMENTAL INVESTIGATION

25
1W
2W
3W

20

15
P-P o [Torr]

10

0
0 2000 4000 6000 8000
Time [s]

(a)

1.2
1W
2W
3W

1.0

0.8
T-To [K]

0.6

0.4

0.2

0.0
0 2000 4000 6000 8000
Time [s]

(b)

Figure 2.8: Effect of heat input on pressure (a) and temperature (b) during
self-pressurization. (Liquid fill level = 26.5%, Bottom heating configuration)

60
2.5 RESULTS EXPERIMENTAL INVESTIGATION

+0.29 +0.03
these first 9 cases, at time = 7000 s, ∆P = 15.93 −0.25 Torr and ∆T = 0.77 −0.04 K.

The data reproducibility is thus deemed excellent.

The effect of applied heat load on the pressure and temperature rise inside the

tank will now be examined. As shown in Fig. 2.8, as the applied heat load is reduced

both the pressure rise and temperature rise decrease. This trend should be quite

intuitive given that less energy is available to raise the sensible energy of the system

and vaporize the liquid. The trend is not quite linear – a doubling of the heat load

does not double the pressure or temperature rise. At 7000 s, the pressure rise after

applying 1 W, 2 W, and 3 W is 9.1 Torr, 16.0 Torr, and 23.3 Torr respectively. The

corresponding temperature rise is 0.46 K, 0.79 K, and 1.16 K. The non-linearity is

due to the fact that at constant ambient temperature the heat loss to the surrounding

varies among the tests. As the applied heat load increases and the sensible energy

of the system rises, the driving potential for heat loss between the test tank and the

ambient surroundings increases as well. Hence at higher applied heat loads, the test

tank is losing more heat than at lower heater powers.

Next the effect of liquid fill level on the pressurization behavior is investigated.

As indicated in Fig. 2.9a, when the bottom heater is activated, the pressure rise is

largest for the lowest fill level. This result can be best understood in terms of how the

applied energy is partitioned in the system. Some of the incident energy is used to

raise the sensible energy of the wall, liquid, and vapor, while some is used for liquid

vaporization. The energy partitions are coupled together; if the sensible energy of

the wall rises faster than that of the liquid or vapor phases, the two phases would
61
2.5 RESULTS EXPERIMENTAL INVESTIGATION

20
26.5%
50.0%
73.5%

15
P-P o [Torr]

10

0
0 2000 4000 6000 8000
Time [s]

(a)

1.0
26.5%
50.0%
73.5%

0.8

0.6
P-P o [Torr]

0.4

0.2

0.0
0 50 100 150 200 250 300
Time [s]

(b)

Figure 2.9: Effect of liquid fill level on pressure (a) during self-pressurization.
Inset of the pressure response (b). (Heat load = 2 W, Bottom heating config-
uration)

62
2.5 RESULTS EXPERIMENTAL INVESTIGATION

1.0
26.5%
50.0%
73.5%

0.8

0.6
T-To [K]

0.4

0.2

0.0
0 2000 4000 6000 8000
Time [s]

(a)

0.05
26.5%
50.0%
73.5%

0.04

0.03
T-To [K]

0.02

0.01

0.00
0 50 100 150 200 250 300
Time [s]

(b)

Figure 2.10: Effect of liquid fill level on temperature (a) during self-
pressurization. Inset of the temperature response (b). (Heat load = 2 W,
Bottom heating configuration)

63
2.5 RESULTS EXPERIMENTAL INVESTIGATION

tend to gain energy at the expense of the wall. At lower fill levels, there is less liquid

in the tank to absorb the incident heat load. Consequently, more energy is available

for vaporization and for raising the sensible energy of the vapor as suggested in Fig.

2.10.

It is also apparent from Fig. 2.9b and 2.10b that there is a lag in both the

temperature and pressure response of the system. The lag is due to the finite amount

of time it takes for heat to reach the vapor phase. For a fixed heater location,

as the liquid fill level increases, the distance between the phase boundary and the

applied heat source increases as well. For the 26.5%, 50% and 73.5% fill levels the

distance between the interface and the top of lower heater is 1.625”, 5.625”, and

9.625” respectively. Based on these distances, the conduction time scales through the

acrylic wall are much larger than the response lag noted in Fig. 2.9b and 2.10b, which

suggests that the incident energy is being transported to the vapor phase through the

bulk liquid. When the warmer fluid adjacent to the heater rises due to buoyancy, it

flows along a cooler acrylic wall since heat is only being applied locally to a band 1”

wide. When the warmer low-conductivity fluid comes into contact with the higher-

conductivity tank wall, the fluid will give up some of its energy to the acrylic and

slow down due to the reduced buoyancy force. This reduction in the buoyancy force

as the fluid flows up along the wall will generally result in the thermal response being

slower than a buoyant time scale.

Additional insight can be gained by observing how the temperature of the outer

tank wall evolves in time during the 2 hour period of self-pressurization. In Fig.
64
2.5 RESULTS EXPERIMENTAL INVESTIGATION

16

14

12
Distance Along Side Wall [in]

10
1800 s
3600 s
5400 s
8 7200 s

6
Approximate Interface Location

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
T-To [K]

Figure 2.11: Temperature profiles along the outer tank wall for the bottom
heating configuration. (Liquid fill level = 50.0%, Heat load = 2 W)

2.11, the time history of wall temperature is presented. The tank is 50% full and

2 W is applied to the heater. For the temperature profiles presented in Fig. 2.11,

only the bottom heater is active. After 2 hours of heat addition, the temperature of

the outer acrylic wall near the heater band has risen by approximately 1 K. Farther

away from the heater, after 2 hours, the wall temperature has risen by 0.5 K. The

wall temperature in regions far from the heater is increasing because of both axial

conduction through the acrylic and heat transfer into the acrylic from the liquid and

vapor phases. Since the acrylic has a higher conductivity than both of the two bulk

phases, warm liquid or vapor adjacent to the wall would tend to lose heat to the

acrylic.

65
2.5 RESULTS EXPERIMENTAL INVESTIGATION

2.5.2 Pressure Control

In addition to studying how various factors effect the self-pressurization behavior

of the liquid-vapor system, the effectiveness of using a mixing jet to control tank

pressure was also investigated. In these pressure control tests, after the tank has

self-pressurized for 2 hours, fluid is withdrawn from the bottom of the tank and

pumped through a heat exchanger before re-entering the tank through an axially-

aligned nozzle. During these 90 minute pressure control tests, the band heaters remain

active supplying heat at the same rate as the preceding 2 hour self-pressurization test.

The mixing loop is a closed system and thus the rate of mass leaving the tank through

the bottom outlet ports is equal to the rate of mass re-entering the tank through the

jet nozzle. The pump, heat exchanger, and flow meter were located on the floor of

the lab beneath the table on which the test tank sat. This provided for a sufficient

hydrostatic head that prevented any cavitation in the fluid line. The temperature of

the secondary fluid in the heat exchanger was a controlled parameter that was used

to vary the degree of subcooling of the liquid re-entering the tank.

Before discussing the subcooled mixing results, we first consider the case when

only the pump is activated. In this quasi mixing-only case, the temperature of the jet

is not being controlled by the heat exchanger. Prior to turning on the jet, both the

fluid in the lines and the fluids in the heat exchanger are stationary and assumed to be

in thermal equilibrium with the mean lab temperature. Once the jet is activated, both

the pressure and temperature inside the tank initially decrease as shown in Fig. 2.12.

66
2.5 RESULTS EXPERIMENTAL INVESTIGATION

40

35

30
P-P o [Torr]

25

20

15

10

5
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(a)

2.5

2.0
T-To [K]

1.5

1.0

0.5
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(b)

Figure 2.12: Effect of axial jet mixing on pressure (a) and temperature (b).
(Liquid fill level = 26.5%, Heat load = 2 W, Bottom heating configuration, Jet
speed = 24.2 cm/s)

67
2.5 RESULTS EXPERIMENTAL INVESTIGATION

The temporary reduction is due to the passive subcooling of the fluid re-entering the

test tank. During the 2 hour self-pressurization experiment immediately preceding

this quasi mixing-only test, heat is being added to the band heater adjacent to the

liquid phase. Because of buoyant convection, the temperature in the liquid phase

stably stratifies with the cooler fluid settling to the bottom of the tank. When the

pump is first turned on, this cooler fluid is drawn into the fluid lines which are

subcooled themselves because they have been sitting close to the initial temperature

of the tank for the previous 2 hours. When this subcooled fluid re-enters the tank

and impinges on the liquid-vapor interface, condensation is promoted and heat is

transferred from the vapor into the bulk liquid – both of which can lead to a pressure

collapse in the ullage. Fig. 2.12a indicates that the pressure collapse is only temporary

and that after the initial reduction in pressure, when the liquid becomes fully mixed,

the ullage pressure continues to rise as before.

This should be expected given that the liquid subcooling was not sustained during

the test. After the initially subcooled liquid re-enters the tank, it mixes with the

warmer bulk liquid. This warmer liquid is eventually drawn into the fluid line before

being pumped back into the tank. While passive subcooling can certainly occur –

the fluid circulating in the line can lose heat to the ambient environment – these heat

losses are expected to be smaller and not comparable in magnitude to the energy

supplied to the band heater. Hence the pressure continues to rise even though mixing

is being performed. Moreover, the power supplied to the pump is an energy source

into the system which if unmitigated can further cause the tank pressure to increase.
68
2.5 RESULTS EXPERIMENTAL INVESTIGATION

So while axial jet mixing is able to thermally destratify the bulk liquid, it is not

particularly effective at reducing tank pressure over the long term. To sustain the

initial pressure collapse, energy must be removed from the liquid-vapor system.

Next, a sustained pressure reduction is investigated by making use of a subcooled

liquid jet. Here, subcooling the liquid is not passively performed as in the previous

case, but is actively maintained by circulating a cold thermally-conditioned fluid

through the heat exchanger. As the test liquid is pumped through the heat exchanger,

it continuously loses heat to the colder fluid and thus the subcooling is maintained.

First the effect of jet speed on the pressure reduction is considered. As expected,

the pressure decays faster for faster jet speeds. As indicated in Fig. 2.13a, after 90

minutes of subcooled jet mixing, the pressure has barely decreased for the slowest jet

speed. There are two factors that reduce the effectiveness of a subcooled jet mixer.

First, the colder jet is trying to penetrate into a warmer bulk liquid. In order to

reach the liquid-vapor interface, the jet’s momentum must overcome the retarding

buoyancy forces which inhibit the jet’s upward flow by forcing the colder incoming

liquid stream to the bottom of the tank. This phenomena would be present even if

the warmer bulk liquid were quiescent. The retarding effect of buoyancy is further

compounded by natural convection in the liquid. In the present experiment, a toroidal

natural convection vortex is established in the bulk liquid which causes warm fluid

near the heater to rise along the tank wall, flow radially inward along the free surface,

and then get pulled downward near the centerline of the tank before resupplying the

upwardly traveling fluid near the heater. The downward natural convection flow
69
2.5 RESULTS EXPERIMENTAL INVESTIGATION

20

10

0
P-P o [Torr]

-10

-20

7.2 cm/s
-30 9.3 cm/s
11.4 cm/s
13.5 cm/s
15.7 cm/s
24.1 cm/s
-40
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(a)

1.25

1.00

0.75

0.50

0.25
T-To [K]

0.00

-0.25

-0.50 7.2 cm/s

9.3 cm/s
11.4 cm/s
13.5 cm/s
-0.75
15.7 cm/s
24.1 cm/s

-1.00
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(b)

Figure 2.13: Effect of jet speed on pressure (a) and temperature (b). (Liquid fill
level = 26.5%, Heat load = 2 W, Bottom heating configuration, Jet temperature
= 20 o C)

70
2.5 RESULTS EXPERIMENTAL INVESTIGATION

near the central axis of the tank suppresses the upwardly traveling cold jet flow.

Once again, the jet’s momentum must be large enough to overcome the suppressive

natural convection flow in order for the jet to reach the interface. Once the colder

jet finally reaches the phase boundary, the liquid removes heat from the vapor phase,

condensation begins, and the tank pressure decays.

The pressure histories shown in Fig. 2.13 indicate that for moderate jet speeds,

the ullage pressure decays slowly at first and then is followed by a rapid collapse.

During the period of slow decay, the forced jet flow is interacting with and competing

with the strong natural convection in the bulk liquid. Even though the subcooled

jet has not yet reached the interface, an energy exchange is taking place between the

toroidal natural convection vortex and the colder jet fluid that is mainly isolated at

the bottom of the tank. The natural convection vortex pulls some of the colder fluid

up towards the interface which leads to a heat transfer out of the ullage and thus a

pressure reduction. As the natural convection vortex continues to exchange energy

with the incoming subcooled liquid, it shrinks in size which allows the cold jet flow

to penetrate deeper into the bulk. When the jet penetrates far enough so that the

cold fluid directly impinges the free surface, the ullage pressure rapidly decreases.

For the fastest flow rate shown in Fig. 2.13, we see that there is no slow decay.

The jet’s momentum is large enough to overcome the retarding effects of buoyancy

and natural convection. As the jet speed decreases, the duration of slow pressure

decay increases as expected. For the slowest jet speed, the jet’s momentum is not

sufficient to overcome buoyancy and natural convection. In this case, it is unlikely

71
2.5 RESULTS EXPERIMENTAL INVESTIGATION

3500 ∆t = 4732.1 Ri + 593.55 Ri - 182.86

2

2
R =1

3000

2500

2000
∆t [s]

1500

1000

500

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Ri

(a)

Figure 2.14: Effect of Richardson number on the onset of rapid pressure col-
lapse.

that the jet impinges the free surface. As a result, a rapid pressure collapse is not

observed.

For the cases were a rapid pressure collapse followed a slow period of pressure de-

cay, the onset of the rapid depressurization can be characterized in terms of Richard-

son number. The Richardson number is a dimensionless measure of the strength

of the interaction between forced jet mixing and buoyancy. For small Richardson

numbers, the jet’s momentum is stronger than the retarding buoyancy forces in the

liquid. For large Richardson numbers, buoyancy can suppress the upward motion of
72
2.5 RESULTS EXPERIMENTAL INVESTIGATION

the mixing jet. In Fig. 2.14, the time lag between turning on the jet and observing

a rapid depressurization is characterized as a function of Richardson number. The

correlation does an excellent job of fitting the data and suggests that the time lag is

a quadratic function of Ri.

The corresponding temperature histories for the pressure collapse cases described

above are shown in Fig. 2.13b. For the slowest jet speed, the temperature in the

ullage decreases almost monotonically. As the jet speed increases, there is initially

only a slight decrease in ullage temperature. When the cold jet fluid finally reaches the

liquid-vapor interface and the pressure rapidly collapses, the temperature in the ullage

decreases rapidly as well. The point where the rapid temperature decay commences

corresponds closely to when the pressure rapidly collapses. As noted previously, as

jet speed increases, the transition from slight temperature decay to rapid collapse

occurs earlier due to the greater momentum of the liquid jet. In contrast to the

depressurization behavior however, for the faster jet speeds, the temperature is not

monotonically decreasing. In all of the cases except the two slowest speeds, the rapid

temperature reduction is followed by a temperature jump which occurs later in time

as the speed decreases. It is suspected that this temperature jump is being caused by

buoyant heating of the ullage from the non-wetted wall. When the cold jet impinges

the interface, condensation occurs and the ullage pressure collapses. The resulting

heat and mass flux out of the ullage, through the interface, acts as an energy sink

causing a rapid drop in ullage temperature. As ullage pressure begins to asymptote

to its final steady state value, the condensation rate decreases and the energy sink
73
2.5 RESULTS EXPERIMENTAL INVESTIGATION

at the interface is reduced. As the temperature of the ullage decreases, however, the

heat transfer rate between the ullage and the wall increases. Energy leaking into the

tank from the wall rises to the top of the tank due to buoyancy, eventually leading

to a stably stratified ullage.

Thus, the onset of the thermal jump, which occurs over a time scale of minutes, is

caused by condensation. While the temperature of the ullage is dropping, the ullage

is losing energy through the interface but gaining energy through the walls. As the

condensation rate slows, heat leaking into the tank causes the temperature to rise

locally. Since the jet is still active and flowing along the liquid side of the interface

a steady state is reached where the amount of energy entering the ullage through

the walls is balanced by the flux of energy leaving the ullage through the interface.

Consequently, the ullage pressure does not rise.

In addition to investigating the effect of jet speed on the depressurization behav-

ior, the effect of jet subcooling was also studied. The depressurization histories for

different jet subcoolings are shown in Fig. 2.15 which shows that the pressure col-

lapses more rapidly as the temperature of the jet decreases. This is not surprising

since a colder jet removes energy more effectively from the liquid-vapor system than a

warmer one. The three curves asymptote to different values. If jet mixing completely

homogenizes the thermal field in the tank, the ullage pressure, neglecting the hydro-

static component, would equal the saturation pressure. Since saturation pressure is

an increasing function of temperature, a warmer liquid jet causes the ullage pressure

in a well-mixed tank to asymptote to a higher value. The asymptotic response is

74
2.5 RESULTS EXPERIMENTAL INVESTIGATION

25

0 o
17.5 C
o
20.0 C
o
22.5 C
P-P o [Torr]

-25

-50

-75
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(a)

1.0

0.5

0.0
T-To [K]

o
17.5 C
o
20.0 C
o
22.5 C
-0.5

-1.0

-1.5
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(b)

Figure 2.15: Effect of jet subcooling on pressure (a) and temperature (b). (Liq-
uid fill level = 26.5%, Heat load = 2 W, Bottom heating configuration, Jet
speed = 24.2 cm/s)

75
2.5 RESULTS EXPERIMENTAL INVESTIGATION

16

14

12
Distance Along Side Wall [in]

Seconds After Activating Jet

10
0s
1800 s
3600 s
8 5400 s

6
Approximate Interface Location

0
-1 0 1 2 3 4 5
T-To [K]

Figure 2.16: Temperature profiles along the outer tank wall for the bottom
heating configurations. (Liquid fill level = 50.0%, Heat load = 2 W, Jet speed
= 24.2 cm/s, Jet temperature = 20 o C)

indicative of a steady-state whereby the heat removed by the subcooled jet and lost

to or gained from the ambient environment exactly balances the applied heat load.

The effect of subcooling on ullage temperature is shown in Fig. 2.15b, For the

warmest jet, the ullage temperature decreases and is quite noisy which can be at-

tributed to convection in the vapor. For the colder jets, the temperature decreases

rapidly initially and then jumps to a higher value which we attribute to thermal inver-

sion in the vapor. Both the rate and magnitude of decay increase with jet subcooling

because as the temperature difference between the jet and system increases, the liquid

jet can more effectively remove energy from the system.

Additional insight can be gained by plotting the time evolution of the wall tem-

76
2.5 RESULTS EXPERIMENTAL INVESTIGATION

perature profiles during subcooled jet mixing. The wall profiles shown in Figs. 2.16

correspond to when the tank is 50% full under a liquid heating configuration. The

temperature of the wall is hottest near where the heat load is applied. The liquid

jet, subcooled to below its initial temperature, removes energy from the wall as well

as from the liquid-vapor system. Consequently, after 90 minutes of subcooled jet

mixing, the temperature of the wall adjacent to the liquid phase has been reduced to

below its initial temperature. Interestingly, the temperature of the wall far from the

heat source and adjacent to the vapor phase has not changed appreciably after 90

minutes. This suggests that any cooling of the wall adjacent to the vapor is driven by

axial conduction through the acrylic which has previously been shown to be a slow

process.

Even though the entire acrylic wall is blanketed in foam insulation, realistically

the insulation is not perfect and some heat is inevitably transferred between the wall

and the ambient environment whose average temperature is nearly constant and equal

to the initial temperature. Thus when the wall temperature is greater (less) than its

initial value, the wall can locally lose (gain) heat to (from) the environment. While

these heat additions and subtractions may cancel each other out, a zero heat flux

condition is not enforced on the outer wall of the tank. The best the insulation can

do is to suppress the thermal link between the tank and its surroundings.

In addition to studying various jet speeds and subcoolings, the focus now will be

on a particular jet speed and temperature and the effects of heat addition and liquid

fill level on the depressurization behavior will be investigated. For these cases, the
77
2.5 RESULTS EXPERIMENTAL INVESTIGATION

jet speed was 24.1 cm/s and the temperature was set to 20 o C.

The depressurization histories for three different heat inputs are presented in Fig.

2.17. The depressurization rates are not easily discernible in Fig. 2.17a because

at time = 0, the liquid-vapor system is in a different thermodynamic state prior to

turning on the jet owing to the three different applied heat loads. Nonetheless, the

depressurization rate can be deduced from the data by computing:

[P (0s) − Po ] − [P (600s) − Po ]
(2.2)
600s

where time = 0 corresponds to when the jet is first activated and the 600 s time interval

was chosen arbitrarily. For the 1 W, 2 W, and 3 W cases, the depressurization rate is

0.040 Torr/s, 0.046 Torr/s, and 0.053 Torr/s, respectively. The depressurization rate

increases with heat load because the rate of decay is a strong function of the heat

removal rate which is initially higher for the 3 W case. As indicated in Fig. 2.17b, as

more heat is applied to the system, at the end of the self-pressurization phase of the

experiment, the system temperature is higher. Thus, when the jet is first activated,

the driving potential for heat exchange – the temperature difference between the bulk

fluid and the subcooled jet – is largest for the 3 W test. If the upward motion of

the jet is not restricted by buoyancy or natural convection, the large temperature

differences result in a higher heat removal rate and consequently a faster pressure

reduction.

Similar trends are noted for the ullage temperature. Defining the temperature

78
2.5 RESULTS EXPERIMENTAL INVESTIGATION

30

20

10

1W
P-P o [Torr]

2W
3W
0

-10

-20

-30
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(a)

1.5

1.0

0.5
T-T o [K]

0.0

-0.5 1W
2W
3W

-1.0
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(b)

Figure 2.17: Effect of heat input on pressure (a) and temperature (b) during
subcooled jet mixing. (Liquid fill level = 26.5%, Bottom heating configuration,
Jet speed = 24.2 cm/s, Jet temperature = 20 o C)

79
2.5 RESULTS EXPERIMENTAL INVESTIGATION

reduction rate similarly to eqn. (2.2), one finds that for the 1 W, 2 W, and 3 W

cases, the temperature is decaying at 1.90×10-3 K/s, 2.05×10-3 K/s, and 2.45×10-3

K/s respectively. For reasons outlined above, the rate of reduction is again greatest

for the 3 W case.

Next the effects of liquid fill level on the depressurization behavior for the liquid

heating configuration will be studied. For the lowest fill level shown in Fig. 2.18,

the pressure response is immediate – the ullage pressure begins to collapse as soon as

the subcooled jet is activated. The response is more delayed for the higher fill levels.

For the 50% case, the pressure decreases slightly for the first 500 s before rapidly

collapsing. At a liquid fill level of 73.5%, the response is significantly delayed. Rapid

depressurization doesn’t commence until 2800 s after turning on the jet. These delays

are significantly longer than the L/U transit times of 0.84 s and 1.26 s for the 50%

and 73.5% cases respectively. In order for the subcooled jet to reach the interface

and initiate the rapid pressure collapse, the incoming liquid stream must first remove

enough energy from the bulk liquid to weaken the natural convection vortex that’s

inhibiting the jet’s upward motion. For the highest fill level, the bulk liquid has

absorbed more of the incident heat load than in the lower fill level cases. Since

the subcooled jet would have to remove more sensible energy from the bulk liquid

before reaching the interface the pressure response would be delayed. Additionally,

for the 73.5% fill level case, the centrally located T1 thermistor is submerged in the

liquid phase and is directly in the path of the incident jet stream. Fluid that should

otherwise reach the interface stagnates on the thermistor. This inhibits the upward
80
2.5 RESULTS EXPERIMENTAL INVESTIGATION

20

10

0
P-P o [Torr]

26.5%
50.0%
-10 73.5%

-20

-30
0 1000 2000 3000 4000 5000 6000
Seconds After Activating Jet

(a)

1.0

0.5

0.0
T-To [K]

26.5%
50.0%
-0.5 73.5%

-1.0

0 1000 2000 3000 4000 5000 6000

Seconds After Activating Jet

(b)

Figure 2.18: Effect of liquid fill level on pressure (a) and temperature (b) during
subcooled jet mixing. (Heat load = 2 W, Bottom heating configuration, Jet
speed = 24.2 cm/s, Jet temperature = 20 o C)

81
2.5 RESULTS EXPERIMENTAL INVESTIGATION

motion of the liquid jet and also delays the pressure collapse response.

