T HE C ITY C OLLEGE OF N EW Y ORK

D OCTORAL T HESIS

The effects of geometric, flow, and

boiling parameters on bubble growth
and behavior in subcooled flow boiling

Author: Supervisor:

Randy Samaroo Dr. Masahiro Kawaji

A thesis submitted in fulfillment of the requirements

for the degree of Doctor of Philosophy

in the

Department of Mechanical Engineering

June 17, 2016





ProQuest Number: 10159915




All rights reserved

INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.






ProQuest 10159915

Published by ProQuest LLC (2016 ). Copyright of the Dissertation is held by the Author.


All rights reserved.
This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.


ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
 2016
c
Randy Samaroo
All Rights Reserved

ii
This manuscript has been read and accepted for the Graduate Faculty in Engineering in satisfaction of
the dissertation requirement for the degree of Doctor of Philosophy

A PP-c, 7,
Masahiro Kawaji, Chair of Examining Committee Date

fiZe---1

Ardie D. i-Assrotiale-ijean for Academic Affairs Date

EXAMINING COMMITTEE

Prof. Masahiro Kawaji, mentor, Dept. of Mechanical Engineering, The City College of New York

Prof. Yiannis Andreopoulos, Dept. of Mechanical Engineering, The City College of New York

Prof. Taehun Lee, Dept. of Mechanical Engineering, The City College of New York

Prof. Jing Fan, Dept. of Mechanical Engineering, The City College of New York

Dr. Matteo Bucci, Dept. of Nuclear Science and Engineering, Massachusetts Institute of Technology

THE CITY COLLEGE OF THE CITY UNIVERSITY OF NEW YORK

 THE CITY COLLEGE OF NEW YORK

Abstract
The effects of geometric, flow, and boiling parameters on bubble growth and
behavior in subcooled flow boiling

by Randy Samaroo

Dr. Masahiro Kawaji

Department of Mechanical Engineering

Air bubble injection and subcooled flow boiling experiments have been performed to
investigate the liquid flow field and bubble nucleation, growth, and departure, in part
to contribute to the DOE Nuclear HUB project, Consortium for Advanced Simulation
of Light Water Reactors (CASL). The main objective was to obtain quantitative data
and compartmentalize the many different interconnected aspects of the boiling process
- from the channel geometry, to liquid and gas interactions, to underlying heat transfer
mechanisms.

The air bubble injection experiments were performed in annular and rectangular ge-
ometries and yielded data on bubble formation and departure from a small hole on
the inner tube surface, subsequent motion and deformation of the detached bubbles,
and interactions with laminar or turbulent water flow. Instantaneous and ensemble-
average liquid velocity profiles have been obtained using a Particle Image Velocimetry
technique and a high speed video camera. Reynolds numbers for these works ranged
from 1,300 to 7,700.

Boiling experiments have been performed with subcooled water at atmospheric pres-
sure in the same annular channel geometry as the air injection experiments. A second
flow loop with a slightly larger annular channel was constructed to perform further
boiling experiments at elevated pressures up to 10 bar. High speed video and PIV
measurements of turbulent velocity profiles in the presence of small vapor bubbles
on the heated rod are presented. The liquid Reynolds number for this set of experi-
ments ranged from 5,460 to 86,000. It was observed that as the vapor bubbles are very
small compared to the injected air bubbles, further experiments were performed using
a microscopic objective to obtain higher spatial resolution for velocity fields near the
heated wall. Multiple correlations for the bubble liftoff diameter, liftoff time and bub-
ble history number were evaluated against a number of experimental datasets from
previous works, resulting in a new proposed correlations that account for fluid prop-
erties that vary with pressure, heat flux, and variations in geometry.

v
For my mother and father, who pushed me to do my best
and for Hema, who encouraged me both in success and in
failure

vi
 Acknowledgments
This thesis is the culmination of years of work, seizing of opportunities, and humble
acceptance of assistance from others. I give my deepest gratitude to those who helped
make my work possible and provided me with support and encouragement through-
out the process.

First and foremost, a special thanks to my mentor, Professor Masahiro Kawaji, who
gave me this opportunity, guided me in my research and enabled me to develop a
deeper understanding of the subject. During my time here, he contributed to an ex-
cellent research environment, permitted me to make mistakes, and encouraged me
to learn from them. Additionally, I thank my committee members, Professor Taehun
Lee, Professor Yiannis Andreopoulos, Professor Jing Fan, and Dr. Matteo Bucci for
their interest in my work.

Every result described in this thesis was achieved with assistance from fellow lab-
mates. I would like to thank those who assisted in the beginning phases of this work,
in particular Saman Reshadi and Joseph Kreynin, as their efforts helped pave the way
for my efforts. I offer further regards and thanks to all members of the lab who have
contributed in any small way to my project and helped to shape the effort into what is
has become.

I am grateful to the funding from the CASL project that allowed me to pursue graduate
studies, as well as funding from other projects that helped keep the lab functioning and
well-stocked with tools and equipment.

Finally, I would like to acknowledge friends and family who supported me in my time
at CCNY. I offer great to thanks Mom, Dad, Sean, and Nadia for their constant love
and support. I also offer my sincerest thanks to Freddy Hernandez-Alvarado, Dinesh
Kalaga and Jorge Pulido, who shared some of my greatest and worst moments, and

vii
made me a better person overall. I especially thank Hema for all her love, companion-
ship, and encouragement.

viii
Contents

Page

Abstract iv

Acknowledgments vii

Contents ix

List of Figures xiii

List of Tables xix

List of Symbols xxi

List of Dimensionless Numbers xxiii

1 Introduction 2
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Applications of Flow Boiling and Two Phase Flow . . . . . . . . . 4
1.2 Objectives and Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . 10

2 Review of Previous Works 14

2.1 Historical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Contemporary Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.1 Experimental Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.2 Numerical Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

ix
3 Theoretical Background 33
3.1 Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.1 Single-Phase Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.1.2 Two-Phase Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.1 Single-phase flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.2 Two-phase flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Existing Correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.1 Wall Temperature and Heat Flux . . . . . . . . . . . . . . . . . . . 55
3.3.2 Bubble liftoff diameter . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.3 Bubble behavior after liftoff . . . . . . . . . . . . . . . . . . . . . . 61

4 Experimental Apparatuses and Cases 64

4.1 Experiments and Geometries Treated . . . . . . . . . . . . . . . . . . . . . 64
4.1.1 Air Bubble Injection in Annular and Rectangular Geometries . . 65
4.1.2 Subcooled Flow Boiling at Ambient and Elevated Pressures . . . 69
4.2 Operational Limits of Experiments . . . . . . . . . . . . . . . . . . . . . . 73
4.2.1 Air Injection Flow Loop . . . . . . . . . . . . . . . . . . . . . . . . 73
4.2.2 Atmospheric Pressure Boiling Loop . . . . . . . . . . . . . . . . . 74
4.2.3 Elevated Pressure Boiling Loop . . . . . . . . . . . . . . . . . . . . 75
4.3 Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.1 High Speed Video . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3.2 Particle Image Velocimetry . . . . . . . . . . . . . . . . . . . . . . 79
4.3.3 Temperature Measurements . . . . . . . . . . . . . . . . . . . . . . 81
4.4 Experimental Conditions Investigated . . . . . . . . . . . . . . . . . . . . 81
4.4.1 Relevant scales for modeling . . . . . . . . . . . . . . . . . . . . . 84

x
 4.4.2 Comparison of experimental conditions with previously pub-
lished data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.5 Post-Processing of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.5.1 Experimental Uncertainties . . . . . . . . . . . . . . . . . . . . . . 94

5 Results and Discussion 96

5.1 Single-Phase Velocity Profiles . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2 Air Bubble Injection Experiments . . . . . . . . . . . . . . . . . . . . . . . 99
5.2.1 Annular Test Section Data . . . . . . . . . . . . . . . . . . . . . . . 99
5.2.1.1 Effect of Air Bubble Injection on Liquid Velocity . . . . 99
5.2.1.2 Air Bubble Size, Shape and Position . . . . . . . . . . . . 103
5.2.2 Rectangular Test Section Data . . . . . . . . . . . . . . . . . . . . . 110
5.2.2.1 Effect of Air Bubble Injection on Liquid Velocity . . . . 110
5.2.2.2 Air Bubble Size, Shape and Position . . . . . . . . . . . . 115
5.3 Subcooled Flow Boiling Experiments . . . . . . . . . . . . . . . . . . . . . 119
5.3.1 Atmospheric Pressure Experiments . . . . . . . . . . . . . . . . . 120
5.3.1.1 Bubble Size, Shape and Behavior . . . . . . . . . . . . . 121
5.3.1.2 Effect of Bubble Growth on Liquid Flow Field . . . . . . 124
5.3.2 Elevated Pressures Experiments . . . . . . . . . . . . . . . . . . . 130
5.3.2.1 Bubble Liftoff Diameter . . . . . . . . . . . . . . . . . . . 131
5.3.2.2 Bubble Liftoff Time . . . . . . . . . . . . . . . . . . . . . 147
5.3.2.3 Bubble Behavior Before and After Liftoff . . . . . . . . . 154

6 Conclusions 169
6.1 Air Bubble Injection Experiments . . . . . . . . . . . . . . . . . . . . . . . 170
6.2 Subcooled Flow Boiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.3 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

xi
References 180

A Literature Review Comparison 192

B Experimental Procedures 199

B.1 Limitations of Experimental Techniques . . . . . . . . . . . . . . . . . . . 199
B.2 Calibration Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
B.2.1 Image Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
B.2.2 Flow Instrumentation Calibration . . . . . . . . . . . . . . . . . . 206
B.2.3 Thermocouple Calibration . . . . . . . . . . . . . . . . . . . . . . . 206
B.3 Flow Loop Operating Procedures . . . . . . . . . . . . . . . . . . . . . . . 209
B.4 Data Acquisition Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . 212

xii
List of Figures

Page

1.1 Annual electricity generation by fuel type, 2000-2040 (trillion kWh) . . . 5

1.2 Annual energy-related carbon dioxide emissions, 2000-2040 (million met-
ric tons) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Battery life and capacity variation with temperature . . . . . . . . . . . . 8
1.4 Current and projected energy consumption in the U.S. by end use . . . . 9

3.1 Upward flow in an annulus . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2 Flow configuration for rectangular channel . . . . . . . . . . . . . . . . . 37
3.3 Force balance on a growing bubble in the X- and Y-directions . . . . . . . 43
3.4 Schematic diagram of bubble nucleation phenomenon . . . . . . . . . . . 46
3.5 Diagram of convective heat transfer over a flat plate . . . . . . . . . . . . 49
3.6 Comparison of temperature distributions in pool and flow boiling cases 51
3.7 Evolution of flow boiling from ONB to DNB . . . . . . . . . . . . . . . . 52

4.1 Schematic diagram of the air injection flow loop . . . . . . . . . . . . . . 66

4.2 Rectangular test section configuration . . . . . . . . . . . . . . . . . . . . 68
4.3 Cross section of the rectangular test section . . . . . . . . . . . . . . . . . 68
4.4 Schematic diagram of the subcooled flow boiling loop . . . . . . . . . . . 70
4.5 Cross section of the subcooled flow boiling annular test section . . . . . 70
4.6 High pressure boiling flow loop schematic . . . . . . . . . . . . . . . . . 72

xiii
4.7 Comparison of High Speed Video and PIV exposure for air injection
experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.8 Laser configuration for PIV measurements . . . . . . . . . . . . . . . . . 80
4.9 Experimental conditions on the subcooling number versus Zuber num-
ber coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.10 Experimental conditions on the subcooling number versus Boiling num-
ber coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.11 Calculation of bubble size, shape, and position . . . . . . . . . . . . . . . 89

5.1 Single-phase time-averaged PIV results . . . . . . . . . . . . . . . . . . . 96

5.2 Comparison of single-phase measured and theoretical velocity profiles
(case A-01) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.3 Dimensionless axial velocities and fluctuations compared with the law
of the wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4 PIV measurements near the injection point (z = 0.07mm, case A-01) . . . 100
5.5 PIV measurements above the injection point (z = 3.98mm, case A-01) . . 100
5.6 Nondimensional velocity profiles near the injection point (z = 0.07mm,
case A-01) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.7 PIV measurements near the injection point (z = 0.02mm, case A-02) . . . 102
5.8 PIV measurements above the injection point (z = 4.40mm, case A-02) . . 102
5.9 Nondimensional axial velocity profile near the injection point (z = 0.02mm,
case A-02) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.10 Bubble coalescence after injection, case A-01 . . . . . . . . . . . . . . . . 104
5.11 Trajectory, size, and shape for coalescing bubbles, case A-01 . . . . . . . 105
5.12 Bubble motion after injection, case A-02 . . . . . . . . . . . . . . . . . . . 107
5.13 Ensemble averaged trajectory, size, and shape (10 bubbles, case A-02) . . 108

xiv
5.14 Instantaneous air injection rate estimated from the variation of the bub-
ble volume at the injection hole, case A-02 . . . . . . . . . . . . . . . . . . 109
5.15 Sample PIV exposure for rectangular geometry, case R-01 . . . . . . . . . 111
5.16 Ensemble-averaged PIV measurements near the injection point (z=0.03mm,
case R-01) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.17 Ensemble-averaged PIV measurements above the injection point (z=3.00mm,
case R-01) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.18 Ensemble-averaged PIV measurements above the injection point (z=3.00mm,
case R-02) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.19 Ensemble-averaged PIV measurements above the injection point (z=3.00mm,
case R-03) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.20 Ensemble-averaged trajectory, size, and shape (10 bubbles, case R-01) . . 116
5.21 Sample bubbles for cases with high confinement number, Co . . . . . . . 116
5.22 Comparison of experimental data correlations from Wellek and Okawa
et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.23 Comparison of developed correlations with experimental data . . . . . . 119
5.24 Sample images from HSV and PIV experiments . . . . . . . . . . . . . . . 120
5.25 Ensemble-averaged bubble size and aspect ratio during growth, (5 bub-
bles, case B-01) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.26 Ensemble-averaged entrained bubble size and trajectory, (10 bubbles,
case B-02) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.27 Ensemble-averaged PIV measurements at a bubble nucleation site ele-
vation, case B-03 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.28 Nondimensional axial velocity profile at a bubble nucleation site eleva-
tion, case B-03 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.29 Ensemble-averaged PIV measurements at a bubble nucleation site ele-
vation, case B-04 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

xv
5.30 Nondimensional axial velocity profile at the bubble nucleation site ele-
vation, case B-04 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.31 Ensemble-averaged PIV measurements at a bubble nucleation site ele-
vation, case B-06 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.32 Nondimensional axial velocity profile at a bubble nucleation site eleva-
tion, case B-06 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.33 Ensemble-averaged PIV measurements at a bubble nucleation site ele-
vation, case B-07 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.34 Nondimensional axial velocity profile at a bubble nucleation site eleva-
tion, case B-07 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.35 Nondimensional axial velocity profile at a bubble nucleation site eleva-
tion, case B-05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.36 Nondimensional axial velocity profile for a sliding bubble passing through
the measurement elevation, case B-08 . . . . . . . . . . . . . . . . . . . . 130
5.37 Timelapse of two sliding bubbles growing and departing (Case P05-01) . 132
5.38 Liftoff diameter versus time, grouped by case . . . . . . . . . . . . . . . . 133
5.39 Bubble aspect ratio at liftoff . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.40 Liftoff diameter versus the inverse of dimensionless wall temperature,
grouped by case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.41 Nondimensional liftoff diameter versus Reynolds number, grouped by
case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
5.42 Effect of bubble growth on momentum boundary layer . . . . . . . . . . 139
5.43 Comparison of Situ et al.’s correlation with present and previous exper-
imental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.44 Comparison of experimentally measured bubble liftoff data with exist-
ing correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.45 Results of proposed correlation for the nondimensional liftoff diameter . 146

xvi
5.46 Liftoff time versus Reynolds number, grouped by case . . . . . . . . . . . 148
5.47 Liftoff diameter versus Nusselt number, grouped by case . . . . . . . . . 149
5.48 Correlations for bubble growth time . . . . . . . . . . . . . . . . . . . . . 152
5.49 Proposed correlation for bubble liftoff time . . . . . . . . . . . . . . . . . 153
5.50 Nondimensionalized individual bubble diameter during growth . . . . 156
5.51 Individual bubble aspect ratio during growth . . . . . . . . . . . . . . . . 157
5.52 Nondimensionalized bubble motion behavior after liftoff . . . . . . . . . 160
5.53 Nondimensionalized bubble radius and motion behavior after liftoff . . 161
5.54 Bubble aspect ratio after liftoff . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.55 Comparison of experimentally measured bubble history data with ex-
isting correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.56 Results of proposed correlation for the bubble history number for col-
lapsing bubbles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.57 Results of proposed correlation for the bubble history number for grow-
ing bubbles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

B.1 Optical correction configurations investigated . . . . . . . . . . . . . . . 203

B.2 PIV calibration process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
B.3 Operating procedures for the air injection flow loop . . . . . . . . . . . . 209
B.4 Operating procedures for the atmospheric pressure boiling flow loop . . 210
B.5 Operating procedures for the elevated pressure boiling flow loop . . . . 211
B.6 Procedure for obtaining high speed video (HSV) or PIV data . . . . . . . 213
B.7 Sample high speed video image (P01-03: P = 1bar, Re = 55281, Ja =
4.2, N ux = 87.6) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
B.8 Change in time-averaged streamwise and spanwise velocity profiles
(Re=6555, air injection rate = 5.4mL min−1 ) . . . . . . . . . . . . . . . . . . 216

xvii
B.9 Percent change in RMS fluctuations in streamwise and spanwise veloc-
ity profiles (Re=6555, air injection rate = 5.4mL min−1 ) . . . . . . . . . . . 217

xviii
List of Tables

Page

3.1 Existing correlations for heat flux . . . . . . . . . . . . . . . . . . . . . . . 56

3.2 Existing correlations for bubble diameters . . . . . . . . . . . . . . . . . . 59
3.3 Existing correlations for bubble behavior . . . . . . . . . . . . . . . . . . 63

4.1 Operating limits for each experiment . . . . . . . . . . . . . . . . . . . . . 75

4.2 Experimental conditions investigated for air injection . . . . . . . . . . . 83
4.3 Experimental conditions investigated for subcooled flow boiling . . . . . 84
4.4 Kolmogorov scales for experimental cases . . . . . . . . . . . . . . . . . . 85
4.5 High speed video and PIV recording parameters and uncertainties . . . 95

5.1 Summary of results for air injection experiments . . . . . . . . . . . . . . 117

5.2 Average error values for each correlation and dataset (%) . . . . . . . . . 146

6.1 Effect of various flow boiling parameters on bubble size, shape and be-
havior (∼ indicates no appreciable change) . . . . . . . . . . . . . . . . . 177

A.1 Summary of fundamental literature reported for subcooled flow boiling 194
A.2 Summary of literature reported for measurements of flow field and two-
phase interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
A.3 Summary of literature reported for subcooled flow boiling at elevated
pressures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
A.4 Summary of literature reported for microlayer thickness measurement
or estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

xix
A.5 Summary of literature investigating CHF or dryout conditions . . . . . . 198

B.1 Thermocouple temperature constants and correction for current applied 208

xx
List of Symbols

Symbol Name Unit

Latin letters
A flow area m2
b empirical asphericity constant
Bo Bond number
Cl lift coefficient
Cp specific heat capacity kJ kg−1 K−1
d, D bubble diameter m or mm
Dh hydraulic diameter m
D+ dimensionless bubble diameter
Err error %
f friction factor
gc gravitational acceleration m s−2
G mass flux kg m−2 s−1
hx local heat transfer coefficient kW m−2 K−1
hf g Latent heat of vaporization kJ kg−1
Ja Jakob number
k thermal conductivity W m−1 K−1
L test section length m
Nb boiling number
NqG dimensionless heat flux
N ux local Nusselt number
Nz Zuber number
P pressure kPa or bar
Pr Prandtl number
qw00 wall heat flux kW m−2
r radial position m
r+ turbulent wall coordinate
Ra aspect ratio (=w/h)
Re liquid Reynolds number
t time s or ms
◦
T temperature C
v velocity m s−1
V voltage V
W weight for error calculation
x vapor quality
z axial bubble position m

xxi
Greek letters
α thermal diffusivity m2 s−1
Symbol Name Unit
(cont’d)
γ ratio of physical properties
 surface roughness m
θ nondimensional temperature
κ law of the wall coefficient
ν kinematic viscosity m2 s−1
ρ density kg m−3
σ surface tension N m−1
τ shear stress N m−2
ζh heated length m
Subscripts
avg average
b bubble
bulk bulk fluid
e effective
hr heater rod
in inlet
l liquid
lo lift-off
NW non-weighted
sat saturation
sub subcooling
TC thermocouple
v vapor
w wall
W weighted

xxii
List of Dimensionless Numbers

uDh
Reynolds number Re = ν

ρl vb2 Db
Weber number We = σ

(ρl −ρv )gDb2

Bond/Eötvös number Bo ≡ Eo = σ
q
σ
Confinement number Co = 2 g(ρ −ρ )
Dh l g

00 D
qw,x
hx Dh h
Local Nusselt number N ux = kl
= kl (Tw,x −Tbulk )

Cp,l (Tw −Tsat )

Jakob number Ja = hf g

00
qw
Boiling number Nb = Ghf g

∆hsub,in ∆ρ
Subcooling number Nsub = hf g ρv
= xin ∆ρ
ρv

00 ζ z
qw ∆ρ
Zuber number NZu = h T
Avl,in hf g ρv ρl
= (x(zT ) − xin ) ∆ρ
ρv

+ Dlo σ Dlo uτ
Non-dimensional Liftoff Diameter Dlo = ρl α2l
or ν

Db (t)
Bubble history number β(t) = Db,lo

tαl tαl
Fourier number Fo = d2b
or (uτ /ν)2

Tw −Tbulk ∆Tw +∆Tsub

Dimensionless wall temperature θ= Tw −Tsat
= ∆Tw

xxiii
Chapter 1

Introduction

Heat transfer to a liquid flowing across a heated surface is considered single phase
convection if both the wall temperature and the liquid temperature are below the sat-
uration temperature of the liquid. However, as the wall temperature begins to exceed
the local saturation temperature, boiling can occur, depending on a number of factors
- such as surface properties, and flow rate. When boiling occurs at the heated sur-
face, but the bulk liquid temperature is below saturation, this phenomenon is called
subcooled flow boiling [1].

1.1 Motivation

Subcooled flow boiling is a critical process for many different industrial applications
where enhanced heat transfer is required. Any components that generate high heat
fluxes during normal operation will likely utilize subcooled flow boiling as a cooling
method due to the high heat transfer coefficients in boiling processes. While heating a
single phase fluid requires a temperature gradient to occur, the very nature of heating
the fluid decreases the temperature difference over time, reducing the efficiency of the
process unless the temperature of the heat source increases. This difficulty is overcome
with phase change materials, where large amounts of energy can be transferred at

2
Chapter 1. Introduction 3

constant temperature difference, increasing the overall heat transfer coefficient.

Further increasing the efficiency of boiling processes creates additional layers of com-
plexity. Flow boiling, as opposed to pool boiling, permits significantly higher advec-
tion of thermal mass in the form of vapor bubbles away from the heated surface and
the constant replacement of heated fluid. As the heat generation increases and the flow
patterns change due to increased interactions between the liquid and vapor phases,
the boiling structure continually changes with increasing rate of bubble generation,
further increasing the heat transfer coefficient. If this trend is allowed to continue, the
system reaches a point of diminishing returns, in that increasing the heat generation
rate does not produce more bubbles, and the heat transfer coefficient suddenly de-
creases. Typically the cause of this phenomenon is an insulating vapor layer that sits
between the liquid and the heated surface. The heat flux at this point of diminishing
returns is called the Critical Heat Flux, or CHF. Understanding this phenomenon has
been the topic of several studies, and attempts have been made to artificially increase
the CHF by changing flow characteristics or the heater structure with promising re-
sults.

While enhancing CHF is a key overall goal of many researchers, it is highly depen-
dent on improving the understanding of the underlying hydrodynamics and conju-
gate heat transfer; the difficulty of this task cannot be understated. Single phase adia-
batic flows are well established and understood for simple geometries. Their behavior
can be predicted by the Navier-Stokes equations, which serve as a basis for all fluid
dynamics problems [2]. These complicated partial differential equations (PDEs) lack
a complete set of solutions and have been given both scientific and purely mathemat-
ical treatments for centuries. In fact, the Clay Mathematics Institute as part of their
Millennium Prize Problems has offered an award to anyone able to prove the exis-
tence and smoothness or lack thereof of pressure and velocity field solutions p(x, t)
4 Chapter 1. Introduction

and ui (x, t) to the Navier-Stokes equations [3]. Despite the lack of complete solutions
for single phase flows, computational models can be used to predict flow behavior
in virtually any geometry by discretizing the equations. As the fluid deviates from
this predicted behavior, as is the case in convective boiling, these equations must be
extended to address the complexities; here, dimensionless parameters play a signifi-
cant role in classifying flow regimes and behaviors. This effectively bridges the gap
between experimental work, numerical modeling, and theory.

To formulate improved closure relations with reduced empiricism, it is necessary to

improve the physical understanding of subcooled flow boiling phenomena such as
the nucleation, growth and detachment of vapor bubbles from a heated wall. Also
important are the bubble behavior in highly turbulent subcooled liquid flow includ-
ing the bubble deformation, break-up, coalescence and condensation, conjugate heat
transfer, and partition of energy transferred from the wall between liquid heating and
liquid-to-vapor phase change.

1.1.1 Applications of Flow Boiling and Two Phase Flow

While the aim of this work is to improve the understanding of the physics behind
subcooled flow boiling and two-phase flows, this information is of little value without
being used in commercial applications. The pressurization that occurs with boiling
processes allows for a very efficient storage of heat. Phase change materials provide
an adequate transfer mechanism between disorderly forms of energy produced by
combustion or volumetric heat generation and can be converted to harnessable energy
in the form of electricity generated by rotating machinery. Increasing the efficiency of
phase change heat transfer correlates directly to savings in energy costs for energy
sources that use heat transfer as an operating principle. Even when considering the
end uses of the electricity generated such as in powering electronics, as the size of
Chapter 1. Introduction 5

devices reduces, as is the present trend with mobile devices, the heat dissipated per
unit area increases significantly. Utilizing phase change heat transfer has been shown
to result in better performance in high energy density electronics due to the reduced
electrical resistance at lower temperatures. In the same vein as power generation,
phase change heat transfer is used in HVAC, refrigeration and cryogenic applications
to enhance the coefficient of performance (COP) of the thermodynamic cycles.

Power Generation As the population grows, the global energy demand is expected
to rise significantly. The anticipated demand for power must be met through the use
of increased energy production and increased efficiency in the generation and trans-
mission processes. Figure 1.1 shows the current and projected power generation in
the United States by source [4]. While the push for renewable energy technologies has
been prolific in recent years, their widespread adoption is not anticipated in the U.S.
energy market, likely due to the majority of the population being in locations that do
not lend themselves to renewable resources (i.e. areas with high sunlight intensity,
proximity to hills, rivers, etc.). With this consideration, the continued dependence of
the U.S. economy on steam generation is all but certain.

F IGURE 1.1: Annual electricity generation by fuel type, 2000-2040 (trillion

kWh)
6 Chapter 1. Introduction

F IGURE 1.2: Annual energy-related carbon dioxide emissions, 2000-2040

(million metric tons)

With this increased production of energy, there will also be a corresponding increase
in carbon emissions. Figure 1.2 shows the projected carbon emissions for various sce-
narios [4]; the ’Reference case’ corresponds to the projected source distribution shown
in Figure 1.1. Unfortunately, in the Reference case, overall carbon emissions are ex-
pected to increase by the year 2040, but note that the projections in the report do not
consider improvements in efficiency in power generation plants. By these projections,
there is a strong motivation to improve the efficacy of phase change heat transfer, as
a marginal increase in the efficiency of power generation will yield a reduction in fuel
consumption and carbon emissions, as well as any associated economical benefits.

A portion of the work performed as part of this thesis was funded by the Department
of Energy project, entitled the Consortium for Advanced Simulation of Light Water Reac-
tors (CASL) and led by Oak Ridge National Laboratory / UT-Battelle, LLC, through a
subcontract under contract DE-AC05-00OR22725. In the CASL project, subcooled flow
boiling is being investigated both experimentally and numerically in order to develop
constitutive relations for two-phase flow models to be used in predictive simulation
of Light Water Reactors. In the Modeling and Numerical Methods (MNM) focus area
of CASL, City College of New York had aimed at providing benchmark data on air
Chapter 1. Introduction 7

bubble injection and subcooled flow boiling for validating Interface Tracking Meth-
ods (ITM). The validated ITM simulation results will be further utilized in providing
high-fidelity closure relations for subchannel codes and 3-D Computational Fluid Dy-
namics (CFD) models of single-phase and two-phase flows in fuel rod bundles of a
PWR. Despite the specific nuclear applications that this work treats, the universality
of the physics is appreciated and understood, as the improved understanding of sub-
cooled flow boiling can be applied to all forms of steam generation plants.

Electronics Cooling Thermal dissipation has long been the limiting factor to how
compact electronics can become. Technology companies have been pushing these
limits by varying the materials used in devices and implementing innovative design
considerations such as incorporating device housings into heatsinks, a trend that has
turned metal housings from an aesthetic consideration to a functional necessity in high
end mobile devices. In time, the industry will again hit the physical limit for thermal
dissipation and will need to engineer new solutions to an ongoing problem, one of
which will involve the use of phase change materials.

In electronics, electrical energy is transformed into heat volumetrically, with most of

the heat being generated by switching devices such as MOSFETs, ICs, etc [5]. With this
consideration, most electronic devices have dedicated heatsinks to central/graphics
processing units (CPU/GPUs), which are comprised almost entirely of switching de-
vices and thus generate the most heat. While large scale heat generation is undesired,
it is necessary, as increased power consumption leads to better computational per-
formance. Even so, it is widespread knowledge that the number of transistors in a
dense integrated circuit doubles approximately every two years, commonly known
as Moore’s law [6], and this trend is expected to continue in the near future. For any
given device, the maximum power consumption, and thus performance, is limited to
8 Chapter 1. Introduction

how much power can be dissipated, PD,M ax :

TM ax − Ta
PD,M ax =
RT

where TM ax , Ta , and RT are the maximum allowable operating temperature of the

device, ambient temperature, and thermal resistance of the device, respectively.

Having an active cooling system permits the device to constantly replenish the ambi-
ent fluid to keep the temperature difference high. In the case of boiling heat transfer,
the temperature difference would be constant for the duration of the boiling process,
allowing the device to operate with higher performance for a longer duration.

A similar treatment can be given to battery technology, which is a rapidly growing

industry and critical in the development of portable devices. Whereas integrated cir-
cuit engineers have a singular aversion to heat generation, battery design is twofold in
that the electrochemical processes can be enhanced by increased temperature causing
increased capacity at the expense of the lifespan of the battery. This effect is shown in
Figure 1.3 [7].

F IGURE 1.3: Battery life and capacity variation with temperature

From Figure 1.3, it is clear that there is an optimal operating temperature for batter-
ies. However, in modern designs, a close proximity between the battery and high heat
Chapter 1. Introduction 9

dissipation components is inevitable, and there is a resulting feedback loop that af-
fects the overall performance of the device. Such a problem lends itself to a solution
that takes advantage of an efficient heat transfer mechanism, particularly one that can
maintain the device at a temperature that optimizes the battery performance.

Heat Exchangers for HVAC and Refrigeration (HVACR) Where two-phase flows
are used in power generation by moving heat from a hot source to a cooler one, they
can similarly be used to move heat in the reverse direction at the expense of energy. A
device that performs such a task is considered refrigerator, and working fluids used in
these thermodynamic cycles are called refrigerants. Such cycles are used for temper-
ature control systems like air conditioning and refrigeration units; while most homes
only have a few of these devices, they are the largest user of residential power and
nearly the largest user for commercial purpose. Figure 1.4 shows the breakdown of
current and projected U.S. residential and commercial energy intensities for various
end uses [4]. Although consumption for HVACR uses as a percentage of total con-
sumption is expected to decrease significantly, it is still likely to remain as one of the
primary end uses for energy in the U.S.

( A ) Residential ( B ) Commercial

F IGURE 1.4: Current and projected energy consumption in the U.S. by

end use
10 Chapter 1. Introduction

In current refrigeration cycle designs, serpentine coils are used to maximize the surface
area for heat transfer for boiling. The diameters and bend radii in such tubes are often
very small, creating localized boiling or condensation caused by bubble breakup or
oscillations, in addition to centrifugal forces that cause separation of liquid and vapor
phases [8]. The resulting flow patterns are highly complex with significant turbulence
transfer between phases. Improving the understanding of vapor bubble dynamics
under turbulent conditions would result in a substantial increase in HVACR cycle per-
formance.

The uses of convective boiling and multiphase flows are not limited to the applications
presented here. These cases demonstrate that there is a market need for improving the
efficiency of boiling heat transfer and that a better understanding of the physics can
help cut energy consumption and costs in the long term.

1.2 Objectives and Thesis Overview

To formulate improved closure relations with reduced empiricism, it is necessary to

improve our physical understanding of subcooled flow boiling phenomena such as
the nucleation, growth and detachment of vapor bubbles from a heated wall, bub-
ble behavior in highly turbulent subcooled liquid flow including the bubble deforma-
tion, break-up, coalescence and condensation, conjugate heat transfer, and partition
of energy transferred from the wall between liquid heating and liquid-to-vapor phase
change.

Appreciating that the benefits of improving the understanding of boiling phenomena

are significant, the primary objective of this work is to compartmentalize the many
different interconnected aspects of the boiling process - from the channel geome-
try, to liquid and gas interactions, to underlying heat transfer mechanisms. The initial
Chapter 1. Introduction 11

direction of the experimental work was guided by the benchmarks and milestones
associated with the CASL project, with a focus on obtaining liquid velocity field and
turbulence data in the vicinity of a single isolated air or vapor bubble. The data ob-
tained from these experiments would establish a foundation for future experiments
that extend beyond the CASL project.

The objectives of the experimental work expanded to investigate the effects of param-
eters outside of the original scope, including the effects of varying geometry, liquid
flow, and other boiling parameters on bubble size and behavior. Multiple experiments
were performed with a wide range of parameters varied: the growth and detachment
of a single air bubble in a laminar and turbulent water stream, and the growth and
detachment of a water vapor bubble under subcooled flow boiling conditions at var-
ious pressures. Both experiments are performed primarily using a vertical annular
test section, with additional experiments performed in rectangular channels. Bub-
ble growth and detachment is investigated using high-speed videography, whereas
the liquid flow and bubble-liquid interactions are measured using Particle Image Ve-
locimetry (PIV) techniques with additional micro-PIV measurements performed for
the subcooled flow boiling case. Further subcooled flow boiling experiments were per-
formed on a high pressure flow loop with a maximum pressure of 10 bar under various
flow rates and heat fluxes. For theses experiments, thermocouples were embedded on
the heater rod surface, allowing wall temperature measurements. By varying not only
the experimental conditions, but also the experimental techniques and the parameters
measured, the complex behavior of subcooled flow boiling and bubble behavior can
be broken down and made more manageable.

By broadening the scope of the experimental work, a more holistic understanding of

two-phase flows can be obtained, resulting in improved models that can better predict
boiling phenomena. To achieve this broader scope, the following secondary objectives
12 Chapter 1. Introduction

were established:

• Assess the current predictive models for boiling and identify limitations and key
parameters that often go overlooked but have strong physical significance

• Create a database of relevant experimental work that spans a wide range of con-
ditions with some overlap between data sets to allow for a level of redundancy
to ensure high fidelity results. Additionally, large gaps in the data, conditions
where experimental data is unavailable or incomplete, should be identified to be
investigated later

• Devise, construct and perform a series of experiments aimed at quantifying the

effects of each parameter while minimizing the impact of varying other param-
eters. Although it is very difficult to control each parameter individually due to
the physical coupling of heat transfer and fluid mechanics, multiple parameters
can be varied in such a way that analyzing groups of experiments allows for the
isolation of a single parameter

• Using the results of the experiments, propose new models that seek to better
predict boiling behavior, ensuring that it is also capable of predicting the range
of the existing experimental database. In the models, each aspect of the boiling
process should be considered, allowing for a more complete understanding

At the beginning of this thesis work, a rigorous literature review was performed, high-
lighting the current state of understanding of subcooled flow boiling phenomena.
From this review, an experimental database is compiled and relevant existing mod-
els are identified for further investigation. To fill in gaps in the experimental database,
the aforementioned experiments have been conducted using multiple flow loops. The
data from the experiments are added to the database, which is then used to evalu-
ate the predictive capabilities of the existing correlations. New correlations for bubble
growth, liftoff and behavior after departure are then developed using the experimental
Chapter 1. Introduction 13

database and the predictive capability is compared with the existing correlations using
each experimental dataset. The results are overall improved correlations for predicting
subcooled flow boiling and bubble behavior over a wide range of conditions. While
these results do not provide a complete understanding of boiling phenomena, they do
represent a marked improvement over existing knowledge.
Chapter 2

Review of Previous Works

It has been established that there is a need to improve the current physical under-
standing of convective boiling behavior. As such, the current state of research should
be reviewed as thoroughly as possible to provide a complete picture of what is known
about subcooled flow boiling, but more importantly what is unknown as of now. A
brief history of research in convective boiling is presented to illustrate the natural evo-
lution of experimental work in synchrony with technological development. In addi-
tion to the historical overview, a survey of contemporary works is also performed.
Useful information that can be extracted from the review includes previous exper-
imental conditions, measured parameters, but possibly the most important are pa-
rameters not measured. This helps to identify gaps in the data and shortcomings in
existing knowledge.

2.1 Historical Background

Prior to the 1930’s, work done on heat transfer to fluids was purely theoretically based,
due to the limitations of measurement techniques of the time. Despite the inability to
perform sophisticated experiments, significant progress was made in the field - specif-
ically in convection, which couples fluid dynamics and heat transfer. Of significance

14
Chapter 2. Review of Previous Works 15

for the purpose of fluid mechanics is the Blasius boundary layer solution, published in
1908 [9], which develops the theoretical formulation of the viscous boundary layer in
flow over a semi-infinite flat plate. This formulation is critical to the advancement of
the field of heat transfer due to the highly coupled nature of the temperature distribu-
tion to the velocity field. Pohlhaussen in 1921 [9], expanded on this breakthrough to
devise an analytic approach to predicting the two-dimensional, steady state tempera-
ture distribution in a flow over a flat plate. Building on the Blasius solution, Falkner
and Skan in 1931 [10,11] developed a generalized solution for two-dimensional bound-
ary layer flow around a variable angle wedge. While these solutions are specific to
flow over a flat geometry, they serve as a framework for future work that focuses on
heat convection in any geometry.

The first preliminary pool boiling investigations were performed in the 1930’s, spear-
headed by Max Jakob and W. Fritz [12–14] in Germany. Jakob is credited with the
development and advancement of heat transfer theory in application to phase change
systems. Their work dealt with photographic experiments that produced bubble mea-
surement data, including departure radius, contact angle, and departure frequency.
Due to the limitations at the time in photography and computational analysis, how-
ever, the correlations produced by this work were highly simplified and crude and
neglected key parameters necessary to capture the phenomena adequately [12, 14].
Among the missing were the surface temperature and heat flux, which are fundamen-
tal to the rate of heat transfer and thus the rate of bubble growth. At around the same
time, forced convective experiments were performed by Nukiyama [15] in Japan. This
work resulted in the development of the boiling curve, which relates the wall temper-
ature and wall heat flux, and identifies the different stages of boiling.