Similar delayed responses are noted in the temperature histories shown in Fig.

2.18b. No temperature jump is observed for the highest fill level but at this level the

thermistor is submerged. The noisier response for this case is due to temperature

fluctuations brought about by buoyancy / forced jet interactions.

2.5.3 Visualization

In this section, the term “visualization” is used loosely to describe the temperature

maps obtained by fully populating the tank with thermistors. One-to-one quantita-

tive comparisons between this last visualization test and the previous quantitative

measurements are not appropriate for several reasons. First, as noted earlier, the

leak rate into the tank was slightly larger when fully populated with thermistors.

Secondly, and this is particularly true during subcooled mixing, the thermistors in-

terfered with the flow field. If a thermistor was in the path of the incoming jet,

fluid would stagnate on the thermistor instead of being carried to the interface. This

tended to retard the depressurization rate. Thirdly, the flow meter was changed out

prior to running the visualization tests. The replacement was a higher capacity flow

meter which resulted in a decreased pressure drop and thus larger throughput for

the same amount of power supplied to the pump. Even with these differences from

previous tests, the qualitative temperature maps are nevertheless useful and can be

used to support some of the hypotheses laid out in previous sections.

To construct the temperature map, the plane containing the thermistors was tri-

82
2.5 RESULTS EXPERIMENTAL INVESTIGATION

Figure 2.19: Data triangulation of the plane containing the thermistors.

83
2.5 RESULTS EXPERIMENTAL INVESTIGATION

angulated as shown in Fig. 2.19. Every vertex in the triangulated plane corresponds

to a thermistor or RTD location. Here, axisymmetry is assumed which effectively

doubled the number of off-axis vertices. The triangulation and pointwise tempera-

tures were then read into tecplot

R
and an internal contouring algorithm was used to

construct a temperature field by interpolating the vertex values.

In this fully-populated tank, the fill level is 26.5% and 2 W is applied to the bot-

tom heater. The ullage pressure history corresponding to this case, shown in Fig.

2.20, exhibits characteristics similar to previous bottom heating test cases. The tem-

perature histories at two locations in the tank are shown in Fig. 2.20. Thermistor T1

is recording temperature in the ullage at the same location as previous temperature

measurements. Not surprisingly then, the temperature trace is similar to previous

cases whereby the rapid decay beginning when the jet is first turned on is followed by

a temperature jump. The histories presented in Fig. 2.20b also indicate that the tem-

perature in the liquid above the heating band (T13) is rising faster than temperature

in the vapor. Hence during self-pressurization heat is being transferred from the liq-

uid into the vapor. During subcooled mixing, both the liquid and vapor temperatures

decay at nearly the same rate initially. After approximately 750 s of subcooled jet

mixing, the vapor temperature jumps while liquid temperature continues its decay.

The absence of a jump in the liquid is due to both a stable stratification prior to

jet activation as a result of the bottom-heating configuration and a strong subcooled

mixing jet which tends to homogenize the bulk and isolate any local hot spots to

regions near the heater.

84
2.5 RESULTS EXPERIMENTAL INVESTIGATION

20

10

0
P-P o [Torr]

-10

-20

-30

-40
0 2000 4000 6000 8000 10000 12000
Time [s]

(a)

1.5
B

1.0

0.5

0.0 A
T-To [K]

-0.5

-1.0

-1.5 T1
T13

-2.0
0 2000 4000 6000 8000 10000 12000
Time [s]

(b)

Figure 2.20: Pressure (a) and temperature (b) histories. (Fill level = 26.5%,
Heat load = 2 W, Bottom heating configuration, Jet flow = 800 mL/min, Jet
temperature = 20 o C)

85
2.5 RESULTS EXPERIMENTAL INVESTIGATION

(A) (B)

Time = 3300 s Time = 7190 s

296.60
296.55
296.50
296.40
296.20
296.00
295.80
295.75

86
2.5 RESULTS EXPERIMENTAL INVESTIGATION

(C) (D)

Time = 7380 s Time = 8550 s

295.80
295.60
295.40
295.20
295.00
294.80
294.60
294.40
294.20
294.00
293.80

Figure 2.21: Temperature contours at various times during self-pressurization

and subcooled jet mixing. (a) ∆Tmax = 0.708 K, ∆Tmin = 0.052 K. (b) ∆Tmax =
1.126 K, ∆Tmin = 0.273 K. (c) ∆Tmax = 0.772 K, ∆Tmin = -1.502 K. (d) ∆Tmax
= 0.528 K, ∆Tmin = -1.788 K.

87
2.5 RESULTS EXPERIMENTAL INVESTIGATION

In order to further investigate this thermal response, the temperature field, mea-

sured at discrete points and then interpolated onto the triangulated domain, is pre-

sented in Fig. 2.21 at time stamps A, B, C, and D as labeled in Fig. 2.20. While

there appears to be quite a variation in temperature along the interface, this is al-

most certainly an artifact of constructing a field map by linearly interpolating between

discrete points in the domain. Interpolating between discrete points also produces

a choppiness in the temperature field that is unphysical. Nevertheless, these maps

provide a useful qualitative description of the thermal response in the system.

At 3300 s into the self-pressurization experiment and continuing until subcooled

mixing is about to begin, the temperature fields shown in Figs. 2.21a and 2.21b indi-

cate that the liquid is stably stratified with the fluid at the bottom of the tank colder

than fluid above the heater. In the ullage, vapor near the phase boundary is warmer

than the bulk vapor which indicates that prior to turning on the jet, the vapor was in

a unstable thermal configuration. Heat penetrates into the ullage and is carried up-

wards by conduction, buoyancy, and the flux of molecules leaving the phase boundary.

The temperature maps also indicate that the ullage is warmer than its bounding walls

and hence the vapor loses energy to the acrylic during self-pressurization. Nearer to

the interface the wall temperature increases due to its proximity to the warmer liquid

phase. The colder walls at the top of the tank confirm that axial conduction is not

dominant and suggests that an increase in the temperature of the wall adjacent to

the vapor phase is due primarily to heat lost from the ullage.

During subcooled jet mixing, temperature maps are shown in Figs. 2.21c and
88
2.5 RESULTS EXPERIMENTAL INVESTIGATION

2.21d. During the period of rapid pressure decay, the ullage is essentially isothermal.

After the temperature jump, heat entering the ullage from the wall rises to the top

of the tank which leads to a stably stratified ullage as shown in Fig. 2.21d.

Further insight can be gained by plotting the time history of vapor temperature

shown in Fig. 2.22b. Prior to turning on the jet, the hottest vapor temperature is near

the interface. Not surprising then, the warmest part of the non-wetted wall is adjacent

to the interface. At the end of the 90 minute pressure control test, the hottest vapor

has risen to the top of the tank. It is also interesting to note the degree of thermal

stratification during both the self-pressurization and pressure control phases of the

experiment. After self-pressurizing for two hours the difference between the minimum

and maximum ullage temperatures is only 0.2 K. After the jump, the stratification is

much more pronounced with the difference between minimum and maximum vapor

temperatures now 2.25 K. This pronounced, and now stable, stratification during the

pressure control test is indicative of the ability of the subcooled jet to remove heat

from the ullage. It is interesting to note that temperatures jump to higher values

at different times. For thermistors away from the cold interface, the times at which

thermistor temperatures jump to a higher value is a strong function of how close they

are to the hottest parts of the non-wetted wall. This suggests that these temperature

jumps are local in nature and are being driven by wall heating.

Stratification is also present in the liquid during self-pressurization as indicated

in the temperature histories shown in Fig. 2.22a. The temperature of liquid below

the heater (T14 and T15) rises much more slowly that fluid above the heater due
89
2.5 RESULTS EXPERIMENTAL INVESTIGATION

1.5

1.0

0.5

0.0
T-To [K]

-0.5

-1.0

-1.5

T12
-2.0 T13
T14
T15

-2.5
0 2000 4000 6000 8000 10000 12000
Time [s]

(a)

1.0
J J
K
J
K K
J J
K H H
F
J J
K K
H H
F F
EI EI
G
J
K K H
F F
EI EI
G G D
J J
K H
F
H
F
EI EI
G G D D A
J
K
J
K K H
F
H
F
EI EI
G G D D A A
C C
J
K
J
K H H
FI
H
FI
G
E EI
G G
D D D
A A
C A
C C
B B B
J J
K H
FI FI E
G D A A
C C B
J K H H
FI E
G E
G D D A
C C B B
J HJ K
K H
FI FI
E
G E
G D D A
C A
C B B
J HK H
FI E
G D D A A
C B B
J HK FI G
E E D A
C C B B
0.5 J H
K K FI G
E
FI G D A A
C B
J H
K E
FI G
E D DA CA C B B
J H
K
J H
K FI G
E
FI G D DA CA CB B B
I
F E
G D A C B
J H
J H
K E E D
FI G A CA C B A A A A A A A A A A A A
J H
K
K FI G
E
FI G D
A C
D
A C B B B D
D B
E
D B D
B D
C
E C
C C
B
B D B
D
C C B C
B D
C C
B D B
B C
A
B
C A
B A A A
B C A A A A B
J H
K E
FI G
D D
A C B B B C
F B D E GC E D D D D C C
B C B C
B C
B C
B C A
B C
J H E
FI G
D A CB B B F C
F G G G E E E E E D D D D D D D D D
K E
FI G
A C
D A C F H C A F F E E E
J H
K
J H
K E
FI G
E D
A C B B F F G G G G G G E E E E E E
H
JI G
K E
D
A
A C
FI G
D
C
B B B C H F F F F F G G G G
E E E E
H
F
E
JI G
D C H HI
K
K
H
F
E
D
JI G
A
C
B B B
A
C
HI HI I I
F F F
F
G G G G
G G G
F
H
F
D
E
C
A
G
B I H H I F F F
G
JI
K
F
D
H
A
E
C
G
B F F F
0.0 G
D
H
C
A
B
E
K
F
JI H I I F F
H H I I
HI H H I I
F H H HI I
K
G
E H HI I
A H
D
C H HI HI
B I HI HI I
J H HI
E
T-To [K]

-0.5
B
I
G D G K
A
K
E
C
D K
K K
K
J K
K
K
A T1 K K
G K K K
-1.0 B T2 A
K
E
K
K K K
D
C T3 K K
K K
J K
D T4 G
A
E T5 K
E
F T6 J
K
G T7
-1.5
H T8 J

I T9 J
J T10 J
J
K T11 J J
J J
J J
J J J J J J J J
J J J J J
-2.0
0 2000 4000 6000 8000 10000 12000
Time [s]

(b)

Figure 2.22: Temperature histories in the liquid (a) and vapor (b). (Fill level =
26.5%, Heat load = 2 W, Bottom heating configuration, Jet flow = 800 mL/min,
Jet temperature = 20 o C)

90
2.5 RESULTS EXPERIMENTAL INVESTIGATION

to the buoyant rise of warm fluid. One of the thermistor measurements above the

heater (T12) is located in the bulk 3.875” away from the wall. This indicates that the

temperature rise in the liquid is not confined to a thin thermal boundary layer near

the wall. The natural convection vortex is pulling warmer fluid up from the heater

and recirculating it along the interface and then downward near the tank centerline.

91
 Chapter 3

Theory

3.1 What is an Interface?

l
Microscopically, an interface is a thin (ǫ = L
<< 1) 3D transition region separat-

ing two bulk phases. The diffuse 3D interface, over which properties are continuous

but undergo rapid changes, becomes a 2D singular surface as ǫ → 0. In this limiting

macroscopic viewpoint, the interface is treated as a Gibbsian dividing surface over

which interfacial properties can be assigned.

In Fig. 3.1, a cylindrical vessel contains two homogeneous bulk phases separated

by an interface. The density variation is plotted to the right. As the interface is

approached from above, the density rapidly changes from its value in the vapor phase

to its value in the liquid phase over a short distance, l. If the interface is treated as a

2D dividing surface, the density of each homogeneous bulk phase is assumed constant

92
3.1 WHAT IS AN INTERFACE? THEORY

ρv

Vapor
z

L l
r

Liquid

ρl

Figure 3.1: Sketch of a liquid-vapor system and corresponding density profile.

up to the interface:

ρ(z = 0+ ) = ρv (3.1)

ρ(z = 0− ) = ρl (3.2)

By neglecting the continuous variation of density across the interface, a surface excess

or deficiency is created between the actual density and the bulk phase values. The

surface excess, shown in Fig. 3.1 as the shaded region in the density plot, can be

expressed as:
Z Z Z Z 
ρs dA = ρ dV − ρv dV + ρl dV (3.3)
Σ V Vv Vl

The areal surface density, ρs , when integrated over the surface, Σ, gives the mass

excess or deficiency of the system. It should be noted that ρs is not the actual value

93
3.2 MATHEMATICAL DESCRIPTION THEORY

∂Ω
y
Ω-
R
x

Vapor

^
n

Ω+

Liquid

Figure 3.2: Geometric description of the two-phase system.

of ρ at the interface. Rather, it is the value that is assigned to the interface to account

for the 3D continuous property variation.

3.2 Mathematical Description

Consider a spherical vapor bubble centered in a pool of liquid as sketched in Fig.

3.2. The domain, Ω, is divided into a liquid region, Ω+ and a vapor region, Ω− . The

vapor region is bounded by a 2D non-Euclidean surface, ∂Ω. The level set method,

which can be used to capture the interfacial dynamics, embeds information about the

2D surface into a function, φ, defined in 3D space.

To make this clearer, the interface sketched in Fig. 3.2 can be described by the

94
3.2 MATHEMATICAL DESCRIPTION THEORY

zeroth isocontour of the following function

φ = x2 + y 2 + z 2 − R2 . (3.4)

The normal to the surface can be defined in terms of this level set function:

∇φ
n̂ = (3.5)
|∇φ|

where the normal vector points in the direction of increasing φ – into the liquid for the

configuration depicted in Fig. 3.2. Once the normal is defined, the mean curvature

of the surface can be computed from

1
H = (∇ · n̂) (3.6)
2

where the curvature is positive if the interface curves away from the normal as it does

in Fig. 3.2.

In what follows, the surface projection operator

Is = I − n̂n̂ (3.7)

will be used. The operator projects a vector defined in 3D onto the surface. The

95
3.2 MATHEMATICAL DESCRIPTION THEORY

surface gradient, for example, is given by

∇s = Is · ∇. (3.8)

The interface can deform in time. Associating the zeroth isocontour with the interface

for all time requires that

dφ
=0 (3.9)
dt

or after expanding the total derivative

∂φ
+ vs · ∇φ = 0 (3.10)
∂t

where vs is the interfacial velocity. The advection equation for φ can be rewritten

using eqn. (3.5):

∂φ ∇φ
+ vs · |∇φ| = 0 (3.11)
∂t |∇φ|
∂φ
+ (vs · n̂)|∇φ| = 0. (3.12)
∂t

In the derivations that follow, the Heaviside function




1 φ > 0

H(φ) = (3.13)



0 φ < 0

96
3.2 MATHEMATICAL DESCRIPTION THEORY

and the Dirac function

∂H
δ(φ) = (3.14)
∂φ

will be used. Various operations will be carried out the Heaviside and Dirac functions.

The gradient of the Heaviside function is:

∂H
∇H = ∇φ
∂φ
∂H ∇φ
= |∇φ|
∂φ |∇φ|
(3.15)
∂H
= |∇φ|n̂
∂φ

= δ(φ)|∇φ|n̂.

The time derivative of the Heaviside function is given by:

∂H ∂H ∂φ
=
∂t ∂φ ∂t
∂H (3.16)
=− (vs · n̂)|∇φ|
∂φ

= −δ(φ)|∇φ|(vs · n̂).

The 3D analog to the 1D Dirac function defined by eqn. (3.14) is

δ(x) = ∇H · n̂, (3.17)

97
3.2 MATHEMATICAL DESCRIPTION THEORY

which after making use of eqn. (3.15) can be rewritten as

δ(x) = δ(φ)|∇φ|. (3.18)

The Heaviside and Dirac functions are useful when defining quantities that vary over

the entire domain. The density, for example, can be expressed as:

ρ = ρv χ− + ρl χ+ + ρs δ(x) (3.19)

where

χ− = 1 − H(φ) (3.20)

χ+ = H(φ) (3.21)

Thus,

x ∈ Ω− , ρ = ρv (3.22)

x ∈ Ω+ , ρ = ρl (3.23)

x ∈ ∂Ω, ρ = ρs . (3.24)

The following operations will be needed in subsequent sections and follow directly

98
3.3 CONSERVATION LAWS THEORY

from the previous discussion:

∇χ− = −∇H = −δ(x) · n̂ (3.25)

∇χ+ = ∇H = δ(x) · n̂ (3.26)

∂χ− ∂H
=− = δ(x)(vs · n̂) (3.27)
∂t ∂t
∂χ− ∂H
= = −δ(x)(vs · n̂) (3.28)
∂t ∂t

3.3 Conservation Laws

Using the operators defined in Section 3.2, the balance equations for mass, mo-

mentum, and energy can now be derived. The derivation shall be explicit for the

mass conservation equation but abbreviated for the other conservation laws.

3.3.1 Mass Conservation

Beginning with the following decompositions:

ρ = ρv χ− + ρl χ+ + ρs δ(x) (3.29)

ρv = ρv vv χ− + ρl vl χ+ + ρs vs δ(x) (3.30)

the continuity equation for the entire liquid-vapor-interfacial system can be expressed

as:
∂ρ
+ ∇ · (ρv) = 0 (3.31)
∂t

99
3.3 CONSERVATION LAWS THEORY

Substituting eqns. (3.29) and (3.30) into the continuity equation yields

∂ρv ∂χ−
χ− + ρv + χ− ∇ · (ρv vv ) + ρv vv · ∇χ− +
∂t ∂t
∂ρl ∂χ+ (3.32)
χ+ + ρl + χ+ ∇ · (ρl vl ) + ρl vl · ∇χ+ +
∂t ∂t
∂ρs ∂δ
δ(x) + ρs + δ(x)∇ · (ρs vs ) + ρs vs · ∇δ = 0
∂t ∂t

Grouping terms together and making use of eqns. (3.25)-(3.28) yields

 
− ∂ρv
χ + ∇ · (ρv vv ) +
∂t
 
+ ∂ρl
χ + ∇ · (ρl vl ) +
∂t
  (3.33)
∂ρs
δ(~x) + ∇ · (ρs vs ) + ρl (vl − vs ) · n̂ − ρv (vv − vs ) · n̂ +
∂t
dδ
ρs = 0.
dt

Due to the arbitrariness of Ω± , the terms post-multiplying χ± must go to zero sepa-

rately. Hence in the bulk phases,

∂ρv
x ∈ Ω− : + ∇ · (ρv vv ) = 0 (3.34)
∂t
∂ρl
x ∈ Ω+ : + ∇ · (ρl vl ) = 0. (3.35)
∂t

Bedeaux et al.188 rigorously show that

dδ
=0 (3.36)
dt

100
3.3 CONSERVATION LAWS THEORY

and so the proof will be omitted here. Furthermore, they demonstrated that equations

of the form

Aδ(~x) + B · ∇δ = 0 (3.37)

imply that

A = B · n̂ = 0. (3.38)

Hence, from eqn. (3.33)

∂ρs
+ ∇ · (ρs vs ) + ρl (vl − vs ) · n̂ − ρv (vv − vs ) · n̂ = 0 (3.39)
∂t

If the surface excess contributions are neglected, i.e.:

∂ρs
+ ∇ · (ρs vs ) = 0 (3.40)
∂t

then the mass flux on either side of the interface are equal:

ρl (vl − vs ) · n̂ = ρv (vv − vs ) · n̂ (3.41)

For convenience, let the condensation mass flux be defined as:

j ≡ ρl (vl − vs ) · n̂ = ρv (vv − vs ) · n̂ (3.42)

101
3.3 CONSERVATION LAWS THEORY

3.3.2 Balance of Linear Momentum

In addition to eqns. (3.29) - (3.30), the following decompositions can be made:

ρvv = ρv vv vv χ− + ρl vl vl χ+ ρs vs vs δ(x) (3.43)

T = Tv χ− + Tl χ+ + Ts δ(x) (3.44)

where Tl and Tv are the stress tensors in the bulk liquid and vapor phases and Ts

is the surface stress tensor. Substituting these decompositions into the momentum

balance
∂
(ρv) = −∇ · (ρvv + T) + ρg (3.45)
∂t

yields

∂
x ∈ Ω− : (ρv vv ) + ∇ · (ρv vv vv ) = −∇ · Tv + ρv g (3.46)
∂t
∂
x ∈ Ω+ : (ρl vl ) + ∇ · (ρl vl vl ) = −∇ · Tl + ρl g (3.47)
∂t

in the bulk phases and

∂
(ρs vs ) + ∇ · (ρs vs vs ) − ρs g + ∇ · Ts + (Tl − Tv ) · n̂ + j(vl − vv ) = 0 (3.48)
∂t
Ts · n̂ = 0 (3.49)

102
3.3 CONSERVATION LAWS THEORY

at the interface. Eqn. (3.49) implies that the surface stress tensor, Ts , only varies

tangential to the interface and not normal to it. In eqn. (3.48), neglecting terms

associated with the surface mass excess yields

−∇ · Ts = (Tl − Tv ) · n̂ + j(vl − vv ). (3.50)

The following constitutive models are used for the stress tensors:

Ti = pi I + Si i = l, v (3.51)

Ts = −σ Is (3.52)

where p is the pressure and S is the viscous part of the stress tensor,

    
T 2
S = −µ ∇v + ∇v + µ − κ ∇ · v I. (3.53)
3

The rheological properties of the fluid are assumed to be Newtonian.189 For the surface

stress tensor, σ is the surface tension and the viscous contribution to the surface stress

has been neglected. Eqn. (3.50) can be transformed into a more recognizable form

103
3.3 CONSERVATION LAWS THEORY

by noting that

∇ · Ts = −∇ · (σIs )

= −σ∇ · Is − Is · ∇σ

= −σ∇ · (I − n̂n̂) − ∇s σ (3.54)

= σ(∇ · n̂)n̂ − ∇s σ

= 2Hσn̂ − ∇s σ

Hence,

−2Hσn̂ + ∇s σ = (pl − pv )n̂ + (Sl · n̂ − Sv · n̂) + j(vl − vv ). (3.55)

The above vector equation can be decomposed into normal and tangential compo-

nents:

−2Hσ = (pl − pv ) + [(Sl · n̂) · n̂ − (Sv · n̂) · n̂] + j(vl − vv ) · n̂ (3.56)

∇s σ = (Sl · n̂) · t̂ − (Sv · n̂)t̂ (3.57)

where no slip has been assumed at the interface,

vl · t̂ = vv · t̂ (3.58)

104
3.3 CONSERVATION LAWS THEORY

The last term in eqn. (3.56) can be rewritten as follows

 
j(vl − vv ) · n̂ = j (vl − vs ) · n̂ − (vv − vs ) · n̂
 
ρl (vl − vs ) · n̂ ρv (vv − vs ) · n̂
=j − (3.59)
ρl ρv
 
1 1
= j2 −
ρl ρv

For j ≪ 1, terms of O(j 2 ) can be neglected, which results in

−2Hσ = (pl − pv ) + [(Sl · n̂) · n̂ − (Sv · n̂) · n̂] (3.60)

∇s σ = (Sl · n̂) · t̂ − (Sv · n̂)t̂ (3.61)

3.3.3 Conservation of Energy

The energy and energy flux can be decomposed similarly

ρe = ρv ev χ− + ρl el χ+ + ρs es δ(x) (3.62)

Je = Je,v χ− + Je,l χ+ + Je,s δ(x) (3.63)

where, after neglecting potential energy contributions, the total energy is given by:

v2
e=u+ (3.64)
2

105
3.3 CONSERVATION LAWS THEORY

and the energy flux can be expressed as:

Je = ρev + T · v + Jq . (3.65)

The first term in the energy flux corresponds to the flux of energy advected with

the fluid, the second term corresponds to the mechanical work, and the last term

corresponds to the heat flux,

Jq = −k∇T. (3.66)