In the 1940’s, the infancy of nuclear technology facilitated the need for further research
in two-phase heat transfer. One of the biggest contributions to the field during this
16 Chapter 2. Review of Previous Works

time was the work done by Lockhart and Martinelli [16], which resulted in the devel-
opment of a correlation for the pressure drop in multiphase flows in channels. This
correlation serves as a reference point for newer correlations, and is still utilized for
comparison purposes today.

The study of two-phase flows expanded greatly in the 1950’s, with multiple researchers
performing experimental work to predict bubble behavior. Gunther in 1951 [17] per-
formed a photographic study to measure bubble departure size in two phase flows.
Egen in 1957 [18], performed experiments to predict void fractions in channels at high
heat flux and pressures. Griffith in 1958 [19], performed a photographic study to mea-
sure void fractions in flow channels under a variety of elevated pressures and heat
fluxes. His work also served to prove the existence of the Point of Net Vapor Gen-
eration. Novak Zuber, in his PhD thesis and subsequent work in the late 50’s and
60’s [20–22] primarily focused on macroscopic photographic studies of bubble behav-
ior, measuring bubble diameters and departure frequencies in pool boiling. His con-
tributions are significant in that they track the growth and collapse processes, and
give a complete picture of the entire lifespan of the bubble. During the same period,
Groothuis [23] investigated the enhancement of heat transfer in water/air mixtures,
and represented his results by a simple correlation between the Nusselt and Reynolds
numbers. On the opposite end of the spectrum, Gambill and Greene performed exper-
iments at Oak Ridge National Lab [24, 25] to investigate burnout processes, where the
heat transfer surface experiences permanent damage due to inadequate cooling.

The mid 1960’s marked the dawn of the commercial nuclear age, with a major push for
thermal hydraulics research to enhance the efficiency of boiling heat transfer. Research
done in this period continued the trend of macroscopic measurements, with much
of the pioneering work taking time-averaged void fraction measurements. Various
techniques were utilized for this task, including gamma densitometry and differential
Chapter 2. Review of Previous Works 17

pressure transducers. Among the researchers involved in utilizing these techniques

in their experiments were Mauer in 1960 [26], Christensen in 1961 [27], Ferell in 1964
[28], Rouhani in 1965 [29], Bartolemei, Costa, St. Pierre and Whittle in 1967 [30–33],
Koumoutsos and Staub in 1968 [34, 35], and Evangelisti in 1969 [36]. Their work set
a foundation for future researchers to utilize similar techniques [37–51], including the
thermal hydraulics laboratory at CCNY.

In addition to the new experimental techniques being pioneered in this era, new mod-
els for microscale heat transfer emerged. In 1962, Hsu [52] was the first to postulate
the criterion for boiling inception; he stated that in order for a bubble to grow, the
surrounding liquid must exceed the saturation temperature inside the bubble. This
criterion was then applied to calculate the wall temperature using the bubble diam-
eter. Bergles and Rohsenow [53] built on Hsu’s work and proposed a correlation for
the heat flux at the onset of nucleate boiling based on wall temperature and system
pressure. Then in 1966, Chen [54] published his correlation that extended previous
models of heat transfer in pool boiling to convective boiling. This correlation was a
breakthrough in the modeling of heat transfer in convective boiling, breaking down
the problem into macro and microscales to determine the wall temperature, and is
still widely used today. Snyder [55] had previously postulated that a liquid micro-
layer forms between a vapor bubble and and the heater surface. Cooper and Lloyd in
1969 [56] verified this hypothesis, and based on their study multiple experiments have
been performed to investigate the effect of the microlayer on heat transfer enhance-
ment and measure the thickness over time.

In the 1970’s, researchers built on the techniques pioneered in the previous decade,
expanding their experiments to elevated pressures to more closely emulate the con-
ditions in commercial nuclear reactors. The key among these researchers was Ünal
[57, 58], who performed subcooled flow boiling experiments at very high pressures
18 Chapter 2. Review of Previous Works

using a liquid sodium heated pipe. Using a high speed photography technique, Ünal
measured entrained bubbles through a glass tube mounted at the end of the pipe, re-
sulting in one of the first correlations for bubble diameter at high pressures up to 17.7
bar.

Further experimental work was performed in the 1980’s, aiming to provide a complete
picture of boiling phenomena. Advances in high speed videography and storage me-
dia allowed multiple parameters to be measured simultaneously with greater resolu-
tion. Hino and Ueda [59, 60] performed a photographic study at near CHF conditions,
determining the point of net vapor generation, measuring the liquid microlayer and
the bubble frequency by analyzing the fluctuations in the wall temperature. Similarly,
Del Valle and Kenning [61] performed experiments at near CHF conditions, investigat-
ing bubble size, frequency and nucleation site density. These works set the foundation
for future CHF experiments and the understanding of the relation between heat flux
and boiling behavior.

At the same time, laser techniques for flow field measurements were being pioneered.
Among these techniques are laser doppler velocimetry (LDV), speckle velocimetry,
and its successor particle image velocimetry (PIV) [62]. The PIV technique was largely
developed [63–66] to improve the measurement of 2-D fluid velocity fields. The tech-
nique has since been enhanced with the use of additional equipment and processing
algorithms to provide localized 3-D velocity fields and capture a wide range of flow
behavior in varying geometries [67–70]. With these advances, multiple studies have
been performed to analyze the flow patterns in convective boiling. Along with experi-
ments that analyze CHF conditions and measurements at the onset of nucleate boiling,
a better understanding of subcooled flow boiling can be achieved.

The half century of work described above illustrates how far the boundaries of under-
standing of boiling phenomena have been pushed. From theoretical formulations of
Chapter 2. Review of Previous Works 19

single phase liquid flows bounded to the simplest of geometries, to the simultaneous
measurement of microlayer thickness and active nucleation sites at near CHF, tech-
nical advances have facilitated the ability to measure quantities that did not exist in
scientific work a century ago.

2.2 Contemporary Research

With the advances made in experimental techniques, the understanding of convec-

tive boiling has improved greatly. The theories of Hsu [52] and Snyder [55] have been
confirmed, and work continues on developing new theories and expanding the knowl-
edge base of the subject. Increased accessibility of scientific journals in academia per-
mits close comparisons of datasets and evaluation of proposed correlations by other
researchers; it also allows researchers to identify gaps in experimental works and oper-
ating conditions where the published data is outdated, incomplete, or missing. While
much work has been done on quantifying boiling parameters, the works rarely capture
the complete physics of the problem, often due to an inability to make certain mea-
surements. To complement experimental measurements, or to reinforce their findings,
efforts have been placed on performing numerical modeling or computational fluid
dynamics (CFD) simulations. For convective boiling simulations, preference is typ-
ically given to interface tracking, volume of fluid, level set and lattice-Boltzmann s,
which benefit from an ability to deal with complex boundaries such as bubble inter-
faces at lower computational cost.

The literature review that follows categorizes the previous experimental and numeri-
cal works by the operating pressure and the parameters measured. The vast majority
of recent research investigates convective boiling at ambient pressures, with particular
focus on bubble measurements at the moment of bubble departure/liftoff. Additional
20 Chapter 2. Review of Previous Works

experiments have been performed to quantify liquid-vapor and vapor-vapor interac-

tions, and analyze behavior at near or post-CHF conditions.

2.2.1 Experimental Work

The historical background provides some of the underlying details of the progress in
convective boiling research, however, most of the publications that align with topics
of this work are contemporary. Although the list is far from complete, it is important
to establish the parameters to be measured experimentally and establish the operating
conditions to build on the experimental database while providing some redundancy
for validation. Select publications that are closely related to this work are described in
detail, while others are summarized in Appendix A.

Fundamental experiments In many previous studies, subcooled flow boiling has

been investigated experimentally to obtain a better understanding of general boiling
behavior [61, 71–88]. Parameters investigated include macroscale measurements such
as bubble nucleation site density and departure frequency, with additional work focus-
ing on calculating the microlayer thickness from wall temperature response. Overall,
these experiments are performed at atmospheric pressure. For comparison purposes,
these publications and their findings are detailed in Table A.1 in Appendix A.

Klausner et al. [73] observed sliding bubbles in a horizontal square channel of dimen-
sions 25×25mm2 , using refrigerant R113 as the working fluid. A high speed photog-
raphy system was employed to measure bubble sliding distance and departure di-
ameters. Klausner et al. also calculated probability density functions for departure
diameter, using 200 bubble diameter measurements to construct each function. Exper-
imental data was used to evaluate an existing correlation from Mikic et al. [89] for the
bubble radius with good agreement.
Chapter 2. Review of Previous Works 21

Thorncroft et al. [75] conducted an experiment investigating bubble size, growth rate,
and departure frequency for vertical upflow, downflow and pool boiling in a square
duct of sides 12.7mm ID with one side heated, and using FC-87 as the working fluid.
The experiments were conducted at atmospheric pressure, with a mass flux range of
190-666kg m−2 s, heat flux ranging between 1.3-14.6kW m−2 , and inlet liquid subcool-
ing between 1.0-5.0◦C. The experiments provided a general understanding of vapor
bubble behavior in vertical flow boiling, identifying the key differences between up-
flow, downflow, and pool boiling dynamics.

Basu et al. [77] conducted subcooled flow boiling experiments with vertical upflow
over a flat copper surface and a nine rod (zircaloy-4) bundle using water as the work-
ing fluid. Wall temperatures in the rod bundle were interpolated from thermocouple
readings on the inner surfaces of the tubes, used as heated rods. The experiments
were performed at atmospheric pressure, mass fluxes ranging from 124-886kg m−2 s,
heat flux of 1.6-96.3kW m−2 , and subcooling between 1.7-52.5◦C. The resulting wall
superheats were 3-22◦C. The location of ONB during the experiments was determined
from visual observations as well as from the thermocouple readings. Correlations for
nucleation site density were developed that are independent of mass flux and liquid
subcooling, but dependent on contact angle and wall superheat.

Situ et al. [78–81] performed experiments under vertical upflow in an annulus of

38.1mm OD and 19.1mm ID with water as the working fluid, investigating the bub-
ble size, growth rate, and departure frequency. Situ et al. [78] also investigated bubble
and liquid velocities. The experimental data were taken at atmospheric pressure and
mass flux of 466-900kg m−2 s, heat flux of 54-206kW m−2 and inlet temperature of 80.0-
98.5◦C. The results of Situ et al. [78] were the asymptotic limit to bubble frequency and
that the bubble frequency increases with increasing Zuber number, which is a measure
of the change in vapor quality between the inlet and the measurement position. When
22 Chapter 2. Review of Previous Works

the subcooling number is greater than the Zuber number, Nsub > NZu , the system is
considered to be in the subcooled region; if Nsub ≤ NZu , the system is considered to be
in the saturated region [78, 90]. Situ et al.’s experiments led to an empirical correlation
proposed for the non-dimensional bubble liftoff diameter [80] and bubble departure
frequency [81].

Buongiorno et al. [85] performed flow boiling experiments in a vertically oriented rect-
angular test section of dimensions 100mm×30mm, using water as the working fluid.
Liquid pressure varied between 1.05-1.5 bar. The experiments involved using high-
speed video and infrared thermography simultaneously to record the bubble dimen-
sions and motion, as well as the temperature response at the heated wall. The heater
material was a sapphire plate coated with a thin Indium Tin Oxide layer. Reynolds
numbers ranged between 0-1×105 , wall heat flux ranged between 0-2MW m−2 , and liq-
uid subcooling ranged between 0-75◦C. Due to limitations of the high-speed infrared
camera, the maximum capture rate was 1000 Hz and the maximum spatial resolution
was 90µm.

Flow field and liquid-vapor interactions Experiments were also performed that pro-
vided velocity fields and bubble motion behavior under subcooled flow boiling and
bubbly flows [91–102]. These experiments demonstrated the coupling of bubble be-
havior, the liquid flow field and the temperature field. For better comparison of the
measured quantities and the experimental conditions, these publications are summa-
rized in Table A.2

Warrier et al. [92] measured the collapse rates of bubbles and using an energy balance,
formulated a model for the interfacial heat transfer and associated thickening of the
thermal boundary layer. In their experiments, water flowed upward through a near
square channel of area 16.33cm2 at atmospheric pressure. Mass fluxes were between
Chapter 2. Review of Previous Works 23

235-684kg m−2 s, inlet subcooling ranged from 7-46.5◦C and heat flux was between 9.6-
96.3kW m−2 . A thermocouple array connected to micrometers measured the liquid
temperature field, and a high speed video technique was used to measure the con-
tact angle and bubble collapse rates after departure. The bubble history number and
condensation Nusselt numbers were compared with existing correlations, and new
correlations were proposed.

Okawa et al. [94] performed air injection and pool boiling experiments in heated stag-
nant water to measure the effects of temperature on the bubble rise characteristics.
Bubbles were injected using three tubes of diameters 0.1, 0.5, and 1.0mm at room tem-
perature as well as near boiling temperature. The liquid subcooling at the elevated
temperature ranged from 0◦C (pool boiling of saturated liquid) to 12.6◦C. Their work
features multiple validations of existing correlations, as well as the development of a
new correlation for the amplitude of the oscillatory motion of rising bubbles.

Estrada Perez and Hassan [97] conducted two-phase PTV experiments in a vertical
rectangular channel with a heated wall at elevated pressures, using refrigerant HFE-
301 as the working fluid. The pressure was varied from 1-16 bar, Reynolds numbers
ranged from 3309-16549, wall heat flux ranged from 0-64kW m−2 , and the channel
cross-sectional area was 8.70mm×7.6mm. The data from these experiments detailed
the effect of wall heat flux on various turbulent parameters, showing that the velocity
profile deviated from the theoretical law of the wall with increasing heat fluxes; this
deviation was more pronounced at lower Reynolds numbers.

Kim and Park [98] measured the interfacial heat transfer during condensation of bub-
bles by placing a microthermocouple probe in the vicinity of collapsing bubbles. Wa-
ter flowed upward through a rectangular channel of cross section 25mm×15mm and
length 1000mm. A 600mm heated section was situated 100mm above the inlet, and a
teflon coating was applied to the heated surface, except for a 3mm area in the center of
24 Chapter 2. Review of Previous Works

the heated section. A microthermocouple was mounted to a micrometer, similar to the

technique of Warrier et al. [92], and a stereoscopic high speed video system was used
to obtain orthogonal views of the bubble. The operating pressures were slightly above
atmospheric, inlet temperatures were between 78-88◦C, the mass flux was 85kg m−2 s
in all cases, and the heat flux ranged from 60-122kW m−2 . Existing correlations for the
condensation Nusselt number and bubble history are evaluated, and new models are
proposed and compared with experimental data.

A publication based on this thesis work [102] recently reported on the turbulent ve-
locity profiles in an annular geometry under adiabatic air injection into a liquid flow
and subcooled flow boiling at atmospheric pressure, using water as the working fluid.
Bubble size, shape and motion were recorded using high speed video, and a PIV tech-
nique was used for velocity field measurements. In all experiments the annular gap
width was 4.9mm, with air being injected through a 400µm hole in the inner rod at a
rate of 20.6mL min−1 ; for subcooled flow boiling experiments, a cartridge heater was
inserted into the inner rod to produce a single vapor bubble. The liquid Reynolds
numbers were 1625 and 6555 for the air injection experiments, and 18000 for the sub-
cooled flow boiling experiments. For the subcooled flow boiling experiments, the heat
flux applied was 100kW m−2 , and the liquid subcooling was 9◦C. It was shown that the
air bubbles produced at lower liquid Reynolds numbers were larger and experienced
more dramatic oscillations in shape, and that the velocity field remained relatively
unaffected outside of the region around a single 100µm vapor bubble.

Elevated pressure Very limited data is available for subcooled flow boiling at ele-
vated pressures [50, 57, 90, 103–117]. While the experimental work that exists provides
some information about the bubble sizes and behaviors, key boiling parameters such
as the wall temperature are often overlooked, and resulting correlations that require
these parameters often make use of additional correlations to obtain these values.
Chapter 2. Review of Previous Works 25

Other authors [51, 118, 119] evaluated correlations, or made analogous comparisons
with experimental data obtained at elevated pressures. These publications are sum-
marized in Table A.3.

Ünal [57] developed a bubble model in which a bubble would grow on a partially
dried liquid film by evaporation of the film. This model was supplemented with ex-
perimental data taken photographically at elevated pressures of 13.9, 15.8, and 17.7MPa,
heat fluxes of 0.47-10.64MW m−2 , liquid velocities of 0.08-9.15m s−1 , and subcooling of
3-86◦C. In these experiments, a sapphire glass tube of 8mm ID and 20mm length was
connected to the end of a 10m sodium heated pipe. Previously published data points
were also used in the validation of his model. The model yielded correlations to pre-
dict the bubble growth rate, maximum bubble diameter and growth time.

Van der Hagen et al. [90] investigated stability characteristics in BWR conditions. Both
Type-I (hydrostatic head) and Type-II (density wave) stabilities were measured; it was
determined that the high subcooling led to a small Type-I stability margin, and a de-
crease in pressure decreased the Type-II stability margin. The stability criterion were
expressed in terms of the subcooling and Zuber numbers, with Type-I instability oc-
curring around the NZu -equals-Nsub line, and Type-II occurring at low pressure and
high power (high NZu , low Nsub ).

Prodanovic et al. [105] investigated subcooled flow boiling using water in an annular
test section with upward flow at pressures ranging from 1.05-3bar, bulk liquid ve-
locities from 0.08-0.8m s−1 , and subcooling from 10-30◦C. High speed photography
was used to capture the bubble behavior from inception to collapse, recording growth
and condensation rates. Their work resulted in correlations for nondimensionalized
parameters, including maximum bubble diameter, departure diameter, and growth,
departure, and condensation times. All correlations were of the same form, with only
the coefficients changing.
26 Chapter 2. Review of Previous Works

Okawa et al. [110] performed subcooled flow boiling experiments using water in a
circular tube geometry. A sapphire glass tube was partially coated with an ITO coat-
ing and heated sufficiently to activate very few nucleation sites. To record the bubble
growth and motion, a stereoscopic high speed video setup was used to provide bub-
ble size measurements in all three dimensions. The system pressure was slightly ele-
vated, ranging from 121-127kPa, with mass flux and liquid subcooling varying from
94-1435kg m−2 s and 2.2-10◦C, respectively. Their study resulted in a better understand-
ing of bubble rise characteristics after departure from a nucleation site, particularly the
limiting conditions for bubble sliding.

Euh et al. [111] measured the bubble departure frequency under vertical upflow in
an annulus of 38.1mm OD and 19.1mm ID with water as the working fluid. The ex-
periment was performed over a pressure range of 167-346kPa, with test conditions
of 214-1869kg m−2 s mass flux, 61-238kW m−2 heat flux, and 7.5-23.4◦C of subcooling.
Their experiments resulted in a modified version of Situ et al.’s [81] empirical equation
for bubble departure frequency at pressures beyond atmospheric.

Yuan et al. [112] performed subcooled flow boiling experiments using upward flow
in a rectangular test section of dimensions 2mm×50mm, using water as the working
fluid. The liquid pressure ranged between 0.1-1.0MPa, liquid subcooling was between
20-36◦C, and the heat flux was 50-348kW m−2 . The experiment focused on obtaining
visual data using a high speed camera. Parameters measured were maximum bubble
diameter, growth rate, sliding distance and sliding velocity with respect to the varying
pressure. Their work showed a gradual decrease in bubble maximum diameter with
increased pressure, and the increased latent heat required to produce similar bubbles
at elevated pressure resulted in a decrease in growth rate.

Chu et al. [119] evaluated a number of existing correlations for the maximum bub-
ble diameter to fit a number of data sets published by Situ et al. [80], Prodanovic et
Chapter 2. Review of Previous Works 27

al. [105], as well as their own experimental work. Chu et al. performed subcooled
flow boiling experiments using water in an annular flow channel with 11.125mm gap
width and 1,280mm long, of which 700mm was heated. Mass fluxes ranged from 300-
702kg m−2 s, heat fluxes were between 133.4-355.6kW m−2 , and resulting subcooling
was between 1.4-24◦C. Bubble sizes, growth times and nucleation site densities were
measured using a high speed imaging system. Chu et al. performed a rigorous com-
parison of existing models with experimental data sets which resulted in modification
of Prodranovic et al.’s model using Chen’s correlation [54] for the wall superheat.

Sugrue et al. [116] measured the effects of varying orientation angle, heat flux, mass
flux, and pressure on bubble departure diameters in subcooled flow boiling conditions
in a rectangular geometry. A wide range of flow conditions were explored: orientation
angles of 0◦ (horizontal), 30◦ , 45◦ , 60◦ , 90◦ (vertical), and 180◦ ; mass flux values of 250,
300, 350, and 400kg m−2 s; pressures of 101, 202, and 505kPa; subcooling of 10 and 20◦C;
and heat fluxes of 50 and 100kW m−2 . The ranges of Froude number was 0.42-1.06 and
the range of Reynolds numbers covered was 11800-34500. Their results showed that
bubble departure diameters would increase with increasing heat flux, decreasing mass
flux, decreasing subcooling and decreasing pressure.

Microlayer measurements While most of these works investigate bulk properties

of bubbles or liquid flows, other works have been undertaken to gain a better under-
standing of the liquid microlayer that forms between a growing bubble and the heated
surface. A key difficulty in measuring the microlayer is a lack of spatial and temporal
resolutions in measurement techniques due to the transient evaporation of the micro-
layer. Until recently, it has been impossible to measure the liquid microlayer directly,
and previous works have only estimated it as a function of the bubble size and growth
rate [120–128]. These works are described in Appendix A and summarized in Table
A.4.
28 Chapter 2. Review of Previous Works

CHF experiments In addition to the aforementioned works, multiple studies have

been performed to investigate the operational behavior at near or post-CHF conditions
[129–140]. These works are briefly described in Appendix A and experimental studies
are summarized in Table A.5.

As seen in the review above, most experiments have been conducted with water using
an electrically heated tube in the middle of an annular flow channel. In this work, one
air injection and two subcooled flow boiling loops have been constructed to obtain
highly resolved experimental data on the bubble nucleation, growth, departure and
motion, void fraction and velocity distributions near the heater rod.

2.2.2 Numerical Work

When coupled with experimental work, numerical models are a useful tool to predict
parameters that cannot be measured, or even to determine if an experiment is worth
doing. An added benefit of numerical simulations are that local, instantaneous values
are available for the entire domain. As computers become more powerful, the ability of
CFD models to accurately predict fluid behavior improves, and the spatial resolution
has increased to the point where modeling of individual atoms is possible.

It has been mentioned that the lattice-Boltzmann method (LBM) is prevalent in con-
vective boiling simulations. In the method, the fluid is considered to consist of imagi-
nary particles which move and collide within the discrete mesh. This particle behavior
places the LBM technique somewhere between molecular dynamics (MD) modeling,
which tracks particles on an atomistic scale, and CFD, which treats the fluids as a con-
tinuum and solves discretized PDEs to produce a solution. LBM exhibits benefits of
both, giving it the ability to treat highly irregular deforming interfaces, but still main-
tain an average bulk flow. The papers reviewed here give a sense of the current state
of numerical modeling of boiling phenomena.
Chapter 2. Review of Previous Works 29

Házi et al. [141] compared LBM and CFD techniques, highlighting the advantages and
disadvantages of each as well as proposing some improvements that can be made.
Advantages of LBM over CFD include an enhanced stability, simple implementation
of the method and boundary conditions, the lack of need for an interface tracking
algorithm, and a linear scaling of performance with CPU number; however these are
limited to adiabatic applications. At the time, thermal models had low compatibility
with LBM, but later works improved on this.

Yu et al. [142] provided another rigorous description of the current state of lattice-
Boltzmann methods, identifying key differences and shortcomings between CFD and
LBM techniques. Among the topics discussed were force definitions, inlet boundary
conditions, solid boundaries, local mesh refinement, and multi-relaxation-time models
and their associated impacts on computational stability.

Inamuro [143] proposed a lattice-Boltzmann model for incompressible-viscous and

two-phase flows. The model is a lattice-kinetic scheme (LKS) that extends the LBM
with heat transfer capability, with the shortcoming that the viscosity is required to be
high. The method has two probability distribution functions, one which is an order
parameter to distinguish between phases, and the other which calculates the velocity
of the two-phase fluid. In preliminary calculations, density ratios of up to 1000 were
achieved.

Xu et al. [144] recently published a 3-D pseudo potential based lattice-Boltzmann

model with an improved forcing scheme to better handle multiphase flows. In val-
idations, the model was able to handle density ratios up to 700 while maintaining
variable surface tension. Additional simulations of droplet motion in a microchannel
allowed for detailed analysis of surface parameters such as surface tension, channel
wettability, and channel roughness.
30 Chapter 2. Review of Previous Works

The Boltzmann equation Although LBM was developed from the lattice-gas au-
tomata method [142], a more holistic derivation can arrive from the Boltzmann equa-
tion by discretizing it to a rectangular mesh, and applying steps for collisions and
streaming. The Boltzmann equation solves for a particle probability distribution func-
tion (PDF), f (~r, p~, t), where at time t particles have mass m, position ~r and momentum
p~ within the phase space d3~rd3 p~. With a force field, F~ applied on the element, by
Hamilton’s equations of motion, without collisions, the volume element d3~rd3 p~ would
remain constant. However collisions cause the particle probability density function to
change. Factoring this, the Boltzmann equation is a conservation equation that gov-
erns the PDF so that all the particles are accounted for, and it is given by equation
(2.1).

 
∂f p~ ∂f ∂f
= · ∇f + F~ · = (2.1)
∂t m ∂~p ∂t coll

The Boltzmann equation already offers a more simplified approach to modeling than
the typical Navier-Stokes equations, which are second order, nonlinear PDEs, whereas
the Boltzmann equation is first order and linear. However, some shortcomings of the
LBM approach are an inefficiency of solving steady state problems, and more complex
boundary conditions due an inability to explicitly apply no-slip conditions at walls.
This is overcome by applying a "bounce-back" boundary condition, in which the dis-
tribution at the boundary is equal and opposite to the incoming fluid [142].

A significant portion of numerical studies serve to improve the modeling of parti-

cle collisions, and the most prevalent of these works was performed by Bhatnagar,
Gross, and Krook [145], to arrive at the BGK relaxation term, which replaces the col-
lision term altogether. The shortcomings of the BGK relaxation term were identified
by d’Humieres [146] and Lallemand and Luo [147]. In summary, the lattice-Boltzmann
model with the BGK relaxation becomes unstable at high Reynolds numbers, and LBM
Chapter 2. Review of Previous Works 31

is memory intensive due to the storage of velocity distribution functions.

CFD Modeling Outside of LBM techiques, commercial and open-source CFD soft-
ware is also widely used to aid in the prediction of subcooled flow boiling ad two-
phase flow behavior [148, 149].

Chen et al. [148] performed CFD simulations of upward convective subcooled boiling
of refrigerant R-113 using CFX-10 along with an extended FORTRAN code for the wall
boiling and thermal phase change models. The simulations were validated against ex-
perimental work performed at Arizona State University [150, 151]. The resulting sim-
ulations predicted well the temperature profile at the measurement plane. Typically,
commercial CFD software as a standalone package does not perform adequately for
two-phase flows, and are usually expanded using user defined codes as seen in [148].

Langewisch and Buongiorno [149] performed an extensive computational study of

slug flow in capillaries using the open source software Gerris. A total of 125 dif-
ferent combinations of Capillary number and Reynolds number, Ca = [0.005, 2] and
Re = [0, 900] were simulated, and new correlation for film thickness, bubble velocity
and pressure drop were proposed. Although these simulation results were not vali-
dated by experimental data, a reasonable agreement with theoretical predictions was
achieved. Additionally, experimentally verifying this particular study would be very
difficult due to the overall small diameter and even smaller liquid film thickness.

Typically, numerical simulations serve multiple purposes; they can be used as an ac-
companiment for experimental results, or as a precursor for them. Experiments are
limited in that they can only provide data at discrete points, whereas simulations,
if performed correctly, can be used to interpolate the intermediate points, only lim-
ited by computational capacity. In the opposite case, experiments may be difficult or
impossible to perform, and numerical simulations are an invaluable tool to predict
32 Chapter 2. Review of Previous Works

behavior until a point where the experiments can be performed. Numerical models
are developed with a combination of theoretical and empirical formulation. However,
with regards to convective boiling, no complete theoretical basis exists, and numerical
models are developed in tandem with experimental correlations.
Chapter 3

Theoretical Background

3.1 Fluid Dynamics

Both static and moving fluids are studied in the field of fluid mechanics [152] to im-
prove the understanding of the underlying physics and better predict fluid behavior.
Researchers approach this task by performing experiments and numerical simulations.
As the fluid behavior becomes more complex, the ability to predict this behavior is cur-
tailed. This increased complexity of fluid flow behavior can be attributed to a number
of factors: increase in turbulence intensity, varying fluid properties due to temperature
variation, and of course, phase change phenomena.

At the foundation of fluid mechanics are three fundamental conservation laws: Con-
servation of mass (3.1), Conservation of momentum (3.2), and Conservation of energy
(3.3). At every point and time, these three laws must be satisfied.

Dρ
+ ρ∇ · ~v = 0 (3.1)
Dt

D~v
ρ = −∇ρ + ∇ · ø + F~ (3.2)
Dt

33
34 Chapter 3. Theoretical Background

DT Dp
ρCp = ∇ · k∇T + βT + µΦ (3.3)
Dt Dt

where τ , β and Φ are the stress tensor, thermal expansion coefficient, and viscous dis-
sipation, respectively. The stress tensor depends on the constitutive relation used for
the fluid of interest. Similarly, the thermal expansion coefficient is a fluid dependent
parameter, given by β = 1/ρ[∂ρ/∂T ]p . The viscous dissipation term influences advec-
tion and convection of heat, and becomes significant for high velocities or very viscous
flows.

With solutions to these three equations, the entire flow behavior and temperature field
are obtained completely. Solving these equations requires significant assumptions and
very specific geometric considerations that simplify the boundary and initial condi-
tions. This severely limits the number of realistic cases where an analytic solution is
applicable, opening the door to experimental and numerical work.

3.1.1 Single-Phase Flow

Analytic solutions to the Navier-Stokes equation are available for annular geometries,
similar to those utilized in this work. Key assumptions are made however to permit
analytic solutions to the governing equations, these assumptions include but are not
limited to:

• Symmetry or invariance with the azimuthal direction

• Steady state conditions
• Newtonian fluid, which specifies the constitutive relation for normal and shear
stresses
• Properties do not vary with position and time
• Streamlines are parallel, indicating that any mixing or entry effects have sub-
sided

Inherent in the assumption of invariant properties is an implied assumption of adia-

batic and isothermal flow, as fluid properties vary with temperature. This effectively
Chapter 3. Theoretical Background 35

provides a solution for the conservation of energy equation. These assumptions are
taken and applied to an upward annular flow, prescribed in Figure 3.1:

F IGURE 3.1: Upward flow in an annulus

For the flow given in Figure 3.1, the flow acts against gravity, and is subject to the
following boundary conditions: V~ (Ri , z) = V~ (Ro , z) = ~0. Considering parallel stream-
lines and axisymmetry assumptions, the radial and azimuthal velocities are zero. Mak-
ing the appropriate substitutions, the resulting continuity and z-momentum equations
are:

∂Vz
=0 (3.4)
 ∂z
∂P 1 ∂ ∂Vz
− ρgz + µ r =0 (3.5)
∂z r ∂r ∂r

Because both terms in (3.5) are dependent on different variables, the pressure gradient
is constant.

Integrating and applying the boundary conditions at r = Ri and r = Ro yields the

following solution:

  
1 dP ln Ri /r 2 2 2 2
Vz (r) = − ρgz (R − Ri ) − (r − Ri ) (3.6)
4µ dz ln Ri /Ro o
36 Chapter 3. Theoretical Background

The mean value theorem can be applied to equation (3.6) to eliminate fluid dependent
terms such as the viscosity and density, as well as the pressure gradient. The resulting
equation for Vavg is given by equation (3.7).

R Ro
Vz (r)2πrdr Ro2 − Ri2
  
Q Ri 1 dP
Vavg ≡ = = − ρgz + Ro2 + Ri2 (3.7)
Ac π(Ro2 − Ri2 ) 8µ dz ln(Ri /Ro )

The revised velocity equation in terms of the average velocity is given by equation
(3.8)

 
2Vavg ln Ri /r 2 2 2 2
Vz (r) = (R − Ri ) − (r − Ri ) (3.8)
Ro2 −Ri2
+ Ro2 + Ri2 ln Ri /Ro o
ln(Ri /Ro )

The process to obtain an analytical solution for the rectangular geometry is signifi-
cantly more complicated due to the corner effects. Tamayol and Bahrami [153] per-
formed a detailed analysis of flow in polygonal cross sections, modeling the shapes
as hyperellipses with a varying parameter that determines the geometric shape. Their
solution is given as an infinite series with varying coefficients. Although available, the
analytic solution for flow in a rectangular channel is not practical.

While a theoretical profile for the rectangular channel is not practical, an assumption
can be made by approximating the flow as flow between flat parallel plates. This is
a reasonable assumption as long as the width of the channel is large compared to the
gap. In such a case, the flow is symmetric. The revised, simplified geometry is given
in Figure 3.2.
Chapter 3. Theoretical Background 37

F IGURE 3.2: Flow configuration for rectangular channel

The boundary conditions for such a case are the same, the wall conditions remain the
same, u(0) = u(R) = 0, however to solve for the combined pressure drop and gravity
terms, the average velocity is used as a secondary condition.

R dP dP
− ρg Rr r2 − ρg 2
Z  
Q 1 dz
Vavg ≡ = − = − dz R (3.9)
Ac R 0 µ 2 2 12µ

The resulting velocity profile in the simplified rectangular channel is given by equation
(3.10).
6Vavg
Rr − r2


V (r) = 2
(3.10)
R

With simple approximations like the one above, the velocity profile can be estimated
with minimal deviation from the actual profile.

Turbulence and Boundary Layer Flows The turbulence characteristics of flows are
characterized by the local velocity of the fluid. If the velocity is above a particular
threshold value the flow is considered turbulent, however, if the velocity is below this
threshold, the viscous forces dampen the turbulent behavior. A measure of turbu-
lence in a flow is given by the Reynolds number, an exact definition is given in a later
38 Chapter 3. Theoretical Background

section. At low Reynolds numbers, the viscous forces dominate and damp out veloc-
ity fluctuations, rendering the flow laminar. At high Reynolds numbers, the inertial
forces cause the flow to become more chaotic, leading to enhanced mixing, making it a
useful phenomenon in heat and mass transfer processes. Typically, turbulent behavior
is defined by irregular behavior in all three dimensions such that the time history of
velocity at a point in space looks completely random; however, there is a structure to
turbulence, therefore the term random is not accurate.

Turbulent flows contain certain structures called eddies, which is a vague term for a
non-persistent spatial flow pattern [154]. They may have the form of a vortex, an em-
bedded jet, a mushroom shape or another form; they can occur in different length
scales; and they are not necessarily isolated, in that a large eddy can consist of smaller
eddies. This mosaic of eddies to form larger eddies is one of the principal characteris-
tics of turbulence. With the large number of eddies that are in constant motion, there
needs to be a constant supply of energy to sustain this behavior. Kolmogorov [155]
suggested that viscous effects do not affect the largest eddies in a flow pattern, but
that these eddies transmit energy to smaller internal eddies. Those eddies transmit
energy to even smaller eddies, and this process perpetuates until the receiving eddies
are so small that the viscous forces prevent them from growing. This process is coined
the energy cascade.

The energy of the largest eddies scales with the bulk velocity squared, u20 , and the time
that the eddy takes to make one turn is L/u0 . It is assumed that energy is lost by the
larger eddy as it transmits energy to smaller ones, and the rate at which this energy is
transmitted and ultimately dissipated is given by equation (3.11).

u30
= (3.11)
L
Chapter 3. Theoretical Background 39

Through a scaling process, Kolmogorov derived characteristic scales for length, time
and velocity that characterize the eddies at the smallest scales (3.12-3.14)

1/4
ν3

η≡ (3.12)

 ν 1/2
τ≡ (3.13)

vKol ≡ (ν)1/4 (3.14)

These lengths define the physical limits of where turbulence can reach, and is of value
for boiling phenomena, where the length scales for heat transfer and interfacial phe-
nomena can be on the atomistic scale. At high Reynolds numbers, the momentum
transfer due to viscous effects can be neglected everywhere except near the wall,
where the turbulent fluctuations diminish. This effectively divides the flow domain
into two regions, one where momentum is transported by turbulent fluctuations, and
one where it is transported molecularly by viscosity, - a turbulent core layer, and a vis-
cous sublayer, respectively. The thickness of the sublayer scales with the viscosity by
equation (3.15).
ν
δw = (3.15)
uτ
p
where uτ is the wall friction velocity, given as uτ ≡ τw /ρ, and τw ≡ ρνdu/dr − ρu0 v 0
is the wall shear stress.

To obtain a continuous solution, a third layer, the overlap layer is introduced that en-
compasses parts of the core layer and sublayer, however this only ensures continuity,
not smoothness. Schlicting [9] derived the velocity distribution for the overlap layer
and sublayer of turbulent Couette flows, as well as a buffer layer for smoothness, and
noted that this distribution can be applied universally near solid surfaces. The solu-
tion is expressed in terms of the single phase turbulent dimensionless wall distance,
r+ , to give a better understanding of the impact the liquid flow regime has on bubble
40 Chapter 3. Theoretical Background

size and behavior. This dimensionless distance is used in the law of the wall, which
relates the nondimensionalized average velocity to the dimensionless wall distance.

The relation for the buffer layer is more complicated than the relations typically pre-
sented because it provides extra smoothness when transitioning between viscous and
buffer layers and then again from buffer to overlap layers. This is important for the
purposes of boiling because the liftoff diameters of most bubbles falls within the buffer
layer, so oversimplifying the flow patterns in this region may hinder the understand-
ing of key phenomena in boiling.

The dimensionless velocity, u+ is defined as:

u
u+ = p (3.16)
τw /ρl

The dimensionless wall distance, r+ is given as:

p
+ r
τw /ρl
r = (3.17)
ν

The law of the wall is given by equation (3.18):


r+ 0 ≤ r+ < 5 Pure viscous layer





   

  
Λr+ +1 +
 Λ 3 ln √
1 1 √1 arctan 2Λr√3−1 + π 1
ln(1 + κBr+4 )


+ 2 +
+ 3 6
+ 4κ
+ (Λr ) −Λr +1
u =

5 < r+ < 70 Buffer layer








1
ln r+ + C + 70 < r+ Overlap layer


κ
(3.18)
where κ=0.41, B=1.43·10−3 , Λ=0.127 and C + =5.0.
Chapter 3. Theoretical Background 41

An empirical formulation for the wall shear stress τw is related to the friction factor by
equation (3.19) [156]:
2
f ρVavg
τw = (3.19)
8

The Darcy-Weisbach friction coefficient is calculated numerically using the Colebrook

equation [157], given by equation (3.20).

 
p  2.51
1/ f = −2 log + √ (3.20)
3.7Dh Re f

Equation (3.20) is solved numerically for each Reynolds number. The radial position
can then be nondimensionalized using equation (3.17) for easier comparison with the
threshold r+ values (r+ = 5 and 70). The law of the wall is widely used for numerical
simulations of turbulent flows, as a boundary condition.

3.1.2 Two-Phase Flow

In treating the interaction between liquid and gas phases, both bulk interactions and
surface interactions must be treated. Typically, surface interactions are governed by
the surface tension, however, the surface tension varies with the temperature and the
local fluid properties.

Eötvös rule allows the prediction of liquid surface tension at various temperatures,
knowing key physical properties of the liquid. It is given as

σV 2/3 = k(Tc − T ) (3.21)

Substituting the properties for water gives V = M/ρ, M is the molar mass (= 0.0180153
kg/mol), ρ is the liquid density, k is the Eötvös constant, k = 2.1 · 10−7 , and Tc is the
critical temperature, Tc =374◦C. This allows the liquid surface tension to be calculated
42 Chapter 3. Theoretical Background

for a wide range of temperatures.

Using this surface tension, the Young-Laplace equation can be used to calculate the
pressure difference across bubble interfaces. The Young-Laplace equation is given as,

 
1 1
∆p = −σ + (3.22)
R1 R2

Where R1 and R2 are the principal radii of curvature of the bubble. For liquid-vapor
interactions, heat transfer and bubble behavior are highly coupled and the constantly
changing mass of the bubble renders equation (3.22) non-applicable.

With the pressure difference across the bubble interface, ∆P , known, the temperature
inside the bubble can be determined using the Clapeyron equation [158]:

∆P Tsat
T = + Tsat (3.23)
∆ρhf g

Considering the increased pressure inside vapor bubbles, and Hsu’s [52] criterion for
bubble growth, it can be stated that in the vicinity of the bubble diameter at the mo-
ment of liftoff, the liquid temperature is equal to the saturation temperature at the
elevated pressure inside the bubble, as determined from equation (3.23). Coupled
with the measured wall temperature and the bulk fluid temperature, this provides a
third point allowing a second order curve fit to the temperature distribution inside the
fluid.

While the Young-Laplace equation and surface tension describe the behavior of the
bubble interface, the bulk behavior of the bubble is also of significant value for model-
ing purposes. The bubble behavior prior to lift off was treated in multiple works, but
most completely in [80], where a force balance is given.
Chapter 3. Theoretical Background 43

Force Balance on Growing Bubble Prior to liftoff, at every instant, the forces acting
on the bubble normal to the wall are balanced so that the bubble is static, but is still
permitted to grow and its interface allowed to deform. From this balance, the moment
of liftoff can be identified. The force balance acting on the bubble is detailed in Figure
3.1.2 [80].

( A ) X-Direction ( B ) Y-Direction

F IGURE 3.3: Force balance on a growing bubble in the X- and Y-directions

The force balance in the x− and y− directions are given by equations (3.24-3.25). Here
Vb is the volume of the bubble.

dubx
ΣFx = Fsx + Fdux + Fsl = ρv Vb (3.24)
dt
duby
ΣFy = Fsy + Fduy + Fp + Fg + Fqs = ρv Vb (3.25)
dt

where Fsi is the surface tension force in the i−direction [73]

π
Fsx = −Dw σ (cos θr − cos θa )
θa − θb
π(θr − θa )
Fsy = −1.25Dw σ 2 (sin θa + sin θr )
π − (θr − θa )2
(3.26)
44 Chapter 3. Theoretical Background

where Dw , θa , and θr are the bubble contact diameter, advancing and receding contact
angles, respectively. The inclination angle, θi , is the angle that the center of mass makes
with the center of the contact area.

Fdui is the unsteady drag force in the i−direction, also known as the added mass force.

11r˙b 2 11rb r¨b

 
Fdux = −ρl πrb2 + cos θi
2 6
11r˙b 2 11rb r¨b
 
2
Fduy = −ρl πrb + sin θi
2 6
(3.27)

Fsl is the shear lift force [73]

1
Fsl = Cl ρl πrb2 (ul − ub )2
2

where ul − ub is the relative velocity between the liquid and bubble, and Cl is the
coefficient of lift, given by the following correlation:

Cl = 3.877Gs1/2 (Re−2 2 1/4

b + 0.014Gs )

 
 dul  rb
Gs = 
 
dx  ul

2rb (ul − ub )
Reb =
νl

Fp is the pressure force

Fp = ρl gVb
Chapter 3. Theoretical Background 45

Fg is the gravity force

Fg = ρv gVb

Fqs is the quasi-steady drag force, given by Klausner et al. [73] as a correlation. The
quasi-steady drag consists of a steady drag force, Fs , with the bubble Reynolds num-
ber calculated using the instantaneous bubble diameter and the single phase liquid
velocity evaluated at the center of the bubble. This makes the equation quasi-steady,
and provides a measure of the effect the presence of the bubble has on the liquid.

 " #−1/0.65 
2 0.65
12 
Fqs = 6πρl νl (ul − uby )rb + + 0.7960.65
3 Reb 

ubi is the bubble velocity in the i−direction

Eventually, an imbalance in the x−direction force occurs and the bubble lifts off from
the heated surface. At the moment of departure, the contact area reduces to zero caus-
ing the surface tension force to diminish, the inclination angle becomes zero, and the
force balance becomes (3.28).

˙ 2 11rb,lo rb,lo
 
2 11rb,lo ¨ 1 2
−ρl πrb,lo + + Cl ρl πrb,lo (ul − ub,lo )2 = 0 (3.28)
2 6 2

Correlations can then be applied to equation (3.28) to determine the liftoff time and
radius.

Figure 3.4 shows a schematic for the bubble nucleation phenomenon [80].

After nucleation, the bubble grows at the nucleation site until it reaches a certain size
and departs from that site. It then slides along the wall, with vaporization occurring
at the heated surface of the bubble and condensation occurring at the outer surface.
Depending on which of these processes is dominant, the bubble can grow or collapse.
46 Chapter 3. Theoretical Background

F IGURE 3.4: Schematic diagram of bubble nucleation phenomenon

According to Okawa et al. [110], bubbles exhibit one of three behaviors after departure,
(a) sliding, (b) bouncing, and (c) collapsing. Okawa et al. documented these behaviors
for individual bubbles.

In the sliding direction, the force balance is governed by buoyancy, the unsteady drag
force, and the quasi-steady force, as given in equation (3.29).

11r˙b 2 11rb r¨b

 
∆ρgVb − ρl πrb2 + sin θi + ...
2 6
 " #−1/0.65  (3.29)
2 0.65
12 0.65

6πρl νl (ul − ub )rb + + 0.796 =0
3 Reb 

For bouncing bubbles, the governing force balance is the same as equation (3.28).

For a bouncing bubble, once the bubble enters a region with temperature lower than
the saturation temperature inside the bubble, the bubble begins to collapse due to in-
terfacial heat transfer from the vapor to the liquid. As the bubble shrinks in size, the
interactions between the liquid and vapor phase cause the shear lift force Fsl to de-
crease, pushing the bubble closer to the wall, and into a region with local temperature
Chapter 3. Theoretical Background 47

higher than the bubble saturation temperature. The bubble grows in size again due
to the temperature increase, and the process repeats, causing the bubble to bounce
around an equilibrium position and size.

When the temperature gradient is sufficiently high, and the velocity gradient suffi-
ciently low, the bubble will collapse once departed. The rate of collapse is determined
by the flow conditions, however without knowing the temperature and velocity fields,
it must be determined empirically.

Two-Phase Boundary Layer Flows The law of the wall described in Section 3.1 is
only applicable to single phase flows, due to the disturbances bubble nucleation would
cause. For many two-phase numerical simulations, the single phase law of the wall
is still widely used, possibly to the detriment of the results. The single phase law
of the wall is not a valid boundary layer for turbulent bubbly flow, however, it has
been established the two-phase boundary layer has the same structure as in single
phase flow, with a similar logarithmic profile in the turbulent core [159]. From these
findings, a two-phase law of the wall can be derived.

There are some key assumptions that must be made in order for a logarithmic law
in bubbly boundary layers, (a) that a similarity hypothesis holds; and (b) that a lo-
cal equilibrium exists between liquid turbulent energy production and dissipation.
In the wall region, the total turbulent viscosity was assumed to be the sum of shear
and bubble induced components. Troshko and Hassan [159] derived a two-phase log-
arithmic law that is consistent with experimental findings. In their derivation, the
non-dimensionalization remains the same, however it is weighted by phase:

τw
uτ = = [xl νt (∂ul /∂r)] |r=0 (3.30)
ρl
48 Chapter 3. Theoretical Background

where xl is the local liquid volume fraction, and the total turbulent viscosity, νt , is
given as
νt = νtout + νtin = κruτ + κl xv,max (ul − ub )r (3.31)

Here νtout and νtin are the turbulent shear and bubble induced viscosities, respectively,
xv,max = max(xv |30 ≤ r+ ≤ 200), and κl is a non-linearity empirical coefficient that dif-
fers from the single-phase κ. An empirical formula for κl is given by κl = 4.9453e−40.661uτ .
Substituting equation (3.31) into (3.30) yields equation (3.32)

dul βuτ
= (3.32)
dr κr

and

  −1
κl xv,max (ul − ub )
β= 1+ (1 − xv,max )
κuτ

The solution of (3.32) is then given by the two-phase law of the wall (3.33):

1
u+
tp =
+
ln rtp + C+ (3.33)
κ

The form is identical to the logarithmic formulation of the single phase law of the wall,
however, the velocity scale with which normalization is done is given as uτ,tp = βuτ .

3.2 Heat Transfer

The understanding of two-phase flow is incomplete without addressing heat transfer

that occurs in boiling or condensation. The processes permit more efficient transport
of heat, at the cost of increasingly complicated interactions between liquid and vapor
Chapter 3. Theoretical Background 49

phases.

3.2.1 Single-phase flow

Before approaching the two-phase heat transfer problem, a clear definition of single-
phase heat transfer should be detailed. Pohlhausen’s solution provides an analyti-
cal approach to flows over heated flat surfaces [9]. There are key assumptions that
Pohlhausen made in his solution:

• Continuum

• Newtonian fluid

• 2-D heat transfer

• Constant properties

• Steady state

• No dissipation

• No gravity

• No energy generation

Pohlhausen’s problem definition is presented in Figure 3.5 (adapted from [160]).

F IGURE 3.5: Diagram of convective heat transfer over a flat plate

Flow enters at velocity and temperature Vbulk and Tbulk , and a no slip condition is ap-
plied at the wall, as well as a specified wall temperature. As the flow slows down in
the wall region, it also heats up due to the temperature difference between the wall
50 Chapter 3. Theoretical Background

and bulk fluid. This creates two distinct boundary layers, a viscous boundary layer
and thermal boundary layer. At the wall, the conditions are given as V (x, 0) = 0 and
T (x, 0) = Tw (x). At the edge of the boundary layer, conditions are approximately the
same as the bulk fluid.

The dynamics of the temperature distribution in the thermal boundary layer changes
depending on whether the thermal boundary layer is thicker than the viscous layer.
The governing parameter is the Prandtl number, the ratio of the viscous to thermal
diffusion rates.

x
Pr  1 δt > δ δt ≈ √
P rRe
x
Pr  1 δt < δ δt ≈
P r Re1/2
1/3

In terms of the solution, similarly to the the approach of Blassius for flow in a bound-
ary layer [9], a similarity transformation is used to combine independent variables into
a single variable, to reduce the complexity of the governing equations. Pohlhausen’s
solution makes use of Blassius’ solution, and is expressed in terms of the Prandtl num-
y
ber and dimensionless temperature at the wall, θp (η = δ
= 0), which must be tabu-
lated. The resulting temperature profile is given by equation (3.34).

! R ∞ h d2 f iP r
dη
r
Vbulk T − Tw η dη 2
θP (η) ≡ θP y ≡ = 1 − R h iP r (3.34)
νx Tbulk − Tw ∞ d2 f
0 dη 2
dη

where d2 f /dη 2 is tabulated from the Blassius solution. The local heat transfer coeffi-
cient is given by (3.35). r
Vbulk dθp (0)
h(x) = k (3.35)
νx dη

This solution can be applied to internal flows, as long as the thermal and viscous
Chapter 3. Theoretical Background 51

boundary layers are only influenced by a single wall boundary, as is the case where
the flow channel width is significantly larger than the boundary layer thicknesses.

3.2.2 Two-phase flow

While analytic solutions based firmly in theory exist for single phase convective heat
transfer, this typically does not carry over to convective boiling. Most of the early
development in convective boiling theory was formulated by Forster and Zuber [161,
161], followed by Chen [54]. Rather than relate flow boiling heat transfer to well estab-
lished models for single phase heat transfer, Forster and Zuber developed a microcon-
vective mechanism for pool boiling, where the flow was driven by the bubble growth
force. In their model, the Reynolds number scaled with the superheat.

Chen noted a key distinguishing fact in Forster and Zuber’s analysis, that the super-
heat is not constant throughout the boundary layer and that the Reynolds number
must be representative of an effective superheat, because there is a temperature gradi-
ent between the wall and the bulk fluid. Chen suggested that the difference between
the effective and actual superheat is negligibly small in pool boiling, however, it can-
not be neglected in flow boiling. Figure 3.6 (Chen [54]) illustrates the difference in the
two cases.

F IGURE 3.6: Comparison of temperature distributions in pool and flow

boiling cases
52 Chapter 3. Theoretical Background

Taking account of the smaller local superheat around a growing bubble in flow boiling
cases, Chen adjusted Forster and Zuber’s microconvective correlation by adding a
suppression factor to adjust for the effective superheat. Chen’s revised correlation is
detailed in a later section.

For every flow condition, there is an equilibrium wall temperature that corresponds
to the heat flux applied. When boiling occurs, the amount of heat extracted increases
significantly with a slight increase in wall superheat. This enhancement of heat trans-
fer continues until the vapor creates an insulating layer between the heated surface
and liquid flow. The evolution of the heat transfer coefficient with local void fraction
is shown in Figure 3.7 (Yan [117]).

F IGURE 3.7: Evolution of flow boiling from ONB to DNB

Chapter 3. Theoretical Background 53

Points A-C are also considered to be partially boiling. In addition, Figure 3.7 identifies
key events in the boiling process:

• ONB - Onset of Nucleate Boiling - initial bubbles

• OSV - Onset of Significant Void - significant increase in vapor

• FDB - Fully Developed Boiling - wall temperature near constant

• DNB - Departure from Nucleate Boiling - significant increase in wall temperature

The heat flux when DNB occurs is called the critical heat flux, or CHF. Beyond this
point, the heat transfer coefficient no longer increases with increasing wall superheat,
and the insulating vapor layer can lead to material damage of the heater surface. Many
approaches exist to increase the heat flux at which DNB occurs, but the main op-
erational principle of these approaches is by increasing the overall amount of phase
change occurring. By balancing the bubble departure diameter with increased bubble
departure frequency and nucleation site density, more bubbles depart from the more
locations on the heated surface resulting in more vapor generation. Coupling this with
an increased pressure, the higher vapor density permits bubbles to have higher vapor
content while maintaining similar size, further increasing the CHF. Many works have
investigated the phenomena of CHF, with some investigating methods of increasing it
by artificial means [129–140].

While most work focuses on estimating the heat transfer coefficient in flow boiling,
other works have investigated the heat transfer from the bubble to the surrounding
fluid. While these works have typically treated condensation heat transfer after bub-
ble departure [92, 98], there is no reason why the same treatment cannot be applied to
the outer edges of an attached bubble, as is done by Chen and Mayinger [162]. Chen
and Mayinger’s model of the condensation Nusselt number is given in terms of the
bubble history number, which effectively connects the heat transferred across the bub-
ble interface with the size of the bubble.
54 Chapter 3. Theoretical Background

In understanding boiling phenomena, the local thermal quality is also an important

parameter, as it can be used to predict the onset of nucleate boiling or the position
where burnout can occur post-CHF. The local thermal quality is given by equation
(3.36)

Z z
ζh
x(z) = xin + qw00 dz 0 (3.36)
AGhf g 0

∆hsub,in
xin =
hf g

where ζh and A are the heated perimeter, and liquid flow area respectively. The inlet
and outlet thermal qualities are of key importance in correlations for critical heat flux
and heat transfer coefficient.

When boiling occurs, the phase change is driven by a local superheat. As the bubble
grows, it is in a constant balance of energy, with additions coming from the heated
surface, and losses to the surrounding subcooled liquid. The problem of the temper-
ature profile near the wall is two-sided; while Chen’s argument of the thinning of the
thermal boundary layer due to liquid flow still holds, there is a re-thickening of the
boundary layer that occurs with the collapse of a bubble. When the bubble density is
high, both of these effects must be accounted for when modeling temperature varia-
tions.
Chapter 3. Theoretical Background 55

3.3 Existing Correlations

3.3.1 Wall Temperature and Heat Flux

Many of the correlations for heat flux aim to predict one of the two extremes of the
boiling spectrum, at the onset of nucleate boiling or the critical heat flux. If the heat
fluxes at these two points can be accurately predicted, the upper and lower limits of all
boiling conditions would be known. All of the experimental work performed herein
00 00
is in the partial boiling region, therefore correlations for qw,ON B and qCHF should un-

derpredict and overpredict, respectively, the heat flux. Alternatively, for cases where
the bubble nucleation density and rate are low, equations for qw,ON B can be used to
predict the wall temperature. Table 3.1 lists some commonly used correlations for heat
flux for boiling conditions

Bergles and Rohsenow [53] developed an empirical expression for the heat flux at ONB
in terms of system pressure. Rather than considering the variation of fluid properties
with pressure, the pressure is included directly. Similarly, their model acknowledges
that there is a wide range of cavity sizes in commercially available surfaces, so the
heat flux at ONB is typically invariant of the surface. The experimental data used to
develop this correlation were performed on stainless steel and nickel surfaces, using
water, at pressures from 1-136bar.

Sato and Matsumura [163] developed an analytical equation for heat flux at ONB
based on the criterion published by Hsu [52]. Building on the relation from Sato and
Matsumura, Davis and Anderson [164] adjusted the equation by the contact angle, φ.

Chen [54] argued that for flow boiling, the temperature gradient neat the wall is much
larger than in pool boiling cases, and the difference in superheats at the wall and the
bubble centroid cannot be ignored. To account for this, Chen modified the microcon-
vective heat transfer coefficient from Forster and Zuber [165], adding a suppression
56 Chapter 3. Theoretical Background

TABLE 3.1: Existing correlations for heat flux

Author Model or Correlation

00 1.156 0.0234 /2.16
Bergles and Rohsenow [53] qw,ON B = 1083P (1.8∆Tw )P
(1964) P is the pressure in bar
00 kl hf g ρv (∆Tw )2
Sato and Matsumura [163] qw,ON B = 8σTsat
(1964)
00 kl hf g ρv (∆Tw )2
Davis and Anderson [164] qw,ON B = 8(1+cos φ)σTsat
(1966)
φ is the contact angle
00
qw = hmic ∆Tw + hmac ∆Tsub
Chen [54]  k0.79 C 0.45 ρ0.49 g0.25 
(1966) hmic = 0.00122 σl 0.5 µ0.29
pl l c
h0.24 ρ0.24
∆Tw0.24 ∆P 0.75 S
l fg v

0.4 kl
hmac = 0.023Re0.8
l P rl Dh F

K + = C1 (xCHF + C2 )

0.5
Kureta and Akimoto [131] K + ≡ BoCHF Gν σ
ζh zh
(2002) x CHF = x in + BoCHF
  
2 A 
C1 = 6.9 ζζwh − 10 ζζwh + 2 · 10−3
 2  
C2 = −0.75 ζζwh + 0.9 ζζwh − 0.28
qw ”Dd
Situ et al. [81] NqG ≡ αl ρv hf g
= 5.28Ja1.37
(2008)

factor, S = (∆Te /∆T ), which is the ratio of the effective superheat to the total super-
heat. Additionally, a Reynolds number factor, F = (Re/ Rel )0.8 , is added to the macro-
convective heat transfer coefficient originally developed by Dittus and Boelter [166].
This Reynolds number factor is determined experimentally, as it is related to the Mar-
tinelli paremeter, Xtt = ((1 − x)/x)0.9 (ρv /ρl )0.5 (µl /µv )0.1 .

Chen and Fang [167] reported on the application of Chen’s correlation. While widely
used, Chen’s correlation contains many parameters and is often cited with typograph-
ical errors which are carried forward in later works. Chen and Fang presented the cor-
relation in its entirety and pointed out inaccuracies in proposed parametric equations
for S and F . They also identified acceptable replacements for these faulty equations
Chapter 3. Theoretical Background 57

from previous works.

The most widely used parametric equations come from Collier [168], and are given by
equations (3.37-3.38):
1
S= 1.17
(3.37)
1 + 2.53 · 10−6 Rel,sat


 2.35(1/Xtt + 0.213)0.736 1/Xtt > 0.1


F = (3.38)

 1
 1/Xtt ≤ 0.1

Here, an assumption is made that 1/Xtt ≤ 0.1 and the Reynolds number factor F is set
to unity.

Kureta and Akimoto [131] developed a CHF correlation based on the critical quality.
Their experimental results showed that the critical quality was between -0.092 and -
0.014 in the conditions investigated. C1 and C2 are coefficients that account for the
ratio of the heated to wetted perimeters. While the experiments performed by Kureta
and Akimoto used a rectangular geometry with a single side heated, their correlation
was fitted to data with differing heater and flow geometries.

Situ et al. [81] proposed a correlation that relates the Jakob number and the nondimen-
sional heat flux. This correlation was fitted to data sets presented by Basu et al. [77],
Thorncroft et al. [75], and Situ et al.’s own data, with reasonably good agreement,
however, the correlation does not take pressure into consideration, albeit the fluid
properties involved vary with pressure. Additionally, the nondimensional heat flux
makes use of the bubble liftoff diameter. Situ et al.’s correlation was developed for
conditions near ONB at atmospheric pressure using data that ranged in mass flux
from 124-890kg m−2 s, heat fluxes from 11-1130kW m−2 , and inlet subcooling from 6.6-
53◦C. This correlation compared well with data from Basu et al. [77], Situ et al. [81],
and Thorncroft et al.’s [75] upflow and downflow data. Situ et al.’s correlation can also
58 Chapter 3. Theoretical Background

be used to predict the wall temperature for a given bubble diameter.

In every instance of partial boiling, equations for CHF should significantly overpre-
dict the heat flux, effectively making it an upper limit for the heat flux. Opposingly,
equations for heat flux at ONB should underpredict the heat flux. For predicting heat
transfer during intermediate phases, Chen’s correlation is widely used.

3.3.2 Bubble liftoff diameter

Bubble liftoff is the moment that the bubble detaches from the heater surface, and
may not coincide with bubble departure, the moment the bubble detaches from the
nucleation site, as is the case with sliding bubbles. The lifespan of a bubble depends
highly on boiling parameters such as subcooling, superheat, and heat flux, in addition
to flow parameters such as the Reynolds number and system pressure. Interactions
between bubbles can also influence bubble growth or obstruct the liquid flow, and this
behavior is also of interest for CFD models.

Models for the bubble diameter as a function of time, t, have been developed by Zuber
[21] and Mikic et al. [89]. To obtain the maximum bubble diameter using these models,
the time at which the bubble liftoff is observed to occur, tlo , can be supplied as an
input. Further models have been developed to predict the bubble departure diameter
explicitly, including those from Ünal [57], Prodanovic et al. [105], and Situ et al. [80].
The resulting correlations from these publications can be found in Table 3.2.

Zeng et al. [74] compared their experimental results with Zuber’s [21] rudimentary
correlation, which was developed for pool boiling, and is valid for atmospheric and
subatmospheric conditions. The correlation is a function of only the Jakob number
and thermal diffusivity, and can be adjusted to fit experimental conditions. Experi-
ments were performed using R-113 as the working fluid, and experimental data cov-
ered mass fluxes from 149-262kg m−2 s, inlet quality from 0.039-0.169, heat fluxes from
Chapter 3. Theoretical Background 59

TABLE 3.2: Existing correlations for bubble diameters

Author Model or Correlation

Db (t) 2b
√
Zuber [21] 2
= √
π
Ja αl t
(1961) b = empirical asphericity constant
Db (t) 2
Mikic et al. [89] 2
= 32 BA ((t + 3/2
q + 1) − (t ) q− 1)
+ 3/2

(1970) t+ = A2 t
,
h ρ ∆T
A = π7 f gρlvTsat sat , B = 12 α Ja
B2 π l
− 0.709 √a
Dmax = 2.42 · 10 5P
Ünal [57] qbφ
(1975) a = 2ρv hf g √παl , b = 2(1−ρv /ρl ) , γ = kHRklρρHR
∆Tsat kl γ ∆Tsub cHR
l cp,l

0.47
φ = max U0.61bulk
,1
Prodanovic et al. [105]  1.772
∗
Dlo = 236.749 · Ja−0.581 θ−0.8843 ρρvl Nb0.138
(2002) √
352/3
Situ et al. [80] Dlo = π (ul −uνbl)·√Cl b2 Ja2e P rl−1
(2005) b is the same constant as in Zuber’s correlation (1961)
Chu et al. [119]  1.36
∗
Dlo = 12788.5 · Ja−0.28 θ−1.07 ρρvl Nb0.35
(2011)

5.8-16.8kW m−2 , and wall superheat from 5.2-12.0◦C.

Klausner et al. [73] used Mikic et al.’s [89] correlation to predict the time rate of change
of the radius for a spherically growing bubble at a wall. The correlation is highly
dependent on the wall temperature to form a nondimensional time parameter. A key
limitation of this model is that it lacks contributions from the mass flux and pressure.
Experiments were performed using R-113 as the working fluid, and experimental data
covered mass fluxes from 113-287kg m−2 s, inlet quality from 0.002-0.165, heat fluxes
from 11.0-26.0kW m−2 , and wall superheat from 12.0-17.7◦C.

Ünal’s model was developed using an energy balance for a growing bubble to pre-
dict the maximum bubble diameter when a bubble grows on a liquid microlayer. He
also performed a photographic study to measure bubble diameters and growth. The
experimental data used to validate his model covered a pressure from 0.1-17.7MPa,
heat fluxes of 0.47-10.64MW m−2 , liquid velocities of 0.08-9.15m s−1 , and subcooling
of 3-86◦C. His model was validated against 37 bubble diameters, with the maximum
bubble diameters seen in his experiments ranging from 0.08-1.24mm.
60 Chapter 3. Theoretical Background

Prodanovic et al. [105] developed a correlation to predict the maximum bubble di-
ameter. The correlation is comprised of parameters considered by the authors to ac-
count for all relevant variables affecting bubble diameters. These include the Jakob
number, a nondimensional temperature parameter (θ), the boiling number, and the
density ratio - which accounts for the pressure. Experiments were performed using
water with upward flow at pressures ranging from 1.05-3bar, bulk liquid velocities
from 0.08-0.8m s−1 , and subcooling from 10-30◦C. A total of 54 data sets were used to
validate the new model. A comparison with the aforementioned models showed that
Zuber’s [21] and Mikic et al.’s [89] models overestimates the bubble diameter greatly,
while Ünal’s [57] correlation was acceptable.

Situ et al. [80] developed a correlation to predict the maximum bubble diameter. A
total of 91 data points were presented from experiments performed with water in
an annular geometry at near atmospheric pressure, covering mass fluxes from 466-
900kg m−2 s, heat fluxes from 60.7-206kW m−2 , and subcooling from 1.5-20◦C. The cor-
relation relates the lift off diameter to the bubble lift coefficient, fluid properties, Jakob
number and Prandtl number. The lift coefficient, however, is given by a second cor-
relation, equation (3.39) that uses the dimensionless fluid velocity gradient and the
bubble Reynolds number.

p
Cl = 3.877 Gs (Re−2 2 1/4
b + 0.014Gs ) (3.39)

1
Gs =
κu+ (ul − ub )/ul

Dlo (ul − ub )
Reb =
νl

Here, ul − ub is the relative velocity between the liquid phase and the bubble center
Chapter 3. Theoretical Background 61

of mass. For a bubble departing, the bubble velocity is equal to the sliding velocity,
if applicable; otherwise ub = 0. u+ and κ are the nondimensional velocity and con-
stant given by the law of the wall described in Section 3.1. The value of u+ depends
on which categorical turbulent region the bubble center is within; that is to say, u+
depends partially on D+ /2. To provide an initial guess as to the appropriate nondi-
mensionalization to use, D+ /2 is calculated using experimentally measured bubble
diameters. Situ et al.’s [80] correlation can then be solved numerically.

Chu et al. [119] evaluated the correlations from Ünal [57], Prodanovic et al. [105], and
Situ et al. [80] and modified Prodanovic et al.’s [105] correlation by modifying the
coefficients and exponents. Chu et al. [119] compared these correlations with exper-
imental data from the respective studies, along with an additional 125 experimental
data points covering a pressure of 139-152kPa, mass flux from 300-702kg m−2 s, heat
flux from 133.4-355.6kW m−2 , and subcooling from 1.4-24◦C.

3.3.3 Bubble behavior after liftoff

After the moment of bubble liftoff, the bubble becomes subject to the dynamic nature
of the bulk fluid, including the turbulent fluctuations and the temperature gradient.
Whether the bubble slides, bounces, or collapses depends on the temperature and
velocity fields. Independent of these behaviors, there are empirical relations that aim
to predict the bubble shape and collapse rate. The resulting correlations from these
publications can be found in Table 3.3.

Wellek et al. [169] proposed an upper boundary for the aspect ratio, Ra . Okawa et
al. [94] proposed a correlation for the lower limit of the aspect ratio. It was shown
experimentally that the aspect ratio decreased with the Bond number. Moore’s cor-
relation [170] was used to predict the aspect ratio as a function of the Weber number,
W e. These correlations were developed for bubble injection into a static liquid, and the
62 Chapter 3. Theoretical Background

aspect ratio used in those cases is the inverse of the way it is defined here due to the
change in direction of elongation between these cases. The air bubble data collected
in this work cannot be adequately compared with these correlations due to the stark
differences in flow conditions, however, they are listed to illustrate the significant gap
in experimental work that characterizes bubble shape in liquid streams.

Chen and Mayinger [162] proposed a bubble size model that relates the nondimen-
sional bubble history to flow parameters and the bubble liftoff diameter. In their
work, experiments were performed using holographic interferometry and high-speed
cinematography to measure the temperature field around condensing bubbles, with
ethanol, propanol, refrigerant R-113 and water as working fluids. Their model is ap-
plicable to bubbles both before and after liftoff.

Warrier et al. [92] evaluated bubble collapse models, including Chen and Mayinger’s
correlation, and developed a new one to fit with their experimental data. In their ex-
periments, a thermocouple array was used to measure the liquid temperature field
around collapsing bubbles, which were measured using high speed video. Their re-
sulting correlation was a modified form of Chen and Mayinger’s model.

Kim and Park [98] also evaluated numerous bubble collapse models and developed a
new one to fit with their experimental data. Similar to the method of Warrier et al., a
microthermocouple was used to measure the liquid temperature field around collaps-
ing bubbles. A stereoscopic high speed video system was used to obtain orthogonal
views of the bubble, providing more accurate measurements of the bubble size. Their
resulting correlation was a modified form of Chen and Mayinger’s model.

Between correlations for the heat transfer at the wall, bubble departure size, and bub-
ble behavior before and after departure, a complete picture of the bubble lifespan can
be predicted using these models.
Chapter 3. Theoretical Background 63

TABLE 3.3: Existing correlations for bubble behavior

Author Model or Correlation

1
Wellek [169] Ra,upper = 1+0.163Bo 0.757

(1966) Ra,upper is the upper limit of the aspect ratio

1
Okawa et al. [94] Ra,lower = 1+1.97Bo 1.3

(2003) Ra,lower is the lower limit of the aspect ratio

Moore [170] 3 +R −2)[R2 sec−1 R −(R2 −1)1/2 ]2
4(Ra a a
We = a
4/3
a
(1959) 2 −1)3
Ra (Ra

Chen and Mayinger [162] βc = (1 − 0.56Re0.7

b,lo P r
1/3
JaF ob,lo )0.9
(1992) (u −u )D
Reb,lo = l νbl b,lo
tlo αl
F ob,lo = (D 2
b,lo

Zeitoun et al. [171]

βc = (1 − 5.67Re0.61
b,lo α
0.328
Ja0.692 F ob,lo )0.72
(1994)
Warrier et al. [92] 3/2 1/2 2/3
βc = 1 − 1.8Reb,lo P r1/3 JaF ob,lo (1 − 0.72Ja0.9 F ob,lo )
(2002)
Kim and Park [98]
βc = (1 − 0.6695Re0.7
b,lo Ja
0.7957
P r−0.4564 F ob,lo )0.7692
(2011)
Chapter 4

Experimental Apparatuses and Cases

Experiments have been performed in multiple geometries and a wide range of condi-
tions to investigate bubble nucleation, growth and departure behavior. The adiabatic
injection of air into a liquid stream, and subcooled flow boiling of water at multiple
pressures were investigated. In these experiments, annular geometries are used, with
additional air injection experiment performed in a rectangular test section. Bubble
growth and detachment is observed using high-speed videography, while liquid flow
fields are measured using a PIV technique. A microscopic objective is used to mea-
sure the flow field and bubble size in boiling cases, where the bubbles are less than a
millimeter in diameter.

Additional information regarding the calibration, acquisition and analysis procedures,

and uncertainties for all cases can be found in Appendix B.

4.1 Experiments and Geometries Treated

Separate experimental apparatuses were constructed for air injection in an annular ge-
ometry, subcooled flow boiling in a vertical annular test section at atmospheric pres-
sure, and subcooled flow boiling with a vertical annular test section at elevated pres-
sures.

64
Chapter 4. Experimental Apparatuses and Cases 65

4.1.1 Air Bubble Injection in Annular and Rectangular Geometries

Two test sections were used to investigate air injection into a liquid stream. A ded-
icated flow loop with an annular test section and a rectangular test section that was
mounted to the ambient pressure boiling flow loop. With the latter test section, the
liquid was heated to raise the Reynolds number for better comparison with boiling
flows.

Annular Geometry A schematic of the air bubble injection loop is shown in Figure
4.1. The loop was originally designed and constructed by Saman Reshadi. The work-
ing fluid used was deionized water, which was stored in a main tank circulated by
a pump with a flexible impeller made of viton, with a capacity of 22.8L min−1 . This
impeller material was selected because of its compatibility with deionized water and
PIV seed particles, and its working temperature range of 15-80◦C. The actual water
flow rate was measured by a digital flow meter.
66 Chapter 4. Experimental Apparatuses and Cases

F IGURE 4.1: Schematic diagram of the air injection flow loop

The cross section of the air injection test section is also shown in Figure 4.1. It con-
sisted of a 9.53mm O.D. inconel tube with a 0.90mm wall thickness placed inside a
19.04mm I.D. polycarbonate tube. The outer diameter of the polycarbonate tube was
22.2mm. The annular flow channel formed had a gap of 4.8mm, so the hydraulic di-
ameter was 9.51mm. The maximum bulk velocity of water in the annular channel was
then 1.78m s−1 . A 400µm hole was drilled on the inconel tube wall for air bubble injec-
tion at an elevation of 1000mm above the liquid inlet. Using a syringe pump, air was
injected at a rate of 5.4mL min−1 through the injection hole. All measurements were
taken in the vicinity of this point, which corresponded to a length-to-hydraulic diam-
eter ratio (L/Dh) of 105 from the liquid inlet. The water exiting the test section flowed
back into the main tank, which was separated by a baffle plate, preventing continued
entrainment of air bubbles in the liquid stream. These experiments were all performed
Chapter 4. Experimental Apparatuses and Cases 67

at room temperature.

Rectangular Geometry As an extension of the air injection experiments, a rectangu-

lar test section was constructed to obtain better accuracy in spanwise velocity mea-
surements, as well as to investigate the effect of reducing the channel dimension on
bubble-liquid interactions and bubble shape and size. Experiments were conducted
using different flow channel areas: rectangular channels of the same width, but dif-
ferent gap heights. The growth and detachment of a single air bubble in a turbulent
water stream was observed for a 5.1mm gap channel; the same phenomena were ob-
served for the same minichannel with a 1.9mm gap under the same velocity or an
intermediate Reynolds number.

The rectangular test section is comprised of multiple parts, a bottom section with a
cutout for an aluminum heating block, a top section with a glass window to allow laser
light to pass, and a middle plate with the injection hole and a threaded connection.
Additionally, the channel has two optically transparent acrylic sidewalls that allow
recording of the flow/air bubble injection using a high-speed camera. Depending on
the channel size desired, an acrylic plate may be added to reduce the overall channel
gap. A diagram of the rectangular test section is shown in Figure 4.2. It is noted that
the laser is used only in PIV experiments.

The cross section of the test section is shown in Figure 4.3. It consists of a polycar-
bonate rectangular channel with a small hole of diameter 500µm to allow injection of
air bubbles from a syringe pump. The larger minichannel (20mm×5.1mm) has a hy-
draulic diameter of Dh =8.13mm. By mounting a 3.2mm thick polycarbonate plate on
the outer face of the larger minichannel, the gap in the smaller minichannel was re-
duced to 1.9mm, so the hydraulic diameter was also reduced to Dh =3.47mm. In all
cases, air was injected from a 500µm diameter hole at a fixed flow rate of 2.0mL min−1 .
68 Chapter 4. Experimental Apparatuses and Cases

F IGURE 4.2: Rectangular test section configuration

F IGURE 4.3: Cross section of the rectangular test section

A lower gas flow rate was selected to prevent large-scale out-of-plane motion of the
bubbles that would hinder flow field measurements.

To achieve a wider range of Reynolds numbers, the rectangular test section was mounted
onto the subcooled flow boiling loop, which allowed Reynolds numbers of up to Re =
7700, higher than was achievable using the air injection loop, due to lower viscosity of
water at higher temperatures.
Chapter 4. Experimental Apparatuses and Cases 69

4.1.2 Subcooled Flow Boiling at Ambient and Elevated Pressures

To investigate flow boiling over a wide range of conditions, two separate boiling flow
loops were constructed.

Ambient Pressure Boiling Loop A schematic of a subcooled flow boiling loop is

shown in Figure 4.4. The working fluid is degassed, deionized water, which is stored
in a main tank equipped with a 1.5kW immersion heater and circulated by a centrifu-
gal pump with a capacity of 37.9L min−1 . The actual water flow rate is measured by a
digital flow meter. The maximum bulk velocity of water in the annular channel is then
2.96m s−1 . The water exiting the test section flows into an air-cooled condenser where
any steam bubbles would be condensed and the liquid cooled down. The inlet sub-
cooling is controlled by adjusting the cooling water flow rate through the condenser
tank and the power supplied to the immersion heater in the main tank. The bulk
temperature in the main tank is kept constant during steady state experiments.