Substituting the above decompositions into the total energy balance,

∂
(ρe) = −∇ · Je , (3.67)
∂t

yields

∂
x ∈ Ω− : (ρv ev ) + ∇ · (ρv ev vv ) = −∇ · (Tv · vv ) − ∇ · Jq,v (3.68)
∂t
∂
x ∈ Ω+ : (ρl el ) + ∇ · (ρl el vl ) = −∇ · (Tl · vl ) − ∇ · Jq,l (3.69)
∂t

in the bulk phases. At the interface,

∂
(ρs es ) + ∇ · Je,s + j(el − ev ) + (Tl · vl − Tv · vv ) · n̂ + (Jq,l − Jq,v ) · n̂ = 0 (3.70)
∂t
Je,s · n̂ = 0 (3.71)

106
3.3 CONSERVATION LAWS THEORY

which implies the surface energy flux does not vary normal to the surface. Neglecting

terms involving the surface mass excess and surface excess heat flux yields,

−∇ · (Ts · vs ) = j(el − ev ) + (Tl · vl − Tv · vv ) · n̂ + (Jq,l − Jq,v ) · n̂ (3.72)

To simply this expression, one can first take the dot product of eqn. (3.50) with vs ,

−vs · (∇ · Ts ) = (Tl · n̂ − Tv · n̂) · vs + j(vl − vv ) · vs (3.73)

and then subtract the result from eqn. (3.72), resulting in:

  
1
vs · (∇ · Ts ) − ∇ · (Ts · vs ) = j (ul − uv ) + vl · vl − vv · vv − 2vl vs + 2vv vs +
2

(Tl · vl · n̂ − Tl · n̂ · vs ) + (Tv · vv · n̂ − Tv · n̂ · vs ) + (Jq,l − Jq,v ) · n̂

(3.74)

The following terms can be rewritten:

vs · (∇ · Ts ) − ∇ · (Ts · vs ) = Ts : ∇vs (3.75)

107
3.3 CONSERVATION LAWS THEORY

1
(vl · vl − vv · vv − 2vl · vs + 2vv · vs )
2
   
1
= vl · vl − 2vl · vs + vs · vs − vv · vv − 2vv · vs + vs · vs )
2
 
1
= (vl − vs ) · (vl − vs ) − (vv − vs ) · (vv − vs ) (3.76)
2
 
1 jn̂ jn̂ jn̂ jn̂
= · − ·
2 ρl ρl ρv ρv
 
j2 1 1
= − 2
2 ρ2l ρv

By making use of the symmetry property of the stress tensor

(Tl · vl ) · n̂ − (Tl · n̂) · vs = (Tl · n̂) · (vl − vs )

(3.77)
jn̂
= (Tl · n̂) ·
ρl

and

(Tv · vv ) · n̂ − (Tv · n̂) · vs = (Tv · n̂) · (vv − vs )

(3.78)
jn̂
= (Tv · n̂) · .
ρv

The energy balance at the interface becomes:

   
j3 1 1 (Tl · n̂) · n̂ (Tv · n̂) · n̂
Ts : ∇vs = j(ul − uv ) + − +j − + (Jq,l − Jq,v ) · n̂
2 ρ2l ρ2v ρl ρv
(3.79)

108
3.4 EQUATION SUMMARY THEORY

Applying eqn. (3.51) results in

   
j3 1 1 (Sl · n̂) · n̂ (Sv · n̂) · n̂
Ts : ∇vs = j(hl − hv ) + − +j − + (Jq,l − Jq,v ) · n̂.
2 ρ2l ρ2v ρl ρv
(3.80)

For j ≪ 1, terms of O(j 3 ) can be neglected, which leads to

 
(Sl · n̂) · n̂ (Sv · n̂) · n̂
Ts : ∇vs = j(hl − hv ) + j − + (Jq,l − Jq,v ) · n̂. (3.81)
ρl ρv

3.4 Equation Summary

From the derivations presented in section 3.3, one finds in the bulk phases

∂ρ
+ ∇ · (ρv) = 0 (3.82)
∂t
∂
(ρvv) + ∇ · (ρvv) = −∇p − ∇ · S + ρg (3.83)
∂t
   
∂ p
(ρe) + ∇ · ρ e + v = −∇ · (S · v) − ∇ · Jq (3.84)
∂t ρ

and at the interface

j ≡ ρl (vl − vs ) · n̂ = ρv (vv − vs ) · n̂ (3.85)

−2Hσ = (pl − pv ) + [(Sl · n̂) · n̂ − (Sv · n̂) · n̂] (3.86)

∇s σ = (Sl · n̂) · t̂ − (Sv · n̂) · t̂ (3.87)

 
(Sl · n̂) · n̂ (Sv · n̂) · n̂
Ts : ∇vs = j(hl − hv ) + j − + (Jq,l − Jq,v ) · n̂. (3.88)
ρl ρv

109
3.5 EVAPORATION CONDITION THEORY

In addition to the above balance equations, equations of state are also required,

p = p(ρ, T ) (3.89)

u = u(ρ, T ) (3.90)

σ = σ(T ) (3.91)

3.5 Evaporation Condition

In addition to the transport equations and constitutive models for stress and heat

flux, an evaporation condition is required for closure. In what follows, the equilibrium

conditions for the liquid-vapor-interfacial system are discussed and later generalized

for non-equilibrium processes.

Consider a closed tank where the bulk phases and surface can exchange mass and

energy among themselves. Each bulk phase can be treated as an open system where

dUl = Tl dSl − pl dVl + µl dml (3.92)

dUv = Tv dSv − pv dVv + µv dmv (3.93)

dUs = Ts dSs + σ dA + µs dms . (3.94)

Here, it has been assumed that only a single species exists in the tank and that

potential energy contributions are negligible to the first law energy balance. The

110
3.5 EVAPORATION CONDITION THEORY

chemical potential, µ, is defined on a mass basis: i.e.

 
∂ul
µl = . (3.95)
∂ml s,v

Summing eqns. (3.92)-(3.94) yields

dUl + dUv + dUs =Tl dSl + Tv dSv + Ts dSs

− pl dVl − pv dVv + σ dA (3.96)

+ µl dml + µv dmv + µs dms .

Since the tank is closed and rigid

dm = dml + dmv + dms = 0 (3.97)

dV = dVl + dVv . (3.98)

At equilibrium,

dU = dUl + dUv + dUs = 0 (3.99)

dS = dSl + dSv + dSs = 0. (3.100)

111
3.5 EVAPORATION CONDITION THEORY

Hence,

0 = (Tl − Ts ) dSl + (Tv − Ts ) dSv − (pv − pl ) dVv + σ dA + (µl − µs ) dml + (µv − µs ) dmv .

(3.101)

Since these variations are independent, at equilibrium,

Tl = Tv = Ts (3.102)

µl = µv = µs (3.103)
dA
pv − pl = σ (3.104)
dVv

For a spherical ullage surrounded by a pool of liquid:

dA d(4πr 2 ) 2
= 3
= = 2H (3.105)
dVv d(4/3πr ) r

where H is the mean curvature defined according to eqn. (3.6). Hence, the condition

for mechanical equilibrium is

pv − pl = 2σH (3.106)

which is consistent with the jump momentum balance (eqn. (3.86)) when no velocity

gradients are present and the interfacial mass flux is zero.

For a pure system, the chemical potential is the specific Gibbs energy:

µ=g (3.107)

112
3.5 EVAPORATION CONDITION THEORY

and

dg = −sdT + νdp. (3.108)

At equilibrium,

gl = gv (3.109)

dgl = dgv . (3.110)

Thus,

−sl dT + νl dpl = −sv dT + νv dpv (3.111)

where isothermality (eqn. (3.102)) has already been invoked. For a flat interface,

pl = pv = pf lat . (3.112)

Taking differentials

dpl = dpv = dpf lat (3.113)

which allows one to rewrite eqn. (3.111) as

sv − sl
dpf lat = dT. (3.114)
νv − νl

For a curved interface

pv − pl = 2Hσ (3.115)

113
3.5 EVAPORATION CONDITION THEORY

or

dpv − dpl = d(2Hσ) (3.116)

Expanding the differential on the right side of eqn. (3.116) yields

d(2Hσ) = 2σdH + 2Hdσ

(3.117)
∂σ
= 2σdH + 2H dH
∂H

If one neglects the variation of surface tension with respect to curvature, and solves

eqn. (3.116) for dpl , eqn. (3.111) can be rewritten

−sl dT + νl dpv − νl 2σdH = −sv dT + νv dpv (3.118)

or

sv − sl −2νl σ
dpv − dT = dH (3.119)
νv − νl νv − νl
−2νl σ
d(pv − pf lat ) = dH. (3.120)
νv − νl

Integrating from a flat interface (pv − pf lat = 0, H = 0) to an arbitrarily curved one

yields
2νl σH
pv = pf lat − . (3.121)
νv − νl

where it is generally understood that pf lat is the saturation pressure and the curva-

ture term is a correction to that pressure. Eqn. (3.121) is more general than other

114
3.5 EVAPORATION CONDITION THEORY

expressions with a Poynting correction factor for the curvature because liquid incom-

pressibility and ideal gas behavior are not assumed in the above analysis. Regardless,

for most practical situations the correction due to curvature is negligible. For ex-

ample, a parahydrogen system at 20 K with a 1 mm spherical ullage results in a

correction of only 0.07 Pa compared to the saturation pressure of 93414 Pa. Hence

in the subsequent analysis, the curvature correction will be neglected and it will be

assumed that at equilibrium

pv = psat (T ). (3.122)

The analysis presented thus far in section 3.5 assumes equilibrium for the liquid-

vapor-interfacial system. Mass transfer is inherently a non-equilibrium process and to

describe this non-equilibrium behavior an additional constitutive equation involving

the mass flux is needed. The form of this constitutive model has been the subject of

much scholarly debate for more than a century. In what follows, a brief overview of

the different constitutive models is presented.

For non equilibrium processes, continuity of temperature across a phase boundary

is generally not assured. Huang and Joseph190 describe three possible temperature

boundary conditions at the interface. The temperature can be continuous with the

value given by the saturation temperature corresponding either the liquid or vapor

temperature. Alternatively, the temperature can be discontinuous with the surface

value given by the saturation temperature corresponding to the local pressure. In

the stability analysis of Huang and Joseph,190 little difference was found among these

115
3.5 EVAPORATION CONDITION THEORY

different interfacial boundary conditions.

Experimentally, conclusions regarding interfacial discontinuities have been mixed.

Shankar and Deshpande191 used fixed thermocouple locations in their experimental

study of liquid/vapor phase change processes. Their data suggested jumps do ex-

ist though it’s unclear whether their experimental setup was able to resolve steep

thermal gradients near the interface. Ramamurthi et al.192 performed experiments

with LN2 and observed no discrete change or discontinuity in temperatures at the

interface. Hisatake et al.193 studied the evaporation of water in an open vessel. A

thermocouple 0.127 mm in diameter measured a temperature profile across the liquid

vapor interface. Steep gradients on either side of the interface were observed and the

authors drew a continuous line through the data to represent the temperature profile.

The experiments by Ward and colleagues have consistently recorded temperature

discontinuities across evaporating interfaces. Fang and Ward194 measured the tem-

perature during steady state evaporation of liquid water in a hemispherical funnel.

The thermocouple was only 25.4 µm in diameter and the authors claimed to have

measured vapor-side temperatures within one mean free path of the interface. They

found the vapor-side temperature to be larger than the liquid-side temperature by as

much as 7.8 K. Again using a hemispherical funnel, Ward and Stanga195 noted the

jump is greater for evaporation than for condensation and observed a uniform tem-

perature layer ( 0.5 mm thick) below the interface which is indicative of some mixing

process near the interface. Duan et al.196 carried out similar experiments using an

elongated funnel and measured a temperature variation tangential to the interface

116
3.5 EVAPORATION CONDITION THEORY

which they suggested can drive Marangoni convection. Duan and Ward197 noted that

the conductive flux on either side of the interface is too small to supply the necessary

energy for phase change and suggest that a surface thermal capacity contribution or

thermocapillary convection contribution be included in the interfacial energy balance.

The preceding experiments by Ward and colleagues were carried out at relatively low

pressures, on the order of a few hundred Pascals. McGaughey and Ward198 studied

the temperature discontinuity at the surface of an evaporating droplet when the op-

erating pressure was a few thousand Pascals. Their data suggests that as pressure

increases, the temperature discontinuity across the interface decreases.

Ward and Fang199 used statistical rate theory (SRT) to derive a non-linear ex-

pression for the evaporative flux at an interface. SRT applies quantum mechanics

and looks at the probability of a molecule transitioning from one bulk phase to the

other. Fang and Ward200 tested their theory by substituting their experimentally

determined mass flux and temperatures into their expression and solved for the cor-

responding vapor pressure. The agreement with the experimentally measured vapor

pressure was acceptable.

Though agreement of the SRT expression with experimental data from Ward’s

group has generally been good, Bond201 questions the completeness of the statistical

rate theory equation since it doesn’t provide a corresponding energy flux relationship.

Bond also notes that the SRT flux expression is essentially a non-linear version of one

derived using irreversible thermodynamics (IRT). IRT applies the 2nd law of thermo-

dynamics to the liquid-vapor-interfacial system to arrive at an entropy source at the

117
3.5 EVAPORATION CONDITION THEORY

interface. The form of the entropy source can be posited as in Juric and Tryggva-

son202 or rigorously derived using the methods of non equilibrium thermodynamics.

Once the entropy source is known, it can be decomposed into force-flux pairs where

the fluxes are assumed to be linear functions of the forces or driving potentials.

The fluxes are generally coupled and Kjelstrup and Bedeaux203 note that neglect-

ing the coupling between heat and mass transfer at the interface is not justified.

Bedeaux et al.188 derived the force-flux pairs for an interfacial system assuming no

singular mass density. Kovac204 extended this work by including the presence of

multicomponent species. IRT is useful for identifying relevant force-flux pairs but

not capable of deriving closed expressions. IRT relates the forces and fluxes through

coupling coefficients which must be determined by some other means.

Bedeaux and Kjelstrup205 computed the coupling coefficients using Ward’s ex-

perimental data194 and not surprisingly obtained agreement with the experimental

evaporation rate. Bornhorst and Hatsopoulos206 used kinetic theory to evaluate the

coupling coefficients for a discontinuous liquid-vapor-interfacial system. Bedeaux et

al.207 found that in order for IRT to give consistent results with kinetic theory, the

interfacial temperature should be taken as the temperature of the adjacent liquid.

Kinetic theory has often been used to derive expressions for the interfacial mass

flux. In kinetic theory, the statistical behavior of a system of monatomic molecules

is described by a distribution function, f , such that f dc represents the fraction of

molecules with velocity in the range [c, c + dc]. Neglecting the Knudsen layer in Fig.

118
3.5 EVAPORATION CONDITION THEORY

Vapor
fv

fl Knudsen Layer

Liquid

Figure 3.3: Knudsen layer above a liquid-vapor interface.

3.3, the net condensation mass flux at the interface can be expressed as

ZZZ ZZZ
j= mcz fv dc − mcz fl dc (3.123)
cz <0 cz >0

where m is the molecular mass and cz is the molecular velocity in the direction of a

unit vector pointing into the vapor. If the distribution function of the evaporating

molecules and condensing molecules are known, eqn. (3.123) can then be used to

evaluate the interfacial mass flux. Generally, one assumes that molecules leaving the

liquid surface can be described with a Maxwellian distribution. Much work has been

done in trying to describe the distribution function in the vapor. The often used Hertz

Knudsen expression for interfacial mass flux assumes the vapor distribution function

is Maxwellian as well. Schrage208 noted that during condensation, there must be a net

molecular motion towards the interface and modified the Hertz Knudsen relationship

119
3.5 EVAPORATION CONDITION THEORY

by assuming a translating Maxwellian distribution for the bulk vapor molecules. In his

analysis, Schrage also used the concept of an evaporation or condensation coefficient.

The condensation coefficient is typically defined as the fraction of vapor molecules

incident to the interface that get captured by the liquid. The remainder get reflected

back into the vapor phase. The evaporation coefficient is similarly defined though

more complicated conceptually due to lack of a surface to reflect liquid molecules

back into the bulk. At equilibrium the condensation and evaporation fluxes are equal

which imply equality between the condensation and evaporation coefficients.

In the literature, measurements of the evaporation/condensation coefficient have

been scattered. Maa209 experimentally determined the evaporation coefficient to be

unity for several liquids. Eames et al.210 reviewed data on the evaporation coefficient

of water and noted a wide scatter between 0.01 - 1. Davis211 also noted the large vari-

ation in reported coefficients. Paul212 compiled a list of evaporation coefficients for a

variety of fluids and found large differences in reported values. Marek and Straub213

noted that the condensation coefficient for water is generally higher than the evapo-

ration coefficient and equality of the coefficients should not be assumed. Tanasawa214

reviewed the relevant kinetic theory literature with regard to interfacial mass flux

expressions and commented that no theory has proved or disproved the assumption

that the condensation and evaporation coefficients are equal during a net conden-

sation or evaporation process. To complicate matters, the evaporation/condensation

coefficients have often been used interchangeably with the accommodation coefficient,

which Bond201 defines as being 0 or 1 depending on whether reflecting vapor molecules

120
3.5 EVAPORATION CONDITION THEORY

reflect diffusely or specularly.

Notwithstanding the uncertainty in the values of the condensation or evaporation

coefficient, equality is often assumed which results in the Schrage mass flux expression

given by
 1/2  
2σ Mw pv pl
j= √ −√ (3.124)
2 − σ 2πR Tv Tl

where, in this context, σ is the evaporation or condensation coefficient. Barrett and

Clement215 point out that Schrage’s simple mass flux relationship violates strict en-

ergy and momentum conservation at the interface though this lack of conservation

has been disputed by Bond.201 Schrage208 later modified his original approach by

assuming a perturbed distribution function for the condensing vapor molecules. The

perturbed distribution function was used to describe vapor molecules in the Knud-

sen layer where collisions with evaporating molecules could change the bulk vapor

distribution function. Panzarella216 applied Schrage’s modified mass flux expression

to film boiling and noted that Schrage’s modified equation captures asymmetries be-

tween condensation and evaporative processes. Ytrehus217 also showed that basic

asymmetries exist between evaporation and condensation.

Instead of assuming a form of the vapor distribution function, a number of investi-

gators have solved for the distribution function by taking moments of the Boltzmann

equation,218, 219 by solving a linearized version of the Boltzmann equation,220–222 or

by numerically solving the Boltzmann equation.223 Typically these analyses suggest

a temperature jump across the Knudsen layer exists even in the continuum limit as

121
3.5 EVAPORATION CONDITION THEORY

Kn → 0.224

The kinetic theory approach to interfacial mass transfer has given way in recent

years to molecular dynamics (MD) simulations. In a MD analysis, a molecular poten-

tial is assumed and Newton’s laws of motion are solved for an ensemble of molecules

subject to certain constraints. Because of the computational overhead associated with

a MD simulation, these analyses are often restricted to small computational domains

for brief periods of time. Often MD simulations have been used to either aid in the

theoretical understanding of interfacial mass transfer or to examine the assumptions

in the approaches described above. Meland et al.225 performed a MD simulation

on a system undergoing phase change and found that the evaporation and conden-

sation coefficients are not equal outside of equilibrium. Ishiyama et al.226 observed

that at low temperatures evaporating molecules have a half Maxwellian distribution

thus supporting the boundary condition in the kinetic theory analyses. Matsumoto

et al.227 performed an MD simulation to show that few vapor molecules get reflected

and that condensing molecules can drive liquid molecules into the bulk vapor. Xu et

al.228 used MD to show that linearity of the force-flux pairs in IRT is valid even at

moderate temperature gradients. Rosjorde et al.229 using MD simulations to provide

closure for IRT coupling coefficients.

While MD simulations show promise at revealing the fine scale behavior near an

interface, these simulations rely on a assumed potential. As computational resources

improve, perhaps the approach suggested by Sandler230 will yield further insight.

Sandler suggests computing the potential using quantum mechanics and then applying
122
3.5 EVAPORATION CONDITION THEORY

this computed potential to a MD simulation.

The above discussion provides a brief overview of various approaches to determine

a constitutive model for the interfacial mass flux. In this dissertation, Schrage’s model

will be used, equality of the condensation and evaporation coefficients will be assumed,

these coefficients will be assumed constant, and continuity of temperature across the

interface will be enforced, i.e.

Tl = Tv = Ts . (3.125)

Finally, it will be assumed that the mass transfer processes are close enough to equi-

librium that

pl ≈ psat . (3.126)

These assumptions lead to the following constitutive relationship

 1/2
2σ Mw
j= (pv − psat ). (3.127)
2 − σ 2πRT

123
 Chapter 4

Model Development

In order to solve the system of equations derived in Chapter 3, several models are

developed with varying degrees of complexity. First, a homogeneous thermodynamic

model will be developed whereby the state of the liquid-vapor system is confined to

the saturated PVT surface during a self-pressurization or pressure control process.

Next, a multizone model is formulated which relaxes the assumption of homogeneity

within the tank. The results of these models will be compared against each other and

against experimental data in subsequent chapters.

Consider a spherical tank partially full of liquid subjected to external heat loads,

Q̇wv and Q̇wl , as shown in Fig. 4.1. The tank is an open system. During a typical

pressure control process, fluid is withdrawn from the liquid phase and pumped through

a heat exchanger to subcool it before it reenters the tank as a subcooled liquid jet.

The tank, with corresponding plumbing, is a closed system so the rate of mass leaving

the tank equals the rate of mass entering the tank through the jet nozzle.

Integrating the continuity equation in each of the bulk phases around their re-

124
 MODEL DEVELOPMENT

Q wv
Vapor

Q iv M

Q il

Q wl Liquid

mout min

Figure 4.1: Sketch of a liquid-vapor system with heat and mass entering or
leaving the system.

spective control volumes yields

Z  Z
d
ρv dV + ρv (vv − vs ) · n̂ dS = 0 (4.1)
dt Vv I
Z  Z Z Z
d
ρl dV − ρl (vl − vs ) · n̂ dS + ρl vl · n dS − ρl vl · n dS = 0. (4.2)
dt Vl I out in

where the surface integral over I is performed over the interfacial area, the interfacial

normal, n̂, points into the liquid, and the normals to the outlet and inlet are outwardly

125
 MODEL DEVELOPMENT

and inwardly pointing respectively. With the following definitions:

Z
1
ρv = ρv dV (4.3)
Vv Vv
Z
1
ρl = ρl dV (4.4)
Vl Vl
Z Z
j ≡ ρl (vl − vs ) · n̂ dS = ρv (vv − vs ) · n̂ dS (4.5)
I I
Z Z
ṁ = ρl vl · n dS = ρl vl · n dS, (4.6)
out in

one has

Z
dρv
= − j dS (4.7)
dt I
Z
dρl
= j dS. (4.8)
dt I

Summing eqns. (4.7) and (4.8) yields,

d
(ρ Vv + ρl Vl ) = 0. (4.9)
dt v

which implies that even though mass exchange may occur between the bulk phases

and mass may be entering or leaving the tank, the total mass inside the tank is

constant. Integrating eqn. (4.9) in time leads to:

ρv Vv + ρl Vl = ρv,o Vv,o + ρl,o Vl,o (4.10)

126
 MODEL DEVELOPMENT

where the subscript, o, denotes some initial time. Since the total tank volume,Vt , is

comprised of liquid and vapor volumes,

Vt = Vl + Vv . (4.11)

Using eqn. (4.10) and assuming the tank is rigid, one can derive an explicit equation

for the vapor volume,

ρl,o − ρl ρv,o − ρl,o

Vv (ρv , ρl ) = Vt + Vv,o . (4.12)
ρv − ρl ρv − ρl

It will be convenient to use the enthalpy form of the energy balance for the bulk

phases. Taking the dot product of the momentum equation with the bulk phase

velocity, subtracting the result from the total energy balance (eqn. (3.84)), then

neglecting viscous dissipation and potential energy contributions yields

∂
(ρu) + ∇ · (ρuv) = −∇ · Jq − p∇ · v (4.13)
∂t

for the balance of internal energy. Substituting

ρu = ρh − p (4.14)

leads to
∂ dp
(ρh) + ∇ · (ρhv) = −∇ · Jq + . (4.15)
∂t dt
127
 MODEL DEVELOPMENT

Integrating around each bulk phase volume as before yields

Z  Z
d
ρv hv dV + ρv hv (vv − vs ) · n̂ dS =
dt Vv I
Z Z Z (4.16)
dpv
− Jq,v · n dS − Jq,v · n̂ dS + dV
w I Vv dt

Z  Z Z Z
d
ρl hl dV − ρl hl (vl − vs ) · n̂ dS + ρl hl vl · n dS − ρl hl vl · n dS =
dt Vl I out in
Z Z Z Z Z
dpl
− Jq,l · n dS + Jq,l · n̂ dS − Jq,l · n dS + Jq,l · n dS + dV
w I out in Vl dt

(4.17)

where the normals to the wall, w, are outwardly pointing. One can take a mass-

average of the bulk phase enthalpies,

Z
1
hi = ρi hi dV i = l, v. (4.18)
ρ i Vi Vi

One can also define Q̇wl and Q̇wv as the amount of heat entering the bulk liquid and

vapor phases through the wall,

Z
Q̇wl = − Jq,l · n dS (4.19)
w
Z
Q̇wv = − Jq,v · n dS, (4.20)
w

and the amount of heat leaving the vapor phase and entering the liquid phase through

128
 MODEL DEVELOPMENT

the interface as:

Z
Q̇il = Jq,l · n̂ dS (4.21)
I
Z
Q̇iv = Jq,v · n̂ dS. (4.22)
I

The diffusive heat flux across the tank inlet and outlet is assumed small relative to

the convective contribution. Hence,

Z
Jq,l · n dS ≈ 0 (4.23)
out
Z
Jq,l · n dS ≈ 0. (4.24)
in

If the liquid enthalpy over the cross-section of the tank inlet and outlet is assumed

constant, then

Z
ρl hl (vl · n) dS = ṁhin
l (4.25)
in
Z
ρl hl (vl · n) dS = ṁhout
l . (4.26)
out

Finally, for this flat interface configuration:

pv = pl = p. (4.27)

The tank pressure can be decomposed into a mean component and a fluctuation about

129
4.1 THERMODYNAMIC MODEL MODEL DEVELOPMENT

the mean:

p = p(t) + p′ (x, t). (4.28)

For p′ ≪ p,
Z
dp dp
dV ≈ V. (4.29)
V dt dt

Hence,

Z
d dp
(ρ hv Vv ) + j hv dS = Q̇wv − Q̇iv + Vv (4.30)
dt v I dt
Z  
d dp out in
(ρ hl Vl ) − j hl dS = Q̇wl + Q̇il + Vl − ṁ hl − hl . (4.31)
dt l I dt

4.1 Thermodynamic Model

In this model, the temperature in the tank is assumed to be uniform and liq-

uid/vapor thermodynamic quantities are assumed spatially invariant and at satura-

130
4.1 THERMODYNAMIC MODEL MODEL DEVELOPMENT

tion within each bulk phase. This is equivalent to

T = Tsat (4.32)

p = psat (Tsat ) (4.33)

ρv = ρsat
v (Tsat ) (4.34)

ρl = ρsat
l (Tsat ) (4.35)

hv = hsat
v (Tsat ) (4.36)

hl = hsat
l (Tsat ). (4.37)

Defining the evaporation rate across the interface as

Z
M =− j dS, (4.38)
I

one has

d sat
(ρ Vv ) = M (4.39)
dt v
d sat
(ρ Vl ) = −M. (4.40)
dt l

Similar to before, the vapor volume can be expressed as

ρsat sat
l,o − ρl ρsat sat
v,o − ρl,o
Vv (ρsat sat
v , ρl ) = Vt sat
+ V v,o sat
. (4.41)
ρsat
v − ρl ρsat
v − ρl

131
4.1 THERMODYNAMIC MODEL MODEL DEVELOPMENT

Adding eqns. (4.30) and (4.31) and substituting in the saturation quantities described

above yields:

  Z
d sat sat sat sat
ρ h Vv + ρl hl Vl + j(hv − hl ) dS =
dt v v I
  (4.42)
dpsat out in
Q̇w + (Q̇il − Q̇iv ) + Vt − ṁ hl − hl
dt

where the net heat entering the tank

Q̇w = Q̇wl + Q̇wv . (4.43)

Neglecting the viscous work terms at the interface and the energy contribution due

to expansion/contraction of the interface, the integral form of eqn. (3.88) reduces to

Z
j(hv − hl ) dS = Q̇il − Q̇iv . (4.44)
I

Hence,

   
d sat sat sat sat dpsat out in
ρ h Vv + ρl hl Vl = Q̇w + Vt − ṁ hl − hl . (4.45)
dt v v dt

Carrying out the differentiation and making use of eqns. (4.39) - (4.40) leads to

   
dhsat
v dhsat dpsat
ρsat
v V v + ρsat
l V l
l sat sat
+ M hv − hl − out in
Vt = Q̇w − ṁ hl − hl . (4.46)
dt dt dt

132
4.2 MULTIZONE MODEL MODEL DEVELOPMENT

The difference in liquid and vapor enthalpies at the saturation temperature is the

latent heat, L. Expanding the time derivatives using the chain rule, one has

   
dTsat sat dhsat
v sat dhl
sat
d(ρsat
v Vv ) dpsat out in
ρv Vv + ρl Vl + L− Vt = Q̇w − ṁ hl − hl .
dt dTsat dTsat dTsat dTsat
(4.47)

Equation (4.47) represents an evolution equation for the saturation temperature Tsat .

All derivatives with respect to temperature can be evaluated explicitly using eqns.

(4.33)-(4.37) and (4.41). Once the saturation temperature is known, the tank pressure

can be computed from eqn. (4.33).

4.2 Multizone Model

In a multizone formulation, uniformity within each bulk phase is still assumed,

but the assumption that the phases are at the same temperature, Tsat , is relaxed.

The tank is partitioned into three zones: a liquid zone, a vapor zone, and a thin

massless interfacial zone at temperatures Tl , Tv , and Ti respectively. The liquid

phase is assumed incompressible and the vapor is treated as a compressible gas whose

pressure is given by:

p = p(ρv , Tv ) (4.48)

133
4.2 MULTIZONE MODEL MODEL DEVELOPMENT

Assuming uniformity within each bulk phase is equivalent to:

ρv = ρv (4.49)

ρl = ρl (4.50)

hv = hv (4.51)

hl = hl . (4.52)

The integrated continuity equation in the liquid is given by:

d
(ρl Vl ) = −M
dt
dVl (4.53)
ρl = −M
dt
dVv
−ρl = −M
dt

which implies that

dVv M
= . (4.54)
dt ρl

Liquid incompressibility and the volume constraint were used to arrive at eqn. (4.54).

The continuity equation in the vapor is

d
(ρv Vv ) = M (4.55)
dt
dVv dρv
ρv + Vv =M (4.56)
dt dt
M dρv
ρv + Vv =M (4.57)
ρl dt

134
4.2 MULTIZONE MODEL MODEL DEVELOPMENT

which leads to
 
dρv M ρv
= 1− . (4.58)
dt Vv ρl

Uniformity within each bulk phases simplifies the bulk phase energy balances to

 
d i out in dpv
(ρl hl Vl ) + M hl = −ṁ hl − hl + Q̇wl + Q̇il + Vl (4.59)
dt dt
d dpv
(ρv hv Vv ) − M hiv = Q̇wv − Q̇iv + Vv . (4.60)
dt dt

The enthalpy of the bulk phases is assumed to behave as

hv = hv,o + cp,v (Tv − Tv,o ) (4.61)

hl = hl,o + cp,l (Tl − Tl,o ). (4.62)

Substituting these expressions into the energy balances, carrying out the differentia-

tion, and rearranging terms leads to

dTl dpv
ρl cp,l Vl = Mcp,l (Tl − Tli ) − ṁcp,l (Tl − Tin ) + Q̇wl + Q̇il + Vl (4.63)
dt dt
dTv dp v
ρv cp,v Vv = Mcp,v (Tvi − Tv ) + Q̇wv − Q̇iv + Vv (4.64)
dt dt

where once again, liquid incompressibility has been invoked. The equation of state

can be used to explicitly evaluate the pressure derivative. That is,

dpv ∂pv dρv ∂pv dTv

= + . (4.65)
dt ∂ρ dt ∂T dt
135
4.2 MULTIZONE MODEL MODEL DEVELOPMENT

Substituting this expression into the energy balances yields

 
dTl i ∂pv dρv ∂pv dTv
ρl cp,l Vl = Mcp,l (Tl − Tl ) − ṁcp,l (Tl − Tin ) + Q̇wl + Q̇il + Vl +
dt ∂ρ dt ∂T dt

(4.66)
 
∂pv dTv ∂pv dρv
ρv cp,v + Vv = Mcp,v (Tvi − Tv ) + Q̇wv − Q̇iv + Vv . (4.67)
∂T dt ∂ρ dt

In order to solve these energy equations, expressions for the interzonal energy and

mass exchange terms are required. Assuming spatial uniformity across the interface,

the energy jump condition reduces to

Q̇il − Q̇iv
M =− . (4.68)
hiv − hil

where the difference in enthalpies across the interface is assumed to be equal to the

latent heat, L. The interfacial temperature, Ti , can be found by equating the kinetic

mass flux and diffusive mass transfer rate at the interface and implicitly solving for

Ti ,
 1/2
2σ Mw Q̇il (Ti , Tl ) − Q̇iv (Ti , Tv )
Ai (psat (Ti ) − pv ) = . (4.69)
2−σ 2πRTi L

136
4.2 MULTIZONE MODEL MODEL DEVELOPMENT

The interfacial heat powers are evaluated using simple correlations which are problem

specific and will be described later:

Q̇il = hil Ai (Ti − Tl ) (4.70)

Q̇iv = hiv Ai (Tv − Ti ). (4.71)

Once Ti is known, the heat powers are updated and the mass transfer rate through

the interface can be computed using the left side of eqn. (4.69). Details regarding

the numerical implementation will be presented in the next chapter.

137
 Chapter 5

Implementation

In the previous chapter, the homogeneous and zonal models were derived. For

clarity, models were derived for a pure two-phase system. Extension to multicompo-

nent two-phase systems is described further in Appendix C. Model predictions will

be validated using the experimental data presented in Chapter 2. The implementa-

tion of these models and their respective solution algorithms will be described in this

chapter.

5.1 Homogeneous Model

The evolution equation for saturation temperature for a two-phase multicompo-

nent ullage is given by eqn. (C.29). In the ground experiment, the thermal capacity

of the walls is not insignificant and should be accounted for in the models. To ac-

count for the wall and other solid features in the experiment (i.e. heaters, nozzles,

138
5.1 HOMOGENEOUS MODEL IMPLEMENTATION

insulation), the following generic energy equation for a solid region is employed

d
(ρhV ) = Q̇in − Q̇out . (5.1)
dt

Since the mass of the solid regions is constant, assuming constant specific heat, yields

dT
ρV cp = Q̇in − Q̇out . (5.2)
dt

After invoking the homogeneous assumption (T = Tsat ), the evolution equation for

saturation temperature becomes


dTsat sat dhsat dhsat
v
X
ρl Vl l + ρsat v Vv + ρg Vv cpg + ρi Vi cpi +
dt dTsat dTsat i=solid
   (5.3)
d(ρsat
v Vv ) dp in sat
L − Vt = Q̇w + ṁ hl − hl .
dTsat dTsat

Since this represents a system level energy balance, all the inter-region heat transfer

mechanisms, Q̇in and Q̇out sum to zero. The only energy load on the system is Q̇w ,

the load that is externally applied.

Considering the right side of the energy balance, the applied load, Q̇w , is user

specified. The sub-cooled liquid enthalpy is assumed to be of the form

 
hin
l = hsat
l + cpl Tjet − Tsat . (5.4)

139
5.1 HOMOGENEOUS MODEL IMPLEMENTATION

Hence,
   
in sat
ṁ hl − hl = ṁcpl Tjet − Tsat . (5.5)

On the left side of the equation, the functional relationship between Tsat and the satu-

ration densities and enthalpies are known and provided in Appendix A for reference.
d(ρsat
v Vv )
With ρsat
v known and an expression for Vv (eqn. (4.41)), dTsat
can be explicitly

evaluated.

The pressure derivative can be expanded

dp dpsat dpg
= + . (5.6)
dTsat dTsat dTsat

The saturation pressure is given by the liquid-vapor saturation curve and so the above
dpg
derivative can be evaluated explicitly. To evaluate dTsat
, the non-condensable gas is

assumed to behave ideally at Tsat ,

mg Rg Tsat
pg = . (5.7)
Vv

Since the mass of the gas in the ullage is constant,

dpg mg Rg mg Rg Tsat dVv

= − (5.8)
dTsat Vv Vv2 dTsat
 
mg Rg mg Rg Tsat ∂Vv dρsat
v ∂Vv dρsat
l
= − + sat .
Vv Vv2 ∂ρsat
v dTsat ∂ρl dTsat

These derivatives can be evaluated explicitly using property functions (see Appendix

140
5.2 ZONAL MODEL IMPLEMENTATION

A) and eqn. (4.41).

Algorithmically, a forward Euler time stepping routine is employed:

 n
n+1 n dTsat
Tsat = Tsat + ∆t (5.9)
dt

where

 
Q̇w + ṁcpl Tjet − Tsat
dTsat
= dhsat P . (5.10)
dt ρsat sat dhsat d(ρsat
v Vv )
− Vt dTdpsat
l Vl dTsat + ρv Vv dTsat + ρg Vv cpg + ρi Vi cpi + L
l v
dTsat

When the sub-cooled jet mixer is inactive, the ṁcpl (Tjet − Tsat ) term is zero. When

the mixer is on, ṁ is computed from the user specified average jet velocity and Tjet

is computed using a heat exchanger model described in Appendix B. At every time

step, after evolving the saturation temperature, the ullage pressure can be updated

according to:
n+1
mg Rg Tsat
pn+1 = psat (Tsat
n+1
)+ n+1 n+1 . (5.11)
Vv (ρv (Tsat ), ρl (Tsat ))

5.2 Zonal Model

In the multizone model, homogeneity within each phase is still assumed but the

assumption that the entire liquid, vapor, wall system is at the same temperature,

Tsat , is relaxed.

In the solid phases of the tank system (i.e. the acrylic wall, the foam insulation,

and the copper heater rings), a 2D-axisymmetric finite element model is developed
141
5.2 ZONAL MODEL IMPLEMENTATION

to simulate the conduction heat transfer through these regions. The finite element

model is coupled to the evolution equations in the bulk liquid and vapor phases

through appropriate use of thermal boundary conditions, which will be described

below.

The acrylic wall, foam insulation, and copper heaters are discretized into linear

triangular elements using GMSH,231 a open-source grid generator. A typical com-

putational grid is shown in Fig. 5.1. Following the Galerkin approach, the residual

integral for an axisymmetric heat conduction problem is:

Z    
k ∂ ∂T ∂2T ∂T
− N r + k 2 + Q̇ − ρcp dV = 0. (5.12)
V r ∂r ∂r ∂z ∂t

Writing the vector of shape functions, N, in terms of area coordinates L1 and L2

yields:  
 L1 
 
 
N=
 L2 
 (5.13)
 
 
1 − L1 − L2

where the area coordinates are shown in Fig. 5.2. Carrying out the differentiation in

eqn. (5.12) and rearranging terms results in:

Z      Z  
k ∂ ∂T ∂ ∂T ∂N ∂T ∂N ∂T
− Nr +k N dV + k + dV −
V r ∂r ∂r ∂z ∂z V ∂r ∂r ∂z ∂z
Z Z (5.14)
∂T
NQ̇ dV + Nρcp dV = 0.
V V ∂t

In carrying out the above differentiation, the conductivity was assumed spatially
142
5.2 ZONAL MODEL IMPLEMENTATION

Ullage

Liquid

Figure 5.1: Finite element mesh of wall, heaters, and foam insulation.

143
5.2 ZONAL MODEL IMPLEMENTATION

constant within each element. The first integral in eqn. (5.14) is the interelement

contribution and can be transformed into a surface integral using Gauss’s theorem,

Z      Z
k ∂ ∂T ∂ ∂T ∂T
− Nr +k N dV = − Nk dA (5.15)
V r ∂r ∂r ∂z ∂z ∂n

where n is the outwardly pointing normal to the element boundary. To evaluate the

conduction integrals, one can transform the volume integral into a surface integral by

noting that
Z Z
dV = 2πr dA. (5.16)

Hence,

Z   Z  
∂N ∂T ∂N ∂T ∂N ∂T ∂N ∂T
k + dV = 2π + r dA (5.17)
V ∂r ∂r ∂z ∂z ∂r ∂r ∂z ∂z
Z1 1−L
Z 2 
∂N ∂T ∂N ∂T
= 2π + r|J| dL1 dL2 . (5.18)
∂r ∂r ∂z ∂z
0 0

The last equality in the above equation follows from transforming an integral over

the area, dA, to an integral over dL1 dL2 of the canonical linear triangular element.

Noting that

T = NT T (5.19)

144
5.2 ZONAL MODEL IMPLEMENTATION

L2 L1 j

L3

Figure 5.2: 2D linear triangular element with area coordinates L1 , L2 , and

L3 = 1 − L1 − L2 .

where T is the vector of nodal temperatures, allows one to write

Z    Z1 1−L
Z 2  
∂N ∂T ∂N ∂T ∂N ∂NT ∂N ∂NT
k + dV = 2π + r|J| dL1 dL2 T.
V ∂r ∂r ∂z ∂z ∂r ∂r ∂z ∂z
0 0
(5.20)

In the above integrand, the Jacobian of the transformation is defined as

 
3
P 3
P
 Zi ∂N i ∂Ni
Ri ∂L 
i=1 ∂L1 i=1
1
J= 3 3
 (5.21)
 P ∂Ni P ∂Ni

Zi ∂L2 Ri ∂L2
i=1 i=1

where (Zi, Ri ) are the nodal coordinates and dN/dL1 , dN/dL2 can be computed

explicitly from eqn. (5.13).

145
5.2 ZONAL MODEL IMPLEMENTATION

Gradients of the shape functions are computed from

   
∂N ∂N
 ∂z   ∂L1 
  = J−1  . (5.22)
   
∂N ∂N
∂r ∂L2

Finally,
3
X
r= Ri Ni . (5.23)
i=1

The integral is of the form

Z1 1−L
Z 2
g(L1 , L2 ) dL1 dL2 (5.24)
0 0

where
 
∂N ∂NT ∂N ∂NT
g(L1 , L2 ) = 2π + r|J|. (5.25)
∂r ∂r ∂z ∂z

Integration is carried out using Gaussian quadrature,

Z1 1−L
Z 2 N pts
X
g(L1 , L2 ) dL1 dL2 = g(L1,i, L2,i )Wi . (5.26)
0 0 i=1

The integration points and weights are listed in Table 5.1. The digits in Table 5.1 and

Table 5.2 are all significant and required for accurate computation of the numerical

integration.

146
5.2 ZONAL MODEL IMPLEMENTATION

L1,i L2,i 2 · Wi
0.333333333333333 0.333333333333333 0.225000000000000
0.797426985353087 0.101286507323456 0.125939180544827
0.101286507323456 0.101286507323456 0.125939180544827
0.101286507323456 0.797426985353087 0.125939180544827
0.470142064105115 0.470142064105115 0.132394152788506
0.470142064105115 0.059715871789770 0.132394152788506
0.059715871789770 0.470142064105115 0.132394152788506
Table 5.1: Integration points and weights for Gaussian quadrature over a tri-
angular element.232

The volumetric heat source integral is evaluated similarly:

Z Z1 1−L
Z 2
− NQ̇ dV = −2π NQ̇r|J|dL1 dL2 . (5.27)
0 0

Q̇ is zero in all solid regions except within the active heater.

The transient term is evaluated using the consistent approximation. That is,

∂T
= NT Ṫ. (5.28)
∂t

The integral becomes:

Z Z
∂T ∂T
Nρcp dV = 2πρCp N rdA (5.29)
∂t ∂t
1 1−L
Z Z 2 
T
= 2πρcp NN r|J|dL1 dL2 Ṫ. (5.30)
0 0

The density and specific heat were assumed constant over each element. Defining the

147
5.2 ZONAL MODEL IMPLEMENTATION

elemental capacitance matrix, CE , as

Z1 1−L
Z 2 
T
CE = 2πρCp NN r|J|dL1 dL2 , (5.31)
0 0

the elemental stiffness matrix, KE as

Z1 1−L
Z 2  
∂N ∂NT ∂N ∂NT
KE = 2πk + r|J|dL1 dL2 , (5.32)
∂r ∂r ∂z ∂z
0 0

and the elemental force vector, FE , as

Z1 1−L
Z 2
FE = 2π NQ̇r|J|dL1dL2 , (5.33)
0 0

allows one to write the residual equation as

Z
∂T
− Nk dA + KE T − FE + CE Ṫ = 0 (5.34)
∂n

Three types of boundary conditions are required in the analysis: Dirichlet, Neumann,

and mixed conditions. Over the exterior boundary of the foam insulation and over

the wetted and non-wetted regions of the interior wall, mixed boundary conditions of

the type
∂T
−k = h(T − T∞ ) (5.35)
∂n

148
5.2 ZONAL MODEL IMPLEMENTATION

are applied. Along the symmetry axis, Neumann conditions are applied:

∂T ′′
−k =q =0 (5.36)
∂n

Finally a Dirichlet condition is applied to the node on the interior tank wall that

intersects the liquid-vapor interface:

T = Ti (5.37)

Care is taken during mesh generation to ensure a node is located at the initial position

of the interface. During the simulation, the location of the node is fixed.

For the mixed condition, integration over the boundary yields

Z Z
∂T
− Nk dA = Nh(T − T∞ )dA (5.38)
∂n
Z
= 2π Nh(T − T∞ )rdΓ (5.39)

R
where dΓ represents integration over the boundary. The boundary is defined by

two nodes and variation over the boundary is assumed linear. Along the boundary,

 
1
 2 (1 − ξ) 
N=
 
 (5.40)
1
2
(1 + ξ)

where the one dimensional coordinate ξ is shown in Fig. 5.3. In the coordinate, ξ,

149
5.2 ZONAL MODEL IMPLEMENTATION

ξ = −1 ξ=1
i L/2 j

Figure 5.3: 1D linear element

the integral over the boundary is

Z Z1
L
2π Nh(T − T∞ )rdΓ = 2π Nh(T − T∞ )rdξ (5.41)
2
−1
 Z1  Z1
= πLh NNT rdξ T − πLhT∞ Nrdξ (5.42)
−1 −1

where L is the length of the side of the boundary element. Integration is again

performed using Gaussian quadrature:

Z1 N pts
X
g(ξ)dξ = g(ξi)Wi . (5.43)
−1 i=1

Integration points and weights for this one dimensional integral are listed in Table

5.2.

150
5.2 ZONAL MODEL IMPLEMENTATION

ξi Wi
0.0000000000000000000000000 0.4179591836734693877551020
0.4058451513773971669066064 0.3818300505051189449503698
-0.4058451513773971669066064 0.3818300505051189449503698
0.7415311855993944398638648 0.2797053914892766679014678
-0.7415311855993944398638648 0.2797053914892766679014678
0.9491079123427585245261897 0.1294849661688693270611400
-0.9491079123427585245261897 0.1294849661688693270611400
Table 5.2: Integration points and weights for Gaussian quadrature over a 1D
element.

R
For these mixed boundary conditions, πLhT∞ Nr dξ contributes to the elemental
R
force vector and πLh NNT r dξ contributes to the elemental stiffness matrix.

Since only zero flux Neumann conditions are considered in this analysis, these

boundaries do not contribute any terms to the stiffness matrix or force vector. The

direct stiffness method is used to form a global set of equations of the form

CṪ + KT = F (5.44)

A backward Euler method is used to solve the above ODE:

 
n+1
C T − T + ∆tKTn+1 = ∆tFn+1
n
(5.45)

or

(C + ∆tK)Tn+1 = ∆tKTn + ∆tFn+1 (5.46)

The global set of equations is modified233 to ensure that for the node where the

151
5.2 ZONAL MODEL IMPLEMENTATION

Dirichlet condition is applied

T n+1 = T n = Ti . (5.47)

The solver proceeds by inverting C + ∆t K and then solving

Tn+1 = ∆t(C + ∆t K)−1 KTn + ∆t (C + ∆t K)−1 Fn+1 . (5.48)

The finite element equations in the solid regions are coupled to the evolution equations

in the bulk liquid and vapor phases through the mixed boundary conditions.