The cross section of the subcooled flow boiling test section is shown in Figure 4.5. It
consisted of a 9.53mm O.D. inconel tube with a 0.90mm wall thickness placed inside
a 19.04mm I.D. polycarbonate tube. The annular flow channel formed had a gap of
4.8mm, so the hydraulic diameter was 9.51mm. The outer diameter of the polycarbon-
ate tube was 25.4mm for the subcooled flow boiling experiments. Different outer tube
thicknesses between the air injection and subcooled flow boiling experiments were
used due to the different fittings used to center the inner inconel tube. While the air
bubble injection experiment did not experience any thermal expansion, the inconel
tube for the subcooled flow boiling set up had to be allowed to slightly elongate axi-
ally.

A cartridge heater (6.3mm O.D., 254mm length) was placed inside the inconel tube to
provide heating at a maximum heater power of 750W or wall heat flux of 100.3kW m−2 ,
70 Chapter 4. Experimental Apparatuses and Cases

F IGURE 4.4: Schematic diagram of the subcooled flow boiling loop

F IGURE 4.5: Cross section of the subcooled flow boiling annular test sec-
tion
Chapter 4. Experimental Apparatuses and Cases 71

which was sufficient to initiate subcooled boiling at the flow conditions of interest.
The 0.715mm gap between the cartridge heater and inner wall of the inconel tube was
filled with thermally conductive grease. To measure the tube wall temperature, a type
T thermocouple was inserted into a 510µm hole drilled through the inconel tube wall
and embedded using a metal-repair epoxy. However, the wall temperature response
data is unavailable due to the fragility of the thermocouple, which becomes damaged
when the test section undergoes thermal expansion.

Elevated Pressure Boiling Loop A separate flow loop was built to conduct sub-
cooled flow boiling experiments at pressures up to 10bar. The loop was originally
designed for a senior design project, but has since been heavily re-engineered and
modified to better achieve the desired conditions. The loop is pressurized using a
compressed nitrogen gas canister and utilizes a gear pump. A heat exchanger is used
to achieve steady state conditions: for low pressures and high heat fluxes, water is
used as a coolant, otherwise air is used. A schematic diagram of the high pressure
flow loop is shown in Figure 4.6.

The apparatus features a stainless steel annular test section that houses a central stain-
less steel tube, heated by joule heating. The inner and outer tubes are selected to be the
same material to match the thermal expansion coefficients. The test section includes
a fitted and sealed glass window for making visual measurements of key subcooled
flow boiling parameters, such as bubble diameter and nucleation site density.

The inner stainless steel rod is joined to copper electrodes at the top and bottom to
allow for the electrical connection to the power supply. Copper is selected for the low
electrical resistance to prevent heat buildup. Additionally, two thermocouples are spot
welded internally 180◦ apart at a distance of 300mm from the top of the test section.
In order to manufacture the inner rod, it had to be done in three sections, press fit to
allow for full electrical contact and prevent separation during heating processes. The
Chapter 4. Experimental Apparatuses and Cases

F IGURE 4.6: High pressure boiling flow loop schematic

72
Chapter 4. Experimental Apparatuses and Cases 73

stainless steel sections are wetted, while the copper section is exposed to air to allow
for electrical connections outside the loop and prevent overheating. The heater rod
has a 12.7mm OD, and is placed inside a stainless steel tube with 25.4mm ID.

4.2 Operational Limits of Experiments

When determining under which conditions experiments should be performed, the op-
erational limits of the experimental setup should be established, so that a wide spec-
trum of data can be obtained. The operational limits of an experiment are bounded by
the limits of its components. These limits are further reduced by design choices that
increase pressure drop or heat losses, among other parameters.

For each experimental setup, the combined limits of the individual components are
reported in Table 4.1. The pressure, superficial liquid and gas injection velocities,<
jf > and < jg,inj >, and heat flux for each experiment are listed, if applicable. Note
that the cross sectional area used to calculate the liquid superficial velocity is flow
channel area, whereas the area used to calculate the gas injection superficial velocity
is the injection hole area.

4.2.1 Air Injection Flow Loop

For the air injection flow loop, a flexible impeller pump with fixed speed motor was
used. Due to the pressure drop in the length of tubing, the pump provided a maximum
flow rate of 8.86L min−1 through the loop, and with the use of a bypass line, the flow
rate through the test section ranged from zero (air injection into a static liquid) to
8.86L min−1 .
74 Chapter 4. Experimental Apparatuses and Cases

When injecting air, the maximum injection rate was used to minimize the intermit-
tency of bubble injection. For the combination of syringe pump and syringe used,
the range of flow rates is 0-5.4mL min−1 . Despite using the maximum gas flow rate,
there is still a delay between successively formed bubbles. This delay is attributable
to the buildup of pressure inside the syringe and injection line, in order to displace a
small amount of liquid that enters the injection hole via capillary action, as well as to
overcome the surface tension.

4.2.2 Atmospheric Pressure Boiling Loop

For the atmospheric pressure boiling loop, a variable speed centrifugal pump was
used. This permitted a wide range of flow rates, and coupled with the bypass line,
gave a much finer control over flow rate through the test section than the air injec-
tion loop permitted. Although the maximum flow rate provided by the pump is
37.9L min−1 , the flow rate used is much less than this to limit heat losses to the en-
vironment.

The main limitation of the ambient pressure boiling flow loop was the heater ele-
ment in the test section, which was powered using a variable voltage A/C trans-
former. The heater element was 254mm long, relative to the 1500mm test section. The
maximum power supplied to the element was 750W, with an equivalent heat flux of
100.3kW m−2 . While this power is sufficient to achieve subcooled boiling at the area of
interest, heat losses between the storage tank and the test section were high enough to
limit the flow rate.

Air injection experiments in the rectangular geometry were performed on this loop.
In these experiments, the liquid temperature was held constant at 88◦C, enough to
adjust the surface tension and significantly reduce the viscosity of the liquid. This
Chapter 4. Experimental Apparatuses and Cases 75

TABLE 4.1: Operating limits for each experiment

Dh P < jf > < jg,inj > qw00

Experimental Loop
(mm) (bar) (m s−1 ) (m s−1 ) (kW m−2 )
Air injection in annulus 9.53 1 0-0.690 0-0.716 -
Air injection in rectangular channel 8.13 1 0-6.193 0-0.458 0-100.3
Air injection in rectangular channel 3.47 1 0-16.62 0-0.458 0-100.3
Atmospheric pressure flow boiling loop 9.53 1 0-2.955 - 0-100.3
Elevated pressure flow boiling loop 12.70 <1-10 0-3.815 - 0-167.9

results in an increased Reynolds number, permitting comparisons of the turbulence

characteristics between air injection experiments.

4.2.3 Elevated Pressure Boiling Loop

In the elevated pressure boiling loop, a gear pump with 87L min−1 flow rate was cho-
sen due to its tight seal to avoid leaks, the maximum operating temperature is 230◦C.
The loop itself was designed to operate at saturation conditions at pressures of up to
10bar.

The DC power supply has a limit of 14kW output, at a maximum voltage of 10V. The
limiting factor is the resistance of the heater rod, which depends on the material and
dimensions of the rod. Using Ohm’s law, and the readings from the power supply for
the voltage and current, the resistance of the 2m long rod is 0.015ohm, determined at a
wall temperature of 175◦C. By this, the maximum power output at 175◦C is 6700kW,
corresponding to a heat flux of 167.9kW m−2 . The higher heat output allows for boiling
at higher flow rates and thus higher Reynolds numbers can be achieved.

The experimental limits described in Table 4.1 are based on the equipment installed on
each loop, and the test section geometries. Taking account for hydrodynamic and heat
losses, the ranges will be reduced further, however, the amount reduced will vary.
76 Chapter 4. Experimental Apparatuses and Cases

4.3 Experimental Techniques

In the experiments performed, a combination of high speed video and particle image
velocimetry techniques were used in combination with temperature measurements
acquired via thermocouples embedded at discrete points in the flow loops.

The following instruments were used in the experimental work:

• Digital high-speed video camera, Phantom V310, capable of recording 3,250 frames
per second at a full resolution of 1280 x 800 pixels. It was used to capture the bub-
ble images and seed particles in PIV measurements. A halogen lamp was used
to provide adequate backlighting for high-speed video imaging of bubbles.

• Digital high-speed video camera, Photron FAST-CAM Mini UX100 with a micro-
scopic objective (used for elevated pressure boiling experiments)

• PIV system: LaVision FlowMaster 2-D PIV System with Litron Laser Model
NANO-S-35-15-PIV (Nd:YAG); Output: 65mJ, Wavelength: 532nm, Repetition
rate: 15Hz.

• 2-D PIV software package DaVis with a multi-pass cross-correlation processing

capability and automatic local adjustment of interrogation spot size, shape and
orientation based on local velocity gradients

• PIV particles: 10µm diameter hollow glass spheres with SG=1.05

• Type T thermocouples, placed at various points on the experimental loop, with

uncertainties of ±1.0◦C

• Digital Liquid Flow Meter (Great Plains Industries, G2S05N09GMA), with a 3.8-
37.9L min−1 flow range and ±2.0% accuracy (used for air injection loop)

For each experiment, variations of high speed videography or PIV are performed
to obtain either detailed bubble data or liquid flow field data, respectively. In cases
Chapter 4. Experimental Apparatuses and Cases 77

where the area of interest is very small, on the order of 10mm×10mm as in the region
very close to the heated surface or around a microbubble, a microscopic objective is
mounted onto the high speed camera to provide the necessary magnification.

Due to the construction of the site glass used on the high pressure flow loop, it is not
possible to create an optical correction box to permit PIV techniques for that series of
experiments. A comparison of high speed video and PIV exposures is shown in Figure
4.7.

F IGURE 4.7: Comparison of High Speed Video and PIV exposure for air
injection experiment

For temperature measurements, a combination of loop and wall-embedded thermo-

couples are used. Thermocouples are selected over other temperature measurement
instruments because of the low cost and high customization options available from
manufacturers. While not as accurate as resistance temperature detectors (RTDs), they
offer reasonable accuracy without the sensitivity of RTDs. The uncertainty of temper-
ature measurements is detailed in a later section.
78 Chapter 4. Experimental Apparatuses and Cases

4.3.1 High Speed Video

High-speed videography involves using the high-speed camera to record the bubble
shape and motion. In most cases, a 85mm macro lens was used; however, a 10× mi-
croscopic objective was used occasionally to obtain highly magnified views near the
heated surface. To achieve the highest quality images, ample backlighting was applied
using a halogen lamp and a high f-stop aperture setting, maximizing the depth of field
of the image, allowing us to capture as much detail as possible. The high-speed camera
has a maximum frame rate of 100,000 frames per second (fps), with a maximum full-
resolution image at 3200 fps. Because the image quality is very high, the high-speed
video is very useful in identifying interfaces and boundaries.

When investigating subcooled flow boiling at elevated pressures, PIV experiments

could not be performed. The reason for this is the geometry of the sight glass – which
featured two concentric components, a borosilicate inner glass cylinder and a polycar-
bonate outer cylinder for safety purposes in the event the inner cylinder should crack
or shatter. To perform PIV experiments, it would be necessary to match the indices
of refraction of all the components, which would be impossible to do without under-
mining the safety features of the sight glass. Taking note that for cases where the void
fraction was near zero, as seen in the ambient pressure cases covered, the liquid veloc-
ity profiles did not deviate significantly from purely single phase flow cases.

Instead, high speed video was employed to acquire magnified images of growing bub-
bles; a Photron FASTCAM Mini UX100 camera with a microscopic objective was used.
The microscopic objective makes a significant difference in measuring bubble diame-
ters at elevated pressures and flow rates. The images are then processed using soft-
ware supplied with the camera, Photron Fastcam Viewer v0.35. Specific bubble prop-
erties can then be measured: bubble size and position, and distance from the wall.
With these measurements, additional quantities can be derived, such as aspect ratio
Chapter 4. Experimental Apparatuses and Cases 79

and growth rate. Due to the reflections of the growing bubbles on the curved heated
surface, bubble contact angles cannot be measured with reasonable accuracy.

4.3.2 Particle Image Velocimetry

The PIV configuration is shown in Figure 4.8. The laser beam passes through a cylin-
drical lens, causing the beam to spread into a fan shape. The beam plane is adjusted
by rotating the cylindrical lens and focusing the beam until the plane is in the vertical
orientation and the thickness is reduced to a minimum. The high speed video camera
records the area illuminated by the beam plane from a perpendicular angle. A high
speed controller synchronizes the video camera and the pulsing of the laser, and all
parameters (pulse rate/duration/energy, image exposure time, etc.) are controlled us-
ing the commercial software package, DaVis. To maximize the quality of the image,
all ambient lighting should be turned off. The laser has two laser heads, each inde-
pendently timed and capable of 15Hz pulse rate. The maximum pulse rate of the laser
is then 30Hz. Density matching seed particles are then added to the liquid to allow
us to trace their movements between consecutive images. By tracing the displacement
of a cluster of seed particles and knowing the time difference between consecutive
images, the instantaneous velocity of the liquid can be determined. By repeating this
process for several clusters of particles, the instantaneous velocity at every point can
be determined, forming an instantaneous velocity field. Note that the 10x microscopic
objective was used in addition to the macro lens in the subcooled flow boiling experi-
ments to obtain more detailed PIV data near the heated surface.

In the annular air injection experiment, for single-phase turbulent flow below the bub-
ble injection point, 1,500 PIV images were processed. To obtain the ensemble-average
velocity profiles in the air injection experiments, at least 500 images were processed,
80 Chapter 4. Experimental Apparatuses and Cases

F IGURE 4.8: Laser configuration for PIV measurements

followed by applying a phase weighting procedure. Ensemble averaging measure-

ments from 500 images was determined to be sufficient to achieve proper measure-
ment statistics for both the laminar and turbulent cases covered. For the air injection
into the rectangular test section, the averaging was performed over 200 instantaneous
velocity fields, followed by the same phase weighting procedure applied in the annu-
lar channel case.

For the subcooled flow boiling experiment, the averaging was performed over 1,000
instantaneous velocity fields, and no phase weighting was applied, as the bubbles
were small enough that only the velocity vectors nearest the heated wall were affected.

The interrogation window size used by the DaVis software was changed depending
on the flow parameters and the time difference used between pairs of frames. For all
cases, two interrogation window sizes were used: a larger one for providing coarse ve-
locity field measurements, and reducing to a smaller one for refined and more detailed
velocity measurements.
Chapter 4. Experimental Apparatuses and Cases 81

4.3.3 Temperature Measurements

In boiling experiments, thermocouples are placed in the flow loop and embedded in
the heated wall to provide temperature measurements at key locations in the flow
loops. This allows the measurement of inlet and outlet temperatures at the test section
as well as the heat exchanger for regulating the temperature.

For the ambient pressure loop, thermocouples are installed at the test section inlet and
outlet. An additional thermocouple is embedded in the wall of the test section, but the
junction has been damaged by a shearing force from inserting the cartridge heater. De-
spite an inability to directly measure the wall temperature, Chen’s correlation can be
used to provide a rough estimate of the wall temperature. This allows for calculations
of boiling parameter that depend on the wall superheat, such as the Jakob number.

For the elevated pressure loop, thermocouples are similarly placed at the inlet and
outlet of the test section, as well as before and after the heat exchanger. Because
the experimental apparatus uses ohmic heating, no cartridge heater is inserted into
the inner rod, so thermocouples can be embedded without concern for damaging the
junction. In this instance, two wall-embedded thermocouples are installed, allowing
a redundant measurement of the wall temperature. To install them flush onto the
heater surface, the junctions are spot welded, and any gaps are filled with a brazing
technique.

4.4 Experimental Conditions Investigated

For the air injection experiments, multiple experimental conditions were considered
to best understand the effects of flow geometry and liquid flow rates on bubble size
and behavior. Those cases are categorized first by geometry, then by superficial liquid
82 Chapter 4. Experimental Apparatuses and Cases

and gas velocities. For the boiling experiments, cases are first identified by the exper-
imental loop used, followed by the pressure, then by flow rate. A labeling scheme is
also used to more easily identify each case is similar to the one used by Prodanovic et
al. [105]. The scheme adopted has the following format:

X(p)-NN

where

• X denotes the experiment type

– A for air injection in a annular channel

– R for air injection in a rectangular channel

– B for flow boiling in atmospheric pressure loop

– P for flow boiling in elevated pressure loop

• p denotes the pressure in bar, but is only used for the elevated pressure flow
boiling experiments

• NN indicates the specific case for above conditions

For example, case R-02 refers to experiment 02 for air injection in the rectangular chan-
nel, whereas case P03-02 refers to experiment 02 for boiling in an annulus at 3 bar
pressure.

Table 4.2 shows the experimental conditions for the air injection experiments. To best
compare turbulence characteristics between cases, the Reynolds number based on the
bulk liquid velocity and hydraulic diameter is added, as many correlations for bubble
behavior and heat transfer coefficients depend directly on it. Note that for experi-
ments performed in the rectangular geometry, the liquid properties were evaluated
at 88◦C, resulting in higher Reynolds numbers than would otherwise occur at room
temperature.
Chapter 4. Experimental Apparatuses and Cases 83

TABLE 4.2: Experimental conditions investigated for air injection

Dh < jf > < jg > < jg,inj >

Case Exp. Technique Rel
(mm) (m s−1 ) (m s−1 ) (m s−1 )
A-01 HSV/PIV 0.17 4.21E-4 0.72 1625
9.53
A-02 HSV/PIV 0.69 4.21E-4 0.72 6555
R-01 PIV 8.13 0.62 3.27E-4 0.17 7700
R-02 PIV 0.69 8.77E-4 0.17 1300
3.47
R-03 PIV 1.66 8.77E-4 0.17 3300

For the air bubble injection in annular geometry experiments, the bubbles were large
enough that they cast a shadow between the bubble and the inner rod, preventing
bubbles from being measured accurately in PIV exposures. For this reason, when the
experiment reached steady state, both PIV and high speed video were captured to
obtain bubble and flow field data, albeit with a delay between the acquisition times.
With air injection in the rectangular channel, only PIV data was acquired due to large
scale out-of-plane bubble motion.

Table 4.3 shows the experimental conditions for the subcooled flow boiling experi-
ments. In addition to the Reynolds number, useful boiling parameters are also re-
ported, such as the liquid subcooling, wall superheat, and heat flux. For the experi-
ments performed on the ambient pressure flow loop, the wall temperature used to cal-
culate the superheat is estimated from Chen’s correlation, given in Table 3.1. Chen’s
correlation is selected due to its widespread use in the literature [80, 105, 119].

The bubble growth rate for the ambient pressure flow boiling experiments was slow
enough that the bubble size could be measured in the PIV exposures, providing simul-
taneous bubble size and liquid flow field measurements for those cases. Unfortunately,
due to an inability to construct a correction box around the sight glass for the elevated
pressure loop, PIV measurements could not be made, and only high speed video mea-
surements could be made for those cases. Despite this, these cases provided highly
detailed vapor bubble data at elevated pressures.

In some measurements, particularly in the elevated pressure experiments, the average

84 Chapter 4. Experimental Apparatuses and Cases

TABLE 4.3: Experimental conditions investigated for subcooled flow boil-

ing

P < jf > ∆Tsub ∆Tw qw00

Case Exp. Technique Rel
(bar) (m s−1 ) (K) (K) (kW m−2 )
B-01 HSV 1 0.195 14 .0 5.4 74.6 5460
B-02 HSV 1 0.195 11.6 7.4 74.6 5570
B-03 PIV 1 0.260 11.6 7.4 74.6 7500
B-04 PIV 1 0.295 11.0 7.3 87.6 8510
B-05 MicroPIV 1 0.295 11.0 7.3 74.6 8610
B-06 PIV 1 0.370 8. 8 7.8 74.6 10900
B-07 PIV 1 0.590 13.0 7.2 74.6 16680
B-08 MicroPIV 1 0.590 10.3 10.3 74.6 17020
P01-01 HSV 1 0.664 15.5 2.4 84.5 24470
P01-02 HSV 1 0.664 7.3 2.4 40.9 26480
P01-03 HSV 1 1.328 3.0 1.4 40.9 55360
P02-01 HSV 2 0.664 6.4 10.8 46.5 34340
P02-02 HSV 2 0.664 4.6 1.8 83.0 34340
P02-03 HSV 2 0.664 9.1 16.8 82.1 34340
P03-01 HSV 3 0.664 4.5 11.5 80.7 38000
P03-02 HSV 3 1.328 5.2 1.5 80.8 76000
P05-01 HSV 5 0.664 10.8 25.1 43.7 42870
P05-02 HSV 5 1.328 12.4 10.2 43.4 85740
P07-01 HSV 7 0.664 10.3 9.6 49.4 46200
P10-01 HSV 10 0.664 7.2 18.1 55.5 49810

value as well as the standard deviation are presented. While great care is taken to only
consider bubbles that are isolated to minimize any effects from growing or entrained
bubbles in close proximity to the bubble of interest, there are instances where this is
unavoidable. Although the standard deviation is larger in such cases, a sufficient num-
ber of bubbles have been measured and the presented range of values is characteristic
of the behavior for that case.

4.4.1 Relevant scales for modeling

In turbulence, there are many scales that can be used to characterize the flow, depend-
ing on the eddy sizes and rate of rotation. The largest eddies are on the order of the
viscous boundary layer thickness, whereas the smallest eddy sizes are limited by the
fluid viscosity. The Kolmogorov scales are representative of the smallest possible eddy
Chapter 4. Experimental Apparatuses and Cases 85

sizes, where all of the energy transferred to these eddies is dissipated as heat [172].
For modeling purposes, the Kolmogorov scales are useful to determine the resolution
needed in numerical simulations. To this end, the Kolmogorov scales as defined in
Section 3.1 for all experimental cases are given in Table 4.4.

TABLE 4.4: Kolmogorov scales for experimental cases

 η τ vKol Rel
Case
(m2 s−3 ) (×10−5 m) (×10−5 s) (m s−1 ) (-)
A-01 0.5155 3.43 0 132 0.0261 1625
A-02 34.471 1.20 0 16.1 0.0745 6555
R-01 23.315 0.596 10.7 0.0559 7700
R-02 94.671 0.445 5.93 0.0749 1300
R-03 1318.2 0.230 1.59 0.1448 3300
B-01 0.7781 1.502 66.2 0.0227 5460
B-02 0.7781 1.477 65.4 0.0227 5570
B-03 1.8448 1.190 42.5 0.0280 7500
B-04 2.6938 1.075 35.0 0.0307 8510
B-05 2.6938 1.075 35.0 0.0307 8610
B-06 5.3151 0.891 24.6 0.0361 10900
B-07 21.551 0.649 12.5 0.0519 16680
B-08 21.551 0.639 12.4 0.0517 17020
P01-01 23.052 0.655 12.3 0.0532 24470
P01-02 23.052 0.613 11.8 0.0521 26480
P01-03 184.41 0.351 4.05 0.0865 55360
P02-01 23.052 0.523 10.6 0.0494 34340
P02-02 23.052 0.518 10.5 0.0492 34340
P02-03 23.052 0.533 10.7 0.0497 34340
P03-01 23.052 0.477 9.96 0.0479 38000
P03-02 184.41 0.286 3.54 0.0808 76000
P05-01 23.052 0.449 9.56 0.0470 42870
P05-02 184.41 0.270 3.40 0.0792 85740
P07-01 23.052 0.403 8.90 0.0453 46200
P10-01 23.052 0.381 8.57 0.0445 49810

With increasing Reynolds number the Kolmogorov microscales decrease, indicating

that to fully capture the turbulent phenomena, the number of elements in a compu-
tational domain must be increased. For two-phase flows, the deformation of bubbles
86 Chapter 4. Experimental Apparatuses and Cases

induces an additional localized turbulence that must also be considered when per-
forming numerical studies. The magnitude of this deformation is dependent on the
size of the bubble and the fluid properties. As part of this work, both the instantaneous
velocity field and the bubble growth and deformation are measured and expressed in
terms of an ensemble average to provide statistically significant results that can be
used for validation of numerical models.

4.4.2 Comparison of experimental conditions with previously pub-

lished data

To obtain a better understanding of how the present study compares with existing
datasets, Figure 4.9 shows the experimental conditions in subcooling number versus
Zuber number coordinates. Similarly, Figure 4.10 shows the experimental conditions
in subcooling number vs boiling number coordinates. Data from Prodanovic et al.,
Situ et al., Chu et al., and Sugrue et al. [80, 105, 116, 119], are presented.

F IGURE 4.9: Experimental conditions on the subcooling number versus

Zuber number coordinates
Chapter 4. Experimental Apparatuses and Cases 87

The subcooling number is a measure of the heat input required to bring the liquid
to saturation temperature, where as the Zuber number is a measure of the total heat
added to the liquid between the inlet and the measurement plane. Therefore, when
the subcooling number is greater than the Zuber number, Nsub > NZu , the system is
considered to be in the subcooled region; if Nsub ≤ NZu , the system is considered to be
in the saturated region. Additionally, the present cases are within the hydrodynamic
stability criteria described by Van der Hagen et al. [90]. Prodanovic et al.’s and Situ et
al.’s experimental conditions were well distributed between the subcooled and satu-
rated regions, while Situ et al.’s and Sugrue et al.’s data were mostly in the subcooled
region. The range of experimental cases in the present study overlaps partially with
the range of subcooling and Zuber numbers presented by Situ et al. [80], and is entirely
in the subcooled region.

F IGURE 4.10: Experimental conditions on the subcooling number versus

Boiling number coordinates

Likewise, Prodanovic et al.’s experimental data involved significantly higher boiling

numbers than the other studies, whereas the present experimental cases tend to have
lower boiling numbers. Situ et al.’s, Chu et al.’s, and Sugrue et al.’s experimental cases
88 Chapter 4. Experimental Apparatuses and Cases

fell in between. The reason for lower overall boiling numbers in the present study is
due to the higher mass fluxes covered in the experiments.

4.5 Post-Processing of Data

With the HSV and PIV images acquired, a highly involved procedure was followed
to transform the raw image data into a usable, quantified format. In both cases, im-
ages were analyzed frame by frame typically using an algorithm, however, in some
cases measurements must be performed manually in a tedious manner. Among the
measurements that can be extracted from these images are the bubble dimensions and
position for high speed video, and the instantaneous liquid flow field for PIV. Addi-
tional calculations can be performed to obtain derived quantities, such as an estimated
bubble volume, aspect ratio, or velocity from the high speed video images, or in-plane
Reynolds stresses and out-of-plane vorticity from the PIV exposures.

Calculation of Bubble Size, Shape and Orientation Due to the highly complex na-
ture of the bubble generation process and liquid-vapor interactions, the bubble size
and shape can vary greatly during its lifespan. In this case, bubbles are not necessarily
spherical, and experience very strong oscillations in shape after departure or coales-
cence. To quantify this type of behavior, some parameters are defined as follows:

Centroid is the point where the sum of all the displacements of the points on the
interface from the centroid is zero. For a spherical bubble, the centroid would
correspond to the geometric center.

Major Axis is the chord passing through the centroid with the longest length

Minor Axis is the chord passing through the centroid with the shortest length
Chapter 4. Experimental Apparatuses and Cases 89

F IGURE 4.11: Calculation of bubble size, shape, and position

Major/Minor Axis Angle is the angle that the major/minor axis forms with respect
to the horizontal. This angle, coupled with major/minor axis lengths allows the
visualization of size, shape and orientation of the bubble.

Bubble Aspect Ratio is the ratio of the vertical length to the horizontal width of the
bubble. These values are calculated using the axis lengths and their orientation
angles

When images are obtained, a MATLAB script performs post-processing to analyze the
bubble shape and position as a function of time. The procedure for calculating the
bubble centroid and major and minor diameters is shown in Figure 4.11.

When calculating the bubble dimensions, high speed videography is preferred due to
the high contrast in the bubble images as well as being able to track the motion of a sin-
gle bubble over a period of time. The high contrast makes it easier to extract the bubble
contours using software. The level of user input required to perform the calculations
90 Chapter 4. Experimental Apparatuses and Cases

is limited to clicking the interior of the bubble for which dimensions are desired; as
there are multiple moving bubbles, it is the user’s task to identify the bubble of inter-
est between consecutive frames. For ensemble averaging of bubble dimensions, the
dimensions are calculated for each bubble as a function of time, starting at the onset
of injection, and at each timestep the dimensions of multiple bubbles are ensemble
averaged.

For the boiling experiments, the bubble does not experience the large scale oscillations
seen in air injection experiment, and can be treated as elliptical. When measuring
the bubble sizes, the bubble diameter is taken as the square root of the product of
the horizontal (dx ) and vertical (dy ) lengths as shown in equation (4.1), similar to the
method used by Okawa [110]. With these measurements, additional quantities can be
derived, such as aspect ratio and growth rate. Due to the light reflections from the
growing bubbles on the curved heated surface, bubble contact angles could not be
measured with reasonable certainty.

p
db = dx dy (4.1)

As the bubble lifts off from the heated surface, the radial position of the bubble cen-
troid is measured. The horizontal distance between the heated wall and the innermost
edge of the bubble is measured and added to half the horizontal width of the bubble
to get the radial position. Prior to lift off, the radial position is considered to be zero
to discern between bubble growth while attached and motion. The radial position is
taken as: 
 0

t < td
rb = (4.2)
dx
 r+

2
t ≥ td

Typically, the bubble shape near its maximum diameter is crisp as in Figure B.7, how-
ever for newly nucleating bubbles or bubbles on the verge of collapsing, the ability to
Chapter 4. Experimental Apparatuses and Cases 91

identify the edge diminishes. The uncertainty of distance measurements is then taken
as ∆r = SF · 2pixels, varying with the scale factor, SF , as calculated from the image
calibration for the image set.

Flow Field Vector Calculations Once the PIV laser exposures are acquired, the im-
ages are calibrated using the procedures described in Section B.2.1. With the scale
factor and origin defined, the PIV data processing can be performed. The procedure
follows a series of steps: preprocessing of the image, definition of a mask, vector cal-
culation, and vector postprocessing. The preprocessing step is used to adjust image
parameters to increase the overall quality of the image to reduce errors from the cal-
culation process; this can include increasing the contrast, inverting the image, or sub-
tracting an averaged image to reduce motion effects. The mask can be defined to
include or exclude specific elements from the vector calculation, either by geometric
selection or by setting threshold values for the pixel intensity.

The vector calculation is the most important and computationally intensive process.
The DaVis software divides the image into several interrogation windows, detects peaks
in pixel intensity in each window, and then performs a cross correlation with the corre-
sponding window from the subsequent image to obtain a vector for the pixel displace-
ment between images. Coupled with the scale factor and time difference between the
images, the instantaneous velocity vector is calculated for the center of the interroga-
tion window. There is an option to use multiple interrogation windows of reducing
size, such that the average of the vectors in smaller interrogation windows is equal to
the vector calculated in the larger. This results in higher overall accuracy, but takes
significantly longer time.

Vector postprocessing allows for smoothing of vector clusters to remove potentially

erroneous results as well as the interpolation of vectors in areas where the particle
density is low.
92 Chapter 4. Experimental Apparatuses and Cases

The estimated uncertainty in velocity measurements from the PIV software varies with
the magnification of the image and the time difference between pairs of images. The
velocity uncertainty is given by the following equation:

∆x
∆v = (4.3)
∆t

where ∆x is the uncertainty in identifying the exact center of a PIV seeding particle
(0.1 pixel multiplied by the scale factor), and ∆t is the time difference between frames.

For the experiments, the interrogation window size used by the DaVis software was
changed depending on the flow parameters and the time difference used between
pairs of frames. For all cases, two interrogation window sizes were used: a larger
one for providing coarse velocity field measurements, and reducing to a smaller one
for refined and more detailed velocity measurements.

After the calculation of instantaneous velocity fields, additional processing options be-
come available to calculate the average and RMS fluctuations, as well as turbulent and
flow parameters such as in-plane Reynolds stresses and out-of-plane vorticity fields.

Phase Weighted Ensemble Averaging of PIV Data In PIV measurements of air injec-
tion, a laser light sheet was directed perpendicular to the injection surface and the air
bubbles and PIV particles were illuminated at the same time. Instantaneous velocity
profiles were obtained, and phase fraction-averaging of velocity data was performed
to obtain the ensemble-averaged velocity profile and root mean square (rms) fluctua-
tions of the velocity. This averaging procedure is similar to the use of a hot wire probe
to simultaneously measure the phase fraction and liquid velocity in bubbly flow pre-
viously performed by Michiyoshi and Serizawa [173–175]. For their experiments, the
velocity signal was detected only when the probe was in the liquid phase, so that the
ensemble-average velocity was based only on the instantaneous liquid velocity data.
Chapter 4. Experimental Apparatuses and Cases 93

To perform this type of phase fraction averaging on PIV data, bubbles were isolated in
raw PIV exposures and correlated with the processed PIV data to remove any contri-
butions from vectors within the boundaries of a bubble.

The procedure for performing phase-weighted, ensemble averaging of velocity pro-

files overlaps with the calculation of bubble dimensions. For each instantaneous ve-
locity profile, if there are bubbles present, the bubble contours are extracted and any
velocity vectors inside the contours do not contribute to the averaging. The process of
extracting contours differs from the way the contours are extracted from high speed
video frames due to the low contrast of PIV exposures, noting that the laser sheet im-
pinging on a bubble casts a shadow, unlike in uniform backlighting. Because of the
low or inconsistent contrast levels, the boundaries of bubbles needs to be input man-
ually by the user clicking on points along the shape of the bubble, forming a close
estimate of the bubble contour.

Correlation Development Using the data from the experimental databases for the
air bubble injection and subcooled flow boiling experiments, multiple correlations are
proposed to aid in the prediction of a number of bubble parameters and behaviors. For
the air bubble injection experiments, correlations for the Bond number, which scales
with bubble size, and the bubble aspect ratio, which defines the bubble shape, are
proposed. For elevated pressure subcooled flow boiling experiments, correlations are
proposed for the bubble liftoff diameter and time, and bubble history numbers before
and after liftoff.

To develop the correlations, first a form of the correlation is suggested, typically a

power law, which is most common in the literature. The power law contains the rel-
evant parameters that can be measured using the experimental apparati, with expo-
nents starting at -5, which is sufficiently highest than the highest magnitudes seen
in the literature. The multiplicative coefficient is permitted to span a larger range if
94 Chapter 4. Experimental Apparatuses and Cases

needed. Exponents are then incremented on a per parameter basis to obtain the entire
range of equations with exponents between -5 and 5, with intervals of 0.1. The mean
average error (MAE) is then calculated for each equation, and the equation that yields
the minimum is highlighted. To further refine the correlation, the process is repeated
between the exponents from the best performing correlation ±0.1, with an interval
0.001. This method differs from the method of steepest descent, which requires less
computational effort, but does not guarantee the overall minimum value in the span.
By performing the refining step, there is an increased precision in the correlation, with-
out significantly increasing the computational time.

4.5.1 Experimental Uncertainties

The uncertainty of the velocity and distance measurements is highly dependent on

the frame acquisition rate and the magnification of the viewing area. For PIV exper-
iments, the motion of the seeding particles between two consecutive images is used
to determine the local instantaneous velocity. The uncertainty of these measurements
is not easily quantified, as the efficacy of the peak detection algorithm varies with the
software package as well as with image quality. For this application, a rudimentary
calculation can be made, stemming from the uncertainty in the position of a seed par-
ticle, which is a function of the image resolution and magnification. With increasing
acquisition rate or scale factor, the uncertainty of the measurements also increases.

The processing parameters and measurement uncertainties for high speed video and
PIV experiments are presented in Table 4.5.
Chapter 4. Experimental Apparatuses and Cases 95
Interrogation Interrogation
Velocity
Experiment Scale Factor ∆t Recording Rate Window Overlap (1) Window Overlap (2)
Case Uncertainty
Type Size (1) Size (2)
mm/px µs FPS N×N px % N×N px % m s−1

Single Phase PIV 0.015 50-200 15 32 25 8 50 0.075-0.3

A-01 PIV 0.014 200 15 16 50 16 50 0.069

A-01 HSV 0.020 - 4000 - - - - 0.080

A-02 PIV 0.017 100 15 24 25 8 50 0.174

A-02 HSV 0.018 - 4000 - - - - 0.072

R-01 PIV 0.009 50 15 32 50 12 25 0.182

R-02 PIV 0.013 50 15 16 75 16 75 0.252

R-03 PIV 0.010 50 15 32 0 8 0 0.190

B-01 HSV 0.012 - 30 - - - - <0.001

B-02 HSV 0.009 - 1000 - - - - 0.009

B-03 PIV 0.013 100 15 32 50 16 50 0.131

B-04 PIV 0.014 60 15 32 50 16 50 0.234

B-05 MicroPIV 0.002 50 15 48 50 24 50 0.040

B-06 PIV 0.014 60 15 32 50 16 50 0.233

B-07 PIV 0.013 100 15 32 50 16 50 0.131

B-08 PIV 0.014 60 15 32 50 16 50 0.233

B-09 MicroPIV 0.002 50 15 48 50 24 50 0.040

P01-01 HSV 0.006 - 500 - - - - 0.003

P01-02 HSV 0.009 - 4000 - - - - 0.036

P01-03 HSV 0.009 - 4000 - - - - 0.036

P02-01 HSV 0.006 - 4000 - - - - 0.025

P02-02 HSV 0.009 - 4000 - - - - 0.034

P02-03 HSV 0.003 - 1000 - - - - 0.003

P03-01 HSV 0.003 - 4000 - - - - 0.014

P03-02 HSV 0.003 - 8000 - - - - 0.023

P05-01 HSV 0.004 - 4000 - - - - 0.016

P05-02 HSV 0.003 - 4000 - - - - 0.012

P07-01 HSV 0.004 - 100 - - - - <0.001

P10-01 HSV 0.004 - 100 - - - - <0.001

TABLE 4.5: High speed video and PIV recording parameters and uncer-
tainties

Using the experimental apparati and techniques described here, the experimental databases
for air bubble injection and subcooled flow boiling results were constructed, spanning
a wide range of conditions many of which were not previously investigated in other
works. Furthermore, post-processing procedures are applied to aid in the quantifica-
tion of the experimental data, much of which is collected in the form of images and
can be subject to much interpretation.
Chapter 5

Results and Discussion

5.1 Single-Phase Velocity Profiles

A series of preliminary adiabatic PIV experiments were performed using the atmo-
spheric pressure boiling loop. As the test section geometry is the same between the
air injection and flow boiling loops, the velocity profiles obtained under single phase
adiabatic conditions should persist between cases. The single-phase time-averaged
velocity profiles and RMS fluctuations are shown in Figures 5.1a and 5.1b.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.1: Single-phase time-averaged PIV results

For lower superficial liquid velocities, a reasonable comparison can be made with the
theoretical laminar velocity profile. A comparison of the single-phase velocity for case

96
Chapter 5. Results and Discussion 97

F IGURE 5.2: Comparison of single-phase measured and theoretical veloc-

ity profiles (case A-01)

A-01 with the corresponding theoretical profile is shown in Figure 5.2.

The measured velocity profile has generally good agreement in the core, however it is
slightly higher than expected. This discrepancy can be attributed to light reflections off
of the inner wall that illuminate particles out of plane. These particles have a higher
velocity than the particles close to the wall, resulting in a slightly higher measured
average velocity. To overcome this problem, the inner rod must be coated with a non-
reflective paint, however this would affect the heat transfer and surface properties.