On the wetted and non-wetted interior tank wall respectively, the mixed boundary

condition becomes

h(T − T∞ ) = hwl (T − Tl ) (5.49)

h(T − T∞ ) = hwv (T − Tv ). (5.50)

Convection coefficients are determined by first evaluating area averages of tempera-

ture along the wetted and non-wetted parts of the wall

Z
1
Twl = T dA (5.51)
Awl
wl
Z
1
Twv = T dA. (5.52)
Awv
wv

152
5.2 ZONAL MODEL IMPLEMENTATION

A Rayleigh number can be evaluated for both bulk phases

ρ2 Cp gβ(Tavr − Tbulk )L3

Ra = . (5.53)
µk

Transport properties in the liquid are evaluated at the film temperature, 21 (Twl + Tl ).

The characteristic length along the wetted part of the wall is the interface height.

Along the non-wetted part of the wall, the characteristic height is the interface height

subtracted from the tank height. Transport properties in the ullage are evaluated us-

ing a mass-weighted average of the vapor and non-condensable gas properties. Defin-

ing the vapor and gas mass fractions as

ρ = ρv + ρg (5.54)
ρv
ωv = (5.55)
ρ

ωg = 1 − ωv , (5.56)

the conductivity and viscosity can be computed

   
Twv + Tv Twv + Tv
k = ωv kv + ωg kg (5.57)
2 2
   
Twv + Tv Twv + Tv
µ = ωv µ v + ωg µ g . (5.58)
2 2

Once the Rayleigh number is computed, the following Nusselt number correlation234

153
5.2 ZONAL MODEL IMPLEMENTATION

is applied:
1
Nu = 0.12 Ra 3 . (5.59)

Heat transfer coefficients can then be evaluated

k Nu
h= . (5.60)
L

Heat transfer coefficients and bulk temperatures defined in this way are incorporated

into the mixed boundary conditions of the finite element equations. The bulk phase

equations for the multizone model were derived in Chapter 4. After modifying the

equation set to account for the presence of a non-condensable gas (Appendix C), these

equations become

dVv M
= (5.61)
dt ρl
 
dρv M ρv
= 1− (5.62)
dt Vv ρl
dρg M ρg
=− (5.63)
dt Vv ρl


dTl
{ρl cpl Vl } = Mcp,l (Tl − Tli ) + ṁcp,l (Tjet − Tl ) + Q̇wl + Q̇il +
dt
    (5.64)
∂pv dρv ∂pg dρg ∂pv ∂pg dTv
Vl + + +
∂ρv dt ∂ρg dt ∂T ∂T dt

154
5.2 ZONAL MODEL IMPLEMENTATION
   
∂pv ∂pg dTv
ρv cp,v Vv + ρg cp,g Vv − Vv + = Mcp,v (Tvi − Tv ) + Q̇wv − Q̇iv +
∂T ∂T dt
 
∂pv dρv ∂pg dρg
Vv + .
∂ρv dt ∂ρg dt

(5.65)

Q̇wl and Q̇wv are evaluated by integrating the mixed boundary conditions over the

wetted and non-wetted parts of the wall respectively:

Z
Q̇wl = hwl (T − Tl ) dA (5.66)
wl
Z
Q̇wv = hwv (T − Tv ) dA. (5.67)
wv

The jet temperature, Tjet , is computed using the heat exchanger module described in

Appendix B. During the self-pressurization phase of the simulation, ṁ = 0. During

the mixing phase of the experiment

ṁ = ρl |vjet | Anozz . (5.68)

The interfacial temperature, Ti , is computed iteratively using eqn. (4.69). Once Ti is

known, the evaporative mass transfer rate can be evaluated

  12
2σ 1
M= Ai (psat (Ti ) − pv ). (5.69)
2−σ 2πRv Ti

The interfacial heat powers, Q̇il and Q̇iv are evaluated using heat transfer correlations.

155
5.2 ZONAL MODEL IMPLEMENTATION

During self-pressurization, the following Nusselt number correlation is applied

1
Nu = 0.12 Ra 3 . (5.70)

The characteristic length used to evaluate these dimensionless parameters is the tank

radius. Transport properties are evaluated at

1
Tf ilm = (Ti + Tl ) (5.71)
2

on the liquid side of the interface and at

1
Tf ilm = (Ti + Tv ) (5.72)
2

on the vapor side of the interface. Once the Nusselt numbers are known, the heat

transfer coefficients can be determined

k Nu
h= . (5.73)
Rt

The interfacial heat powers are given by

Q̇il = hil Ai (Ti − Tl ) (5.74)

Q̇iv = hiv Ai (Tv − Ti ). (5.75)

156
5.2 ZONAL MODEL IMPLEMENTATION

Since the interface is assumed flat,

Ai = π Rt2 . (5.76)

During jet mixing, the liquid side heat transfer coefficient is replaced by

 
1 1 Cp,l |Ti − Tl |
hil = ρl cp,l |vjet| · 0.0198Pr 3 1− (5.77)
2 L

which is derived based on a vapor condensation coefficient driven by liquid side turbu-

lence from an axisymmetric liquid jet.83 The evolution equations in the bulk phases

along with the finite element equations are evolved implicitly in time. Within the

time integration loop, first the finite element equations are solved and then the bulk

fluid equations. The implicit integration proceeds iteratively until the L2 norm of the

change in the solution falls below 1 × 10−5 .

157
 Chapter 6

Results

In this chapter, the predictive capability of both the thermodynamic and zonal

models will be assessed. Model predictions will be compared with the experimental

data presented in Chapter 2. Initially, comparisons will be made to the data set

corresponding to when the tank is 50% full of liquid and 2 W of heat enters the system.

First the results of the thermodynamic model will be analyzed. A sensitivity analysis

was performed to assess the effects of various parameters on the thermodynamic

solution. The thermodynamic model was initialized using the parameters outlined

for Case 12 in Table 2.3. Originally, the thermodynamic model did not include the

effects of the tank wall. In this formulation, the predicted self-pressurization rate was

significantly larger than the experimental data as shown in Fig. 6.1. By not including

the wall, all of the applied heat load enters the liquid-ullage system which results in

a large pressure rise. In the experiment, during self-pressurization, the wall acts like

a heat sink - absorbing some of the applied load. When the wall is included in the

model, the self-pressurization rate is underpredicted. This is a consequence of the

158
 RESULTS

25
20
15 Data
10 No Wall
Wall
5
0
P-P o [Torr]
-5
-10
-15
-20
-25
-30
-35
-40
0 2500 5000 7500 10000 12500
Time [s]

Figure 6.1: Effect of including the tank wall in the homogeneous model.

thermal inertia of the wall and the homogeneity assumption. Homogeneity requires

that the temperature of the wall changes in a lumped fashion. Homogeneity and the

large thermal capacity of the acrylic enhances the thermal sink effects of the wall

which results in a smaller predicted pressure rise.

During subcooled jet mixing, the same wall effects are present. By not including

the wall, the liquid-ullage system is more responsive to subcooled mixing which results

in a steeper depressurization rate as shown in Fig. 6.1. During subcooled mixing, the

walls act as a heat source. Before the pressure decays, the walls must give up their

sensible energy. Because the walls are modeled as a lumped system, it takes longer

for the wall to give up its energy than if heat transfer in the wall were accounted

for in a local way. Hence, the depressurization response is slower than the measured

data.

159
 RESULTS

25
20
15 Data
10 Insulated
Heat Transfer
5
0
P-P o [Torr]
-5
-10
-15
-20
-25
-30
-35
-40
0 2500 5000 7500 10000 12500
Time [s]

Figure 6.2: Effect of heat exchange with the ambient environment.

In addition to considering the effects of the presence of the wall, the effect of

external heat transfer is also analyzed. In the experiment, the wall is blanketed with

insulation. Ideally, the insulation should act as an adiabatic boundary. Realistically,

it has a non-zero thermal conductivity and heat transfer could occur through the

insulation layer. The average temperature of the ambient environment is essentially

constant. Whether or not the external environment acts as a heat sink or heat source

depends on whether the constant ambient temperature is greater or less than the

liquid-ullage-wall system temperature. In this particular liquid heating case, the

average ambient temperature is 295.75 K and the initial temperature of the system is

295.38 K. Hence, the warmer ambient environment results in an additional heat load

on the system driven by natural convection at the exterior boundary of the insulation

layer.

160
 RESULTS

This additional heat source is reflected in the results shown in Fig. 6.2. When

this heat source is accounted for, the model predicts a larger pressure rise during self-

pressurization and a slower depressurization rate during subcooled mixing. While the

effect of external heat transfer is small, it is non-negligible and will be accounted for

in subsequent analyses.

161
 RESULTS

Element count ∆t [s] ∆p [torr]

545 1.0 0.132
1008 0.1 0.131
1008 1.0 0.132
1008 10.0 0.135
1719 1.0 0.131
Table 6.1: Sensitivity of the zonal solution.

Some of the lack of agreement between the thermodynamic model and the exper-

imental data stems from modeling the liquid-ullage-wall system in a lumped fashion.

The zonal model relaxes the homogeneity assumption by treating the bulk phases as

separate systems and accounting for local heat transfer in the wall.

Before analyzing the results of the zonal model, the sensitivity of the zonal solution

to grid size and time step will be discussed. The zonal model for this liquid heating

case was run for different mesh densities and time steps. After 300 s, the difference

between the current ullage pressure and the initial ullage pressure was computed. The

results are listed in Table 6.1. The solution is fairly constant over the range of mesh

densities and time steps considered. The variation in the solution over all sensitivity

parameters is within 3%. In the results that follow, the finite element grid consisted

of 1008 elements with a denser clustering of nodes near the heaters. During the two-

hour self-pressurization simulation, the timestep was set to 10 s. During subcooled

mixing, for better temporal resolution, the time step was reduced to 0.5 s.

Wall temperature contours after self-pressurizing for 1800 s and 7200 s are shown

in Figs. 6.3 and 6.4 respectively.

162
 RESULTS

Ullage T [K]

296.8
296.6
296.4
296.2
296.0
295.8

Liquid

Figure 6.3: Wall temperature contours during self-pressurization.

Time = 1800 s.

163
 RESULTS

After 1800 s, the hottest part of the system is the submerged heater. Heat has

barely conducted into the wall, which is not surprising given the low thermal dif-

fusivity of the acrylic. After self-pressurizing for two hours, the hottest part of the

system is still the submerged heater but the tank wall adjacent to the bulk liquid

has also increased in temperature as shown in Fig. 6.4. The sensible energy of the

bulk liquid increases due to the heat being supplied by the submerged heater. As the

temperature of the bulk liquid increases, the liquid gives up some of its energy to the

wetted wall through convection resulting in a wall temperature increase.

Temperature contours are shown in Figs. 6.5 and 6.6 when the subcooled jet

mixer is activated. After 2700 s of mixing, the wall adjacent to the bulk liquid has

cooled as indicated in Fig. 6.5. Since the heater is still on during mixing, locally near

the heater, hot spots exist. Away from the heater, convection between the subcooled

liquid and the wall has resulted in a temperature decrease. As time progresses, the

degree of wall cooling increases. After two hours of mixing, the temperature of the

wall adjacent to the liquid has continued to cool.

For the non-wetted part of the wall, thermal equilibration occurs over a much

longer timescale as heat has to conduct either into the lower conductivity vapor or

through the acrylic wall. Consequently, after two hours of subcooled mixing, the

temperature of the top part of the tank wall has not changed appreciably.

In Fig. 6.7, the pressurization behavior predicted by the zonal model is compared

with experimental data. During self-pressurization, the agreement between the model

and the data is excellent. The model is able to predict both the pressure rise and
164
 RESULTS

Ullage T [K]

296.8
296.6
296.4
296.2
296.0
295.8

Liquid

Figure 6.4: Wall temperature contours during self-pressurization.

Time = 7200 s.

165
 RESULTS

Ullage T [K]

296.0
295.8
295.6
295.4
295.2
295.0
294.8
294.6
Liquid 294.4

Figure 6.5: Wall temperature contours during subcooled mixing. Elapsed time
after activating jet = 2700 s.

166
 RESULTS

Ullage T [K]

295.6
295.4
295.2
295.0
294.8
294.6
294.4
294.2
Liquid 294.0
293.8

Figure 6.6: Wall temperature contours during subcooled mixing. Elapsed time
after activating jet = 5400 s.

167
 RESULTS

15

10
Data
5
Zonal
0

P-P o [Torr] -5

-10

-15

-20

-25

-30

-35
0 2500 5000 7500 10000 12500
Time [s]

Figure 6.7: Comparison of zonal pressure prediction with experimental data.

self-pressurization rate. During subcooled mixing, the initial stage of the depressur-

ization is well predicted. In appears that at the end of the subcooled mixing phase

of the experiment, the zonal model and the data are asymptotically approaching

different thermodynamic states. When a final steady state is achieved, the ullage

pressure becomes constant in time and no net condensation or evaporation occurs at

the interface. Moreover, the amount of energy entering and leaving each bulk phase

is balanced. In the liquid, for example, the amount of energy being removed by the

subcooled jet is balanced by the energy entering the bulk liquid through the wetted

walls and through the interface. In the ullage, at equilibrium, the net mass transfer

rate across the interface is zero. The energy entering and leaving the ullage through

the interface and walls is balanced. Hence,

Q̇iv = Q̇wv . (6.1)

168
 RESULTS

Or, after expressing the heat powers in terms of convection coefficients,

hiv Ai (Tv − Ti ) = hwv Awv (Tw − Tv ) (6.2)

where Tw is the average temperature over the non-wetted wall. Solving for the ullage

temperature yields

hiv Ai hwv Awv

Tv = Ti + Tw (6.3)
hiv Ai + hwv Awv hiv Ai + hwv Awv

The final ullage temperature is an average of the interfacial temperature and non-

wetted wall temperature weighted by the product of heat transfer coefficient and area.

This temperature, along with density, are substituted into the equation of state to

determine the steady state equilibrium pressure in the ullage.

After 5400 s of subcooled mixing, the temperature contours shown in Fig. 6.6

suggest that the interfacial temperature is lower than the temperature of the wall

adjacent to the ullage. Since the predicted ullage pressure is asymptoting to a lower

pressure than the experimental data, eqn. (6.3) implies that the zonal model is over-

predicting the heat transfer coefficient between the ullage and the interface. The final

ullage temperature is being weighted more heavily by the colder interface resulting

in a lower steady state pressure.

The lack of agreement in the final ullage pressure is a consequence of not only

the correlations used to derive the heat transfer coefficients but also the assumption

169
 RESULTS

1.00 Data

0.75 Zonal

0.50

T-To [K] 0.25

0.00

-0.25

-0.50

-0.75

-1.00
0 2500 5000 7500 10000 12500
Time [s]

Figure 6.8: Comparison of zonal temperature prediction with experimental

data.

of homogeneity within each bulk phase. A more local representation of the heat flux

between the ullage and wall and between the ullage and interface would yield a higher

fidelity model of the heat being transferred into and out of the ullage.

The breakdown of the homogeneity assumption is also apparent in Fig. 6.8 which

shows a comparison between the time history of the average ullage temperature pre-

dicted by the model and a time history of the local temperature recorded by the

thermistor.

During self-pressurization, the average temperature response predicted by the

zonal model lags the local temperature response recorded by the thermistor. At

the end of the two hour pressurization period, the predicted temperature rise is only

0.26 K compared to the 0.80 K recorded temperature rise. During subcooled mixing,

the zonal model does not capture the jump in temperature at approximately 8000 s.

170
 RESULTS

In Chapter 2, it was posited that this jump was caused by local wall heating as the

condensation flux decreases. Since the zonal model only tracks the average energy of

the ullage, it fails to capture local energy distributions within the bulk phase.

During the self-pressurization phase of this particular liquid heating case, the

primary driver for pressurization is evaporation through the interface. Given the

equation of state,

p = pv (ρv , Tv ) + pg (ρg , Tv ), (6.4)

the pressure rise can be approximated by

 
∂pv ∂pg ∂pv ∂pg
∆p ≈ ∆ρv + ∆ρg + + ∆Tv (6.5)
∂ρv ∂ρg ∂Tv ∂T v

The contributions of these different terms are listed in Table 6.2. The partial deriva-

tives listed in Table 6.2 are evaluated at the initial state of the ullage. The change

in density and temperature are evaluated using the difference between the predicted

state at the end of the two hour self-pressurization period and the initial state of the

ullage. As indicated in the table, the derivative of pressure with respect to density

is significantly larger than the derivative with respect to temperature. Hence, only a

small change in vapor or gas density is required to affect a large change in ullage pres-

sure. The vapor density increases because during evaporation, mass is being added

to the ullage. The gas density is decreasing because as the liquid is vaporizing and

the interface is receding, the ullage expands. Finally, the temperature of the ullage

171
 RESULTS
 
∂pv ∂pg ∂pv ∂pg
∆p ≈ ∆ρv + ∆ρg + + ∆Tv
∂ρv ∂ρg ∂Tv ∂Tv
| {z } | {z } | {z }
3 3
= 11476 Pa·m
kg
· 0.1499 mkg3 + 84774 Pa·m
kg
· −4 · 10−6 mkg3 + 235 Pa
K
· 0.26K

= 1720 Pa - 0.3 Pa + 61.1 Pa

Table 6.2: Contributions to pressure rise.

increases due to heat and mass transfer additions at the ullage boundary as well as

ullage compression. By noting the relative magnitudes of the different contributions

to the total pressure rise, one can deduce that evaporation through the interface is

driving the ullage pressure rise.

An honest assessment of the zonal model requires comparing the model’s predic-

tions to more than just a single data point. In what follows, the model is spot-checked

by comparing predictions to different cases in the experimental test matrix.

In Fig. 6.9, the zonal model is compared against a case corresponding to when the

tank is 26.5% full of liquid and 2 W of heat is entering the system. During mixing, the

jet speed is 0.241 cm/s and the heat exchanger bath temperature is 20 o C and 17.5 o C

in Figs. 6.9a and 6.9b respectively. The model does a reasonable job of predicting the

self-pressurization rate and initial depressurization rate. Similar to the 50% fill level

case, the model predicts a final ullage pressure lower than the experimental data. For

reasons outlined above, the discrepancy is most likely attributed to overpredicting

the heat transfer coefficient between the ullage and the interface.

The zonal model is compared against a lower heat load case is Fig. 6.10. In this

172
 RESULTS

30

Zonal
20 Data

10
P - Po [Torr]

-10

-20

-30

-40
0 3600 7200 10800 14400
Time [s]

(a)

20 Zonal
Data

0
P - Po [Torr]

-20

-40

-60

-80
0 3600 7200 10800 14400
Time [s]

(b)

Figure 6.9: Comparisons when the tank is 26.5% full and a 2 W heat load is
applied. The jet speed is 0.241 cm/s and the bath temperature is (a) 20 o C and
(b) 17.5 o C.

173
 RESULTS

20

Zonal
Data
10

P - Po [Torr]
-10

-20

-30

-40
0 3600 7200 10800 14400
Time [s]

Figure 6.10: Comparisons when the tank is 26.5% full and a 1 W heat load is
applied. The jet speed is 0.241 cm/s and the bath temperature is 20 o C.

case, the tank is 26.5% full of liquid and only 1 W of heat is entering the tank. The self-

pressurization rate and initial depressurization rate agree well with the experimental

data.

Predictions are made for a higher fill level case and comparisons are shown in Fig.

6.11. Here, the tank is 73.5% full of liquid with an incident heat load of 2 W. As

indicated in the figure, there is excellent agreement between the model predictions

and the experimental data during self-pressurization. During subcooled mixing, the

depressurization rate is overpredicted. For this high fill level case, the liquid in the

tank absorbs more incident energy than in the lower fill level cases. The subcooled

jet is trying to penetrate upward into a warmer medium. Adverse buoyancy is hin-

dering the jet’s upward flow. Consequently, the jet has to remove energy from the

bulk liquid before it can impinge the interface causing the pressure collapse. Since

174
 RESULTS

20

Zonal
Data
10

0
P - Po [Torr]

-10

-20

-30
0 3600 7200 10800 14400
Time [s]

Figure 6.11: Comparisons when the tank is 73.5% full and a 2 W heat load is
applied. The jet speed is 0.241 cm/s and the bath temperature is 20 o C.

this adverse buoyancy interaction is an intraphase phenomena and the zonal model

assumes spatial uniformity within each bulk phase, a discrepancy exists between the

model’s predictions and the experimental data.

Finally, comparisons are made in Fig. 6.12 when the tank is 26.5% full and the

jet speed is only 0.114 cm/s. In this case, the self-pressurization rate is reasonably

predicted, but once again, the pressure collapse during subcooled mixing occurs over

a shorter time scale than what the data suggests. For this lower jet flow case, the

momentum of the jet is not strong enough to directly impinge the interface. The

upward motion of the jet is suppressed by buoyant convection in the liquid. Prior to

interface impingement, the buoyantly driven vortices interact with the forced jet flow.

The colder jet exchanges energy with the vortices, weakening them until impingement

can occur. Since the zonal model is unable to resolve this intraphase interaction, a

175
 RESULTS

30

Zonal
Data
20

10

P - Po [Torr]
0

-10

-20

-30
0 3600 7200 10800 14400
Time [s]

Figure 6.12: Comparisons when the tank is 26.5% full and a 2 W heat load is
applied. The jet speed is 0.114 cm/s and the bath temperature is 20 o C.

discrepancy exists between the model and the data.

Generally, the agreement between the model and the data is reasonable during

self-pressurization. During subcooled mixing, the model does an excellent job of

predicting the initial depressurization rate when the interaction between buoyant

convection and forced jet mixing is weak. When the interaction is strong, for low

jet speeds, the model predicts a faster pressure collapse than the data suggests. To

resolve this discrepancy at lower jet speeds, local transport phenomena within the

bulk phases must be accounted for in the model.

176
 Chapter 7

Conclusions

In this dissertation, the effectiveness of using subcooled jet mixing as a means

to control tank pressure was analyzed. A small-scale experiment with a refrigerant

fluid was performed in a normal gravity environment. During the self-pressurization

phase of the experiment, both the liquid fill fraction and heating rate were varied.

For a liquid heating configuration, the data revealed that increasing the heating rate

increased the self-pressurization rate. Increasing the liquid fill fraction decreased the

pressurization rate as more liquid was available to absorb the applied heat load.

After self-pressurizing for two hours, a mixing jet aligned with the tank’s central

axis was used to control tank pressure. Results showed that a mixing jet alone was not

sufficient to control tank pressure. Over a short time scale, after the jet is first turned

on, tank pressure did decrease. However, the decrease was only temporary. When

the bulk liquid became well-mixed, tank pressure began to rise at a rate higher than

self-pressurization alone. The mixing pump provided an additional energy source to

the liquid/vapor system and resulted in the increased pressurization rate.

177
 CONCLUSIONS

In order to sustain the pressure collapse, thereby increasing the effectiveness of

the jet, energy must be removed from the liquid stream. In this pressure control

phase of the experiment, both the jet speed and temperature were varied. For large

jet speeds, the experiment demonstrated that subcooled mixing jets are an effective

way to control pressure. When the jet’s momentum is strong enough to overcome

both adverse buoyancy and natural convection in the liquid, a rapid reduction in

tank pressure can be realized.

As a pressure control strategy, subcooled axial jet mixing will show a gravitational

dependence. In this experiment, when the jet’s momentum was small, an interaction

between the forced jet flow and natural convection in the liquid took place. Before

the pressure collapsed in the ullage, the forced jet would have to weaken the natural

convection vortices in the bulk phase. The interaction is much weaker in a low-

gravity environment due to reduced buoyancy effects. To effectively control tank

pressure when operating in a range of gravitational environments, it is recommended

that either the jet’s momentum to be large enough to overcome buoyancy or other

gravity-independent design solutions, such as a subcooled spray mixer, be considered.

In addition to performing a ground-based experiment, in this dissertation, the gov-

erning equations were used to formulate both a homogeneous thermodynamic model

and a zonal model. Comparing the thermodynamic model with the experimental data

highlighted some of the deficiencies of the thermodynamic modeling approach. First,

it was shown that the effects of the tank wall should not be neglected in the analysis.

The wall has a large thermal inertia and neglecting the wall in these instances will
178
 CONCLUSIONS

result in the model significantly overpredicting both the self-pressurization and de-

pressurization rates. When the wall was included in the analysis, the thermodynamic

model underpredicted the experimental data. This discrepancy was primarily caused

by the homogeneity assumption in the thermodynamic model. The assumption forces

the liquid, ullage, and wall to be at the same temperature which changes in a lumped

fashion based on energy conservation. When thermal non-uniformities are present in

the system, the thermodynamic model fails to capture these inhomogeneous regions

and will not be able to make detailed temperature or pressure predictions. As a design

tool, the thermodynamic model is useful at uncovering trends in the pressurization

behavior but it is limited in its predictive ability.

The zonal model relaxed the restrictive homogeneity assumption by treating the

bulk liquid and ullage as separate zones. Within each zone, all thermodynamic quan-

tities were assumed to be spatially uniform. The liquid zone and ullage zone were

coupled to each other through heat and mass transfer correlations. Additionally, the

zonal model accounted for local heat transfer in the wall by making use of a finite

element model. The finite element model solved the energy equation in the wall and

was coupled to the bulk phase transport equations through heat transfer relationships

between the wetted and non-wetted parts of the interior tank wall and the liquid and

ullage.

Comparisons between the experimental pressure data and the zonal model predic-

tions are excellent for moderate to high jet flow rates. For slower jet speeds, buoyancy

in the liquid adversely affects the effectiveness of a subcooled mixing jet. Natural con-
179
 CONCLUSIONS

vection vortices in the bulk liquid suppress the upward motion of the jet. If the jet is

weak, buoyancy dominates and the cooling jet never reaches the interface. If the jet is

strong, the jet is able to reach the interface with little hindrance from the buoyantly

driven vortices. For intermediate jet flows, an interaction occurs where the subcooled

jet first has to weaken the buoyant flow before the jet can impinge the interface driv-

ing a pressure collapse. Since the zonal model assumes spatial uniformity within each

bulk phase, these buoyancy - forced jet flow interactions are not captured.

For faster jet flows, the interaction is weak and the zonal model does an excellent

job of predicting tank pressurization behavior. In these cases, the model can be used

as a design tool to determine the pressure response of a flightweight cryogenic storage

tank. As a design tool, expediency in obtaining an accurate prediction is paramount,

as it affords the opportunity to evaluate multiple design solutions in a quick and

straightforward manner. There are situations however where a more detailed model

is warranted. In particular, to capture buoyancy - forced jet flow interactions or

ullage position in low-gravity applications, it is recommended that future modeling

efforts include a computational fluid dynamics simulation of the bulk phases. These

simulations, while sacrificing expediency, can provide designers with a more thorough

picture of the complicated transport phenomena occurring inside a storage tank.

In addition to recommending future improvements to the modeling approach,

recommendations to improve future small-scale experiments are threefold. First, a

more robust degassing scheme should be developed. The current method for degassing

the test fluid was slow, time-consuming, and not as robust as desired. Improvements
180
 CONCLUSIONS

in this area would lead to a more rapid turnaround time between experiments. Second,

a non-intrusive way to measure temperature is desirable. In the current experimental

configuration, fully populating the test tank with thermistors resulted in a test cell

with a greater leak rate than if the thermistor ports were capped. Moreover, including

the thermistors in the tank resulted in an interaction with the underlying flow. If

the jet flow stagnated on a thermistor, the measured depressurization rate would be

smaller than if the thermistors were not present. Non-intrusive measurements provide

a means to record local bulk phase temperatures without interfering with the flow.

Finally, the current experiment was performed in a normal gravity environment. In

low gravity environments, buoyancy effects are reduced and even slow subcooled jets

may be effective at reducing tank pressure. Additionally, other effects masked by

buoyancy in 1g would become more prominent. The interaction between the jet and

the ullage itself may not be insignificant. The jet may geyser into the ullage or push

the ullage to one side of the tank. In the ground-based experiment, the liquid vapor

interface was essentially flat. In microgravity, the position of the interface and location

of the ullage is less certain. To fully evaluate these scenarios, it is recommended that

small-scale pressurization experiments be performed in low-gravity environments.

181
 Appendix A

Properties of HFE-7000

For the ground-based experiment, the test fluid is HFE-7000 (C3 F7 OCH3 ) - a

refrigerant with a relatively low normal boiling point (34 o C) which makes it well

suited for performing experiments without elaborate cryogenic thermal controls. The

fluid, provided by 3M, had a stated purity level of 99.5%. The molar mass is 200

g/mol. The critical properties, as reported by Geller235 are

Tc = 437.61 K ± 0.03 K

pc = 2483.3 kPa ± 2 kPa

ρc = 553 kg/m3 ± 0.1 kg/m3 .

Saturation tables provided by the manufacturer were used to develop correlations for

relevant thermodynamic variables. These correlations, along with their coefficient of

determination, and their range of validity are listed below. The density is given in

kg/m3 , the temperature in K, and the internal energy and enthalpy in kJ/mol.

182
 PROPERTIES OF HFE-7000

5
 ik 5
 ik
P T
P T
ρsat
l (T ) = Nk 1 − Tc
ρsat
v (T ) = Nk 1 − Tc
k=1 k=1

N1 =-5.144870514·103 i1 =4 N1 = 6.985225098·103 i1 =4

N2 = 8.203129740·103 i2 =3 N2 =-1.115730773·104 i2 =3

N3 =-5.332926413·103 i3 =2 N3 = 6.941222670·103 i3 =2

N4 = 2.795426861·103 i4 =1 N4 =-2.018442164·103 i4 =1

N5 = 8.526087154·102 i5 =0 N5 = 2.333191030·102 i5 =0

R2 =0.999 R2 =0.999

230 K < T < 388 K 230 K < T < 388 K

Table A.1: Saturation densities for HFE-7000.

5
P 5
P
usat
l (T ) = Nk T ik usat
v (T ) = Nk T ik
k=1 k=1

N1 = 2.35159740·10-9 i1 =4 N1 =-3.08196826·10-9 i1 =4

N2 =-2.73801441·10-6 i2 =3 N2 = 2.93282320·10-6 i2 =3

N3 = 1.37519874·10-3 i3 =2 N3 =-8.75529400·10-4 i3 =2

N4 =-1.07225719·10-1 i4 =1 N4 = 2.19185784·10-1 i4 =1

N5 =-2.13327641·101 i5 =0 N5 =-1.11377115·100 i5 =0

R2 =0.999 R2 =0.999

230 K < T < 388 K 230 K < T < 388 K

Table A.2: Saturation internal energies for HFE-7000.

183
 PROPERTIES OF HFE-7000

5
P 5
P
hsat
l (T ) = Nk T ik hsat
v (T ) = Nk T ik
k=1 k=1

N1 = 2.70825618·10-9 i1 =4 N1 =-3.81657780·10-9 i1 =4

N2 =-3.09843024·10-6 i2 =3 N2 = 3.61708992·10-6 i2 =3

N3 = 1.51386682·10-3 i3 =2 N3 =-1.13113147·10-3 i3 =2

N4 =-1.31220610·10-1 i4 =1 N4 = 2.72161968·10-1 i4 =1

N5 =-1.97620855·101 i5 =0 N5 =-4.14395325·100 i5 =0

R2 =0.999 R2 =0.999

230 K < T < 388 K 230 K < T < 388 K

Table A.3: Saturation enthalpies for HFE-7000.

For the above thermodynamic data, the reference point is given as the saturated

liquid state at 230 K. The pressure-temperature relationship along the saturation

curve is given by the manufacturer as:

−3548.6
ln(p [Pa]) = + 22.978
T [K]

R2 = 0.998

243K <T < Tc

However, a refit of saturation data provided by Geller235 (accuracy ±0.05%) results

in a Wagner-type equation for vapor pressure:

184
 PROPERTIES OF HFE-7000

  a=-8.384714
aτ +bτ 1.5 +cτ 2.5 +dτ 5
p = pc exp 1−τ
b= 2.815690

R2 =0.999 c=-4.166609

241 K < T < 370 K d=-7.663946

where τ =1-T/Tc . This expression results in a better fit of the data especially around

the operating temperatures of the ground experiment. Geller also provided data on

the ideal gas specific heat capacity of HFE-7000. The reported accuracy on this data

is estimated as less than ±0.6%. Fitting a polynomial through his data yields:
  4
o kJ
P
cp (T [K]) kg·K = Nk T ik
k=1

N1 = 3.682659933·10-9 i1 =3

N2 =-5.441919192·10-6 i2 =2

N3 = 4.143476431·10-3 i3 =1

N4 =-1.666666656·10-4 i4 =0

R2 =0.999

260 K < T < 420 K

185
 PROPERTIES OF HFE-7000

Additionally, the manufacturer provided the following expressions for properties

in the liquid:

 
1
β = 0.00219 (A.1)
K
 
W
kl = 0.0798 − 0.000196 T [o C] (A.2)
m·K
 
J
cp l = 1223.2 + 3.0803 T [o C] (A.3)
kg · K

ν [cSt] = (z − 0.7) − exp[−0.7487 − 3.295(z − 0.7) + 0.6119(z − 0.7)2 − 0.3193(z − 0.7)3 ]

(A.4)

where
10.151−4.6006 log(T [K])
z = 1010 .

The range of validity for these correlations are -120 o C to 35 o C, -120 o C to 40 o C,

and -120 o C to 40 o C for the conductivity, specific heat, and kinematic viscosity

respectively. Curve fitting surface tension data provided by the manufacturer yields:

   1.016
mN T
σ = 42.830 1 − . (A.5)
m Tc

Reliable property measurements in the vapor phase have not been reported in the

literature. As such, relevant vapor phase properties must be estimated using a variety

of techniques. To estimate the value of the expansion coefficient and the specific

heat, an equation of state (EOS) is posited for the vapor phase. Then, using classical

186
 PROPERTIES OF HFE-7000

thermodynamics, it can be shown that:

p = p(ν, T ) (A.6)
1 (∂p/∂T )ν
β=− (A.7)
ν (∂p/∂ν)T
Z ∞   2 
o ∂ p
cv = cp − R − T dν (A.8)
ν ∂T 2 T
(∂p/∂T )2ν
cp = cv − T (A.9)
(∂p/∂ν)T

where R is the gas constant.

The EOS that best predicted the P-V-T relationship near the saturation curve

was chosen to represent HFE-7000 vapor. Some of the EOS described below make

use of the accentric factor,236 which is defined as

 
ω = − log lim (psat /pc ) − 1.0. (A.10)
T /Tc →0.7

With the vapor pressure curve defined previously, the accentric factor for HFE-

7000 evaluates to 0.413. Several EOS are evaluated to determine which EOS best

describes the P-V-T behavior of HFE-7000 near the saturation curve. The following

EOS are evaluated:

Ideal Gas
RT
p(ν, T ) = (A.11)
ν

187
 PROPERTIES OF HFE-7000

Redlich-Kwong

5/2
R2Tc
a = 0.42748
pc
RTc
b = 0.08664
pc
RT a
p(ν, T ) = − √ (A.12)
ν − b ν(ν + b) T

Peng-Robinson

(RTc)2
a = 0.45724
pc
RTc
b = 0.0778
pc
  2
2
p
a 1 + (0.37464 + 1.54226ω − 0.2699ω ) 1 − T /Tc
RT
p(ν, T ) = − (A.13)
ν−b ν 2 + 2bν − b2

Virial
 
RT B(T ) C(T )
p(ν, T ) = 1+ + (A.14)
ν ν ν2

Tsonopoulos237–239 provides correlations for the 2nd virial coefficient in terms of the

reduced temperature, T /Tc :

 
RTc
B= Fo (Tr ) + ωF1 (Tr ) (A.15)
pc
0.330 0.1385 0.0121 0.000607
Fo = 0.1445 − − − − (A.16)
Tr Tr2 Tr3 Tr8
0.331 0.423 0.008
F1 = 0.0637 + − − . (A.17)
Tr2 Tr3 Tr8

188
 PROPERTIES OF HFE-7000

Orbey and Vera240 provide a correlation for the 3rd virial coefficient:

 2
RTc
C= (Go (Tr ) + ωG1(Tr )) (A.18)
pc
0.02432 0.00313
Go = 0.01407 + − (A.19)
Tr2.8 Tr10.5
0.01770 0.040 0.003 0.00228
G1 = −0.02676 + + − − . (A.20)
Tr2.8 Tr Tr6 Tr10.5

Pressure eqn. p(νvsat ,290 K) p(νvsat ,294 K) p(νvsat ,298 K) r.m.s. error

Saturation 50173 Pa 59328 Pa 69748 Pa -

Ideal Gas 52082 Pa 61356 Pa 71978 Pa 2060 Pa

Redlich-Kwong 50922 Pa 59805 Pa 69921 Pa 522 Pa

Peng-Robinson 50636 Pa 59427 Pa 69425 Pa 331 Pa

Virial 50296 Pa 59011 Pa 68921 Pa 516 Pa

Table A.6: Comparison of different EOS in predicting P-V-T behavior near the

saturation curve

To assess the abilities of these models to accurately capture the P-V-T behavior

near the saturation curve, the predicted vapor pressures at 290 K, 294 K, and 298

K are compared with the saturation pressure given by eqn. (A.4). The r.m.s. error

is computed for each EOS. For the three representative temperatures, the Peng-

Robinson EOS results in the lowest error as indicated in Table A.6. Accordingly, HFE-

7000 vapor is assumed to behave as Peng-Robinson gas. Carrying out the operations

189
 PROPERTIES OF HFE-7000

described in eqns. (A.7)-(A.9) and evaluating the results at νvsat (298K) results in:

1
β = 0.00372 (A.21)
K
J
cv = 808 (A.22)
kg · K
J
cp = 856 . (A.23)
kg · K

The computed value of cv is within 11% of the independent estimation provided by

the manufacturer (cv = 730 J/kg · K). With the thermodynamic properties in the

vapor now estimated, the vapor viscosity and vapor thermal conductivity must be

determined before the transport equations can be solved. To estimate the vapor

viscosity, a corresponding states method described in Lucas241 is employed:

 1/6
Tc
ζ = 0.176 (A.24)
Mw3 Pc4
 
−1 0
µ[µP ] = ζ 0.807Tr .618 − 0.357 exp(−0.449Tr ) + 0.340 exp(−4.058Tr ) + 0.018 .

(A.25)

In eqn. (A.25), the molar mass is in g/mol, the temperature is in Kelvin, the pressure

is in bars, and the reduced temperature, Tr = T /Tc . This expression is valid for a low

pressure gas. Since the operating pressure is significantly less than the critical pressure

(p/pc ≈ 0.02), the high pressure corrections will be neglected. The viscosity computed

by eqn. (A.25) differs by approximately 7% from the independent estimation provided

by the manufacturer (µ = 109 · 10−7 Pa · s).

190
 PROPERTIES OF HFE-7000

The thermal conductivity for HFE-7000 vapor can be estimated from the Eucken

expression for polyatomic gases:

kMw 9/4
=1+ (A.26)
µcv cv R

where the units of conductivity, molar mass, viscosity, and specific heat are such that

the above equation is dimensionless. Computing the conductivity in this way yields

W
k = 0.0091 m·K .

To improve the ability of his equation to predict thermal conductivity, Eucken

later modified his original expression to:

kMw 1.77
= 1.32 + (A.27)
µcv cv R

W
which yields a conductivity of 0.012 mK . Poling et al.,242 however, notes that experi-

mental values of k typically lie between the values computed from eqns. (A.26) and

(A.27). The thermal conductivity estimated independently by the manufacturer is

W
0.01 mK .

191
 Appendix B

Heat Exchanger Model

During the pressure control phase of the experiment, the liquid withdrawn from

the tank passes through a shell and tube heat exchanger before it re-enters the tank

as a subcooled axial jet. The test fluid, HFE-7000, flows through the tubes in the heat

exchanger while water flows through the shell. It is assumed that the temperature of

the HFE-7000 entering the heat exchanger is equal to the average fluid temperature

at the tank outlet. Consequently, heat transfer to the fluid between the tank and

the heat exchanger is assumed negligible. Given this inlet temperature, the mass

flow rates of the two fluids, and the water inlet temperature, the following analysis

describes a model used to predict the temperature of the two fluids exiting the heat

exchanger. Once again, it is assumed that heat transfer to the HFE-7000 between

the heat exchanger exit and the tank inlet is negligible. In the analysis that follows,

the hot (HFE-7000) and cold (water) fluids correspond to the tube-side, t, and shell-

side, s, of the heat exchanger respectively. Various parameters describing the heat

exchanger is provided in Table 2.1. The heat exchanger is shown notionally in Fig.

192
 HEAT EXCHANGER MODEL

mt
Tjet
Tint

ms
Touts
in
T s

Figure B.1: Sketch of shell and tube heat exchanger.

B.1. Using the parameters listed in Table 2.1, the pitch, PT , and hydraulic diameter,

De can be computed:

PT = ODt + C (B.1)
4PT2
De = − ODt (B.2)
πODt

where C is the clearance between tubes. The equivalent areas of the tubes and shell

are

ID2t
At = Nt π (B.3)
4
IDs CB
As = (B.4)
PT

193
 HEAT EXCHANGER MODEL

where Nt is the number of tubes and B is the baffle spacing in the shell. With the

flow areas known, the flow velocities can be computed according to:

ṁt
Vt = (B.5)
ρt At
ṁs
Vs = (B.6)
ρs As

Given the Prandtl number of the fluids and the Reynolds numbers:

ρt Vt IDt
Ret = (B.7)
µt
ρs Vs De
Res = , (B.8)
µs

the Nusselt numbers can be evaluated using heat transfer correlations

 1/3
IDt Ret Prt
Nut = 1.86 (B.9)
L

Nus = 0.36Re0.55 1/3

s Prs . (B.10)

In the above correlation for the tube-side Nusselt number, a modified Sieder-Tate

equation for laminar flow is employed. For the jet speeds of interest, the flow in the

tubes is laminar. The inner and outer heat transfer coefficients are

Nut kt
hi = (B.11)
IDt
Nus ks
ho = . (B.12)
De
194
 HEAT EXCHANGER MODEL

The total exchange coefficient is given by

1 ODt 1 1
= + . (B.13)
U IDt hi ho

To compute the HFE-7000 outlet temperature, the ratio of thermal capacitances is

needed:
ṁs cps
R= . (B.14)
ṁt cpt

Defining
Tsout − Tsin
S= , (B.15)
Ttin − Tsin

one can determine the outlet temperature by solving the following equation for S:

 √ 
UAo 1 2 − S(R + 1 − R2 + 1)
=√ ln √ . (B.16)
ṁs Cps R2 + 1 2 − S(R + 1 + R2 + 1)

where the effective heat transfer area, Ao is listed in Table 2.1. The water outlet tem-

perature, Tsout can then computed using eqn. (B.15). The HFE-7000 exit temperature

can finally be evaluated using a simple energy balance:

ṁt cpt (Tjet − Ttin ) = ṁs cps (Tsout − Tsin ). (B.17)

This value, Tjet , is the jet temperature used in the homogeneous and zonal models.

195
 Appendix C

Non-Condensable Gases

In Chapter 4, models were developed to describe mass and energy transport in

a pure two-phase system. Modifications to these models due to the presence of a

non-condensable gas will now be discussed.

Consider a tank containing the liquid and vapor phases of a pure species in addition

to a non-condensable gas species in the ullage. Dissolution of the non-condensable

gas into the bulk liquid will be neglected. Mass conservation can be described by:

Z  Z Z Z
d
ρl dV − ρl (vl − vs ) · n̂ dS + ρl vl · n dS − ρl vl · n dS = 0 (C.1)
dt Vl I out in
Z  Z
d
ρv dV + ρv (vv − vs ) · n̂ dS = 0 (C.2)
dt Vv I
Z 
d
ρg dV = 0. (C.3)
dt Vv

196
 NON-CONDENSABLE GASES

As before, one can define,

j = ρl (vl − vs ) · n̂ = ρv (vv − vs ) · n̂ (C.4)

Z
M = − j dS (C.5)
I
Z
1
ρl = ρl dV (C.6)
Vl Vl
Z
1
ρv,g = ρv,g dV (C.7)
Vv Vv
Z Z
ṁ = ρl vl · n dS = ρl vl · n dS (C.8)
in out

which yields,

d
(ρ Vl ) = −M (C.9)
dt l
d
(ρ Vv ) = M (C.10)
dt v
d
(ρ Vv ) = 0. (C.11)
dt g

Similar to the pure two-phase system,the energy balance in the liquid phase is

Z  Z Z Z
d
ρl hl dV − ρl hl (vl − vs ) · n̂ dS + ρl hl vl · n dS − ρl hl vl · n dS =
dt Vl I out in
Z Z Z Z Z
dp
− Jq,l · n dS + Jq,l · n̂ dS − Jq,l · n dS + Jq,l · n dS + dV.
w I out in Vl dt

(C.12)

197
 NON-CONDENSABLE GASES

For the gas mixture in the ullage

Z  Z Z Z Z
d dp
ρhdV + ρh(v−vs )· n̂ dS = − Jq,v ·n dS − Jq,v · n̂ dS + . (C.13)
dt Vv I w I vv dt

The two species in the gas mixture are assumed to mix ideally,

ρh = ρv hv + ρg hg (C.14)

p = pv (ρv , T ) + pg (ρg , T ). (C.15)

For this multicomponent ullage, the energy flux consists of a heat flux, J̃q in addition

to the flux of energy due to species diffusion

X
Jq = J̃q + ji hi (C.16)
i=g,v

where the diffusive flux is defined as:

ji = ρi (vi − v). (C.17)

In the above expression, v is the mass-averaged velocity of the gas mixture. Before

198
 NON-CONDENSABLE GASES

substituting these expressions into eqn. (C.13), it is useful to note that

ρh(v − vs ) = ρv hv (v − vs ) + ρg hg (v − vs ) (C.18)

= ρv hv (vv − vs ) − ρv hv (vv − v) + ρg hg (vg − vs ) − ρg hg (vg − v)

(C.19)
X
= ρv hv (vv − vs ) − (ji hi ). (C.20)

The ρg (vg − vs ) term is zero because there is no mass transfer of the non-condensable

species across the interface. Defining similar averages and terms as in the pure two-

phase system (eqns. (4.18)-(4.22)), and after substituting these expressions into the

ullage energy balance, the following equations for the two phase system results:

Z
d dp
(ρ hv Vv + ρg hg Vv ) + j hv dS = Q̇wv − Q̇iv + Vv (C.21)
dt v I dt
Z  
d dp out in
(ρ hl Vl ) − j hl dS = Q̇wl + Q̇il + Vl − ṁ hl − hl . (C.22)
dt l I dt

The energy flux due to species diffusion, which appears in eqns. (C.16) and (C.20)

cancel with one another. Hence, the heat powers appearing the above equations are

only due to conduction.

199
C.1 HOMOGENEOUS MODEL NON-CONDENSABLE GASES

C.1 Homogeneous Model

Invoking the homogeneity assumption; that is, the entire system is isothermal and

the liquid and vapor phases are defined by their saturation states leads to:

   
d sat sat sat in sat dp
ρl hl Vl + M hl − ṁ hl − hl = Q̇wl + Q̇il + Vl (C.23)
dt dt
 
d dp
ρsat sat
v hv Vv + ρg hg Vv − M hsat
v = Q̇wv − Q̇iv + Vv (C.24)
dt dt

Adding the above equations results in

     
d sat sat sat sat sat sat in sat
ρ h Vl + ρv hv Vv + ρg hg Vv − M hv − hl − ṁ hl − hl =
dt l l
dp
Q̇w + (Q̇il − Q̇iv ) + Vt .
dt
(C.25)

Or, after carrying out the differentiation, applying eqns. (C.9)-(C.11) and the energy

jump condition across the interface yields

   
dhsat
l dhsat
v dhg dpsat
ρsat
l V l +ρsat
v V v +ρg Vv sat sat
−M hv −hl −Vt in sat
= Q̇w + ṁ hl −hl .
dt dt dt dt
(C.26)

Treating the non-condensable gas as an ideal gas with a constant specific heat leads

to
dhg dT
= cp g . (C.27)
dt dt

200
C.2 ZONAL MODEL NON-CONDENSABLE GASES

Since homogenity is invoked,

dhg dTsat
= cp g . (C.28)
dt dt

Hence,

 
dTsat sat dhsatl sat dhsat
v d(ρsat
v Vv ) dp
ρl Vl + ρv Vv + ρg Vv cpg + L − Vt
dt dTsat dTsat dTsat dTsat
  (C.29)
in sat
= Q̇w + ṁ hl − hl .