With increasing axial velocity, there is a clear transition of the axial profile from parabolic
to flat, which is typical as the fluid transitions from laminar to turbulent. Similarly, as
the Reynolds number increases, the magnitude of the velocity fluctuations scale with
it. While the fluctuation magnitude is a metric of the overall turbulence in the flow,
another useful comparison can be made by applying the nondimensionalization pro-
posed in Section 3.1 and comparing with the law of the wall, given by equation (3.18).
All of the time-averaged nondimensional velocity profiles are shown in the law of the
wall coordinates in Figure 5.3.
Chapter 5. Results and Discussion

F IGURE 5.3: Dimensionless axial velocities and fluctuations compared with the law of the wall
98
Chapter 5. Results and Discussion 99

The lower velocity data tend to overshoot the theoretical profile, due to the lower cal-
culated friction velocities. Similarly, higher velocities have generally good agreement
near the wall, with better results in the turbulent core.

With the single-phase velocity profiles established the impact of introducing air bub-
bles can be properly assessed. It is expected that the velocity near the wall will experi-
ence deviations due to the strong influence of bubble injection on the liquid boundary
layer. The deviation from the single-phase velocity can be examined to quantify the
impact of the bubbles rising on the liquid flow. Depending on the relative velocity
between the bubble and the liquid, which will vary with the bubble size and radial
position, the time-averaged velocity profiles and fluctuation magnitudes are expected
to change to incorporate the bubble motion.

5.2 Air Bubble Injection Experiments

5.2.1 Annular Test Section Data

5.2.1.1 Effect of Air Bubble Injection on Liquid Velocity

The ensemble-averaged velocity profiles and RMS fluctuations for laminar flow (case
A-01) at the injection hole are shown in Figures 5.4a and 5.4b. The injection of air
bubbles caused considerable increases in both streamwise and spanwise velocity fluc-
tuations near the inner tube wall, and as far as the center of the annulus. Also, be-
cause the ensemble-average velocity near the wall increased, the mean velocity near
the center of the annulus decreased from 0.25m s−1 to 0.20m s−1 . Further above at y
= 3.98mm, the effect of bubble injection was more uniformly distributed as shown in
Figures 5.5a and 5.5b for the same flow conditions. In these Figures, the single-phase
velocity profile is also shown for comparison purposes.
100 Chapter 5. Results and Discussion

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.4: PIV measurements near the injection point (z = 0.07mm, case
A-01)

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.5: PIV measurements above the injection point (z = 3.98mm,

case A-01)

The velocity data at the injection point elevation can be nondimensionalized to com-
pare with the law of the wall as shown in Figure 5.6.
Chapter 5. Results and Discussion 101

F IGURE 5.6: Nondimensional velocity profiles near the injection point (z

= 0.07mm, case A-01)

The effect of the bubble on the nondimensionalized velocity profile is significant, even
at a radial position far away from the injection point. Similarly, the magnitude of the
fluctuations is large and persists up to a third of the channel gap. In the core, there is a
discrepancy between the measured and expected profiles, this is due to the relatively
low Reynolds number of the flow (Re =1625). Despite the discrepancy, the effect of
the bubble is further emphasized because of the semi-logarithmic scale.

The effect of bubble injection on turbulent flow of liquid is shown in Figures 5.7a and
5.7b (at the bubble injection point) and Figures 5.8a and 5.8b (at 4.40mm above the bub-
ble injection point) for an average air injection rate of 5.4mL min−1 . As in the laminar
flow case, the bubble injection increased the fluctuations in the streamwise and radial
velocities near the inner tube and up to the center of the annular gap. The mean ve-
locities at both elevations were, however, reduced in contrast to the laminar flow case.
This is because the bubble rise velocity was less than the liquid velocity, so the injected
bubbles slowed down the local liquid flow near the inner tube wall. Also, the stream-
wise velocity fluctuations appear to be amplified at the higher elevation contrary to
the laminar flow case.
102 Chapter 5. Results and Discussion

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.7: PIV measurements near the injection point (z = 0.02mm, case
A-02)

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.8: PIV measurements above the injection point (z = 4.40mm,

case A-02)

The same nondimensionalization can be applied to the axial velocity in the turbulent
case. The resulting nondimensional profile is shown in Figure 5.9, indicating good
agreement with the law of the wall, which is to be expected given the higher Reynolds
number. It can be seen that the bubble significantly affects the flow around the injec-
tion point, causing localized backflow in some cases; this effect is confined to a small
area, and diminishes quickly.
Chapter 5. Results and Discussion 103

F IGURE 5.9: Nondimensional axial velocity profile near the injection

point (z = 0.02mm, case A-02)

5.2.1.2 Air Bubble Size, Shape and Position

From the high-speed video images of the injected bubbles, the shape and size of the
bubbles were determined during their growth and after their detachment from the in-
jection hole. The bubble images were also processed to determine the following bubble
characteristics: the bubble centroid position, the major/minor axes of the bubble, and
the orientation of the major/minor axes with respect to the horizontal direction point-
ing away from the inner tube surface. For a laminar flow case, bubble coalescence was
frequently observed as shown in Figure 5.10.

For the particular sequence shown in Figure 5.10, the bubble centroid trajectory in
the streamwise and spanwise directions, the major and minor axes and their orienta-
tions before and after the bubble coalescence are shown in Figures 5.11a, 5.11b and
5.11c. The uncertainties in the bubble centroid position and major/minor axis values
indicated by error bars were calculated from the pixel span of the bubble interface
multiplied by the magnification factor.
Chapter 5. Results and Discussion

F IGURE 5.10: Bubble coalescence after injection, case A-01

104
Chapter 5. Results and Discussion 105

( A ) Trajectory shift ( B ) Axis lengths

( C ) Axis orientations ( D ) Aspect ratio

F IGURE 5.11: Trajectory, size, and shape for coalescing bubbles, case A-01

The coalescence of consecutive bubbles occurred at t = 10ms as indicated by large

shifts in position and size as shown in Figures 5.11a and 5.11b. Before coalescing, the
injected bubble grew in size, steadily increasing in both the major and minor axes. As
the bubble grew large enough to catch up with the tail of the preceding bubble, the
two bubbles coalesced.

By first taking the projection of the major and minor axes in the horizontal and ver-
tical directions, and then taking the ratio of these projected lengths, the aspect ratio
can be obtained. The aspect ratio for the coalescing bubble seen in case A-01 is shown
in Figure 5.11d. At incipience, the bubble is near spherical, becoming slightly more
106 Chapter 5. Results and Discussion

elongated due to the shearing effect from the liquid. Upon coalescence, the bubble be-
comes more elongated, then experiences large scale fluctuations in shape with changes
in aspect ratio having a magnitude of 0.3 at a frequency of roughly 160Hz.

The air bubbles injected into a turbulent liquid stream departed faster and at a higher
frequency due to a higher liquid velocity as shown in Figure 5.12. They did not coa-
lesce, and tended to be near spherical, with small, high frequency oscillations in shape
and size, and becoming increasingly spherical with time.
 Chapter 5. Results and Discussion

F IGURE 5.12: Bubble motion after injection, case A-02

107
108 Chapter 5. Results and Discussion

The ensemble-averaged streamwise and spanwise trajectories of 10 bubbles injected

into turbulent liquid stream (case A-02) are shown in Figure 5.13a. The ensemble-
averaged major and minor axes and their orientations for the turbulent flow case are
shown in Figures 5.13b and 5.13c.

( A ) Bubble trajectory ( B ) Axis lengths

( C ) Axis orientations ( D ) Aspect ratio

F IGURE 5.13: Ensemble averaged trajectory, size, and shape (10 bubbles,
case A-02)

The ensemble averaged bubble aspect ratio for case A-02 is shown in Figure 5.13d.
Again, at incipience, the bubble is near spherical, becoming slightly more elongated
prior to departure. Upon departure, the bubbles begin a rapid oscillation. The ampli-
tude of these oscillations is about ±0.05, which is much smaller than in the laminar
case. The frequency of the oscillations was about 1.5kHz. The significant increase in
the oscillation frequency is due to the generally spherical shape of the bubbles.
Chapter 5. Results and Discussion 109

In the air bubble injection experiments described above, a syringe pump was used to
inject the air at a constant volumetric flow rate, however, there was intermittency ob-
served in the volume of air coming out of the injection hole on the inner tube recorded
by a high-speed video camera. Since the actual volumetric flow rate of air injected into
the liquid stream may not be constant in time, the volume of the air bubble emerging
from the injection hole and growing with time before departure was estimated frame
by frame as shown in Figure 5.14.

F IGURE 5.14: Instantaneous air injection rate estimated from the variation
of the bubble volume at the injection hole, case A-02

The air bubbles formed, grew and departed at a frequency of about 400Hz in turbulent
flow. For a given bubble, the instantaneous air injection rate reached a peak value
before falling to zero which corresponds to the bubble departure. After the bubble
departed, there was a period of zero air flow out of the injection hole until the next
bubble started to form and grow at the injection hole.

Although the time-averaged air injection rate obtained from the instantaneous injec-
tion data shown in Figure 5.14 agreed well with the syringe pump setting, the instan-
taneous air flow rate through the injection rate was not constant. The oscillation in
the instantaneous air flow rate is attributed to the pressure build up upstream of the
hole needed to overcome the surface tension force and push the gas-liquid interface
110 Chapter 5. Results and Discussion

into the liquid stream. When a fully-grown bubble departs, the gas-liquid interface is
likely pushed back into the injection hole, which has a small radius (0.4mm), so the
radius of curvature of the interface becomes as small as the hole radius. The surface
tension force would then resist the expansion of the bubble into the liquid stream until
the air pressure upstream of the injection hole builds up sufficiently to overcome the
surface tension force.

For numerical simulations, the air flow rate through the injection hole and into the liq-
uid stream is an important boundary condition and needs to be properly specified. In
view of the data shown in Figure 5.14, a constant air flow rate or constant air pressure
upstream of the injection hole would not be a suitable boundary condition to be used
in numerical simulations.

5.2.2 Rectangular Test Section Data

5.2.2.1 Effect of Air Bubble Injection on Liquid Velocity

A sample image of water with seed particles flowing upward past air bubbles injected
into the rectangular minichannel is shown in Figure 5.15. A total of 200 instantaneous
velocity profiles were obtained each at a distance of z = 0.3mm and 3.0mm above the air
bubble injection hole (z = 0 is at the top edge of the injection hole), and phase fraction-
weighted averaging of velocity data was performed to obtain the ensemble-average
velocity profile and root mean square fluctuations of the velocity.
Chapter 5. Results and Discussion 111

F IGURE 5.15: Sample PIV exposure for rectangular geometry, case R-01

The purpose of these experiments is to investigate the effect of geometry on liquid flow
and bubble behavior. In considering the impact of geometry on bubble behavior, the
confinement number, Co, is a useful parameter. Kew and Cornwell [176] suggested
that two-phase flows exhibit different flow and heat transfer behavior at confinement
numbers greater than 0.6, Co > 0.6. In the present experiments, the confinement
number was 0.37 in the larger rectangular channel, and 0.87 in the smaller channel.
For comparison, the confinement number was 0.35 in the annular air injection exper-
iments, 0.32 in the atmospheric flow boiling loop experiments, and even lower for
elevated boiling experiments. Cases R-02 and R-03 have confinement numbers that
exceed the threshold proposed by Kew and Cornwell, and these cases should exhibit
differing flow behavior.

For case R-01, the full sized rectangular channel with a height of roughly the same size
as the gap in the annular test section was used. The differences here are the properties
of the liquid, which has been heated to a temperature of 88◦C. For water, the surface
tension difference between 25 and 88◦C is about 20%. There is also a slight reduction in
the liquid density, however, the biggest difference is in the liquid viscosity decreasing
from 0.9 to 0.3 mPa s from 25 to 88◦C. The reduction in viscosity is significant enough
to permit higher Reynolds numbers than the air injection loop is capable of providing,
112 Chapter 5. Results and Discussion

despite a lower flow rate and hydraulic diameter. The time-averaged velocity pro-
file and RMS fluctuations at the injection point are shown in Figures 5.16a and 5.16b,
respectively. Although the superficial velocity in this case is 0.62m s−1 the measured
average velocity is higher, due to the wall effects in the out of plane direction. The
measured velocity has a smooth parabolic profile, however the velocity near the wall
does not reduce to zero, due to motion of the bubble interface.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.16: Ensemble-averaged PIV measurements near the injection

point (z=0.03mm, case R-01)

At an elevation of 3.00mm above the injection point, the average velocity and RMS
fluctuation profiles are given by Figures 5.17a and 5.17b. There is an increase in the
average velocity near the wall attributable to the bubble rising motion. Further in the
center there is a peak in the RMS fluctuation due to the viscous drag of the bubbles
acting on the liquid phase.
Chapter 5. Results and Discussion 113

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.17: Ensemble-averaged PIV measurements above the injection

point (z=3.00mm, case R-01)

For a fully laminar flow at Re = 1300 in the smaller minichannel, case R-02, the ensemble-
average streamwise and spanwise velocities and RMS fluctuations at z = 3.0mm above
the injection point are shown in Figures 5.18a and 5.18b. Here, the effects of the bubble
are minimal, as the liquid and bubble velocities are nearly equal, effectively making
the bubbles seem stationary when compared to the liquid flow.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.18: Ensemble-averaged PIV measurements above the injection

point (z=3.00mm, case R-02)
114 Chapter 5. Results and Discussion

For flow in the smaller minichannel at a higher flow rate such that Re = 3300, case R-
03, the ensemble-averaged streamwise and spanwise velocity profiles and RMS fluc-
tuations are shown in Figures 5.19a and 5.19b obtained at z = 3.0mm above the in-
jection point. As in the larger minichannel case, the bubble injection increased the
velocity fluctuations near the center of the channel. However, the effect of the bub-
bles is much more dominant on the streamwise velocity, where a distinct increase in
ensemble-average velocity near the center can be seen.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.19: Ensemble-averaged PIV measurements above the injection

point (z=3.00mm, case R-03)

The effect of the increased confinement number is not readily evident from the PIV
data. Although the effect of the bubble on the liquid flow can be quantitatively mea-
sured, the most significant impact of the higher confinement number of the smaller
channel size on the liquid velocity field occurs in the out-of-plane direction due to the
large width-to-gap ratio of the channel and cannot be measured using this configura-
tion.
Chapter 5. Results and Discussion 115

5.2.2.2 Air Bubble Size, Shape and Position

From the high-speed video images of the injected bubbles, the shape and size of the
bubbles were determined during their growth and after their detachment from the in-
jection hole. For the turbulent flow case in the larger minichannel, high speed video
imaging was used to obtain data for ten bubbles at varying times and an ensemble
average was taken. The bubbles injected did not coalesce, and tended to be nearly
spherical in shape, with small, high frequency oscillations in shape and size, becom-
ing increasingly spherical with time. The ensemble-averaged streamwise and span-
wise trajectories of 10 bubbles (Re = 7700) are shown in Figure 5.20a. The ensemble-
averaged major and minor axes and aspect ratio for the turbulent flow case are shown
in Figures 5.20b and 5.20c.

Bubbles for cases R-02 and R-03 experienced out of plane oscillatory motions, which
made them difficult to measure over a period of time. For both cases, while air was be-
ing injected, the bubble expanded to fill the entire gap, and once detached the bubble
was elongated in the axial direction. In case R-02 the bubble persisted with a highly
elongated contiguous shape; in case R-03, the duration of bubble growth was short-
ened by the liquid stream to produce smaller but still elongated bubbles. Sample im-
ages taken from cases R-02 and R-03 identifying the boundaries and axis lengths of
bubbles are shown in Figure 5.21.

( A ) Bubble trajectory ( B ) Axis lengths

116 Chapter 5. Results and Discussion

( C ) Aspect ratio

F IGURE 5.20: Ensemble-averaged trajectory, size, and shape (10 bubbles,

case R-01)

( A ) Case R-02 ( B ) Case R-03

F IGURE 5.21: Sample bubbles for cases with high confinement number,
Co

For a comparison of bubble behavior between these cases, there is a characteristic bub-
ble size associated with the flow parameters. This characteristic bubble size can be
used to estimate Bond and Weber numbers to better provide a quick comparison of
the bubble dynamics between cases. Similarly, an equilibrium bubble aspect ratio af-
ter departure can be determined. While correlations from Wellek [169] and Okawa et
al. [94] aim to predict bubble shape via the aspect ratio, the bubble shape is highly
dependent on the conditions under which the bubble is injected (i.e. vertical injection
Chapter 5. Results and Discussion 117

by a sparger or horizontal injection through an injection hole).

Table 5.1 lists the confinement number, characteristic bubble diameter and variance,
aspect ratio, liquid to gas superficial velocity ratio, Bond number, Weber number and
liquid Reynolds number for each case. A trend emerges with the bubble diameter re-
ducing with Reynolds number, and with reducing bubble diameter for fixed geometry,
the aspect ratio more closely approaches unity indicating a spherical shape.

TABLE 5.1: Summary of results for air injection experiments

√
Db,avg varb,avg <jf >
Case Co Ra,avg <jg > Boavg W eavg Rel
(mm) (mm)
A-01 0.35 1.75 0.07 0.87 403.8 0.304 1.534 1625
A-02 0.35 0.94 0.02 1.02 1639 0.082 4.336 6555
R-01 0.37 0.87 0.05 0.72 1896 0.084 15.38 7700
R-02 0.87 1.56 0.11 0.57 786.8 0.270 70.59 1300
R-03 0.87 0.98 0.08 0.53 1893 0.106 160.1 3300

A rudimentary evaluation of correlations from Wellek [169] and Okawa et al. [94] can
be performed using the data from Table 5.1. Their correlations were aimed at predict-
ing the lower and upper limits of bubble aspect ratio for bubbles injected vertically
into a static liquid. Figure 5.22 shows experimental bubble shape data compared with
the correlations for the bounding aspect ratios given in the literature.

As expected, the correlations from Wellek [169] and Okawa et al. [94] generally do not
predict well the bubble aspect ratio where bubbles are injected normal to the flow. Ad-
ditionally, these correlations do not consider the channel geometry or liquid Reynolds
number. As the confinement number increases, there is an evident elongation of the
bubble.

Using the parameters from Table 5.1, predictive correlations can be developed for the
bubble Bond number and aspect ratio that take into account geometry, flow and fluid
properties. The Bond number is the ratio of the body forces acting on the bubble to the
surface forces, and it scales with the bubble diameter. Using a numerical fit, separate
118 Chapter 5. Results and Discussion

F IGURE 5.22: Comparison of experimental data correlations from Wellek

and Okawa et al.

correlations can be developed to predict the average bubble diameter and aspect ratio.
The resulting correlations are given by equations (5.1-5.2).

The data that the correlations are suitable for is limited to the air bubbles injected
through a hole perpendicular to a vertical upward liquid stream. The correlations were
developed using data covering 50 bubbles, spanning the 5 cases. The mean average
error (MAE) of equations (5.1) and (5.2) compared with the present data are 6% and
3%, respectively. Comparisons of the correlations with obtained experimental data are
shown in Figures 5.23a and 5.23b.

 −1.62
(ρf − ρg )gDb,avg < jf >
Boavg ≡ = 38.1Rel0.78 Co0.94 (5.1)
σ < jg >

 3.31
< jf >
Ra,avg = 1.53Rel−3.27 Co−3.60 (5.2)
< jg >

It was determined that the average bubble Bond number and shape varies with the
combination of liquid Reynolds number, Rel , confinement number, Co, and the ratio of
Chapter 5. Results and Discussion 119

liquid velocity to gas velocity, < jf > /< jg >. For the bubble shape, the confinement
number plays a more significant role in the elongation; for both liquid and gas to
flow in an increasingly narrow channel, the gas assumes a more elongated shape to
minimize drag. Note that in cases where the liquid is static, bubbles are elongated
perpendicular to the motion, and the aspect ratio is the bubble height divided by the
width. Here, the bubbles are observed to be elongated parallel to the direction of
motion, and thus the aspect ratio is the width divided by the height.

( A ) Bond number ( B ) Aspect ratio

F IGURE 5.23: Comparison of developed correlations with experimental

data

5.3 Subcooled Flow Boiling Experiments

Flow boiling experiments were performed on two separate flow loops, one operating
at atmospheric pressure and the other operating at pressures up to 10bar. The atmo-
spheric pressure loop is primarily used to perform PIV experiments to obtain a better
understanding of liquid vapor interactions and the effect of temperature gradient on
liquid flow patterns. The elevated pressure loop was used to perform photographic
studies on bubble size and behavior during growth and after departure. PIV experi-
ments could not be performed on the elevated pressure loop, however flow field data
120 Chapter 5. Results and Discussion

from PIV experiments can be analyzed in conjunction with bubble size and behavior
data to provide a more complete picture of two-phase flow.

5.3.1 Atmospheric Pressure Experiments

To measure the effect of attached vapor bubbles on the liquid flow near the wall, high
speed video measurements have been performed to analyze the growth of several
nucleating bubbles as seen in cases B-01 and B-02, as well as several PIV experiments
to measure the flow field near the wall around a single, isolated vapor bubble in cases
B-03 to B-09. To better discern the effects of altering flow and boiling parameters on the
bubble size at departure, additional experiments were performed using the elevated
pressure flow loop, which can provide more control over heating as well as direct wall
temperature measurements.

A sample high speed video image of a collection of vapor bubbles for case B-01 is
shown in Figure 5.24a. There are several bubbles distributed on the surface, although
many of them are not in the focal plane. The locations of the nucleation sites are not
fixed, that is, there is no artificial cavity to promote nucleation. As bubbles grow in
close proximity to each other, they affect the liquid flow near the wall, and can enhance
downstream bubble growth.

( A ) HSV image, case B-01 ( B ) PIV image, case B-04

F IGURE 5.24: Sample images from HSV and PIV experiments
Chapter 5. Results and Discussion 121

In Figure 5.24a, the bubble highlighted has a diameter of 0.264mm, which is larger than
bubbles typically seen at higher pressures or flow rates. The low liquid flow rate in
this case causes increased development of the thermal boundary layer, which by Hsu’s
criterion for bubble growth [52], will allow for larger bubbles.

A sample PIV image of water with seed particles flowing upward past a vapor bub-
ble is shown in Figure 5.24b. When applicable, phase fraction-weighted averaging
of velocity data was performed to obtain the time-averaged velocity profile and RMS
fluctuations of the velocity for the liquid phase only.

The period for time averaging is 150s for cases B-03 to B-06, and 300s for cases B-07 to
B-09. In all cases, the time averaging period was slightly longer than the nucleation
and growth time of the bubble, as there is a large uncertainty in identifying the onset
of nucleation due to the slow growth needed to produce a single vapor bubble. For
the fully turbulent cases, no phase weighting is applied, as the single bubble is small
enough that only the vectors closest to the wall would be affected. The reason a sin-
gle bubble is desired for PIV experiments is to isolate the effect of one bubble on the
local liquid flow field. Taking this information, the effect of multiple bubbles can be
estimated by superimposing the effect of a single bubble at multiple locations in the
flow.

5.3.1.1 Bubble Size, Shape and Behavior

Figure 5.25a shows the ensemble averaged bubble size evolution with time. Bubbles
in case B-01 grew slowly, departing about 45s after incipience.

After incipience, there is a period of steady and relatively rapid growth lasting about
30s, followed by a tapering of the growth rate and an oscillation of size around an
equilibrium length, and then departure. This behavior was typical of all the bubbles
observed, and a clear trend of rapid growth, followed by tapered growth persisted.
122 Chapter 5. Results and Discussion

Taking the bubble size data from Figure 5.25a, the bubble aspect ratio can be deter-
mined. Occurring with the growth there is a decrease in the bubble sphericity until
the bubble reaches a steady elliptical shape, at which point minor high frequency os-
cillations in shape can be observed until departure as shown in Figure 5.25b.

Due to the video camera’s low frame rate used to capture the bubble images for case
B-01, data for bubbles after departure could not be obtained. To best consider the bub-
ble behavior after departure, additional experiments were performed with a lower
subcooling, resulting in overall larger bubbles. Larger bubbles are preferred for exper-
iments that investigate variations in bubble size and shape because the incremental
changes are more appreciable and easier to measure. The bubbles measured in case
B-02 were entrained in the liquid, and had an average diameter of about 1.0mm.

( A ) Axis lengths ( B ) Aspect ratio

F IGURE 5.25: Ensemble-averaged bubble size and aspect ratio during

growth, (5 bubbles, case B-01)

The bubble size data is shown versus radial centroid position for case B-02 in Figure
5.26a and has been ensemble averaged over 10 bubbles. The observed trend may be
counterintuitive at first glance, as it appears that with motion away from the heated
surface there is an increase in the bubble size that can only be caused by vaporization.
In actuality, the motion of the bubble occurs first, followed by the growth or shrinkage
Chapter 5. Results and Discussion 123

of the bubble shortly thereafter. This change in size of the bubble induces an oscilla-
tory motion in the spanwise direction, which in turn produces a delayed size change.
Throughout the rising process, the bubbles also maintain an elliptical shape similar to
that seen in air injection experiments at high confinement numbers.

The motion of the entrained bubbles is shown in Figure 5.26b, with 1ms between data
points. As the bubbles rise upward, there is an oscillatory motion in the radial posi-
tion, attributable to the change in bubble size shown in Figure 5.26a. Any motion in
the radial direction causes the bubble to move to a different position in the thermal
boundary layer. Motion away from the wall causes the bubble to condense and be
subsequently pushed back towards the wall by means of a reduction in the shear lift
force acting on the bubble and vice versa for motion toward the wall.

( A ) Axis lengths compared to radial bubble po-

( B ) Bubble trajectory
sition

F IGURE 5.26: Ensemble-averaged entrained bubble size and trajectory,

(10 bubbles, case B-02)

By tracing the bubble motion shown in Figure 5.26b and coupling it with the change in
bubble size shown in Figure 5.26a, the behavior of the bubble can be visualized quite
simply. For modeling purposes, the bubble can be idealized as having an oscillatory
motion about an equilibrium position. The amplitude of this motion is governed by
the temperature profile and liquid flow parameters that produce the body forces acting
in the spanwise direction.
124 Chapter 5. Results and Discussion

5.3.1.2 Effect of Bubble Growth on Liquid Flow Field

Further experiments were aimed at isolating the effect of a single bubble on the liquid
flow rate. The effect of multiple bubbles on the instantaneous velocity field can be ide-
alized as the superimposition of the effects of single bubbles of varying sizes discretely
separated.

The spanwise and streamwise velocity profiles for case B-03 are shown in Figure 5.27a,
with corresponding RMS fluctuations shown in Figure 5.27b. The single-phase axial
velocity profile and RMS fluctuations, taken before heating but at the same flow rate, is
shown by a dotted line. There is a significant increase in the Reynolds number, despite
the same flow rate, due to reduced viscosity at a higher temperature. The result is a
decrease in the averaged core velocity, and a reduction in the boundary layer thickness
near the walls. In contrast to this decrease in average velocity, there is an increase in the
turbulent fluctuation in the core by a factor of 2.5. The difference in velocity profiles
between the adiabatic and heated flows is due to the increase in the Reynolds number
causing the flow to transition from laminar to turbulent due solely to heating.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.27: Ensemble-averaged PIV measurements at a bubble nucle-

ation site elevation, case B-03

An appreciable increase in the average streamwise velocity near the wall is observed,
due to motion of bubbles in the boundary layer. As a result, the maximum fluctuations
Chapter 5. Results and Discussion 125

in the streamwise velocities occur at the inner wall, and persist until 0.5mm from the
inner wall. Similar to the air injection cases, the axial velocity profile can be nondi-
mensionalized by the process detailed in Section 3.1. The resulting nondimensional
velocity profile is shown in Figure 5.28, along with the corresponding single-phase
adiabatic profile. The heating of the flow causes a shift of the core profile up and to
the right; with the presence of the bubbly layer, the nondimensional velocity close to
the wall shift upward away from the theoretical profile.

F IGURE 5.28: Nondimensional axial velocity profile at a bubble nucle-

ation site elevation, case B-03

In the sample PIV exposure for case B-04 shown in Figure 5.24b, the bubble diameter
was 0.12mm. The time-averaged velocity profiles and RMS fluctuations are shown in
Figures 5.29a and 5.29b. The nondimensionalized axial velocity at the bubble nucle-
ation site is given in Figure 5.30. Here, the velocity fluctuations in the turbulent core
are of the order of 20%, and the viscous boundary layer is compressed further than
in the adiabatic profile as, as evident in Figure 5.29a, however, large streamwise fluc-
tuations remain near the inner wall. The presence of bubbles has a minimal effect on
the velocity profile in the bulk flow, with upstream departed bubbles being confined
to this viscous boundary layer. Of note is the negative spanwise velocity that indicates
the liquid flows into the wake of the bubble, however, the bubble does not grow large
126 Chapter 5. Results and Discussion

enough to cause backflow, as seen in the air injection cases.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.29: Ensemble-averaged PIV measurements at a bubble nucle-

ation site elevation, case B-04

F IGURE 5.30: Nondimensional axial velocity profile at the bubble nucle-

ation site elevation, case B-04

The flow rate was slightly increased in case B-06 and the liquid subcooling was re-
duced to achieve minimal boiling. In this case, the measured bubble size was roughly
0.08mm. The resulting velocity profile and RMS fluctuations are shown in Figures
5.31a and 5.31b. Additionally, the dimensionless axial velocity is given in Figure 5.32.
Similar to case B-04, the viscous boundary layer is compressed beyond what was ob-
served in the adiabatic case, this effect can be seen in Figure 5.31a. However, here, the
Chapter 5. Results and Discussion 127

negative spanwise velocity is smaller due to the smaller bubble. Compared to the law
of the wall, the profile in the immediate vicinity of the wall is in good agreement with
the theoretical profile, however there is little agreement in the core. This indicates that
the flow regime in this case is in the transition region, but approaching turbulent.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.31: Ensemble-averaged PIV measurements at a bubble nucle-

ation site elevation, case B-06

F IGURE 5.32: Nondimensional axial velocity profile at a bubble nucle-

ation site elevation, case B-06

To further increase the Reynolds number to 17000, the flow rate was increased for
case B-07, resulting in fully turbulent flow. The bubbles observed in this case were
very small, typically around 0.06mm. Figures 5.33a-5.34 show the average velocity
128 Chapter 5. Results and Discussion

profiles, RMS fluctuations and nondimensional velocity for case B-07, respectively.

( A ) Velocity profiles ( B ) RMS fluctuations

F IGURE 5.33: Ensemble-averaged PIV measurements at a bubble nucle-

ation site elevation, case B-07

F IGURE 5.34: Nondimensional axial velocity profile at a bubble nucle-

ation site elevation, case B-07

Similar to previous cases, the Prandtl number is still higher than unity, P r ≈1.76, caus-
ing the thermal boundary layer thickness to further reduce with the viscous bound-
ary layer, resulting in yet smaller bubble diameters. The measured profile compares
very well with the adiabatic profile in the core, and there is minimal variation in the
spanwise velocity. Comparing the velocity fluctuations shows that the heated case
has significantly higher fluctuations in the core, despite the average velocity being in
Chapter 5. Results and Discussion 129

good agreement. In the nondimensional profile, the measured velocity matches almost
identically with the law of the wall, except for a single velocity vector near the wall.
The spatial resolution of the camera with the lens configuration used is not sufficient
to determine the effect of the bubble on the liquid velocity any closer to the wall.

To better understand the effect of the bubble on the near-wall velocity field, further
experiments were performed by mounting a microscopic objective onto the high speed
camera to perform Micro-PIV measurements. As the bubble grows, the liquid velocity
in the immediate vicinity of the bubble is affected by the motion of the bubble interface.
Figures 5.35 and 5.36 show the nondimensionalized velocity profiles from Micro-PIV
experiments for cases B-05 and B-08 respectively. Figure 5.36 shows the compounded
effect of a sliding, growing bubble.

F IGURE 5.35: Nondimensional axial velocity profile at a bubble nucle-

ation site elevation, case B-05
130 Chapter 5. Results and Discussion

F IGURE 5.36: Nondimensional axial velocity profile for a sliding bubble

passing through the measurement elevation, case B-08

A large deviation from the law of the wall is observed, showing an increase in the
liquid velocity at the elevation where the bubble is formed. Considering the RMS
fluctuations in velocity during the bubble growth period, an upward shift from the
law of the wall occurs. Typically, there is a region of influence over which the static
bubble affects the liquid velocity field as the liquid diverts around the bubble; this
region is roughly 1.5 times the maximum bubble diameter. The sliding bubble shows
a more significant effect on the liquid velocity, as the deviation from the law of the wall
expands further than the bubble diameter, and the nondimensional velocity matches
that of the bubble in Figure 5.36.

5.3.2 Elevated Pressures Experiments

Building on the results from the PIV and high speed video experiments performed on
the atmospheric pressure flow loop, experiments were performed to quantify the ef-
fects of varying flow and heat transfer parameters on bubble growth and behavior at
higher pressures than atmosphere. These experiments were performed on the elevated
Chapter 5. Results and Discussion 131

pressure flow loop, which had added benefit of better thermal boundary layer devel-
opment due to the longer heated length, as well as a higher energy density heater.
For each case, the bubble diameter, aspect ratio and position were measured at ev-
ery frame. In some cases, the growing bubbles slide, and the sliding distance is also
recorded where applicable. A series of images of two typical sliding bubbles for case
P05-01 are shown in Figure 5.37.

In a typical case, with large enough separation between bubbles, the growth rates
and behaviors are roughly uniform. Conversely, a nearby upstream bubble can allow
the downstream bubble to grow larger than it normally would, as seen in the upper
bubble. In Figure 5.37, while both bubbles slide for the duration of their growth, the
gap between them reduces, and the growth rate of the downstream bubble increases.

While the problem of predicting subcooled flow boiling is very multifaceted, the present
study attempts to isolate the impact of each flow boiling parameter to study its effects.

5.3.2.1 Bubble Liftoff Diameter

Figure 5.38 shows the liftoff diameter plotted against the liftoff time, to give an indi-
cation of the growth rate for bubbles in each case. The error bars shown in this section
indicate one standard deviation from the average value. Expectedly, cases with large
bubbles experience a large variation in diameter, while some smaller bubbles experi-
ence large variations in liftoff time. Due to this, the technique of taking an average
diameter may not be appropriate for larger bubbles.

Looking at the 1 bar cases (P01-01, P01-02, and P01-03) in detail, the bubble liftoff
diameter is significantly smaller with longer growth time for case P01-01, where the
subcooling is large. Cases P01-02 and P01-03 yield similar liftoff diameters, indicating
a lesser dependency between bubble growth and mass flux.
Chapter 5. Results and Discussion
F IGURE 5.37: Timelapse of two sliding bubbles growing and departing (Case P05-01)
132
 Chapter 5. Results and Discussion

F IGURE 5.38: Liftoff diameter versus time, grouped by case

133
134 Chapter 5. Results and Discussion

Comparing the liftoff diameters for the 2 bar cases, which have similar flow rates,
subcooling and superheats, but different heat fluxes, Figure 5.38 indicates that the heat
flux alone is not a significant contributing factor to liftoff diameter and growth time,
although the literature suggests it may be a significant factor in active nucleation site
density or bubble frequency by reducing the waiting times [77, 78, 81, 111]. P02-02 has
a high local heat transfer coefficient, which results in very short growth times relative
to the other cases which is to be expected, because at steady state conditions, a higher
proportion of the heat generated by the heated tube will go to phase change rather
than increasing the temperature of the heated surface or bulk fluid.

At 3 bar, the cases show smaller overall liftoff diameters than the 1 and 2 bar cases,
with bubbles in P03-01 generally having longer growth times than P03-02, which has
the higher local heat transfer coefficient. This is a similar effect seen in P02-02.

At 5 bar, the cases P05-01 and P05-02 show longer growth times than lower pressure
cases, with liftoff diameters similar to the 3 bar cases. These cases have lower local
heat transfer coefficients relative to the other cases. This, coupled with the additional
energy required to produce bubbles of similar sizes to 3 bar cases is indicative of a
lower nucleation site density, as seen in Figure 5.37.

At the highest pressures tested, 7 and 10 bar, slightly larger bubbles are seen than in
other elevated pressure cases, with significantly longer liftoff times. The nucleation
site densities in these cases were observed to be very high, which limits the effects of
liquid flow very close to the wall.

Prior to liftoff, the bubble has a deformable interface, effectively pinned to the nucle-
ation site at the contact line. The bubble is subjected to numerous forces, all balancing
until the moment of liftoff. Though the forces are constantly changing in magnitude,
the bubble interface reacts to rebalance the forces at every moment. For modeling
purposes, the size and shape of the bubble at the moment of liftoff serves as a useful
Chapter 5. Results and Discussion 135

initial condition for two-phase flows. Figure 5.39 shows the bubble aspect ratio at the
moment of liftoff versus the liftoff diameter.

A clear trend emerges from Figure 5.39. At the moment of liftoff, bubbles tend to be
elongated due to the shear lift force from the liquid flow, resulting in a higher than
unity aspect ratio.

Effect of liquid subcooling and wall superheat To further investigate the effect of
the liquid subcooling and wall superheat on bubble liftoff diameter and liftoff time,
the dimensionless wall temperature is considered:

Tw − Tbulk ∆Tw + ∆Tsub 1 ∆Tw

θ= = , = (5.3)
Tw − Tsat ∆Tw θ ∆Tw + ∆Tsub

Figure 5.40 shows the bubble liftoff diameter versus the inverse of the nondimensional
wall temperature. The inverse is selected because 1/θ varies between 0 and 1, if ∆Tw
increases or ∆Tsub decreases, 1/θ increases up to unity. Additionally, as the wall su-
perheat goes to zero and 1/θ decreases, it is expected that the driving temperature
difference for boiling diminishes and bubble growth rate will decrease, increasing the
liftoff time.
Chapter 5. Results and Discussion

F IGURE 5.39: Bubble aspect ratio at liftoff

136
Chapter 5. Results and Discussion 137

F IGURE 5.40: Liftoff diameter versus the inverse of dimensionless wall

temperature, grouped by case

From Figure 5.40, a clear trend emerges looking at cases on a same-pressure basis: as
the liquid subcooling decreased and 1/θ increases, the liftoff diameter increases.

Effect of liquid flow rate The effect of liquid flow rate on the bubble liftoff diame-
+ Dlo uτ
ter can be investigated by plotting the nondimensional liftoff diameter, Dlo = ν
,
against the liquid Reynolds number as shown in Figure 5.41. The shear stress and
viscosity are calculated for each case based on the bulk temperature properties. This
allows the reasonable comparison of bubble diameters for cases with different operat-
ing pressures.
138 Chapter 5. Results and Discussion

F IGURE 5.41: Nondimensional liftoff diameter versus Reynolds number,

grouped by case

A significant majority of bubbles have liftoff diameters that approach the lower end of
+
the buffer layer, with average Dlo between 5 and 10. This is of importance in the ap-
plication of Situ et al.’s correlation for the liftoff diameter, which relies on predicting
the turbulent region that the bubble diameter falls within. It is important to note that
as the pressure increases, the Prandtl number at the saturation temperature decreases.
Additionally, the size of the thermal boundary layer relative to the viscous boundary
layer is linked to the Prandtl number; as P r → 1, the two layers become similar in
size, which contributes to the near uniform nondimensional departure diameters. The
exceptions to this are cases P01-02, and P01-03, which have smaller temperature gradi-
ents near the wall, and thus thicker thermal boundary layers, permitting evaporation
Chapter 5. Results and Discussion 139

of liquid over a longer range.