C.2 Zonal Model

In a multizone formulation, uniformity within each bulk phase is still assumed,

but the assumption that the phases are at the same temperature, Tsat , is relaxed. The

tank is again partitioned into three zones: a liquid zone, an ullage zone, and a thin

massless interfacial zone at temperatures Tl , Tv , and Ti respectively. The liquid phase

is assumed incompressible. The ullage contains both vapor and a non-condensable

gas. The two gases are assumed to mix ideally,

p = p(ρv , Tv ) + pg (ρg , Tv ) (C.30)

201
C.2 ZONAL MODEL NON-CONDENSABLE GASES

Similar to the pure liquid vapor system, the integrated continuity equation in the

liquid is given by:

d
(ρl Vl ) = −M
dt
dVl (C.31)
ρl = −M
dt
dVv
−ρl = −M
dt

which implies that

dVv M
= . (C.32)
dt ρl

Liquid incompressibility and the volume constraint were used to arrive at eqn. (C.32).

The continuity equation for the vapor in the ullage is

d
(ρv Vv ) = M (C.33)
dt
dVv dρv
ρv + Vv =M (C.34)
dt dt
M dρv
ρv + Vv =M (C.35)
ρl dt

which leads to
 
dρv M ρv
= 1− . (C.36)
dt Vv ρl

202
C.2 ZONAL MODEL NON-CONDENSABLE GASES

Since no dissolution of the non-condensable gas is assumed to occur, the continuity

equation for the non-condensable gas in the ullage is

d
(ρg Vv ) = 0 (C.37)
dt
dVv dρg
ρg + Vv =0 (C.38)
dt dt
M dρg
ρg + Vv =0 (C.39)
ρl dt

which leads to
dρg M ρg
=− . (C.40)
dt Vv ρl

Uniformity within each bulk phase simplifies the bulk phase energy balances to

 
d i out in dp
(ρl hl Vl ) + M hl = −ṁ hl − hl + Q̇wl + Q̇il + Vl (C.41)
dt dt
d dp
(ρv hv Vv + ρg hg Vv ) − M hiv = Q̇wv − Q̇iv + Vv . (C.42)
dt dt

The enthalpy of the bulk phases is assumed to behave as

hv = hv,o + cp,v (Tv − Tv,o ) (C.43)

hg = hg,o + cp,g (Tg − Tg,o ) (C.44)

hl = hl,o + cp,l (Tl − Tl,o ). (C.45)

203
C.2 ZONAL MODEL NON-CONDENSABLE GASES

Substituting these expressions into the energy balances, carrying out the differentia-

tion, and rearranging terms leads to

dTl dp
ρl cp,l Vl = Mcp,l (Tl − Tli ) − ṁcp,l (Tl − Tin ) + Q̇wl + Q̇il + Vl (C.46)
dt dt
dTv dp
(ρv cp,v + ρg cp,g )Vv = Mcp,v (Tvi − Tv ) + Q̇wv − Q̇iv + Vv (C.47)
dt dt

where once again, liquid incompressibility has been invoked. The equation of state

can be used to explicitly evaluate the pressure derivative. That is,

   
dp ∂pv dρv ∂pv dTv ∂pg dρg ∂pg dTg
= + + + . (C.48)
dt ∂ρ dt ∂T dt ∂ρ dt ∂T dt

Substituting this expression into the energy balances yields


dTl
{ρl cpl Vl } = Mcp,l (Tl − Tli ) + ṁcp,l (Tin − Tl ) + Q̇wl + Q̇il +
dt
    (C.49)
∂pv dρv ∂pg dρg ∂pv ∂pg dTv
Vl + + +
∂ρv dt ∂ρg dt ∂T ∂T dt

   
∂pv ∂pg dTv
ρv cp,v Vv + ρg cp,g Vv − Vv + = Mcp,v (Tvi − Tv ) + Q̇wv − Q̇iv +
∂T ∂T dt
 
∂pv dρv ∂pg dρg
Vv + .
∂ρv dt ∂ρg dt

(C.50)

These equations represent the integral energy balances in the liquid and ullage

when a non-condensable gas is present.

204
 Bibliography

[1] J.A. Salzman. Fluid management in space-based systems. In Engineering,

Construction, and Operations in Space, 5th Int. Conference on Space, volume 1,

pages 521–526, 1996.

[2] F.P. Chiaramonte and J.A. Joshi. Workshop on critical issues in microgravity

fluids, transport, and reaction processes in advanced human support technology.

NASA TM 2004-212940, 2004.

[3] G.B. Sanders. ISRU: An overview of NASA’s current development activities and

long-term goals. AIAA 2000-1062, 2000.

[4] G.A. Landis and D.L. Linne. Mars rocket vehicle using in situ propellants. J.

Spacecraft and Rockets, 38(5):730–735, 2001.

[5] J.T. Howell, J.C. Mankins, and J.C. Fikes. In-space cryogenic propellant depot step-

ping stone. Acta Astronautica, 59:230–235, 2006.

[6] D.J. Chato. Low gravity issues of deep space refueling. AIAA 2005-1148, 2005.

[7] J.F. Schuster. Long term cryogenic facility systems study. In Microgravity Fluid

Management Symposium, NASA CP-2465, pages 17–30, 1986.

205
 [8] N.S. Brown. Advanced long-term cryogenic storage systems. In Microgravity Fluid

Management Symposium, NASA CP-2465, pages 7–16, 1986.

[9] NASA Exploration Systems Architecture Study. NASA TM-2005-214062, 2005.

[10] S.M. Motil and M.L. Meyer. Cryogenic fluid management technologies for advanced

green propulsion systems. AIAA 2007-343, 2007.

[11] R.C. Van Meerbeke. Thermal Stratification and Sloshing in Liquid Helium Trailers,

volume 13 of Advances in Cryogenic Engineering, pages 199–206. Plenum Press, 1968.

[12] R.L. DeWitt and R.J. Boyle. Thermal performance of an integrated thermal pro-

tection system for long-term storage of cryogenic propellants in space. NASA TN

D-8320, 1977.

[13] D.W. Plachta, R.J. Christie, J.M. Jurns, and P. Kittel. Passive ZBO storage of liquid

hydrogen and liquid oxygen applied to space science mission concepts. Cryogenics,

46:89–97, 2006.

[14] C.S. Guernsey, R.S. Baker, D. Plachta, and P. Kittel. Cryogenic propulsion with zero

boil-off storage applied to outer planetary exploration. AIAA 2005-3559, 2005.

[15] B. Gebhart. Random convection under conditions of weightlessness. AIAA J.,

1(2):380–383, 1963.

[16] P.G. Grodzka and T.C. Bannister. Natural convection in low-g environments. AIAA

74-156, 1974.

206
[17] R.C. Mitchell, J.A. Stark, and R.C. White. Zero-g Hydrogen Tank Venting Systems,

volume 12 of Advances in Cryogenic Engineering, pages 72–81. Plenum Press, 1967.

[18] A. Hofmann. Theory to boil-off gas cooled shields for cryogenic storage vessels.

Cryogenics, 44:159–165, 2004.

[19] R. Cunnington. Reducing Boil Off Losses in Cryogenic Storage Systems to the

Minimum, volume 29 of Advances in Cryogenic Engineering, pages 767–776. Plenum

Press, 1984.

[20] D.E. Hill and R.J. Salvinski. A thermodynamic system for zero-g venting, storage,

and transfer of cryogenic propellants. J. Spacecraft and Rockets, 4(7):955–957, 1967.

[21] R.P. Warren and J.W. Anderson. A System for Venting a Propellant Tank in the

Absense of Gravity, volume 12 of Advances in Cryogenic Engineering, pages 63–71.

Plenum Press, 1967.

[22] M.W. Liggett. Space-based LH2 propellant storage system: Subscale ground testing

results. Cryogenics, 33(4):438–442, 1993.

[23] T. Lak and C. Wood. Zero g thermodynamic vent system final report. Report no.

SSD 94M0038, Rockwell Aerospace, 1994.

[24] R.H. Flachbart, L.J. Hastings, and J.J. Martin. Testing of a spray bar zero gravity

cryogenic vent system for upper stages. AIAA 99-2175, 1999.

[25] R.H. Flachbart, L.J. Hastings, A. Hedayat, S.L. Nelson, and S.P. Tucker. Testing

of a spray-bar thermodynamic vent system in liquid nitrogen. In Proceedings of the

2005 Cryogenic Engineering Conference, Keystone, CO, 2005.

207
[26] A. Hedayat, S.L. Nelson, R.H. Flachbart, and S.P. Tucker. Liquid nitrogen (oxygen

simulant) thermodynamic vent system test data analysis. In Proceedings of the 2005

Cryogenic Engineering Conference, Keystone, CO, 2005.

[27] L.J. Hastings, R.H. Flachbart, J.J. Martin, A. Hedayat, M. Fazah, T. Lak, H. Nguyen,

and J.W. Bailey. Spray bar zero-gravity vent system for on-orbit liquid hydrogen

storage. NASA TM 2003-212926, 2003.

[28] J.H. Stark and M.H. Blatt. Cryogenic zero-gravity prototype vent system. NASA

CR 98079, 1967.

[29] W.H. Sterbentz. Liquid propellant thermal conditioning system final report. NASA

CR-72365, 1968.

[30] T.J. Van Overbeke. Thermodynamic vent system test in a low earth orbit simulation.

AIAA 2004-3838, 2004.

[31] L.J. Hastings, S.P. Tucker, R.H. Flachbart, A. Hedayat, and S.L. Nelson. Marshall

space flight center in-space cryogenic fluid management program overview. AIAA

2005-3561, 2005.

[32] L.J. Hastings, D.W. Plachta, L. Salerno, and P. Kittel. An overview of NASA efforts

on zero boil off storage of cryogenic propellants. Cryogenics, 41:833–839, 2002.

[33] D.W. Plachta. Results of an advanced development zero boil-off cryogenic propellant

storage test. AIAA 2004-3837, 2004.

[34] D.W. Plachta. Hybrid thermal control testing of a cryogenic propellant tank. NASA

TM 1999-209389, 1999.
208
[35] A. Hedayat, L.J. Hastings, C. Bryant, and D.W. Plachta. Cryogenic propellant long-

term storage with zero boil off. In Proceedings of the 2001 Cryogenic Engineering

Conference, Madison, WI, 2001.

[36] L.J. Salerno and P. Kittel. Cryogenics and the human exploration of mars. Cryogenics,

39:381–388, 1999.

[37] D.W. Plachta and P. Kittel. An updated zero boil-off cryogenic propellant storage

analysis applied to upper stages or depots in an leo environment. NASA TM 2003-

211691, 2003.

[38] J.A. Clark. A Review of Pressurization, Statification, and Interfacial Phenomena,

pages 259–283. Int. Advances in Cryogenic Engineering. Plenum Press, 1965.

[39] D.J. Chato. The role of flight experiments in the development of cryogenic fluid

management technologies. NASA TM 2006-214261, 2006.

[40] R.H. Knoll, G.R. Smolak, and R.R. Nunamaker. Weightlessness experiments with

liquid hydrogen in aerobee sounding rockets; uniform radiant heat addition - flight 1.

NASA TM X-484, 1962.

[41] J.G. McArdle, R.C. Billon, and D.A. Altmos. Weightlessness experiments with liquid

hydrogen in aerobee sounding rockets; uniform radiant heat addition - flight 2. NASA

TM X-718, 1962.

[42] R.R. Nunamaker, E.L. Corpas, and J.G. McArdle. Weightlessness experiments with

liquid hydrogen in aerobee sounding rockets; uniform radiant heat addition - flight 3.

NASA TM X-872, 1963.

209
[43] J.C. Aydelott, E.L. Corpus, and R.P. Gruber. Comparison of pressure rise in a hydro-

gen dewar for homogeneous, normal-gravity, quiescent, and zero gravity conditions -

flight 9. NASA TM X-1052, 1965.

[44] K.L. Abdalla, T.C. Frysinger, and C.R. Androcchio. Pressure-rise characteristics for a

liquid hydryogen dewar for homogeneous, normal gravity, quiescent, and zero-gravity

tests. NASA TM X-1134, 1965.

[45] R.F. Lacovic, F.C. Yeh, Jr. S.V. Szabo, R.J. Brun, A.J. Stofan, and J.A. Berns.

Management of cryogenic propellants in a full scale orbiting space vehicle. NASA TN

D-4571, 1968.

[46] R.D. Bradshaw. Evaluation and application of data from low gravity orbital experi-

ment phase I - final report. NASA CR-109847, 1970.

[47] J. Navickas and R.A. Madsen. Propellant Behavior During Venting in an Orbiting

Saturn S-IVB Stage, volume 13 of Advances in Cryogenic Engineering, pages 188–198.

Plenum Press, 1968.

[48] R.K. Allgeier. Subcritical cryogenic storage development and flight test. NASA TN

D-4293, 1968.

[49] A.B. Yanke. Titan/Centaur flight evaluation TC-5. Technical Report Report

CASD/LVP-76-066, General Dynamics Convair, 1977.

[50] R. Lacovic. Propellant management report for the Titan/Centaur TC-5 extended

mission. NASA TM-73749, 1977.

210
[51] P.M. Ordin, S. Weiss, and H. Christenson. Pressure-Temperature Histories of Liquid

Hydrogen Under Pressurization and Venting, volume 5 of Advances in Cryogenic

Engineering, pages 481–486. Plenum Press, 1960.

[52] M.D. Bentz. Tank pressure control in low gravity by jet mixing. NASA CR 191012,

1993.

[53] R.N. Eberhardt, D.A. Fester, W.A. Johns, and J.S. Marino. An Experiment to

Evaluate Liquid Hydrogen Storage in Space, volume 27 of Advances in Cryogenic

Engineering, pages 1107–1116. Plenum Press, 1982.

[54] M.M. Hasan, C.S. Lin, R.H. Knoll, and M.D. Bentz. Tank Pressure Control Experi-

ment: Thermal phenomena in microgravity. NASA TP 3564, 1996.

[55] M.D. Bentz, J.M. Albayyari, R.H. Knoll, M.M. Hasan, and C.S. Lin. Tank Pressure

Control Experiment: Results of three space flights. AIAA 1997-2816, 1997.

[56] J.W. Rose. Condensation of a vapour in the presence of a non-condensing gas. Int.

J. Heat and Mass Transfer, 12:233–237, 1969.

[57] W.J. Minkowycz and E.M. Sparrow. Condensation heat transfer in the presence of

noncondensable, interfacial resistance, superheating, variable properties, and diffu-

sion. Int. J. Heat and Mass Transfer, 9:1125–1144, 1966.

[58] S.C. Fan, J.C. Chu, and L.E. Scott. Thermal Stratification in Closed Cryogenic

Containers, volume 14 of Advances in Cryogenic Engineering, pages 249–257. Plenum

Press, 1969.

211
[59] C. Beduz, R. Rebiai, and R.G. Scurlock. Thermal Overfill and the Surface

Vaporisation of Cryogenic Liquids Under Storage Conditions, volume 29 of Advances

in Cryogenic Engineering, pages 795–804. Plenum Press, 1984.

[60] R.T. Swim. Temperature Distribution in Liquid and Vapor Phases of Helium in

Cylindrical Dewars, volume 5 of Advances in Cryogenic Engineering, pages 498–504.

Plenum Press, 1960.

[61] J.W. Tatom, W.H. Brown, L.H. Knight, and E.F. Coxe. Analysis of Thermal

Stratification of Liquid Hydrogen in Rocket Propellant Tanks, volume 9 of Advances

in Cryogenic Engineering, pages 265–272. Plenum Press, 1964.

[62] B.D. Neff and C.W. Chiang. Free Convection in a Container of Cryogenic Fluid,

volume 12 of Advances in Cryogenic Engineering, pages 112–124. Plenum Press,

1967.

[63] T.M. Lovrich, S.H. Schwartz, and L.A. Holmes. Flow visualization of thermal strati-

fication with localized heat sources. J. Spacecraft and Rockets, 11(9):664–669, 1974.

[64] B.H. Anderson and M.J. Kolar. Experimental investigation of the behavior of a

confined fluid subjected to non uniform source and wall heating. NASA TM D-2079,

1963.

[65] S.P. Das, S. Chakraborty, and P. Dutta. Studies on thermal stratification phenomena

in LH2 storage vessel. Heat Transfer Engineering, 25(4):54–66, 2004.

212
[66] H.-C. Ji, S.H. Schwartz, T.N. Lovrich, J.I. Hochstein, and L.A. Holmes. Experimental

verification of scaling parameters for thermal stratification. J. Thermophysics and

Heat Transfer, 6(3):522–530, 1992.

[67] L. Manson. A Technique for the Simulation of Thermal Behavior of Fluids in a

Low Gravity Field, volume 10 of Advances in Cryogenic Engineering, pages 297–304.

Plenum Press, 1965.

[68] R. Neff. A Survey of Stratification in a Cryogenic Liquid, volume 5 of Advances in

Cryogenic Engineering, pages 460–466. Plenum Press, 1960.

[69] M.H. Bourgarel, M.P. Segel, and J.P. Huffenus. Study of Stratification Similtude

Laws in Liquid Hydrogen, volume 12 of Advances in Cryogenic Engineering, pages

103–111. Plenum Press, 1967.

[70] M.H. Blatt. Empirical correlations for pressure rise in closed cryogenic containers. J.

Spacecraft and Rockets, 5(6):733–735, 1968.

[71] L.E. Scott, R.F. Robbins, D.B. Mann, and B.W. Birmingham. Temperature stratifi-

cation in a non venting liquid helium dewar. J. Research NBS, 64C(1):19–23, 1960.

[72] R.J. Stochl and R.H. Knoll. Thermal performance of a liquid hydrogen tank multilayer

insulation system at warm boundary temperatures of 630, 530, and 152 R. NASA

TM 104476, 1991.

[73] M.M. Hasan, C.S. Lin, and N.T. Van Dresar. Self-pressurization of a flightweight

liquid hydrogen storage tank subjected to low heat flux. NASA TM 103804, 1991.

213
[74] N.T. Van Dresar, C.S. Lin, and M.M. Hasan. Self-pressurization of a flightweight

liquid hydrogen tank: Effects of fill level at low wall heat flux. NASA TM 105411,

1992.

[75] J.C. Aydelott. Self-pressurization of liquid hydrogen tankage. Master’s thesis, Cornell

U., 1967.

[76] J.C. Aydelott. Normal gravity self-pressurization of 9-inch (23 cm) diameter spherical

liquid hydrogen tankage. NASA TN D-4171, 1967.

[77] J.C. Aydelott and C.M. Spuckler. Effect of size on normal-gravity self-pressurization

of spherical liquid hydrogen tankage. NASA TN D-5196, 1969.

[78] J.C. Aydelott. Effect of gravity on self-pressurization of spherical liquid hydrogen

tankage. NASA TN D-4286, 1986.

[79] D.H. Liebenberg and F.J. Edeskuty. Pressurization Analysis of a Large-Scale Liquid

Hydrogen Dewar, pages 284–289. Int. Advances in Cryogenic Engineering. Plenum

Press, 1965.

[80] S.C. Huntley. Temperature-Pressure-Time Relationships in a Closed Cryogenic

Container, volume 3 of Advances in Cryogenic Engineering, pages 342–352. Plenum

Press, 1960.

[81] B.R. Bullard. Liquid propellant thermal conditioning system test program final re-

port. NASA CR-72971, 1972.

[82] S.M. Dominick. Mixing induced condensation inside propellant tanks. AIAA 84-0514,

1984.
214
[83] C.S. Lin, M.M. Hasan, and N.T. Van Dresar. Experimental investigation of jet-

induced mixing of a large hydrogen storage tank. NASA TM 106629, 1994.

[84] C.S. Lin, M.M. Hasan, and T.W. Nyland. Mixing and transient interface condensation

of a liquid hydrogen tank. NASA TM 106201, 1993.

[85] O.S. Jones, J.S. Meserole, and A. Fortini. Measurements of jet-induced pressure decay

in a thermally stratified tank. J. Spacecraft and Rockets, 31(2):290–296, 1994.

[86] J.S. Meserole, O.S. Jones, and A.F. Fortini. Mixing-induced fluid destratification and

ullage condensation. AIAA 1987-2018, 1987.

[87] O.S. Jones, J.S. Meserole, and M.D. Bentz. Correlation of ullage condensation rate

with mixing intensity in propellant tanks. AIAA 1991-2543, 1991.

[88] K.J. McNaughton and C.G. Sinclair. Submerged jets in short cylindrical flow vessels.

J. Fluid Mechanics, 25(2):367–375, 1966.

[89] J.C. Mollendorf and B. Gebhart. An experimental and numerical study of the vis-

cous stability of a round laminar vertical jet with and without thermal buoyancy for

symmetric and asymmetric disturbances. J. Fluid Mechanics, 61(2):367–399, 1973.

[90] J.C. Mollendorf and B. Gebhart. Thermal buoyancy in round laminar vertical jets.

Int. J. Heat and Mass Transfer, 16:735–745, 1973.

[91] E.P. Symons and T.L. Labus. Experimental investigation of an axisymmetric fully

developed laminar free jet. NASA TN D-6304, 1971.

215
 [92] T.L. Labus and E.P. Symons. Experimental investigation of an axisymmetric free jet

with an initially uniform velocity profile. NASA TN D-6783, 1972.

[93] E.P. Symons and J.V. Staskus. Interface stability during liquid inflow to partially

full, hemispherical ended cylinders during weightlessness. NASA TM X-2348, 1971.

[94] S.G. Berenyi, R.C. Nussle, and K.L. Abdalla. An experimental investigation of the

effect of gravity on a forced circulation pattern in spherical tanks. NASA TN D-4409,

1968.

[95] J.C. Aydelott. Axial jet mixing of ethanol in cylindrical containers during weight-

lessness. NASA TP 1487, 1979.

[96] J.C. Aydelott. Axial jet mixing of ethanol in spherical containers during weightless-

ness. NASA TM X-3380, 1976.

[97] E.A. Fox and V.E. Gex. Single phase blending of liquids. AIChE J., 2(4):539–544,

1956.

[98] H. Fossett and L.E. Prosser. The application of free jets to the mixing of fluids in

bulk. Proc. Inst. of Mech. Engineers, 160(2):224–232, 1949.

[99] N. Okita and Y. Oyama. Mixing characteristics in jet mixing. Jap. Chem. Eng.,

1(1):94–101, 1963.

[100] A.G.C. Lane and P. Rice. An investigation of liquid jet mixing employing an inclined

side entry mixer. Trans. IChemE, 60:171–176, 1982.

216
[101] L.J. Poth and J.R. Van Hook. Control of the thermodynamic state of space-stored

cryogens by jet mixing. J. Spacecraft and Rockets, 9(5):332–336, 1972.

[102] L.J. Poth, J.R. Van Hook, D.M. Wheeler, and C.R. Kee. A study of cryogenic

propellant mixing techniques final report vol. I: Mixer design and experimental in-

vestigations. NASA CR 73908, 1968.

[103] I.H. Lehrer. A new model for free turbulent jets of miscible fluids of different density

and jet mixing time criterion. Trans. IChemE, 59:247–252, 1981.

[104] R.K. Grenville and J.N. Tilton. A new theory improves the correlation of blend time

data from turbulent jet mixed vessels. Trans. IChemE Part A: Chem. Eng. Res.

Design, 74:390–396, 1996.

[105] A.W. Patwardhan and S.G. Gaikwad. Mixing in tanks agitated by jets. Trans.

IChemE Part A: Chem. Eng. Res. Design, 81:211–220, 2003.

[106] B.K. Revill. Jet mixing. In N. Harnby, M.F. Edwards, and A.W. Nienow, editors,

Mixing in the Process Industries. Butterworth Heinemann, 2nd edition, 1992.

[107] M.R. Helmick, B.C. Khoo, and A.A. Sonin. Vapor condensation on a turbulent liquid

interface. In Cryogenic Fluid Management Technology Workshop Vol. I: Presentation

Material and Discussion, NASA CP-10001, pages 301–315, 1986.