Despite the marked difference in Reynolds number between the experimental cases,
there is no discernible effect of the liquid Reynolds number on the nondimensional
bubble diameter, indicating that the effect is dominated by the close relative thick-
nesses of the viscous and thermal boundary layers. To better visualize this behavior,
Figure 5.42 shows a bubble in a thin thermal boundary layer, indicated by a dashed
line, followed by growth in a thicker thermal boundary layer, shown as a solid line.
The resulting impact on the momentum boundary layer is also indicated, although not
necessarily to scale.

F IGURE 5.42: Effect of bubble growth on momentum boundary layer

As the bubble grows in size, the projected area of the bubble in the spanwise direc-
tion increases, subjecting the bubble to higher liquid velocities and associated forces.
Additionally, the change in the momentum boundary layer causes a corresponding
change in the quasi-steady force, Fqs . Similarly, the shear lift and buoyancy forces (Fsl
and Fp − Fg , respectively) increase with the bubble size.

Correlations of bubble liftoff diameter

140 Chapter 5. Results and Discussion

Nondimensional heat flux Situ et al. [81] proposed a correlation that relates the
Jakob number and the nondimensional heat flux as given by equation (5.4), which
includes the departure diameter. This correlation was fitted to data sets presented by
Basu et al. [77], Thorncroft et al. [75], and Situ et al.’s own data, with reasonably good
agreement, however, the correlation does not take pressure into consideration, albeit
the fluid properties involved vary with pressure.

qw ”Dd
NqG ≡ = 5.28Ja1.37 (5.4)
αl ρv hf g

To better compare other experimental results from Prodanovic et al. [105], Situ et al.
[80], and Chu et al. [119] using equation (5.4), the wall superheat in each of these cases
must be estimated. This is done using Chen’s correlation [54], given in Table 3.1.

In addition to those cases, data from Sugrue et al. [116] is also compared. Sugrue et al.
provided wall temperature and bubble size data at varying heat fluxes and flow rates,
so the Jakob number can be calculated directly without any additional correlations.

A comparison of equation (5.4) with the aforementioned experimental results and the
present experimental data is shown in Figure 5.43. Situ et al.’s correlation was devel-
oped for conditions near ONB at atmospheric pressure using data that ranged in mass
flux from 124 to 890kg m−2 s, heat fluxes from 11 to 1130kW m−2 , and inlet subcooling
from 6.6 to 53◦C. This correlation compared well with data from Basu et al. [77], Situ
et al. [81], and Thorncroft et al.’s [75] upflow and downflow data.

The 1 bar cases fit well within ±75% of Situ et al.’s [81] correlation, with higher pres-
sure data tending to fall below the curve. Likewise, data from Situ et al. [80] and
Chu et al. [119] also deviate from the curve. The reason for deviation in those cases
is likely due to Chen’s correlation overpredicting the wall superheat, resulting in a
 Chapter 5. Results and Discussion

F IGURE 5.43: Comparison of Situ et al.’s correlation with present and previous experimental data
141
142 Chapter 5. Results and Discussion

shift to higher Jakob number values. Data from Sugrue et al., despite having condi-
tions covered by Situ et al’s [81] correlation fell well below the curve. For the present
study, however, the deviation comes from smaller liftoff diameters at higher pressures
at fixed heat fluxes, leading to lower nondimensional heat fluxes, NqG .

Other correlations Multiple correlations have been developed to predict bubble liftoff
diameters, as seen in Table 3.2. A database containing experimentally measured bub-
ble departure/liftoff diameters was constructed by including data from Prodanovic
et al. [105], Situ et al. [80], Chu et al. [119], Sugrue et al. [116] and the present work.
The predictive capabilities of correlations from Zuber [21], Mikic et al. [89], Ünal [57],
Prodanovic et al. [105], Situ et al. [80], and Chu et al. [119] are evaluated against the
database. For each correlation, the average error between the predicted and experi-
mentally measured liftoff diameters is calculated to best evaluate the correlation. For
consistency, parameters that require the wall superheat are estimated using Chen’s
correlation [54], when the measured wall temperature is not available. Additionally,
bubbles are assumed not to slide. The average error is calculated in two ways: the er-
ror for each case is counted equally, or the error in higher pressure cases are weighted
by the pressure, as given by equation 5.6. The latter method of weighing the error pro-
vides better metrics for determining which correlation best treats elevated pressure
data, which is currently limited in availability. A correlation that has a significantly
higher pressure-weighted error than non-weighted error can be considered unsuitable
for use with elevated pressure data. Figure 5.44 shows the predicted versus measured
bubble liftoff diameters for the evaluated correlations. Dotted lines indicate ±75% for
each correlation.

|Dd,i,pred − Dd,i,exp |
Erri = × 100 (5.5)
Dd,i,exp
Chapter 5. Results and Discussion 143

Pn
Err × W
Erravg = Pn i
i=1
(5.6)
i=1 W

 
 1 Non-weighted 
 
where W =
 P [bar]
 Weighted 


In general, existing correlations did not provide significant accuracy in predicting boil-
ing data at elevated pressures, however, Prodanovic et al.’s correlation yielded the best
results when weighting is applied. Their correlation compared well with experimen-
tal results from their own work and from Chu et al., but gave mixed results for other
cases.

One of the key limitations of Zuber’s and Mikic et al.’s correlations is that they both
require the liftoff time of the bubble to predict the liftoff diameter. This is suitable for
comparison with experimental data, however, to use these correlations in a predictive
capacity would require an additional correlation for the liftoff time, which would be
subject to additional inaccuracies.

Although Ünal’s correlation was developed using very high pressure boiling data,
key measurements were missing, and bubbles were measured at unknown times after
liftoff. As a result, the correlation overpredicted the liftoff diameters for the present
study and Situ et al., and underpredicted some data from Prodanovic et al. and Sugrue
et al.

Chu et al.’s correlation is a modified form of Prodanovic et al.’s correlation, and gave
improved results for those cases and Situ et al.’s experimental data. These improve-
ments did not extend to elevated pressure cases, leading to a reduction in accuracy
when error is weighted by pressure.

Of the correlations evaluated in the present study, Situ et al.’s correlation yielded re-
sults that matched fairly well with all cases, and the results were relatively unaffected
144 Chapter 5. Results and Discussion

( A ) Zuber’s Correlation, b = 23.5 ( B ) Mikic et al.’s Correlation

( C ) Ünal’s Correlation ( D ) Prodanovic et al.’s Correlation

( E ) Situ et al.’s Correlation, b = 0.57 ( F ) Chu et al.’s Correlation

F IGURE 5.44: Comparison of experimentally measured bubble liftoff data

with existing correlations
Chapter 5. Results and Discussion 145

when weighted by pressure. It also predicted data from the present study signifi-
cantly better than other correlations, which tended to overpredict, indicating that it
could inherently handle high pressure data. This correlation also considers most of
the parameters treated in the present study, with the exception of the liquid subcool-
ing.

Note that correlations from Zuber and Situ et al. contain an empirical asphericity con-
stant which accounts for variations in shape. Zeng et al. [74] set b = 1.73 in Zuber’s
correlation to provide the best fit to their data, here b = 23.5 to best fit with the exper-
imental data. By setting b = 0.57 for Situ et al.’s correlation, the average error values
improve overall for the datasets analyzed here.

+
A new correlation for Dlo is proposed in equation (5.7), making use of a θ parameter to
account for the wall temperature gradient, and the Zuber number to add consideration
for the geometry and heat flux. Additionally, a parameter from Ünal’s correlation, √a ,
b

was added to provide contributions from the heater rod material properties. Figure
5.45 shows the predicted versus measured dimensionless liftoff diameter.

 1.89
Dlo uτ a
+
Dlo ≡ = 23.5θ1.91 Re0.97 Ja−0.61 P r−3.91 NZu
0.57
√ (5.7)
ν b

Table 5.2 summarizes the average error values for each correlation and dataset. Dis-
crepancies between previously reported errors and the errors reported in Table 5.2 can
be attributed to the way each correlation is applied; here, values are calculated at the
saturation temperature with the exception of the macroscale heat transfer coefficient in
Chen’s [54] correlation, because liquid around the growing bubble is at superheated
temperature. The key limitation of the existing correlations is that they were devel-
oped based mainly on Chen’s highly idealized, and somewhat outdated wall temper-
ature model. Chen’s correlation requires a few approximations that may significantly
affect the results. The majority of previous works utilize Chen’s correlation, however
146 Chapter 5. Results and Discussion

few detail the approximations made. When experimentally determined wall tempera-
ture values are supplied as inputs, as in Sugrue et al. and the present study, the ability
of correlations to accurately predict bubble behavior is poor.

TABLE 5.2: Average error values for each correlation and dataset (%)
Experimental Dataset
Correlation Prodanovic et al. Situ et al. Chu et al. Sugrue et al. Present Study Average for all datasets
ErrN W ErrW ErrN W ErrW ErrN W ErrW
Zuber 31 31 - - - 187 343 165 316
Mikic et al. 96 77 - - - 373 786 332 723
Ünal 34 31 117 42 37 557 675 242 433
Prodanovic et al. 56 56 199 20 128 266 179 179 155
Situ et al. 74 69 69 57 71 85 89 75 81
Chu et al. 31 29 141 19 56 162 124 114 105
Proposed correlation 36 31 24 28 33 51 39 37 34

F IGURE 5.45: Results of proposed correlation for the nondimensional

liftoff diameter

The proposed correlation for the nondimensional bubble liftoff diameter is useful for
predicting the sizes of bubbles relative to the momentum boundary layer thickness.
The correlation has a strong dependence on the Prandtl number, which determines
the relative sizes of the momentum and thermal boundary layers. Additionally, the
Chapter 5. Results and Discussion 147

parameters with which the liftoff diameter is nondimensionalized (uτ /ν) scale with
the liquid Reynolds number, so it is expected that the Reynolds number appears with
an exponent near unity.

Most of the experimental datasets surveyed fit well within 75% of the proposed cor-
relation, with minor deviations in some cases. Despite this, the overall mean average
error (MAE) is 37%, improving to 34% when weighted by pressure, a significant im-
provement over existing models. The particular datasets evaluated were selected to
cover differing operating conditions while maintaining some overlap between them.
While other correlations yield better results in specific cases, the proposed correlation
provided suitable results for a wide range of conditions: 1 < P < 10bar, 0.1 < Ja <
100, 5 × 10−6 < NZu < 200, 10−5 < Nb < 10−3 .

5.3.2.2 Bubble Liftoff Time

While the liftoff diameter is controlled by a balance of radial and axial forces acting on
a bubble, the liftoff time after bubble nucleation is governed additionally by the rate
of bubble growth. The force balance has been treated in several previous papers, but
with particular details by Situ et al. [80], where a correlation for the shear lift coefficient
presented earlier in equation (3.39) is given. While the shear lift coefficient is inversely
related to the bubble Reynolds number, it has a higher dependence on the liquid shear
gradient, which increases with liquid Reynolds number. This results in a net force on
the bubble being achieved sooner, holding thermal parameters constant. To illustrate
this effect, the liftoff time is compared with the liquid Reynolds number in Figure 5.46.
148 Chapter 5. Results and Discussion

F IGURE 5.46: Liftoff time versus Reynolds number, grouped by case

At each pressure, the average liftoff time decreases with increasing liquid Reynolds
number. As described in section 3.2.1, the momentum boundary layer becomes thin-
ner with increasing Reynolds number, exposing a growing bubble to higher liquid
velocity and causing a higher shear force sooner at smaller bubble sizes. The key ex-
ceptions to this are elevated pressure conditions, where the Prandtl number near the
saturation temperature is close to unity. This causes both the thermal and momentum
boundary layers to have roughly the same size. As the vapor bubble diameter grows
to reach the edge of superheated liquid layer near the wall, it is still within the low
liquid velocity region of the momentum boundary layer. Without the shear forces re-
sulting from high liquid velocity within the thermal boundary layer, bubbles will tend
not to liftoff quickly; instead the contact area reduces over time by the evaporation of
the microlayer, as detailed by Yabuki et al. [128]. When the bubble has a smaller con-
tact diameter with the wall, the shear force required to cause liftoff is further reduced.
Additional information detailing the force balance on a bubble at the moment of liftoff
Chapter 5. Results and Discussion 149

can be found in Appendix B.

Effect of heat flux To study the effect of heat flux on bubble growth, the liftoff time
is plotted against the local Nusselt number in Figure 5.47. At higher Nusselt numbers,
it is expected that bubbles will grow and depart faster, as more heat will be convected
from the heated surface and used for phase change.

F IGURE 5.47: Liftoff diameter versus Nusselt number, grouped by case

In Figure 5.47, a weaker trend can be seen with liftoff time decreasing with increasing
local Nusselt number. The liftoff times are longer at higher pressures, 5, 7, and 10
bar, possibly due to a combination of higher vapor densities and the relatively thicker
thermal boundary layer thickness that occurs at lower Prandtl numbers. The higher
vapor densities would require more heat to vaporize liquid to produce bubbles of the
same size. Additionally, a thicker thermal boundary layer means that more heat is
required to reduce the contact area enough to allow the lower liquid flow to shear the
bubble from the heated surface.
150 Chapter 5. Results and Discussion

There are multiple correlations in the literature relating the Nusselt number with the
Reynolds number, but these relations tend to focus on single-phase flows [135, 156].
From the data in Section 5.3.1.2, the liquid phase velocity profiles are relatively unaf-
fected by the bubbles on the heated surface. Only as bubble diameters extend beyond
the viscous layer do they experience stronger interactions with the bulk fluid. Simi-
larly, as bubbles nucleate and depart, the Nusselt number varies as a function of time
and the bubble distribution on the heated surface. Without the consideration of nu-
cleation site density or local void fraction parameters to bridge the gap between the
heat transfer immediately at the heated surface and the turbulent mixing in the bulk
fluid, correlations that relate the Nusselt number to Reynolds number are limited in
capability when it comes to two-phase flows.

Correlation of liftoff time While the bubble liftoff diameter is a highly studied pa-
rameter for experimental works, the liftoff time is typically treated in a limited capac-
ity, as part of the bubble departure frequency. The time between bubble departures is
defined as the sum of the growth time, the time between nucleation and departure, and
the waiting time, the period after departure preceding the nucleation of the following
bubble. Bubble departure either precedes or coincides with bubble liftoff, depending
on if the bubble slides remaining in contact with the heated surface. Therefore, the
minimum value of the bubble liftoff time, tlo , must be the bubble growth time, tg .

Of the works that measured the bubble departure frequency, none have investigated
the bubble liftoff time explicitly. A correlation has been developed by Basu et al. [177]
to predict the bubble growth time, validated by experiments involving flow over a
heated copper plate and flow through a 3×3 heated rod bundle. Experimental pres-
sures ranged from 1.03 to 3.2 bar, mass fluxes from 124 to 926kg m−2 s−1 , and heat
Chapter 5. Results and Discussion 151

fluxes from 25 to 900kW m−2 . The correlation is given in equation (5.8).

tg αl 1 Dd2
F og = = , tg = (5.8)
Dd2 45Ja exp(−0.02Jasub ) 45αl Ja exp(−0.02Jasub )

Basu et al.’s correlation is given in terms of the Fourier number, using the bubble de-
parture diameter as the characteristic lengthscale. Here, Ja is the typical Jakob number
based on the wall temperature, whereas Jasub is the Jakob number calculated using the
liquid subcooling. Additionally, Basu et al.’s experiments did not yield wall temper-
ature data and the wall superheat was estimated using the nucleation cavity size and
the minimum requirements for boiling to occur at ONB.

Using the experimentally measured liftoff diameters as inputs, it is expected that equa-
tion (5.8) should underpredict the liftoff times measured in this work. Figure 5.48a
shows the predicted bubble liftoff times versus the experimentally measured bubble
growth time for Basu et al.’s correlation. Basu et al.’s correlation generally predicts
growth times that are smaller than the measured liftoff times, with the exception of
some cases, particularly those at lower pressures with low Jakob numbers.

Another correlation that can be used to predict the bubble growth time is Zuber’s cor-
relation, given in Table 3.2, which predicts the bubble diameter as a function of time.
For predicting the growth time, the correlation can be used with the liftoff diameter
supplied as the input. Figure 5.48b shows the measured liftoff time versus the growth
time predicted by Zuber’s correlation.

Like Basu et al.’s correlation, Zuber’s correlation also predicts growth times that are
less than the measured liftoff times in most cases. For modeling purposes, Zuber’s cor-
relation can be used in tandem with correlations for bubble liftoff diameter to establish
a lower limit for the bubble liftoff time.
152 Chapter 5. Results and Discussion

( A ) Basu et al.’s correlation ( B ) Zuber’s correlation

F IGURE 5.48: Correlations for bubble growth time

Since there are no models that aim to predict the nondimensional bubble liftoff time,
a new correlation can then be proposed, using similar parameters to the correlation
for bubble liftoff diameter, given in equation (5.7). The liftoff time is included in the
dimensionless Fourier number at liftoff, F olo , with one caveat; the characteristic length
used in this case scales with the momentum boundary layer thickness, uτ /ν ∼ δ. As
the Prandtl number decreases with increasing pressure, the thermal and momentum
boundary layers become of similar size, and the required conditions for bubble liftoff
change. Because of this, the proposed correlation for the Fourier number at liftoff is
not valid for pressures above 5 bar, where the Prandtl number at saturation conditions
is close to unity, P r < 1.1. The new correlation is given in equation (5.9).

 −0.21
tlo αl −2.50 ρv
F olo ≡ = 1.18Ja1.01 P r2.98 θ0.46 Re−4.62 NZu (5.9)
(uτ /ν)2 ρl

Equation (5.9) differs from the proposed correlation for the nondimensional liftoff di-
ameter in that it includes the density ratio, (ρv /ρl ), which increases with higher pres-
sures. From equation (5.9), if the Reynolds number increases, it results in a signifi-
cantly smaller Fourier number, indicating that the liftoff time decreases much faster
than the momentum boundary layer thickness. Additionally, as expected, the liftoff
Chapter 5. Results and Discussion 153

time is strongly affected by the Prandtl number and nondimensional wall tempera-
ture, θ, which relate inversely with the thermal boundary layer thickness. Figure 5.49
shows the liftoff time predicted in equation (5.9) plotted against the experimentally
measured liftoff time from the present study and Prodanovic et al.’s [105] data.

Due to limitations of the new proposed correlation in handling low Prandtl number
data, and the very large range of values of tlo measured across the different experi-
mental conditions, some of the higher pressure data and data where the subcooling is
exceptionally high are significantly underpredicted. Despite this, the overall mean av-
erage error across all data is 57%; ignoring the high pressure cases (P07-01 and P10-01),
the error for the correlation reduces to 51%.

F IGURE 5.49: Proposed correlation for bubble liftoff time

154 Chapter 5. Results and Discussion

5.3.2.3 Bubble Behavior Before and After Liftoff

While most previous studies focused on bubble characteristics at the moment of liftoff,
very few considered the bubble behavior over the entire lifetime. Characterizing this
behavior is crucial to understanding the coupling of two-phase flows and heat transfer,
especially for numerical simulations that aim to capture the complex interactions be-
tween vapor and liquid phases and the heated surface. This behavior has been treated
passingly by many authors, with very little data published at elevated pressure. Situ et
al. [78] tracked vapor bubbles’ radial tracjectories as a function of time but the data pre-
sented was limited to one flow condition at atmospheric pressure. Okawa et al. [110]
measured bubble size, shape and motion evolution before and after liftoff, providing
complete histories for the bubbles. The bubbles were categorized as sliding, bouncing
(departing and returning towards the wall) and collapsing.

An important phenomenon that separates flow boiling from adiabatic air injection
is the condensation of the vapor bubble once entrained in the flow. This behavior
becomes increasingly complex at higher pressures. Higher vapor densities reduce the
elasticity of bubbles and cause them to behave more rigidly.

Similar to the nondimensionalization performed on the liftoff diameter in Figure 5.41,

the same method can be applied to subsequent diameters and radial bubble positions.
By visualizing the data in this fashion, a better understanding of the exact position
of the bubble interface relative to the turbulent core regions can be achieved. Here,
the bubble lifespan is is divided into two parts: nucleation and growth which covers
inscipience to the moment of liftoff, and after liftoff which covers the moment of liftoff
to until the bubble collapses or leaves the viewing area.

Nucleation and growth Figure 5.50 shows the nondimensional bubble diameter evo-
lution against time normalized by the liftoff time for each bubble. After nucleation,
Chapter 5. Results and Discussion 155

the rate of bubble growth is relatively high, it then slows down and tapers off around
T = 0.7Td . After the rapid growth period ends, the bubble maintains the same overall
size until liftoff. This trend persists through almost every case, with exceptions caused
by interactions from passing or sliding bubbles.

During the growth period, the bubble has a deformable interface, effectively pinned
to the nucleation site at the contact line. The bubble is subject to numerous forces, all
balancing until the moment of liftoff. Though the forces are constantly changing, the
bubble interface reacts to rebalance the forces at every moment. To better understand
the behavior of growing bubbles, it is necessary to consider the variation of the aspect
ratio. Figure 5.51 shows the evolution of the bubble aspect ratio versus the normalized
time.

A clear trend emerges from Figure 5.51. At inscipience, bubbles begin with a near
spherical shape, with aspect ratio averaging around unity. At the moment of liftoff,
bubbles tend to be elongated, resulting in a higher than unity aspect ratio. However,
the elongation is in the radial direction rather than the axial direction as shown by
Okawa et al. [110] and Samaroo et al. [102]. This difference can be attributed to the
smaller overall bubble sizes seen in the present experiments; because the edge of the
bubble interface is typically at the boundary of the viscous and buffer layers, the effect
of the liquid drag force is minimal.
Chapter 5. Results and Discussion

( A ) Atmospheric pressure cases (P01-01, P01-02, and P01-03) ( B ) Elevated pressure cases
F IGURE 5.50: Nondimensionalized individual bubble diameter during growth
156
 Chapter 5. Results and Discussion

F IGURE 5.51: Individual bubble aspect ratio during growth

157
158 Chapter 5. Results and Discussion

After liftoff Figure 5.52 shows the nondimensional radial bubble centroid motion as
a function of the normalized time. In most cases, bubble centroid starts in the buffer
zone when the bubbel lifts off or moves quickly from the wall into the buffer zone.
Within the buffer zone, typical bubbles decelerate, as indicated by a decrease in slope.
From this point, bubbles can follow one of three trajectories: they can move into the
fully turbulent core region, stay in the buffer zone, or decelerate enough to move back
into the viscous layer. Bubbles that follow the latter motion are shown in the small
inset of Figure 5.52, which indicates the time particular bubbles reattach to the wall,
as indicated in the The type of motion the bubbles follow appears to be determined
by the number and size of bubbles already entrained in the fluid at the measurement
elevation. The larger the void fraction in the buffer zone near the departing bubble,
the higher the tendency for that bubble to change radial direction and move toward
the wall after liftoff. In some cases (P01-01, P07-01, and P10-01), the recording rate
was set low to capture the slow bubble growth, resulting in an inability to capture the
rapid bubble behavior after departure.

With the motion of the bubbles, there is also an accompanying change in size as the
bubble moves through subcooled liquid. Figure 5.53 shows the nondimensionalized
bubble radius after liftoff versus the nondimensional radial position of the bubble cen-
troid. On the left are the liftoff diameters, and the diagonal line indicates the position
where the base of a spherical bubble interface would be in contact with the wall. To
obtain the location of the outer or inner edge of the bubble, assuming sphericity, add
or subtract the nondimensional radius from the radial centroid location.

Bubbles in atmospheric pressure cases collapse rapidly after liftoff while simultane-
ously moving into the buffer zone. A similar motion can be seen at higher pressures,
although the bubble collapsing is diminished. At 5 bar conditions, bubbles enter the
buffer zone with roughly the same size as during liftoff. The smaller overall size of
Chapter 5. Results and Discussion 159

these bubbles results in a smaller surface area for convective heat transfer in the bulk
fluid; the higher vapor density means that more heat must be lost to produce a notice-
able change in size. Note that Figure 5.53 does not distinguish between bubbles that
move uniformly away from the wall and bubbles that change direction mid-route, nor
does it take account for the eccentricity of the bubbles.

In addition to the motion and change in size of the bubbles, the bubbles also change
shape while they move. Figure 5.54 shows the time evolution of the aspect ratio af-
ter liftoff. Bubbles experience oscillations in shape while moving, during which the
direction of elongation alternates between radial and axial. The magnitude of the
oscillations decays over time as the bubble approaches a near spherical shape. The
average aspect ratio for each case is shown on the inset on the right. Bubbles in case
P05-01 have a larger average aspect ratio after liftoff, despite the increased vapor den-
sity at 5bar, likely due to the high wall superheat. The bubbles depart with a highly
elongated shape, and become more spherical over time.
Chapter 5. Results and Discussion

F IGURE 5.52: Nondimensionalized bubble motion behavior after liftoff

160
 Chapter 5. Results and Discussion

F IGURE 5.53: Nondimensionalized bubble radius and motion behavior after liftoff
161
Chapter 5. Results and Discussion

F IGURE 5.54: Bubble aspect ratio after liftoff

162
Chapter 5. Results and Discussion 163

Correlations for bubble history number

After liftoff Existing correlations for the bubble history number can also be eval-
uated. The bubble history number during collapse, βc , is a useful metric for deter-
mining the evolution of interfacial area in a collapsing bubble. Several works have
investigated vapor bubble collapse with very limited results [92, 98, 162, 171]. Due to
the tedious nature of measuring bubble sizes from the moment of liftoff until they
collapse or leave the viewing area, the data available for the development of useful
correlations for the bubble history number has been limited. In addition to measuring
bubble diameters at the moment of liftoff, the bubble diameter for the entire lifespan is
measured, resulting in data that was previously unavailable. With this data, existing
correlations for the bubble history can be properly evaluated to determine their suit-
ability for handling bubble collapse at elevated pressures, where thermal mass of the
bubble is significantly higher and is a dominant factor in interfacial heat transfer.

A key parameter in determining the bubble history number is the Fourier number
2
based on the bubble liftoff diameter, F ob,lo = tlo αl /Db,lo , which scales with time. The
time used to calculate the Fourier number, tF o , in this case is taken as tF o = t − tlo such
that at the moment of liftoff F o = 0, and βc = 1. A severe limitation of the applicability
of these existing correlations, however, is a common fractional exponent that may re-
sult in non-real values of βc for low ratios of the bubble Reynolds number and Fourier
number. Relatively large Fourier numbers are very common in high pressure boiling
experiments due to the smaller vapor bubble sizes. Figure 5.55 shows the predicted
versus measured bubble history numbers for a number of correlations presented in
Table 3.3 for each elevated pressure case. A total of 643 data points were used to eval-
uate each correlation, as well as develop a new model for bubble collapse. Instances
where the predicted βc is non-real do not appear in the plots, and the error associated
with those points is not calculated.
164 Chapter 5. Results and Discussion

( A ) Chen and Mayinger’s Correlation ( B ) Zeitoun et al.’s Correlation

132 of 643 points (20% of data) 643 of 643 points (100% of data)

( C ) Warrier et al.’s Correlation ( D ) Kim and Park’s Correlation

459 of 643 points (71% of data) 141 of 643 points (22% of data)
F IGURE 5.55: Comparison of experimentally measured bubble history
data with existing correlations

Chen and Mayinger’s correlation does a good job of predicting bubble collapse at am-
bient pressures, as well as a few elevated pressure cases. These data points, however,
have the commonality of low Fourier numbers, F ob,lo < 0.01. For times large enough
or bubble liftoff diameters small enough to cause the Fourier number to dip below
this threshold value, the discriminant of the correlation becomes negative and pro-
duces non-real values of βc for 80% of the data points. The resulting MAE for this
correlation was 32% for the 20% of data points that produced real values.

Zeitoun et al.’s correlation exhibits the opposite behavior of that seen in Chen and
Chapter 5. Results and Discussion 165

Mayinger’s correlation. The introduction of the thermal diffusivity term significantly

dampens the effect of the Fourier number. The result is a highly reduced sensitivity
to the Fourier number that while reducing the average error, does so at the expense of
predicting low Fourier number data poorly. Of the correlations evaluated, Zeitoun’s
correlation performed the best, with an MAE of 30% for all of the data.

Warrier et al.’s correlation is subject numerical instability that causes very large values
of βc occurring typically at elevated pressures. While the correlation predicts well
the bubble history number at ambient pressures, it is incapable of handling the high
Fourier numbers that occur at elevated pressures, resulting in a MAE of over 2480%
for 71% of data points.

Kim and Park’s correlation is very similar to that of Chen and Mayinger’s, and is
subject to the same benefits and shortcomings. While performing slightly better with
elevated pressure data, the accuracy for ambient pressure bubbles is reduced enough
to cause an increased overall error of 58% for 22% of data.

A new correlation is now proposed that provides an improved accuracy to Zeitoun

et al.’s correlation, with an increased sensitivity to the time evolution of the bubble
diameter. Additionally, to avoid the possibility of a negative discriminant, the correla-
tion uses different terms from previous models. Rather than use the Fourier number,
which is problematic based on existing correlations, a nondimensional time parameter
based on the time to bubble liftoff, t/tlo is used. The time is given as a fraction of the
liftoff time and the ratio (t/tlo ) is greater than unity, unlike the Fourier number which
has values from 10−4 < F ob,lo < 10, contributing to the overall volatility of these cor-
relations. Despite removing contributions from the Fourier number, the bubble liftoff
diameter is still present in the Reynolds number term, although weakly, so the cor-
relation is still representative of bubble growth. The proposed correlation is given in
166 Chapter 5. Results and Discussion

equation (5.10).

 0.25
Db (t > tlo ) t
βc ≡ = 2.48θ−0.57 Re−0.68
b,lo P r
−0.10
Ja0.31 + 0.638 (5.10)
Db,lo tlo

Figure 5.56 shows a comparison between predicted and measured bubble history num-
bers for the correlation proposed in equation (5.10). Overall, the correlation predicts
well the bubble behavior at elevated pressure reasonably well, and can capture a wide
range of bubble behavior after liftoff, including bubbles that grow even after depar-
ture. Despite this, it is limited in its ability to predict rapid bubble collapse, and sig-
nificant bubble growth after liftoff. The MAE for this new correlation against the data
presented is 22%, an improvement over Zeitoun et al.’s correlation.

F IGURE 5.56: Results of proposed correlation for the bubble history num-
ber for collapsing bubbles
Chapter 5. Results and Discussion 167

Before liftoff Although the literature investigates bubble collapse rates exclusively, a
similar approach can be taken to predict bubble growth rates. The same optimization
method is applied to fit a correlation to bubble growth data, consisting of 507 data
points, spanning all elevated pressure loop cases. The resulting model for the bubble
history number during growth, βg , is given by equation (5.11).

 0.42
Db (t < tlo ) t
βg ≡ = 0.58 Ja−0.05 P r−0.49 Re0.11 NZu
0.18
(5.11)
Db,lo tlo

Here, the decision to use the nondimensional time parameter provides an added con-
venience in that this parameter can only have values between 0 and 1. A comparison
of the proposed model with bubble growth data is shown in Figure 5.57 with dotted
lines representing ±25%, and the resulting MAE for equation (5.11) is 20%.

F IGURE 5.57: Results of proposed correlation for the bubble history num-
ber for growing bubbles
168 Chapter 5. Results and Discussion

The majority of the data points that the proposed correlation underpredicts tend to
represent bubbles near the liftoff diameter under conditions having liquid Reynolds
numbers on the lower end of the cases covered. While this trend is evident, the liq-
uid Reynolds number does not occur in equation (5.11), however, the liquid temper-
ature profile near the wall would change with Reynolds number. As bubbles mature,
the local superheat around the bubble will vary as detailed by Chen [54]. For lower
Reynolds numbers, the thermal boundary layer will be thicker than at higher Reynolds
numbers where mixing effects are more dominant. This thicker thermal boundary
layer will cause more rapid growth and permit larger bubbles. This phenomenon is
not adequately captured by equation (5.11), although the inclusion of a nondimen-
sional wall temperature term helps minimize errors in the prediction to an extent.

With the four new correlations proposed in this Section, a complete model of bubble
departure size as well as behavior before and after liftoff can be achieved. Many pre-
vious works have performed limited investigations of the different phases of vapor
bubble lifespan, with most of the experiments focusing on bubble size at the moment
of liftoff. Within these experiments, very few have treated boiling at elevated pres-
sures, and even fewer have measured the wall temperature directly. Without these
measurements, understanding of nucleate boiling is incomplete and poorly validated
correlations would be used to predict subcooled flow boiling.
Chapter 6

Conclusions

Air injection and subcooled flow boiling experiments have been conducted to better
understand the subcooled flow boiling phenomenon and also contribute detailed new
data for validation of Interface Tracking Methods (ITM) in the DOE Nuclear HUB
project, Consortium for Advanced Simulation of Light Water Reactors (CASL). Atmo-
spheric pressure experiments have been performed using a vertical annular test sec-
tion with a 9.53mm OD inconel tube placed inside a 19.04mm ID polycarbonate tube. A
high-speed imaging system and PIV technique for flow visualization have been used
to obtain highly resolved data useful for ITM and CFD code validation. The conditions
for these experiments ranged from laminar to turbulent (Re=1625-6555) for air bubble
injection experiments and in the transition region to fully turbulent (Re=5460-17020)
for subcooled flow boiling experiments.

As an extension of the air injection experiments, a rectangular channel was constructed

to investigate the effect of altering the channel shape and dimensions on liquid-bubble
interactions. In these experiments the main channel was 20mm × 5.1mm, however, an
acrylic plate was mounted to the top of the channel to reduce the gap to 20mm ×
1.9mm. To expand the Reynolds number range, the liquid was also heated to 88◦C,
reducing the liquid viscosity substantially. The resulting conditions spanned multiple
flow regimes from fully laminar to transition to fully turbulent (Re=1300, 3300, 7700).

169
170 Chapter 6. Conclusions

To further investigate the effect of elevated pressure on bubble growth and behavior
during growth and after liftoff, additional forced convective subcooled flow boiling
experiments were performed using an elevated pressure loop. This experimental loop
had an annular test section with a 12.7mm OD stainless steel tube inside a 25.4mm
ID stainless steel tube. A borosillicate sight glass was installed at the measurement
plane to perform high speed video measurements of the bubble behavior. A total of
12 experimental cases were covered with 5 runs in each case, treating pressures from
1 to 10 bar, with bubbles for each case being measured for the entire lifespan. Liquid
turbulence in these experiments was much higher than in the atmospheric pressure
cases (Re=24,470-85,740) due to the expanded operating limits permitted by this flow
loop. Parameters measured included bubble size, shape and motion before and after
liftoff.

6.1 Air Bubble Injection Experiments

The air bubble injection experiments involving PIV measurements yielded ensemble-
averaged velocity distributions as air bubbles were injected from a single hole on the
inconel tube. As expected, the bubble injection strongly affected the instantaneous
velocity profile and turbulence level in the region between the inconel tube wall and
midway through the annulus. The shape and motion of the air bubbles which were
injected and then departed from the tube wall were also quantified. In the laminar flow
case, the coalescence of two bubbles successively injected could be seen. In turbulent
flow, the bubble departure frequency increased and no coalescence was observed.

Additional air injection experiments were performed in a rectangular geometry to in-

vestigate the effect of reducing the channel dimensions on the bubble dynamics. Ex-
periments were conducted under laminar, transition, and turbulent flow conditions
Chapter 6. Conclusions 171

and ensemble-average velocity distributions were obtained as air bubbles were in-
jected from a single hole on the surface of the test section. As expected, the bubble
injection strongly affected the liquid velocity profile and turbulence level in the region
between the injection wall surface and midway through the minichannel. It was deter-
mined that the confinement number in addition to the Reynolds number would play
a significant role in the bubble size and shape.

The findings of the air injection experiments can be summarized as follows:

• The injection of air bubbles causes significant streamwise and spanwise veloc-
ity fluctuations at the injection elevation. In some cases, immediately near the
wall the velocity fluctuation is larger than the average value, indicating that the
motion of the bubble interface can cause local backflow.

• Downstream of the injection plane, the motion of the bubble primarily affects
the streamwise velocity. The zone of influence of the fully entrained bubbles on
the liquid flow field was equal to approximately 1.5 times the average bubble
diameter.

• Localized deviations from the single-phase velocity profiles were observed, and
the deviations were greater at the injection site where the bubbles were station-
ary.

• Bubble coalescence was frequently observed under laminar flow conditions. The
bubble velocity immediately after detachment was slower than the rate of ad-
vancement of the subsequently injected bubble interface.

• Holding the channel geometry constant, higher liquid Reynolds numbers re-
sulted in smaller, more frequent bubbles due to increased liquid shear. Smaller
bubbles experienced smaller amplitude but higher frequency oscillations in shape.

• For low confinement numbers, the bubble trajectories and shape variations were
172 Chapter 6. Conclusions

nearly uniform between individual bubbles. Variations in shape and rise path
were reduced for high Reynolds numbers

• At high confinement numbers, the bubbles had highly irregular shapes and tended
to have large amplitude, out-of-plane motions.

• Although the supplied air injection rate was continuous, delays were observed
between subsequent bubbles formed, attributable to a necessary buildup of air
pressure to overcome the liquid surface tension inside the injection hole.

• Preliminary correlations were developed to predict bubble size based on the liq-
uid Reynolds number and confinement number. It was determined that the ratio
of liquid to gas superficial velocities was not a major contributing factor to the
bubble size and shape. The direction of air injection, perpendicular or parallel
to the liquid stream (tested by Okawa et al. [94]), has a significant effect on the
bubble size and shape, as well as variations in liquid velocity field. Previous
experimental works investigated vertical injection into a vertical flow, and cor-
relations developed in this work do not predict bubbles injected in parallel with
the liquid flow.

With these results, a better understanding of individual bubble rise motion and fluc-
tuations in shape has been obtained. For modeling purposes, the adiabatic injection
of air is an excellent starting point for the validation of numerical simulations, as the
mass of such bubbles is conserved and the relevant Kolmogorov scales, detailed in
Appendix B, were comparable with those seen in flow boiling cases. This results in a
similar size computational domain but with less complexity from thermal effects.
Chapter 6. Conclusions 173

6.2 Subcooled Flow Boiling

Vapor bubbles were observed to be near spherical in shape in the moments after incip-
ience, but became slightly more elliptical as they grew. Once the bubbles departed, an
oscillatory motion was observed, along with a corresponding change in bubble size
and shape, as shown in Section 5.3.1.1. Though the motion and size of the bubble
oscillated, the shape of the bubble remained elliptical.

The vapor bubbles forming on the heated tube wall affected the turbulent velocity
profile in the liquid but the zone of influence was found to be limited by the liquid
Reynolds number. In all cases, it was found that static bubbles affected liquid veloci-
ties in the area about 1.5 times the size of the bubble diameter, as seen in Section 5.3.1.2
causing deviations from expected behavior. Further effects of heating included higher
velocity fluctuations throughout the channel due to the increase turbulent effect, as
well as steepening viscous boundary layers. The most significant difference was seen
in case B-03, which had a low liquid flow rate, but the effect of heating was significant
enough to cause the flow to shift from the laminar to transition regime.