[108] P.D. Thomas and F.H. Morse. Analytic Solution for the Phase Change in a Suddenly

Pressurized Liquid-Vapor System, volume 8 of Advances in Cryogenic Engineering,

pages 550–562. Plenum Press, 1963.

217
[109] J.-H. Chun, M.A. Shimko, and A.A. Sonin. Vapor condensation onto a turbulent

liquid II: Condensation burst instability at high turbulence intensities. Cryogenics,

29(9):1333–1338, 1986.

[110] A.A. Sonin, M.A. Shimko, and J.-H. Chun. Vapor condensation onto a turbulent

liquid I: The steady condensation rate as a function of liquid-side turbulence. Int. J.

Heat and Mass Transfer, 29(9):1319–1332, 1986.

[111] J.S. Brown, B.C. Khoo, and A.A. Sonin. Rate correlation for condensation of pure

vapor on turbulent, subcooled liquid. Int. J. Heat and Mass Transfer, 33(9):2001–

2018, 1990.

[112] J.S. Brown. Vapor Condensation on Turbulent Liquid. PhD thesis, MIT, 1991.

[113] D.J. Chato, J. Marchetta, J.I. Hochstein, and M. Kassemi. Approaches to validation

of models for low gravity fluid behavior. NASA TM 2005-213832, 2005.

[114] C.S. Lin, N.T. Van Dresar, and M.M. Hasan. Pressure control analysis of cryogenic

storage systems. J. Propulsion and Power, 20(3):480–485, 2004.

[115] C.K. Forester. Non-equilibrium Storage and Expulsion of Single-Phase Cryogens,

volume 12 of Advances in Cryogenic Engineering, pages 82–91. Plenum Press, 1967.

[116] D. H. Riemer. Cryogenic Tank Stratification: A Simpler Approach, volume 31 of

Advances in Cryogenic Engineering, pages 957–962. Plenum Press, 1986.

[117] C.H. Panzarella and M. Kassemi. On the validity of purely thermodynamic descrip-

tions of two-phase cryogenic fluid storage. J. Fluid Mechanics, 484:41–68, 2003.

218
[118] Y.S. Cha, R.C. Niemann, and J.R. Hull. Thermodynamic analysis of helium boil-off

experiments with pressure variations. Cryogenics, 33:675–679, 1993.

[119] Y. Rotenberg. Numerical Simulation of Self-Pressurization in a Small Cryogenic Tank,

volume 31 of Advances in Cryogenic Engineering, pages 963–972. Plenum Press, 1986.

[120] Z. Li, L. Xu, H. Sun, Y. Xiao, and J. Zhang. Investigation on performances of non-loss

storage for cryogenic liquefied gas. Cryogenics, 44:357–362, 2004.

[121] P.N. Estey, D.H. Lewis, and M. Connor. Prediction of a propellant tank pressure

history using state space methods. J. Spacecraft and Rockets, 20(1):49–54, 1983.

[122] M. Epstein and H.K. Georgius. A Generalized Propellant Tank-Pressurization

Analysis, pages 290–302. Int. Advances in Cryogenic Engineering. Plenum Press,

1965.

[123] P. Schallhorn, D.M. Campbell, S. Chase, J. Piquero, C. Fortenberry, X. Li, and

L. Grob. Upper stage tank thermodynamic modeling using SINDA/FLUINT. AIAA

2006-5051, 2006.

[124] A. Hedayat, L.J. Hastings, J.W. Bailey, R.H. Flachbart, and K.A. Holt. Thermo-

dynamic venting system modeling and comparison with liquid hydrogen test data.

AIAA 2003-4450, 2003.

[125] H. Nguyen. Zero-g thermodynamic venting system (TVS) performance prediction

program. Technical report, Rockwell Aerospace, 1994.

[126] E.L. Knuth. Non stationary phase changes involving a condensed phase and a satu-

rated vapor. Physics of Fluids, 2(1):84–86, 1959.

219
[127] E.L. Knuth. Evaporation and condensation in one-component systems. ARS Journal,

32:1424–1426, 1962.

[128] A.F. Schmidt, J.R. Purcell, W.A. Wilson, and R.V. Smith. An Experimental Study

Concerning the Pressurization and Stratification of Liquid Hydrogen, volume 5 of

Advances in Cryogenic Engineering, pages 487–497. Plenum Press, 1960.

[129] M.P. Segel. Experimental Study of the Phenomena of Stratification and

Pressurization of Liquid Hydrogen, pages 308–313. Int. Advances in Cryogenic Engi-

neering. Plenum Press, 1965.

[130] T. Bailey, R. Vandekoppel, G. Skartvedt, and T. Jefferson. Cryogenic propellant

stratification analysis and test data correlation. AIAA J., 1:1657–1659, 1963.

[131] T.E. Bailey and R.F. Fearn. Analytical and Experimental Determination of

Liquid-Hydrogen Temperature Stratification, volume 9 of Advances in Cryogenic

Engineering, pages 254–264. Plenum Press, 1964.

[132] D.M. Tellep and E.Y. Harper. Approximate analysis of propellant stratification.

AIAA J., 1(8):1954–1956, 1963.

[133] J.M. Ruder. Stratification in a pressurized container with side wall heating. AIAA

J., 2:135–137, 1964.

[134] J.H. Robbins and Jr. A.C. Rogers. An analysis on predicting thermal stratification

in liquid hydrogen. J. Spacecraft and Rockets, 3:40–45, 1966.

[135] G.C. Vliet. Stratification with bottom heating. J. Spacecraft and Rockets, 3(7):1142–

1144, 1966.
220
[136] D.O. Barnett. Liquid Nitrogen Stratification Analysis and Experiments in a Partially

Filled Spherical Container, volume 13 of Advances in Cryogenic Engineering, pages

174–187. Plenum Press, 1968.

[137] D.R. Kirk, J. Oliveira, and P.A. Schallhorn. Cryogenic propellant stratification in a

rotating and reduced gravity environment. AIAA 2007-954, 2007.

[138] D.O. Barnett, T.W. Winstead, and L.S. McReynolds. An Investigation of

Liquid-Hydrogen Stratification in a Large Cylindrical Tank of the Saturn

Configuration, pages 314–324. Int. Advances in Cryogenic Engineering. Plenum Press,

1965.

[139] R.W. Arnett and D.R. Millhiser. A theoretical model for predicting thermal strat-

ification and self pressurization of a fluid container. In Proc. of the Conference on

Propellant Tank Pressurization and Stratification, NASA Marshall Space Flight Cen-

ter, 1965.

[140] R.W. Arnett and R.O. Voth. A computer program for the calculation of thermal

stratification and self-pressurization in a liquid hydrogen tank. NASA CR 2026,

1972.

[141] S. Venkat and S.A. Sherif. Self-pressurization and thermal stratification in a liquid

hydrogen tank under varying gravity conditions. AIAA 2004-1341, 2004.

[142] S. Gursu, S.A. Sherif, T.N. Veziroglu, and J.W. Sheffield. Analysis and optimization of

thermal statification and self-pressurization effects in liquid hydrogen storage systems

part I: Model development. J. Energy Resources Technology, 115:221–227, 1993.

221
[143] S. Gursu, S.A. Sherif, T.N. Veziroglu, and J.W. Sheffield. Analysis and optimiza-

tion of thermal stratification and self-pressurization effects in liquid hydrogen storage

systems part II: Model results and conclusions. J. Energy Resources Technology,

115:228–231, 1993.

[144] S.A. Nikitin and V.I. Polezhaev. Convection and heat transfer in a spherical vessel,

partly filled with liquid, under low-gravity conditions. Izv. Akad. Navk SSSR, Mekh.

Zhidk. Gaza, 2:154–159, 1976.

[145] S.G. Cherkasov. Natural convection in a vertical cylindrical vessel with heat supplied

to its side and free surfaces. Izv. Akad. Navk SSSR, Mekh. Ahidk. Gaza, 6:51–56,

1984.

[146] C.S. Lin and M.M. Hasan. Numerical investigation of the thermal statification in

cryogenic tanks subjected to wall heat flux. NASA TM 103194, 1990.

[147] A. Sengupta. Comparison of computed to measured liquid hydrogen stratification.

In Proc. of the 2001 ASME Fluids Engineering Division Summer Meeting, pages

629–636, 2001.

[148] Z. Tanyun, H. Zhongping, and S. Li. Numerical Simulation of Thermal Stratification

in Liquid Hydrogen, volume 41 of Advances in Cryogenic Engineering, pages 155–161.

Plenum Press, 1996.

[149] H.Z. Barakat and J.A. Clark. Transient natural convection in closed containers.

Technical report, Heat Transfer Laboratory, U. of Michigan, 1965. Tech Report No.

2.

222
[150] J. Navickas. Prediction of a liquid tank thermal stratification by a finite difference

computing method. AIAA 88-2917, 1988.

[151] G.D. Grayson. Coupled thermodynamic-fluid-dynamic solution for a liquid-hydrogen

tank. J. Spacecraft and Rockets, 32(5):918–921, 1995.

[152] G.D. Grayson and J. Navickas. Interaction betweeen fluid-dynamic and thermody-

namic phenomena in a cryogenic upper stage. AIAA 93-2753, 1993.

[153] M.M. Hasan and C.S. Lin. Axisymmetric confined turbulent jet directed towards the

liquid surface from below. NASA TM 101409, 1989.

[154] C.S. Lin and M.M. Hasan. Effect of Liquid Surface Turbulent Motion on the Vapor

Condensation in a Mixing Tank, pages 1526–1537. Transport Phenomena in Heat

and Mass Transfer. Elsevier, 1992.

[155] J.I. Hochstein, P.M. Gerhart, and J.C. Aydelott. Computational modeling of jet

induced mixing of cryogenic propellants in low g. NASA TM 83703, 1984.

[156] J.I. Hochstein, H.-C. Ji, and J.C. Aydelott. Temperature fields due to jet induced

mixing in a typical OTV tank. AIAA 87-2017, 1987.

[157] M.C. Wendl, J.I. Hochstein, and G.P. Sasmal. Modeling of jet-induced geyser forma-

tion in a reduced gravity environment. AIAA 91-0803, 1991.

[158] R.J. Thornton and J.I. Hochstein. Microgravity geyser and flow field prediction.

AIAA 2000-0858, 2000.

223
[159] B.D. Simmons, J.G. Marchetta, and J.I. Hochstein. Reduced gravity cryogenic pro-

pellant tank re-supply simulation and geyser prediction. AIAA 2005-1150, 2005.

[160] J.G. Marchetta, C.E. Kain, B. Simmons, and R.H. Benedetti. Simulation and predic-

tion of jet induced geyser formation in microgravity propellant tanks. AIAA 2006-937,

2006.

[161] J.G. Marchetta and R.H. Benedetti. Three dimensional modeling of jet-induced gey-

sers in low gravity. AIAA 2007-955, 2007.

[162] F.R. Menter. Zonal two-equation k-w turbulence models for aerodynamic flows. AIAA

93-2906, 1993.

[163] D.J. Chato and D.A. Jacqmin. Modeling the restraint of liquid jets by surface tension

in microgravity. AIAA 2001-0931, 2001.

[164] D. Jacqmin. Calculation of two-phase Navier Stokes flows using phase-field modeling.

J. Computational Physics, 155:96–127, 1999.

[165] D.J. Chato. Influence of turbulence on the restraint of liquid jets by surface. AIAA

2002-0758, 2002.

[166] D.J. Chato. Parametric investigation of liquid jets in low gravity. AIAA 2003-997,

2003.

[167] S.K. Mukka and M.M. Rahman. Analysis of fluid flow and heat transfer in a liq-

uid hydrogen storage vessel for space applications. In M.S. El-Genk, editor, Space

Technology and Applications International Forum - STAIF 2004, volume 699, pages

37–44, Albuquerque, NM, 2004. AIP Conference Proceedings.

224
[168] S.H. Ho and M.M. Rahman. Three-dimensional analysis of liquid hydrogen cryogenic

storage tank. AIAA 2005-5712, 2005.

[169] C.S. Lin and M.M. Hasan. Self-pressurization of a spherical liquid hydrogen storage

tank in a microgravity environment. NASA TM 105372, 1992.

[170] J.I. Hochstein, H.-C. Ji, and J.C. Aydelott. Effect of subcooling on the on-orbit

pressurization rate of cryogenic propellant tankage. AIAA 86-1253, 1986.

[171] J.I. Hochstein, H.-C. Ji, and J.C. Aydelott. Prediction of self-pressurization rate of

cryogenic propellant tankage. J. Propulsion and Power, 6(1):11–17, 1990.

[172] J.I. Hochstein. Numerical modelling of cryogenic propellant behavior in low g. In

Microgravity Fluid Management Symposium, NASA CP-2465, pages 85–100, 1986.

[173] G.D. Grayson, A. Lopez, F.O. Chandler, L.J. Hastings, and S.P. Tucker. Cryogenic

tank modeling for the Saturn AS-203 experiment. AIAA 2006-5258, 2006.

[174] C.W. Hirt. Modeling phase change and homogeneous bubbles. FSI-01-TN57, Flow

Science Inc., 2001.

[175] H. Merte, J.A. Clark, and H.Z. Barakat. Finite difference solution of stratification

and pressure rise in containers. Technical report, Heat Transfer Laboratory, U. of

Michigan, 1968. Tech Report No. 4.

[176] H. Merte et al. Transient pressure rise of a liquid-vapor system in a closed container

under variable gravity. 4th Int. Heat Transfer Conference, Paris, France, 1970.

225
[177] Y.V. Val’tsiferov and V.I. Polezhaev. Convective heat transfer in a closed axisymmet-

ric vessel with curvilinear generatrix in the presence of phase boundaries and phase

transitions. Izv. Akad. Navk SSSR, Mekh. Ahidk. Gaza, 6:126–134, 1975.

[178] G. Tunc, H. Wagner, and Y. Bayazitoglu. Space shuttle upgrade liquid oxygen tank

thermal stratification. AIAA 2001-3082, 2001.

[179] N.V. Amirkhanyan and S.G. Cherkasov. Theoretical analysis and procedure for the

calculation of thermophysical processes occuring in a cryogenic vessel under condi-

tions of non-vented storage. High Temperature, 39(6):905–911, 2001.

[180] C.H. Panzarella and M. Kassemi. Self-pressurization of large spherical cryogenic tanks

in space. J. Spacecraft and Rockets, 42:299–308, 2005.

[181] J.M. Albayyari. Cryogenic fuel tanks pressure reduction: A low-g fluid mixing exper-

iment. Int. J. Fluid Mechanics Research, 29(2):135–145, 2002.

[182] C.S. Lin. Numerical studies of the effects of jet-induced mixing on liquid-vapor in-

terface condensation. NASA TM 182285, 1989.

[183] C.S. Lin and M.M. Hasan. Vapor condensation on liquid surface due to laminar

jet-induced mixing: The effects of system parameters. NASA TM 102433, 1990.

[184] M.M. Hasan and C.S. Lin. Buoyancy effects on the vapor condensation rate on a

horizontal liquid surface. NASA TM 102437, 1990.

[185] C. Panzarella, D. Plachta, and M. Kassemi. Pressure control of large cryogenic tanks

in microgravity. Cryogenics, 44:475–483, 2004.

226
[186] M.J. Assael, S. Botsios, K. Gialov, and I.N. Metaxa. Thermal conductivity of poly-

methyl methacrylate (pmma) and borosilicate crown glass bk7. Int. J. Thermophysics,

26(5):1595–1605, 2005.

[187] O. Almanza, M.A. Rodrigues-Perez, and J.A. de Saja. Measurement of the thermal

diffusivity and specific heat capacity of polyethylene foams using the transient plane

source technique. Polymer International, 53:2038–2044, 2004.

[188] D. Bedeaux, A.M. Albano, and P. Mazur. Boundary conditions and non-equilibrium

thermodynamics. Physica, 82A:438–462, 1976.

[189] G. Marchionni, P. Maccone, and G. Pezzin. Thermodynamic and other physical

properties of several hydrofluoro-compounds. J. Flourine Chemistry, 118(1-2):149–

155, 2002.

[190] A. Huang and D.D. Joseph. Instability of the equilibrium of a liquid below its vapor

between heated horizontal plates. J. Fluid Mech., 242:235–247, 1992.

[191] P.N. Shankar and M.D. Deshpande. On the temperature distribution in liquid-vapor

phase change between plane liquid surfaces. Phys. Fluids A, 2(6):1030–1038, 1990.

[192] K. Ramamurthi, S.S. Kumar, and B.S. Chaitanya. Interface conditions governing

evaporation of stored liquids in the presence of non-condensable gas. Int. J. Heat and

Mass Transfer, 49:3583–3594, 2006.

[193] K. Hisatake, S. Tanaka, and Y. Aizawa. Evaporation rate of water in a vessel. J.

Appl. Phys., 73(11):7395–7401, 1993.

227
[194] G. Fang and C.A. Ward. Temperature measured close to the interface of an evapo-

rating liquid. Phys. Rev. E, 59(1):417–428, 1999.

[195] C. A. Ward and D. Stanga. Interfacial conditions during evaporation or condensation

of water. Phys. Rev. E, 64:051509, 2001.

[196] F. Duan, V.K. Badam, F. Durst, and C.A. Ward. Thermocapillary transport of

energy during water evaporation. Phys. Rev. E, 72:056303, 2005.

[197] F. Duan and C. A. Ward. Surface-thermal capacity of d2o from measurements made

during steady-state evaporation. Phys. Rev. E, 72:056304, 2005.

[198] A.J.H. McGaughey and C.A. Ward. Temperature discontinuity at the surface of an

evaporating droplet. J. App. Physics, 91(10):6406–6415, 2002.

[199] C. A. Ward and G. Fang. Expression for predicting liquid evaporation flux: Statistical

rate theory approach. Phys. Rev. E, 59(1):429–440, 1999.

[200] G. Fang and C.A. Ward. Examination of the statistical rate theory expression for

liquid evaporation rates. Phys. Rev. E, 59(1):441–453, 1999.

[201] M. Bond. Non-equilibrium evaporation and condensation. Master’s thesis, U. of

Victoria, 2004.

[202] D. Juric and G. Tryggvason. Computations of boiling flows. Int. J. Multiphase Flow,

24(3):387–410, 1998.

228
[203] S. Kjelstrup and D. Bedeaux. Heat and mass transfer across phase boundaries: Es-

timates of coupling coefficients. In Physical, Mathematical, and Natural Sciences,

volume 86, 2005.

[204] J. Kovac. Non-equilibrium thermodynamics of interfacial systems. Physica, 86A:1–24,

1977.

[205] D. Bedeaux and S. Kjelstrup. Transfer coefficients for evaporation. Physica A,

270:413–426, 1999.

[206] W.J. Bornhorst and G.N. Hatsopoulos. Analysis of liquid vapor phase change by the

methods of irreversible thermodynamics. J. Appl. Mech., pages 840–846, 1967.

[207] D. Bedeaux, L.J.F. Hermans, and T. Ytrehus. Slow evaporation and condensation.

Physica A, 169:263–280, 1990.

[208] R. W. Schrage. A Theoretical Study of Interphase Mass Transfer. Columbia Univer-

sity Press, 1953.

[209] J.R. Maa. Evaporation coefficient of liquids. I & EC Fundamentals, 6(4):504–518,

1967.

[210] I.W. Eames, N.J. Marr, and H. Sabir. The evaporation coefficient of water: A review.

Int. J. Heat and Mass Transfer, 40(12):2963–2973, 1997.

[211] E.J. Davis. A history and state of the art of accomodation coefficients. Atmospheric

Research, 82:561–578, 1986.

[212] B. Paul. Compilation of evaporation coefficients. ARS Journal, 32:1321–1328, 1962.

229
[213] R. Marek and J. Straub. Analysis of the evaporation coefficient and the condensation

coefficient of water. Int. J. Heat and Mass Transfer, 44:39–53, 2001.

[214] I. Tanasawa. Advances in Condensation Heat Transfer, volume 21 of Advances in

Heat Transfer, pages 55–137. Academic Press, 1991.

[215] J. Barrett and C. Clement. Kinetic evaporation and condensation rates and their

coefficients. J. Colloid and Interface Sci., 150(2):352–363, 1992.

[216] C.H. Panzarella. Nonlinear Analysis of Horizontal Film Boiling. PhD thesis, North-

western University, 1998.

[217] T. Ytrehus. Asymmetries in evaporation and condensation kundsen layer problems.

Phys. of Fluids, 26(4):939–949, 1983.

[218] L.D. Koffman, M.S. Plesset, and L. Lees. Theory of evaporation and condensation.

Phys. of Fluids, 27(4):876–880, 1984.

[219] D.A. Labuntsov and A.P. Kryukov. Analysis of intensive evaporation and condensa-

tion. Int. J. Heat and Mass Transfer, 22:989–1002, 1979.

[220] Y.-P. Pao. Temperature and density jumps in the kinetic theory of gases and vapors.

Phys. of Fluids, 14(1):1340–1346, 1971.

[221] Y. Onishi and Y. Sone. Kinetic theory of evaporation and condensation for a cylin-

drical condensed phase. Phys. of Fluids, 26(3):659–664, 1983.

[222] Y. Sone and Y. Onishi. Kinetic theory of evaporation and condensation. J. Phys.

Soc. Japan, 35(6):1773–1776, 1973.

230
[223] A.V. Gusarov and I. Smurov. Gas-dynamic boundary conditions of evaporation and

condensation: Numerical analysis of the knudsen layer. Phys. of Fluids, 14(12):4242–

4255, 2002.

[224] J. B. Young. The condensation and evaporation of liquid droplets in a pure vapour at

arbitrary knudsen number. Int. J. Heat and Mass Transfer, 34(7):1649–1661, 1991.

[225] R. Meland, A. Frezzotti, T. Ytrehus, and B. Hafskjold. Nonequilibrium molecular-

dynamics simulation of net evaporation and net condensation, and evaluation of the

gas-kinetic boundary condition at the interphase. Phys. of Fluids, 16(2):223–243,

2004.

[226] T. Ishiyama, T. Yano, and S. Fujikawa. Molecular dynamics study of kinetic boundary

condition at an interface between argon vapor and its condensed phase. Phys. of

Fluids, 16(8):2899–2906, 2004.

[227] M. Matsumoto, K. Yasuoka, and Y. Kataoka. Molecular simulation of evaporation

and condensation. Fluid Phase Equilibria, 104:431–439, 1995.

[228] J. Xu, S. Kjelstrup, D. Bedeaux, A. Rosjorde, and L. Rekvig. Verification of onsager’s

reciprocal relations for evaporation and condensation using non-equilibrium molecular

dynamics. J. Colloid Interface Sci., 299:452–463, 2006.

[229] A. Rosjorde, S. Kjelstrup, D. Bedeaux, and B. Hafskjold. Nonequilibrium molecular

dynamics simulations of steady-state heat and mass transport in condensation. ii.

transfer coefficients. J. Colloid and Interface Sci., 240:355–364, 2001.

231
[230] S. I. Sandler. Quantum mechanics: A new tool for engineering thermodynamics.

Fluid Phase Equilibria, 210:147–160, 2003.

[231] C. Geuzaine and J.-F. Remacle. Gmsh: a three-dimensional finite element mesh

generator with built-in pre- and post-processing facilities. Int. J. Numerical Methods

in Engineering, 79(11):1309–1331, 2009.

[232] T. J. R. Hughes. The Finite Element Method. Dover, 2000.

[233] L. J. Segerlind. Applied Finite Element Analysis. Wiley, 1984.

[234] J.D. Means and R.D. Ulrich. Transient convective heat transfer during and after gas

injection into containers. J. Heat Transfer, 97.

[235] V. Geller. Thermodynamic properties of 3M mixtures. Thermophysics research cen-

ter, 1997.

[236] K.S. Pitzer. J. Am. Chem. Soc., 77:3427, 1955.

[237] C. Tsonopoulos. AIChE J., 20:263, 1974.

[238] C. Tsonopoulos. AIChE J., 21:827, 1975.

[239] C. Tsonopoulos. AIChE J., 24:1112, 1978.

[240] H. Orbey and J.H. Vera. AIChE, (29):107, 1983.

[241] K. Lucas. Phase equilibria and fluid properties in the chemical industry. Dechema,

page 573, 1980.

[242] B.E. Poling, J.M. Prausnitz, and J.P. O’Connell. The Properties of Gases and Liquids.

232