As the liquid navigated around a growing bubble, there were clear shifts in the radial
velocity profile that scaled with the bubble size. Slightly upstream of the bubble, the
liquid moved in the outward radial direction, with a corresponding inward motion
occurring in the immediate wake of the bubble. Simultaneously, the axial and radial
velocities near the wall in the bubble nucleation plane were equal to the velocity of
the bubble interface. In some cases, a negative axial velocity was observed upstream
of a growing bubble due to the rapid bubble growth. Depending on the size of the
bubble, only one or two velocity vectors near the wall were affected by the presence of
the bubble. Further Micro-PIV experiments were performed to provide more details
in the vicinity of a single bubble, showing that for a stationary bubble, the effect on
the velocity field diminished at a distance of 1.5 times the liftoff diameter, consistent
174 Chapter 6. Conclusions

with the regular PIV experiments. For sliding bubbles, the liquid velocity matched the
velocity of the bubble interface, as expected, and this effect persisted throughout the
viewing area.

The bubble liftoff diameters measured in the elevated pressure experiments fit well
for the most part with the expected trends presented by Sugrue et al. [116], in that
the bubble liftoff diameters increase with decreasing subcooling and pressure. It was
observed that the bubble liftoff diameters increased with decreasing non-dimensional
wall temperature, θ, indicating that the temperature gradient plays a significant role in
determining the liftoff diameter. While there was no discernible effect of the Reynolds
number on the dimensionless liftoff diameter, it was observed that the bubble liftoff
time decreased with increasing Reynolds number. Although comparing the bubble
liftoff time and the local Nusselt number did not provide a clear link between bubble
growth rate and heat flux, the experimental data at atmospheric pressure fit reasonably
well with Situ et al.’s [81] correlation relating the non-dimensional heat flux and Jakob
number.

Along with the experimental data presented, multiple correlations for the liftoff diam-
eter were also evaluated against an experimental database of 265 points including the
present results and results from previous works, with mixed results. The average er-
ror was measured in two ways, one that weighed each measurement equally, and one
that increased the weight of elevated pressure data. Of the correlations tested, Situ
et al.’s [80] correlation provided the best accuracy, but required an implicit numerical
solution to calculate and had slightly diminishing accuracy with increasing pressure.
Other correlations performed well with the limited datasets under which they were
developed, but failed to predict well the current experimental results as other datasets.
With these considerations, a new correlation was developed to predict liftoff diameter,
taking into consideration multiple parameters that are often ignored, but are needed
Chapter 6. Conclusions 175

to provide a complete picture of subcooled flow boiling phenomena.

Similarly, the bubble liftoff time was investigated. The bubble growth time, by defini-
tion is a lower limit to the bubble liftoff time. Correlations from Basu et al. [177] and
Zuber [21] were evaluated to determine if they fulfill this criterion, with acceptable
results. Furthermore, Situ et al. [80] detailed the force balance on a bubble prior to
liftoff, indicating in the formulation that the bubble liftoff time would decrease with
increasing Reynolds number, holding heat transfer terms constant. This concept was
confirmed experimentally, resulting in a new proposed correlation. The liftoff time
was shown to decrease with increasing Reynolds number and density ratio, and in-
creasing Jakob number and non-dimensional wall temperature. Additionally, liftoff
times for bubbles at elevated pressures (7 and 10 bar) were significantly longer than
at lower pressures. This increase is partially attributable to the reduction in Prandtl
number that occurs with increasing pressure. As the Prandtl number decreases, the
thermal and momentum boundary layers become of similar size, limiting the shear
force for the liquid flow acting on the bubble. For this lower shear force to remove
the bubble from the heated surface, additional heat is required to vaporize the liquid
microlayer between the bubble and the heated surface. The measured liftoff time is
a complicated parameter and is highly dependent on a number of other parameters
such as nucleation site density and departure frequency, which were not investigated
in this work.

Additional correlations for the bubble history number, which describes the rate of
growth and collapse of bubbles before and after liftoff, were developed using the ex-
perimental database from this work. Few works investigated the bubble growth rate,
so previous correlations and experimental data were unavailable. The collapse rate
has garnered attention in some select works [92, 98, 162, 171, 178], due to the close re-
lation to interfacial heat transfer after bubble liftoff. None of these works, however,
176 Chapter 6. Conclusions

attempt to investigate the bubble collapse at elevated pressures. To further complicate

matters, the correlations that were developed in these works suffered from one of two
problems, either they were invalid for a significant range of flow conditions, or were
insensitive or too sensitive to key changes in fluid parameters. To address these limi-
tations in existing models, two new correlations were proposed for the bubble history
number during growth and collapse.

Using these correlations, the trends predicted by Sugrue et al. [116] could be verified
quantitatively. The effect of varying each parameter on bubble growth and behavior
could be isolated and treated separately, however, some parameters could not be fully
decoupled. Table 6.1 lists the parameters varied in the subcooled flow boiling exper-
iments and their effects on the bubble liftoff diameter, aspect ratio at liftoff, nondi-
mensionalized radial bubble motion, and bubble history numbers during the growth
and collapse phases. The Table summarizes the effects of increasing each experimen-
tal parameters on the bubble size, shape and behavior during subcooled flow boiling.
Note that these effects are only applicable to vertical upward subcooled flow boiling
of water.

The trends from Table 6.1 were determined from the correlations developed. Some
observations of interest can be made from these results:

• Increasing the system pressure causes a general reduction in bubble diameters

and a corresponding increase in the bubble aspect ratio. The reductions in bub-
ble growth and collapse rates are due to the increased pressure, which requires
higher interfacial heat transfer and vaporization of more liquid to produce simi-
lar sized bubbles seen in lower pressures.

• Increased wall heat flux had a small effect on the bubble liftoff diameter, and was
not a dominant factor. From the plotted data, it was determined that the bubble
aspect ratio would increased by increasing the heat flux, likely due to a thicker
Chapter 6. Conclusions 177

TABLE 6.1: Effect of various flow boiling parameters on bubble size,

shape and behavior (∼ indicates no appreciable change)
Experimental Parameter Increased
Pressure Heat Flux Wall Superheat Subcooling Liquid Inlet Velocity
00
P qw ∆Tw ∆Tsub < jf,in >
Bubble
liftoff
↓ ↑ ↑ ↓ ↓
diameter
Dlo
Bubble
liftoff
↑ ↓ ↓ ↑ ↓
time
tlo
Bubble
aspect ratio
↑ ↑ ↑ ↓ ↓
at liftoff
Ra,lo
Nondimensional
bubble radial
∼ ↑ ↑ ↓ ↓
motion
Rb+
Bubble
growth rate ↓ ∼ ↑ ↓ ∼
∂βg /∂t
Bubble
collapse rate ↓ ∼ ↓ ↑ ∼
−∂βc /∂t

thermal boundary layer. Similarly, bubbles moved farther away from the wall
under higher heat fluxes for the same reason.

• Increasing the wall superheat affected the bubble as would be expected, by in-
creasing the bubble diameter and aspect ratio, and permitting a thicker thermal
boundary layer that allowed the bubble to persist over a larger distance from the
wall.

• Similarly increasing the liquid subcooling had adverse effects on the bubble
growth, reducing the overall diameter and increasing the collapse rate of the
bubbles.

• A higher liquid inlet velocity caused a reduction in the viscous boundary layer
thickness, as seen in the PIV results. As the bubble diameters tend not to exceed
the buffer zone, a reduction in the boundary layer thickness causes a reduction in
the liftoff diameter. Because the Prandtl number is greater than unity, the thermal
178 Chapter 6. Conclusions

boundary layer thickness can be considered to be less than that of the viscous
layer. This results in a sharper temperature gradient near the wall, reducing the
radial distance over which the bubble can persist.

Enhancing heat transfer in boiling applications is done by increasing the overall rate
of phase change, and is not limited to single bubble growth and departure. Achieving
a careful balance of the nucleation site density and bubble departure frequency while
also maximizing the bubble diameters would lead to heat transfer enhancement. The
data presented in this work were used to develop new correlations for single bub-
ble growth and behavior under various operating conditions. Compared to previ-
ous correlations which lacked the capability to capture departure behavior at elevated
pressures or bubble growth after departure, the current correlations show a marked
improvement in overall accuracy and applicability.

6.3 Future work

Despite the advances made in recent years in the area of subcooled flow boiling re-
search, experimental data for elevated pressures is still lacking. Without additional
experimental studies, relevant models for predicting subcooled flow boiling or two-
phase flows, such as ITM and CFD models cannot be properly validated. From the
work performed as part of this thesis, two additional aspects of subcooled flow boil-
ing are identified to be investigated experimentally:

• Bubble nucleation site densities

• Bubble departure frequencies

Using a combination of the nucleation site densities and departure frequencies, the
heat transferred from the heater rod to the fluid can be partitioned between heating
of liquid flow and phase change. Furthermore, experimental techniques should aim
Chapter 6. Conclusions 179

to measure quantities that are often estimated or extrapolated, particularly if the pa-
rameter is critical to existing numerical models. Important parameters, and suggested
methods to measure them include but are not limited to:

• Liquid temperature distributions

– Traverse a thermocouple through the bulk fluid near the wall

• Heater surface temperature distributions

– Apply a thermochromatic coating to the heater surface

– Use precision machining to mount multiple themocouples or RTDs near a

nucleation site

• Vapor quality

– Use gamma densitometry in combination with visual techniques

Modifications should be made to improve the measurement capabilities of the flow

loops to better understand heat transfer during convective boiling. Future exper-
iments to be performed on the experimental flow loops used in this work should
aim at providing simultaneous measurements of bubble departure frequency, depar-
ture/liftoff diameters, and nucleation site density. These three quantities can provide
a complete picture of the heat transfer enhancement due to boiling, and serve as a
foundation for future boiling models.
References

[1] M. Shoji and V.K. Dhir. Handbook of Phase Change: Boiling and Condensation. CRC
Press, 1999.
[2] G.K. Batchelor. An Introduction to Fluid Mechanics. Cambridge Mathematical
Library, 1967.
[3] C.L. Fefferman. Existence and smoothness of the navier-stokes equation. PDF,
2000. Clay Mathematics Institute - Millenium Prize Problems.
[4] Office of Integrated U.S. Energy Information Administration and Interna-
tional Energy Analysis. Annual energy outlook 2015 with projections to 2040.
Technical report, U.S. Department of Energy, April 2015.
[5] N. Seshasayee. Understanding thermal dissipation and design of a heatsink.
Technical report, Texas Instruments, 2011.
[6] 50 Years of Moore’s Law. http://www.intel.com/content/www/us/en/silicon-
innovations/moores-law-technology.html, 2015.
[7] Discover Energy Corp. Temperature effects on battery performance and life.
[8] H.L. Wu, X.F. Peng, P. Ye, and Y. Eric Gong. Simulation of refrigerant flow boiling
in serpentine tubes. International Journal of Heat and Mass Transfer, 2007.
[9] H. Schlicting and K. Gersten. Boundary Layer Theory. Springer, 1999.
[10] V. Falkner and S.W. Skan. Some approximate solutions of the boundarylayer
equations. The London, Edinburgh, and Dublin Philosophical Magazine and Journal
of Science, 12(80):865–896, 1931.
[11] M. Keshtkar. The falkner-skan flow over a wedge with variable parameters.
Journal of Applied Science and Agriculture, 2014.
[12] M. Jakob and W. Fritz. Frosh. Geb. Ingenieurwes, 2:434, 1931.
[13] M. Jakob. Kondensation und verdampfung. Zeitschriff des Vereins Deutscher In-
gerieure, 1932.
[14] W. Fritz. Berechnung des maximalvolume von dampfblasen. Phys. Z., 36(379-
388), 1939.
[15] S. Nukiyama. Maximum and minimum values of heat transmitted from metal
to boiling water under atmospheric pressue. Journal of the Japanese Society of
Mechanical Engineers, 1934.
[16] R.W. Lockhart and R.C. Martinelli. Proposed correlation of data for isothermal,
two-phase, two-component flow in inpipes. Chem. Prog., 45:39–48, 1949.

180
REFERENCES 181

[17] F.C. Gunther. Photographic study of surface-boiling heat transfer to water in

forced convection. Transaction of the ASME, 73:115–123, 1951.
[18] J.W. Chastain R.A. Egen, D.A. Dingee. Vapor formation and behavior in boiling
heat transfer. Technical Report BMI-1163, Battelle Memorial Institute, 1957.
[19] P. Griffith, J.A. Clark, and W.M. Rohsenow. Void volumes in subcooled boiling
systems. ASME Paper 58-HT 19, U.S. National Heat Transfer Conference, Chicago,
IL, 1958.
[20] N. Zuber. Hydrodynamic Aspects of Boiling Heat Transfer. PhD thesis, University
of California Los Angeles, 1959.
[21] N. Zuber. The dynamics of vapor bubbles in nonuniform temperature fields.
International Journal of Heat and Mass Transfer, 2(1–2):83 – 98, 1961.
[22] N. Zuber. Nucleate boiling. the region of isolated bubbles and the similarity
with natural convection. International Journal of Heat and Mass Transfer, 6(1):53 –
78, 1963.
[23] H. Groothuis and W.P. Hendal. Heat transfer in two-phase flow. Chemical Engi-
neering Science, 1959.
[24] W.R. Gambill and N.D. Greene. Boiling burnout with water in vortex flow. Chem-
ical Engineering Progress, 1958.
[25] W.R. Gambill, R.D. Bundy, and R.W. Wansbrough. Heat transfer, burnout and
pressure drop for water for swirl flow through tubes with internal twisted tapes.
1961.
[26] G.W. Mauer. A method of predicting steady-state boiling vapour fraction in
reactor coolant channels. Technical Report WAPDBT-19, Westinghouse Electric
Corp. Atomic Power Division, 1960.
[27] H. Christensen. Power to void transfer functions. Technical Report ANL-6385,
Argonne National Laboratory, Argonne IL, 1961.
[28] J.K. Ferell. A study of convection boiling inside a channel. PhD thesis, North Car-
olina State University - Raleigh NC, 1964.
[29] University of Exeter. Void measurement in the region of sub-cooled and low quality
boiling. Symposiym on Two-Phase Flow, 1965.
[30] G.G. Bartolemei and V.M. Chanturiya. Experimental study of true void fraction
when boiling sub-cooled water in vertical tubes. Thermal Engineering, 14:123–
128, 1967.
[31] Measure de la perte de pression per acceleration et etude de l’apperition pression. CENG,
Grenoble Fr, 1967.
[32] C.C. St-Pierre and S.G. Bankoff. Vapor volume profiles in developing two-phase
flow. Int. J. Heat Mass Transfer, 10:237–249, 1967.
[33] R.H. Whittle and R. Forgan. Correlation for the minima in the pressure drop
versus flowrate curves for subcooled water flowing in narrow heated channels.
182 REFERENCES

Nuclear Engineering and Design, 6:89–99, 1967.

[34] N. Koumoutsos, R. Moissis, and A. Spyridonos. A study of bubble departure in
forced-convection boiling. Journal of Heat Transfer, pages 223–230, 1968.
[35] F.W. Staub. The void fraction in subcooled boiling - prediction of the initial point
of net vapor generation. Journal of Heat Transfer, pages 151–157, 1968.
[36] R. Evangelisti and P. Lupoli. The void fraction in an annular channel at atmo-
spheric pressure. Int. J. Heat Mass Transfer, 12:699–711, 1969.
[37] G.E. Dix. Vapor void fraction for forced convection with subcooled boiling at low flow
rates. PhD thesis, University of California - Berkeley, Berkeley CA, 1971.
[38] R. Martin. Measurement of the local void fraction at high pressure in a heating
channel. Nuclear Engineering and Design, 48:125–137, 1972.
[39] K. Sekoguchi, H. Fukui, and Y. Sato. Flow boiling in subcooled and low quality
regions heat transfer and local void fraction. In Proceedings of the 5th International
Heat Transfer Conference, Tokyo Japan, volume 4, pages 180–184, 1974.
[40] R.A.M. Al-Hayes and R.H.S. Winterton. Bubble diameter on detactment in flow-
ing liquids. Int. J. Heat Mass Transfer, 24:223–230, 1981.
[41] Z. Edelman and E. Elias. Void fraction distribution in low flow rate subcooled
boiling. Nuclear Engineering and Design, 66:375–382, 1981.
[42] G.G. Bartolemei, V.G. Brantov, Y.U. Molochnikov, Y.V Kharitonov, V.A Solodkii,
G.N. Batashova, and V.N. Milhailov. An experimental investigation of true vol-
umetric vapour content with subcooled boiling in tubes. Thermal Engineering,
29:132–135, 1982.
[43] D.A. Labuntsov, A.G. Lobachev, B.A. Kol’chugin, and E.A. Zakharova. The main
principles of variation in vapour contents of equilibrium and non-equilibrium
two-phase flows in channels of different geometry. Thermal Engineering, 31:506–
508, 1984.
[44] R.D. McLeod. Investigation of subcooled void fraction growth in water under
low pressure and low flowrate conditions. Master’s thesis, Carelton University,
Ottawa, Ontario, Canada, 1986.
[45] J.T. Rogers, M. Salcudean, Z. Abdullah, D. McLeod, and D. Poirier. The onset of
significant void in up-flow boiling of water at low pressure and velocities. Int. J.
Heat Mass Transfer, 30:2247–2260, 1987.
[46] T. Dougherty, C. Fighetti, G. Reddt, B. Yang, T. Jafri, E. McAssey, and Z. Qureshi.
Flow boiling in vertical down-flow. In Proceedings of 10th International Heat Trans-
fer Conference, volume 6, pages 9–14, 1990.
[47] M. Salcudean E.L. Bibeau. Subcooled void growth mechanisms and prediction
at low pressure and low velocity. Int. J. Multiphase Flow, 20(837-863), 1994.
[48] M. Salcudean E.L. Bibeau. A study of bubble ebullition in forced-convective
subcooled nucleate boiling at low pressure. Int. J. Multiphase Flow, 37(2245-2259),
REFERENCES 183

1994.
[49] M. Shoukri O. Zeitoun. Axial void fraction profiles in low pressure subcooled
flow boiling. Int. J. Heat Mass Transfer, 40:869–879, 1997.
[50] G.P. Celata, M. Cumo, and A. Mariani. Experimental evaluation of the onset of
subcooled flow boiling at high liquid velocity and subcooling. Int. J. Heat Mass
Transfer, 40:2879–2885, 1997.
[51] M.D. Bartel, M. Ishii, Y. Mi, T. Masukawa, R. Situ, and M. Mori. Experimental
investigation of subcooled boiling under bwr conditions. In Proceedings of the
NURETH-9, San Francisco, 1999.
[52] Y.Y. Hsu. On the size range of active nucleation cavities on a heating surface.
ASME J Heat Transfer, 1962.
[53] A.E. Bergles and W.M. Rohsenow. The determination of forced-convection
surface-boiling heat transfer. Journal of Heat Transfer, 1964.
[54] J. Chen. Correlation for boiling heat transfer to saturated fluids in convective
flow. I&EC Process Design and Development, 1966.
[55] N.W. Snyder and T.T. Robin. Mass-transfer model in subcooled nucleate boiling.
AMSE J Heat Transfer, 1968.
[56] M.G. Cooper and A.J.P. Lloyd. The microlayer in nucleate pool boiling. Interna-
tional Journal of Heat and Mass Transfer, 12(8):895 – 913, 1969.
[57] H.C. Ünal. Maximum bubble diameter, maximum bubble-growth time and
bubble-growth rate during the subcooled nucleate flow boiling of water up to
17.7mn/m2 . Int. J. of Heat Mass Transfer, 1976.
[58] H.C. Ünal. Void fraction and incipient point of boiling during the subcooled
nucleate boiling of water. Int. J Heat Mass Transfer, 1977.
[59] R. Hino and T. Ueda. Studies on heat transfer and flow characteristics in sub-
cooled flow boiling - part 1. boiling characteristics. Int J Multiphase Flow, 1984.
[60] R. Hino and T. Ueda. Studies on heat transfer and flow characteristics in sub-
cooled flow boiling - part 2. flow characteristics. Int J Multiphase Flow, 1984.
[61] M.V.H. DelValle and D.B.R. Kenning. Subcooled flow boiling at high heat flux.
Int. J. Heat Mass Transfer, 28:1907–1920, 1985.
[62] J.W. Goodman. Laser Speckle and Related Phenomena. Springer, Berlin, 1975.
[63] R. Grousson and S. Mallick. Study of flow pattern in a fluid by scattered laser
light. Applied Optics, 1977.
[64] R. Meynart. Digital image processing for speckle flow velocimetry. Review of
Scientific Instruments, 1982.
[65] R.J. Adrian. Scattering particle characteristics and their effect on pulsed laser
measurements of fluid flow: speckle velocimetry vs particle image velocimetry.
Applied Optics, 1984.
184 REFERENCES

[66] R.J. Adrian. Image shifting technique to resolve directional ambiguity in double-
pulsed velocimetry. Applied Optics, 1986.
[67] C.D. Meinhart, S.T. Wereley, and J.G. Santiago. Piv measurements of a mi-
crochannel flow. Experiments in Fluids, 1999.
[68] C.D. Meinhart, S.T. Wereley, and J.G. Santiago. A piv algorithm for estimating
time-averaged velocity fields. Journal of Fluids Engineering, 2000.
[69] F. Scarano and M.L. Riethmuller. Iterartive multigrid approach in piv image
processing with discrete window offset. Experiments in Fluids, 1999.
[70] J. Sheng, H. Meng, and R.O. Fox. A large eddy piv method for turbulence dissi-
pation rate estimation. Chemical Engineering Science, 55(20):4423 – 4434, 2000.
[71] S.G. Kandlikar. A general correlation for saturated two-phase flow boiling heat
transfer inside horizontal and vertical tubes. J. of Heat Transfer, 1990.
[72] S.G. Kandlikar. Heat transfer mechanisms during flow boiling in microchannels.
ASME J of Heat Transfer, 2004.
[73] J.F. Klausner, R. Mei, D.M. Bernhard, and L.Z. Zeng. Vapor bubble departure in
forced convection boiling. Int. J. Heat Mass Transfer, 1992.
[74] L.Z. Zeng, J.F. Klausner, D.M. Bernhard, and R. Mei. A unified model for the
prediction of bubble detachment diameters in boiling systems - ii. flow boiling.
Int. J. Heat Mass Transfer, 1993.
[75] G.E. Thorncroft, J.F. Klausner, and R. Mei. An experimental investigation of
bubble growth and detachment in vertical upflow and downflow boiling. Int. J.
Heat Mass Transfer, 41:3857–3871, 1998.
[76] R.H.S. Winterton and J.S. Munaweera. Bubble size in two-phase gas-liquid bub-
ble flow in ducts. Chemical Engineering and Processing, 40:437–447, August 2001.
[77] N. Basu, G.R. Warrier, and V.K. Dhir. Onset of nucleate boiling and active nucle-
ation site density during subcooled flow boiling. ASME J. Heat Transfer, 124:717–
728, 2002.
[78] R. Situ, Y. Mi, M. Ishii, and M. Mori. Photographic study of bubble behaviors
in forced convection subcooled boiling. Int. J. Heat Mass Transfer, 47:3659–3667,
2004.
[79] R. Situ, T. Hibiki, X. Sun, Y. Me, and M. Ishii. Flow structure of subcooled boiling
flow in an internally heated annulus. Int J. of Heat and Mass Transfer, 2004.
[80] R. Situ, T. Hibiki, M. Ishii, M. Mori, J. Y. Tu, G.H. Yeoh, and G.C. Park. Bubble
lift-off in forced convective subcooled boiling flow. In Boiling. Begell House,
2006.
[81] R. Situ, M. Ishii, T. Hibiki, J.Y. Tu, G.H. Yeoh, and M. Mori. Bubble departure
frequency in forced convective subcooled boiling flow. Int. J. Heat Mass Transfer,
51:6268–6282, 2008.
[82] J. Kim and M.H. Kim. On the departure behaviors of bubble at nucleate pool
REFERENCES 185

boiling. International Journal of Multiphase Flow, 32(10-11):1269–1286, oct 2006.

[83] M.V. Sardeshpande and V.V. Ranade. Two-phase flow boiling in small channels:
A brief review. Sadhana, 38(6):1083–1126, dec 2013.
[84] B. Xiao, G. Jiang, D. Zheng, L. Chen, and B. Liu. Calculation of active nucleation
site density in boiling systems. Research Journal of Applied Sciences, 2013.
[85] E. Baglietto and M. Christon. Demonstration & assessment of advanced mod-
elling capabilities for multiphase flow with sub-cooled boiling. CASL Internal
Document, 2013.
[86] J.J. Xu, B.D. Chen, and T.Z. Xie. Experimental and theoretical analysis of bubble
departure in narrow rectangular channel. Progress in Nuclear Energy, 2014.
[87] X. Fang, Z. Zhou, and H. Wang. Heat transfer correlation for saturated flow
boiling of water. Applied Thermal Engineering, 2015.
[88] Y. Cao, Z. Kawara, T. Yokomine, and T. Kunugi. Experimental study on be-
haviors of two successive bubbles in subcooled flow boiling at high degrees of
subcooling. Thermal Science and Engineering, 2015.
[89] B.B. Mikic, W.M. Rohsenow, and P. Griffith. On bubble growth rates. Int. J. Heat
Mass Transfer, 1970.
[90] T.H.J.J. van der Hagen and D.D.B. van Bragt A.J.C. Stekelenburg. Reactor ex-
periments on type-i and type-ii bwr stability. Nuclear Engineering and Design,
2000.
[91] R. Roy, A. Hasan, and S. Kalra. Temperature and velocity fields in turbulent
liquid flow adjacent to a bubbly boiling layer. Int .J. Multiphase Flow, 19:765–795,
1993.
[92] G.R. Warrier, V.K. Dhir, and L.A. Momoda. Heat transfer and pressure drop in
narrow rectangular channels. Experimental Thermal and Fluid Science, 26(1):53 –
64, 2002.
[93] H.M. Choi, T. Kurihara, H. Monji, and G. Matsui. Measurement of parti-
cle/bubble motion and turbulence around it by hybrid {PIV}. Flow Measurement
and Instrumentation, 12(5–6):421 – 428, 2002.
[94] T. Okawa, T. Tanaka, I. Kataoka, and M. Mori. Temperature effect on single
bubble rise characteristics in stagnant distilled water. International Journal of Heat
and Mass Transfer, 46(5):903 – 913, 2003.
[95] R. Maurus, V. Ilchenko, and T. Sattelmayer. Automated high-speed video anal-
ysis of the bubble dynamics in subcooled flow boiling. International Journal of
Heat and Fluid Flow, 25(2):149–158, apr 2004.
[96] R. Maurus and T. Sattlemayer. Bubble and boundary layer behavior in subcooled
flow boiling. International Journal of Thermal Sciences, 2006.
[97] C.E. Estrada-Perez and Y.A. Hassan. Ptv experiments of subcooled boiling flow
through a vertical rectangular channel. International Journal of Multiphase Flow,
186 REFERENCES

36:691–706, 2010.
[98] S.J. Kim and G.C. Park. Interfacial heat transfer of condensing bubble in sub-
cooled boiling flow at low pressure. Int J of Heat and Mass Transfer, 2011.
[99] C. Hutter, K. Sefiane, T.G. Karayiannis, A.J. Walton, R.A. Nelson, and D.B.R.
Kenning. Nucleation site interaction between artificial cavities during nucle-
ate pool boiling on silicon with integrated micro-heater and temperature micro-
sensors. International Journal of Heat and Mass Transfer, 2012.
[100] S.M. Kim and I. Mudawar. Universal approach to predicting two-phase fric-
tional pressure drop for mini/micro-channel saturation flow boiling. Int J of
Heat and Mass Transfer, 2013.
[101] C. Clark, M. Griffiths, S.W. Chen, T. Hibiki, M. Ishii, I. Kinoshita, and Y. Yoshida.
Experimental study of void fraction in an 8x8 rod bundle at low pressure and
low liquid flow conditions. Int J of Multiphase Flow, 2014.
[102] R. Samaroo, N. Kaur, K. Itoh, S. Banerjee, and M. Kawaji. Turbulent flow char-
acteristics in an annulus under air bubble injection and subcooled flow boiling
conditions. Nuclear Engineering and Design, 268:203–214, 2014.
[103] Y. Chin, M.S. Lakshminarasimhan, Q. Lu, D.K. Hollingsworth, and L.C. Witte.
Convective heat transfer in vertical asymmetrically heated narrow channels.
Journal of Heat Transfer, 124(6):1019, 2002.
[104] M.S. Lakshminarasimhan, Q. Lu, Y. Chin, D.K. Hollingsworth, and L.C. Witte.
Fully developed nucleate boiling in narrow vertical channels. Journal of Heat
Transfer, 2005.
[105] V. Prodanovic, D. Fraser, and M. Salcudean. Bubble behavior in subcooled flow
boiling of water at low pressures and low flow rates. International Journal of
Multiphase Flow, 28(1):1 – 19, 2002.
[106] X. Huo, L. Chen, Y.S. Tian, and T.G. Karayiannis. Flow boiling and flow regimes
in small diameter tubes. Applied Thermal Engineering, 2004.
[107] L. Chen, Y.S. Tian, and T.G. Karayiannis. The effect of tube diameter on vertical
two-phase flow regimes in small tubes. International Journal of Heat and Mass
Transfer, 2006.
[108] M.M. Mahmoud and T.G. Karayiannis. Heat transfer correlation for flow boiling
in small to micro tubes. International Journal of Heat and Mass Transfer, 2013.
[109] M.M. Mahmoud and T.G. Karayiannis. Flow patter transition models and cor-
relations for flow boiling in mini-tubes. Experimental Thermal and Fluid Science,
2016.
[110] T. Okawa, T. Ishida, I. Kataoka, and M. Mori. Bubble rise characteristics after the
departure from a nucleation site in vertical upflow boiling of subcooled water.
Nuclear Engineering and Design, 235(10-12):1149–1161, may 2005.
REFERENCES 187

[111] D. Euh, B. Ozar, T. Hibiki, M. Ishii, and C.H. Song. Characteristics of bubble de-
parture frequency in a low-pressure subcooled boiling flow. J. of Nuclear Science
and Technology, 47:608–617, 2010.
[112] D. Yuan, L. Pan, D. Chen, H. Zhang, J. Wei, and Y. Huang. Bubble behavior
of high subcooling flow boiling at different system pressure in vertical narrow
channel. Applied Thermal Engineering, 31(16):3512–3520, 2011.
[113] D. Chen, L. Pan, and S. Ren. Prediction of bubble detachment diameter in flow
boiling based on force analysis. Nuclear Engineering and Design, 2012.
[114] B. Ozar and M. Ishii. Investigation of one-dimensional interfacial area transport
for vertical upward air-water two-phase flow in an annular channel at elevated
pressures. Nuclear Engineering and Design, 263:362–79, 2013.
[115] A. Miglani, D. Joo, S. Basu, and R. Kumar. Nucleation dynamics and pool boiling
characteristics of high pressure refrigerant using thermochromic liquid crystals.
International Journal of Heat and Mass Transfer, 60:188–200, may 2013.
[116] R. Sugrue, J. Buongiorno, and T. McKrell. An experimental study of bubble
departure diameter in subcooled flow boiling including the effects of orienta-
tion angle, subcooling, mass flux, and pressure. Nuclear Engineering and Design,
279:182–188, 2014.
[117] J. Yan, Q. Bi, Z. Liu, G. Zhu, and L. Cai. Subcooled flow boiling heat transfer
of water in a circular tube under high heat fluxes and high mass fluxes. Fusion
Engineering and Design, 2015.
[118] M.D. Bartel, M. Ishii, T. Masukawa, Y. Mi, and R. Situ. Interfacial area measure-
ments in subcooled flow boiling. Nuclear Engineering and Design, 210(1–3):135 –
155, 2001.
[119] I.C. Chu, H.C. No, and C.H. Song. Bubble lift-off diameter and nucleation fre-
quency in vertical subcooled boiling flow. Journal of Nuclear Science and Technol-
ogy, 2011.
[120] S.C.M. Yu and C.P. Tso. Simulation of fiber optic sensors in determination of thin
liquid film thicknesses. Advances in Engineering Software, 1995.
[121] S.C.M. Yu, C.P. Tso, and R. Liew. Analysis of thin film thickness determination
in two-phase flow using multifiber optical sensor. Applied Math Modelling, 1996.
[122] B.B. Bayazit, D.K. Hollingsworth, and L.C. Witte. Heat transfer enhancement
caused by sliding bubbles. Journal of Heat Transfer, 125(3):503, 2003.
[123] X. Li, D.K. Hollingsworth, and L.C. Witte. The thickness of the liquid microlayer
between a cap-shaped sliding bubble and a heated wall: Experimental measure-
ments. Journal of Heat Transfer, 128(9):934, 2006.
[124] D.K. Hollingsworth, X. Li, and L.C. Witte. The thickness of the liquid microlayer
between a sliding bubble and a heated wall: Comparison of models to experi-
mental data. Journal of Heat Transfer, 130(11):111501, 2008.
188 REFERENCES

[125] N. Borgetto, C. Galizzi, F. André, and D. Escudié. A thickness measurement

technique based on low-coherence interferometry applied to a liquid film with
thermal gradient. Experimental Thermal and Fluid Sciences, 2010.
[126] N. Borgetto, F. André, C. Galizzi, and D. Escudié. Simultaneous film thickness
measurement and wall temperature assessment by low-coherence interferome-
try. Experimental Thermal and Fluid Sciences, 2013.
[127] T. Yabuki, T. Saitoh, and O. Nakabeppu. Contribution of microlayer evaporation
to bubble growth in pool saturated boiling of water. In International Conference
on Nanochannels, Microchannels and Minichannels, 2014.
[128] T. Yabuki, R. Samaroo, O. Nakabeppu, and M. Kawaji. Mems sensor measure-
ment of surface temperature response during subcooled flow boiling in a rect-
angular flow channel. Experimental Thermal and Fluid Science, pages 0894–1777,
February 2015.
[129] M. Caira, G. Caruso, and A. Naviglio. A correlation to predict chf in subcooled
flow boiling. Internation Communications in Heat and Mass Transfer, 1995.
[130] W. Liu, H. Nariai, and F. Inasaka. Prediction of critical heat flux for subcooled
flow boiling. International Journal of Heat and Mass Transfer, 43(18):3371 – 3390,
2000.
[131] M. Kureta and H. Akimoto. Critical heat flux correlation for subcooled boiling
flow in narrow channels. International Journal of Heat and Mass Transfer, 2002.
[132] W. Zhang, T. Hibiki, K. Mishima, and Y. Mi. Correlation of critical heat flux
for flow boiling of water in mini-channels. International Journal of Heat and Mass
Transfer, 2006.
[133] H. Zhang, I. Mudawar, and M.M. Hasan. Photographic study of high-flux sub-
cooled flow boiling and critical heat flux. International Communications in Heat
and Mass Transfer, 34(6):653–660, jul 2007.
[134] S.J. Kim, I.C. Bang, J. Buongiorno, and L.W. Hu. Surface wettability change
during pool boiling of nanofluids and its effect on critical heat flux. Int J of Heat
and Mass Transfer, 2007.
[135] K. Hata and S. Masuzaki. Critical heat fluxes of subcooled water flow boiling
in a short vertical tube at high liquid reynolds number. Nuclear Engineering and
Design, 2010.
[136] J.M. Le Corre, S.C. Yao, and C.H. Amon. A mechanistic model of critical heat
flux under subcooled flow boiling conditions for application to one- and three-
dimensional computer codes. Nuclear Engineering and Design, 2010.
[137] J.M. Le Corre, S.C. Yao, and C.H. Amon. Two-phase flow regimes and mech-
anisms of critical heat flux under subcooled flow boiling conditions. Nuclear
Engineering and Design, 2010.
[138] C. Gerardi, J. Buongiorno, L.W. Hu, and T. McKrell. Study of bubble growth in
water pool boiling through synchronized, infrared thermometry and high-speed
REFERENCES 189

video. Int J of Heat and Mass Transfer, 2010.

[139] Y. Chen. Heat Transfer in Film Boiling of Flowing Water, Heat Transfer - Theoretical
Analysis, Experimental Investigations and Industrial Systems. InTech, 2011.
[140] Y.J. Chung, K.H. Bae, K.K. Kim, and W.J. Lee. Boiling heat transfer and dryout
in helically coiled tubes under different pressure conditions. Annals of Nuclear
Energy, 2014.
[141] G. Házi, A.R. Imre, G. Mayer, and I. Farkas. Lattice boltzmann methods for
two-phase flow modeling. Annals of Nuclear Energy, 200.
[142] D. Yu, R. Mei, L.S. Luo, and W. Shyy. Viscous flow computations with the
method of lattice boltzmann equation. Progress in Aerospace Sciences, 2003.
[143] T. Inamuro. Lattice boltzmann methods for viscous fluid flows and for two-
phase fluid flows. Fluid Dynamics Research, 2006.
[144] A.Xu, T.S. Zhao, L.An, and L.Shi. A three-dimensional pseudo-potential-based
lattice boltzmann model for multiphase flows with large density ratio and vari-
able surface tension. International Journal of Heat and Fluid Flow, 2015.
[145] P.L. Bhatnagar, E.P. Gross, and M. Krook. A model for collision processes in
gases i: Small amplitude processes in charged and neutral one-component sys-
tems. Physical Review, 1954.
[146] D. dH́umieres. Generalized lattice-boltzmann equations. Rarefied Gas Dynamics,
Progress in Astronautics and Aeronautics, AIAA, 1992.
[147] P. Lallemand and L.S. Luo. Theory of the lattice boltzmann method: dispersion,
dissipation, isotropy, galilean invariance, and stability. Phys. Rev., 2000.
[148] E. Chen, Y. Li, and X. Cheng. Cfd simulation of upward subcooled boiling flow
of refrigerant-113 using the two-fluid model. Applied Thermal Engineering, 2009.
[149] D.R. Langewisch and J. Buongiorno. Prediction of film thickness, bubble veloc-
ity, and pressure drop for capillary slug flow using a cfd-generated database. Int
J of Heat and Fluid Flow, 2015.
[150] R.P. Roy, S. Kang, J.A. Zarate, and L. Laporta. Turbulent subcooled boiling flow
experiments and simulations. ASME J of Heat Transfer, 2002.
[151] S. Kang and R.P. Roy. Vapor phase measurements in subcooled boiling flow.
Journal of Heat Transfer, 2002.
[152] M. Potter and D. Wiggert. Schaum’s Outlines of Fluid Mechanics. McGraw Hill,
2008.
[153] A. Tamayol and M. Bahrami. Laminar flow in microchannels with noncircular
cross section. J. of Fluids Engineering, 2010.
[154] R. Panton. Incompressible Flows. John Wiley & Sons, 2005.
[155] A.N. Kolmogorov. C. R. Acad. Sci. U.R.S.S, 1941.
[156] Y.A. Cengel. Heat and Mass Transfer. McGraw Hill, 2007.
190 REFERENCES

[157] C.F. Colebrook. Turbulent flow in pipes, with particular reference to the transi-
tion region between smooth and rough pipe laws. J, Inst. Civil Engr 11, 1938.
[158] Y.A. Cengel and M.A. Boles. Thermodynamics: an engineering approach. McGraw
Hill, 2008.
[159] A.A. Troshko and Y.A. Hassan. Law of the wall for two-phase turbulent bound-
ary layers. International Journal of Heat and Mass Transfer, 44(4):871 – 875, 2001.
[160] L.M. Jiji. Heat Convection. Springer, 2006.
[161] H.K. Forster and N. Zuber. Growth of a vapor bubble in a superheated liquid.
Journal of Applied Physics, 25:474, 1954.
[162] Y.M. Chen and F. Mayinger. Measurements of heat transfer at phase interface of
condensing bubble. Int. J. of Multiphase Flow, 1992.
[163] T. Sato and H. Matsumura. On the conditions of incipient subcooled boiling
with forced convection. Bull. JSME, 1964.
[164] E.J. Davis and G.H. Anderson. The incipience of nucleate boiling in forced con-
vection flow. AIChE J., 1966.
[165] H.K. Forster and N. Zuber. Journal of the American Institute of Chemical Engineers,
1955.
[166] F.W. Dittus and L.M.K. Boelter. Heat transfer in automobile radiators of the
tubular type. University of California Publications in Engineering, Berkeley, 1930.
[167] W. Chen and X. Fang. A note on the chen correlation of saturated flow boiling
heat transfer. International Journal of Refrigeration, 2014.
[168] J.G. Collier. Heat transfer in the post dryout region and during quenching and reflood-
ing. Hemisphere, 1982.
[169] R.M. Wellek, A.K. Agrawal, and A.H.P. Skelland. Shape of liquid drops moving
in liquid media. AIChE J., 1966.
[170] D.W. Moore. The rise of a gas bubble in a viscous liquid. J. Fluid Mech., 1959.
[171] O. Zeitoun, M. Shoukri, and V. Chatoorgoon. Measurement of interfacial area
concentration in subcooled liquid-vapor flow. Nuclear Engineering and Design,
1994.
[172] H. Tennekes and J.L. Lumley. A first course in turbulence. The MIT Press, 1972.
[173] I. Michiyoshi and A. Serizawa. Turbulence in two-phase bubbly flow. Nuclear
Engineering and Design, 95:253–267, 1986.
[174] A. Serizawa, I. Kataoka, and I. Michiyoshi. Turbulence structure of air-water
bubbly flow - i: Measuring technique. Int. J. Multiphase Flow, 2:221–233, 1975.
[175] A. Serizawa, I. Kataoka, and I. Michiyoshi. Turbulence structure of air-water
bubbly flow - ii: Local properties. Int. J. Multiphase Flow, 2:235–246, 1975.
[176] P. Kew and K. Cornwell. Correlations for the prediction of heat transfer in small-
diameter channels. Applied Thermal Engineering, 1997.
REFERENCES 191

[177] N. Basu, G.R. Warrier, and V.K. Dhir. Wall heat flux partitioning during sub-
cooled flow boiling: Part 1-model development. ASME J. Heat Transfer, 127:131–
140, 2005.
[178] S.S. Bertsch, E.A. Groll, and S.V. Garimella. A composite heat transfer correlation
for saturated flow boiling in small channels. International Journal of Heat and Mass
Transfer, 2009.
Appendix A

Literature Review Comparison

Although a historical background and key relevant contemporary works were detailed
in Chapter 2, not all experimental works were listed. This Appendix organizes a wide
range of experimental works on boiling and two-phase flows, highlighting the ex-
perimental conditions investigated, the experimental techniques used and parameters
measured, and summarizing the important results from each work. The works are
then categorized and grouped together based on the focus of each.

In addition to the works reviewed, a collection of publications related to this work is

listed below in chronological order:

• Kawaji, M., R. Samaroo, J. Kreynin, T. Lee, and S. Banerjee. Subcooled Flow Boiling
Experiments and Numerical Simulation for Virtual Reactor Development. Proc. of the
15th International Topical Meeting on Nuclear Reactor Thermal - Hydraulics,
NURETH-14, Toronto, Ontario, Canada. September 25-30, 2011.

• Samaroo, R., D. Das, N. Kaur, K. Itoh, M. Kawaji, T. Lee, and S. Banerjee. Air
Bubble Injection and Subcooled Flow Boiling Experiments for Numerical Simulation of
a Virtual Reactor. Proc. of 2012 Japan-U.S. Seminar on Two-Phase Flow Dynamics,
Tokyo, Japan. June 7-12, 2012.

• Samaroo, R., M. Kawaji, T. Lee, and S. Banerjee. Subcooled Flow Boiling Experi-
ments for Numerical Simulation of a Virtual Reactor. Proc. of the 15th International

192
Appendix A. Literature Review Comparison 193

Topical Meeting on Nuclear Reactor Thermal - Hydraulics, NURETH-15, Pisa,

Italy. May 12-17, 2013.

• Samaroo, R., and M. Kawaji. Air Bubble Injection into a Liquid Stream in a Minichan-
nel. Proc. of The 11th International Conference on Nanochannels, Microchannels,
and Minichannels, Sapporo, Hokkaido, Japan. June 16-19, 2013.

• Samaroo, R., N. Kaur, K. Itoh, T. Lee, S. Banerjee, and M. Kawaji. "Turbulent

Flow Characteristics in an Annulus under Air Bubble Injection and Subcooled
Flow Boiling Conditions." Nuclear Engineering and Design 268 (2014): 203-14.

• Samaroo, R., P. Rahman, and M. Kawaji. Investigation of Subcooled Flow Boiling

Heat Transfer at 1 - 3 Bar. Proc. of The 1st Thermal and Fluid Engineering Summer
Conference, TFESC, New York City, USA. August 9-12, 2015.
Appendix A. Literature Review Comparison

TABLE A.1: Summary of fundamental literature reported for subcooled flow boiling
Geometry
00
Dh Working P G qw ∆Tsub ∆Tw Measured
Authors Findings
Flow Direction Fluid Parameters
[mm] [kPa] [kg m−2 s] [kW m−2 ] [◦C] [◦C]
Del Valle and Rectangular Bubble size More nucleation sites
Kenning [61] 0.6cm2 Water 101.3 0-2000 77500 84 - Departure frequency with increasing
(1985) Upflow Nucleation site density ∆Tw
Square Comparison with
Klausner et al. [73] Bubble size
25 R-113 101.3 113-287 11-26 - 12-17.7 Mikic et al.’s [89]
(1992) Bubble growth rate
Horizontal correlation
Square
Zeng et al. [74] Bubble size Comparison with
25 R-113 101.3 113-287 11-26 - 12-17.7
(1993) Bubble growth rate Zuber’s [21] correlation
Horizontal
Square Bubble size Identified key differences
Thorncroft et al. [75]
12.7 FC-87 101.3 190-666 1.3-14.6 1-5 - Bubble growth rate between upflow, downflow
(1998)
Up/Down/Pool Departure frequency and pool boiling
Circular
Bubble diameter Comparison of experimental data
Winteron and Munaweera [76] 7,10.9,13.2
Water,1-hexanol,PP1 101.3 2100-700 - - - Gas and liquid superficial velocities with existing correlations and
(2001) Air injection into
Turbulent Energy dissipation previously published data
upward liquid stream
Square/Nine-rod bundle Location of ONB was
Inner surface temperature
with heated wall/rods determined from visual
Basu et al. [77] of heater rods,
16.33cm2 flow area Water 101.3 124-886 1.6-96.3 1.7-52.5 3-22 thermocouple measurements.
(2002) location of ONB
1.11 cm OD rods Correlation for nucleation
nucleation site density
Upflow site density was developed
Developed empirical correlations
Annulus Bubble size
Situ et al. [78, 80, 81] for non-dimensional bubble
19.1ID, 38.1OD Water 101.3 466-900 54-206 1.5-20 - Bubble growth rate
(2004,2005,2008) diameter and departure
Upflow Departure frequency
frequency
Simultaneous infrared thermography The surface temperature reduced
Rectangular and high speed video in the presence of a growing bubble
Buongiorno et al. [85]
10×30 Water 105-150 Re=0-105 0-2000 0-75 - to measure bubble size and however temporal resolution may not be
(2014)
Upflow motion as well as wall be high enough to capture
temperature response rapidly high departure frequencies
Stereoscopic measurement
A force analysis showed
Rectangular of bubble size
Xu et al. [86] that the bubble growth
2×8 Water 101.3 138-710 26.3-215.4 8.4-28.3 - contact and inclination
(2014) force is small compared
Vertical to ±45◦ angles and contact
to the other forces
diameter
After departure of
Measurement of
Annulus the first bubble
Cao et al. [88] instantaneous bubble
20ID, 8OD Water 101.3 67-293 2390 max 30-50 - the second will grow
(2015) diameters for
Upflow larger than it
successive bubbles
ordinarily would
194
 TABLE A.2: Summary of literature reported for measurements of flow field and two-phase interactions
Geometry
00
Dh Working P G qw ∆Tsub ∆Tw Measured
Authors Findings
Flow Direction Fluid Parameters
−2 −2 ◦ ◦
[mm] [kPa] [kg m s] [kW m ] [ C] [ C]
Annular Local velocity and void Liquid turbulence structure
Roy et al. [91] 35-43 70-95
11.37 R-113 219-253 Re:24400-33400 30-80 fraction measurement was highly affected by
(1993) inlet temperature wall temperature
Upflow using hot film anemometry the bubbly sublayer
Thermal boundary layer thickness
Near-square Temperature field measurement
Warrier et al. [92] increase and interfacial heat
16.33cm2 Water 103 235-684 9.6-96.3 7-46.5 8-14 Bubble contact angle and
(2002) transfer were estimated by
Upflow collapse rate
bubble collapse rate
Increased buoyancy effect
from larger bubbles was
attenuated with resistance
Hybrid particle image
Circular from deformations in shape.
Choi et al. [93] velocimetry/tracking
40 Water 101.3 No liquid flow - - - Bubbles with significant
(2002) for flow field measurements
Upflow deformations produce twice
around injected bubbles
the turbulence intensity
as a similarly sized
solid particle
Comparison with existing
Circular High speed video
Okawa et al. [94] No liquid correlations and development
0.1, 0.5, 1 Water 101.3 - 0-12.6 - for measuring oscillatory
(2003) flow of a new correlation for
Upflow bubble motion
motion amplitude
Particle image velocimetry The impact on the flow
Square
Maurus and Sattelmayer [96] to measure the impact of extends beyond the bubbly
Appendix A. Literature Review Comparison

40×40 Water 101.3 250-2000 175-1100 10-30 15-53

(2006) bubbly sublayer on liquid layer, but depends on
Horizontal
boundary layer the magnitude of boiling
The velocity profile deviated
from the theoretical law
Rectangular PIV measurements of
Estrada-Perez et al. [97] Re:3309, 9929, of the wall with increasing
8.7×7.6 HFE-301 100-1600 0-64 - - the effect of wall heat
(2010) and 16,549 heat flux, with a stronger
Upflow flux on turbulent parameters
effect at lower Reynolds
numbers
Rectangular Traversing microthermocouple used
Kim and Park [98] 78-88 Development of a new
25×15 Water 112 85 60.1-122.8 - to measure temperature around
(2011) Inlet temperature interfacial heat transfer model
Upflow a collapsing bubble
Bubble departure diameter, The bubble diameter and wall
Pool
Hutter et al. [99] No liquid departure frequency, superheat were influenced by the
40×37 FC-72 50-100 - 0 1.5-14
(2012) flow and nucleation frequency close proximity of the
Horizontal
variations with cavity spacing nucleation sites
8×8 square rod bundle
140×140 Observed increase in distribution
Clark et al. [101] < jf >=0-1m s−1 Impedance sensor measurement
with Drod =12.7 Air-Water 101.3 No heating - - parameter at low mixture
(2014) < jg >=0.3-10m s−1 of void fraction
Prod =16.7 volumetric flux
Upflow
Air bubbles injected into
PIV measurements of flow field
low Reynolds number flows were
Annular around adiabatically injected
Samaroo et al. [102] Re:1625,6555 (air injection) larger and coalesced more often.
19.1ID, 9.52OD Water/Air 101.3 100 9 - air bubbles. Micro PIV
(2014) Re:18000 (boiling) Single small vapor bubbles
Upflow measurements of flow field
affected a small region near
around vapor bubble
the heated wall
195
Appendix A. Literature Review Comparison

TABLE A.3: Summary of literature reported for subcooled flow boiling at elevated pressures
Geometry
00
Dh Working P G qw ∆Tsub ∆Tw Measured
Authors Findings
Flow Direction Fluid Parameters
[mm] [kPa] [kg m−2 s] [kW m−2 ] [◦C] [◦C]
Correlations developed for
Circular Bubble diameter
Ünal [57] bubble growth rate
8 Water 13900, 15800, 17700 0.08-9.15m s−1 470-10640 3-86 - Bubble growth rate
(1975) maximum diameter
Upflow Growth time
and growth times
Circular
Celata et al. [50] Pressure drop in test section Comparison with correlations
8 Water 1000-2500 4400-8400 0-14000 120-194 -
(1997) Heat flux at ONB for heat flux and wall superheat
Upflow
Determination of
Dodewaard
Van der Hagen et al. [90] Subcooling, Zuber, Type I (hydrostatic head)
natural-cirulation Water 2,000-13,000 45-1,300kg s−1 45-190MWth 3.5-13 -
(2000) and Flashing numbers and Type II (density wave)
BWR
stability criterion
Chin et al. and Rectangular Calculation of heat transfer
Wall temperature measurements using
Lakshminarasimhan et al. 1,2,4cm2 R-113 120-210 60-5260 4.4-121 12.6-37 44-52 coefficient enhancement due to
Liquid Crystal Thermography
[103, 104] (2002,2005) Upflow boiling
Correlations for bubble
Bubble maximum diamter
Annular maximum and lift-off
Prodanovic et al. [105] Lift-off Diameters
22ID, 12.7OD Water 105-300 0.08-0.8m s−1 100-1200 10-30 - diameter and times
(2002) Growth time
Upflow were developed using a
Lift-off time
single form equation
Huo et al. [106] (2004)
Circular
Chen et al. [107] (2006) Photographic study of Development of modified
0.52-4.26 R134a-R245fa 100-14000 100-700 1.7-158 - -
Mahmoud et al. [108, 109] flow behavior flow regime map
Upflow
(2013,2016)
Bubble diameter Limiting conditions for
Circular
Okawa et al. [110] Bubble motion bubble sliding
20 Water 121-127 94-1435 54-338 2.2-10 -
(2005) Growth rate Bubble motion after
Upflow
Collapse rate departure
Annulus Improvement of Situ et al.’s [81]
Euh et al. [111]
19.1ID, 38.1OD Water 167-346 214-1869 61-238 7.5-23.4 - Departure frequency correlation for dimensionless
(2010)
Upflow bubble departure frequency
Decrease in maxmimum diameter
Rectangular
Yuan et al. [112] Bubble diameter with increased pressure
2×0.5 Water 100-1000 76.6-602.7 50-348 20-36 -
(2011) Growth rate Increased vapor pressure
Upflow
results in decreased growth rate
Developed of a dimensionless
Rectangular
Chen et al. [113] bubble growth model which
45×2 Water 120-335 214.58-702.38 83.62-334.52 Ti n:79.4-111.60◦C - Bubble detachment diameter
(2012) predicts bubble departure time
Upflow
and contact diameter
Conductivity probe was used
Annular
Ozar et al. [114] < jf >=0.23-3.36m s−1 to measure bubble velocities and local Measured IAC compared well
38.1ID, 19.1OD Water and air 126-589 - - -
(2013) < jg0 >=0.04-3.43m s−1 void fraction. Interfacial area concentration with the IATE
Upflow
was also measured
Flat plate Thermochromic liquid crystal Heat transfer coefficient
Miglani et al. [115] No liquid
- R134a 8167 2.5-12 0 4.9-7.9 was used to measure wall was enhanced when temperature drop
(2013) flow
Horizontal temperature response under bubble occurred at the surface
Bubble departure diameters increase
Rectangular
Sugrue et al. [116] Bubble diameter with increasing heat flux,
14.3×19.9 Water 101-505 250-400 50-100 10,20 2-6
(2014) Growth rate and decreasing mass flux,
Orientation angle:0◦ ,30◦ ,45◦ ,60◦ ,90◦
subcooling and pressure
Measurement of wall temperature
Circular allowed calculation of heat
Yan et al. [117] Wall temperature
9 Water 3000, 4200, 5000 6000, 8000, 10000 5000-12500 - 2-40 transfer coefficient at HHHM
(2015) Inlet vapor quality
Upflow conditions. Development of a
modified Chen’s correlation
Entrained bubble size
Annular Comparison with multiple
Bartel et al. [51, 118] Interfacial velocity
38.1ID, 19.1OD Water 101.3 400-1953 0-193 2-8.9 - experimental studies, many
(1999,2001) Local time-averaged void fraction
Vertical at elevated pressures
Pressure difference in bubbles
Annulus Comparison with multiple correlations [57, 80, 105]
Chu et al. [119] Bubble size
31.75ID, 9.5OD Water 139-152 300-702 133.4-355.6 1.4-24 - Improvement on correlation for
(2011) Departure Frequency
Upflow bubble departure diameter
196
 TABLE A.4: Summary of literature reported for microlayer thickness measurement or estimation
Geometry
00
Dh Working P G qw ∆Tsub ∆Tw Measured
Authors Findings
Flow Direction Fluid Parameters
[mm] [kPa] [kg m−2 s] [kW m−2 ] [◦C] [◦C]
Microlayer thickness under
Tilted tank Bubble diameter sliding bubbles was estimated
Bayazit et al. [122] No liquid
- FC-87 101.3 1.6 5 8-10 Bubble growth rate using wall temperature response
(2003) flow
Upflow Wall temperature response Enhancement of heat transfer
Appendix A. Literature Review Comparison

coefficient was observed

Simultaneous high speed video
Tilted tank and microlayer measurement Microlayer thickness was measured
Li et al. [123] (2006)
- FC-87 101.3 Re=819-2413 - - - using fiber optic probe. directly using the fiber
Hollingsworth et al. [124] (2008)
Upflow Li et al. also used TLC optic probe
temperature measurement
Rectangular MEMS sensor measurement of Wall temperature response
Yabuki et al. [128]
20×5.1 Water 101.3 240 90.4-105 - 18-20.5 wall temperature at high was used to estimate
(2014)
Upflow sampling frequency microlayer thickness
197
Appendix A. Literature Review Comparison

TABLE A.5: Summary of literature investigating CHF or dryout conditions

Geometry
00
Dh Working P G qw ∆Tsub ∆Tw Measured
Authors Findings
Flow Direction Fluid Parameters
[mm] [kPa] [kg m−2 s] [kW m−2 ] [◦C] [◦C]
Outlet quality was measured
Rectangular at CHF conditions
Kureta and Akimoto [131] New CHF correlation was
4.2-5.28 Water 101.3 846-15100 2000-13000 10-60 - Heated length was varied
(2002) proposed
Upflow to allow flow to develop
in some cases
Dry patches expanded and
Rectangular Photographic study was performed combined at near CHF condition
Zhang et al. [133]
5×2.5 FC-72 138-152 (outlet) 0.5-1.5m s−1 30-48 10-30 (outlet) - to observe the evolution CHF occurred when the
(2007)
Horizontal of CHF dry patch covered the
surface
Nanoparticles deposited onto the
Wire/Flat Plate Water with 0.1% concentration heater surface during boiling
Kim et al. [134] No liquid
Dwire =0.051,A=5×45 Water with nanoparticles 101.3 500 - - of nanoparticles was boiled to CHF causing an ehancement of CHF
(2007) flow
No flow Heat surface was measured after The deposition was caused by
microlayer evaporation
Circular Correlations were developed
Hata and Masuzaki [135] Critical wall temperature
6 Water 812.1-1181.5 17.2-42.4m s−1 - 80.9-147.6 - depending on the inlet and
(2010) and heat flux were recorded
Upflow outlet temperatures and velocities
Circular Wall temperature measurements
Chen et al. [139] Heat transfer coefficient
6.7×20 Water 100-6000 23-1462 4.7-7.9 - - were taken at varying
(2011) varied with inlet conditions
Upflow stages of DNB
Wall temperature response was
Pool
Gerardi et al. [138] No liquid Simultaneous high speed measured using infrared camera
- Water 101.3 2-100 0 0.9-10.5
(2010) flow video and thermography Bubble growth and waiting time
No flow
were recorded
Dryout conditions of water
Helical tube
Chung et al. [140] Dryout conditions were measured in helical coils was
12mm Water 1000-5800 176.84-530.52 2587.89 - -
(2014) by interpolating the wall temperature investigated. Increased centrifugal
Upflow
forces complicate the flow
198
Appendix B

Experimental Procedures

B.1 Limitations of Experimental Techniques

The limitations of the visualization techniques employed stem from two components:
the high speed camera, and the pulsing laser. All visualization experiments must be
planned with considerations for these limits in mind. If the time or length scales of
the physical phenomena being investigated are outside of these limits, sacrifices must
be made to either bring the phenomena into the limits or the amount of data must be
altered to allow the phenomena to be recorded.

While the camera is capable of recording video at 100,000 frames per second (FPS), it
does so by sacrificing image resolution. There is also a limit to the storage memory
on the camera itself at 8GB, which determines the overall length of the video. For
example, if 10 seconds of video can be captured at 1000 FPS, then only 1 second can be
captured at 10000 FPS. The resolution of the Phantom V310 is 1280x800 at up to 3250
FPS, whereas the resolution of the Photron Mini UX100 is 1200x1024 at up to 4000 FPS.

A similar limit exists for the pulsed laser; the laser has two pulsing heads, with each
head capable of a maximum pulse rate of 15Hz. For PIV acquisition, the timing be-
tween the heads can be adjusted to a minimum of 1µs, or a maximum of 1/30s. For
high flow rates where particles move large distances in a short time, the timing should

199
200 Appendix B. Experimental Procedures

be small; for low flow rates, the timing can be larger, but does not necessarily have to
be. A significant limitation of PIV arises when the physical process that needs to be
recorded takes a very short time relative to the timing between pulses. In this case, the
laser will only pulse once or twice during the process, limiting the amount of data that
can be acquired.

To determine which technique to employ, high speed video versus PIV, the order of
magnitude of the characteristic timescale should be known. Due to physical limita-
tions of the pulse laser, the PIV technique cannot capture any physical phenomena
with a timescale of less than 100ms with a sufficient number of images (>3 images). To
put this in context, if a bubble nucleates, grows and departs within 100ms, PIV would
not be the recommended technique. High speed video, while excelling in document-
ing the nucleation, growth and departure of a bubble, does not provide any liquid
velocity data.

When increasing the magnification in an image, there is a tradeoff between the magni-
fication and the depth of field, the depth over which the image is in focus. To compen-
sate for this effect, brighter lighting is required to produce sharper images. While this
is not a problem in high speed video acquisition, it is an issue for micro-PIV imaging
due to a limit on the brightness of the laser pulse and the safety risks of operating the
laser at high power output. For PIV imaging, if the depth of field is smaller than the
thickness of the laser sheet, a disproportionate number of particles will move out of
the viewing plane between images, reducing the accuracy of the results.

There are also some imaging issues that are caused by the flow geometry. Because
much of the work is performed in annular geometry with a metal inner rod, there
is some reflection of laser light off of the heater rod surface that illuminates particles
slightly out of plane. This affects particles near the wall, resulting in slightly higher
velocity readings near the wall than would otherwise be measured. This effect cannot
Appendix B. Experimental Procedures 201

be readily quantified without the use of fluorescing particles that would be illuminated
differently with reflected light, allowing those particles to be filtered out.

When acquiring data, it is imperative to consider the limitations of the measurement

techniques and apply any working knowledge of the experimental conditions to de-
termine if the phenomena of interest fall within the limits of the technique.

B.2 Calibration Procedures

Prior to performing the experiment, instrumentation must be calibrated to achieve

accuracy in measurements. This is particularly important for visual measurements, as
simple changes in the camera focus will render previous image calibrations invalid.
For thermocouples, the calibration only needs to be done once, since the calibration
coefficients should not vary over time.

B.2.1 Image Calibration

For each experiment performed, image calibration was required to permit accurate
measurements of bubble sizes and liquid velocity fields. The largest problem encoun-
tered in acquiring images was correcting for the curvature of the outer tube in annular
geometries. Without any correction applied, for observations near the outer tube, there
were large discrepancies between the measured velocity fields and theoretical values.

Optical Correction Due to the geometry of the annular test sections, the curvature
and thickness of the outer tube plays a significant role in optical measurements. As
measurements are taken in the vicinity of the inner wall of the outer tube, the optical
path in the outer tube wall through which the observations are made becomes much
greater, and the distortion becomes more significant. For the purposes of calibration, it
202 Appendix B. Experimental Procedures

is imperative to correct for the curvature of the test sections to further enhance the ac-
curacy of the experimental results near the outer wall. To do this, an optical correction
box was constructed to be small enough to not inhibit the camera lens from focusing on
the area of interest within the test section. Figure B.1 illustrates the problem posed by
the outer tube curvature, where the outer tube is a polycarbonate material for ambient
pressure flow loops, and a combination of borosilicate glass and polycarbonate for the
elevated pressure loop. Note that for the elevated pressure loop an optical correction
box could not be installed without either obstructing the camera lens or undermining
the safety functions of the sight glass.

To measure the impact of the optical correction box, several tabletop experiments were
performed to quantify the correction attributable to the optical correction box. Two dif-
ferent cases were treated, as shown in Figure B.1, a half-tube filled with water enclosed
in an otherwise empty optical correction box (equivalent to no correction applied), and
a half-tube enclosed in an optical correction box where both are filled with water. A
low power laser beam was traversed on a micrometer stage through a half tube assem-
bly enclosed inside an optical correction box filled with water. A camera recorded the
perceived position of the beam at each step and the centroid of the beam was calcu-
lated. For a perfectly square channel, the perceived position of the beam would vary
linearly with the actual position.

Without correction, the deviations of the observed position from the actual position
were significant, up to 33% occurring at roughly 2/3 of the tube diameter, and seem-
ingly random variation elsewhere. Even the axial position experienced distortion due
to curvature, and a slight angle between the test section and the camera being am-
plified at the extreme edge of the test section. With correction, this accuracy was im-
proved significantly, with the observed position of the beam falling within the accept-
able tolerance of the actual position (an acceptable tolerance being 1/10th of the beam
Appendix B. Experimental Procedures 203

F IGURE B.1: Optical correction configurations investigated

thickness). Figures B.2a and B.2b show the results of these calibration experiments
and the impact of the correction box on radial and axial distortions respectively, with
a solid line indicating the optimal case of no curvature.
204 Appendix B. Experimental Procedures

( A ) Radial distortion due to curvature ( B ) Axial distortion due to curvature

Calibration of PIV Images With the optical correction box installed, the high speed
video image is calibrated prior to data acquisition. The temperature of the liquid
within the correction box was allowed to equilibrate with the liquid in the test sec-
tion, allowing for matched index of refraction throughout. Within the DaVis software,
this procedure is performed by selecting two points and specifying the distance be-
tween them. Because the optical correction box prevents distortion in both the axial
and radial directions, the two points can be taken in either and the same scaling can be
applied in both directions. For calibration, the horizontal distance between the outer
edge of the inner rod and the inner edge of the outer tube is considered. Figure B.2
illustrates this procedure within the DaVis software.
Appendix B. Experimental Procedures 205

F IGURE B.2: PIV calibration process

Calibration of High Speed Video Images As with the PIV images, a similar pro-
cedure is applied to high speed video images. When using the Phantom high speed
camera in the ambient pressure loops, the images are calibrated using the same two
points – the outer edge of the inner rod, and the inner edge of the outer tube. Because
both the air injection and ambient pressure boiling test sections had correction boxes
installed, this two point calibration can be applied without concern for distortion.
206 Appendix B. Experimental Procedures

When using the Photron camera in the elevated pressure flow loop, the same principle
is applied, however, an optical correction box could not be installed on the test section.
Instead, vertically spaced markings on the inner rod are used for calibration, noting
that the region near the inner rod considered to have negligible distortion, as shown
in Figures B.2a and B.2b.

B.2.2 Flow Instrumentation Calibration

Flow Meters Due to the operating temperature restrictions for digital flow meters,
the only digital flow meter used was installed on the air injection loop. For calibration
purposes, the digital flow meter was installed in series with an analog flow meter that
was calibrated by the manufacturer. No additional calibration was performed on the
analog flow meters.

Syringe Pump for Air Injection To obtain an accurate readout of the rate of injection
from the syringe pump, the inner diameter of the syringe was supplied as an input to
the pump. This rate wass then verified by measuring the length of time required to in-
ject a fixed amount of air. The injection rate was verified experimentally my measuring
the volume change in bubbles injected over a period of time.

B.2.3 Thermocouple Calibration

The calibration procedures varied, depending on the experimental loop of interest.

Problems experienced in the elevated pressure loop did not occur in the ambient pres-
sure loop, so the calibration procedure differs for the two loops.
Appendix B. Experimental Procedures 207

Ambient Pressure Subcooled Flow Boiling For experiments performed in the ambi-
ent pressure subcooled flow boiling loop, thermocouples were calibrated by the man-
ufacturer, Omega Engineering inc, according to NIST specifications. To assure that the
thermocouples were operational for each experiment, readings were taken with the
experimental setup drained of all liquid and compared with an analog thermometer.
In each case, the thermocouples were within the manufacturer specified accuracy of
the thermometer reading, and no additional calibration was needed.

Elevated Pressure Subcooled Flow Boiling Due to the Ohmic heating technique ap-
plied in the elevated pressure boiling loop, there were a number of additional calibra-
tion procedures necessary prior to the operation of the experiment. Once this calibra-
tion was complete, no additional calibration would be needed unless the experimental
apparatus was disassembled or the test section removed. The procedure involves cal-
ibrating the loop-embedded thermocouples, and then calibrating the wall-embedded
thermocouples against the loop-embedded ones.

Wall thermocouple readings experienced a significant deviation from their actual read-
ings due to the DC voltage applied to the test section. The reason for this is due to the
spot welding of the thermocouple junctions to the heated surface. As the voltage ap-
plied to the test section increased, there was a corresponding increase or decrease in
the thermocouple measurements, depending on the polarity of the current through
the test section. To compensate for this deviation of measurement, a calibration tech-
nique was applied to separate the effects of current and temperature. First the tem-
perature was calibrated: cold water was circulated in the loop and the temperature
was allowed to equilibrate with cold water flowing through the heat exchanger, and
then the thermocouple voltages were recorded. This process was repeated at different
temperatures to determine the temperature calibration curve for each wall embedded
thermocouple.
208 Appendix B. Experimental Procedures

To calibrate for the effect of current, varying voltages were applied to the heater rod,
the thermocouples were allowed to reach an equilibrium state, and measurements
recorded. The polarity of the current through the heater rod was then reversed, and
the thermocouples were again allowed to reach equilibrium, and measurements were
again taken. Taking the average of these two measurements yields the actual voltage
output from the thermocouples for the specific temperature, based on temperature
alone. To quantify the effect of the current, the difference between the measured ther-
mocouple voltage for either polarity and the average of the measurements for both
polarities was taken and divided by the voltage applied to the heater rod to obtain
them. This yields a scaling constant that when multiplied by the voltage applied pro-
vides a correction for the thermocouples.

The temperature for any given thermocouple with measured voltage from the thermo-
couple, VT C , and applied heater rod voltage VHR is given by the following equation,

T = f (VT C ) + g(VHR ) = A1 VT C + A2 + BVHR (B.1)

where the calibration constants, A1 , A2 , and B for each thermocouple are shown in Ta-
ble B.1. The difference in Temperature constants between the loop thermocouples and
wall thermocouples is attributable to the loop thermocouples being Type K (chromel-
alumel), and wall thermocouples being Type T (copper-constantan).

TABLE B.1: Thermocouple temperature constants and correction for cur-

rent applied

Temperature Constants Current Constant

Thermocouple
A1 A2 B
Tinlet 40.1 1064 0
Toutlet 40.1 1064 0
Twall1 41.25 1083 -5.48
Twall2 41.25 1083 163.7
Appendix B. Experimental Procedures 209

F IGURE B.3: Operating procedures for the air injection flow loop

Once the thermocouples were calibrated, experiments could be performed; calibra-

tion of visualization techniques is done after the experiments are performed and the
accompanying videos are recorded.

B.3 Flow Loop Operating Procedures

Achieving the desired experimental conditions involves a different process for each
flow loop. For the adiabatic air injection loop, this process is quite simple, as the
condition is determined only by the conjunction of two flow rates, liquid and gas.
For the boiling experiments, arriving at the chosen conditions is more involved as a
careful balance of flow rate and heat input must be achieved. In the case of the elevated
pressure loop, this process is even more complicated, as system is closed and the heat
removed must be considered as well.

Figures B.3, B.4, and B.5 give overviews of the procedures for operating the air injec-
tion loop, ambient pressure boiling flow loop, and the elevated pressure boiling loop,
respectively.
210 Appendix B. Experimental Procedures

F IGURE B.4: Operating procedures for the atmospheric pressure boiling

flow loop
Appendix B. Experimental Procedures 211

F IGURE B.5: Operating procedures for the elevated pressure boiling flow
loop
212 Appendix B. Experimental Procedures

When the chosen conditions are achieved, steady state conditions must be ensured
before acquiring images. The system was judged to be in steady state if the changes in
flow rate and temperature readings were within 1% over a reasonable period of time.
Typically, a period of 5 minutes was sufficient to reach steady state.

B.4 Data Acquisition Procedures

An overview of the procedure behind obtaining high speed video or PIV data is de-
tailed in Figure B.6. In both processes, well-defined images are required to provide
the best quality results. This is particularly true for PIV, which relies on algorithmic
detection of seed particles to calculate the flow field. If the focus or particle density is
insufficient, the results of the PIV processing will be unreliable.
Appendix B. Experimental Procedures 213

F IGURE B.6: Procedure for obtaining high speed video (HSV) or PIV data

High Speed Video Acquisition For high speed video acquisition, a shadowgraphy
technique is applied, with halogen lamp lighting applied behind the test section to
record shadows of bubbles. This shadowgraphy method yields high contrast images
of bubbles to permit easier identification of bubbles in postprocessing. A sample im-
age is shown in Figure B.7.
214 Appendix B. Experimental Procedures

F IGURE B.7: Sample high speed video image (P01-03: P = 1bar, Re =

55281, Ja = 4.2, N ux = 87.6)

The image acquisition frame rate and exposure time of the high speed video is deter-
mined by the time scale of the phenomena to be observed. In air injection experiments
for example, high gas flow rates required a much higher frame rate to adequately
capture the motion of bubbles and the exposure time was kept small intentionally to
prevent motion blur but still permit a high contrast image. For subcooled flow boiling
experiments with low bubble growth rates, a slower frame rate can be used without
running the danger of missing bubbles entirely potentially at the expense of not being
able to capture the departure. Still a short exposure time with high intensity lighting
is necessary to maintain a crisp, high contrast image.

PIV Image Acquisition Prior to image acquisition, seeding particles are added to the
fluid. To test for sufficient seeding, sample images are taken prior to the experiment
and inspected for uniform and dense seeding. Because the seeding particles have a
Appendix B. Experimental Procedures 215

similar density to the liquid, there may be some areas in the flow loop that do not
have adequate seeding – due to liquid stagnation and settling of particles in those
regions. If these regions are of interest, this lack of seeding can be compensated for by
adding more particles or allowing more light into the camera by increasing the laser
intensity, opening the camera aperture, or increasing the exposure time.

For PIV imaging using the dual head pulse laser, the timing between pulses, and thus
images, is adjusted with the flow rate, depending on the level of turbulence encoun-
tered, and depending on the growth rate of bubbles. For short timescale phenomena
such as high turbulence intensity or high bubble growth rates, the timing between
pulses, t1−2 , is kept well below the 1/15s between same head pulses. For such cases
where t1−2 <1/30s, sequential pairs of images are correlated (e.g. 1+2, 3+4, and so
on). For laminar flows, the laser timing can be adjusted so that there is an equal 1/30s
between each pulse, and consecutive images can be correlated (e.g. 1+2, 2+3, and
so on). For treating liquid flows with simultaneous bubble motion, a careful balance
must be struck between capturing the liquid velocity field and the bubble motion: too
short a pulse differential and the bubble history will be limited to the 1/15s pair of
images; too long and the particles will drift too far for any cross-correlation results to
be meaningful.

A continuous laser was purchased to overcome this problem of pulse timing in order
to simultaneously measure the bubble motion with the liquid flow field. For very high
turbulence, as seen in most gas flows, the laser output is insufficient for the low t1−2
required, however, for liquid-vapor flows the continuous laser worked well to capture
the phenomena of interest.

Sufficient Number of Profiles to Average One of the main concerns for perform-
ing ensemble averaging of velocity profiles is the minimum number of instantaneous
profiles needed. Averaging too few profiles results in poor statistics, therefore it is
216 Appendix B. Experimental Procedures

important to determine the sufficient number of images to average to obtain proper

statistics. To do this, a base case was considered: air injection into a turbulent liquid
stream in an annular geometry. This case was chosen because the large bubbles cou-
pled with the liquid turbulence resulted in large fluctuations in both the streamwise
and spanwise directions. To determine the sufficient number of images to average,
the phase-weighted ensemble averaging of instantaneous profiles was performed in
batches of 100, and then those batches were averaged. Upon the inclusion of addi-
tional batches of 100 images, the change in both the ensemble averaged velocities and
the root-mean-square (RMS) fluctuations were examined; once they fall below some
threshold values, the number of images processed was considered to be sufficient.
From this procedure, it was determined that 500 images were sufficient for performing
ensemble averaging. Figures B.8a and B.8b show the difference in ensemble average
velocity between 200 and 100 images, and 500 and 400 images. Figure B.9a and B.9b
show the percent difference in RMS fluctuations between 200 and 100 images, and 500
and 400 images.

( A ) 100-200 images ( B ) 400-500 images

F IGURE B.8: Change in time-averaged streamwise and spanwise velocity

profiles (Re=6555, air injection rate = 5.4mL min−1 )
Appendix B. Experimental Procedures 217

( A ) 100-200 images ( B ) 400-500 images

F IGURE B.9: Percent change in RMS fluctuations in streamwise and span-

wise velocity profiles (Re=6555, air injection rate = 5.4mL min−1 )

As the number of images increases, the change in ensemble averaged velocities and
the percent change in RMS fluctuations approach zero. This ensures that the PIV data
is processed and averaged with proper statistics.

Thermocouple Measurements Thermocouple measurements were recorded by con-

necting the thermocouple leads to the data acquisition system and using National
Instruments SignalExpess software. From this software, the sampling rate and cali-
bration constants could be set for each thermocouple. For this experimental work a
sampling rate of 50Hz was sufficient for obtaining average temperatures at the inlet
and outlet, as well at at the heater surface. Setting the sampling rate too high would
result in a very noisy signal, causing higher uncertainty.

All subcooled flow boiling experiments were performed at steady state conditions,
meaning that the average temperature variation throughout the entire acquisition pro-
cess was <1% of the measured value for a period of time. To achieve such conditions,
a coolant fluid (air or water) was supplied to the heat exchanger at a rate that would
bring the equilibrium temperature to the desired range – the amount of heat extracted
218 Appendix B. Experimental Procedures

by the coolant air or water in addition to the heat losses inherent in the experimental
setup operating at the equilibrium temperature would equal the heat added by the
heater rod.